Description
Growing political and academic interest in entrepreneurship and business demography, and particularly the role and value of new businesses in national economies, is prompting various research projects on these topics. One of the main issues faced by researchers and policy makers is the current lack of international comparability of data on business start-up rates, which are often seen as key indicators of entrepreneurship and economic dynamism.
The International Comparability of
Business Start-up Rates
Final Report
Steven Vale
January 2006
2
3
The International Comparability of Business Start-up Rates
Final Report
Table of Contents
0. Executive Summary
1. Introduction
2. Data Sources and Existing International Comparisons
3. Factors Affecting Comparability
4. Methods to Improve Comparability
5. A Harmonised Methodological Framework and Start-up Indicators?
6. Conclusions
7. References
Annex 1. Glossary of Terms: Proposals for Harmonised Terminology
Annex 2. Inventory of Data on Business Start-ups by Country
Annex 3. Defining Business Populations: Comparing Point in Time and Live During
Period Estimates
Annex 4. Business Start-up Data for Selected Countries: Comparisons of National
Sources
Annex 5. Business Closures
Author: Steven Vale
OECD Statistics Directorate / Office for National Statistics, UK
Contact Details
Please send any comments or questions to the author at: [email protected]
Comments or questions to the OECD should be addressed to:
Nadim Ahmad – [email protected] (Business demography)
Tim Davis – [email protected] (Entrepreneurship indicators)
4
5
0. Executive Summary
Growing political and academic interest in entrepreneurship and business demography, and
particularly the role and value of new businesses in national economies, is prompting various
research projects on these topics. One of the main issues faced by researchers and policy
makers is the current lack of international comparability of data on business start-up rates,
which are often seen as key indicators of entrepreneurship and economic dynamism. The
International Consortium for Dynamic Entrepreneurship Benchmarking, led by the Danish
government agency FORA, has responded by providing funding for a five month consultancy
at the OECD to study this topic. The consultant appointed for this task was Steven Vale, on
secondment from the UK Office for National Statistics.
The objectives of the project were:
• The compilation of existing evidence on comparative start-up rates;
• The comparison of results and identification of reasons for differences in results, in
particular methodological and statistical differences;
• Drawing up lessons for future studies to improve comparability and to ensure that
results are meaningful.
The underlying question that this project has aimed to answer is; “How comparable are
existing data on business start-up rates from different OECD countries?” The short answer is;
“Not very”, so this report looks at the reasons why data are not comparable, and what can be
done to improve comparability in the future.
This report starts by examining the existing sources of business start-up data for different
countries (an inventory of sources is included in Annex 2), and assessing previous
international projects and papers that have aimed to produce comparable data for groups of
countries. Where there are several data sources for a particular country, they have been
studied to gain a better understanding why they often differ (see Annex 4). The conclusion
from this work is that there are a number of factors that affect the comparability of business
start-up data, some of which may have been overlooked in previous international
comparisons, resulting in the true variability of data between countries being masked by
methodological differences.
Section 3 develops these ideas into a typology of the factors affecting international
comparisons of business start-up rates, describing each factor, and its potential impact in
detail. Start-up rates are based on two components, the numerator (new businesses), and the
denominator (a population). Some factors affect just one of these, others affect both. In total,
nine factors have been identified:
Numerator factors:
• Purity – to what extent are “pure births” (i.e. new combinations of production factors)
distinguished from reactivations and other creations?
• Timing – at what point in the creation process is a start-up measured?
• Periodicity – over what period are start-ups measured, and how does this affect the
measurement of very short-lived businesses?
Denominator factors:
• Type of Population – businesses or people?
• Temporal basis – is the population measured at a specific point in time, or does it
consist of all units that were present at any time during a given period?
6
Factors affecting both:
• Source – are the data taken from a register, a census or a survey? How reliable is the
source?
• Units – what is the entity about which the data are produced?
• Coverage - to what extent are certain types of business included or excluded based on
specific attributes (e.g. economic activity or legal form)?
• Thresholds – what explicit or implicit size thresholds apply to the source?
Section 4 looks at how these factors can affect data comparisons in practice. It shows that
adjustments to compensate for differences in specific factors can sometimes help to improve
comparability, but have to be made with care, based on a detailed understanding of the data
sources and methods. In this sense, although not perfect, informed adjustments can at least
give approximate results, and can warn against drawing false conclusions based on the raw
data alone.
The goal of more comparable data is the theme of Section 5, which links this project to wider
OECD work to develop a methodological framework for business demography. This section
also looks at the pros and cons of different types of business start-up indicators, and
recommends focussing on one key indicator, supplemented by several secondary indicators.
The conclusions of this report are that:
• Simple comparisons of start-up rates from different sources are often misleading.
• The availability of data on business start-up rates varies considerably between countries.
• Where metadata exist, they are not always easy to find or understand. A harmonised
terminology is proposed in Annex 1, and a common metadata template is needed.
• Some previous international comparisons do not fully recognise all comparability issues,
but have provided useful models for assembling data from different countries.
• To assess the comparability of business start-up rates it is necessary to decompose them
into numerator and denominator components, and consider the factors that affect each.
• The factors that have the most impact are usually the purity of the data in the numerator,
the temporal basis of the denominator, and the coverage of both.
• The larger a “new” business is, the less likely it is to be a pure birth. Increasing purity
leads to a considerable reduction in the employment attributed to new businesses.
• Analytical adjustments can help to compensate for differences in specific comparability
factors, but risk introducing noise into the data, so have to be made with care.
• Statistical business registers are the best sources for business start-up data, as they are
already subject to a degree of harmonisation, particularly within Europe.
• A clearly defined key indicator would improve data comparability. Secondary indicators
could give additional insights to more specialist data users.
• Data producers are often more influenced by national data requirements than international
comparability. The OECD has a role to communicate international needs.
• It is important to find out what data users really want, and what they use start-up data for.
This knowledge can then inform the future development of indicators.
• The short term priority is the identification of “quick wins”, i.e. actions that increase the
international comparability of data from individual countries for minimal cost.
• A step-by step approach may not result in fully comparable data as quickly as some users
might want, though alternative, more radical, approaches may take at least as long, as
they would require considerable changes to methods and sources in many countries.
• The goal of internationally comparable business start-up rates is not an easy one, but is
possible.
7
1. Introduction
There is growing international interest in the topics of business dynamics and
entrepreneurship, particularly from policy makers and academic researchers. Business start-
up rates are seen as providing key indicators for both purposes. They are also used as a
measure of economic dynamism, and have been linked to improvements in productivity
through the notion of creative destruction
1
.
So far the focus has mainly been on producing national data to inform national policies and
research, however, there is a growing interest in international comparisons, particularly for
benchmarking purposes. To facilitate international comparisons, it is necessary to determine
measures of business start-ups that will show the real differences between countries, and not
just reflect differences in national methodologies, as has often been the case in the past.
For this reason, the International Consortium for Dynamic Entrepreneurship Benchmarking,
led by the Danish government agency FORA, has provided funding for a five month
consultancy at the OECD to study the international comparability of business start-up data.
The consultant appointed for this task was Steven Vale, on secondment from the UK Office
for National Statistics.
The objectives of the project were agreed at the outset as being:
• The compilation of existing evidence on comparative start-up rates;
• The comparison of results and identification of reasons for differences in results, in
particular methodological and statistical differences;
• Drawing up lessons for future studies to improve comparability and to ensure that results
are meaningful.
The underlying question that this project has aimed to answer is; “How comparable are data
on business start-up rates from different OECD countries?” Figure 1.1 shows business start-
up rate data for a number of countries, including two sources for the United States, as
published by those countries or Eurostat. Is this chart a valid comparison of business start-up
rates for these countries?
Figure 1.1 – Raw Business Start-up Rate Data for Selected Countries
0%
5%
10%
15%
20%
25%
A
u
s
t
r
a
l
i
a
C
a
n
a
d
a
D
e
n
m
a
r
k
F
i
n
l
a
n
d
F
r
a
n
c
e
G
e
r
m
a
n
y
I
t
a
l
y
L
a
t
v
i
a
N
e
t
h
e
r
l
a
n
d
s
N
e
w
Z
e
a
l
a
n
d
N
o
r
w
a
y
P
o
r
t
u
g
a
l
S
l
o
v
a
k
i
a
S
p
a
i
n
S
w
e
d
e
n
U
K
U
S
(
S
U
S
B
)
U
S
(
B
L
S
)
2000 2001 2002
Sources: National statistical office and Eurostat publications and internet sites
1
Although the focus of this report is on business start-ups, the comparability issues affecting the
complimentary indicator of business closures are set out in Annex 5.
8
This report will show that the comparison in Figure 1.1 is not particularly valid, but that through
an understanding of the data and metadata, meaningful comparisons are possible. To reach
this conclusion, this report decomposes the questions above into a number of sub-questions,
corresponding to the different sub-tasks undertaken within this project:
• What data are available for each OECD country? - The project started by making an
inventory of data sources by country, initially through Internet searches, but also through
discussions with contacts in different countries. A copy of this inventory is included as
Annex 2.
• What metadata are available with these data? – The availability and quality of metadata
for each data source were assessed within the inventory.
• What comparisons or compendiums of data from different countries exist? – A trawl was
made of databases, literature and other sources combining data on start-up rates from
more than one country. Section 2 considers how others have tried to collect and compare
data from different countries, with varying degrees of success.
• Do the data seem comparable? – The above steps gave an initial view as to the degree of
data comparability. The conclusion was that methodological differences frequently mask
the real variations between countries.
• Do the metadata confirm comparability or explain the differences? – This initial view on the
comparability of data was tested using the available metadata, to determine how helpful
these metadata are in highlighting and explaining methodological differences. Annex 4
includes short studies on the comparability of sources within selected countries, on the
assumption that differences in data relating to the same country must be purely
methodological. This work led to the development of the framework of factors affecting the
comparability of start-up rates proposed in Section 3.
• Are there other explanations for differences in data? – The extent to which variations
between countries could be explained by political, social and cultural factors was briefly
considered, though this question is not considered further in this report, as it is more
appropriate to look at these issues when the data have been compiled or corrected to
remove methodological differences.
• How can comparability be improved for existing data? – Methods to make adjustments to
existing data to improve comparability are considered in Section 4, where examples are
used to illustrate how data can be adjusted, and some of the potential pitfalls.
• What is the scope for improving comparability at source? – Finally, Section 5 considers
the extent to which it is possible to recommend changes to the ways the source data are
produced to improve comparability, and proposes a set of standard indicators, within a
harmonised methodological framework.
There is a strong link between this project and other OECD work on business demography,
where this report will feed into the development of a wider methodological framework
including business survival, growth and closure. There are also links to OECD work on
entrepreneurship where there are plans to develop a set of harmonised indicators, including
business start-up rates. Outside the OECD there are links to Eurostat work on business
demography and the factors of business success, as well as to various international groups
concerned with business demography, entrepreneurship and statistical business registers.
9
2. Data Sources and Existing International Comparisons
Most OECD countries have produced indicators on business start-up rates, usually derived
from data held in statistical business registers. However, the methodology used has often
been driven by national considerations, rather than a desire for international comparability. A
quote from a recent Australian paper on establishing a conceptual framework for business
demography (ABS (2004)) illustrates this perfectly; “Whilst international comparability of the
data is considered to be important, the overriding requirement is the provision of data in the
Australian context”. This is not stated as clearly by other national data providers, but appears
to be a widely held view
2
. Understanding the methodological differences between data from
different countries is therefore a vital pre-condition to any meaningful comparisons.
2.1 An Inventory
The first step in this project was the compilation of an inventory of the different sources of data
on start-ups in the OECD member countries (see Annex 2 for a summary version). This
inventory is based on searches of the internet during autumn 2005, and thus will miss any
sources made available after that date, or sources that are only available in other formats.
Linguistic limitations may also mean that some sources not available in English or French
have been missed.
The inventory includes information on metadata, where available, to try to gain a better
understanding of how comparable the different data sets really are. The availability of
metadata varies from source to source, from virtually none to detailed papers describing every
aspect of the source, definitions and methodology. The lack of standards in the presentation
of metadata, and the availability of more detailed information only in the national language
often made the task of understanding the metadata more difficult, and may have contributed
to any errors in interpretation. This highlights the need for the uniform application of metadata
standards to help data users to better understand differences in data, particularly when
making international comparisons.
Whilst international comparisons can be problematic, some countries have several data-sets
available, based on different sources, which often give rather different measures of business
start-ups at the national level. The assumption in this project is that any variation between
sources relating to the same country must be purely methodological, i.e. linked to differences
in definitions, coverage, thresholds, or any of the other factors affecting comparability
identified in Section 3. This assumption has been tested on data for several countries (see
Annex 4), where it has proved generally possible to explain differences in data in terms of the
methodology used to produce them.
2.2 Other International Comparisons
Before starting to compare data for different countries, it is useful to see what can be learned
from previous work in this area. There have been several attempts over recent years to
provide internationally comparable business start-up data, either by international
organisations with an interest in harmonised statistics, or by individual countries seeking to
benchmark their data in a meaningful way. Some of the main work in this area is summarised
below, with an assessment of the level of comparability achieved.
2
For example, the conflicting requirements of national and international users of United Kingdom data
are considered in detail in Vale and Powell (2003).
10
• Demography of Small and Medium-sized Enterprises (DOSME) – Eurostat
The DOSME project was funded by the European Union from the mid-1990s until 2003 to
produce data on business demography and factors affecting business success in twelve
central and eastern European countries
3
, as they made the transition to a market economy.
The project was based on a series of surveys, which effectively created several panels of
businesses over time, and allowed the study of start-up and exit rates, survival, and the
characteristics of the entrepreneur. The result was a firm-level dataset that, subject to
confidentiality constraints, provides a useful resource for research on the development of the
business economy in these countries during this transition period. Full information about this
project is contained on the DOSME web site -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm.
In terms of producing comparable data on business start-ups, this project was quite
successful in developing and applying standard methodologies. However, the survey-based
approach, differences in the administrative sources used, as well as coverage and general
data quality issues, do cause some problems. The final stage of the project included finding
ways to overcome some of these issues analytically, based on the variables available in the
dataset, and even managed a reasonably robust comparison of data with those from the more
recent Eurostat business demography project. It must, however, be remembered that the
DOSME project observed these countries during an atypical period in their economic
development.
• Firm-level Data Project – OECD / World Bank
This project attempted to create harmonised firm-level databases in ten OECD member
countries
4
, with the aim of using these to produce comparable data on business dynamics.
Researchers in each country were responsible for running standard analyses of their micro-
data, with the resulting aggregates being shared for further cross-country analyses. The
project is described in detail, along with some of the resulting analyses, in various papers
linked to the project home page within the OECD web site:http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html.
The data cover different periods between 1978 and 1998 depending on what was available at
the national level, with the widest coverage in the early 1990’s. They are based on a variety of
sources, and not all countries were able to produce start-up rates in line with the project
definitions, for example some countries were not able to use the standard threshold of one
employee. Comparisons with more recent Eurostat data have highlighted these and other
quality issues (e.g. Brandt (2004)), often linked to improvements to the coverage and
maintenance procedures of statistical business registers during the 1990’s.
The World Bank has recently funded work to extend this approach to cover a further fourteen,
mostly developing, countries
5
. This is documented in two papers by Bartelsman, Haltiwanger
and Scarpetta (Bartelsman et al (2004), and Bartelsman et al (2005)). Various threshold and
3
Albania, Bulgaria, Czech Republic, Estonia, Former Yugoslav Republic of Macedonia (FYROM),
Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic and Slovenia.
4
Canada, Denmark, Finland, France, Germany, Italy, the Netherlands, Portugal, United Kingdom and
United States
5
Argentina, Brazil, Chile, Colombia, Estonia, Hungary, Indonesia, Latvia, Mexico, Romania, Slovenia,
South Korea, Chinese Taipei and Venezuela
11
coverage issues that might affect data comparability, particularly for business start-ups, are
noted in those papers.
It could be argued that the original OECD firm-level data project was a little too ahead of its
time, and that the resulting data are subject to a number of comparability issues that could
not realistically be resolved at the time; indeed some of these have only recently started to be
resolved at the national level. Having said this, many of the analytical techniques used seem
to have been robust enough to give plausible results despite the limitations of the basic data.
Also, putting data issues to one side, the approach of distributed analyses of standardised
micro-data seems worth pursuing in any future projects of this nature, as it avoids data
confidentiality issues, and makes use of national knowledge about the data.
• Business Environment and Firm Entry – NBER / World Bank
This study (Klapper et al (2004)) is published as a Working Paper of the US National Bureau
of Economic Research (NBER), acknowledging financial support from the World Bank. It is
available on the NBER website athttp://www.nber.org/papers/w10380. It compares business
start up data for over twenty European countries using data, mainly on corporate businesses,
from the Amadeus database compiled by the private sector business data provider, Bureau
Van Dijk. The results are also compared to US data sourced from Dun and Bradstreet, though
comparisons may be affected by differences in the way the sources are compiled.
The results are broadly in line with other sources, though some results such as an average
start-up rate of 3.46% for Italy compared to 11.13% for Finland seem to be at odds with
Eurostat figures (8.35% and 7.48% respectively). This is almost certainly due to the restriction
to corporate businesses, and raises additional comparability issues related to variations in the
propensity of businesses to incorporate. This will differ between countries depending on the
cost and complexity of registration procedures, tax incentives, reporting burdens and possibly
even cultural factors. Variations in the extent of re-registration in national systems, for
example when a business changes its name, may also affect comparability.
• Eurostat Business Demography Project
This project brings together data for European Union countries (plus Norway and Romania)
on business start-ups (births) closures (deaths), survival and growth, produced by national
statistical offices using a common methodology. So far it has been run on a voluntary basis,
which has resulted in a lack of data for some of the larger countries, particularly Germany and
France, though it will soon become a legal requirement through the forthcoming revision to the
Structural Business Statistics Regulation.
In terms of data comparability, this is probably the most successful international project to
date, as the methodology to be followed at the national level is very detailed, and was tested
and refined using pilot studies. The methodology is based on the use of business register
data. The registers themselves are subject to a considerable degree of harmonisation due to
the existence for over ten years of a European Union regulation on statistical business
registers
6
, which requires minimum standards of contents and coverage. Unfortunately this
does not mean that the data can be considered fully comparable yet, as different national
thresholds affect the smallest size-classes, and matching procedures to separate pure births
6
Council Regulation (EEC) No 2186/93 of 22 J uly 1993 on Community co-ordination in drawing up
business registers for statistical purposes -http://europa.eu.int/eur-lex/lex/LexUriServ/LexUriServ.do?uri=CELEX:31993R2186:EN:HTML
12
from other creations are affected by the availability and quality of key matching variables, as
well as the use of different matching tools.
Data and summary methodology resulting from this project are available via the Eurostat web
site (http://epp.eurostat.cec.eu.int). A more detailed methodological manual has been
produced, but not yet been published
7
.
• Global Entrepreneurship Monitor
The Global Entrepreneurship Monitor (GEM) project collects data on various aspects of
entrepreneurship through a series of coordinated household surveys in a gradually increasing
number of countries world-wide. More information on the project and participants can be
found at;http://www.gemconsortium.org/. One of the key outputs of the GEM work is an
indicator of “Total Entrepreneurial Activity” (TEA)
8
, which measures those respondents who
have recently started a business, or have started taking steps towards setting up a new
business. The TEA index is therefore not strictly a measure of business start-up rates, but
should provide a reasonable indicator.
Figure 2.1 - Comparing GEM TEA Rates and Eurostat Business Start-up Rates
2000
0%
5%
10%
15%
0% 5% 10% 15%
Eurostat Start-up Rate
G
E
M
T
E
A
2001
0%
5%
10%
15%
0% 5% 10% 15%
Eurostat Start-up Rate
G
E
M
T
E
A
2002
0%
5%
10%
15%
0% 5% 10% 15%
Eurostat Start-up Rate
G
E
M
T
E
A
Source: GEM 2004 Global Report and Eurostat web site
7
Business Demography Recommendations Manual, Eurostat, latest draft December 2004.
8
This is also referred to as the “Early Stage Prevalence Rate” in the 2005 GEM report.
Correlation Coefficients
2000 =0.341
2001 =0.489
2002 =0.763
13
Figure 2.1 shows that the degree of correlation between the GEM TEA rates and start-up
rates for countries contributing to the Eurostat business demography project seems to be
increasing over time, possibly reflecting data quality and methodological improvements in both
sources. Eurostat start-up data are used here because they provide the most reliable cross-
country comparisons currently available. TEA rates could also be compared with business
start-up rates from other sources and countries, but the current lack of harmonisation of start-
up rate methodologies used by these sources would cause distortions and increase the risk of
misleading results.
The TEA rates are roughly comparable in magnitude to the Eurostat start-up rates in 2000
and 2001, but appear to drop in 2002, probably due to methodological changes in the GEM
data. The relatively small GEM sample sizes in many countries
9
may affect the reliability of
these data, however, the increasing degree of correlation between data from these sources
could be seen as a positive indicator of the quality of both data sets.
• National Benchmarking – Canada and New Zealand
Two papers have been identified that consider the international comparability of business
start-up data in the context of benchmarking national data. The first, Baldwin et al (2002),
looks at different sources of data within Canada, and explains the differences in terms of the
methodologies used. The paper then considers how these methodological issues could affect
international data comparisons.
The second paper, Mills and Timmins (2004), seeks to establish if business dynamics in New
Zealand are really as different to those of other OECD countries as previous studies have
indicated. It concludes that when measurement differences, particularly relating to the
coverage of very small businesses, are taken into account, the New Zealand data are not very
different to those of other countries.
Both of these papers are useful in identifying some of the reasons why existing estimates of
start-up data may not be comparable across countries, and have informed the development of
the factors of comparability set out in Section 3 below. They clearly show that comparisons of
data from different sources must include comparisons of the metadata.
9
GEM sample sizes increased from 2,000 to over 15,000 people per year in the UK during this period,
which could be expected to help improve comparability with register-based business data such as
those from Eurostat, however they remained stable at around 2,000 per year for each of the other
countries included in the charts in Figure 2.1.
14
15
3. Factors Affecting Comparability
This section of the report aims to identify the different factors affecting the comparability of
data on business start-ups, and to highlight the main issues involved. At first glance, the
number and range of factors that affect comparability can make the task of compiling
comparable data appear to be virtually impossible. The aim of this report is not to discourage
the reader from trying to make comparisons, but to explore in detail the factors affecting
comparability. If these are better understood, they may be more easily overcome, or it will at
least be possible to make more informed decisions about which ones have little enough
impact that they can safely be ignored.
J ust as comparability is often listed in typologies of the components of statistical data quality,
so it is possible to develop a typology of the factors affecting comparability. Looking at this in
another way, such a typology can also provide a list of the reasons why data may not be
comparable. Focussing specifically on the area of the international comparability of business
start-up rates, these factors can be defined either in terms of the numerator (the number of
new businesses), the denominator (the population or stock), or both (assuming the
denominator is based on businesses)
10
.
The approach of separating numerator and denominator factors is based on the study of
differences between data sources within countries (see Annex 4). This work clearly shows the
range of factors that can affect data comparability between sources that are attempting to
measure the same phenomenon for the same country. It also demonstrates that there is a
complex interaction between these factors.
The three charts in Figure 3.1 below are taken from Annex 4, where they, and similar charts
for nine other countries, are discussed in detail, and the reasons for the differences are
explained. They compare United States data from various sources, and demonstrate clearly
how start-up rate indicators that appear to be similar are actually quite different when they are
split into their components.
The Business Employment Dynamics quarterly data set is a clear outlier in terms of start-up
rates, though the annualised data from this source show that this is almost entirely due to
periodicity and data purity issues. The remaining data sources appear to give fairly
comparable measures of start-up rates, typically between 10% and 13%, though these mask
the differences in the populations of new and existing businesses used to derive these rates.
10
Business start-ups can also be measured in terms of employment creation rather than numbers of
new businesses (see Baldwin et al (2002)). This measure is less sensitive to the inclusion or exclusion
of very small units, but is more sensitive to the type of unit used (new establishments of existing
enterprises can be very large), and the inclusion of events other than pure births (which tend to involve
larger businesses). This approach is not considered further in this section for the purely pragmatic
reasons that more data are available on counts of businesses than on employment, and that
employment of new businesses can be rather difficult to measure accurately. It is, however, revisited in
Section 5 of this report, which considers possible supplementary indicators.
16
Figure 3.1 – A Comparison of Different Sources of Start-up Rates, New Businesses and
Business Populations in the US
a) Start-up Rates
0%
5%
10%
15%
20%
25%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
Business Employment Dynamics - Summed Quarterly Data Firm Size Data
Business Employment Dynamics - Annualised Data Statistics of US Businesses
Longitudinal Business Database OECD Firm-level Data
b) New Businesses
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Business Employment Dynamics - Summed Quarterly Data Firm Size Data
Business Employment Dynamics - Annualised Data Statistics of US Businesses
Longitudinal Business Database OECD Firm-level Data
c) Business Populations
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Business Employment Dynamics Statistics of US Businesses
Longitudinal Business Database OECD Firm-level Data
Firm Size Data
17
The typology approach has been followed below, resulting in a set of nine factors affecting the
comparability of business start-up rates, each of which is considered in more detail in
Sections 3.1 to 3.9. This typology has been developed based on reactions to earlier drafts
proposed in Vale (2005(a)), and Ahmad and Vale (2005).
Numerator factors:
• Purity – to what extent are “pure births” distinguished from reactivations and other
creations?
• Timing – at what point in the creation process is a start-up measured?
• Periodicity – over what period are start-ups measured, and how does this affect the
measurement of very short-lived businesses?
Denominator factors:
• Type of Population – businesses or people?
• Temporal basis – is the population measured at a specific point in time, or does it
consist of all units that were present at any time during a given period?
Factors affecting both:
• Source – are the data taken from a register, a census or a survey? How reliable is the
source?
• Units – what is the entity about which the data are produced?
• Coverage - to what extent are certain types of business included or excluded based on
specific attributes (e.g. economic activity or legal form)?
• Thresholds – what explicit or implicit size thresholds apply to the source?
Various other factors can be identified as affecting comparability of start-up data, such as the
size of national economies, demand and supply constraints, the impact of tax, subsidy and
other policies, the nature of the political system, and a wide range of other economic, political,
social and cultural factors. None of these factors relate to the data production methodology,
and many of them account for the sort of variation in data that users are really interested in.
Indeed if they were all eliminated, the data would be identical for each country, and of no real
use to anyone. For this reason, this report only focuses on the nine methodological factors of
comparability listed above. If these can be understood, and their impact reduced, data users
have a much better chance to observe the non-methodological factors in a less biased way.
3.1 Purity
It is often relatively easy to measure business entries, i.e. those businesses that are present
in a given period but were not present in the previous period. It is rather more difficult to
separate out pure births (sometimes referred to as creations ex nihilo) from entries due to re-
registrations, reactivations, take-overs and other demographic events
11
, i.e. those entries that
are either the continuation of an activity previously carried out under a different unit, or a
reactivation of a business that has been active in the recent past, but was dormant (or not
recorded) in the previous period. The term “purity” is therefore used to refer to the extent to
which business start-ups have been split into pure births and other entries.
11
For a typology of demographic events affecting businesses see Eurostat (2003).
18
The impact of separating pure births from other entries can be considerable. Figure 3.2 uses
data for France from the Agence Pour la Création d’Entreprises (APCE), which show that
around one third of all new businesses recorded by that source are not considered to be
creations ex nihilo.
Figure 3.2 Separating Pure Births from Other Creations in France
0
50
100
150
200
250
300
350
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
APCE Total Creations APCE Creations "Ex nihilo"
Business entries are usually derived from registrations with administrative sources, so can be
affected by administrative requirements to re-register in the event of certain changes, e.g. a
sole proprietor converting to a corporation. As legal and administrative requirements vary
considerably from country to country, and are likely to continue to do so, data on entries can
never be fully comparable between countries, whereas, at least in theory, data on pure births
can be.
Re-registrations, and other entries that are not pure births, can often be identified using data
matching techniques. A new unit that has a number of characteristics in common with a
previously existing unit, (e.g. name, address, economic activity, employees), is unlikely to be a
pure birth. Typically such matching will be automatic, or semi-automatic, based on rules or
algorithms to determine the likelihood that two units actually represent the same business in
the real world.
Methodology for matching to determine which creations are pure births has been developed in
the context of the Eurostat business demography project, though this has highlighted the need
to tune matching techniques to suit national data sources, and the danger of over-matching,
i.e. too many “false” matches. Experience in a number of countries shows that the larger a
business creation is (measured in terms of persons employed or turnover), the less likely it is
to be a pure birth.
Reactivations can also be difficult to deal with conceptually. A business that is dormant for a
few months (possibly due to seasonal activities) before re-starting would not be considered to
be a pure (ex nihilo) birth. However, if the period of dormancy was ten years or more, it would
be harder to argue that the business creation could be treated as a continuation of the
previous activity. A threshold may therefore be required. For the European Union this is
currently set at two years, whereas for data from the US Census Bureau longitudinal
database, all reactivations are excluded from the category of pure births, regardless of the
period of dormancy. The longer the period, the greater the delay in producing definitive data
19
on business closures, thus some sort of compromise is needed. This should be informed by a
better understanding of the reasons for dormancy, and the possibilities of adjusting for
reactivations using modelling based on historic data.
In countries where it is possible to link employers and employees over time, these links can
be used to help determine pure births. This method has been tested in New Zealand, where, if
at least 70% of employees appear to move from an old registration to a new one, it is
assumed that the new business is not a pure birth. Taken together with work to identify when
sites are transferred between businesses, this has resulted in around 20% of entries now
being confirmed not to be pure births. These businesses tend to be the larger entries,
accounting for around 60% of the employment attributed to entries (Mead (2005)).
Similar work in Canada is reported in Baldwin et al (2002), which also showed that using
linked employer-employee data from the Longitudinal Employment Analysis Program (LEAP)
file can reduce business start-up rates from an annual average of 18.5% to around 14.5%.
The fall was considerably more pronounced in terms of the employment attributed to start-
ups, which dropped from an annual average of 11.8% to just 4% (or 2.5% depending on how
and when employment was measured). These results are complemented in that paper by
survey data showing that firms entering the manufacturing sector by acquiring an existing
plant accounted for only 0.8% of plants, but 3.2% of employment, whereas those firms
entering manufacturing with a new plant accounted for 7% of plants, but only 2.1% of
employment.
Both studies appear to call into question the importance of business start-ups in terms of job
creation, demonstrating that where more advanced linkage techniques are used, the
employment that can be attributed to pure births is rather lower than previously thought. This
is backed up by findings in several countries participating in the Eurostat business
demography project, e.g. Cella and Viviano (2004). Pure births with more than 20 employees
seem to be quite rare in most countries, and tend to be limited to cases of inward investment,
a few labour intensive service activities, or manufacturing activities that have traditionally been
associated with high entry thresholds.
The figures from France, Canada and New Zealand clearly illustrate that any work to
distinguish pure births from other entries will result in lower start-up rates, therefore the
amount of such work undertaken should be considered when comparing data from different
sources. The potential impact on trends is less obvious. Total entries may well show similar
trends to pure births in the short term, but are more likely to be affected when administrative
sources and systems change.
3.2 Timing
This issue concerns differences in the point at which data sources record a business start-up.
This can vary from the time a person starts thinking about creating a new business to the time
a new business makes its first sales, reaches a certain financial or employment threshold, or
survives for a certain period. For some new businesses the time intervals between these
events is very short, for others it can be measured in years, whereas a third category do not
meet all the criteria, so could be measured as a start-up by one source, but not by another.
This third category demonstrate that sources that record start-ups at an early point in the
process tend to show higher start-up rates, particularly for very small businesses. These are,
of course, accompanied by higher exit rates.
20
Typically, the point at which a start-up is recorded is determined by the nature of the data
source. Surveys of people or households can measure intentions, administrative sources are
linked to more concrete legal or fiscal obligations, and surveys of businesses are typically
directed at those that have at least a certain level of economic activity.
There is a clear link here with the discussion on thresholds below, as sources with higher
thresholds are likely to record businesses at a later point in the start-up process than sources
with lower thresholds. For example, a business will only register with an administration
responsible for taxation of employee earnings when it takes on its first paid employee. This
could be some time after it has registered to pay sales or value-added tax.
It is also important to know whether certain sources allow pre-registration, i.e. where a
business can be registered in advance of actually starting activity. This is typically more
common for regulatory sources than taxation sources, but can happen for both. Ideally both a
registration and a start date are needed, but in practice it is usually necessary to use other
indicators of whether a business has really started such as tax returns, sales or employment.
A related issue concerns lags, i.e. the time difference between events taking place in the real
world, and being recorded in the data source. For statistical business registers the lags in
recording business start-ups depend on the source of the information, typically administrative
or tax registers. Figure 3.3 shows an analysis of business start-up lags for the British
statistical business register resulting from the use of value added tax (VAT) registration data.
Figure 3.3 – VAT Registration Lags Affecting the British Statistical Business Register
0%
5%
10%
15%
20%
25%
30%
-
5
00
5
0
1
0
0
1
5
0
2
0
0
2
5
0
3
0
0
3
5
0
4
0
0
4
5
0
5
0
0
5
5
0
6
0
0
Lag in Days
This chart shows that almost 80% of start-ups are notified within 100 days, and that very few
have a lag of more than a year
12
. Lags will obviously vary considerably depending on the
nature of the source and the frequency of updates. If the source records only the date of
notification, and start-up rates and lags are stable over time the impact will be negligible. If the
source attempts to record the actual start-up date, it will be necessary to either wait until the
impact of the remaining lags is insignificant before producing start-up data for a given period,
or to make adjustments based on the estimated effect of lags. Thus the main impact on
comparability due to lags will be in the most recent periods.
12
One reason for lags of more than a year is retrospective registration of businesses found not to have
declared their revenue to the tax authorities.
21
3.3 Periodicity
This issue concerns whether the data are sub-annual, annual, or less frequent. The majority
of the sources identified in Annex 2 concern annual data, though quarterly and monthly data
sets are available for some countries. In a few cases, data availability is linked to economic
censuses at intervals of five years.
For data with a periodicity of greater than one year it is difficult to construct start-up rates that
can be compared to annual data, as the proportion of short-lived firms that will be missed
increases considerably. In J apan, annualised average rates are calculated for the years
between censuses (Takahashi (2000)), but these mask the natural year on year variability
usually observed in start-up data.
If sub-annual data include counts of start-ups, they can simply be added to produce annual
totals, though these totals will be higher than those based on annual snap-shots due to better
coverage of businesses that survive for less than one year. If sub-annual start-up data are
only available in the form of birth rates, it is clearly more difficult to estimate the annual rate
without further information about the net change in the population.
Work to convert quarterly establishment start-up data from the Business Employment
Dynamics series produced by the US Bureau of Labor Statistics to an annual basis has
resulted in differences of over 40% between annualised start-ups and the sum of start-ups for
the four separate quarters. This is a result of both the removal of short-lived businesses, and
improvements to the purity of the start-up estimates by better linkage of establishments over
time, and is documented in Pinkston and Spletzer (2004).
This leads towards questions about the value of data for very short-lived businesses. Is a
business that only lasts for a month or two, perhaps with no employees, and possibly even no
sales, of any real interest? Would it be more meaningful to only consider start-ups that remain
active for at least a year, or some longer period? In terms of current data availability, often
based on annual snap-shots of the population of businesses, this becomes a rather difficult
question. Many of the businesses that are live for less than a year will be excluded altogether,
but those that, by chance, are live on the day the snap-shot is taken, will be included. This
could cause certain biases, for example a common reference date in a number of data sets is
31 December / 1 J anuary. Short-lived businesses with activities related to the Christmas
period are likely to be included, but, for the northern hemisphere, short-lived businesses with
certain tourism or agriculture-related activities could be under-represented.
Possible solutions include the recording of start-up and closure dates to allow a more
accurate measure of the period of survival, or only counting business start-ups that are
present in at least two consecutive periods. The use of dates is the more attractive and
flexible option, but it relies on the availability of accurate information. Linking the timing of the
start-up to an administrative event, such as coming into scope of an administrative source
might help, as the source is likely to record that date. Given the lack of harmonisation of sales
related taxation systems, administrative sources that record when a business takes on its first
paid employee are likely to be most appropriate in terms of international comparability.
Annual data may not be fully comparable if they refer to different periods. Typically the period
is the calendar year, but other periods such as March to March (United States) and J uly to
J uly (Australia) are also used. For strict comparisons on a calendar year basis, such data sets
would need to be apportioned between years, though in practice this may not be necessary if
start-up rates are fairly stable over time.
22
Finally, where data are annual, they may not reflect an exact calendar year. If the
observations are not taken on exactly the same day each year, there will be an impact on data
comparability both between countries and over time.
3.4 Type of Population
Two basic types of population can be used as the denominator for calculating business start-
up rates. The population of businesses is the most frequently used, however for some
countries and sources, particularly where household surveys are used to measure business
start-ups and entrepreneurship, the denominator can also be based on the human population.
Business populations can vary considerably in the way they are defined. Most of the issues
are covered in the sections on coverage and thresholds below, but one specific point to note
here is the extent to which the population includes non-active units. The requirement in the
Eurostat business demography methodology for population units to be active in terms of
having turnover and/or employment at some point during the reference period is rather more
restrictive than taking, for example, a count of all current registrations.
Both types of population raise potential issues for international comparability, particularly
where there are large differences in the structure of the population between countries. For
example, using the total human population of a country as a denominator is likely to give
higher start-up rates for countries with a higher proportion of the population considered to be
of working age, than those with higher proportions of children or retired people.
It may also be necessary to have some knowledge about under-coverage due to factors such
as illegal immigration and undeclared workers to either make informed adjustments to the
population, or to be able to safely assume that the impact of under-coverage on comparability
is negligible. The issue of undeclared workers is closely related to underground businesses,
that is, those businesses that are active but which are not registered to avoid tax payments or
compliance with labour laws for example, an issue that affects both the numerator and
denominator.
Another approach is to use the population of working age, or of those people considered to be
economically active, if these populations can be defined consistently across countries.
However, even if a consistent definition is used, structural differences in national economies,
political or cultural differences (e.g. the rate of participation of women in economic activity, or
the ease with which a new business can be established
13
), can affect comparability. In such
cases however it might be preferable not to try to correct for such differences, as they,
arguably, form part of the phenomena to be observed, nevertheless, it is always helpful to be
aware of their potential impact when trying to interpret data from different countries.
For some specific purposes other sub-sets of the human population may be used, an example
of this is the use of the population of unemployed persons for analyses designed to illustrate
the extent to which unemployment encourages entrepreneurship. Great care is needed to
accurately interpret data using such sub-populations, as, in practice, only a proportion of new
businesses are actually started by people who were previously unemployed.
13
The World Bank and the International Finance Corporation, have financed work on an indicator
ranking countries on this topic, see:http://www.doingbusiness.org/EconomyRankings/Default.aspx?direction=asc&sort=2
23
3.5 Temporal Basis
If the denominator is based on the human population, point in time estimates are generally
used, i.e. the number of people on a specific date. Where it is based on a business
population, two variants have developed. The traditional approach, followed in most of the
data sets studied, is to use point in time business population data. This is consistent with
human demography, and allows a “stocks and flows” approach to business demography.
An alternative approach is to use the population of businesses that were considered to be in
scope at any point during a given reference period. This approach is favoured by Eurostat in
their business demography data collections, partly because it ties in with the approach used to
collect financial variables (e.g. turnover for a given period), and partly because it was thought
at one time to be easier for countries that did not have accurate birth dates for units in their
business registers.
It is clear that a live during period population will be larger than one on a point in time basis.
The extent of the difference will depend on various factors, but mainly on the length of the
period, and the degree of churn (i.e. entries plus exits) in the business population. As a result,
data compiled using a point in time population will not be directly comparable with those
based on a live during period approach.
One further comparability issue with the live during period approach is that a proportion of
business entries will be due to new businesses taking over the activities from businesses
recorded as exits
14
. Technically, many of these cases should be considered as the continuity
of a previous business, and should not be recorded as entries and exits. However, as most
data sources are based either directly or indirectly on registrations and de-registrations with
administrative or tax sources, it is unlikely that all such take-over cases are recorded as
business continuity, particularly for small businesses.
This will inevitably result in a certain amount of duplication in live during period populations, as
such businesses will appear in them at least twice. The extent of duplication will vary from
country to country and between sources, depending on the nature of the source and register
maintenance procedures. This, in turn, will affect the comparability of indicators based on live
during period populations.
There is, however, also a danger with the point in time approach, in that those short-lived
businesses discussed in the section on periodicity above, that enter and exit in the period
between two reference points may not be included in counts of start-ups, or the associated
business populations. This risk is theoretically removed using the live during period approach,
but in practice is only really solved for either approach by holding accurate birth and death
dates, the recording of some measure of activity (e.g. turnover), or frequent observations of
the whole population.
It is often possible to estimate a live during period population by adding the total number of
business entries during a period to the point in time estimate for the start of that period.
Similarly a point in time population can be estimated from live during period data, though
movements into and out of scope, and the degree of duplication in live during period
populations, can affect such estimates.
To illustrate this, point in time and live during period populations of businesses can be broken
down into a number of components, which can then be re-aggregated in different ways to give
different types of population estimates. The basic components are shown in Figure 3.4 below.
14
Around 15% in the French data shown in Figure 3.2.
24
Figure 3.4 - A Simple Model for Business Populations
Key: PA
t
=The population at the start of period t
PB
t
=The population at the end of period t
S
t
=businesses present in both populations (i.e. “survivors”)
L
t
=businesses that are in population PA
t
, but not PB
t
(i.e. “leavers”)
J
t
=businesses that are not in population PA
t
, but are in PB
t
(i.e. “joiners”)
J L
t
=businesses that are not present in PA
t
or PB
t
, but would be present in an
intermediate population (i.e. they join and leave within period t)
The population of businesses considered in scope at the start of the period (PA
t
), sometimes
referred to as the opening stock, can be defined as: PA
t
=S
t
+L
t
. Similarly the population at
the end of the period (PB
t
), or closing stock, can be defined as: PB
t
=S
t
+J
t
. Businesses in
the sub-set J L
t
do not appear in either population.
The population of businesses live at any time during period t (P
t
) can be defined as: P
t
=S
t
+
L
t
+J
t
+J L
t
, or by substitution as: P
t
=PA
t
+J
t
+J L
t
, or: P
t
=PB
t
+L
t
+J L
t
. Thus to convert
from a point in time to a live during period population, it is necessary to know, or have
reasonable estimates for J L
t
and either L
t
or J
t
. In practice, J L
t
is rarely available from
published data sources, and such businesses are usually ignored as they are not present in
PA
t
or PB
t
. The size, and hence the importance of J L
t
will depend on the length of period t. If t
is one month, it is relatively safe to assume that J L
t
is very small. If PA
t
and PB
t
are derived
from economic censuses with a five year interval, however, J L
t
will be much larger. These
relationships assume that there is no duplication within P
t
or between J
t
and L
t
.
It is possible to produce estimates for J L
t
, and to use these as a basis for converting
population data from live during period to point in time, or vice versa, if duplication is assumed
to be negligible. This approach is considered in Annex 3, which explores in much more detail
the relationships between different types of business population.
As stated above, a live during period approach will result in a higher denominator and lower
start-up rates. Typically for most OECD member countries start-up rates based on live during
period business populations are between 1% and 2% lower than those based on point in time
populations. Thus care must be taken in any comparisons that data collected on different
bases are not mixed. The point in time approach is conceptually simpler, and is less affected
by duplication issues, so is more likely to result in comparable data than the live during period
approach.
t
S
t
L
t
J
t
J L
t
PA
t
PB
t
25
3.6 Source
The main source for publicly available data on business start-ups is usually some sort of
register, either an administrative register maintained by a tax or regulatory agency, or a
statistical business register maintained by a national statistical institute. The main advantage
of this sort of source is usually comprehensive coverage of the population of interest, though
systematic biases may also be present due to the sort of coverage and threshold issues
identified below.
In theory, census data can be at least as good, and sometimes better, if they have less scope
restrictions, but the cost of running a census of businesses every year makes this approach
unrealistic for most countries. Data from less frequent censuses may still be of interest, but as
discussed in the section on periodicity above, they raise major comparability issues.
Survey data have also been used by some countries, most notably in the DOSME
15
project for
countries of Central and Eastern Europe. This approach can be useful when registers are not
sufficiently developed, and has the advantage of being able to collect more information on
entrepreneurship than is available from other sources, but it also suffers from the usual
constraints of survey errors and sample size limitations when detailed data breakdowns are
required.
The reliability of the source needs to be taken into account. This takes us back to the
components of the quality of statistical data, which have been well documented elsewhere
16
,
but it is clear that data from a comprehensive, frequently updated statistical business register
are likely to be more reliable than those from a small scale survey or study. The quality of the
data in the source clearly has an impact on most of the other factors of comparability identified
here, for example poor quality information on economic activity will have an impact on the
comparability of coverage.
It is also often the case that data from an official source (e.g. a national statistical institute) are
at least perceived to be more reliable than those from a commercial organisation. However,
this sort of generalisation is not always helpful, and a detailed understanding of the respective
methods used is necessary to make an informed judgement.
3.7 Units
The notion of a “business” is rather vague. Statistical institutes around the world have
historically tried to define the units for business statistics based on the sources of
administrative data available to them. The starting points are typically the unit that has some
sort of legal or tax obligation, often referred to as a “legal unit”, and the unit that corresponds
to a physical location from which a business operates, often referred to as a “local unit” or an
“establishment”.
Over time there have been attempts at the international standardisation of these units,
particularly in the European Union, where a regulation on statistical units
17
has gone part of
15
Demography Of Small and Medium-sized Enterprises – see:http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
16
For example:http://forum.europa.eu.int/Public/irc/dsis/qis/library?l=/public&vm=detailed&sb=Title
17
Council Regulation (EEC) No 696/93 of 15 March 1993 on the statistical units for the observation
and analysis of the production system in the Community (Official J ournal of the European
Communities No L 076, 30/03/1993, p. 1),http://europa.eu.int/eur-lex/lex/LexUriServ/LexUriServ.do?uri=CELEX:31993R0696:EN:HTML
26
the way towards harmonising the units used, and, has at least succeeded in harmonising the
terminology. Thus data from European Union countries will refer to enterprises, local units or
enterprise groups in a basically consistent way. There have been proposals to study the
demography of local units and enterprise groups, but, at least for now, business start-up data
for these countries are usually at the enterprise level
18
.
The enterprise is defined for European Union countries in the statistical units regulation as
“the smallest combination of legal units that is an organisational unit producing goods or
services, which benefits from a certain degree of autonomy in decision-making, especially for
the allocation of its current resources. An enterprise carries out one or more activities at one
or more locations. An enterprise may be a sole legal unit.”
Unfortunately it has been demonstrated that this definition is not always applied consistently
(e.g. in Herczog et al (1998)), particularly for more complex enterprises (e.g. those with more
than one legal unit) so it can not be assumed that data on units labelled as enterprises are
fully comparable in practice. This is largely due to differences in the legal, administrative and
tax frameworks across countries. A business that is organised as a single legal unit in one
country might prefer to organise itself into several legal units in another country to optimise
the way it interacts with government, employees and the market. For statistical purposes it is
necessary to recognise both as the same sort of entity, though creating the necessary
statistical structures through business profiling is a task that is difficult to automate, so is
therefore very expensive.
Outside the European Union there is a much greater freedom in terms of the choice (and
terminology) of units. In the United States, the establishment, which is closer to the European
local unit, is the main unit used for business statistics purposes. The term “firm” is used for an
aggregation of establishments under common control in a given geographic area, or sharing a
particular economic activity. Enterprises are defined as “business organizations consisting of
one or more domestic establishments that were specified under common ownership or
control”
19
, thus making them closer to the European concept of the enterprise group. Similar
terminology and definitions are used in Canada, though the term “business” is sometimes
used instead of “firm”. It is noted in Baldwin et al (2002), that “international studies must
recognize that the level at which a “firm” is defined varies across countries”.
The term enterprise is also used in most other OECD member countries, with slight variations
in the definition. It is defined in the System of National Accounts
20
, the key international
methodological framework for economic statistics, as “an institutional unit in its capacity as a
producer of goods and services; an enterprise may be a corporation, a quasi-corporation, a
non-profit institution, or an unincorporated enterprise.” In practice, the unit referred to as the
enterprise is often equivalent to, or very closely linked to, the national definition of the legal
unit. For a more detailed study of the different types and definitions of units used, see Choi
and Ward (2004).
Despite all of the above differences, it is likely that the vast majority (often at least 95%) of
business start-ups have a very simple structure, with just one site. This means that, in most
cases, all of the units above have a one to one relationship, and are in fact different views of
the same entity.
18
Several countries are, however considering the potential of local unit / establishment data to help
determine enterprise continuity.
19
US Census Bureau -http://www.census.gov/csd/susb/defterm.html
20
The System of National Accounts (1993) is promoted by the United Nations Statistics Division, and is
available via their web site:http://unstats.un.org/unsd/sna1993/introduction.asp
27
Unfortunately, it is not quite as simple as might appear from the above paragraph to compare
start-up rates for enterprises and establishments. There are two main complicating factors.
The first is that the total population of active enterprises will typically have higher proportions
of multi-site and complex businesses than the population of enterprise start-ups, thus if
enterprise data are to be converted to an establishment basis, the denominator needs to be
increased to take account of this. How much of an increase is likely to depend on a number of
factors including fiscal policy and the economic size and geography of the country. For the
United Kingdom this would reduce start-up rates by up to 2%. The second factor works in the
opposite direction, because a proportion of establishment start-ups will be new sites of
existing enterprises
21
. These would need to be added to the numerator, increasing the start-up
rate by up to 3%. The net result is therefore likely to be that establishment start-up rates are
slightly higher than those for enterprises.
3.8 Coverage
The coverage of data on new businesses and the business population depends heavily on the
source. In most cases this is a national statistical business register. If this register does not
include all legal forms or all economic activities, it logically follows that the data on new
businesses will have at least the same restrictions. Sometimes, even if the register does
include certain categories, there may be reasons for excluding them from counts of new
businesses. These reasons may be linked to quality concerns, the policy of the statistical
institute, customer requirements, or just tradition.
Categories most frequently considered to be out of scope in terms of economic activity are
agriculture, forestry, fishing and public administration. Units with the legal forms of central or
local government are also often excluded. The existence of a number of different
classifications of economic activity and legal form further complicates matters, as specific
categories of units may be treated differently according to the classification system used.
Fortunately the examples of frequently excluded categories above are also areas where
international classification systems are relatively well harmonised.
The treatment of businesses that move into and out of scope needs to be determined and
specified. The Eurostat approach attempts to exclude entries due solely to changes in
economic activity or other characteristics from data on pure births, whereas this distinction is
not necessarily made in other data sets. The comprehensiveness of the source obviously has
a major bearing on the ease of identifying such cases.
As is the case for units, the greatest degree of harmonisation in coverage, at least in theory,
exists between the Member States of the European Union, mainly due to the minimum
requirements set out in a regulation on statistical business registers
22
, which is currently being
revised with the aim of extending and further harmonising the scope of these registers.
Despite this, the data on business demography currently published by Eurostat has one of the
most restricted scopes of the data sets studied. Economic activities such as health, education
and personal services, and all non-market legal forms are excluded from most analyses.
These exclusions are largely driven by data quality concerns, which suggest that the actual
21
Between 19% and 28% depending on the year based on comparisons of data on births at original
locations (new firms), and secondary locations (new sites of existing firms) from the US Small Business
Administration - Seehttp://www.sba.gov/advo/research/data_uspdf.xls, worksheets dyn_00 to dyn_02.
22
Council Regulation (EEC) No 2186/93 of 22 J uly 1993 on Community co-ordination in drawing up
business registers for statistical purposes -http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/2186-93en.htm
28
level of harmonisation of business registers is still somewhat below that required by the
regulation.
There is a tendency when comparing data that differ in scope to look for the lowest common
denominator, i.e. the core set of data that are available for all sources. This can, however, be
problematic in some cases. For example, data from the United States typically exclude railway
operators, a category that is not always readily and separately identifiable in data from other
countries. In cases like this, alternative approaches could include modelling for the missing
categories, or simply ignoring minor scope exclusions in some sources if their impact is
considered to be trivial.
Where a population of businesses is used as the denominator for start-up rates, it is obviously
better to try to define this in as close a way as possible to that used for new businesses, i.e.
the coverage should be the same for both in terms of economic activity, legal form and other
criteria. Any differences will increase the possibility of other changes having an impact on the
birth rate
23
.
By definition, administrative and statistical registers, in all countries, will exclude businesses
operating exclusively in the non-observed or informal economy. They will also understate size
variables for businesses operating partially in this way. For developed countries, the economic
importance of such businesses is generally not thought to be significant, particularly in terms
of total economic activity. These businesses may, however, be of interest in the context of
measuring entrepreneurship or determining small and medium-sized enterprise (SME) policy.
It is therefore useful to have reasonably reliable estimates of the impact of such businesses,
perhaps from comparing labour force and business employment survey statistics or consumer
expenditure and declared business income.
Variations in geographical coverage may also affect within country comparability. This is most
likely to be a problem when there are autonomous or detached territories that may be
included in one source but not another, or when there has been a boundary change. The most
notable examples found in the course of this project are German data that exclude the former
East Germany, and United Kingdom data that exclude Scotland and Northern Ireland. In both
cases, start up rates for the included areas are, according to other sources, different to those
for the excluded areas. Thus these sources are not strictly comparable with other national
sources, and are not really suitable for international comparisons, as they are not fully
representative of the national situation.
3.9 Thresholds
There are no clearly defined, internationally agreed minimum size criteria for business activity.
Most data from the United States include only businesses with employees, whereas certain
J apanese data, and some international comparisons,
24
include only corporate businesses.
These sources therefore contain only a limited proportion of smaller businesses.
The European Union requires that all businesses with a labour input of at least one person
half-time are included in statistical business registers, and recommends covering smaller
businesses if possible. Some countries require all businesses to be registered regardless of
23
For example, if the population used for the denominator includes all legal forms, but the data on new
businesses used for the numerator exclude central and local government units. A re-organisation of
local government that creates many new units in that sector would have the effect of increasing the
denominator but not the numerator, thus artificially reducing the birth rate.
24
E.g. Klapper et al, (2004)
29
size, but even these are unlikely to record very low levels of business activity such as
individuals who occasionally sell second-hand or surplus goods to neighbours, via markets, or
through internet auction sites.
Some of the smallest “businesses”, particularly those with a labour input of less than one
person half-time, may be registered, but of little interest statistically. Hobby businesses
operated for reasons other than profit maximisation, and the provision of goods or services for
a few hours per week to supplement a main income are examples of this. In Volfinger (2004)
the statistical relevance of Hungarian enterprises with a turnover of less than one thousand
Euros is questioned. These accounted for 9% of active enterprises in 2002.
Thresholds can be helpful in terms of excluding such types of businesses, if they can be
applied uniformly across countries. An alternative can be to ensure that data have a strong
size dimension with classes based on quality and comparability criteria, so that certain
classes can be flagged as less comparable than others. Similarly, thresholds can provide a
route to exclude “pseudo-enterprises”, sometimes also referred to as “false self-employed”
where a person acts as an employee of an enterprise, in that they effectively work for that
enterprise every day over a long period of time, but for legal or tax purposes he or she is
technically self-employed. These issues are considered in more detail in Vale and Powell
(2002) and Vale (2005(b)), and their impact on European Union data in Brandt (2004).
In practice, thresholds are likely to be determined by the administrative sources that supply
data to statistical business registers. In many cases, data from sales or value added tax
registrations are used, with thresholds varying from zero up to GBP 60,000 in the case of the
United Kingdom. Where higher thresholds exist, data are often supplemented from other
sources to mitigate the impact, so it is often impossible to define the actual threshold applying
to a particular data set in terms of a single variable.
Particular care should be taken with thresholds related to sales or value added, as it is quite
possible in certain economic activities, e.g. software development, for a business to have
employees but no sales for a year or more, while it is developing products. An additional
complication is that monetary based thresholds are affected by differences in inflation and
fiscal policy at the national level, both of which can be expected to affect comparisons of birth
rates across countries and over time.
Thresholds relating to labour input are often more appropriate, but again it is important to
know how it is measured, e.g. in terms of wage-related costs, head counts of employees, or
full-time equivalents, as this could also have an impact, albeit probably small, on
comparability.
The quality of size variables can have a considerable impact on comparability when
thresholds are used. Unfortunately the quality of data is often lowest for relatively small and
new businesses, the categories that are often of the most interest. The methods of allocating
size (and other) variables in statistical systems, in the absence of full information on certain
businesses, can vary considerably. Some attempts to standardise these processes have been
introduced in the Eurostat methodology, including the use of turnover per head ratios to
estimate missing size variables.
The impact of thresholds varies depending on the use of data, and is usually much lower
when measuring economic or financial variables than for those based on counts of
businesses. It may also be the case that data subject to different thresholds can display the
same trends, even if those trends are less marked and the levels are different. This is
illustrated by the graph in Figure 3.5, showing two sources of data on business start-up rates
for the United Kingdom.
30
Figure 3.5 – UK Business Birth Rates – A Comparison of Data Sources
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
1995 1996 1997 1998 1999 2000 2001 2002 2003
Value-added tax registrations - Source: Small Business Service
New Businesses - Source: Barclays
In Figure 3.5, the source with the lower threshold, Barclays, shows a higher level and greater
volatility, as might be expected given the typically more dynamic nature of the smallest size
classes. The Small Business Service data are based on value-added tax registrations with a
threshold that has remained more or less constant in real terms at around GBP 60,000 in
2005 prices over this period. They clearly resemble a smoothed version of the Barclays data
series, albeit at a lower level. It would therefore appear to be possible to model one series
from the other. Whilst the birth rates themselves are not directly comparable, the underlying
data might be considered sufficiently comparable for some purposes. Also, the fact that both
sources show similar trends over time helps to validate the quality of the data sets.
31
4. Methods to Improve Comparability
This section is concerned with the extent to which existing data can be made more
comparable by performing various transformations. It follows three worked examples, the first
comparing data from the United States and Eurostat, the second comparing the Eurostat data
with experimental estimates from Australia, and the third comparing national data from France
and Germany with Eurostat data for other European countries. The transformations are made
on the basis of information available in the existing metadata, in some cases this has been
supplemented through contacts with those responsible for the source.
The approach of transforming existing data sets to try to improve comparability is not ideal,
and is unlikely to result in perfectly comparable data. These examples show that it can,
however, lead to some improvements in comparability. It should be seen as a short-term
measure, whilst waiting for the results of longer-term improvements such as the international
implementation of methodological standards.
4.1 Example 1 – United States and Europe
An obvious first step when looking at data from different countries, each with several sources,
is to use the comparability factors in Section 3 above to determine which sources give the
best trade-off in terms of the level of comparability already present and data quality. Thus, for
the first example, a comparison of data from the United States and Europe, it would seem
logical to use the Firm Size dataset from the US Small Business Administration, and the data
from the Eurostat business demography project. These sources use a similar unit, and both
define populations on a “live during period” basis. Figure 4.1 shows a basic comparison of the
raw data.
Figure 4.1 – A Comparison of US and Eurostat Business Start-up Rates
0%
2%
4%
6%
8%
10%
12%
14%
16%
U
n
i
t
e
d
S
t
a
t
e
s
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
N
e
t
h
e
r
l
a
n
d
s
P
o
r
t
u
g
a
l
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
1998
1999
2000
2001
Sources: United States – Firm Size Data – Small Business Administration
EU Mean – Mean start-up rate for the European countries shown
Other countries – Eurostat (The full Eurostat data set includes several other countries,
but only those for which data are available for at least three of the above years are
shown)
32
The data seem to indicate very little difference between start-up rates in the United States and
the European countries shown. However, there remain several methodological differences
between the US and European data. Perhaps the most important is that the US data only
include employer firms, i.e. businesses with at least one employee. The Eurostat database
includes a breakdown by size class, with a category for zero employee enterprises.
Subtracting this category from both the births and the population of active enterprises
therefore gives an estimate of start-up rates for employer businesses, as shown in Figure 4.2.
The data in Figure 4.2 show considerably more variation within Europe, and result in much
lower EU mean rates. Unfortunately, this comparison represents a backwards step in
comparability compared to Figure 4.1 for most of the countries shown. This is because, in
trying to correct for thresholds, new problems of coverage have been introduced. The start-up
rates for the European countries now only include those businesses that have employees
from the start. They do not include businesses that start with no employees, and then take on
employees as they expand. These businesses are, however, captured in the US data. This
also explains some of the increased variability in the European data, as, for example,
coverage of non-employer enterprises is much higher in Italy than in the United Kingdom or
Luxembourg, due to higher size-thresholds in the data sources for the latter two countries.
This makes it more likely that new enterprises will be identified before they take on employees
in Italy, and will thus be missing from the start-up rates shown in Figure 4.2.
Figure 4.2 – Start-up Rates for Employer Businesses – A Backwards Step?
0%
2%
4%
6%
8%
10%
12%
14%
U
n
i
t
e
d
S
t
a
t
e
s
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
N
e
t
h
e
r
l
a
n
d
s
P
o
r
t
u
g
a
l
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
1998
1999
2000
2001
Sources – as for Figure 4.1 above
Thus, the conclusion from Figure 4.2 is that merely removing the non-employer business
births from the Eurostat data causes extra distortions, and certainly does not improve
comparability. This clearly shows the danger of attempting to improve comparability of existing
data analytically without a proper understanding of the complex interactions of comparability
factors. However, in the context of this worked example, it is still possible to see the
adjustment made in Figure 4.2 as a step towards more comparable data, because, even
though it reduces comparability, it also opens up several possibilities to improve comparability
above the initial level in Figure 4.1.
One such possibility to improve comparability would be to determine the proportion of start-
ups in the US that previously existed as non-employer businesses, and remove them from the
33
US start-up rates. Data are not currently available to make this adjustment, but may result
from preliminary work to link employer and non-employer universes reported in Davis et al
(2005).
Alternatively, and perhaps as an interim measure, a study of cohorts of non-employer births
could be carried out in several European countries, to see how many subsequently became
employers, and how long it was before they made this transition. The results could then be
used to model the missing data, raise the European estimates, and hence to improve
comparability with the US data. The proportion of businesses making the transition, and the
timing, are likely to vary between countries and over time, so this approach is probably most
suitable for countries that want to make one-off comparisons with the US, rather than for wider
cross-country comparisons over time.
A better solution, however, given the considerable variability of European data in Figure 4.2,
would be to define the business population for those countries so that it only included
employers, and measure entries into that population (as recommended in the OECD Business
Demography Framework). The Eurostat data do not currently support this approach, but it
would be relatively easy to adapt the current Eurostat methodology to produce the necessary
figures.
It may still be necessary to interpret any resulting figures with care, as they could be affected
by variations in the propensity to incorporate between countries. Most new businesses start
as either a corporation or a sole-proprietorship. In the case of a corporation, the entrepreneur
is normally considered to be an employee, whereas in the later, he or she is not. Thus the
choice of legal form, which could be affected by national fiscal and administrative burden
considerations, can determine whether a start-up is included in the population of employers or
not. There are very few data on this subject at present, though the overall impact on start-up
rates is thought to be quite small. The methodologically purest long-term solution, therefore,
would be to define a lower threshold in terms of total labour input (e.g. 0.5 person), which
would be independent of issues of legal form. Unfortunately this is not really feasible for the
main indicator on business start-up rates, as it would require major (and hence expensive)
changes in several countries, particularly the United States.
Another methodological difference between the US and European data is purity. The Eurostat
methodology requires extensive matching to determine which start-ups are pure births,
whereas the metadata for this US source (Armington (1998)) make clear that no attempts are
made to track the survival of individual firms. The US data will therefore include an unknown
proportion of start-ups that are not pure births in the European sense. The proportion of start-
ups that are not pure births varies from source to source, as the propensity to re-register will
be determined by legal or other requirements that are usually source specific. In Europe this
proportion is usually around 20%, though French national data suggest figures of between 30
and 40%
25
. This proportion typically increases for larger businesses. Applying this to the US
data in Figure 4.2 would reduce start-up rates to around 6-8%, well below those of the United
Kingdom and Luxembourg, the countries for which the difference in thresholds compared to
the US data is likely to be least significant.
25
See Figure 3.2 in Section 3.1 above.
34
4.2 Example 2 – Australia and Europe
The second example considers the comparability of data from Australia and Eurostat. In this
case, the Australian approach is very similar to the Eurostat methodology in terms of
thresholds and some elements of purity. Figure 4.3 shows a basic comparison of the raw
data.
Figure 4.3 – A Comparison of Australian and Eurostat Business Start-up Rates
0%
2%
4%
6%
8%
10%
12%
14%
16%
A
u
s
t
r
a
l
i
a
E
U
M
e
a
n
S
p
a
i
n
I
t
a
l
y
L
i
t
h
u
a
n
i
a
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
S
l
o
v
a
k
i
a
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
2001
2002
Sources: Australia – Experimental Estimates, Entries and Exits of Business Entities – Australian
Bureau of Statistics.
EU Mean – Mean start-up rate for the European countries shown.
Other countries – Eurostat (The full Eurostat data set includes several other countries, but
only those with data available for both years, and no known coverage issues, are shown).
Figure 4.3 appears to show that Australian start-up rates are almost 1% above the mean for
the European countries shown. However, for a true comparison, it is necessary to make
adjustments to compensate for several methodological differences.
The main difference is that the Australian population data used as the denominator for these
start-up rates are on a point in time basis, whereas those from Eurostat are “live during
period”. The Australian point in time population is defined as the population from the previous
observation, adjusted for reactivations, plus entries in the previous period, minus exits in the
previous period. This relationship is shown in Figure 4.4, taken from the Australian
publication.
Figure 4.4 - The Relationship between Populations, Entries and Exits in the Australian
Data
35
Entries and exits are measured on a monthly basis, so any short-lived businesses that enter
and then exit between the two reference points, are included in both the entry and exit figures,
and thus cancel themselves out in this model. This aids comparison with the Eurostat data,
which also include such short-lived businesses.
Building on the approach introduced in Section 3.5 above (and developed further in Annex 3),
the components of the population can be shown as in Figure 4.5.
Figure 4.5 - The Components of the Business Population
PA
t
=The opening population for period t
PB
t
=The closing population for period t
B
t
=Births in period t that survive into t+1
D
t
=Deaths in period t that were live in t-1
BD
t
=Birth and Death within period t
The Australian relationship above can therefore be re-written as PB
t
=PA
t
+(B
t
+BD
t
) – (D
t
+
BD
t
). The Eurostat population, defined on a “live during period” basis, can be expressed as
PA
t
+B
t
+BD
t
(i.e. all businesses live at the start of the period plus all births during the
period). Thus adding the birth data to the opening population data for Australia will give a
good estimate of a “live during period” population
26
.
Table 4.1 – Converting Australian Start-up Data to a Live During Period Basis
Period Opening
Population
Entries Start-up
Rate
Live During Period
Population
Live During Period
Start-up Rate
2001-2 2,935,700 334,266 11.39% 3,269,966 10.22%
2002-3 2,941,666 329,907 11.21% 3,271,573 10.08%
Source: Authors calculations using data from the Australian Bureau of Statistics.
Other differences between the Australian and European data are that the Australian data
cover all economic activities, and use a J uly to J uly reference period, whereas the European
data have a more restrictive coverage, and are on a calendar year basis. In terms of the
coverage by economic activity, it is possible to use breakdowns by industry in the Australian
publication to get a close match to the European coverage (sections C-K of the International
Standard Industrial Classification (ISIC Rev. 2)).
26
Note – this approach does not work in cases where births and deaths are not measured on a regular
basis during the period, as, although it becomes easier to measure B
t
, it becomes much more difficult to
quantify BD
t
, and hence total births in the period.
t
D
t
B
t
BD
t
PA
t
PB
t
36
Table 4.2 – Adjusting for Coverage of Economic Activity
Period ISIC
Sections
Opening
Population
Entries Start-up
Rate
Live During Period
Population
Live During Period
Start-up Rate
2001-2 C-K 2,246,229 267,601 11.91% 2,513,830 10.65%
Other 689,471 66,665 9.67% 756,136 8.82%
Total 2,935,700 334,266 11.39% 3,269,966 10.22%
2002-3 C-K 2,299,104 248,833 10.82% 2,547,937 9.77%
Other 642,562 81,074 12.62% 723,636 11.20%
Total 2,941,666 329,907 11.21% 3,271,573 10.08%
Source: Authors calculations using data from the Australian Bureau of Statistics.
The final step is to adjust for the difference in time periods. This requires certain assumptions,
which risk introducing noise into the data, but should still result in a net improvement to
comparability in this case. The first assumption is that the population of businesses (for ISIC
sections C to K) at 1 J anuary 2002 is exactly halfway between the 1 J uly populations for 2001
and 2002. This gives a value of 2,272,667. The second assumption is that the entries are
following a linear trend, thus all things being equal, the number of entries on 1 J anuary 2002,
the mid-point of the period, should be equal to the annual total for 2001-2 divided by 365, i.e.
733.153. Similarly the number of births on 1 J anuary 2003 would be 681.734, and the total for
2002 would be ((733.153 +681.734) / 2) x 365, i.e. 258,217. These figures give a “live during
period” population of 2,530,884, and hence a “live during period” start-up rate of 10.20%.
Based on the available metadata, differences due to timing, type of population and source are
likely to be negligible, as well as not being easy to quantify. There still, however, remain
questions over purity and units. In terms of purity, the Australian data seek to identify where a
re-registration is really the continuation of an existing business, but having done this, all
remaining new businesses are treated as genuine entries. In the Eurostat methodology, a
further (probably quite small) proportion of these would be removed. These are new
businesses that do not meet the requirement of being a new combination of factors of
production (e.g. a new business formed by splitting off part of the activity of an existing
business).
Regarding units, the Australian data are based on tax registrations (i.e. legal units) rather than
enterprises. This is discussed in ABS (2005), where data for large and complex business
entities (for J une 2004), show that 67,000 tax units have been combined to form 30,000 “type
of activity units”. On the assumption that the European data correspond to the Eurostat
definition of the enterprise, it would therefore be necessary to reduce the Australian
population by around 37,000 businesses, and the entries by a rather smaller proportion (as
relatively few new businesses tend to be complex from the outset). However, as discussed in
Section 3.7 above, the enterprise definition is not yet applied fully and consistently in all
European countries, therefore the value of any adjustment to the Australian data is doubtful.
It is therefore perhaps easiest to assume that the differences in purity, which would reduce the
numerator of the Australian data, and units, which would reduce the denominator, are both
relatively small, and would largely cancel each other out.
The combined impact of the adjustments made to the Australian data is shown in Figure 4.6.
As we now only have one year for which data are reasonably comparable, it would be
dangerous to draw too many conclusions, though it is interesting to see that the start-up rate
estimate for Australia is now very slightly below the mean value for the European countries
shown (10.20% and 10.36% respectively).
37
Figure 4.6 – More Comparable Start-up Rates for Europe and Australia (2002 Data)
0%
2%
4%
6%
8%
10%
12%
14%
16%
A
u
s
t
r
a
l
i
a
E
U
M
e
a
n
S
p
a
i
n
I
t
a
l
y
L
i
t
h
u
a
n
i
a
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
S
l
o
v
a
k
i
a
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
4.3 Example 3 – France and Germany Compared to Other European Countries
Data on business start-up rates for France and Germany have not yet been published by
Eurostat, though both countries are now taking an active part in the Eurostat business
demography project, and will be supplying data for future publications. This example looks at
how existing national data for France and Germany could be compared to those from
Eurostat. Figure 4.7 shows a comparison of the raw data, which does seem to show that there
are comparability issues, particularly for Germany.
Figure 4.7 – Comparing French and German Start-up Rates with Eurostat Data for Other
European Countries
0%
5%
10%
15%
20%
F
r
a
n
c
e
G
e
r
m
a
n
y
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
N
e
t
h
e
r
l
a
n
d
s
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
2001
2002
Sources: France – Créations d’Entreprises, INSEE
Germany - Start-ups and Liquidations in Germany, Institut für Mittelstandsforschung, Bonn
EU Mean – Mean start-up rate for the European countries shown.
Other countries – Eurostat (The full Eurostat data set includes several other countries, but
only OECD members with no known coverage issues, are shown).
38
The French data are produced by the national statistical institute (INSEE). They are checked
for purity (removing around one third of all entries), and measure entries to the French
business register (SIRENE), regardless of the length of survival. Thus they can be considered
as comparable to the Eurostat data in terms of purity, timing, periodicity, type of population
and source. The main differences concern coverage, and the temporal basis of the population
of businesses used as the denominator. Differences in units are likely to be negligible, as are
differences in thresholds, except when compared to the United Kingdom and, to a lesser
extent, Luxembourg.
The coverage of the French data in terms of economic activities is wider than for the Eurostat
data. Some broad economic activity breakdowns are available via the INSEE web site
(www.insee.fr), which can be re-aggregated to match the coverage of the Eurostat data. If this
is done, the start-up rates increase by around 0.4%, assuming that the ratio of pure births to
other entries is constant across activities.
The French data use a point in time (1 J anuary) business population as the denominator. To
convert this to an estimate of the live during period population, it is necessary to add total
entries (i.e. pure births and other entries) to this population. This can be done on the same
basis as for the Australian data in Section 4.2. The results of these two conversions are
shown in Table 4.3
Table 4.3 – Adjusting French Data for Temporal Basis and Coverage
Period ISIC
Sections
Opening
Population
Entries Pure
Births
Birth
Rate
Live During
Period
Population
Live During Period
Birth Rate
2001 C-K 1,927,602 226,019 147,364 7.64% 2,153,621 6.84%
Other 490,348 42,600 27,775 5.66% 532,948 5.21%
Total 2,417,950 268,619 175,140 7.24% 2,686,569 6.52%
2002 C-K 1,964,295 224,722 147,642 7.52% 2,189,017 6.74%
Other 504,491 43,737 28,735 5.70% 548,228 5.24%
Total 2,468,786 268,459 176,378 7.14% 2,737,245 6.44%
Source: Authors calculations using data from the INSEE web site.
The French data, however, also identify a proportion of entries as taking over the activities of
existing enterprises (referred to as “reprises”). These account for around 15% of entries
(15.36% in 2001 and 14.85% in 2002), and will duplicate business activity recorded in the
opening population (or possibly in other entries). Thus, in accordance with principles of
business continuity, and to avoid artificially inflating the live during period population with
duplicates, they should be removed from that population. The result of this adjustment is
shown in Table 4.4.
Table 4.4 – Removing Duplication in the Live During Period Population
Period ISIC
Sections
Opening
Population
Corrected
Entries
Pure
Births
Corrected Live During
Period Population
Corrected Live During
Period Birth Rate
2001 C-K 1,927,602 191,302 147,364 2,118,904 6.95%
2002 C-K 1,964,295 191,351 147,642 2,155,646 6.85%
Source: Authors calculations using data from the INSEE web site.
Turning to the German data, these are based on notifications of new businesses for turnover
tax purposes, supplied via the statistical business register. The register data are adjusted by
the Institut für Mittelstandsforschung (IfM), to remove new sites of existing businesses,
registrations purely for tax or administrative purposes that do not result in new business
39
activity, and registrations for activities carried out as a second job by the entrepreneur.
Business registrations due to the movement of a legal unit from one district to another or a
change in ownership or legal form are also removed. These adjustments result in
approximately 62% of notifications being considered as real births (compared to the French
figure of around 65%). Thus the data can be considered to have been corrected for purity.
In terms of timing, periodicity, type of population and source, the German data can be
considered as comparable to those from Eurostat.
The German data use a point in time population, which can be converted to a live during
period basis in a similar way to the French data above. However, the German data do not
separately identify the different categories of entries that are not real births, thus the data in
Table 4.5 below add all registrations to the opening population, which may overstate the
population, and slightly under-estimate the live during period birth rate.
Table 4.5 – Adjusting German Data for Temporal Basis
Period Opening
Population
Entries Pure
Births
Birth
Rate
Live During Period
Population
Live During Period
Birth Rate
2001 2,920,293 728,978 454,700 15.57% 3,649,271 12.46%
2002 2,926,570 723,333 451,800 15.44% 3,649,903 12.38%
Source: Authors calculations using data from IfM and the German Federal Statistical Office.
The units used are effectively the sub-set of legal units that are considered to be economically
relevant. This is unlikely to have any real impact on the number of entries, but may mean that
the population of businesses is slightly overstated, again leading to a very slight under-
estimation of start-up rates.
The data cover all economic activities except the “liberal professions”
27
, most health services
and some insurance services that are not subject to turnover tax. This is still a slightly wider
coverage than the Eurostat data. Detailed breakdowns of economic activity that would allow
more exact comparisons are not currently available, though based on the evidence for France
above, the impact is likely to be small.
The population of businesses is subject to a threshold (16,617 Euros during this period),
which means that some smaller businesses are excluded. Smaller businesses typically have
higher entry and exit rates, thus the impact of this threshold is likely to be a slight under-
estimation of start-up rates when compared to all other European countries except the United
Kingdom (where the threshold was around 90,000 Euros).
Although it is not possible to quantify the impact of these factors accurately, it is clear that the
net effect will be a slight under-estimation of business start-up rates. The rates calculated in
Table 4.5 should therefore be seen as minimum estimates, but it is unlikely that the real
values are more than 0.5% higher.
The revised estimates of business start-up rates for France and Germany are shown in Figure
4.8. These can now be seen as rather more comparable with the Eurostat data. The German
rates are still above the mean of the Eurostat rates, but are no longer the highest. This could
be expected, as the German data include the former East Germany, which may have an
upward influence on the national rate, as start-up rates published by Eurostat for the former
communist countries of Eastern and Central Europe (e.g. Hungary) are generally higher than
27
The liberal professions can generally be defined as occupations requiring special training in the arts
or sciences. These include lawyers, notaries, accountants, architects, engineers and pharmacists.
40
those for Western European countries. The adjusted data for France are slightly lower than
the raw data, but still not the lowest for the countries present.
Figure 4.8 – More Comparable Start-up Rates for European Countries
0%
5%
10%
15%
20%
F
r
a
n
c
e
G
e
r
m
a
n
y
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
N
e
t
h
e
r
l
a
n
d
s
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
2001
2002
4.4 Summary
The three examples above show that although it is not always possible to be precise, it is
clear that adjustments to compensate for differences in specific factors can sometimes help to
improve comparability. It is also clear that in compensating for one factor, it is possible to
affect others, and to introduce noise into the data, thus adjustments have to be made with
care, based on a detailed understanding of the data sources and methods.
Having said that, adjustments based on estimates of the impact of a specific factor of
comparability can still help to determine whether differences in start-up rates are likely to be
significant or not. This is no substitute for having real, comparable data, but, in the first
example above, at least it should caution an analyst against making statements that US start-
up rates are definitely higher than those in the European Union.
41
5. A Harmonised Methodological Framework and Start-up Indicators?
The discussions on factors of comparability in Section 3, and possible ways to improve the
comparability of existing data in Section 4, lead to the conclusion that the best way to get
really comparable data across countries is to harmonise as far as possible the underlying
methodology. Thus what is needed is a standard methodological framework that can be
applied in all countries, and which leads to a set of indicators of business start-ups that can be
used with confidence for cross-country comparisons. This section discusses how this might be
achieved.
5.1 Towards a Harmonised Methodological Framework
The idea of developing a harmonised methodological framework for indicators of business
demography and dynamics is not new. It has been attempted with varying degrees of success
in some of the projects outlined in Section 2, though often either the focus has been on a
limited group of countries that already share certain characteristics (as in the DOSME project),
or a common legal framework for statistics (as in the Eurostat project), or the methodology
was not detailed enough to generate real comparability.
The OECD is in the process of developing a new framework, taking into account what has
worked and not worked in the past, as well as a more detailed knowledge and understanding
of the methodological issues than in many previous projects. As business start-up rates are
an important component of business dynamics, this report will also feed into the new OECD
framework. The approach of using factors of comparability introduced in this report is being
broadened to cover business demography as a whole, and to inform the decisions on the
preferred methodology.
5.2 Different Types of Indicators
Within this harmonised methodological framework it would be possible to envisage several
different types of business start-up indicators. The traditional version, showing the number of
new businesses as a percentage of the population of businesses, is clearly the key indicator
for business start-ups, but, as this report shows, it is also not a particularly easy indicator to
define in a way that results in fully comparable data across countries. A range of other
indicators have been proposed over the years, each of these has certain merits, but none
seem to offer a full solution to the problem of international comparability.
Several studies have argued that it is better to measure start-ups by only including those
businesses that survive for a certain length of time. For the OECD Firm-level Data project, that
period was at least one year, whereas the authors of Baldwin et al (2002) recommend using
periods of up to five years. They show that start-up rates from different sources in Canada
vary more in the short-run than in the long-run, and hence recommend that a longer-run view
should be used for international comparisons.
The Eurostat business demography project goes somewhat in the opposite direction, and
seeks to include all start-ups, no matter how short-lived they are, and attempts to tackle
comparability issues through the harmonisation of sources and methods. Despite this, there is
some evidence from the limited Eurostat data currently available that would seem to support
the view of Baldwin and colleagues. If birth rates are plotted against two-year survival rates for
the European countries for which data are available, there appears to be a fairly weak
negative correlation, indicating a limited increase in convergence over time. Figure 5.1 shows
data for births in 2000, where the correlation coefficient is -0.47.
42
Figure 5.1 – Birth Rate v. 2 Year Survival Rate for Births in 2000
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
4% 6% 8% 10% 12% 14% 16% 18%
Bi rth Rate
2
Y
e
a
r
S
u
r
v
i
v
a
l
R
a
t
e
Source: Constructed from Eurostat data
There are, however, a few problems with using long-run entry rates. The first is that few policy
makers would be prepared to wait for five years for data. The second is that in countries with
genuinely dynamic business populations perhaps fuelled by very low entry and exit costs and
a strong entrepreneurial culture, it is to be expected that short-run start-up rates would be
higher and survival rates lower, reflecting an increased degree of experimentation on the part
of entrepreneurs. Long-run entry rates are less useful for identifying the extent to which this
particularly interesting phenomenon varies between countries. The real challenge is to
separate this sort of genuine variation between countries from the noise in the data due to
methodological differences.
The third problem with using long-run entry rates is that they do not affect all of the factors of
comparability equally. They should resolve most timing issues and can help to smooth the
effects of differences in periodicity, thresholds, and possibly units, though it is difficult to see
how they will have much impact on the other factors. There is even a risk that they could
actually aggravate the impact of different temporal bases for the population in that data
compiled on a live during period basis are likely to have lower birth rates (due to the higher
population), and lower survival rates (due to the inclusion of more short-lived units) than data
compiled on a point in time basis, thus the long-run birth rates could potentially be more
divergent than those for the short-run.
Despite these potential problems, there is still a role for long-run entry rates in conjunction
with short-run data. If they are sufficiently comparable between countries, they can give
another view on business dynamics, thus helping to give a better overall picture of the real
differences between countries.
A different approach to business start-up indicators is to consider the impact of new
businesses in terms of employment creation. This is of great interest to policy makers and
researchers concerned with the impact of encouraging entrepreneurship. As noted in Baldwin
et al (2002), employment based measures are less influenced by differences in thresholds,
but much more sensitive to purity. This is because thresholds tend to affect the smallest
businesses, excluding those with no or very few paid employees, whereas new businesses
that are not pure births tend to have more employees. A new business created by the merger
of two large corporations, bringing together thousands of employees, could swamp data on
employment creation.
43
Given the interest in employment creation, however, it is still useful to have a measure of the
impact of business start-ups, as long as this measure is sufficiently reliable and comparable.
Thus the Eurostat approach of trying first to obtain harmonised data on births, then
complementing these with data on employment creation, seems worth pursuing.
In a few cases, business start-up rates have been calculated using a human population as the
denominator. This approach relies on a suitably harmonised definition of the population used,
and can be affected by various social and cultural factors as discussed in Section 3.4 above,
but produces an indicator that is perhaps better focussed on entrepreneurship propensity.
Most of the data sets studied in this report are at the level of the business, firm or enterprise,
which despite their definitional differences can generally be seen as the same unit as far as
business start-ups are concerned, given that the overwhelming majority of these have very
simple structures. The exceptions are the establishment level data available for the United
States, J apan and a few other countries.
Although establishment level data cause comparability problems, they also have clear
benefits in terms of studying business dynamics at the local or regional level, a topic which is
generating interest in various countries, notably the United States and the United Kingdom.
Thus indicators at the level of the establishment (or local unit in Europe) are worth
consideration, particularly if it is possible to determine whether new establishments are due to
pure enterprise births, other enterprise creations, or the opening of a new site by an existing
enterprise; although measures based on establishments do provide other complications
28
.
5.3 Proposed Indicators
Based on the findings of this report, it is clear that a single indicator of business start-up rates
is unlikely to meet all requirements, therefore a system based on a key indicator, supported by
a range of complementary indicators is proposed.
• The Key Start-up Indicator
The key indicator of business start-up rates should try to meet potentially conflicting
requirements. Firstly it should be meaningful and easy to interpret for non-specialists, thus it
should be based on concepts and methodologies that are as simple as possible. Secondly it
should be designed to maximise international comparability. This second requirement could
lead to a purist view that the indicator should be designed without any reference to existing
data sources, or a pragmatic view that it should be built around the data that are currently
available. The purist view is likely to delay the availability of comparable data, whereas the
pragmatic view would not necessarily lead to optimal methodological solutions. Thus the
challenge is to try to find an acceptable compromise between all of the different requirements
and views.
The best option for the numerator therefore seems to be a count of new businesses, and for
the denominator, the population of active businesses. New businesses should be split into
pure births and other creations, along the lines of the Eurostat definition as the rate of other
creations will vary between countries depending on national registration systems and
practices, whereas pure births are much more suitable for international comparability
purposes. As purity has a major impact on comparability, the key indicator should focus on
pure births, though information on other creations may have some value in terms of quality
28
For more detail see the OECD Business Demography Framework
44
assurance, and comparing the impact of national systems on the business community. The
method to determine pure births proposed by Eurostat, based on automatic matching and
limited clerical checking of large units seems suitable, and has the advantage of already being
in place in around half of the OECD member countries.
In terms of timing and thresholds, the point at which a new business takes on its first paid
employee seems to be the easiest to measure in a consistent way across countries. This does
not mean that non-employer businesses are of no interest, just that they should be considered
in a secondary indicator. This means that births are defined as entries into the population of
employers, regardless of whether the business previously existed with no employees or not.
An important issue to resolve concerns businesses that fluctuate between having employees
or not, either on a seasonal basis, or in response to market conditions. The simplest approach
is to consider a business that leaves and re-enters the population of employers within a given
period as being a reactivation, and therefore not a pure birth. Eurostat currently recommend a
two year threshold for reactivations, which has the advantages of being relatively easy to
implement, and that it provides definitive data on pure deaths more quickly than if a longer
period (or no threshold) is used.
In terms of periodicity and temporal basis, annual data seem most appropriate, based as
closely as possible on the calendar year. A comparison of point in time populations at the start
and the end of the year to determine entries and exits is the easiest approach to implement.
This allows the construction of a simple equation as used in Australia (see Section 4.2),
whereby the population at a particular point in time is defined as the population from the
previous observation, plus entries in the previous period, minus exits in the previous period.
This sort of stock and flow approach is analogous to that used for human demography, and is
easy for non-specialists to understand. The potential duplication issue raised in Section 3.5 is
also reduced, thus improving data comparability, if a point in time population is used.
The remaining question in terms of periodicity is the treatment of short-lived enterprises. It is
methodologically preferable to include all of these. This would require the use of dates to
denote when a start-up occurred, or regular (at least monthly) observations of the population.
The best source for business start-up indicators seems to be national statistical business
registers. The units and coverage of these registers are gradually becoming more
harmonised, particularly in Europe. The unit of interest is usually the “business”, but this
concept is not really defined in its own right, hence the use of the enterprise, as defined in the
System of National Accounts seems most appropriate
29
. For the purposes of this indicator, the
definitions of the enterprise in the International Standard Industrial Classification (ISIC), and
the European Union regulation on statistical units, should be regarded as sufficiently similar.
Coverage should be defined as all “market” enterprises operating in the national economy.
The term “market” should be considered as excluding the government sector and non-profit
institutions serving households. In terms of economic activity, the best solution seems to be to
request a breakdown to at least the section-level of the ISIC.
This indicator should not necessarily be seen as permanent. It is designed more from the
pragmatic than the purist point of view, based on data that are currently available. The reason
for this is to try to get a comparable dataset as quickly as possible. This sort of relatively
simple indicator may well prove to be the best approach in the longer term, but it may also be
possible to improve it based on feedback from data users and the experiences of data
providers.
29
This definition is given in Section 3.7 above.
45
• Complementary Indicators
As the key start-up indicator proposed above is unlikely to be ideal for all purposes, a number
of complementary indicators could be envisaged (see also the OECD Business Demography
Framework). These complementary indicators are presented in approximate order of priority:
o An indicator of business start-ups using the working-age population of the country
as the denominator. A consistent definition of this population would need to be
applied, but this may be a useful secondary indicator, particularly for studying
entrepreneurship. It is also, perhaps, more relevant for economies in transition,
where the population of businesses starts low, but grows rapidly. In these
circumstances, start-up rates based on the business population could give a false
impression of the volume of start-ups.
o An indicator of the start-up rate for non-employer businesses: This indicator would
be rather problematic, as it would currently be heavily affected by the wide range of
thresholds used in national sources. More methodological work would be needed
to ensure real comparability, mainly to define a suitable threshold based on some
notion of labour-input that could be applied in all countries. The interest in this type
of business is, however, probably sufficient to justify this work. In the short term,
however, the development of indicators based on information sourced from
business registers, even without threshold adjustments, should be encouraged.
o An indicator of start-ups in terms of employment created: This could be developed
alongside the key indicator proposed above, but would need to be tested for
robustness, as it could be heavily influenced by relatively small differences in
purity.
o An indicator of start-ups in terms of businesses that survive for a minimum period:
Whereas the key indicator would aim to include all start-ups, no matter how short-
lived, a comparative measure of their durability would be useful. Thus start-up
rates defined in terms of businesses that survive for at least two years, or at least
five years, could be envisaged.
o An indicator of start-ups at the site level: Establishments from North America are
probably sufficiently similar to local units in Europe to consider the possibility of a
site-level start-up indicator. Ideally this would have two components, new sites due
to pure enterprise births, and new sites created by existing enterprises. Both are of
interest for studying employment dynamics and the impact of entrepreneurship at
the regional and local levels.
o An indicator of the start-up rate of non-market businesses: Non-profit institutions
serving households are a recognised category of institutional units in national
accounts. They have a clear role in society, and their activities have been referred
to as “social entrepreneurship”, thus measures of their dynamics could be of
interest for socio-economic policy making.
46
47
6. Conclusions
The basic question underlying this project, as stated in the introduction to this report, was;
“How comparable are data on business start-up rates from different OECD countries?” The
short answer, based on the factors of comparability above is: “Not very”. This is because the
comparability factors show that simple comparisons of start-up rates from the different
countries and sources listed in Annex 2 would be misleading and of little value. The longer
answer is, however, rather more positive. Even though data are not currently very
comparable, it seems relatively easy to make a number of improvements to comparability in
the short-term, both analytically and at source.
A number of more detailed conclusions can also be drawn from this report:
• The availability of data on business start-up rates varies considerably between countries.
Some have several sources and long time series that continue up to the present, whereas
others have limited sources, data for only a few years or data series that are not being
continued. For a few OECD member countries, no data sources have been found, and the
availability of data is also very limited for non-OECD countries.
• The availability of metadata is even more variable. Even where metadata exist, they are
not always easy to find or understand, even for specialists. A common metadata template,
based on the factors of comparability above, would make a significant contribution to the
understanding of the data and the reliability of international comparisons. As a first step
towards this, Annex 1 includes proposals for harmonised terminology.
• Some previous international comparisons have suffered from a lack of detailed
understanding of comparability issues. Having said that, they have, however, provided
some useful models for assembling comparable data. The distributed data analysis
models introduced in the OECD Firm-level data study, and, more recently, the Eurostat
business demography project, seem to provide the best route to obtaining harmonised
analyses whilst retaining the detailed knowledge of the source that is necessary for
accurate interpretations of the data.
• To assess comparability of business start-up rates it is necessary to decompose them into
numerator and denominator components, and consider the factors that affect each of
these. Start-up rates that might appear comparable at first glance may be much less so
when they are decomposed in this way.
• A total of nine factors affecting the comparability of business start-up rates have been
identified. Some of these are specific to the numerator or denominator, whereas the others
affect both. The factors that have the most impact are usually the purity of the data in the
numerator, the temporal basis of the denominator, and the coverage of both, though this
varies considerably depending on the data sources being compared.
• The larger a new business is, the less likely it is to be a pure birth in the sense of being a
genuinely new combination of production factors. There is thus a direct relationship
between the amount of work done to improve purity, and the resulting observed impact of
new businesses in terms of employment creation.
• It is possible to make certain analytical adjustments to start-up data to compensate for
differences in specific comparability factors. Unfortunately when compensating for one
factor, it is possible to affect others, and to introduce noise into the data, thus adjustments
have to be made with care, based on a detailed understanding of the data sources and
methods. However, adjustments based on estimates of the impact of a specific factor of
48
comparability can still help to determine whether differences in start-up rates are likely to
be significant or not.
• It would be preferable for any adjustments to be made at source, or through discussion
with those who have a detailed understanding of the data, to reduce the risk of these
adjustments having a negative impact on comparability.
• The degree of harmonisation of data sources has a considerable impact on the
comparability of the resulting data. In this respect, statistical business registers are
perhaps the best sources for business start-up data, as they are already subject to a
degree of harmonisation, of methods, coverage and contents, particularly within Europe.
• A basic key indicator, well defined and relatively easy to implement in all countries, is
necessary to improve data comparability. This approach is in line with the forthcoming
OECD methodological framework for business demography. The key indicator should be
supplemented by a number of complementary indicators that give additional insights to
more specialist data users.
• International organisations such as the OECD need to try to influence the mind-set of
those producing national data. These data producers are often more aware of, and
influenced by national data requirements than they are of the needs for international
comparability.
It is clear that fully comparable data sets can not be produced for all OECD member countries
without more work to develop a suitable methodological framework, and considerable efforts
on the part of those countries. As discussed in Section 5, the former is already in the course
of development at the OECD, but the latter is something that would be rather unrealistic to
expect, at least in the short-term. The focus should therefore be on incremental development
towards more harmonised indicators, whilst promoting longer term convergence within an
agreed methodological framework. The immediate priority is therefore the identification of
“quick wins”, i.e. actions that could increase the international comparability of data from
individual countries for minimal cost. This could include exploring the potential for countries to
supply certain additional data that could be used to make more informed adjustments to their
start-up rates.
At the same time, it is important to increase our understanding of what the users of the data
really want, and what they will use the data for. This knowledge can then inform the future
development of indicators, ensuring that they are as relevant as possible. This sort of step-by
step approach towards a clear goal through incremental improvements may not result in fully
comparable data as quickly as some users might want, but is likely to be more acceptable to
OECD member countries, and perhaps more timely than any more radical approach. For this
reason, it is likely to provide the quickest route to more comparable data on business start-
ups.
Thus in summary, although data on business start ups are not currently very comparable
between countries, there are a number of relatively quick and easy steps that could be taken
to improve their comparability. Clear and comprehensive metadata are vital, as is a detailed
methodological framework that balances user needs and the pragmatic concerns of data
producers. The goal of internationally comparable business start-up rates is not an easy one,
but is possible.
49
7. References
The references below are to papers cited in the main body of this report. References to data
sources are included in Annexes 2 and 4.
ABS (2005), “A Statistical View of Counts of Businesses in Australia”, information paper
published by the Australian Bureau of Statistics, October 2005.http://www.abs.gov.au/Ausstats/abs@.nsf/0/97e9b20f99363f92ca2570920075085a?OpenDoc
ument
ABS (2004), “Business Entries and Exits, A Conceptual Framework”, paper produced by the
Australian Bureau of Statistics, October 2004.
Ahmad, Nadim (forthcoming), “A Framework for Business Demography Statistics”, OECD
proposals for data harmonisation.
Ahmad, Nadim and Steven Vale (2005), “Moving Towards Comparable Business
Demography Statistics”, paper presented at the OECD Structural Business Statistics Expert
Meeting, Paris, November 2005.http://www.oecd.org/dataoecd/54/13/35563231.pdf
Armington, Catherine (1998), “Statistics of US Businesses – Microdata and Tables”, paper
prepared for the US Small Business Administration.http://www.sba.gov/advo/research/rs190tot.pdf
Baldwin, J ohn R., Desmond Beckstead and Andrée Girard (2002), “The Importance of Entry
to Canadian Manufacturing with an Appendix on Measurement Issues”, Statistics Canada
research paper.http://www.statcan.ca/english/research/11F0019MIE/11F0019MIE2002189.pdf
Bartelsman, Eric, J ohn Haltiwanger and Stefano Scarpetta (2005), “Measuring and Analyzing
Cross-country Differences in Firm Dynamics”, paper presented to the 2
nd
EUKLEMS
Consortium Meeting, Helsinki, J une 2005.http://www.euklems.net/meetings/CM_Helsinki/Bartelsman,_Haltiwanger_&_Scarpetta_(2005)
.pdf
Bartelsman, Eric, J ohn Haltiwanger and Stefano Scarpetta (2004), “Microeconomic Evidence
of Creative Destruction in Industrial and Developing Countries”, World Bank Policy Research
Working Paper.http://siteresources.worldbank.org/INTWDR2005/Resources/creative_destruction.pdf
Brandt, Nicola (2004), “Business Dynamics in Europe”, OECD Science, Technology and
Industry Working Paper.http://www.olis.oecd.org/olis/2004doc.nsf/43bb6130e5e86e5fc12569fa005d004c/d7330a8070
18cccbc1256e55005051c7/$FILE/J T00159809.DOC
Cella, Patrizia and Caterina Viviano (2004), “Register Entries / Exits and Demographic Flows:
Some Comparisons for Statistical Aggregates”, paper presented to the 18
th
International
Roundtable on Business Survey Frames, Beijing, October 2004.http://forum.europa.eu.int/irc/DownLoad/kfecA5J -mpGIXl0SR1OAauLw2xPI-
SvMwc82kKbj0B2SBp2UxVqIlDcEc04rLSm9uxKFCBTAaIZ3lqM0jNc9dF5L_IbkP/S8-2%20-
%20Register%20entries-exits%20and%20demographic%20flows%20-
%20some%20comparisons%20for%20statistical%20aggregates.doc
50
Choi, Bongho and Denis Ward (2004), “Analysis of Statistical Units Delineated by OECD
Member Countries”, paper presented to the 18
th
International Roundtable on Business Survey
Frames, Beijing, October 2004.http://forum.europa.eu.int/irc/DownLoad/kYedA1J Sm_GMsYREH2D5UbAjBu320H8WVHgZtF
cfKeSCBQKuSxDZA0zr4s3K81qMzfT5uGS67YAxCZLHjBgRxd9Rm0df4e1D/S3-1%20-
%20ANALYSIS%20OF%20STATISTICAL%20UNITS%20DELINEATED%20BY%20OECD%2
0MEMBER%20COUNTRIES.doc
Davis, Steven J ., J ohn Haltiwanger, Ron J armin, C. J . Krizan, J avier Miranda, Alfred Nucci
and Kristin Sandusky (2005), “Measuring the Dynamics of Young and Small Businesses:
Integrating the Employer and Nonemployer Universes”, preliminary paper presented to the
US National Bureau of Economic Research Conference on Research in Income and Wealth,
Bethesda, April 2005.http://www.nber.org/confer/2005/CRIWs05/haltiwanger.pdf
Eurostat (2003), Eurostat Manual of Recommendations for Business Registers, Chapter 13.http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/embs/registers/chapter13.doc
GEM (2004), Global Entrepreneurship Monitor 2004 Global Reporthttp://www.gemconsortium.org/document.asp?id=364
Herczog, Aimée, Hans van Hooff and Ad Willeboordse (1998), “The Impact of Diverging
Interpretations of the Enterprise Concept”, prepared for Eurostat by Statistics Netherlands.http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/embs/ent_concept/section1.html
Klapper, Leora, Luc Laeven and Raghuram Rajan (2004), “Business Environment and Firm
Entry: Evidence from International Data”, US National Bureau of Economic Research Working
Paper.http://www.nber.org/papers/w10380
Mead, Geoff (2005), “Statistics New Zealand Business Frame Strategy and Developments
Related to Statistics on SMEs and the Support of Longitudinal Business Statistics”, paper
presented at the OECD Structural Business Statistics Expert Meeting, Paris, November 2005.http://www.oecd.org/dataoecd/9/44/35485064.pdf
Mills, Duncan and J ason Timmins (2004), “Firm Dynamics in New Zealand: Comparative
Analysis with OECD Countries”, paper presented to the 2004 conference of the New Zealand
Association of Economists.http://www.nzae.org.nz/conferences/2004/88-Mills-Timmins.pdf
Pinkston, J oshua C., and J ames R. Spletzer (2004), “Annual Measures of Gross J ob Gains
and Gross J ob Losses”, US Bureau of Labor Statistics Monthly Labor Review, November
2004.http://www.bls.gov/opub/mlr/2004/11/art1full.pdf
Takahashi, Masao (2000), “Business Demography and the J apanese Business Survey
Frame”, paper presented to the 14th International Roundtable on Business Survey Frames
Auckland, November 2000.http://forum.europa.eu.int/irc/DownLoad/kveFAjJ ZmSGspYM195H5EFCl6eTNvOz6Vt5McKbY
N63r0IIuHVQp4CmHyIxc1GjlFVXmUpoo2tSfBIMGtOpIxcLHbI/Paper%20J apan%20-
%20session7.pdf
Vale, Steven (2005(a)), “International Data on Business Start-ups: Factors Affecting
Comparability”, paper presented to the 19
th
International Roundtable on Business Survey
Frames, Cardiff, October 2005.http://forum.europa.eu.int/irc/DownLoad/kXeeAJ J AmjGIcxMS0S9p_GZ2GvLRcOhH4rLSm9ux
KEH2EUGq8jov3OoGlUXySA_6pkAUF-S0DF-9gqmf-
Hj0fcg/Comparability%20RT%20Paper.doc
51
Vale, Steven (2005(b)), “The Coverage of Micro-Enterprises in Business Registers”, paper
presented at the OECD Structural Business Statistics Expert Meeting, Paris, November 2005.http://www.oecd.org/dataoecd/32/46/35506105.pdf
Vale, Steven and Claire Powell (2003), “Developments in Business Demography
- Reconciling Conflicting Demands”, paper presented to the Comparative Analysis of
Enterprise (micro) Data Conference, London, September 2003.http://www.statistics.gov.uk/events/CAED/abstracts/downloads/vale.pdf
Vale, Steven and Claire Powell (2002), “Estimating Under-coverage of Very Small
Enterprises”, paper presented to the 16
th
International Roundtable on Business Survey
Frames, Lisbon, October 2002.http://forum.europa.eu.int/irc/DownLoad/kgecA1J SmRGTex2OFEBEGfEvoFle1HrMxrNLRFur
O5CG0NNpCl6eTNvOz64c5SN2VkAUF-0RCF-9gqmf-H/ses3_UK_Paper.pdf
Volfinger, Zsolt (2004), “Coverage of the Hungarian Business Register”, paper presented to
the 18
th
International Roundtable on Business Survey Frames, Beijing, October 2004.http://forum.europa.eu.int/irc/DownLoad/kjecAJ J UmfG1uvhdvqlF0uAePVRfj3jMhqKGf0phOGF
-HOBF7z6zLRjGpRmu-AZ-um3THrGuypb4pqOIjE5Tzc1L/S5-3%20-
%20Coverage%20of%20the%20Hungarian%20Business%20Register.doc
52
53
Annex 1 – Glossary of Terms: Proposals for Harmonised
Terminology
This report notes that comparability of data on business start-ups, and business demography
more generally, is hampered by inconsistent presentation of metadata. To reduce this
problem, the creation of a standard metadata template is proposed. As a first step towards
this, a harmonised terminology is needed. This Annex proposes standard terms and
definitions, using a notation which relates events to a particular time period, t.
These terms and definitions have been used throughout this report, so this Annex also acts as
a glossary.
• Births (B
t
) – A birth is the creation of a combination of production factors with the
restriction that no other national businesses are involved in the event. Births do not include
entries into the population due to reactivations, mergers, break-ups, split-offs or other
restructuring of a group of businesses linked by ownership or control. Births also exclude
entries into a population resulting from changes to characteristics of existing businesses.
(Note – this is largely based on, and fully consistent with the Eurostat definition for
enterprise births).
• Churn – Total churn is defined as the sum of businesses joining and leaving the
population during a given period, i.e. entries plus exits (E
t
+X
t
). Pure churn excludes
entries and exits that are due to events other than births and deaths (i.e. B
t
+D
t
).
• Closing Stock (PB
t
) – The population at the end point of the period. This is usually
equivalent to the population at the start point for the following period (PA
t+1
).
• Deaths (D
t
) - A death is the dissolution of a combination of production factors with the
restriction that no other domestic businesses are involved in the event. Deaths do not
include exits from the population due to temporary inactivity, mergers, take-overs, break-
ups or other restructuring of a group of businesses linked by ownership or control. Deaths
also exclude exits from a population resulting from changes to characteristics of
businesses which remain active. (Note – this is largely based on, and fully consistent with
the Eurostat definition for enterprise deaths).
• Entries (E
t
) – All businesses that join the population during the period, regardless of
whether they are still present at the end of the period.
• Exits (X
t
) – All businesses that leave the population during the period, regardless of
whether they were present at the start of the period.
• J oiners (J
t
) – Businesses that are present in the population at the end of the period, but
were not present at the start of the period.
• J oiners and leavers within period (J L
t
) – Businesses that are not present in the population
at the start or the end of the period, but are present in at least one observation of the
population between these two points (or would be if such observations were made).
• Leavers (L
t
) – Businesses that are present in the population at the start of the period, but
are not present at the end of the period.
54
• Opening Stock (PA
t
) – The population at the start point of the period. This is usually
equivalent to the population at the end point for the previous period (PB
t-1
).
• Other Entries (OE
t
) – All entries that are not births
• Other Exits (OX
t
) – All exits that are not deaths
• Population – All businesses that meet certain predefined criteria.
o Live during Period (P
t
) - All businesses that meet certain predefined criteria at any
time during a specified time period.
o Point in Time – All businesses that meet certain predefined criteria at a specific
temporal reference point.
• Purity – The degree to which pure births and deaths are distinguished from other
demographic events.
• Survivors (S
t
) All businesses that are in the population at both the start and the end of the
period
55
Annex 2 - Inventory of Data on Business Start-ups by Country
Introduction
This Annex provides an inventory of the available data and metadata found for each OECD
country. It also includes information from other countries for which data have been found in
the course of this work. In some cases, data are also available for groups of countries (e.g.
the members of the European Union), via international agencies.
The information contained in this Annex has been compiled based on searches of the Internet
in Autumn 2005, and the author’s knowledge of sources. The focus is on official data sources,
usually from National Statistical Institutes, though other sources are considered for some
countries.
The following pages list the sources found for each country, using a standard template for
each source. They include links to the data where possible, but do not attempt to explicitly
assess the comparability or any of the other dimensions of the quality of the data. Summary
metadata, including coverage and definitions, are included, as well as information on the
availability of more detailed metadata.
Several countries have participated in the Eurostat business demography data collections, the
DOSME (Demography of Small and Medium-sized Enterprises) project, and/or the OECD
firm-level data project, so data are available via those routes. These sources are included for
each country if appropriate, but to avoid repetition, information on the coverage and definitions
used are provided separately at the end.
1. Australia
One data source available
Title – Web publication “Experimental Estimates, Entries and Exits of Business Entities”
Source – Australian Bureau of Statistics, 2005
Internet address –http://www.ausstats.abs.gov.au/Ausstats/subscriber.nsf/Lookup/2EB3AE08FFBC9AD4CA257
0280078B69E/$File/8160055001_2001-02,%202002-03%20and%202003-04.pdf
Contents – Register-based population, entry, survival and exit estimates.
Breakdowns – The data are broken down by economic activity, size, geography and “type of
business entity” (legal form). Size breakdowns are not given for survival data.
Metadata – Some metadata are included in the publication, further information is contained in
an additional paper, “Business Entries and Exits – A Conceptual Framework”, available from
the Australian Bureau of Statistics on request. See also the paper “Development of Statistics
on Business Demography and Continuity in Australia” -http://forum.europa.eu.int/irc/DownLoad/kxeFAAJ UmZGMYjKH--
EUNFa3pFKPjEWCqF4EiCwmAUM8GSHYRf6dTHzryIxJ 9UZ-bYR3R1H-
BbbkSskDDjYv4G8BZM/ses3_Australia_Paper1.pdf
56
Period covered – The population estimates are as at 1 J uly for 2001 to 2003. Entries and
exits are available for the periods 1 J uly to 30 J une 2001/2 to 2003/4. One and two year
survival rates are available for entries in 2001/2 and one year survival rates for entries in
2002/3.
Coverage – ANZSIC division M (government) and non-market government entities are
excluded. The threshold for registration is generally 50,000 Australian Dollars (approx.
€31,000), with some exceptions, and some voluntary registrations.
Definitions
• The population estimates are on a point in time basis.
• The unit used is the legal unit, i.e. entities registered for an Australian Business
number (ABN).
• Entries are defined as the allocation of a tax role within the Australian Business
Register (ABR). This excludes inactive businesses, changes in legal form, and
reactivations. A check is also made for newly created businesses that take over the
activity of one or more existing businesses. These are excluded where identified.
• Exits are defined as the cancellation of all tax roles within the ABR, with similar
inclusions and exclusions to those for entries.
2. Austria
One data source available
Title – Unternehmensneugründungen in Österreich 1993-2004
Source – Wirtschaftskammern Österreich (WKO)
Internet address –http://portal.wko.at/wk/dok_detail_file.wk?AngID=1&DocID=344536&DstID=1721&StID=17871
2
Contents – Start year stock and new registrations for 1993 to 2004
Breakdowns – New registrations are broken down by economic activity, legal form,
geography and (for natural persons) sex and age of the entrepreneur.
Metadata – There are some descriptive metadata in the introduction to the publication.
Period covered – 1993 to 2004
Coverage – Registrations with the WKO
Definitions
• The population estimates are on a point in time basis (start of the year), and consist of
active WKO registrations. The population data are not subject to the same adjustments
as those for new registrations.
• The basic unit is the registration at the WKO, which can be considered as a legal unit.
However, the corrections made bring the unit used for analysis much closer to the
enterprise.
• New registrations are adjusted to remove re-registrations, dormant units, and multiple
registrations for the same enterprise. Adjustments are also made for registration lags.
57
3. Belgium
Two data sources available
a) Title – Démographie des entreprises (1998-2004)
Source – Statistics Belgium
Internet address –http://statbel.fgov.be/figures/d422_fr.asp
Contents – Value-added tax (VAT) stock, new registrations, de-registrations and liquidations
Breakdowns – The data are broken down by legal form, economic activity and geography.
Metadata – Limited to table headings and notes. More detailed metadata exist in the
publication “Démographie des entreprises 2002”, which also contains more comprehensive
data breakdowns, but just for 2002.
Period covered – The population estimates are as at 31 December for 1998 to 2003.
Registrations and de-registrations are available for the years 1998 to 2003. Liquidations are
available for the years 1998 to 2004.
Coverage – Some specific legal, medical, financial, social and personal services are exempt
from value-added tax. There is no VAT registration threshold. Public sector entities are
included if they are registered for VAT.
Definitions
• The stock data are on a point in time basis, and include those VAT registrations
marked as active at the end of the year.
• The unit used is the VAT registration, which is broadly equivalent to the legal unit.
• Entries are defined as VAT registrations whose year of creation is the reference year,
including those that have been removed from the VAT register before the end of that
year.
• Exits are defined as VAT registrations whose year of removal from the VAT register is
the reference year.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2000. Data
for births exist for 1998 and 2000 (not 1999), and data for deaths exist for 1998 and 1999.
58
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
4. Canada
Three data sources available
a) Title – Business Dynamics in Canada, 2001 (supplemented for 2002 by data prepared for
FORA).
Source – Statistics Canada – Longitudinal Employment Analysis Program (LEAP) Database.
Internet address –http://www.statcan.ca:8096/bsolc/english/bsolc?catno=61-534-X (note:
$25 charge)
Contents – This publication includes data on business populations, birth and death rates, and
survival.
Breakdowns – Business populations, births and deaths are broken down by size and
economic activity categories based on knowledge intensity. Business populations are also
broken down by geography.
Metadata – The publication contains a chapter on methodology.
Period covered – The population estimates are available for 1991 to 2001, births are
available from 1992 to 2001, and deaths from 1991 to 2000. Additional data on population and
births for 2002, and deaths for 2001 is taken from a short report prepared by Statistics
Canada for FORA.
Coverage – All employers in Canada, public and private (i.e. data do not include businesses
with no employees). Non employing businesses are included in the report prepared for FORA,
but only for three years, and have considerable variation in the rates, which could call into
question the validity of these data.
Definitions
• The unit used is the firm, which, at the national level is equivalent to the legal unit.
• Births are defined as firms that are not present on the LEAP database in year t, but are
present in year t+1. The birth rate is the number of new enterprises in t+1 divided by
the total number of firms observed in year t+1.
• Deaths are defined as firms that are present on the LEAP database in year t, but are
not present in year t+1. The death rate is the number of enterprises operating in t, but
not in t+1, divided by the total number of firms observed in year t+1. Note this is
different to the rates calculated in many other countries where the population in year t
is the denominator.
b) Title – Self-Employment Entry and Exit Flows
Source – Statistics Canada – Paper by Zhengxi Lin, Garnett Picot and J anice Yates
Internet address –http://www.statcan.ca/english/research/11F0019MIE/11F0019MIE1999134.pdf
59
Contents – This paper contains a table with counts of self-employed persons, and rates for
entry and exit, as well as other related data and analyses.
Breakdowns – No breakdowns of the entry and exit data are given in the paper.
Metadata – The paper contains descriptive metadata on sources and definitions.
Period covered – The population estimates are available for 1981 to 1995, entries are
available from 1982 to 1995, and exits from 1981 to 1994.
Coverage – All persons declaring income from self employment in their annual tax returns to
revenue Canada.
Definitions
• The unit used is the person completing a tax return.
• Self-employment entries are income-tax filers who report earnings from self-
employment in one year but not the previous year
• Self-employment exits are income-tax filers who report earnings from self-employment
in one year but not the next.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Canada are available from 1984 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
5. Czech Republic
Two data sources available
a) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
60
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2001 and 2002, and data for deaths exist for 2000 and 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
6. Denmark
Three data sources available
a) Title – Statistical Yearbook
Source – Statistics Denmark
Internet address
2005 (data for 2001) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2005.aspx
2003 (data for 2000) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2003.aspx
2001 (data for 1999) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2001.aspx
61
2000 (data for 1998) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2000.aspx
Contents – Counts of new enterprises
Breakdowns – Data are broken down by economic activity
Metadata – Very limited
Period covered – Birth counts are available for 1998 to 2001
Coverage – Data exclude agriculture and public administration
Definitions
• The unit used is the enterprise
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2001. Data
for births exist for 1998 to 2001, and data for deaths exist for 1997 to 2000.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Denmark are available from 1981 to 1994.
62
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
7. Finland
Three data sources available
a) Title – Enterprise openings and closures
Source – Statistics Finland
Internet address –http://www.stat.fi/til/aly/index_en.html Data only accessible via Finnish
version -http://www.stat.fi/til/aly/index.html
Contents – Counts of enterprise openings and closures. Stock figures are available
separately from the StatFin database, but may not have the same coverage.
Breakdowns – Data are broken down by economic activity, legal form and geography.
Metadata – Mostly in Finnish
Period covered – Data are available from 1999 to 2004
Coverage – The openings and closures data are derived from Statistics Finland’s business
register. They only cover those enterprises engaged in business activity that are liable to pay
value-added tax or act as employers. Foundations, housing companies, voluntary
associations, public authorities and religious communities are excluded. The data cover state-
owned enterprises, but not those owned by municipalities.
Definitions
• The unit used is the enterprise
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
63
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Finland are available from 1989 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
8. France
Three data sources available
a) Title – Créations d'entreprises
Source – INSEE
Internet address –http://www.insee.fr/fr/ffc/chifcle_liste.asp?theme=9&soustheme=1&souspop=
Contents – Counts of enterprise creations, split into new creations, resumptions and re-
activations
Breakdowns – Breakdowns by economic activity, size and legal form are available.
Metadata – Limited, e.g. some key definitions.
Period covered – Data are available from 1993 to 2004
Coverage – The data cover all of France, including the overseas départements.
Definitions
• The unit is assumed to be the enterprise.
• Three categories of enterprise creation are identified:
o Pure creations (creations “ex nihilo”) where the new enterprise does not take
over the activities of a previously existing enterprise.
o Reactivations, where a person who has previously been self-employed re-
starts a self-employed activity.
o Resumptions, where a new business takes over an activity previously carried
out by another enterprise.
64
b) Title – La Création en Chiffres
Source – Agence Pour la Création d’Entreprises (APCE)
Internet address –http://www.apce.com/index.php?rubrique_id=261&type_page=I
Contents – Counts of enterprise creations, split into new creations (“ex nihilo”), resumptions
and re-activations
Breakdowns – None
Metadata – Very limited
Period covered – Data are available from 1993 to 2004
Coverage – No information
Definitions – None given – the data are very similar to, but not the same as, those from
INSEE.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for France are available from 1990 to 1996.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
9. Germany
Three data sources available
a) Title – Business Notifications
Source – Federal Statistics Office, Germany
Internet address –http://www.destatis.de/themen/e/thm_unternehmen.htm
65
Contents – Business registrations, modifications and de-registrations, and counts of
businesses liable to pay turnover tax.
Breakdowns – The data are broken down by economic activity.
Metadata – Descriptive metadata are available via the website
Period covered – Registration and de-registration data are available for 2001 to 2003. Data
on the population of businesses liable for tax are available for 2002 and 2003.
Coverage – The registration and de-registration data are assumed to cover the whole
economy. The population of businesses liable to pay turnover tax covers businesses with a
turnover of at least €16,620 per year. It covers most economic activities, with exceptions for
certain health, public administration, insurance and agricultural activities.
Definitions
• The unit for registrations and de-registrations is effectively the local unit as “the
obligation to report business registrations and de-registrations applies to enterprises,
branch offices and dependent sub-offices”.
• Registration is required when a new activity is started or a business is taken over, be it
through purchase or succession, a partner entering the business, a change in legal
form, or a relocation of the business to a different registration district.
• De-registration is required when a business is shut down completely or in part, or is
sold, a partner withdraws from the business, the legal form is changed, or the business
is relocated to a different registration district.
b) Title – Start-ups and Liquidations in Germany 1991 - 2004
Source – Institut für Mittelstandsforschung Bonn
Internet address –http://www.ifm-bonn.org/dienste/gruendungen-engl.htm
Contents – Counts of business start-ups and liquidations. Some enterprise population totals
in Table 1 of:http://www.ifm-bonn.org/dienste/kap-2.pdf
Breakdowns – The start-up and liquidation data are broken down into the former East and
West Germany.
Metadata – Limited metadata available via the web site.
Period covered – Start-up and liquidation data are available for 1991 to 2004. Data on the
population of businesses are available for 1994 and 1996 to 1999 (IfM have provided
estimates for the missing years).
Coverage – The population of businesses is subject to a threshold (€17,500 since 2003), and
covers all economic activities except the “liberal professions”, most health services and some
insurance services that are not subject to value added tax.
Definitions
• The units used are effectively the sub-set of legal units that are considered to be
economically relevant.
66
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Germany are available from 1978 to 1998.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex. Note – data for Germany only cover the former West Germany.
10. Greece
No data sources found
11. Hungary
Three data sources are available
a) Title – Enterprises and Non-profit organisations
Source – Hungarian Central Statistical Office
Internet address –http://portal.ksh.hu/portal/page?_pageid=38,341368&_dad=portal&_schema=PORTAL
Contents – Annual counts of registered economic corporations and unincorporated
enterprises, as well as quarterly counts of new registrations
Breakdowns – Both are broken down by legal form, the population data are also broken
down by economic activity.
Metadata – Information on definitions and sources is available on the web site. See also the
paper “Coverage of the Hungarian Business Register” at:http://forum.europa.eu.int/irc/DownLoad/kjecAJ J UmfG1uvhdvqlF0uAePVRfj3jMhqKGf0phOGF
-HOBF7z6zLRjGpRmu-AZ-um3THrGuypb4pqOIjE5Tzc1L/S5-3%20-
%20Coverage%20of%20the%20Hungarian%20Business%20Register.doc
Period covered – Population are available from 1994 to 2004. Annual data on new
registrations can be constructed by adding the four quarterly totals for 2001 to 2004.
67
Coverage – The data cover all businesses that hold an active registration and tax number in
the administrative register, including most government bodies. There is no registration
threshold in Hungary, so part-time businesses are included. Approximately 75% of
registrations are considered to be economically active by the Hungarian statistical office. All
economic activities are covered, though NACE division L (public administration) is excluded
from the counts broken down by activity.
Definitions
• The unit is referred to as the enterprise, but the definition is closer to that of a legal
unit.
• The population data are point in time estimates for the end of the year.
b) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
c) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2000 to 2002, and data for deaths exist for 1999 to 2001.
68
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
12. Iceland
Two data sources available
a) Title – Registered Enterprises and Organisations 1995-2001
Source – Statistics Iceland
Internet address –http://www.statice.is/?pageid=1198&src=/temp_en/fyrirtaeki/fyrirtaeki.asp
Contents – Counts of registered enterprises, new registrations and “new depreciation”
Breakdowns – The data are broken down by legal form.
Metadata – Only brief footnotes are available.
Period covered – The data are available for 1995 to 2001.
Coverage – The data seem to cover all legal forms.
Definitions
• The population figures appear to be point in time estimates, as at the end of the year.
• The unit used appears to be the legal unit.
b) Title – Enterprises / New Registrations by Economic Activity
Source – Statistics Iceland
Internet address –http://www.statice.is/?pageid=1198&src=/temp_en/fyrirtaeki/fyrirtaeki.asp
Contents – Counts of enterprises and new registrations
Breakdowns – The data are broken down by economic activity (NACE section).
Metadata – Only brief footnotes are available.
Period covered – The enterprise population data are available for 1999 to 2004, the new
registrations data are available from 1995 to 2004.
Coverage – The data seem to cover all economic activities.
Definitions
• Non available
13. Ireland
No data sources found, though Ireland are starting to supply data for the Eurostat business
demography project
69
14. Italy
Three data sources available
a) Title – Movimprese
Source – InfoCamere
Internet address –http://www.infocamere.it/movi_search.htm
Contents – Counts of total registrations and active registrations, new registrations, cessations
and changed registrations at the Italian chamber of commerce.
Breakdowns – The data are broken down by economic activity and geography.
Metadata – A glossary and other metadata are available on the web site (in Italian)
Period covered – The data are available for 1995 to 2004.
Coverage – The data do not cover NACE section L (public administration), and presumably
do not cover government units.
Definitions
• The unit used is the legal unit
• The population data are point in time, and appear to relate to the end of the year
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
c) Title – OECD Firm-Level Data Project
Source – OECD
70
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Italy are available from 1987 to 1993.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
15. Japan
One data source found
Title – Establishment and Enterprise Census
Source – Statistics bureau of J apan
Internet address –http://forum.europa.eu.int/irc/DownLoad/kveFAjJ ZmSGspYM195H5EFCl6eTNvOz6Vt5McKbY
N63r0IIuHVQp4CmHyIxc1GjlFVXmUpoo2tSfBIMGtOpIxcLHbI/Paper%20J apan%20-
%20session7.pdf
There are also some data on establishment and enterprise populations athttp://www.stat.go.jp/english/data/jigyou/kekka.htm.
Contents – Table 3 of the paper at the first address above includes counts of existing
establishments (i.e. survivors), “newly-organised establishments” and “abolished
establishments” based on data from the 1989, 1994 and 1999 establishment and enterprise
censuses.
Breakdowns – The data for 1999 are broken down by economic activity and employment size
band.
Metadata – The paper contains definitions and descriptive metadata.
Period covered – The data are available for 1989, 1994 and 1999. Annualised “opening” and
“abolishment” rates are also given for the periods between censuses.
Coverage – The data do not cover sole-proprietor businesses in agriculture, forestry and
fishing activities, or any businesses classified to domestic services, foreign governments or
international agencies. Several other specific exclusions are listed in the paper.
Definitions
• The unit used is the establishment
• The population data are on a point in time basis
• A newly-organized establishment is defined as an establishment that had been newly-
organized or had moved into the present place since the date of the preceding census.
71
• An abolished establishment is defined as an establishment that had moved to a
different place or had been closed since the date of the preceding census.
16. Korea
No data sources found, though there are some counts of establishments at:http://kosis.nso.go.kr/cgi-
bin/SWS_1021.cgi?KorEng=2&A_UNFOLD=1&TableID=MT_ETITLE&TitleID=HA&FPub=4&U
serID=
17. Luxembourg
Two data sources available
a) Title – Démographie des Entreprises
Source – STATEC
Internet addresses –http://www.statistiques.public.lu/stat/TableViewer/tableView.aspx?ReportId=258http://www.statistiques.public.lu/stat/TableViewer/tableView.aspx?ReportId=259http://www.statistiques.public.lu/stat/tableviewer/document.aspx?FileId=209
Contents – Counts of the stock of enterprises at the start of the year, new enterprises created
during the year in the framework of the policy of economic diversification, and requests for
authorisation for establishments. Note – from the numbers given, the new enterprise data
would only seem to account for a small proportion of all enterprise births.
Breakdowns – The stock and new enterprise data are broken down by very broad categories
of economic activity. The requests for authorisation for establishments are broken down by
nationality of the requestor (Luxembourgish or foreign).
Metadata – Some metadata are available by following the information links within the tables
on the web site.
Period covered – Data on the stock of enterprises are available for 2002 to 2004. Data on
new enterprises are available for 1990, and 2000 to 2004. Data on requests for authorisation
for establishments are available for 1990 and 1995 to 2004.
Coverage – There is no specific information on coverage.
Definitions
• The data on the stock of enterprises are on a point in time basis (1 J anuary of the
reference year).
• The unit used is assumed to be the enterprise for the stock and new enterprise data,
and the local unit for the data on requests for authorisation for establishments.
b) Title – Business demography indicators
Source – Eurostat
72
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
18. Mexico
No data sources found
19. Netherlands
Three data sources available
a) Title – Establishment and Closure of Businesses
Source – Statistics Netherlands
Internet address –http://statline.cbs.nl/StatWeb/table.asp?PA=07223eng&D1=a&D2=0&D3=(l-11)-
l&DM=SLEN&LA=en&TT=2
Contents – Counts (and employment) of the stock of businesses (as at 1 J anuary),
businesses opening, and businesses closing.
Breakdowns – The data are available broken down by economic activity.
Metadata – Metadata are available by clicking on the table headings on the web site.
Period covered – The data are available from 1993 to 2002 (closures only to 1996).
Coverage – The data exclude certain NACE categories (Sections A, B, E, L, M and N, and
divisions 70, 73, 91 and 92). On this basis it is assumed that most government activity is also
excluded.
Definitions
• The stock of businesses is a point in time estimate
• The unit used is the “business” which is assumed to be close to the enterprise, as the
terms are both used in the metadata.
• Establishment of a business is the formation of a new enterprise. This implies that the
statistical criteria for enterprises (autonomy, description and external orientation) have
to be met. Moreover, the enterprise has to be economically active, i.e. at least one
73
person works in the enterprise for at least 15 hours a week. The enterprise has to be a
new one, i.e. not the continuation of one or more existing enterprises.
• Closure of enterprises implies discontinuation of all activities, hence
no continuation of activities by other enterprises.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1999 to 2001. Data
for births exist for 1998 to 2002, and data for deaths exist for 1998 to 2000.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for the Netherlands are available from 1987 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
20. New Zealand
One data source available
Title – SMEs in New Zealand: Structure and Dynamics - 2005
74
Source – New Zealand Ministry of Economic Development
Internet address –http://www.med.govt.nz/irdev/ind_dev/smes/2005/index.html
Contents – Counts of the stock of enterprises in February of each year (Figure 1 of the
publications for 2001 to 2005), and enterprise births and deaths during the year (table
underlying Figure 15 of the 2005 publication).
Breakdowns – Stock data are broken down by employee size band. There are no
breakdowns of the birth and death data.
Metadata – The 2005 publication contains extensive metadata, including a glossary.
Period covered – The stock data are available for 2000 to 2004. The data on births and
deaths are available for 1998 to 2004.
Coverage – The data exclude agriculture production (ANZSIC subdivision A01). They
also exclude businesses of “little economic significance”, i.e. those that fail to meet at least
one of the following criteria:
• greater than $30,000 (approx €17,500) annual taxable expenses or sales
• rolling mean employee count of greater than three
• in a tax-exempt industry (except for residential property leasing and rental)
• part of a group of enterprises
• registered for tax and involved in agriculture or forestry.
Definitions
• The stock data are on a point in time basis, with a February reference date.
• The unit used appears to be the legal unit, though the term ‘enterprise’ is used.
• Data on the entry and exit of firms include administrative changes such as
restructuring and changes of ownership, as well as genuine business start-ups and
closures.
21. Norway
Two data sources available
a) Title – Statbank Norway / Enterprises
Source – Statistics Norway
Internet address –http://statbank.ssb.no/statistikkbanken/default_fr.asp?PLanguage=1
Contents – Count and employment data on the population of enterprises, new enterprises
and enterprise “drop-outs”. (Limited data on survival are also available at -http://www.ssb.no/english/subjects/10/01/fordem_en/tab-2004-12-01-01-en.html).
Breakdowns – The data are broken down by geography, economic activity, legal form and
size band.
Metadata – Detailed methodological notes and definitions are available athttp://www.ssb.no/vis/foretak_en/about.html
75
Period covered – The data on the population of enterprises are available for 2001 to 2005.
The data on new enterprises and enterprise drop-outs are available for 2001 to 2004.
Coverage – Enterprises classified to public administration, agriculture, forestry and fishing are
excluded, as are central and local government units.
Definitions
• A new enterprise in a given period is an enterprise registered with dates that indicate
start-up in this period.
• The number of new established enterprises is the number of a new enterprises
corrected for the change of ownership. That is - new enterprises that take over existing
activity are not counted as new established enterprise, but only as a new enterprise.
• A discontinuance of an activity is counted as a drop-out. If all of the establishment is
closed down, and is not taken over by another enterprise, the drop-out is also
classified as a closure.
• The population of enterprises is a point in time estimate as of 1st J anuary.
• The unit used is the enterprise
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2001. Data
for births exist for 1998 to 2001, and data for deaths exist for 1999 and 2000.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
22. Poland
Two data sources available
a) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
76
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
b) Title – Entry and Exit Rates in the Polish manufacturing
Source – National Bank of Poland
Internet address –http://www.fcee.urv.es/departaments/economia/recerca/grit/Catala/web/papers/Rogowski-
Socha.pdf
Contents – This paper compares data from several national sources on business entry and
exit. Most sources concentrate only on manufacturing, but whole economy entry and exit rates
from the REGON register are included in Table 10.
Breakdowns – The data are broken down by broad economic activity.
Metadata – The paper contains some limited metadata, mainly describing the source.
Period covered – Entry and exit rates are available for 1998 to 2003
Coverage – The data are claimed to cover the whole economy, but the breakdown by broad
economic activity does not include data for agriculture, business services, public
administration, health, education or personal services, so these activities may be excluded. A
warning is given that around a half of the businesses included in REGON were inactive in
1999, dropping to 30-40% in 2003. This could imply that entry rates as a proportion of active
businesses should be much higher, but it is likely that a proportion of entries are themselves
inactive.
Definitions
• The data appear to use a point in time population.
• The unit used is referred to as the enterprise, but is not defined.
23. Portugal
Two data sources available
Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
77
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
b) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Portugal are available from 1983 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
24. Slovakia
Two data sources available
a) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
78
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2000 to 2002, and data for deaths exist for 2000 and 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
25. Spain
Two data sources available
a) Title – Demografía de las Empresas
Source – Instituto Nacional de Estadística
Internet address –http://www.ine.es/inebase/cgi/um?M=/t37/p201&O=inebase&N=&L=0
Contents – Counts of enterprises, creations (split between pure births and reactivations), and
cessations.
Breakdowns – The population, creation and cessation counts are broken down by economic
activity, legal form and size. The INEbase data warehouse allows more detailed breakdowns
of total creations and cessations by the same variables.
Metadata – Some descriptive metadata are available via the web site above.
Period covered – The population counts are available for 1999 to 2005. Data on creations
and cessations are available for 1998 to 2004, but the split of creations into pure births and
reactivations is only present for 2001 to 2004.
79
Coverage – The data appear to exclude agriculture, forestry, fishing and public administration
activities, as well as central and local government units.
Definitions
• The population counts are on a point in time basis (1 J anuary).
• The unit used is the enterprise.
• Creations (altas) are defined as new registrations in DIRCE (the statistical business
register) that start their activities in the reference year.
• Cessations (bajas) are defined as units that cease activity in DIRCE in the reference
year.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2000 to 2002, and data for deaths exist for 2000 and 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
26. Sweden
Two data sources available
a) Title – Nystartade företag (New enterprise starts)
Source – Statistics Sweden
Internet address –http://www.scb.se/templates/tableOrChart____27185.asphttp://www.scb.se/templates/Standard____36176.asp
Contents – Counts of new enterprises, and the stock of enterprises (and local units) at 1
J anuary each year.
Breakdowns – The data on new enterprises are broken down by broad economic activity
categories. The series on the stock of enterprises is not broken down, though annual
publications provide detailed breakdowns by size (employment and turnover), legal form,
economic activity and geography for individual years.
Metadata – Descriptive metadata about the Swedish statistical business register are available
athttp://www.scb.se/templates/Listning2____31034.asp , including some definitions.
80
Period covered – The data on new enterprises are available for 1996 to 2004. The data on
the stock of enterprises are available for 1971 to 2004, though include some discontinuities
(e.g. in 1996) due to changes in the scope of the administrative sources used for the statistical
business register.
Coverage – The data on the stock of enterprises cover all economic activities and legal
forms. It is not clear whether the data on new enterprises have the same coverage.
Definitions
• The stock of enterprises is on a point in time basis (1 J anuary).
• The unit used is the enterprise, though enterprises are defined as active legal units.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2002.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
27. Switzerland
One data source available
Title – Démographie des entreprises
Source – Office Fédéral de la Statistique, Switzerland
Internet address –http://www.bfs.admin.ch/bfs/portal/fr/index/themen/industrie_und_dienstleistungen/unternehm
en/blank/medienmitteilungen.htmlhttp://www.bfs.admin.ch/bfs/portal/fr/index/themen/industrie_und_dienstleistungen/uebersicht/
blank/publikationen.html?publicationID=853
Contents – Count and employment data for new enterprises. The second internet address
above gives counts of the stock of enterprises for years in which a census of enterprises has
been carried out.
81
Breakdowns – The data on new enterprises are broken down by economic activity. There are
various breakdowns available for data from censuses of enterprises.
Metadata – Methodological notes are included in the press release on new enterprises.
Period covered – Counts of new enterprises are available for 1999 to 2003. Data on the
stock of enterprises are available for 1985, 1995, 1998 and 2001.
Coverage – Economic activities in agriculture, forestry, fishing and public administration
(NACE sections A, B and L) are not covered. National and local government units are also
excluded.
Definitions
• The stock data are on a point in time basis
• The unit used is the enterprise
• New enterprises are defined as “ex nihilo” creations, i.e. pure births, with reactivations
and take-overs excluded.
28. Turkey
One data source available
Title – Newly Established and Liquidated Companies and Firms
Source – State Institute of Statistics
Internet address –http://www.die.gov.tr/english/SONIST/SIRKET/sirket.html, orhttp://www.die.gov.tr/TURKISH/SONIST/SIRKET/sirket.html (Turkish version)
Contents – Counts of newly established companies and co-operatives and newly established
firms, as well as liquidations of both.
Breakdowns – Some breakdowns by economic activity and geography are available for more
recent data.
Metadata – Some definitions are available in the SIS Data Dictionary
(http://www.die.gov.tr/TURKISH/SOZLUK/dataa.html).
Period covered – Data are available for 1995, and 1997 to 2004.
Coverage – All economic activities seem to be covered, though the counts for agriculture look
rather low. Central and local government units do not seem to be covered.
Definitions
• Firms are defined as “business establishments excluding companies and
cooperatives.”
• Companies are defined as “a number of persons forming an establishment for
commercial purposes as a result of economic and social joining.”
• Co-operatives are defined as “legal entity operating without fixed capital that may be
established by public institutions, provincial special administrations, municipalities,
associations or societies, whose aim is to provide certain economic benefits to
shareholders, especially in relation to their occupation and livelihood through aid and
solidarity.”
82
• Newly established companies and co-operatives, and liquidations, are those
announced in the Turkish Trade Register Gazette.
29. United Kingdom
Four data sources available
a) Title – Value-Added Tax Registrations and De-registrations
Source – Department for Trade and Industry – Small Business Service
Internet address –http://www.sbs.gov.uk/sbsgov/action/layer?r.l2=7000000243&r.l1=7000000229&r.s=tl&topicId
=7000011757
Contents – Counts of the stock of value-added tax (VAT) registered businesses, new
registrations and de-registrations.
Breakdowns – Data are broken down by economic activity and geography.
Metadata – A paper on the methodology used is available via the web site.
Period covered – The data are available for 1994 to 2003. A previous series from 1980 to
1993 is also available, but the data are not directly comparable due to large changes in the
VAT threshold.
Coverage – The data cover all economic activities and legal forms, though coverage is limited
for certain activities that are exempt from VAT, particularly in the education and health
sectors.
Definitions
• The stock data are on a point in time basis (1 J anuary).
• The unit used is the VAT registration, which approximates to the legal unit.
b) Title – Barclays Small Business Surveys
Source – Barclays
Internet address –http://www.business.barclays.co.uk/BRC1/jsp/brccontrol?task=articleFWgroup&value=6502&t
arget=_self&site=bbb
Contents – Data on business start-ups and closures, as well as the total population of
businesses are contained in a series of quarterly reports.
Breakdowns – The data are broken down in different ways each year, including by
geography, economic activity, and sex of the entrepreneur.
Metadata – Limited metadata are available within the reports.
Period covered – Data are available from 1995 to 2004, though for latter years they are
increasingly broadly rounded estimates.
83
Coverage – The data only cover England and Wales. They are based on business current
account openings and closures at Barclays, multiplied by estimates of their share of the
business banking market. This makes it unlikely that central and local government activities
will be included. Businesses that do not operate via business current accounts are also
excluded.
Definitions
• The unit used is the business account, which will be close to the definition of the
enterprise.
• The population data are on a point in time basis.
c) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2003. Data
for births exist for 1998 to 2003, and data for deaths exist for 1997 to 2003.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
d) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for the United Kingdom are available from 1986 to 1997, except 1992.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
84
30. United States
Six data sources available
a) Title – Statistics of US Businesses / Dynamic Data
Source – US Census Bureau
Internet address –http://www.census.gov/csd/susb/susbdyn.htm
Contents – Counts of establishment stock, births, deaths, expansions and contractions, and
associated employment changes.
Breakdowns – The data are broken down by economic activity, size band (based on
employment) and geography.
Metadata – Papers with descriptive metadata and definitions are available via the web site.
Period covered – Data are available for 1995 to 2001.
Coverage – Businesses without employees are excluded. All economic activities are covered
except crop and animal production (NAICS 111,112), rail transportation (NAICS 482), National
Postal Service (NAICS 491), pension, health, welfare, and vacation funds (NAICS 525110,
525120, 525190), trusts, estates, and agency accounts (NAICS 525920), private households
(NAICS 814), and public administration (NAICS 92). Governmental establishments are
excluded except for wholesale liquor establishments (NAICS 4228), retail liquor stores (NAICS
44531), Federally-chartered savings institutions (NAICS 522120), Federally-chartered credit
unions (NAICS 522130), and hospitals (NAICS 622).
Definitions
• The stock data are on a point in time basis (first quarter).
• The unit used is the establishment, defined as “a single physical location where
business is conducted or where services or industrial operations are performed.” This
is broadly equivalent to the local unit.
• Other units referred to are:
o Enterprise - A business organization consisting of one or more domestic
establishments that were specified under common ownership or control. The
enterprise and the establishment are the same for single-establishment firms.
o Firm - A business organization consisting of one or more domestic
establishments in the same state and industry that were specified under
common ownership or control. The firm and the establishment are the same for
single-establishment firms. For each multi-establishment firm, establishments
in the same industry within a state will be counted as one firm.
• Establishment births are establishments that have zero employment in the first quarter
of the initial year and positive employment in the first quarter of the subsequent year.
• Establishment deaths are establishments that have positive employment in the first
quarter of the initial year and zero employment in the first quarter of the subsequent
year.
b) Title – Firm Size Data
Source – US Small Business Administration
85
Internet address –http://www.sba.gov/advo/research/data.html
Contents – Counts of the population of firms, births and deaths. Employment data are also
available.
Breakdowns – Data on the population of firms are broken down by size band (employment)
and economic activity. There are no breakdowns of the data on firm births and deaths.
Metadata – Extensive metadata are available in the paper “Statistics of U.S. Businesses –
Microdata and Tables”, available on the website.
Period covered – Data on the population or firms are available for 1988 to 2002. Data on
births and deaths are available for 1989 to 2001.
Coverage – The coverage is as for source 1 above, as the firm level data are derived from
the US Census Bureau establishment-level Statistics of US Businesses.
Definitions
• The population counts cover all businesses that had an active payroll at any point
during the year, so can be considered as “live during period” data.
• The unit used is the firm, which is defined as “the largest aggregation of business legal
entities under common ownership or control”, so corresponds most closely to the
European definition of the Enterprise Group (truncated or all-residential rather than
global).
• Firm birth and death definitions correspond to those for establishments in source 1
above.
c) Title – Business Employment Dynamics, Quarterly Data
Source – Bureau of Labor Statistics
Internet address –http://www.bls.gov/bdm/home.htm
Contents – Counts and rates for establishment openings and closures each quarter.
Breakdowns – The data can be broken down by economic activity
Metadata – Descriptive metadata and definitions are available via the web site.
Period covered – Data are currently available from quarter 3 of 1992 to quarter 4 of 2004
inclusive.
Coverage – The data exclude business with no employees, central and local government
units, and some non-profit organizations. Certain economic activities are also excluded
(religious organizations, some small farms, the Armed Forces and railways).
Definitions
• No stock data are given, but these can be estimated from birth counts and rates (or
death counts and rates) for the same quarter. These can then be used to calculate
annual birth and death rates. Note; Birth and death data give slightly different stock
figures, but these are all within the margins of error associated with the use of rounded
86
data, and are unlikely to affect the annual birth and death rate estimates by more than
0.2%.
• The unit used is the establishment, which is broadly equivalent to the local unit.
• Openings are either establishments with positive third month employment for the first
time in the current quarter, with no links to the prior quarter, or with positive third month
employment in the current quarter following zero employment in the previous quarter.
• Closings are either establishments with positive third month employment in the
previous quarter, with no positive employment reported in the current quarter, or with
positive third month employment in the previous quarter followed by zero employment
in the current quarter.
d) Title – Business Employment Dynamics, Annualised Data
Source – Bureau of Labor Statistics
Internet address –http://www.bls.gov/opub/mlr/2004/11/art1full.pdf
Contents – This paper gives annualised versions of the data in source c) above, by removing
businesses that enter and exit within the year, and those entries that are really the
continuation of a previous registration.
Breakdowns – The data are not broken down in any way.
Metadata – The metadata for source c) mostly apply. The paper contains information on the
method for annualising the data.
Period covered – Data are available for 1998 to 2001 inclusive.
Coverage and Definitions – As for source c).
e) Title – Longitudinal Business Database
Source – US Census Bureau
Internet address –http://www.ces.census.gov/ces.php/abstract?paper=101647
Contents – The database contains linked records of establishments and firms over time. It
can be used to produce data on business dynamics. The internet address above is that of a
paper describing the database, which includes data on births and deaths. A second paper is
available with more detailed analyses for the retail sector – see:http://www.ces.census.gov/ces.php/abstract?paper=101704
Breakdowns – The data in the paper are not broken down in any way, but the database
would allow a range of detailed breakdowns.
Metadata – The paper contains descriptive metadata.
Period covered – The paper presents stock, birth and death data for 1976 to 1999.
Coverage – The source data cover establishments with paid employees. Economic
activity coverage is the same as for source number 1 above.
87
Definitions
• The stock data are on a point in time basis
• The unit used is the establishment
• Births are records that were active in one year, but not the previous year, adjusted for
reactivations
• Deaths are records that were active in one year, but not the next, adjusted for
reactivations
f) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for the United States are available from 1989 to 1996.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
31. Brazil
One data source available
Title – Estatísticas do Cadastro Central de Empresas - 2001
Source – Instituto Brasileiro de Geografia e Estatística (IBGE)
Internet address –http://www.ibge.gov.br/home/estatistica/economia/cadastroempresa/2000/Publicacao_comple
ta.pdfhttp://www.ibge.gov.br/home/estatistica/economia/cadastroempresa/2001/cempre2001.pdfhttp://www.ibge.gov.br/home/estatistica/economia/cadastroempresa/2002/cempre2002.pdf
Contents – Counts of enterprises and local units. The 2001 and 2002 publications also
contain data on births and deaths in specific sections on business demography.
Breakdowns – Data are broken down by economic activity, size and geography.
Metadata – Some descriptive metadata are available within the publications (in Portuguese).
See also the paper “Brazilian Enterprise Birth and Death rates by economic activity from 1997
to 2001” at:http://forum.europa.eu.int/Public/irc/dsis/businessurvey/library?l=/2003_rome/sessions7simpro
vingsbrsc&vm=detailed&sb=Title
88
Period covered – Data on the stock of enterprises are available for 1999 to 2002. Data on
births and deaths are available from 1997 to 2002.
Coverage – All economic activities and legal forms (including public administration) are
covered in the stock figures. The births and deaths for 1997 to 2001 cover “manufacturing”
(ISIC sections C +D), “trade” (ISIC section G), and services (ISIC sections H +I +J ). Birth
and death data for 2002 also include a category of “others” (ISIC sections A +B +E +Q)
Definitions
• The stock of enterprises is a point in time estimate at 31 December.
• The unit used is referred to as the enterprise, though is equivalent to the legal unit
• A birth in a given year is defined by the existence of an enterprise identification
number in the Business Register that was not found in the preceding year
• A death is the absence of an enterprise identification number that was found in the
previous year
• Birth and death rates were calculated dividing the number of births and deaths in each
year by the population of enterprises of the previous year.
• The birth and death study has not considered mergers and acquisitions as separate
demographic events. Also, as the business register is mainly based on administrative
records, if an enterprise fails to submit an administrative form in a certain year, this can
result in a false death followed by a false birth.
32. China
Data are being prepared from the 2004 economic census. Start-up rates are provisionally
estimated to be between 20% and 30%
Some data for corporate registrations in Hong Kong are available athttp://www.info.gov.hk/cr/key/index.htm
Eurostat Business Demography Indicators
Introduction
Eurostat have started a project to collect harmonised data on business demography from the
Member States of the European Union (EU). Romania and Norway have also participated on
a voluntary basis. The first data collections were in 2001, initially on a pilot basis. Data are
now available from 1997 to 2003, though not all EU countries have participated, and those
that have, have not provided data for all periods. Current data availability is shown in the table
below:
1997 1998 1999 2000 2001 2002
P B D P B D P B D P B D P B D P B D
Belgium X X X X X X X X
Czech
Republic X X X X X X X
Denmark X X X X X X X X X X X X X
Estonia X X X X X X X X X
Spain X X X X X X X X X X X X X X X X
Italy X X X X X X X X X X X X X X X X
Latvia X X X X X X X X X
89
Lithuania X X X X X X X X X
Luxembourg X X X X X X X X X X X X X X X X
Hungary X X X X X X X X X
Netherlands X X X X X X X X X X
Portugal X X X X X X X X X X X X X X X X
Slovenia X X X X X X X X X
Slovakia X X X X X X X X X
Finland X X X X X X X X X X X X X X X X
Sweden X X X X X X X X X X X X X X X X X
United
Kingdom X X X X X X X X X X X X X X X X X
Romania X X X X X X X X X
Norway X X X X X X X X X X X
Key:
P =Population data
B =Births data
D =Deaths data
X =Data available
Some data on survival and growth are also available.
The provision of data on business demography is likely to become compulsory for EU Member
States when the current draft revision of the EU Structural Business Statistics Regulation
comes into force, probably in 2006.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. Countries supplying data are also requested to provide
information on how closely they have followed the recommended methodology.
Coverage
The published data cover all economic activities in NACE sections C to K, except
management activities of holding companies (class 74.15). This means that agriculture,
forestry, fishing, public administration, health, education, other community, social and
personal service activities, activities of households, and extra-territorial organizations and
bodies are excluded. All legal forms are covered except central and local government, and
non-profit organisations serving households.
The coverage of data from individual countries is also influenced by the coverage of their
business registers, particularly in terms of size thresholds. These are in turn influenced by
national administrative sources, which vary considerably from country to country. For
example, the current value-added tax threshold in the United Kingdom is around €85,000,
whereas it is zero, or close to zero in most other countries. These differences can be partly
offset (as in the UK) by using a range of sources to improve coverage, but they still lead to
noticeable differences in data for the smallest size classes.
Definitions
• The statistical unit is the enterprise, however, the methodology used recognises that
some countries only hold data at the level of the legal unit, and attempts to
compensate for this through matching routines.
• The population of active enterprises consists of all enterprises that had either turnover
or employment at any time during the reference period, i.e. it is on a “live during
period” basis.
90
• Enterprise births are defined as the creation of a combination of production factors
with the restriction that no other enterprises are involved in the event. They include
enterprises started by a person who previously performed the same activity, but as an
employee, and newly born national or foreign subsidiaries that are real enterprises
(legal units rather than just local units or branches) with autonomy of decision making,
where new production factors are created, rather than transferred from another unit.
The following categories are excluded:
o Enterprises that are created by merging production factors or by splitting them
into two (or more) enterprises (break-ups, mergers, split-offs, restructuring)
o Newly created enterprises that simply take over the activity of a previously
created enterprise (take-over)
o Any creations of additional legal units/enterprises solely for the purpose of
providing a single production factor (e.g. the real estate or personnel) or an
ancillary activity (see note below) for an existing enterprise.
o An enterprise that is registered when an existing enterprise changes legal
form. E.g. a successful sole proprietor moves operations from his home to
another location and at the same time changes the legal form of the enterprise
to a limited liability company.
o Reactivated enterprises if they restart activity within 2 calendar years.
o Temporary associations and joint ventures that do not involve the creation of
new factors of production.
• Enterprise deaths are defined as the dissolution of a combination of production factors
with the restriction that no other enterprises are involved in the event. Events leading
to the closure of an enterprise that are not considered to be deaths are:
o Enterprises that close down due to merging or breaking-up of production
factors (break-ups, mergers, restructuring)
o Enterprises whose activity is taken over by another enterprise (take-over)
o Enterprises that are deleted due to a change of legal form, e.g. a successful
sole proprietor moving operations from his home to another location and at the
same time changing the legal form of the enterprise to a limited liability
company is a case that should be excluded.
o Reactivated enterprises if they restart activity within 2 calendar years.
The DOSME Project
Introduction
DOSME (Demography of Small and Medium-sized Enterprises) was the name given to a
series of projects to develop statistical business registers and data on business demography
and factors of success in a group of Central and Eastern European countries. The DOSME
projects were financed by the European Union, through Eurostat. Data were collected from
samples of new and existing enterprises between 1994 and 2000, and have been
consolidated into a single database held by Eurostat.
Details of the data collections, methodology and publications from this series of projects are
all available on the DOSME web site
30
. Some of the countries involved have continued the
data collections since the end of the project, but most have now started to participate in the
Eurostat Business Demography Project. Data are available for Bulgaria, Czech Republic,
Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia for 1995 to
2000. Some data are also available for Albania and Macedonia for earlier periods.
30
http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
91
Coverage – NACE Rev. 1 sections A, B and L (agriculture, forestry, fishing and public
administration) are excluded, as are central and local government units. The data are based
on survey results weighted using business register counts to give whole population estimates.
This limits the coverage of the final data sets in terms of the availability and quality of detailed
breakdowns.
Definitions
• The stock data used are point in time estimates, though approximations to “live during
period” populations have been possible by adding births during a particular year to the
stock at the start of that year.
• The unit used is theoretically the enterprise, but in practice it is usually the legal unit.
As the surveys focussed mainly on small units, the impact of this is probably
negligible.
• Births and deaths are defined as registrations and de-registrations with the relevant
administrative sources. It is recognised that these may be subject to lags, and that
some types of businesses were not required to register in certain countries.
The OECD Firm-Level Data Project
Introduction
The OECD firm-level project involved bringing together data from ten OECD countries (United
States, Germany, France, Italy, United Kingdom, Canada, Denmark, Finland, the Netherlands
and Portugal). It aimed to use on a common analytical framework, including the
harmonisation, to the extent possible, of key concepts (e.g. entry, exit, or the definition of the
unit of measurement) as well as the definition of common methodologies for studying firm-
level data.
The data were used to analyse firm demographics, resulting in a number of papers available
via the project web site:http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
The data were derived from business registers (Canada, Denmark, France, Finland,
Netherlands, United Kingdom and United States) or social security databases (Germany and
Italy). Data for Portugal were drawn from an employment-based register containing
information on both establishments and firms. These databases allow firms to be tracked
through time because addition or removal of firms from the registers (at least in principle)
reflects the actual entry and exit of firms.
Coverage
While some data sources included even single-person businesses, others omitted firms
smaller than a certain size, usually in terms of the number of employees, but sometimes in
terms of other measures such as sales (as is the case in the data for France and Italy).
Analyses based on the data from this project typically exclude single-person businesses.
However, because smaller firms tend to have more volatile firm dynamics, remaining
differences in the threshold across different country datasets should be taken into account in
the international comparison.
The data were compiled on an annual basis, covering varying time spans. The German,
Danish and Finnish register data cover the longest time periods, while data for the other
92
countries are available for shorter periods of time or, although available for longer periods,
include significant breaks in definitions or coverage.
Special efforts were made to organise the data along a common industry classification (ISIC
Rev.3). In countries where the data collection by the statistical agency varied across major
sector (e.g., construction, industry, services), a firm that switched between major sectors
could not be tracked as a continuing firm but ended up creating an exit in one sector and an
entry in another. Most countries have been able to provide firm demographic data across
most sectors of the economy, with the exception that public services are often not included
(the United Kingdom is a special case where data only refer only to manufacturing).
Definitions
• Unit of observation: Data used in the study refer to the firm as the unit of reference,
with the exception of Germany where data are only available with reference to
establishments, and Finland where data are reported with reference to both firms and
establishments.
• Firm entry: The number of firms entering a given industry in a given year. It comprises
firms observed as (out, in, in) the register in time (t – 1, t, t +1).
• Firm exit: The number of firms that leave the register. It comprises firms observed as
(in, in, out) the register in time (t – 1, t, t +1).
• Continuing firms: The number of firms that were in the register in a given year, as well
as in the previous and subsequent year. It comprises firms observed as (in, in, in) the
register in time (t – 1, t, t +1).
• One-year firms: The number of firms that were present in the register for only one
year. It comprises firms observed as (out, in, out) the register in time (t – 1, t, t +1).
93
Annex 3 - Defining Business Populations: Comparing Point in Time
and Live During Period Estimates
1. Introduction
The definition of a population of businesses can have a significant impact on any data derived
from it. This Annex looks at different ways in which business populations have been defined in
terms of the time dimension. It focuses specifically on the role of populations in business start-
up rates, but obviously has the potential for wider application. It concludes by proposing a
model to help improve the international comparability of business data. The terminology used
in this Annex is consistent with that proposed in Annex 1.
There are two main approaches to defining business populations with respect to time. A “point
in time” population is a relatively simple concept, and consists of all businesses deemed to be
in scope at a given point in time, usually on a specific reference day. A “live during period”
population, however, consists of all businesses that were in scope at any point during a given
reference period. This Annex aims to explain the nature of the differences between these two
approaches and look at ways to estimate the impact of using different populations. The focus
is on deriving methods to convert data compiled on one basis to provide more comparable
estimates.
It is clear that a live during period population will be larger than one constructed on a point in
time basis. The extent of the difference will depend on various factors, but mainly on the
length of the period, and the degree of churn (i.e. joiners and leavers) in the business
population. As a result, business start-up data compiled using a point in time population are
not likely to be comparable with those based on a live during period approach.
Differences in populations compiled using the two approaches are typically in the range of
10% to 15% for annual data. The choice of population can therefore affect start-up rates by up
to 2%, with the impact highest when start-up rates themselves are high. This issue is of
particular relevance for international comparability purposes, as business demography data
from Eurostat are compiled using live during period populations, whereas almost all other
sources favour the point in time approach.
2. Components of the population
Point in time and live during period populations of businesses can be broken down into a
number of components, which can then be re-aggregated in different ways to give different
types of population estimates. The basic components are shown in Figure 1 below.
94
Figure 1: A Simple Model for Business Populations
In this model:
• PA
t
=The population at the start of period t
• PB
t
=The population at the end of period t
• S
t
=businesses present in both populations (i.e. “survivors”)
• L
t
=businesses that are in population PA
t
, but not PB
t
(i.e. “leavers”)
• J
t
=businesses that are not in population PA
t
, but are in B
t
(i.e. “joiners”)
• J L
t
=businesses that are not present in PA
t
or PB
t
, but would be present in an
intermediate population (i.e. they join and leave within period t)
The population of businesses considered in scope at the start of the period (PA
t
), sometimes
referred to as the opening stock, can be defined as: PA
t
=S
t
+L
t
. Similarly the population at
the end of the period (PB
t
), or closing stock, can be defined as: PB
t
=S
t
+J
t
. As PB
t
=PA
t+1
, it
follows that: PA
t
=S
t-1
+J
t-1
, PB
t
=S
t+1
+L
t+1
, and that PB
t
=PA
t
+J
t
– L
t.
Businesses in the
sub-set J L
t
do not appear in either population.
Total entries to the population are defined as E
t
=J
t
+J L
t
, and total exits as X
t
=L
t
+J L
t
, so P
t
=PA
t
+E
t
=PB
t
+X
t
. Thus to convert from a point in time to a live during period population, it
is necessary to know, or have reasonable estimates for E
t
or X
t
, or J L
t
and one of L
t
or J
t
. In
practice, J L
t
is rarely available from published data sources, and such businesses are usually
ignored as they are not present in PA
t
or PB
t
, thus there is a risk that P
t
could be
underestimated. If they are included, data usually take the form of E
t
and / or X
t
so they
appear in both, rather than as a separate group. The latter is usually the case in data sets
based on population observations at a series of intermediate points between PA
t
and PB
t
. The
size, and hence the importance of J L
t
will depend on the length of period t. If t is one month, it
is relatively safe to assume that J L
t
is very small. If PA
t
and PB
t
are derived from economic
censuses with a five year interval, however, J L
t
will be much larger.
Eurostat “live during period” enterprise survival data covering 48 observations for 18 countries
over 4 years show that on average 87.23% of births in a given year are also active in the
following year. This indicates that the size of J L
t
is typically 14.64% of that of J
t
, where t is one
year. Thus for annual estimates, where data on J
t
are available, but data on J L
t
are not, it
would be reasonable to define entries as E
t
=J
t
(1 +0.1464) and estimate P
t
or PA
t
using: P
t
=
PA
t
+J
t
(1 +0.1464). The value of 0.1464 is the best current estimate of the ratio of J L
t
to J
t
,
for European Union countries, and will obviously vary over time and space, so this value is
replaced by c in the remainder of this Annex.
t
S
t
L
t
J
t
J L
t
PA
t
PB
t
95
3. Other Models
3.1 OECD Firm-level Data Project
Figure 2: The OECD Firm-level Data project Model
Note – different notation is used in this model:
S
t
=Survivors for year t (defined as in / in / in, for t-1 / t / t+1)
X
t
=Exits for year t (defined as in / in / out, for t-1 / t / t+1)
E
t
=Entries for year t (defined as out / in / in, for t-1 / t / t+1)
O
t
=One year firms (defined as out / in / out, for t-1 / t / t+1)
A point in time approach was used for this data collection project. Firms that entered and
exited between observations are not recorded, and “one year firms” are also excluded from
most analyses due to data quality concerns. This effectively defines firm entries and exits only
in terms of firms that existed for at least one year.
This project also took the relatively unusual approach of identifying entries for period t as firms
that appeared during the period t-1 to t, whereas exits are defined as those that disappeared
between t and t+1.
t -1 t +1 t
S
t
O
t
X
t-1
E
t
X
t
O
t-1
E
t+1
O
t+1
96
3.2 Eurostat Business Demography
Figure 3: The Eurostat Business Demography Model
Notation:
t =Reference year
N
t
=Population of enterprises active at any point during t
R
t
=Real births in t
X
t
=Other entries in t
D
t
=Real deaths in t
Y
t
=Other exits in t
The Eurostat model in Figure 3 is based on the “live during period approach”. This project is
ongoing, and will be extended to all European Union (EU) countries following a revision to the
EU Structural Business Statistics Regulation. The definitions of the various populations do not
take into account the length of survival. A business can be a birth and a death in the same
period, and will also be counted in the population of active enterprises for that period. A
methodology based on checking for reactivations (within two years), then matching, then
manual inspection of large units, is used to separate “real” (i.e. pure) births and deaths from
other entries and exits.
3.3 National Data Models
A wide variety exists in the models used to define the populations used for national business
demography data. Some of the differences are purely down to terminology or notation, and
most models can be seen as derivatives of one or more of those presented above. For
example, most sources in the United States use models similar to that shown in Figure 1,
whereas the Australian model mixes elements from Figure 1 and Figure 3.
4. Purity of Entries and Exits
A complicating factor for many population models is that for business demography purposes,
it is often of interest to split total entries (E
t
) and exits (X
t
) into a number of sub-components
linked to different types of demographic events. A basic distinction for many data users, as
t
N
t
D
t
X
t
R
t
Y
t
97
seen in the Eurostat model, is to split entries into pure births
31
(also referred to as creations
ex-nihilo), and other entries (e.g. due to restructuring, re-registration, reactivation, merger,
break-up or split-off), with a similar split for exits into pure deaths and other exits. If more
detail is required, and available, each different demographic event can form a separate sub-
component of entries and exits.
32
The model in Figure 4 assumes a simple split of entries and exits on the basis of purity, i.e.
separating pure births (B
t
) from other entries (OE
t
), and pure deaths (D
t
) from other exits
(OX
t
). It also assumes that the length of survival of the business is not relevant, in that
businesses are included whether or not they survive into the subsequent period.
Figure 4: Developing the Concept of Purity
Unfortunately, one problem with the live during period approach is that a proportion of other
entries (OE
t
) will be due to new businesses taking over the activities of businesses recorded
as other exits (OX
t
). Technically, many of these cases should be considered as the continuity
of a business, and should not be recorded as entries and exits. However, as most data
sources are based either directly or indirectly on registrations and de-registrations with
administrative or tax sources, it is unlikely that all such take-over cases are recorded as
business continuity, particularly for small businesses. This will vary from country to country
and between sources, depending on the nature of the source and register maintenance
procedures. The way in which business continuity is treated in the source will therefore affect
the degree of duplication in a live during period population. This, in turn, will affect the
comparability of indicators based on live during period populations.
31
Defined in the Eurostat Business Demography Methodological Manual as: “the creation of a
combination of production factors with the restriction that no other enterprises are involved in the event.
Births do not include entries into the population due to: mergers, break-ups, split-off or restructuring of a
set of enterprises. It does not include entries into a sub-population resulting only from a change of
activity”.
32
A typology of demographic events is proposed in Chapter 13 of the Eurostat Business Registers
Manual of Recommendations, see:http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/embs/registers/chapter13.doc
t
E
t
OX
t
OE
t
B
t
D
t
X
t
=X
t
E
t
=
98
5. Towards a Standard Model for Business Populations
So far, this Annex has concentrated on understanding and explaining the different models
used to define populations for business demography purposes. The next logical step is to look
for ways to move towards common standards, with the aim of improving international data
comparability.
A major constraint is that any changes to current methods are likely to have costs both
financially and in terms of comparability of data series over time. Thus an approach is
required that minimises the impact of any change, whilst maximising comparability and
standardising terminology. It also needs to recognise the requirements of splitting entries
based on both purity and survival through an observation point.
Figure 5: A Standard Model?
t
S
t
D
t
OE
t
BS
t
OX
t
PA
t
PB
t
BD
t
BOX
t
OED
t
OEOX
t
B
t
OES
t
SD
t
SOX
t
B
t
=
OE
t
=
OED
t
BD
t
=D
t
BOX
t
OEOX
t
=OX
t
99
The point in time population at the start of period t (PA
t
) can be defined by rearranging the
equation P
t
=PA
t
+E
t
from Section 2, to give PA
t
=P
t
– E
t
. The live during period population
(P
t
) is defined in the same way as in the Eurostat model (referred to there as N
t
). E
t
gives the
total number of new businesses in period t, and is equivalent to B
t
+OE
t
(pure births plus
other entries) in Figure 4. Therefore PA
t
can be defined as PA
t
=P
t
– (B
t
+OE
t
), or using the
Eurostat notation, PA
t
=N
t
– (R
t
+X
t
).
33
Eurostat do not currently publish data on X
t
, but as
this is a by-product of the data production process it is relatively easy to obtain. In this way,
consistent measures of PA
t
can be defined based on data from a wide range of models.
PA
t
can also be defined as the survivors throughout period t plus the leavers in period t that
were live in the previous period. From Figure 1; PA
t
=S
t
+L
t
, however, in Figure 5, L
t
is split
into SD
t
(survivors from the previous period that were pure deaths in t) and SOX
t
(survivors
from the previous period that were other exits in t), so in terms of Figure 5, PA
t
=S
t
+SD
t
+
SOX
t
.
Either P
t
or PA
t
can be used as the basis for business demography statistics. The point in time
approach (PA
t
) is currently the more popular of the two, and has the advantages of being a
relatively simple concept to explain, and being analogous to the population basis used for
human demography statistics. It also largely avoids the potential duplication issues that can
affect the comparability of P
t
(see Section 4 of this Annex). For these reasons, the use of PA
t
is recommended, though it should be remembered that it is relatively simple to substitute P
t
if
required.
Having defined a stock population, the next step is to determine the dynamic populations of
entries and exits. In Figure 1, the sum of the entries in period t is defined as J
t
+J L
t
(or
J
t
(1+c)). The total number of entries in period t from Figure 5 is B
t
+OE
t
, which is equivalent to
R
t
+X
t
in the Eurostat notation (Figure 3). Thus a standard measure of entries (E
t
) can be
calculated reasonably easily from all sources.
Unfortunately, national factors (e.g. units, sources, coverage, definitions, thresholds etc.), and
other movements into and out of scope, can influence total entries, reducing the value of this
variable for international comparison purposes. A more comparable variable should be the
number of pure births, but this assumes perfect knowledge to separate pure births from other
entries. Much progress in developing a standard methodology for this process has been made
through the Eurostat business demography project, such that reasonable estimates of B
t
, and
OE
t
are now possible for a number of European countries.
Figure 5 breaks down B
t
and OE
t
(and the corresponding variables for leavers, D
t
and OX
t
)
into three components. B
t
can be seen as consisting of businesses that survive into t+1 (BS
t
),
and those that do not. The latter category can be broken down into pure deaths (BD
t
) and
other exits (BOX
t
). If similar components are derived for OE
t
, D
t
and OX
t
(note: some
components are shared), these components can be re-grouped to form populations J
t
(=BS
t
+
OES
t
), L
t
(=SD
t
+SOX
t
), and J L
t
(=BD
t
+BOX
t
+OED
t
+OEOX
t
).
Similar issues apply to the way the population of leavers is defined. D
t
offers purity, and the
potential for greater comparability, whereas L
t
may be easier to measure in practice. To
facilitate data conversion, it will also be necessary to calculate an equivalent to the value c to
express J L
t
as a proportion of L
t
.
One issue not covered so far is how to deal with reactivations, i.e. businesses that leave a
population (by closing temporarily) then re-join it. Recording these as a death followed by a
33
The robustness of the relationship described by these equations depends on the extent of duplication
in P
t
(see the last paragraph of Section 4). The higher the degree of duplication, the less robust the
relationships.
100
birth does not fit well with the purity approach, particularly if the period of closure is short. For
Eurostat business demography purposes, a two year threshold is applied, so that periods of
closure of less than two years do not result in deaths and births.
If the period of temporary closure does not include a point at which the population is
observed, i.e. the temporary closure starts and ends between the dates of PA
t
and PB
t
, it may
not be recorded. This means that, if t is one year, in theory a business can be inactive for over
11 months, yet still be recorded as having survived throughout the period, whereas a business
that closes for a few days either side of the PB
t
reference date would be recorded as a leaver
in t, and a joiner in t+1.
To improve consistency, a rule that a business has to be out of the population for at least two
consecutive observations to be considered a pure death and birth seems reasonable where t
is one year. Thus a reactivation that was out of the population for just one period (e.g. PA
t
)
would be included in the populations for other events (SOX
t-1
, and OES
t
) rather than those for
pure births and deaths. The only real problem with this approach is that it introduces a lag for
data on deaths, though this can be at least partially overcome by estimation.
For completeness, in addition to the total population, entries and exits, it is useful to determine
the population of businesses that survive throughout the period (or are at least present at the
start and the end of the period), S
t
. S
t
can be defined simply (from Figure 1) as PB
t
– J
t
. To
relate this to the Eurostat populations in Figure 3, it is necessary to refer to PB
t
as PA
t+1
,
which has been shown above to correspond to N
t+1
– (R
t+1
+X
t+1
). J
t
has been shown to equal
to (R
t
+X
t
) / (1+c). Thus, by substitution, S
t
=N
t+1
– (R
t+1
+X
t+1
) – ((R
t
+X
t
) / (1+c)).
Having derived the various populations of interest, it is useful to note certain logical
relationships based on a stock and flow basis, which can be used as a quality check, or to
derive a missing population. The basic equation is that opening stock, plus entries, minus
exits should equal closing stock: PA
t
+E
t
– X
t
=PB
t
(it also follows that PA
t
+J
t
– L
t
=PB
t
). In
terms of Figure 5, this equation can also be expressed as PA
t
+(BS
t
+OES
t
) – (SD
t
+SOX
t
) =
PB
t
.
Having defined the populations referred to above, it is then relatively straightforward to apply
them to study business survival and growth rates, though this is beyond the scope of this
Annex. The proposed standard model also implies the introduction of harmonised
terminology, the various elements of which are defined in Annex 1.
6. Conclusions
It seems feasible to apply a standard model for defining business populations that can accept
data from a variety of sources using different methodologies. As a result, it should be possible
to improve the international comparability of data on business populations, business
demography, and small and medium–sized enterprises (SMEs), whilst not imposing significant
additional burdens on data suppliers in national statistical institutes.
This approach will remove, or at least reduce the impact of a number of the different factors
affecting comparability identified in this report. The next logical step is therefore to test the
proposed model, and refine it where necessary, using data from as many countries as
possible.
101
Annex 4 – Business Start-up Data for Selected Countries:
Comparisons of National Sources
Introduction
This Annex looks at intra-country comparability of business start-up data for 10 countries for
which three or more sources allowing the construction of start-up rates have been identified in
the inventory in Annex 2. The purpose of this exercise is to try to explain the differences
between sources in terms of the factors of comparability identified in the main body of this
report. This work is based on the assumption that any differences in data relating to the same
country and the same time period must be purely methodological in nature.
To facilitate comparability the data shown have been summarised, and converted to a
standard format. This has, in some cases, included the calculation of birth and death rates.
1. Canada
Three sources of data on business start-ups have been identified for Canada.
a) Business Dynamics in Canada
Statistics Canada – Longitudinal Employment Analysis Program (LEAP) Database
(supplemented for 2002 by data prepared for FORA).http://www.statcan.ca:8096/bsolc/english/bsolc?catno=61-534-X
This publication includes data on business populations, birth and death rates, and survival. It
also contains a chapter on methodology. Business populations, births and deaths are broken
down by size and economic activity categories based on knowledge intensity. Business
populations are also broken down by geography.
Population estimates are available for 1991 to 2001 and counts of births are available from
1992 to 2001. Additional data on population and births for 2002, and deaths for 2001 is taken
from a short report prepared by Statistics Canada for FORA.
The data cover all employers in Canada, public and private (i.e. data do not include
businesses with no employees). The unit used is the firm, which, at the national level is
equivalent to the legal unit. Births are defined as firms that are not present on the LEAP
database in year t-1, but are present in year t. The birth rate is the number of new enterprises
in t divided by the total number of firms observed in year t.
The data in the publication are protected by copyright, so are not reproduced here.
b) Self-Employment Entry and Exit Flows
Statistics Canada – Paper by Zhengxi Lin, Garnett Picot and J anice Yateshttp://www.statcan.ca/english/research/11F0019MIE/11F0019MIE1999134.pdf
This paper contains a table with counts of self-employed persons, and rates for entry and exit,
as well as other related data, analyses, and descriptive metadata on sources and definitions.
No breakdowns of the entry data are given in the paper. The population estimates are
available for 1981 to 1995, and entries are available from 1982 to 1995.
102
The data cover all persons whose self-employment earnings are the dominant source of
earnings in the year according to their annual tax returns to revenue Canada, but are based
on a 10% sample. The unit is therefore the person completing the tax return. Self-employment
entries are income-tax filers who report earnings from self-employment in one year but not the
previous year.
Population Entries Birth Rate
1981 915,140
1982 931,240 194,750 20.91%
1983 953,350 197,700 20.74%
1984 988,590 208,030 21.04%
1985 990,980 197,280 19.91%
1986 1,019,390 221,760 21.75%
1987 1,069,690 248,630 23.24%
1988 1,099,470 248,370 22.59%
1989 1,125,410 253,710 22.54%
1990 1,159,370 269,500 23.25%
1991 1,191,930 273,190 22.92%
1992 1,253,290 293,330 23.40%
1993 1,334,050 312,620 23.43%
1994 1,400,760 330,810 23.62%
1995 1,471,800 355,940 24.18%
c) OECD Firm-level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
Canada are available for 1984 to 1997.
Annual point in time populations were used, based on the Canadian statistical business
register. New businesses had to be present in both the reference year and the following year
to be counted as a birth in published analyses of the data. “One-year” businesses were
identified separately, but have been included as births in the table below to try to improve
comparability with other sources. Businesses without employees were excluded
Year Population Total Entries Start-up Rate
1984 701,115 128,837 18.38%
1985 729,929 119,424 16.36%
1986 757,980 117,843 15.55%
1987 779,956 116,916 14.99%
1988 781,594 103,998 13.31%
1989 783,415 103,096 13.16%
1990 798,855 121,773 15.24%
1991 796,223 115,662 14.53%
1992 798,215 113,740 14.25%
1993 801,127 114,308 14.27%
1994 808,849 117,512 14.53%
1995 810,336 88,301 14.46%
1996 806,777 92,570 14.72%
1997 825,389 92,686 15.07%
103
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
Self-employment Business Dynamics in Canada OECD Firm-level Data Project
The chart shows an apparently high degree of convergence for data from the OECD Firm-
level project and Business Dynamics in Canada for the years where the data overlap. As
would be expected, birth rates are much higher for self-employment businesses. Data before
1990 show rather more variability, which may be genuine, or may indicate that these data are
less reliable.
b) New Businesses
0
50
100
150
200
250
300
350
400
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
T
h
o
u
s
a
n
d
s
Self-employment Business Dynamics in Canada OECD Firm-level Data Project
104
This chart shows that whilst start-up rates were more or less equivalent for data from the
OECD firm-level project, and Business Dynamics in Canada, the levels of new businesses are
not so close, though both follow a similar, rather stable trend during the period of overlap.
There is no readily apparent explanation for the difference between these sources in the
metadata, though it is almost certainly due to different coverage, possibly of the public sector.
The number of new self-employed businesses shows a rapid growth during the 1990’s,
suggesting that self-employed businesses are becoming increasingly important in Canada.
c) Business Populations
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
M
i
l
l
i
o
n
s
Self-employment Business Dynamics in Canada OECD Firm-level Data Project
The patterns here are quite similar to those for new businesses, with a rapid growth of self-
employment businesses, and steady, parallel trends for the employer businesses covered by
the other two sources.
105
2. Denmark
Three sources of data on business start-ups have been identified for Denmark.
a) Statistical Yearbook
Statistics Denmark:
2005 (data for 2001) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2005.aspx
2003 (data for 2000) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2003.aspx
2001 (data for 1999) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2001.aspx
2000 (data for 1998) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2000.aspx
The yearbooks contain counts of new enterprises broken down by economic activity, for 1998
to 2001, as well as the end-year population (used as the start population for the following year
in the table below). They contain very little metadata. Data exclude agriculture and public
administration, and the unit used is the enterprise.
Year Population New Enterprises Start-up Rate
1998 16,063
1999 279,037 17,734 6.36%
2000 284,446 18,640 6.55%
2001 284,166 16,447 5.79%
2002 281,653
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Danish data for the population of active enterprises are available for 1997 to 2001, and data
on births cover 1998 to 2001.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Danish statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
106
Year Population Births Birth Rate
1997 243,946
1998 245,762 24,755 10.07%
1999 253,887 27,562 10.86%
2000 261,911 26,137 9.98%
2001 261,926 24,275 9.27%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
Denmark are available for 1981 to 1994.
Annual point in time populations were used, taken at the end of November each year from the
Danish pay and performance database. New businesses had to be present in both the
reference year and the following year to be counted as a birth in published analyses of the
data. “One-year” businesses were identified separately, but have been included as births in
the table below to try to improve comparability with other sources. Businesses without
employees were excluded
Year Population Total Entries Start-up Rate
1981 136512 21334 15.63%
1982 136373 18148 13.31%
1983 137585 18162 13.20%
1984 139760 18414 13.18%
1985 140914 17961 12.75%
1986 138496 13609 9.83%
1987 140826 16054 11.40%
1988 138772 15491 11.16%
1989 136380 14099 10.34%
1990 135448 14857 10.97%
1991 133565 15420 11.54%
1992 132187 14778 11.18%
1993 126070 13282 10.54%
1994 126113 12567 9.96%
107
Graphical Comparisons
a) Birth Rates
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
OECD Firm-level Data Statistical Yearbook Eurostat Business Demography
All sources seem to indicate a downwards trend in birth rates over time. The difference
between the data from the Statistical Yearbook and those from Eurostat is not easy to explain
based on the available metadata, but may be related to coverage.
b) New Businesses
0
5
10
15
20
25
30
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Statistical Yearbook Eurostat Business Demography
The number of new businesses in the OECD firm-level data set is generally lower than the
other sources due to the exclusion of non-employer businesses, and of new businesses that
did not survive for at least a year. Again the data from Eurostat are higher than those from the
statistical Yearbook, which is not easy to explain given the greater restrictions of the Eurostat
data in terms of coverage of economic activity, as well as the requirements for a relatively high
level of purity in this source.
108
c) Business Populations
0
50
100
150
200
250
300
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Statistical Yearbook Eurostat Business Demography
The one employee threshold used for the OECD firm-level data clearly has an impact on the
population, appearing to remove around half of the units included in the other sources.
Interestingly, the Statistical Yearbook data show higher population levels than those from
Eurostat. This appears to be rather counter-intuitive compared to the number of new
businesses. The Eurostat population would normally be expected to be larger as it is on a live
during period basis, but the impact of this seems to be outweighed by the relatively limited
coverage of economic activities.
109
3. Finland
Three sources of data on business start-ups have been identified for Finland.
a) Enterprise Openings and Closures
Statistics Finland –http://www.stat.fi/til/aly/index_en.html, Data only accessible via Finnish
version -http://www.stat.fi/til/aly/index.html
This source contains counts of enterprise openings and closures. Stock figures are available
separately from the StatFin database, but may not have the same coverage. The data are
broken down by economic activity, legal form and geography, and are available for 1999 to
2004. Some metadata are available, mostly in Finnish.
The openings data are derived from Statistics Finland’s business register. They only cover
those enterprises engaged in business activity that are liable to pay value-added tax or act as
employers. Foundations, housing companies, voluntary associations, public authorities and
religious communities are excluded. The data cover state-owned enterprises, but not those
owned by municipalities. The unit used is the enterprise
Year Stock Openings Birth Rate
1999 219,516 21,460 9.78%
2000 222,817 22,361 10.04%
2001 224,847 21,942 9.76%
2002 226,593 22,190 9.79%
2003 228,422 23,886 10.46%
2004 253,617 24,756 9.76%
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Finnish data for the population of active enterprises exist for 1997 to 2002 and data on births
cover 1998 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Finnish statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
110
Year Population Births Birth Rate
1997 229,786
1998 234,521 19,659 8.38%
1999 233,380 17,581 7.53%
2000 233,451 16,614 7.12%
2001 235,746 16,841 7.14%
2002 237,065 17,174 7.24%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
Finland are available for 1989 to 1997.
Annual point in time populations were used, taken from the Finnish statistical business
register. The coverage of the register improved in 1994 for smaller enterprises, which may
account for at least part of the peak in birth rates in 1994/5. New businesses had to be
present in both the reference year and the following year to be counted as a birth in published
analyses of the data. “One-year” businesses were identified separately, but have been
included as births in the table below to try to improve comparability with other sources.
Businesses without employees were excluded.
Year Population Total Entry Start-up Rate
1989 231,311 46,791 20.23%
1990 252,426 30,207 11.97%
1991 210,501 17,271 8.20%
1992 209,982 25,745 12.26%
1993 179,549 17,961 10.00%
1994 176,804 21,830 12.35%
1995 194,092 25,018 12.89%
1996 202,085 18,883 9.34%
1997 207,008 14,991 7.24%
111
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
Eurostat Business Demography Enterprise Openings and Closures OECD Firm-level Data
The Enterprise Openings and Closures data show a fairly similar trend to those from Eurostat
during the period of overlap. The higher levels of the former are likely to be due mainly to the
use of a point in time population, and the greater purity of the Eurostat data. The fluctuations
in the OECD firm-level data look odd compared to the relative smoothness of the other two
series. This may be partly due to the turbulence in the Finnish economy in the early 1990’s,
but it also seems likely that there is rather a lot of “noise” in the data.
b) New Businesses
0
5
10
15
20
25
30
35
40
45
50
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
Eurostat Business Demography Enterprise Openings and Closures OECD Firm-level Data
Again the OECD firm-level data show much more variability, with the figure for 1989 looking
particularly implausible. The levels also seem rather high given that the metadata state that
businesses without employees were excluded, particularly as that the Eurostat data indicate
that around 60% of births had no employees. As mentioned above, purity is likely to account
for much of the difference in levels between the Enterprise Openings and Closures and the
112
Eurostat data, though the limited coverage of economic activities in the latter will also play a
part.
c) Business Populations
0
50
100
150
200
250
300
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
Eurostat Business Demography Enterprise Openings and Closures OECD Firm-level Data
In terms of the business population, the OECD firm-level data show much more stability,
particularly for latter years, though the level still looks too high if businesses with no
employees are really excluded. The other two sources are close in both trend and level,
suggesting that the limited coverage of the Eurostat data is cancelled out by the effects of
using a live during period population. The 2004 figure for the Enterprise Openings and
Closures data looks rather high, and represents a significant deviation from the trend.
Unfortunately it is too early to tell whether or not it will be confirmed by the Eurostat data.
113
4. France
Three sources of data on business start-ups have been identified for France
a) Créations d'entreprises
INSEE:http://www.insee.fr/fr/ffc/chifcle_liste.asp?theme=9&soustheme=1&souspop=
This source contains counts of enterprise creations, split into new creations, resumptions and
re-activations, as well as some population data. Data are broken down by economic activity,
size and legal form. Some metadata are available, for example key definitions. Data are
available from 1993 to 2004, and cover all of France, including the overseas départements.
Three categories of enterprise creation are identified:
• Pure creations (creations “ex nihilo”) where the new enterprise does not take over the
activities of a previously existing enterprise.
• Reactivations, where a person who has previously been self-employed re-starts a self-
employed activity.
• Resumptions, where a new business takes over an activity previously carried out by
another enterprise.
Year Stock Creations Creation Rate Pure Creations Pure Creation Rate
1993 2,307,638 272,264 11.80% 169,620 7.35%
1994 292,847
1995 283,608
1996 273,811 170,233
1997 269,430 165,277
1998 264,601
1999 266,919
2000 270,043 174,718
2001 2,417,950 268,619 11.11% 175,140 7.24%
2002 2,468,786 268,459 10.87% 176,378 7.14%
2003 2,498,082 291,986 11.69% 197,675 7.91%
2004 2,568,647 318,757 12.41% 222,747 8.67%
b) La Création en Chiffres
Agence Pour la Création d’Entreprises (APCE):http://www.apce.com/index.php?rubrique_id=261&type_page=I
This source contains counts of enterprise creations, split into new creations (“ex nihilo”),
resumptions and re-activations. No breakdowns are given, and metadata are very limited.
Data are available from 1993 to 2004, and are very similar to, but slightly higher than those
from INSEE, perhaps suggesting a slight timing or coverage difference.
114
Year Total Creations Creations Ex Nihilo
1993 273,462 170,919
1994 294,131 183,764
1995 284,853 178,923
1996 275,275 171,628
1997 271,088 166,850
1998 266,446 166,190
1999 268,919 169,674
2000 272,072 176,754
2001 270,564 177,015
2002 270,206 178,008
2003 293,840 199,399
2004 320,265 223,995
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
France are available for 1990 to 1996.
Annual point in time populations were used, taken at the end of the year from a fiscal
database and an enterprise survey. New businesses had to be present in both the reference
year and the following year to be counted as a birth in published analyses of the data. “One-
year” businesses were identified separately, but have been included as births in the table
below to try to improve comparability with other sources. Manufacturing businesses with an
annual turnover of less than 3.8 million French Francs, and service businesses with a turnover
of less than 1.1 million French Francs are excluded.
Year Population Total Entry Start-up Rate
1990 474,118 100,596 21.22%
1991 477,666 66,814 13.99%
1992 505,580 77,098 15.25%
1993 488,757 68,424 14.00%
1994 516,730 90,544 17.52%
1995 505,871 58,727 11.61%
1996 493,432 50,560 10.25%
115
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
OECD Firm-level Data SIRENE Creations SIRENE Pure Creations
The SIRENE data clearly show the impact of correcting for purity when the other comparability
factors are held constant. The pure creation rate is consistently almost 4% lower than the total
creation rate. As with other countries there is considerable variability in the start-up rate from
the OECD firm-level data, and for the one year of overlap, 1993, the rate is higher than that
for the SIRENE data. This could be a result of the high turnover threshold used for French
data in this source.
b) New Businesses
0
50
100
150
200
250
300
350
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
OECD Firm-level Data SIRENE Creations SIRENE Pure Creations
APCE Creations APCE Creations "ex nihilo"
For this chart, it is possible to add the APCE Creations data. As can be seen, the APCE data
very closely follow those from SIRENE, such that the APCE creations “ex nihilo” seems to be
116
a good indicator for the missing variables in the SIRENE pure creations series. For the four
years of overlap, the trend followed by the OECD firm-level data seems reasonably correlated
to those of the other sources, with the difference in levels clearly attributable to the threshold
in the OECD data.
c) Business Populations
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data SIRENE
This chart is less interesting given the lack of overlap. Both sources follow a stable to slightly
rising trend, and the clear impact of the threshold in the OECD data is again visible.
117
5. Germany
Three sources of data on business start-ups have been identified for Germany
a) Business Notifications
Federal Statistics Office, Germany:http://www.destatis.de/themen/e/thm_unternehmen.htm
This source contains data on business registrations, modifications and de-registrations, as
well as counts of the stock of businesses liable to pay turnover tax. Registration data are
available for 2001 to 2003. Data on the population of businesses liable for tax are available for
2002 and 2003. The data are broken down by economic activity.
Some descriptive metadata are available via the website. The data on new registrations are
assumed to cover the whole economy. The population of businesses liable to pay turnover tax
covers businesses with a turnover of at least €16,620 per year. Most economic activities are
covered, with the exception of certain health, public administration, insurance, and agricultural
activities.
The unit for registrations and de-registrations is effectively the local unit as “the obligation to
report business registrations and de-registrations applies to enterprises, branch offices and
dependent sub-offices”. Registration is required when a new activity is started or a business is
taken over, be it through purchase or succession, a partner entering the business, a change in
legal form, or a relocation of the business to a different registration district. Thus quite a high
proportion of registrations will not be pure births.
Year Population Business Registrations Birth Rate
2001 728,978
2002 2,926,570 723,333 24.72%
2003 2,915,482 810,706 27.81%
b) Start-ups and Liquidations in Germany
Institut für Mittelstandsforschung (IfM), Bonn:http://www.ifm-bonn.org/dienste/gruendungen-
engl.htm
This source contains counts of business start-ups and liquidations based on notifications of
new businesses. The data are adjusted to remove new sites of existing businesses,
registrations purely for tax or administrative purposes that do not result in new business
activity, and registrations for activities carried out as a second job by the entrepreneur.
Enterprise population totals for some years are available in Table 1 of:http://www.ifm-
bonn.org/dienste/kap-2.pdf. These do not provide a full series, but IfM have been able to
provide additional data (1992, 2000 – 2003) or suggest appropriate approximations (1991,
1993 and 1995) to fill the gaps.
The start-up data are broken down into the former East and West Germany, and are available
for 1991 to 2004. The population of businesses is subject to a threshold (€17,500 since 2003),
and covers all economic activities except the “liberal professions”
34
, most health services and
some insurance services that are not subject to VAT. The units used are effectively the sub-
set of legal unit that are considered to be economically relevant.
34
The liberal professions can generally be defined as occupations requiring special training in the arts
or sciences. These include lawyers, notaries, accountants, architects, engineers and pharmacists.
118
Year Population Start-ups Birth Rate
1991 2,572,202 531,000 20.64%
1992 2,631,812 494,000 18.77%
1993 2,709,443 486,000 17.94%
1994 2,787,074 493,000 17.69%
1995 2,775,000 528,000 19.03%
1996 2,762,925 507,000 18.35%
1997 2,797,759 507,100 18.13%
1998 2,859,983 512,800 17.93%
1999 2,886,268 493,100 17.08%
2000 2,909,150 471,700 16.21%
2001 2,920,293 454,700 15.57%
2002 2,926,570 451,800 15.44%
2003 2,915,482 507,900 15.54%
2004 572,600
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data, which
cover only the former West Germany, are available for 1978 to 1998, though data on entries
for 1978 are clearly incomplete.
Annual point in time populations were used, taken from a social security database. The unit
used is referred to as the “plant”, thus is likely to be closer to the definition of the local unit
than to that of the enterprise. Only businesses with one or more employees are included, and
certain public sector units are considered out of scope. New businesses had to be present in
both the reference year and the following year to be counted as a birth in published analyses
of the data. “One-year” businesses were identified separately, but have been included within
“Total Entry” in the table below to try to improve comparability with other sources.
Year Population Total Entry Start-up Rate
1978 1,320,297 2,292 0.17%
1979 1,339,660 160,550 11.98%
1980 1,369,687 161,363 11.78%
1981 1,384,396 159,594 11.53%
1982 1,399,054 162,219 11.59%
1983 1,405,631 165,856 11.80%
1984 1,418,422 153,286 10.81%
1985 1,436,305 168,097 11.70%
1986 1,448,569 166,603 11.50%
1987 1,449,059 169,702 11.71%
1988 1,470,122 177,681 12.09%
1989 1,503,586 172,989 11.51%
1990 1,503,757 162,153 10.78%
1991 1,539,597 183,197 11.90%
1992 1,572,557 183,180 11.65%
1993 1,589,724 180,054 11.33%
119
1994 1,602,366 174,098 10.87%
1995 1,618,343 175,963 10.87%
1996 1,626,563 172,455 10.60%
1997 1,632,956 175,693 10.76%
1998 1,638,470 182,290 11.13%
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
OECD Firm-level Data Start-ups and Liquidations in Germany Business Notifications
There is clearly a problem with the OECD Firm-level Data for 1978, otherwise this series
seems quite stable over time. The IfM Start-ups and Liquidations data show consistently
higher start-up rates. This is to be expected as they include non-employer businesses (subject
to a turnover threshold), which generally have higher entry and exit rates than employer
businesses. It is also possible that the coverage of the former East Germany in the IfM data
may also contribute to higher start-up rates, as it has been observed in the Eurostat business
demography project that start-up rates are slightly higher in the former communist countries of
Eastern and Central Europe than they are in Western Europe.
The Business Notifications data have even higher rates. The reasons for this are likely to be
due to the units used (local units rather than enterprises), and that many of the apparent start-
ups are really re-registrations of existing business activities.
120
b) New Businesses
0
100
200
300
400
500
600
700
800
900
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Start-ups and Liquidations in Germany Business Notifications
The exclusion of non-employer businesses, and businesses in the former East Germany, is
clearly apparent in the OECD Firm-level data in this chart. The IfM data show a little more
variability over time. This could be due to non-employer business patterns showing a greater
responsiveness to the economic cycle, though there is no hard evidence for this theory. Again
the Business Notifications data show the highest levels reflecting the units used and the
inclusion of many re-registrations.
c) Business Populations
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data Start-ups and Liquidations in Germany Business Notifications
The OECD Firm-level and IfM data sets show a very consistent and slowly rising trend, though
the level of the former is low due to the coverage and threshold limitations. Unfortunately the
shortness of the Business Notifications series limits the conclusions that can be drawn about
this source, though the population estimates are the same as those provided by IfM for the
two years available.
121
6. Hungary
Three sources of data on business start-ups have been identified for Hungary.
a) Enterprises and Non-profit Organisations
Hungarian Central Statistical Office:http://portal.ksh.hu/portal/page?_pageid=38,341368&_dad=portal&_schema=PORTAL
This source contains annual point in time (end year) counts of registered economic
corporations and unincorporated enterprises, as well as quarterly counts of new registrations.
Both are broken down by legal form, the population data are also broken down by economic
activity. Data on new registrations are available from 2001 to 2004.
The data cover all businesses that hold an active registration and tax number in the
administrative register, including most government bodies. There is no registration threshold
in Hungary, so part-time businesses are included. Approximately 75% of registrations are
considered to be economically active by the Hungarian statistical office. All economic activities
are covered, though NACE division L (public administration) is excluded from the counts
broken down by activity. The unit is referred to as the enterprise, but the definition is currently
closer to that of a legal unit.
Year Population New Enterprises Birth Rate
2000 1,175,480
2001 1,207,831 125,233 10.37%
2002 1,236,890 115,878 9.37%
2003 1,263,990 106,471 8.42%
2004 1,286,993 103,271 8.02%
b) Demography of Small and Medium-sized Enterprises (DOSME) Study
http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc. For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
This source contains data on enterprise populations, births, deaths, survival and factors of
success. The data are broken down by economic activity, size, legal form and characteristics
of the entrepreneur. The final report describes the methodology used to produce the data it
contains. Other descriptive metadata is available on the project web site. The key feature of
this source is that data are based on surveys of businesses rather than directly on the
business register. This will introduce certain survey errors in addition to other methodological
differences.
Data on births are available for Hungary from 1994 to 2001. Known problems with the
observation of business closures in this project led to the construction of trend adjusted
closure data, which has resulted in a certain smoothing of the population data, as shown in
the chart below.
122
The impact on the population of enterprises of using trend-adjusted deaths data
350
400
450
500
550
1994 1995 1996 1997 1998 1999 2000
T
h
o
u
s
a
n
d
s
Raw Population Population Based on Trend-adjusted Deaths
The data below cover NACE sections C to K, to improve comparability with the Eurostat data,
though data for other economic activities are also available.
Year Population Births Birth Rate
1994 456,376 96,654 21.18%
1995 486,975 87,193 17.91%
1996 495,722 63,805 12.87%
1997 488,747 46,993 9.61%
1998 497,401 61,777 12.42%
1999 504,922 59,953 11.87%
2000 520,505 67,432 12.96%
c) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Hungarian data are available for 2000 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Hungarian statistical business
register with economic activities in NACE sections C to K (production, construction, trade and
most services), except class 74.15, management activities of holding companies. All legal
forms are covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
Year Population Births Birth Rate
2000 526,553 71,395 13.56%
2001 542,288 68,963 12.72%
2002 576,609 83,817 14.54%
123
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
DOSME Eurostat Business Demography Central Statistical Office (KSH)
The data on birth rates seem to show an overall downward trend from the immediate post-
communist period in the mid-1990’s. This is fairly typical of other European countries making
the transition to market economies at this time. The trough in 1997 in the DOSME data may
well be genuine, but there is a risk that it may also be partially due to survey errors, or over-
smoothing of the population for inconsistencies in closure data. It is interesting to note,
however, that the Eurostat data series appears to take over where the DOSME data ended.
The divergence between the Eurostat and KSH data between 2001 and 2002 looks
suspicious, but could be due to differences in coverage and purity.
b) New Businesses
0
20
40
60
80
100
120
140
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
DOSME Eurostat Business Demography Central Statistical Office (KSH)
124
Again the Eurostat data seem to take over where the DOSME data end in 2000. The main
difference in this chart, however, is that the KSH data are rather higher than those from the
other sources. This is to be expected, however, as the KSH data have a wider coverage of
legal forms and economic activities, as well as a lower degree of purity than the Eurostat data.
c) Business Populations
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
DOSME Eurostat Business Demography Central Statistical Office (KSH)
Here, the apparent continuation from the DOSME to the Eurostat data is most striking,
particularly given that the DOSME populations are on a point in time basis, whereas the
Eurostat ones are live during period, and could be expected to be around 10% higher. One
clue might be in the fact that the KSH populations are so much higher. This will be partly due
to the wider coverage noted above, but also to the inclusion of a relatively large proportion of
inactive units (possibly up to 25%). This may also be a problem, on a lesser scale, in the
DOSME data.
125
7. Italy
Three sources of data on business start-ups have been identified for Italy
a) Movimprese
InfoCamere:http://www.infocamere.it/movi_search.htm
This source contains counts of total registrations, active registrations, new registrations,
cessations and changed registrations at the Italian chamber of commerce. The data are
broken down by economic activity and geography, and are available for 1995 to 2004.
A glossary and other metadata are available on the web site (in Italian). The data do not cover
NACE section L (public administration), and presumably do not cover government units. The
unit used is the legal unit. The population data are point in time, and appear to relate to the
end of the year, so have been carried over as start-year populations for the following year inn
the table below.
Year
Active Registrations
at 1 J anuary
New Registrations Birth Rate
1995 350,498
1996 3,578,931 505,354 14.12%
1997 3,806,838 1,260,364 33.11%
1998 4,704,107 408,475 8.68%
1999 4,727,504 390,074 8.25%
2000 4,774,264 403,408 8.45%
2001 4,840,366 421,451 8.71%
2002 4,897,933 417,204 8.52%
2003 4,952,053 389,342 7.86%
2004 4,995,738 425,510 8.52%
2005 5,061,859
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Italian data are available for 1998 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Italian statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
126
Year Population Births Birth Rate
1998 3,596,450 409,272 11.38%
1999 3,677,890 278,104 7.56%
2000 3,760,098 291,856 7.76%
2001 3,833,049 294,866 7.69%
2002 3,853,598 283,463 7.36%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for Italy
are available for 1987 to 1993.
Annual point in time populations were used, taken from a social security database. Only
businesses with one or more employees are included. New businesses had to be present in
both the reference year and the following year to be counted as a birth in published analyses
of the data. “One-year” businesses were identified separately, but have been included as
births in the table below to try to improve comparability with other sources.
Year Population Total Entries Start-up Rate
1987 1,115,036 118,676 10.64%
1988 1,150,278 123,394 10.73%
1989 1,177,162 141,112 11.99%
1990 1,191,290 116,359 9.77%
1991 1,191,651 105,252 8.83%
1992 1,195,573 105,222 8.80%
1993 1,151,733 92,444 8.03%
127
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
35%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
OECD Firm-level Data InfoCamere Eurostat Business Demography
There is clearly a problem for the InfoCamere data for 1997, and possibly the Eurostat data
for 1998. The most likely cause is a large increase in the scope of the source. From 1999
onwards these two sources seem much more stable, producing closely comparable rates,
though with possibly slightly diverging trends.
b) New Businesses
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data InfoCamere Eurostat Business Demography
Again the InfoCamere and Eurostat data seem to follow similar, if slightly diverging trends
after 1999. The difference in levels will be mainly due to a mixture of coverage and purity,
though the difference in units (legal unit for InfoCamere, enterprise for Eurostat) may also play
128
a small part. The OECD firm-level data show a lower level again, which is due to the exclusion
of businesses without employees in this source.
c) Business Populations
0
1
2
3
4
5
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data InfoCamere Eurostat Business Demography
The higher levels in the InfoCamere population data will be due mainly to wider coverage,
and, to a much lesser extent, to the difference in units when compared to the Eurostat data,
as noted above. The difference between these sources would be even greater if the Eurostat
data were on a point in time basis like those from InfoCamere. The rapid growth in the
InfoCamere population between 1997 and 1998 lends weight to the theory that there was a
significant increase in the coverage of this source around that time, as suggested above. The
OECD firm-level data are relatively low again due to the exclusion of non-employer
businesses.
129
8. The Netherlands
Three sources of data on business start-ups have been identified for the Netherlands.
a) Establishment and Closure of Businesses
Statistics Netherlands:http://statline.cbs.nl/StatWeb/table.asp?PA=07223eng&D1=a&D2=0&D3=(l-11)-
l&DM=SLEN&LA=en&TT=2
This source contains counts (and employment) of the stock of businesses (as at 1 J anuary),
businesses opening, and businesses closing. The data are available broken down by
economic activity, and are available for 1993 to 2002. The data exclude certain NACE
categories (Sections A, B, E, L, M and N, and divisions 70, 73, 91 and 92). On this basis it is
assumed that most government activity is also excluded.
Metadata are available by clicking on the table headings on the web site. The unit used is the
“business” which is assumed to be close to the enterprise, as the terms are both used in the
metadata. Establishment of a business is the formation of a new enterprise. This implies that
the statistical criteria for enterprises (autonomy and external orientation) have to be met.
Moreover, the enterprise has to be economically active, i.e. at least one person works in the
enterprise for at least 15 hours a week. The enterprise has to be a new one, i.e. not the
continuation of one or more existing enterprises.
Year Population New Businesses Birth Rate
1993 376,300 29,000 7.71%
1994 382,080 26,000 6.80%
1995 386,360 33,000 8.54%
1996 406,585 34,000 8.36%
1997 425,780 31,000 7.28%
1998 452,450 30,000 6.63%
1999 464,620 31,000 6.67%
2000 473,095 39,000 8.24%
2001 482,295 40,000 8.29%
2002 486,575 38,000 7.81%
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Dutch data are available for 1998 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Dutch statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
130
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
Year Population Births Birth Rate
1999 523,243 49,999 9.56%
2000 534,339 50,475 9.45%
2001 541,538 52,053 9.61%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for the
Netherlands are available for 1987 to 1997.
Annual point in time populations were used, taken from the general business register. Only
businesses with one or more employees are included. New businesses had to be present in
both the reference year and the following year to be counted as a birth in published analyses
of the data. “One-year” businesses were identified separately, but have been included as
births in the table below to try to improve comparability with other sources.
Year Population Total Entry Start-up Rate
1987 589,220 48,250 8.19%
1988 640,787 74,612 11.64%
1989 689,359 56,486 8.19%
1990 732,253 82,990 11.33%
1991 759,372 78,554 10.34%
1992 799,563 89,530 11.20%
1993 852,417 107,235 12.58%
1994 915,444 102,139 11.16%
1995 944,909 98,678 10.44%
1996 909,841 101,896 11.20%
1997 932,260 107,013 11.48%
131
Graphical Comparisons
a) Birth Rates
0%
2%
4%
6%
8%
10%
12%
14%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
OECD Firm-level Data Establishment and Closure of Businesses Eurostat Business Demography
The OECD firm-level data show rather a lot of variation, particularly in the earlier years, but
tend to stabilise at around 11% towards the end of the series. This rate is slightly higher than
that from the Eurostat data, due to the interaction of greater purity and the use of a live during
period population reducing the Eurostat
rates, whilst the threshold of one employee
would be expected to reduce the firm-level
rates. The data from the Establishment and
Closure of Business source seem to exhibit
a cyclical pattern, which interestingly
shows a strong negative correlation to the
real GDP growth rate data for the
Netherlands (sourced from Eurostat). This
may be coincidence, but could be worth
further investigation.
b) New Businesses
0
20
40
60
80
100
120
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Establishment and Closure of Businesses Eurostat Business Demography
0%
2%
4%
6%
8%
10%
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
GDP Growth Start-up Rate
132
The number of new businesses in the OECD firm-level data set is surprisingly high compared
to the other sources, particularly as the metadata state that only businesses with employees
are included. The coverage in terms of economic activity of this source is higher, but it is likely
that purity and the unit of observation are more important factor in explaining the difference.
The other two sources seem more comparable, with the differences probably due to the more
restricted coverage and the threshold of 15 hours labour input per week in the Establishment
and Closure of Businesses source. The apparent cyclical pattern in the data from this source
is once again evident in this chart.
c) Business Populations
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
M
i
l
l
i
o
n
s
OECD Firm-level Data Establishment and Closure of Businesses Eurostat Business Demography
As for new businesses, the population data from the OECD firm-level project is rather high,
again casting doubt on the statement in the metadata that non-employer businesses are
excluded. Units and coverage, particularly the inclusion of non-active units, are likely to
account for a significant part of the difference. The Eurostat population at least 10% higher
than that from the Establishment and Closure of Businesses source due mainly to it being on
a live during period basis. Relatively small differences in coverage and threshold will also play
a minor role.
133
9. United Kingdom
Four sources of data on business start-ups have been identified for the United Kingdom.
a) Value-Added Tax Registrations and De-registrations
UK Department for Trade and Industry – Small Business Servicehttp://www.sbs.gov.uk/sbsgov/action/layer?r.l2=7000000243&r.l1=7000000229&r.s=tl&topicId
=7000011757
This source contains counts of the stock of businesses registered for value-added tax (VAT),
as well as new registrations and de-registrations. The data are broken down by economic
activity and geography, and are available on a calendar year basis for 1994 to 2003. A
previous series from 1980 to 1993 is also available, but the data are not directly comparable
due to large changes in the VAT threshold.
A paper on the methodology used is available via the web site. The data cover all economic
activities and legal forms, though coverage is limited for certain activities that are exempt from
VAT, particularly in the education and health sectors. The stock data are on a point in time
basis (1 J anuary). The unit used is the VAT registration, which approximates to the legal unit.
The data are sourced from the UK statistical business register, so a proportion of registrations
are statistical rather than purely administrative, particularly for larger businesses.
Year Stock at 1/1 Registrations Birth Rate
1994 1,629,120 169,210 10.39%
1995 1,623,575 164,910 10.16%
1996 1,627,905 169,590 10.42%
1997 1,645,950 185,950 11.30%
1998 1,683,675 184,770 10.97%
1999 1,719,330 178,450 10.38%
2000 1,744,380 179,585 10.30%
2001 1,767,530 168,445 9.53%
2002 1,783,135 175,700 9.85%
2003 1,794,920 189,890 10.58%
2004 1,810,460
b) Barclays Small Business Surveys
http://www.business.barclays.co.uk/BRC1/jsp/brccontrol?task=articleFWgroup&value=6502&t
arget=_self&site=bbb
This commercial source contains data on business start-ups and closures, as well as the total
population of businesses in a series of quarterly reports. The data are broken down in
different ways each year, including by geography, economic activity, and sex of the
entrepreneur, and are available from 1995 to 2004, though for latter years they are
increasingly broadly rounded estimates.
Limited metadata are available within the reports. The data only cover England and Wales,
and are based on business current account openings and closures at Barclays, multiplied by
estimates of their share of the business banking market. This makes it unlikely that central
and local government activities will be included, and the extent of coverage of non-profit
organisations is unclear. Businesses that do not operate via business current accounts are
134
also excluded by definition. The population data are on a point in time basis. The unit used is
the business account, which is expected to be fairly close to the definition of the enterprise,
particularly for new businesses.
Year Stock at 1/1 Births Birth Rate
1995 2,656,570 471,406 17.74%
1996 2,680,924 477,630 17.82%
1997 2,621,702 476,690 18.18%
1998 2,655,889 454,628 17.12%
1999 2,721,198 438,727 16.12%
2000 2,773,646 438,745 15.82%
2001 2,770,000 342,000 12.35%
2002 2,720,000 373,500 13.73%
2003 2,687,000 446,300 16.61%
2004 2,800,000 452,800 16.17%
c) Eurostat business demography indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form. UK
data for the population of active enterprises are available for 1997 to 2003, and data on births
cover 1998 to 2003.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the UK statistical business register with
economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. Separate data for
some countries (including the UK) are also available for NACE sections M, N and O (health,
education, community, social and personal services). These have been added in to the table
below to improve coverage. This means that agriculture, forestry, fishing, public
administration, activities of households, and extra-territorial organizations and bodies remain
excluded. All legal forms are covered except central and local government, and non-profit
organisations serving households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
Year
Population of
Active Enterprises Births Birth Rate
1997 1,898,810
1998 1,958,750 256,285 13.08%
1999 2,016,395 257,840 12.79%
2000 2,041,685 242,595 11.88%
2001 2,084,540 244,105 11.71%
2002 2,115,325 242,945 11.48%
2003 2,183,125 281,460 12.89%
135
d) OECD Firm-level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for most
countries were drawn from either statistical or administrative business registers, usually at the
level of the enterprise or firm. The UK data, however, were taken from a series of frames for
an annual survey of production businesses. The units used (“reporting units”) were designed
for data collection purposes, and tended to change as business structures evolved, making
them less stable over time than enterprises, and the coverage was determined by survey
requirements, which varied over time, so was rather less comprehensive than that of the
business register, particularly for smaller businesses.
Annual point in time populations were used, based on survey frames usually drawn around
October of each year (variations in the frame date may cause some minor comparability
issues). New businesses had to be present in both the reference year and the following year
to be counted as a birth in published analyses of the data. “One-year” businesses were
identified separately, but have been included as births in the table below to try to improve
comparability with other sources.
Year Stock Entries During Year Entry Rate
1986 148741 23872 16.05%
1987 150778 24114 15.99%
1988 154956 27315 17.63%
1989 158131 38646 24.44%
1990 151945 20775 13.67%
1991 147984 14952 10.10%
1992 x x x
1993 148057 50897 34.38%
1994 157975 31526 19.96%
1995 174825 47639 27.25%
1996 166981 21316 12.77%
1997 169826 21218 12.49%
Note: Data for 1992 are missing, but it looks likely that births for this year are included in 1993 figures.
136
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
35%
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
VAT Registrations Barclays Eurostat OECD Firm-level Data
The OECD Firm-level data show considerably more variability than those from the other
sources. This is likely to be due to changes in coverage between years, as well as the
instability over time of the unit used (the reporting unit). The break in 1992 is likely to be linked
to the introduction of a new statistical business register around this time, which led to a new
definition and numbering system for reporting units. The data only really seem to show
plausible rates at the start and the end of the period covered.
The Barclays data also show more variability than the remaining two sources. This could
reflect the greater coverage of very small businesses, which are known to be more volatile
than their larger counterparts. The pronounced trough in 2001 could be at least partly due to
the “foot and mouth disease” epidemic amongst farm animals in the UK in that year. This had
a particularly strong effect on rural businesses.
VAT Registration data show a much more stable trend, but interestingly do seem to follow a
similar pattern to the Barclays data, looking rather like a smoothed version, albeit at a lower
absolute level. This could be taken as a positive indication of the quality of the two data sets.
Start-up rates for the Eurostat data again follow similar trends, with the increased coverage of
small businesses more than compensating for the use of live during period population figures
when compared to the VAT Registrations series.
137
b) New Businesses
0
100
200
300
400
500
600
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
VAT Registrations Barclays Eurostat OECD Firm-level Data
The OECD Firm-level data show much lower levels because they only include manufacturers,
and exclude many smaller businesses. The other sources have a much more complete
coverage of economic activities. Both the Barclays, and to a lesser extent, the Eurostat
sources, have a higher coverage of small businesses than the VAT registrations data, due to
the high VAT registration threshold in the UK. As is the case for birth rates, the Barclays,
Eurostat and VAT registrations data show similar patterns, though the fluctuations in the
Barclays data are much more exaggerated.
c) Business Populations
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
VAT Registrations Barclays Eurostat OECD Firm-level Data
Barclays data show the highest population, despite only covering England and Wales. The
data from this source would be around 11 - 12% higher if they covered all of the UK (i.e.
including Scotland and Northern Ireland). The higher population is due to a much greater
coverage of very small businesses (low turnover, no employees) than the other sources, and
no restrictions in terms of economic activities.
VAT Registrations data are lower than those from Eurostat, despite originating from the same
business register, due to the interaction of three coverage related issues, and one basic
138
difference in methodology. VAT registrations data have more comprehensive coverage in
terms of economic activities (particularly agriculture) and legal forms (non-profit institutions),
but this is more than compensated for by a lower coverage of very small businesses, and the
fact that the Eurostat population data are on a “live during period” basis, whereas the VAT
registrations population includes only those registrations live on a specific date (1 J anuary).
139
10. United States
Five sources of data on business start-ups have been identified for the United States.
a) Statistics of US Businesses / Dynamic Data
US Census Bureau -http://www.census.gov/csd/susb/susbdyn.htm
This source contains counts of the stock of establishments, births, deaths, expansions and
contractions, and associated employment changes. The data are broken down by economic
activity, size band (based on employment) and state, and are available for 1995 to 2001.
Papers with descriptive metadata and definitions are available via the web site. Businesses
without employees are excluded. All economic activities are covered except crop and animal
production (NAICS 111,112), rail transportation (NAICS 482), National Postal Service (NAICS
491), pension, health, welfare, and vacation funds (NAICS 525110, 525120, 525190), trusts,
estates, and agency accounts (NAICS 525920), private households (NAICS 814), and public
administration (NAICS 92). Governmental establishments are excluded except for wholesale
liquor establishments (NAICS 4228), retail liquor stores (NAICS 44531), Federally-chartered
savings institutions (NAICS 522120), Federally-chartered credit unions (NAICS 522130), and
hospitals (NAICS 622).
The stock data are on a point in time basis (businesses with employees in the first quarter),
though counts of establishments that had employees in any quarter of the year are also
available. The unit used is the establishment, which is defined as “a single physical location
where business is conducted or where services or industrial operations are performed.” This is
broadly equivalent to the European definition of the local unit. Establishment births are defined
as establishments that have zero employment in the first quarter of the initial year and positive
employment in the first quarter of the subsequent year. Establishment deaths are
establishments that have positive employment in the first quarter of the initial year and zero
employment in the first quarter of the subsequent year. The definitions of births and deaths
are thus quite broad, and correspond to all recorded creations and closures respectively.
Year Population Establishment Births Birth Rate
1995 5,878,957 697,457 11.86%
1996 5,970,420 822,582 13.78%
1997 6,120,714 719,616 11.76%
1998 6,187,599 713,002 11.52%
1999 6,248,411 709,079 11.35%
2000 6,297,423 727,320 11.55%
2001 6,345,890 787,309 12.41%
b) Firm Size Data
US Small Business Administration –http://www.sba.gov/advo/research/data.html
This source contains counts of the population of firms, births and deaths. Employment data
are also available. The data on the population of firms are broken down by size band
(employment) and economic activity. There are no breakdowns of the data on firm births and
deaths. Data on the population or firms are available for 1988 to 2002. Data on births and
deaths are available for 1989 to 2001.
140
Extensive metadata are available in the paper “Statistics of U.S. Businesses – Microdata and
Tables”, available on the website. The coverage is basically the same as source 1 above, as
the firm level data are derived from the US Census Bureau establishment statistics. The
population counts cover all businesses that had an active payroll at any point during the year,
so can be considered as “live during period” data. The unit used is the firm, which is defined
as “the largest aggregation of business legal entities under common ownership or control”, so
corresponds most closely to the European definition of the Enterprise Group (truncated or all-
residential rather than global). Firm birth and death definitions correspond to those for
establishments in source 1 above.
Year Employer Firms Firm Births Birth Rate
1988 4,954,645
1989 5,021,315 584,892 11.65%
1990 5,073,795 541,141 10.67%
1991 5,051,025 544,596 10.78%
1992 5,095,356 564,504 11.08%
1993 5,193,642 570,587 10.99%
1994 5,276,964 594,369 11.26%
1995 5,369,068 597,792 11.13%
1996 5,478,047 590,644 10.78%
1997 5,541,918 589,982 10.65%
1998 5,579,177 579,609 10.39%
1999 5,607,743 574,300 10.24%
2000 5,652,544 585,140 10.35%
2001 5,657,774 569,750 10.07%
2002 5,697,759
Note: The population data have been taken from a different table to the data on births and deaths. The
assumption (in the absence of any evidence to the contrary) is that they are on a comparable basis.
c) Business Employment Dynamics
Bureau of Labor Statistics –http://www.bls.gov/bdm/home.htm, and Pinkston and Spletzer
(2004) -http://www.bls.gov/opub/mlr/2004/11/art1full.pdf
This source contains counts and rates for establishment openings and closures each quarter.
The data can be broken down by economic activity, and are currently available from quarter 3
of 1992 to quarter 4 of 2004 inclusive.
Descriptive metadata and definitions are available via the web site. The data exclude
business with no employees, central and local government units, and some non-profit
organizations. Certain economic activities are also excluded (religious organizations, some
small farms, the Armed Forces and railways). The unit used is the establishment, which is
broadly equivalent to the European definition of the local unit. Openings are either
establishments with positive third month employment for the first time in the current quarter,
with no links to the prior quarter, or with positive third month employment in the current quarter
following zero employment in the previous quarter. Closings are either establishments with
positive third month employment in the previous quarter, with no positive employment
reported in the current quarter, or with positive third month employment in the previous
quarter followed by zero employment in the current quarter.
141
No stock data are given, but they can be estimated from openings counts and rates (or
closures counts and rates) on a quarterly basis. These can then be used to calculate annual
birth and death rates. The data on openings and closures give slightly different stock figures.
This is due to rounding of the counts of openings and closures (to the nearest thousand), and
the rates (to one decimal place). The derived stock figures based on openings and closures
for each year are within the margins of error associated this level of rounding. The impact on
the annual opening and closure rate estimates is less than 0.2%.
The paper by Pinkston and Spletzer explores the impact of short-lived businesses on the
data, and gives annualised data for 1998 to 2001. Their method removes very short-lived
businesses, and false start-ups due to businesses that have previously been in the population
of employers, but were temporarily absent. The effect on start-up rates is dramatic.
Year Population Births Birth Rate Annualised Births Annualised Birth Rate
1993 5,419,807 1,171,000 21.61%
1994 5,544,268 1,223,000 22.06%
1995 5,738,196 1,242,000 21.64%
1996 5,828,816 1,306,000 22.41%
1997 5,902,142 1,326,000 22.47%
1998 6,045,896 1,344,000 22.23% 778,826 12.99%
1999 6,096,898 1,409,000 23.11% 804,022 13.19%
2000 6,200,692 1,405,000 22.66% 809,301 13.09%
2001 6,268,227 1,363,000 21.74% 790,237 12.67%
2002 6,344,799 1,374,000 21.66%
2003 6,378,568 1,355,000 21.24%
2004 6,535,698
Note: The population is calculated as the median value of the intersection between the ranges of
possible values based on the births and deaths data for the second quarter of each year (i.e. it is the
estimated population as at 1 April). Births and deaths are for the period 1 April year t to 31 March year
t+1. 1 April is used as a reference date in an attempt to improve comparability with source 1, which is
calculated on a March to March basis.
d) Longitudinal Business Database
US Census Bureau -http://www.ces.census.gov/ces.php/abstract?paper=101647
This database contains linked records of establishments and firms over time. It can be used to
produce data on business dynamics. The internet address above is that of a paper describing
the database, which includes data on births and deaths. A second paper is available with
more detailed analyses for the retail sector – see:http://www.ces.census.gov/ces.php/abstract?paper=101704
The data in the paper are not broken down in any way, but the database would allow a range
of detailed breakdowns. The paper presents stock, birth and death data for 1976 to 1999.
The paper contains descriptive metadata. The source data cover establishments with paid
employees. Economic activity coverage is the same as for source 1 above. The stock data are
on a point in time basis, and the unit used is the establishment. Births are records that were
active in one year, but not the previous year, adjusted for reactivations. Deaths are records
that were active in one year, but not the next, adjusted for reactivations.
Year Population Births Birth Rate
1976 4,945,528 824,563 16.67%
142
1977 5,125,942 844,422 16.47%
1978 5,152,243 683,598 13.27%
1979 5,330,266 681,813 12.79%
1980 5,283,897 610,991 11.56%
1981 5,244,139 649,292 12.38%
1982 5,294,765 702,036 13.26%
1983 5,586,606 755,528 13.52%
1984 5,833,945 779,039 13.35%
1985 5,981,692 771,830 12.90%
1986 6,098,536 763,103 12.51%
1987 6,174,220 851,033 13.78%
1988 6,228,218 717,030 11.51%
1989 6,388,877 797,117 12.48%
1990 6,645,560 933,622 14.05%
1991 6,729,082 799,454 11.88%
1992 6,759,906 787,850 11.65%
1993 6,860,000 746,635 10.88%
1994 6,973,457 760,594 10.91%
1995 7,077,456 754,795 10.66%
1996 7,167,943 766,265 10.69%
1997 7,305,127 894,978 12.25%
1998 7,351,196 754,708 10.27%
1999 7,405,245 828,164 11.18%
Note: The population data are those establishments considered to be active in the longitudinal
database. Births and deaths have been adjusted to remove all reactivations.
e) OECD Firm-level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for the
United States are available for 1989 to 1996.
Annual point in time populations were used, taken from the prototype longitudinal business
database. Only businesses with one or more employees are included. New businesses had to
be present in both the reference year and the following year to be counted as a birth in
published analyses of the data. “One-year” businesses were identified separately, but have
been included as births in the table below to try to improve comparability with other sources.
Year Population Total Entry Start-up Rate
1989 4648625 528711 11.37%
1990 4798181 526578 10.97%
1991 4867411 503077 10.34%
1992 4981011 550934 11.06%
1993 5051860 505943 10.01%
1994 5137618 515718 10.04%
1995 5224433 519906 9.95%
1996 5311984 530919 9.99%
143
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
Longitudinal Business Database Statistics of US Businesses
Firm Size Data Business Employment Dynamics - Summed Quarterly Data
OECD Firm-level Data Business Employment Dynamics - Annualised Data
The Business Employment Dynamics quarterly data set is a clear outlier in terms of birth
rates. The annualised data set clearly show that this is almost entirely due to periodicity and
data purity. The remaining data sources appear to give fairly comparable measures of start-up
rates, typically between 10% and 13%.
Based on the table under the population chart below, if the population of firms active in March
was used as the denominator for the Firm Size data set, this would have the effect of
increasing the birth rate for this source by around 1.5%, taking it to a similar, or very slightly
higher level than that for Statistics of US Businesses.
This suggests the interesting conclusion that where data are otherwise comparable, the
choice of firm or establishment as the unit of observation makes little difference to business
start-up rates. The increased volatility usually associated with establishments is, in this case,
cancelled out by the higher population of establishments, i.e. both the numerator and the
denominator are higher for establishments, but the rate is almost identical. Taking this a step
further, this suggests that, for international comparison purposes, differences in units may not
be a major obstacle. Unfortunately these conclusions are only based on data for one country
for a limited period of time, so it remains to be seen how safe they are in a wider context.
144
b) New Businesses
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Longitudinal Business Database Statistics of US Businesses
Firm Size Data Business Employment Dynamics - Summed Quarterly Data
OECD Firm-level Data Business Employment Dynamics - Annualised Data
For births, the clear outlier is again the Business Employment Dynamics quarterly data. This is
due to issues of purity (i.e. including virtually all establishment creations), and a much greater
chance of including short-lived businesses.
The number of births from the Longitudinal Business Database seems close to those from
Statistics of US Businesses, but this masks two differences between the data that seem to
largely cancel each other out. All things being equal, the number of births from the
Longitudinal Business Database should be higher, simply because the population for that
source is more comprehensive, however, this seems to be balanced by the greater extent of
data matching within this source, to remove all reactivations and other “false” births.
The spike in the Longitudinal Business Database series in 1997 is acknowledged as
suspicious, and could perhaps be linked to the very similar spike in the data from Statistics of
US Businesses the year before. If this is the case, it could indicate the presence of a lag
between these two sources.
The gap between the Statistics of US Businesses series and that from the Firm Size Data
source indicates the proportion of new establishments created by existing firms, assuming
that very few new firms have more than one site when they are created. It is interesting to
note that the spike in the Statistics of US Businesses data for 1996 is not present in the Firm
Size Data series. This might indicate that it was more of a source processing issue than a real
increase. Similarly the increasing divergence between these sources for 2001, combined with
the fact that Statistics of US Businesses is the only source to show an increase between 2000
and 2001, might suggest similar processing issues.
The data for the OECD Firm-level series are taken from a prototype of the Longitudinal
Business Database, which explains the similar trends, however there is clearly a difference in
coverage.
145
c) Business Populations
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Business Employment Dynamics Statistics of US Businesses OECD Firm-level Data
Longitudinal Business Database Firm Size Data
All five sources show a similar trend, but there is considerable variation in the levels. The
series from the Longitudinal Business Database is a clear outlier. This is likely to be due to
coverage, as this source includes certain economic activities that are excluded in the other
sources, particularly farms, public administration and state education.
Statistics of US Businesses and Business Employment Dynamics have very similar coverage,
but are based on two different business frames. Frame maintenance and timing of updates is
likely to account for the slight differences between them.
The Firm Size Data source is derived from the same register as Statistics of US Businesses,
with identical coverage, hence the similar trend. It shows a lower level, which is to be
expected as this is the only source based on the firm, which can be an aggregate of the
individual establishments used in the other sources. This effect would be greater if it was not
partly cancelled out by the inclusion of all firms that had an active payroll during the year,
rather than just on the March reference date used for Statistics of US Businesses. The table
below is derived from US Census Bureau data, and shows that the population of firms would
be about 13% lower if it only included those with a payroll in March.
Year Employer Firms
(whole year)
Firms with no
employees in March
Employer Firms
(March)
% Difference
1997 5,541,918 719,978 4,821,940
-12.99%
1998 5,579,177 711,899 4,867,278
-12.76%
1999 5,607,743 709,074 4,898,669
-12.64%
2000 5,652,544 726,862 4,925,682
-12.86%
2001 5,657,774 703,837 4,953,937
-12.44%
2002 5,697,759 770,041 4,927,718
-13.51%
146
Conclusions
In one sense it is easier to compare different data sources from one country, than data from
different countries, as, at least in theory, they should give the same answer when all of the
differences in methodology have been removed. However, this exercise proves the benefit of
highlighting those differences in methodology which might otherwise have been put down to
genuine variations if data from different countries were being compared. This, in turn, gives a
better understanding of the factors affecting comparability, which can then be used to improve
international comparisons of data by separating out genuine variations from those caused by
methodological differences.
It is possible to explain most of the apparent discrepancies between the data from the
different sources considered for each country above by a close study of the metadata
available for each source, and by making assumptions (of varying degrees of robustness) of
the impact of the main differences in methodology. The lack of standardisation of metadata, in
terms of content, terminology and presentation, sometimes combined with a certain lack of
clarity, particularly for non-specialists, makes this task rather more difficult than it should be.
For several of the countries above the explanations and assumptions have been verified by,
those responsible for the source, or are partly based on additional information from national
experts. This shows that the metadata necessary for better informed international
comparisons could be made available relatively easily.
147
Annex 5 – Business Closures
Introduction
The focus of this report has been on the comparability of data on business start-up rates, as
this is perhaps the key indicator for studies of business demography, and one of the most
important for entrepreneurship. Business start-ups, however, only give part of the picture. To
properly understand and interpret the data, it is necessary to know the extent to which new
businesses are adding to economic activity or replacing businesses that have closed.
Measures of business closure rates are therefore a very useful complement to start-up
indicators.
This interdependency between start-up and closure rate data has been recognised in the
Eurostat business demography project, where there has been a specific effort to develop
methodologies for these indicators that closely mirror each other. This approach seems both
logical and successful, and implies that the factors of comparability affecting business start-up
rates proposed in the main body of this report are also likely to be relevant for closure rates.
This Annex explores this hypothesis further, looking at similarities and differences in the ways
the factors can be applied.
The chart below shows business closure rate data for a number of countries, including two
sources for the United States, as published by those countries or Eurostat (for more
information on the sources, see Annex 2). It complements that on start-up rates in the
introduction to the main body of this report (Figure 1.1). It shows a similar degree of variability,
though there are also more gaps in the data.
Business Closure Rate Data for Selected Countries
0%
5%
10%
15%
20%
25%
A
u
s
t
r
a
l
i
a
C
a
n
a
d
a
D
e
n
m
a
r
k
F
i
n
l
a
n
d
G
e
r
m
a
n
y
I
t
a
l
y
L
a
t
v
i
a
N
e
t
h
e
r
l
a
n
d
s
N
e
w
Z
e
a
l
a
n
d
N
o
r
w
a
y
P
o
r
t
u
g
a
l
S
l
o
v
a
k
i
a
S
p
a
i
n
S
w
e
d
e
n
U
K
U
S
(
S
U
S
B
)
U
S
(
B
L
S
)
2000 2001 2002
Sources: National statistical office and Eurostat publications and internet sites
As is the case for Figure 1.1, the chart above is not a particularly meaningful or valid
comparison, as the variability shown is due rather more to methodological differences than to
real variations between countries.
148
Factors Affecting the Comparability of Business Closure Rates
This section considers how far the factors of comparability for business start-up rates
identified in Section 3 of this report can be applied to business closure rate data.
• Purity
It is clear from the discussions on the purity of start-up data that the separation of “pure
deaths” from other exits will have a significant impact on data from many sources. It is
logically easier to consider purity of start-ups and closures together rather than separately,
because in many cases, apparent closures and start-ups can be linked, proving that these
businesses have in fact continued to operate, despite appearances to the contrary. There is
thus likely to be a strong correlation between the ratios of pure births to total entries, and pure
deaths to total exits, for any given source.
As with start-ups, reactivations can be difficult to deal with conceptually. A business that is
dormant for a few months before re-starting would not normally be considered to be a pure
death, however longer periods are not so easy to deal with. If a threshold is applied for start-
ups, it is logical that the same threshold should be applied for closures, otherwise businesses
will not be treated in a consistent way in the two data sets.
There is a specific problem for closures, however, in that the longer the period of time allowed
for potential reactivations, the greater the lag in the production of definitive closure rate data.
Eurostat apply the same two-year reactivation threshold for closures as they do for start-ups,
so the lag for definitive closure data is two years longer than for start-up data relating to the
same period. This is partly resolved by the release of provisional estimates until the definitive
closure rates can be calculated.
• Timing
This issue is often more significant for business closures than for start-ups, as the closure
process can take many years in some cases, and reporting of closures to administrative and
fiscal bodies tends to be rather slower than for start-ups. An entrepreneur might consider a
business to be closed from the day he or she stops taking on new work or ceases trading. For
accounting purposes there is likely to be a further period during which payments are sought
from debtors and made to creditors, until the business accounts can be finalised. There may
then be a further period for administrative or fiscal purposes during which any outstanding
obligations are either fulfilled or written-off. Finally there may also be some sort of legal
procedure, which may take place before, during or after the above. The point at which a
closure is recorded will therefore be determined by the nature of the data source.
As for start-ups, a closely related issue is that of lags, the different events above may not be
notified immediately. Work on lags in the value-added tax data used for the British statistical
business register, described in Section 3.2 of this report, revealed whilst registration lags had
a median value close to forty days, that for de-registrations was nearly seventy days.
• Periodicity
The issues affecting closure rate data are similar to those described for start-up data.
149
• Type of Population
Virtually all business closure rate data currently available use businesses rather than people
as the population, so this factor is unlikely to be important for the comparability of existing
closure rate data.
• Temporal Basis
The differences between point in time and live during period populations are described in
Section 3.5 of this report, and, in more detail, in Annex 3. The use of a live during period
population will result in a higher denominator and lower closure rates. The issues affecting
closure rates are the same as those affecting start-up rates.
• Source
Where the source of closure data is a statistical or administrative business register, the issues
are largely the same as for start-up data, however it is important to know how closures are
defined in these sources, at what point in the process closures are identified, and with what
lags (see the comments on timing above). This is often more complicated for statistical
business registers, as these tend to be updated from a number of statistical and
administrative sources, all of which may have different definitions and lags for closures.
There is a fundamental problem with survey data on closures, in that if a business has closed,
it is often difficult to make contact to confirm this, so it can be difficult to differentiate between
closures, businesses that choose not to respond, and those that can not be contacted
because they have moved to an unknown address. The wider coverage of a census can help
to reduce this problem, but is unlikely to eliminate it.
• Units
The issues affecting closure rates are similar to those affecting start-up rates. It should be
noted that establishment or legal unit closures do not necessarily equate to enterprise
closures.
• Coverage
In some cases, closures may be indicated when a business is still active, but has moved out
of scope of the source. Sometimes this may be entirely due to a change in the source rather
than any change on the part of the business. It is also possible that businesses that do not
respond to, or comply with the requirements of a particular source, may be treated as closed
by that source, usually after a certain number of periods of non-compliance or non-response.
• Thresholds
Similarly, a business can appear to close if it moves out of scope of a particular source by
dropping below a certain threshold. Such businesses, however, often continue to operate,
albeit at a lower level. An example may be an artisan who reaches retirement age and stops
full-time activities, but continues his or her business on a part-time basis, perhaps just for a
limited range of customers.
Businesses that operate close to the threshold used for a particular source, e.g. sole
proprietors that take on employees only when market conditions allow, are likely to move in
and out of scope of that source, possibly quite frequently. They would normally be considered
to be reactivations, as discussed in the section on purity above, rather than pure deaths
150
followed by pure births. As for start-ups, the international application of a standard threshold
would considerably improve the comparability of business closure rates.
• Other Factors
As with business start-ups, various other factors can affect the international comparability of
closure rates, including the complexity of administrative procedures, the impact of tax, subsidy
and other policies, the nature of the political system, and a wide range of other economic,
political, social and cultural factors. These factors relate more the sort of variation in data that
users are really interested in, than to data production methodology. Thus, as for start-ups, this
Annex only focuses on methodological factors of comparability. If these can be better
understood, and eliminated where possible, data users have a much better chance to observe
the non-methodological factors in a less biased way.
Conclusion
This annex demonstrates that the factors of comparability derived for business start-up rates
in the main body of this report, can also be applied relatively easily to data on business
closures. There are some differences in terms of the relative importance of the different
factors, for example identifying a business closure, and attributing it to a specific point in time
is often more of a problem than identifying when start-ups occur. However, these differences
are relatively minor, so it is recommended to treat closures as complementary to start-ups in
terms of developing data collection and comparison methodology.
doc_408540456.pdf
Growing political and academic interest in entrepreneurship and business demography, and particularly the role and value of new businesses in national economies, is prompting various research projects on these topics. One of the main issues faced by researchers and policy makers is the current lack of international comparability of data on business start-up rates, which are often seen as key indicators of entrepreneurship and economic dynamism.
The International Comparability of
Business Start-up Rates
Final Report
Steven Vale
January 2006
2
3
The International Comparability of Business Start-up Rates
Final Report
Table of Contents
0. Executive Summary
1. Introduction
2. Data Sources and Existing International Comparisons
3. Factors Affecting Comparability
4. Methods to Improve Comparability
5. A Harmonised Methodological Framework and Start-up Indicators?
6. Conclusions
7. References
Annex 1. Glossary of Terms: Proposals for Harmonised Terminology
Annex 2. Inventory of Data on Business Start-ups by Country
Annex 3. Defining Business Populations: Comparing Point in Time and Live During
Period Estimates
Annex 4. Business Start-up Data for Selected Countries: Comparisons of National
Sources
Annex 5. Business Closures
Author: Steven Vale
OECD Statistics Directorate / Office for National Statistics, UK
Contact Details
Please send any comments or questions to the author at: [email protected]
Comments or questions to the OECD should be addressed to:
Nadim Ahmad – [email protected] (Business demography)
Tim Davis – [email protected] (Entrepreneurship indicators)
4
5
0. Executive Summary
Growing political and academic interest in entrepreneurship and business demography, and
particularly the role and value of new businesses in national economies, is prompting various
research projects on these topics. One of the main issues faced by researchers and policy
makers is the current lack of international comparability of data on business start-up rates,
which are often seen as key indicators of entrepreneurship and economic dynamism. The
International Consortium for Dynamic Entrepreneurship Benchmarking, led by the Danish
government agency FORA, has responded by providing funding for a five month consultancy
at the OECD to study this topic. The consultant appointed for this task was Steven Vale, on
secondment from the UK Office for National Statistics.
The objectives of the project were:
• The compilation of existing evidence on comparative start-up rates;
• The comparison of results and identification of reasons for differences in results, in
particular methodological and statistical differences;
• Drawing up lessons for future studies to improve comparability and to ensure that
results are meaningful.
The underlying question that this project has aimed to answer is; “How comparable are
existing data on business start-up rates from different OECD countries?” The short answer is;
“Not very”, so this report looks at the reasons why data are not comparable, and what can be
done to improve comparability in the future.
This report starts by examining the existing sources of business start-up data for different
countries (an inventory of sources is included in Annex 2), and assessing previous
international projects and papers that have aimed to produce comparable data for groups of
countries. Where there are several data sources for a particular country, they have been
studied to gain a better understanding why they often differ (see Annex 4). The conclusion
from this work is that there are a number of factors that affect the comparability of business
start-up data, some of which may have been overlooked in previous international
comparisons, resulting in the true variability of data between countries being masked by
methodological differences.
Section 3 develops these ideas into a typology of the factors affecting international
comparisons of business start-up rates, describing each factor, and its potential impact in
detail. Start-up rates are based on two components, the numerator (new businesses), and the
denominator (a population). Some factors affect just one of these, others affect both. In total,
nine factors have been identified:
Numerator factors:
• Purity – to what extent are “pure births” (i.e. new combinations of production factors)
distinguished from reactivations and other creations?
• Timing – at what point in the creation process is a start-up measured?
• Periodicity – over what period are start-ups measured, and how does this affect the
measurement of very short-lived businesses?
Denominator factors:
• Type of Population – businesses or people?
• Temporal basis – is the population measured at a specific point in time, or does it
consist of all units that were present at any time during a given period?
6
Factors affecting both:
• Source – are the data taken from a register, a census or a survey? How reliable is the
source?
• Units – what is the entity about which the data are produced?
• Coverage - to what extent are certain types of business included or excluded based on
specific attributes (e.g. economic activity or legal form)?
• Thresholds – what explicit or implicit size thresholds apply to the source?
Section 4 looks at how these factors can affect data comparisons in practice. It shows that
adjustments to compensate for differences in specific factors can sometimes help to improve
comparability, but have to be made with care, based on a detailed understanding of the data
sources and methods. In this sense, although not perfect, informed adjustments can at least
give approximate results, and can warn against drawing false conclusions based on the raw
data alone.
The goal of more comparable data is the theme of Section 5, which links this project to wider
OECD work to develop a methodological framework for business demography. This section
also looks at the pros and cons of different types of business start-up indicators, and
recommends focussing on one key indicator, supplemented by several secondary indicators.
The conclusions of this report are that:
• Simple comparisons of start-up rates from different sources are often misleading.
• The availability of data on business start-up rates varies considerably between countries.
• Where metadata exist, they are not always easy to find or understand. A harmonised
terminology is proposed in Annex 1, and a common metadata template is needed.
• Some previous international comparisons do not fully recognise all comparability issues,
but have provided useful models for assembling data from different countries.
• To assess the comparability of business start-up rates it is necessary to decompose them
into numerator and denominator components, and consider the factors that affect each.
• The factors that have the most impact are usually the purity of the data in the numerator,
the temporal basis of the denominator, and the coverage of both.
• The larger a “new” business is, the less likely it is to be a pure birth. Increasing purity
leads to a considerable reduction in the employment attributed to new businesses.
• Analytical adjustments can help to compensate for differences in specific comparability
factors, but risk introducing noise into the data, so have to be made with care.
• Statistical business registers are the best sources for business start-up data, as they are
already subject to a degree of harmonisation, particularly within Europe.
• A clearly defined key indicator would improve data comparability. Secondary indicators
could give additional insights to more specialist data users.
• Data producers are often more influenced by national data requirements than international
comparability. The OECD has a role to communicate international needs.
• It is important to find out what data users really want, and what they use start-up data for.
This knowledge can then inform the future development of indicators.
• The short term priority is the identification of “quick wins”, i.e. actions that increase the
international comparability of data from individual countries for minimal cost.
• A step-by step approach may not result in fully comparable data as quickly as some users
might want, though alternative, more radical, approaches may take at least as long, as
they would require considerable changes to methods and sources in many countries.
• The goal of internationally comparable business start-up rates is not an easy one, but is
possible.
7
1. Introduction
There is growing international interest in the topics of business dynamics and
entrepreneurship, particularly from policy makers and academic researchers. Business start-
up rates are seen as providing key indicators for both purposes. They are also used as a
measure of economic dynamism, and have been linked to improvements in productivity
through the notion of creative destruction
1
.
So far the focus has mainly been on producing national data to inform national policies and
research, however, there is a growing interest in international comparisons, particularly for
benchmarking purposes. To facilitate international comparisons, it is necessary to determine
measures of business start-ups that will show the real differences between countries, and not
just reflect differences in national methodologies, as has often been the case in the past.
For this reason, the International Consortium for Dynamic Entrepreneurship Benchmarking,
led by the Danish government agency FORA, has provided funding for a five month
consultancy at the OECD to study the international comparability of business start-up data.
The consultant appointed for this task was Steven Vale, on secondment from the UK Office
for National Statistics.
The objectives of the project were agreed at the outset as being:
• The compilation of existing evidence on comparative start-up rates;
• The comparison of results and identification of reasons for differences in results, in
particular methodological and statistical differences;
• Drawing up lessons for future studies to improve comparability and to ensure that results
are meaningful.
The underlying question that this project has aimed to answer is; “How comparable are data
on business start-up rates from different OECD countries?” Figure 1.1 shows business start-
up rate data for a number of countries, including two sources for the United States, as
published by those countries or Eurostat. Is this chart a valid comparison of business start-up
rates for these countries?
Figure 1.1 – Raw Business Start-up Rate Data for Selected Countries
0%
5%
10%
15%
20%
25%
A
u
s
t
r
a
l
i
a
C
a
n
a
d
a
D
e
n
m
a
r
k
F
i
n
l
a
n
d
F
r
a
n
c
e
G
e
r
m
a
n
y
I
t
a
l
y
L
a
t
v
i
a
N
e
t
h
e
r
l
a
n
d
s
N
e
w
Z
e
a
l
a
n
d
N
o
r
w
a
y
P
o
r
t
u
g
a
l
S
l
o
v
a
k
i
a
S
p
a
i
n
S
w
e
d
e
n
U
K
U
S
(
S
U
S
B
)
U
S
(
B
L
S
)
2000 2001 2002
Sources: National statistical office and Eurostat publications and internet sites
1
Although the focus of this report is on business start-ups, the comparability issues affecting the
complimentary indicator of business closures are set out in Annex 5.
8
This report will show that the comparison in Figure 1.1 is not particularly valid, but that through
an understanding of the data and metadata, meaningful comparisons are possible. To reach
this conclusion, this report decomposes the questions above into a number of sub-questions,
corresponding to the different sub-tasks undertaken within this project:
• What data are available for each OECD country? - The project started by making an
inventory of data sources by country, initially through Internet searches, but also through
discussions with contacts in different countries. A copy of this inventory is included as
Annex 2.
• What metadata are available with these data? – The availability and quality of metadata
for each data source were assessed within the inventory.
• What comparisons or compendiums of data from different countries exist? – A trawl was
made of databases, literature and other sources combining data on start-up rates from
more than one country. Section 2 considers how others have tried to collect and compare
data from different countries, with varying degrees of success.
• Do the data seem comparable? – The above steps gave an initial view as to the degree of
data comparability. The conclusion was that methodological differences frequently mask
the real variations between countries.
• Do the metadata confirm comparability or explain the differences? – This initial view on the
comparability of data was tested using the available metadata, to determine how helpful
these metadata are in highlighting and explaining methodological differences. Annex 4
includes short studies on the comparability of sources within selected countries, on the
assumption that differences in data relating to the same country must be purely
methodological. This work led to the development of the framework of factors affecting the
comparability of start-up rates proposed in Section 3.
• Are there other explanations for differences in data? – The extent to which variations
between countries could be explained by political, social and cultural factors was briefly
considered, though this question is not considered further in this report, as it is more
appropriate to look at these issues when the data have been compiled or corrected to
remove methodological differences.
• How can comparability be improved for existing data? – Methods to make adjustments to
existing data to improve comparability are considered in Section 4, where examples are
used to illustrate how data can be adjusted, and some of the potential pitfalls.
• What is the scope for improving comparability at source? – Finally, Section 5 considers
the extent to which it is possible to recommend changes to the ways the source data are
produced to improve comparability, and proposes a set of standard indicators, within a
harmonised methodological framework.
There is a strong link between this project and other OECD work on business demography,
where this report will feed into the development of a wider methodological framework
including business survival, growth and closure. There are also links to OECD work on
entrepreneurship where there are plans to develop a set of harmonised indicators, including
business start-up rates. Outside the OECD there are links to Eurostat work on business
demography and the factors of business success, as well as to various international groups
concerned with business demography, entrepreneurship and statistical business registers.
9
2. Data Sources and Existing International Comparisons
Most OECD countries have produced indicators on business start-up rates, usually derived
from data held in statistical business registers. However, the methodology used has often
been driven by national considerations, rather than a desire for international comparability. A
quote from a recent Australian paper on establishing a conceptual framework for business
demography (ABS (2004)) illustrates this perfectly; “Whilst international comparability of the
data is considered to be important, the overriding requirement is the provision of data in the
Australian context”. This is not stated as clearly by other national data providers, but appears
to be a widely held view
2
. Understanding the methodological differences between data from
different countries is therefore a vital pre-condition to any meaningful comparisons.
2.1 An Inventory
The first step in this project was the compilation of an inventory of the different sources of data
on start-ups in the OECD member countries (see Annex 2 for a summary version). This
inventory is based on searches of the internet during autumn 2005, and thus will miss any
sources made available after that date, or sources that are only available in other formats.
Linguistic limitations may also mean that some sources not available in English or French
have been missed.
The inventory includes information on metadata, where available, to try to gain a better
understanding of how comparable the different data sets really are. The availability of
metadata varies from source to source, from virtually none to detailed papers describing every
aspect of the source, definitions and methodology. The lack of standards in the presentation
of metadata, and the availability of more detailed information only in the national language
often made the task of understanding the metadata more difficult, and may have contributed
to any errors in interpretation. This highlights the need for the uniform application of metadata
standards to help data users to better understand differences in data, particularly when
making international comparisons.
Whilst international comparisons can be problematic, some countries have several data-sets
available, based on different sources, which often give rather different measures of business
start-ups at the national level. The assumption in this project is that any variation between
sources relating to the same country must be purely methodological, i.e. linked to differences
in definitions, coverage, thresholds, or any of the other factors affecting comparability
identified in Section 3. This assumption has been tested on data for several countries (see
Annex 4), where it has proved generally possible to explain differences in data in terms of the
methodology used to produce them.
2.2 Other International Comparisons
Before starting to compare data for different countries, it is useful to see what can be learned
from previous work in this area. There have been several attempts over recent years to
provide internationally comparable business start-up data, either by international
organisations with an interest in harmonised statistics, or by individual countries seeking to
benchmark their data in a meaningful way. Some of the main work in this area is summarised
below, with an assessment of the level of comparability achieved.
2
For example, the conflicting requirements of national and international users of United Kingdom data
are considered in detail in Vale and Powell (2003).
10
• Demography of Small and Medium-sized Enterprises (DOSME) – Eurostat
The DOSME project was funded by the European Union from the mid-1990s until 2003 to
produce data on business demography and factors affecting business success in twelve
central and eastern European countries
3
, as they made the transition to a market economy.
The project was based on a series of surveys, which effectively created several panels of
businesses over time, and allowed the study of start-up and exit rates, survival, and the
characteristics of the entrepreneur. The result was a firm-level dataset that, subject to
confidentiality constraints, provides a useful resource for research on the development of the
business economy in these countries during this transition period. Full information about this
project is contained on the DOSME web site -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm.
In terms of producing comparable data on business start-ups, this project was quite
successful in developing and applying standard methodologies. However, the survey-based
approach, differences in the administrative sources used, as well as coverage and general
data quality issues, do cause some problems. The final stage of the project included finding
ways to overcome some of these issues analytically, based on the variables available in the
dataset, and even managed a reasonably robust comparison of data with those from the more
recent Eurostat business demography project. It must, however, be remembered that the
DOSME project observed these countries during an atypical period in their economic
development.
• Firm-level Data Project – OECD / World Bank
This project attempted to create harmonised firm-level databases in ten OECD member
countries
4
, with the aim of using these to produce comparable data on business dynamics.
Researchers in each country were responsible for running standard analyses of their micro-
data, with the resulting aggregates being shared for further cross-country analyses. The
project is described in detail, along with some of the resulting analyses, in various papers
linked to the project home page within the OECD web site:http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html.
The data cover different periods between 1978 and 1998 depending on what was available at
the national level, with the widest coverage in the early 1990’s. They are based on a variety of
sources, and not all countries were able to produce start-up rates in line with the project
definitions, for example some countries were not able to use the standard threshold of one
employee. Comparisons with more recent Eurostat data have highlighted these and other
quality issues (e.g. Brandt (2004)), often linked to improvements to the coverage and
maintenance procedures of statistical business registers during the 1990’s.
The World Bank has recently funded work to extend this approach to cover a further fourteen,
mostly developing, countries
5
. This is documented in two papers by Bartelsman, Haltiwanger
and Scarpetta (Bartelsman et al (2004), and Bartelsman et al (2005)). Various threshold and
3
Albania, Bulgaria, Czech Republic, Estonia, Former Yugoslav Republic of Macedonia (FYROM),
Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic and Slovenia.
4
Canada, Denmark, Finland, France, Germany, Italy, the Netherlands, Portugal, United Kingdom and
United States
5
Argentina, Brazil, Chile, Colombia, Estonia, Hungary, Indonesia, Latvia, Mexico, Romania, Slovenia,
South Korea, Chinese Taipei and Venezuela
11
coverage issues that might affect data comparability, particularly for business start-ups, are
noted in those papers.
It could be argued that the original OECD firm-level data project was a little too ahead of its
time, and that the resulting data are subject to a number of comparability issues that could
not realistically be resolved at the time; indeed some of these have only recently started to be
resolved at the national level. Having said this, many of the analytical techniques used seem
to have been robust enough to give plausible results despite the limitations of the basic data.
Also, putting data issues to one side, the approach of distributed analyses of standardised
micro-data seems worth pursuing in any future projects of this nature, as it avoids data
confidentiality issues, and makes use of national knowledge about the data.
• Business Environment and Firm Entry – NBER / World Bank
This study (Klapper et al (2004)) is published as a Working Paper of the US National Bureau
of Economic Research (NBER), acknowledging financial support from the World Bank. It is
available on the NBER website athttp://www.nber.org/papers/w10380. It compares business
start up data for over twenty European countries using data, mainly on corporate businesses,
from the Amadeus database compiled by the private sector business data provider, Bureau
Van Dijk. The results are also compared to US data sourced from Dun and Bradstreet, though
comparisons may be affected by differences in the way the sources are compiled.
The results are broadly in line with other sources, though some results such as an average
start-up rate of 3.46% for Italy compared to 11.13% for Finland seem to be at odds with
Eurostat figures (8.35% and 7.48% respectively). This is almost certainly due to the restriction
to corporate businesses, and raises additional comparability issues related to variations in the
propensity of businesses to incorporate. This will differ between countries depending on the
cost and complexity of registration procedures, tax incentives, reporting burdens and possibly
even cultural factors. Variations in the extent of re-registration in national systems, for
example when a business changes its name, may also affect comparability.
• Eurostat Business Demography Project
This project brings together data for European Union countries (plus Norway and Romania)
on business start-ups (births) closures (deaths), survival and growth, produced by national
statistical offices using a common methodology. So far it has been run on a voluntary basis,
which has resulted in a lack of data for some of the larger countries, particularly Germany and
France, though it will soon become a legal requirement through the forthcoming revision to the
Structural Business Statistics Regulation.
In terms of data comparability, this is probably the most successful international project to
date, as the methodology to be followed at the national level is very detailed, and was tested
and refined using pilot studies. The methodology is based on the use of business register
data. The registers themselves are subject to a considerable degree of harmonisation due to
the existence for over ten years of a European Union regulation on statistical business
registers
6
, which requires minimum standards of contents and coverage. Unfortunately this
does not mean that the data can be considered fully comparable yet, as different national
thresholds affect the smallest size-classes, and matching procedures to separate pure births
6
Council Regulation (EEC) No 2186/93 of 22 J uly 1993 on Community co-ordination in drawing up
business registers for statistical purposes -http://europa.eu.int/eur-lex/lex/LexUriServ/LexUriServ.do?uri=CELEX:31993R2186:EN:HTML
12
from other creations are affected by the availability and quality of key matching variables, as
well as the use of different matching tools.
Data and summary methodology resulting from this project are available via the Eurostat web
site (http://epp.eurostat.cec.eu.int). A more detailed methodological manual has been
produced, but not yet been published
7
.
• Global Entrepreneurship Monitor
The Global Entrepreneurship Monitor (GEM) project collects data on various aspects of
entrepreneurship through a series of coordinated household surveys in a gradually increasing
number of countries world-wide. More information on the project and participants can be
found at;http://www.gemconsortium.org/. One of the key outputs of the GEM work is an
indicator of “Total Entrepreneurial Activity” (TEA)
8
, which measures those respondents who
have recently started a business, or have started taking steps towards setting up a new
business. The TEA index is therefore not strictly a measure of business start-up rates, but
should provide a reasonable indicator.
Figure 2.1 - Comparing GEM TEA Rates and Eurostat Business Start-up Rates
2000
0%
5%
10%
15%
0% 5% 10% 15%
Eurostat Start-up Rate
G
E
M
T
E
A
2001
0%
5%
10%
15%
0% 5% 10% 15%
Eurostat Start-up Rate
G
E
M
T
E
A
2002
0%
5%
10%
15%
0% 5% 10% 15%
Eurostat Start-up Rate
G
E
M
T
E
A
Source: GEM 2004 Global Report and Eurostat web site
7
Business Demography Recommendations Manual, Eurostat, latest draft December 2004.
8
This is also referred to as the “Early Stage Prevalence Rate” in the 2005 GEM report.
Correlation Coefficients
2000 =0.341
2001 =0.489
2002 =0.763
13
Figure 2.1 shows that the degree of correlation between the GEM TEA rates and start-up
rates for countries contributing to the Eurostat business demography project seems to be
increasing over time, possibly reflecting data quality and methodological improvements in both
sources. Eurostat start-up data are used here because they provide the most reliable cross-
country comparisons currently available. TEA rates could also be compared with business
start-up rates from other sources and countries, but the current lack of harmonisation of start-
up rate methodologies used by these sources would cause distortions and increase the risk of
misleading results.
The TEA rates are roughly comparable in magnitude to the Eurostat start-up rates in 2000
and 2001, but appear to drop in 2002, probably due to methodological changes in the GEM
data. The relatively small GEM sample sizes in many countries
9
may affect the reliability of
these data, however, the increasing degree of correlation between data from these sources
could be seen as a positive indicator of the quality of both data sets.
• National Benchmarking – Canada and New Zealand
Two papers have been identified that consider the international comparability of business
start-up data in the context of benchmarking national data. The first, Baldwin et al (2002),
looks at different sources of data within Canada, and explains the differences in terms of the
methodologies used. The paper then considers how these methodological issues could affect
international data comparisons.
The second paper, Mills and Timmins (2004), seeks to establish if business dynamics in New
Zealand are really as different to those of other OECD countries as previous studies have
indicated. It concludes that when measurement differences, particularly relating to the
coverage of very small businesses, are taken into account, the New Zealand data are not very
different to those of other countries.
Both of these papers are useful in identifying some of the reasons why existing estimates of
start-up data may not be comparable across countries, and have informed the development of
the factors of comparability set out in Section 3 below. They clearly show that comparisons of
data from different sources must include comparisons of the metadata.
9
GEM sample sizes increased from 2,000 to over 15,000 people per year in the UK during this period,
which could be expected to help improve comparability with register-based business data such as
those from Eurostat, however they remained stable at around 2,000 per year for each of the other
countries included in the charts in Figure 2.1.
14
15
3. Factors Affecting Comparability
This section of the report aims to identify the different factors affecting the comparability of
data on business start-ups, and to highlight the main issues involved. At first glance, the
number and range of factors that affect comparability can make the task of compiling
comparable data appear to be virtually impossible. The aim of this report is not to discourage
the reader from trying to make comparisons, but to explore in detail the factors affecting
comparability. If these are better understood, they may be more easily overcome, or it will at
least be possible to make more informed decisions about which ones have little enough
impact that they can safely be ignored.
J ust as comparability is often listed in typologies of the components of statistical data quality,
so it is possible to develop a typology of the factors affecting comparability. Looking at this in
another way, such a typology can also provide a list of the reasons why data may not be
comparable. Focussing specifically on the area of the international comparability of business
start-up rates, these factors can be defined either in terms of the numerator (the number of
new businesses), the denominator (the population or stock), or both (assuming the
denominator is based on businesses)
10
.
The approach of separating numerator and denominator factors is based on the study of
differences between data sources within countries (see Annex 4). This work clearly shows the
range of factors that can affect data comparability between sources that are attempting to
measure the same phenomenon for the same country. It also demonstrates that there is a
complex interaction between these factors.
The three charts in Figure 3.1 below are taken from Annex 4, where they, and similar charts
for nine other countries, are discussed in detail, and the reasons for the differences are
explained. They compare United States data from various sources, and demonstrate clearly
how start-up rate indicators that appear to be similar are actually quite different when they are
split into their components.
The Business Employment Dynamics quarterly data set is a clear outlier in terms of start-up
rates, though the annualised data from this source show that this is almost entirely due to
periodicity and data purity issues. The remaining data sources appear to give fairly
comparable measures of start-up rates, typically between 10% and 13%, though these mask
the differences in the populations of new and existing businesses used to derive these rates.
10
Business start-ups can also be measured in terms of employment creation rather than numbers of
new businesses (see Baldwin et al (2002)). This measure is less sensitive to the inclusion or exclusion
of very small units, but is more sensitive to the type of unit used (new establishments of existing
enterprises can be very large), and the inclusion of events other than pure births (which tend to involve
larger businesses). This approach is not considered further in this section for the purely pragmatic
reasons that more data are available on counts of businesses than on employment, and that
employment of new businesses can be rather difficult to measure accurately. It is, however, revisited in
Section 5 of this report, which considers possible supplementary indicators.
16
Figure 3.1 – A Comparison of Different Sources of Start-up Rates, New Businesses and
Business Populations in the US
a) Start-up Rates
0%
5%
10%
15%
20%
25%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
Business Employment Dynamics - Summed Quarterly Data Firm Size Data
Business Employment Dynamics - Annualised Data Statistics of US Businesses
Longitudinal Business Database OECD Firm-level Data
b) New Businesses
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Business Employment Dynamics - Summed Quarterly Data Firm Size Data
Business Employment Dynamics - Annualised Data Statistics of US Businesses
Longitudinal Business Database OECD Firm-level Data
c) Business Populations
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Business Employment Dynamics Statistics of US Businesses
Longitudinal Business Database OECD Firm-level Data
Firm Size Data
17
The typology approach has been followed below, resulting in a set of nine factors affecting the
comparability of business start-up rates, each of which is considered in more detail in
Sections 3.1 to 3.9. This typology has been developed based on reactions to earlier drafts
proposed in Vale (2005(a)), and Ahmad and Vale (2005).
Numerator factors:
• Purity – to what extent are “pure births” distinguished from reactivations and other
creations?
• Timing – at what point in the creation process is a start-up measured?
• Periodicity – over what period are start-ups measured, and how does this affect the
measurement of very short-lived businesses?
Denominator factors:
• Type of Population – businesses or people?
• Temporal basis – is the population measured at a specific point in time, or does it
consist of all units that were present at any time during a given period?
Factors affecting both:
• Source – are the data taken from a register, a census or a survey? How reliable is the
source?
• Units – what is the entity about which the data are produced?
• Coverage - to what extent are certain types of business included or excluded based on
specific attributes (e.g. economic activity or legal form)?
• Thresholds – what explicit or implicit size thresholds apply to the source?
Various other factors can be identified as affecting comparability of start-up data, such as the
size of national economies, demand and supply constraints, the impact of tax, subsidy and
other policies, the nature of the political system, and a wide range of other economic, political,
social and cultural factors. None of these factors relate to the data production methodology,
and many of them account for the sort of variation in data that users are really interested in.
Indeed if they were all eliminated, the data would be identical for each country, and of no real
use to anyone. For this reason, this report only focuses on the nine methodological factors of
comparability listed above. If these can be understood, and their impact reduced, data users
have a much better chance to observe the non-methodological factors in a less biased way.
3.1 Purity
It is often relatively easy to measure business entries, i.e. those businesses that are present
in a given period but were not present in the previous period. It is rather more difficult to
separate out pure births (sometimes referred to as creations ex nihilo) from entries due to re-
registrations, reactivations, take-overs and other demographic events
11
, i.e. those entries that
are either the continuation of an activity previously carried out under a different unit, or a
reactivation of a business that has been active in the recent past, but was dormant (or not
recorded) in the previous period. The term “purity” is therefore used to refer to the extent to
which business start-ups have been split into pure births and other entries.
11
For a typology of demographic events affecting businesses see Eurostat (2003).
18
The impact of separating pure births from other entries can be considerable. Figure 3.2 uses
data for France from the Agence Pour la Création d’Entreprises (APCE), which show that
around one third of all new businesses recorded by that source are not considered to be
creations ex nihilo.
Figure 3.2 Separating Pure Births from Other Creations in France
0
50
100
150
200
250
300
350
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
APCE Total Creations APCE Creations "Ex nihilo"
Business entries are usually derived from registrations with administrative sources, so can be
affected by administrative requirements to re-register in the event of certain changes, e.g. a
sole proprietor converting to a corporation. As legal and administrative requirements vary
considerably from country to country, and are likely to continue to do so, data on entries can
never be fully comparable between countries, whereas, at least in theory, data on pure births
can be.
Re-registrations, and other entries that are not pure births, can often be identified using data
matching techniques. A new unit that has a number of characteristics in common with a
previously existing unit, (e.g. name, address, economic activity, employees), is unlikely to be a
pure birth. Typically such matching will be automatic, or semi-automatic, based on rules or
algorithms to determine the likelihood that two units actually represent the same business in
the real world.
Methodology for matching to determine which creations are pure births has been developed in
the context of the Eurostat business demography project, though this has highlighted the need
to tune matching techniques to suit national data sources, and the danger of over-matching,
i.e. too many “false” matches. Experience in a number of countries shows that the larger a
business creation is (measured in terms of persons employed or turnover), the less likely it is
to be a pure birth.
Reactivations can also be difficult to deal with conceptually. A business that is dormant for a
few months (possibly due to seasonal activities) before re-starting would not be considered to
be a pure (ex nihilo) birth. However, if the period of dormancy was ten years or more, it would
be harder to argue that the business creation could be treated as a continuation of the
previous activity. A threshold may therefore be required. For the European Union this is
currently set at two years, whereas for data from the US Census Bureau longitudinal
database, all reactivations are excluded from the category of pure births, regardless of the
period of dormancy. The longer the period, the greater the delay in producing definitive data
19
on business closures, thus some sort of compromise is needed. This should be informed by a
better understanding of the reasons for dormancy, and the possibilities of adjusting for
reactivations using modelling based on historic data.
In countries where it is possible to link employers and employees over time, these links can
be used to help determine pure births. This method has been tested in New Zealand, where, if
at least 70% of employees appear to move from an old registration to a new one, it is
assumed that the new business is not a pure birth. Taken together with work to identify when
sites are transferred between businesses, this has resulted in around 20% of entries now
being confirmed not to be pure births. These businesses tend to be the larger entries,
accounting for around 60% of the employment attributed to entries (Mead (2005)).
Similar work in Canada is reported in Baldwin et al (2002), which also showed that using
linked employer-employee data from the Longitudinal Employment Analysis Program (LEAP)
file can reduce business start-up rates from an annual average of 18.5% to around 14.5%.
The fall was considerably more pronounced in terms of the employment attributed to start-
ups, which dropped from an annual average of 11.8% to just 4% (or 2.5% depending on how
and when employment was measured). These results are complemented in that paper by
survey data showing that firms entering the manufacturing sector by acquiring an existing
plant accounted for only 0.8% of plants, but 3.2% of employment, whereas those firms
entering manufacturing with a new plant accounted for 7% of plants, but only 2.1% of
employment.
Both studies appear to call into question the importance of business start-ups in terms of job
creation, demonstrating that where more advanced linkage techniques are used, the
employment that can be attributed to pure births is rather lower than previously thought. This
is backed up by findings in several countries participating in the Eurostat business
demography project, e.g. Cella and Viviano (2004). Pure births with more than 20 employees
seem to be quite rare in most countries, and tend to be limited to cases of inward investment,
a few labour intensive service activities, or manufacturing activities that have traditionally been
associated with high entry thresholds.
The figures from France, Canada and New Zealand clearly illustrate that any work to
distinguish pure births from other entries will result in lower start-up rates, therefore the
amount of such work undertaken should be considered when comparing data from different
sources. The potential impact on trends is less obvious. Total entries may well show similar
trends to pure births in the short term, but are more likely to be affected when administrative
sources and systems change.
3.2 Timing
This issue concerns differences in the point at which data sources record a business start-up.
This can vary from the time a person starts thinking about creating a new business to the time
a new business makes its first sales, reaches a certain financial or employment threshold, or
survives for a certain period. For some new businesses the time intervals between these
events is very short, for others it can be measured in years, whereas a third category do not
meet all the criteria, so could be measured as a start-up by one source, but not by another.
This third category demonstrate that sources that record start-ups at an early point in the
process tend to show higher start-up rates, particularly for very small businesses. These are,
of course, accompanied by higher exit rates.
20
Typically, the point at which a start-up is recorded is determined by the nature of the data
source. Surveys of people or households can measure intentions, administrative sources are
linked to more concrete legal or fiscal obligations, and surveys of businesses are typically
directed at those that have at least a certain level of economic activity.
There is a clear link here with the discussion on thresholds below, as sources with higher
thresholds are likely to record businesses at a later point in the start-up process than sources
with lower thresholds. For example, a business will only register with an administration
responsible for taxation of employee earnings when it takes on its first paid employee. This
could be some time after it has registered to pay sales or value-added tax.
It is also important to know whether certain sources allow pre-registration, i.e. where a
business can be registered in advance of actually starting activity. This is typically more
common for regulatory sources than taxation sources, but can happen for both. Ideally both a
registration and a start date are needed, but in practice it is usually necessary to use other
indicators of whether a business has really started such as tax returns, sales or employment.
A related issue concerns lags, i.e. the time difference between events taking place in the real
world, and being recorded in the data source. For statistical business registers the lags in
recording business start-ups depend on the source of the information, typically administrative
or tax registers. Figure 3.3 shows an analysis of business start-up lags for the British
statistical business register resulting from the use of value added tax (VAT) registration data.
Figure 3.3 – VAT Registration Lags Affecting the British Statistical Business Register
0%
5%
10%
15%
20%
25%
30%
-
5
00
5
0
1
0
0
1
5
0
2
0
0
2
5
0
3
0
0
3
5
0
4
0
0
4
5
0
5
0
0
5
5
0
6
0
0
Lag in Days
This chart shows that almost 80% of start-ups are notified within 100 days, and that very few
have a lag of more than a year
12
. Lags will obviously vary considerably depending on the
nature of the source and the frequency of updates. If the source records only the date of
notification, and start-up rates and lags are stable over time the impact will be negligible. If the
source attempts to record the actual start-up date, it will be necessary to either wait until the
impact of the remaining lags is insignificant before producing start-up data for a given period,
or to make adjustments based on the estimated effect of lags. Thus the main impact on
comparability due to lags will be in the most recent periods.
12
One reason for lags of more than a year is retrospective registration of businesses found not to have
declared their revenue to the tax authorities.
21
3.3 Periodicity
This issue concerns whether the data are sub-annual, annual, or less frequent. The majority
of the sources identified in Annex 2 concern annual data, though quarterly and monthly data
sets are available for some countries. In a few cases, data availability is linked to economic
censuses at intervals of five years.
For data with a periodicity of greater than one year it is difficult to construct start-up rates that
can be compared to annual data, as the proportion of short-lived firms that will be missed
increases considerably. In J apan, annualised average rates are calculated for the years
between censuses (Takahashi (2000)), but these mask the natural year on year variability
usually observed in start-up data.
If sub-annual data include counts of start-ups, they can simply be added to produce annual
totals, though these totals will be higher than those based on annual snap-shots due to better
coverage of businesses that survive for less than one year. If sub-annual start-up data are
only available in the form of birth rates, it is clearly more difficult to estimate the annual rate
without further information about the net change in the population.
Work to convert quarterly establishment start-up data from the Business Employment
Dynamics series produced by the US Bureau of Labor Statistics to an annual basis has
resulted in differences of over 40% between annualised start-ups and the sum of start-ups for
the four separate quarters. This is a result of both the removal of short-lived businesses, and
improvements to the purity of the start-up estimates by better linkage of establishments over
time, and is documented in Pinkston and Spletzer (2004).
This leads towards questions about the value of data for very short-lived businesses. Is a
business that only lasts for a month or two, perhaps with no employees, and possibly even no
sales, of any real interest? Would it be more meaningful to only consider start-ups that remain
active for at least a year, or some longer period? In terms of current data availability, often
based on annual snap-shots of the population of businesses, this becomes a rather difficult
question. Many of the businesses that are live for less than a year will be excluded altogether,
but those that, by chance, are live on the day the snap-shot is taken, will be included. This
could cause certain biases, for example a common reference date in a number of data sets is
31 December / 1 J anuary. Short-lived businesses with activities related to the Christmas
period are likely to be included, but, for the northern hemisphere, short-lived businesses with
certain tourism or agriculture-related activities could be under-represented.
Possible solutions include the recording of start-up and closure dates to allow a more
accurate measure of the period of survival, or only counting business start-ups that are
present in at least two consecutive periods. The use of dates is the more attractive and
flexible option, but it relies on the availability of accurate information. Linking the timing of the
start-up to an administrative event, such as coming into scope of an administrative source
might help, as the source is likely to record that date. Given the lack of harmonisation of sales
related taxation systems, administrative sources that record when a business takes on its first
paid employee are likely to be most appropriate in terms of international comparability.
Annual data may not be fully comparable if they refer to different periods. Typically the period
is the calendar year, but other periods such as March to March (United States) and J uly to
J uly (Australia) are also used. For strict comparisons on a calendar year basis, such data sets
would need to be apportioned between years, though in practice this may not be necessary if
start-up rates are fairly stable over time.
22
Finally, where data are annual, they may not reflect an exact calendar year. If the
observations are not taken on exactly the same day each year, there will be an impact on data
comparability both between countries and over time.
3.4 Type of Population
Two basic types of population can be used as the denominator for calculating business start-
up rates. The population of businesses is the most frequently used, however for some
countries and sources, particularly where household surveys are used to measure business
start-ups and entrepreneurship, the denominator can also be based on the human population.
Business populations can vary considerably in the way they are defined. Most of the issues
are covered in the sections on coverage and thresholds below, but one specific point to note
here is the extent to which the population includes non-active units. The requirement in the
Eurostat business demography methodology for population units to be active in terms of
having turnover and/or employment at some point during the reference period is rather more
restrictive than taking, for example, a count of all current registrations.
Both types of population raise potential issues for international comparability, particularly
where there are large differences in the structure of the population between countries. For
example, using the total human population of a country as a denominator is likely to give
higher start-up rates for countries with a higher proportion of the population considered to be
of working age, than those with higher proportions of children or retired people.
It may also be necessary to have some knowledge about under-coverage due to factors such
as illegal immigration and undeclared workers to either make informed adjustments to the
population, or to be able to safely assume that the impact of under-coverage on comparability
is negligible. The issue of undeclared workers is closely related to underground businesses,
that is, those businesses that are active but which are not registered to avoid tax payments or
compliance with labour laws for example, an issue that affects both the numerator and
denominator.
Another approach is to use the population of working age, or of those people considered to be
economically active, if these populations can be defined consistently across countries.
However, even if a consistent definition is used, structural differences in national economies,
political or cultural differences (e.g. the rate of participation of women in economic activity, or
the ease with which a new business can be established
13
), can affect comparability. In such
cases however it might be preferable not to try to correct for such differences, as they,
arguably, form part of the phenomena to be observed, nevertheless, it is always helpful to be
aware of their potential impact when trying to interpret data from different countries.
For some specific purposes other sub-sets of the human population may be used, an example
of this is the use of the population of unemployed persons for analyses designed to illustrate
the extent to which unemployment encourages entrepreneurship. Great care is needed to
accurately interpret data using such sub-populations, as, in practice, only a proportion of new
businesses are actually started by people who were previously unemployed.
13
The World Bank and the International Finance Corporation, have financed work on an indicator
ranking countries on this topic, see:http://www.doingbusiness.org/EconomyRankings/Default.aspx?direction=asc&sort=2
23
3.5 Temporal Basis
If the denominator is based on the human population, point in time estimates are generally
used, i.e. the number of people on a specific date. Where it is based on a business
population, two variants have developed. The traditional approach, followed in most of the
data sets studied, is to use point in time business population data. This is consistent with
human demography, and allows a “stocks and flows” approach to business demography.
An alternative approach is to use the population of businesses that were considered to be in
scope at any point during a given reference period. This approach is favoured by Eurostat in
their business demography data collections, partly because it ties in with the approach used to
collect financial variables (e.g. turnover for a given period), and partly because it was thought
at one time to be easier for countries that did not have accurate birth dates for units in their
business registers.
It is clear that a live during period population will be larger than one on a point in time basis.
The extent of the difference will depend on various factors, but mainly on the length of the
period, and the degree of churn (i.e. entries plus exits) in the business population. As a result,
data compiled using a point in time population will not be directly comparable with those
based on a live during period approach.
One further comparability issue with the live during period approach is that a proportion of
business entries will be due to new businesses taking over the activities from businesses
recorded as exits
14
. Technically, many of these cases should be considered as the continuity
of a previous business, and should not be recorded as entries and exits. However, as most
data sources are based either directly or indirectly on registrations and de-registrations with
administrative or tax sources, it is unlikely that all such take-over cases are recorded as
business continuity, particularly for small businesses.
This will inevitably result in a certain amount of duplication in live during period populations, as
such businesses will appear in them at least twice. The extent of duplication will vary from
country to country and between sources, depending on the nature of the source and register
maintenance procedures. This, in turn, will affect the comparability of indicators based on live
during period populations.
There is, however, also a danger with the point in time approach, in that those short-lived
businesses discussed in the section on periodicity above, that enter and exit in the period
between two reference points may not be included in counts of start-ups, or the associated
business populations. This risk is theoretically removed using the live during period approach,
but in practice is only really solved for either approach by holding accurate birth and death
dates, the recording of some measure of activity (e.g. turnover), or frequent observations of
the whole population.
It is often possible to estimate a live during period population by adding the total number of
business entries during a period to the point in time estimate for the start of that period.
Similarly a point in time population can be estimated from live during period data, though
movements into and out of scope, and the degree of duplication in live during period
populations, can affect such estimates.
To illustrate this, point in time and live during period populations of businesses can be broken
down into a number of components, which can then be re-aggregated in different ways to give
different types of population estimates. The basic components are shown in Figure 3.4 below.
14
Around 15% in the French data shown in Figure 3.2.
24
Figure 3.4 - A Simple Model for Business Populations
Key: PA
t
=The population at the start of period t
PB
t
=The population at the end of period t
S
t
=businesses present in both populations (i.e. “survivors”)
L
t
=businesses that are in population PA
t
, but not PB
t
(i.e. “leavers”)
J
t
=businesses that are not in population PA
t
, but are in PB
t
(i.e. “joiners”)
J L
t
=businesses that are not present in PA
t
or PB
t
, but would be present in an
intermediate population (i.e. they join and leave within period t)
The population of businesses considered in scope at the start of the period (PA
t
), sometimes
referred to as the opening stock, can be defined as: PA
t
=S
t
+L
t
. Similarly the population at
the end of the period (PB
t
), or closing stock, can be defined as: PB
t
=S
t
+J
t
. Businesses in
the sub-set J L
t
do not appear in either population.
The population of businesses live at any time during period t (P
t
) can be defined as: P
t
=S
t
+
L
t
+J
t
+J L
t
, or by substitution as: P
t
=PA
t
+J
t
+J L
t
, or: P
t
=PB
t
+L
t
+J L
t
. Thus to convert
from a point in time to a live during period population, it is necessary to know, or have
reasonable estimates for J L
t
and either L
t
or J
t
. In practice, J L
t
is rarely available from
published data sources, and such businesses are usually ignored as they are not present in
PA
t
or PB
t
. The size, and hence the importance of J L
t
will depend on the length of period t. If t
is one month, it is relatively safe to assume that J L
t
is very small. If PA
t
and PB
t
are derived
from economic censuses with a five year interval, however, J L
t
will be much larger. These
relationships assume that there is no duplication within P
t
or between J
t
and L
t
.
It is possible to produce estimates for J L
t
, and to use these as a basis for converting
population data from live during period to point in time, or vice versa, if duplication is assumed
to be negligible. This approach is considered in Annex 3, which explores in much more detail
the relationships between different types of business population.
As stated above, a live during period approach will result in a higher denominator and lower
start-up rates. Typically for most OECD member countries start-up rates based on live during
period business populations are between 1% and 2% lower than those based on point in time
populations. Thus care must be taken in any comparisons that data collected on different
bases are not mixed. The point in time approach is conceptually simpler, and is less affected
by duplication issues, so is more likely to result in comparable data than the live during period
approach.
t
S
t
L
t
J
t
J L
t
PA
t
PB
t
25
3.6 Source
The main source for publicly available data on business start-ups is usually some sort of
register, either an administrative register maintained by a tax or regulatory agency, or a
statistical business register maintained by a national statistical institute. The main advantage
of this sort of source is usually comprehensive coverage of the population of interest, though
systematic biases may also be present due to the sort of coverage and threshold issues
identified below.
In theory, census data can be at least as good, and sometimes better, if they have less scope
restrictions, but the cost of running a census of businesses every year makes this approach
unrealistic for most countries. Data from less frequent censuses may still be of interest, but as
discussed in the section on periodicity above, they raise major comparability issues.
Survey data have also been used by some countries, most notably in the DOSME
15
project for
countries of Central and Eastern Europe. This approach can be useful when registers are not
sufficiently developed, and has the advantage of being able to collect more information on
entrepreneurship than is available from other sources, but it also suffers from the usual
constraints of survey errors and sample size limitations when detailed data breakdowns are
required.
The reliability of the source needs to be taken into account. This takes us back to the
components of the quality of statistical data, which have been well documented elsewhere
16
,
but it is clear that data from a comprehensive, frequently updated statistical business register
are likely to be more reliable than those from a small scale survey or study. The quality of the
data in the source clearly has an impact on most of the other factors of comparability identified
here, for example poor quality information on economic activity will have an impact on the
comparability of coverage.
It is also often the case that data from an official source (e.g. a national statistical institute) are
at least perceived to be more reliable than those from a commercial organisation. However,
this sort of generalisation is not always helpful, and a detailed understanding of the respective
methods used is necessary to make an informed judgement.
3.7 Units
The notion of a “business” is rather vague. Statistical institutes around the world have
historically tried to define the units for business statistics based on the sources of
administrative data available to them. The starting points are typically the unit that has some
sort of legal or tax obligation, often referred to as a “legal unit”, and the unit that corresponds
to a physical location from which a business operates, often referred to as a “local unit” or an
“establishment”.
Over time there have been attempts at the international standardisation of these units,
particularly in the European Union, where a regulation on statistical units
17
has gone part of
15
Demography Of Small and Medium-sized Enterprises – see:http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
16
For example:http://forum.europa.eu.int/Public/irc/dsis/qis/library?l=/public&vm=detailed&sb=Title
17
Council Regulation (EEC) No 696/93 of 15 March 1993 on the statistical units for the observation
and analysis of the production system in the Community (Official J ournal of the European
Communities No L 076, 30/03/1993, p. 1),http://europa.eu.int/eur-lex/lex/LexUriServ/LexUriServ.do?uri=CELEX:31993R0696:EN:HTML
26
the way towards harmonising the units used, and, has at least succeeded in harmonising the
terminology. Thus data from European Union countries will refer to enterprises, local units or
enterprise groups in a basically consistent way. There have been proposals to study the
demography of local units and enterprise groups, but, at least for now, business start-up data
for these countries are usually at the enterprise level
18
.
The enterprise is defined for European Union countries in the statistical units regulation as
“the smallest combination of legal units that is an organisational unit producing goods or
services, which benefits from a certain degree of autonomy in decision-making, especially for
the allocation of its current resources. An enterprise carries out one or more activities at one
or more locations. An enterprise may be a sole legal unit.”
Unfortunately it has been demonstrated that this definition is not always applied consistently
(e.g. in Herczog et al (1998)), particularly for more complex enterprises (e.g. those with more
than one legal unit) so it can not be assumed that data on units labelled as enterprises are
fully comparable in practice. This is largely due to differences in the legal, administrative and
tax frameworks across countries. A business that is organised as a single legal unit in one
country might prefer to organise itself into several legal units in another country to optimise
the way it interacts with government, employees and the market. For statistical purposes it is
necessary to recognise both as the same sort of entity, though creating the necessary
statistical structures through business profiling is a task that is difficult to automate, so is
therefore very expensive.
Outside the European Union there is a much greater freedom in terms of the choice (and
terminology) of units. In the United States, the establishment, which is closer to the European
local unit, is the main unit used for business statistics purposes. The term “firm” is used for an
aggregation of establishments under common control in a given geographic area, or sharing a
particular economic activity. Enterprises are defined as “business organizations consisting of
one or more domestic establishments that were specified under common ownership or
control”
19
, thus making them closer to the European concept of the enterprise group. Similar
terminology and definitions are used in Canada, though the term “business” is sometimes
used instead of “firm”. It is noted in Baldwin et al (2002), that “international studies must
recognize that the level at which a “firm” is defined varies across countries”.
The term enterprise is also used in most other OECD member countries, with slight variations
in the definition. It is defined in the System of National Accounts
20
, the key international
methodological framework for economic statistics, as “an institutional unit in its capacity as a
producer of goods and services; an enterprise may be a corporation, a quasi-corporation, a
non-profit institution, or an unincorporated enterprise.” In practice, the unit referred to as the
enterprise is often equivalent to, or very closely linked to, the national definition of the legal
unit. For a more detailed study of the different types and definitions of units used, see Choi
and Ward (2004).
Despite all of the above differences, it is likely that the vast majority (often at least 95%) of
business start-ups have a very simple structure, with just one site. This means that, in most
cases, all of the units above have a one to one relationship, and are in fact different views of
the same entity.
18
Several countries are, however considering the potential of local unit / establishment data to help
determine enterprise continuity.
19
US Census Bureau -http://www.census.gov/csd/susb/defterm.html
20
The System of National Accounts (1993) is promoted by the United Nations Statistics Division, and is
available via their web site:http://unstats.un.org/unsd/sna1993/introduction.asp
27
Unfortunately, it is not quite as simple as might appear from the above paragraph to compare
start-up rates for enterprises and establishments. There are two main complicating factors.
The first is that the total population of active enterprises will typically have higher proportions
of multi-site and complex businesses than the population of enterprise start-ups, thus if
enterprise data are to be converted to an establishment basis, the denominator needs to be
increased to take account of this. How much of an increase is likely to depend on a number of
factors including fiscal policy and the economic size and geography of the country. For the
United Kingdom this would reduce start-up rates by up to 2%. The second factor works in the
opposite direction, because a proportion of establishment start-ups will be new sites of
existing enterprises
21
. These would need to be added to the numerator, increasing the start-up
rate by up to 3%. The net result is therefore likely to be that establishment start-up rates are
slightly higher than those for enterprises.
3.8 Coverage
The coverage of data on new businesses and the business population depends heavily on the
source. In most cases this is a national statistical business register. If this register does not
include all legal forms or all economic activities, it logically follows that the data on new
businesses will have at least the same restrictions. Sometimes, even if the register does
include certain categories, there may be reasons for excluding them from counts of new
businesses. These reasons may be linked to quality concerns, the policy of the statistical
institute, customer requirements, or just tradition.
Categories most frequently considered to be out of scope in terms of economic activity are
agriculture, forestry, fishing and public administration. Units with the legal forms of central or
local government are also often excluded. The existence of a number of different
classifications of economic activity and legal form further complicates matters, as specific
categories of units may be treated differently according to the classification system used.
Fortunately the examples of frequently excluded categories above are also areas where
international classification systems are relatively well harmonised.
The treatment of businesses that move into and out of scope needs to be determined and
specified. The Eurostat approach attempts to exclude entries due solely to changes in
economic activity or other characteristics from data on pure births, whereas this distinction is
not necessarily made in other data sets. The comprehensiveness of the source obviously has
a major bearing on the ease of identifying such cases.
As is the case for units, the greatest degree of harmonisation in coverage, at least in theory,
exists between the Member States of the European Union, mainly due to the minimum
requirements set out in a regulation on statistical business registers
22
, which is currently being
revised with the aim of extending and further harmonising the scope of these registers.
Despite this, the data on business demography currently published by Eurostat has one of the
most restricted scopes of the data sets studied. Economic activities such as health, education
and personal services, and all non-market legal forms are excluded from most analyses.
These exclusions are largely driven by data quality concerns, which suggest that the actual
21
Between 19% and 28% depending on the year based on comparisons of data on births at original
locations (new firms), and secondary locations (new sites of existing firms) from the US Small Business
Administration - Seehttp://www.sba.gov/advo/research/data_uspdf.xls, worksheets dyn_00 to dyn_02.
22
Council Regulation (EEC) No 2186/93 of 22 J uly 1993 on Community co-ordination in drawing up
business registers for statistical purposes -http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/2186-93en.htm
28
level of harmonisation of business registers is still somewhat below that required by the
regulation.
There is a tendency when comparing data that differ in scope to look for the lowest common
denominator, i.e. the core set of data that are available for all sources. This can, however, be
problematic in some cases. For example, data from the United States typically exclude railway
operators, a category that is not always readily and separately identifiable in data from other
countries. In cases like this, alternative approaches could include modelling for the missing
categories, or simply ignoring minor scope exclusions in some sources if their impact is
considered to be trivial.
Where a population of businesses is used as the denominator for start-up rates, it is obviously
better to try to define this in as close a way as possible to that used for new businesses, i.e.
the coverage should be the same for both in terms of economic activity, legal form and other
criteria. Any differences will increase the possibility of other changes having an impact on the
birth rate
23
.
By definition, administrative and statistical registers, in all countries, will exclude businesses
operating exclusively in the non-observed or informal economy. They will also understate size
variables for businesses operating partially in this way. For developed countries, the economic
importance of such businesses is generally not thought to be significant, particularly in terms
of total economic activity. These businesses may, however, be of interest in the context of
measuring entrepreneurship or determining small and medium-sized enterprise (SME) policy.
It is therefore useful to have reasonably reliable estimates of the impact of such businesses,
perhaps from comparing labour force and business employment survey statistics or consumer
expenditure and declared business income.
Variations in geographical coverage may also affect within country comparability. This is most
likely to be a problem when there are autonomous or detached territories that may be
included in one source but not another, or when there has been a boundary change. The most
notable examples found in the course of this project are German data that exclude the former
East Germany, and United Kingdom data that exclude Scotland and Northern Ireland. In both
cases, start up rates for the included areas are, according to other sources, different to those
for the excluded areas. Thus these sources are not strictly comparable with other national
sources, and are not really suitable for international comparisons, as they are not fully
representative of the national situation.
3.9 Thresholds
There are no clearly defined, internationally agreed minimum size criteria for business activity.
Most data from the United States include only businesses with employees, whereas certain
J apanese data, and some international comparisons,
24
include only corporate businesses.
These sources therefore contain only a limited proportion of smaller businesses.
The European Union requires that all businesses with a labour input of at least one person
half-time are included in statistical business registers, and recommends covering smaller
businesses if possible. Some countries require all businesses to be registered regardless of
23
For example, if the population used for the denominator includes all legal forms, but the data on new
businesses used for the numerator exclude central and local government units. A re-organisation of
local government that creates many new units in that sector would have the effect of increasing the
denominator but not the numerator, thus artificially reducing the birth rate.
24
E.g. Klapper et al, (2004)
29
size, but even these are unlikely to record very low levels of business activity such as
individuals who occasionally sell second-hand or surplus goods to neighbours, via markets, or
through internet auction sites.
Some of the smallest “businesses”, particularly those with a labour input of less than one
person half-time, may be registered, but of little interest statistically. Hobby businesses
operated for reasons other than profit maximisation, and the provision of goods or services for
a few hours per week to supplement a main income are examples of this. In Volfinger (2004)
the statistical relevance of Hungarian enterprises with a turnover of less than one thousand
Euros is questioned. These accounted for 9% of active enterprises in 2002.
Thresholds can be helpful in terms of excluding such types of businesses, if they can be
applied uniformly across countries. An alternative can be to ensure that data have a strong
size dimension with classes based on quality and comparability criteria, so that certain
classes can be flagged as less comparable than others. Similarly, thresholds can provide a
route to exclude “pseudo-enterprises”, sometimes also referred to as “false self-employed”
where a person acts as an employee of an enterprise, in that they effectively work for that
enterprise every day over a long period of time, but for legal or tax purposes he or she is
technically self-employed. These issues are considered in more detail in Vale and Powell
(2002) and Vale (2005(b)), and their impact on European Union data in Brandt (2004).
In practice, thresholds are likely to be determined by the administrative sources that supply
data to statistical business registers. In many cases, data from sales or value added tax
registrations are used, with thresholds varying from zero up to GBP 60,000 in the case of the
United Kingdom. Where higher thresholds exist, data are often supplemented from other
sources to mitigate the impact, so it is often impossible to define the actual threshold applying
to a particular data set in terms of a single variable.
Particular care should be taken with thresholds related to sales or value added, as it is quite
possible in certain economic activities, e.g. software development, for a business to have
employees but no sales for a year or more, while it is developing products. An additional
complication is that monetary based thresholds are affected by differences in inflation and
fiscal policy at the national level, both of which can be expected to affect comparisons of birth
rates across countries and over time.
Thresholds relating to labour input are often more appropriate, but again it is important to
know how it is measured, e.g. in terms of wage-related costs, head counts of employees, or
full-time equivalents, as this could also have an impact, albeit probably small, on
comparability.
The quality of size variables can have a considerable impact on comparability when
thresholds are used. Unfortunately the quality of data is often lowest for relatively small and
new businesses, the categories that are often of the most interest. The methods of allocating
size (and other) variables in statistical systems, in the absence of full information on certain
businesses, can vary considerably. Some attempts to standardise these processes have been
introduced in the Eurostat methodology, including the use of turnover per head ratios to
estimate missing size variables.
The impact of thresholds varies depending on the use of data, and is usually much lower
when measuring economic or financial variables than for those based on counts of
businesses. It may also be the case that data subject to different thresholds can display the
same trends, even if those trends are less marked and the levels are different. This is
illustrated by the graph in Figure 3.5, showing two sources of data on business start-up rates
for the United Kingdom.
30
Figure 3.5 – UK Business Birth Rates – A Comparison of Data Sources
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
1995 1996 1997 1998 1999 2000 2001 2002 2003
Value-added tax registrations - Source: Small Business Service
New Businesses - Source: Barclays
In Figure 3.5, the source with the lower threshold, Barclays, shows a higher level and greater
volatility, as might be expected given the typically more dynamic nature of the smallest size
classes. The Small Business Service data are based on value-added tax registrations with a
threshold that has remained more or less constant in real terms at around GBP 60,000 in
2005 prices over this period. They clearly resemble a smoothed version of the Barclays data
series, albeit at a lower level. It would therefore appear to be possible to model one series
from the other. Whilst the birth rates themselves are not directly comparable, the underlying
data might be considered sufficiently comparable for some purposes. Also, the fact that both
sources show similar trends over time helps to validate the quality of the data sets.
31
4. Methods to Improve Comparability
This section is concerned with the extent to which existing data can be made more
comparable by performing various transformations. It follows three worked examples, the first
comparing data from the United States and Eurostat, the second comparing the Eurostat data
with experimental estimates from Australia, and the third comparing national data from France
and Germany with Eurostat data for other European countries. The transformations are made
on the basis of information available in the existing metadata, in some cases this has been
supplemented through contacts with those responsible for the source.
The approach of transforming existing data sets to try to improve comparability is not ideal,
and is unlikely to result in perfectly comparable data. These examples show that it can,
however, lead to some improvements in comparability. It should be seen as a short-term
measure, whilst waiting for the results of longer-term improvements such as the international
implementation of methodological standards.
4.1 Example 1 – United States and Europe
An obvious first step when looking at data from different countries, each with several sources,
is to use the comparability factors in Section 3 above to determine which sources give the
best trade-off in terms of the level of comparability already present and data quality. Thus, for
the first example, a comparison of data from the United States and Europe, it would seem
logical to use the Firm Size dataset from the US Small Business Administration, and the data
from the Eurostat business demography project. These sources use a similar unit, and both
define populations on a “live during period” basis. Figure 4.1 shows a basic comparison of the
raw data.
Figure 4.1 – A Comparison of US and Eurostat Business Start-up Rates
0%
2%
4%
6%
8%
10%
12%
14%
16%
U
n
i
t
e
d
S
t
a
t
e
s
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
N
e
t
h
e
r
l
a
n
d
s
P
o
r
t
u
g
a
l
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
1998
1999
2000
2001
Sources: United States – Firm Size Data – Small Business Administration
EU Mean – Mean start-up rate for the European countries shown
Other countries – Eurostat (The full Eurostat data set includes several other countries,
but only those for which data are available for at least three of the above years are
shown)
32
The data seem to indicate very little difference between start-up rates in the United States and
the European countries shown. However, there remain several methodological differences
between the US and European data. Perhaps the most important is that the US data only
include employer firms, i.e. businesses with at least one employee. The Eurostat database
includes a breakdown by size class, with a category for zero employee enterprises.
Subtracting this category from both the births and the population of active enterprises
therefore gives an estimate of start-up rates for employer businesses, as shown in Figure 4.2.
The data in Figure 4.2 show considerably more variation within Europe, and result in much
lower EU mean rates. Unfortunately, this comparison represents a backwards step in
comparability compared to Figure 4.1 for most of the countries shown. This is because, in
trying to correct for thresholds, new problems of coverage have been introduced. The start-up
rates for the European countries now only include those businesses that have employees
from the start. They do not include businesses that start with no employees, and then take on
employees as they expand. These businesses are, however, captured in the US data. This
also explains some of the increased variability in the European data, as, for example,
coverage of non-employer enterprises is much higher in Italy than in the United Kingdom or
Luxembourg, due to higher size-thresholds in the data sources for the latter two countries.
This makes it more likely that new enterprises will be identified before they take on employees
in Italy, and will thus be missing from the start-up rates shown in Figure 4.2.
Figure 4.2 – Start-up Rates for Employer Businesses – A Backwards Step?
0%
2%
4%
6%
8%
10%
12%
14%
U
n
i
t
e
d
S
t
a
t
e
s
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
N
e
t
h
e
r
l
a
n
d
s
P
o
r
t
u
g
a
l
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
1998
1999
2000
2001
Sources – as for Figure 4.1 above
Thus, the conclusion from Figure 4.2 is that merely removing the non-employer business
births from the Eurostat data causes extra distortions, and certainly does not improve
comparability. This clearly shows the danger of attempting to improve comparability of existing
data analytically without a proper understanding of the complex interactions of comparability
factors. However, in the context of this worked example, it is still possible to see the
adjustment made in Figure 4.2 as a step towards more comparable data, because, even
though it reduces comparability, it also opens up several possibilities to improve comparability
above the initial level in Figure 4.1.
One such possibility to improve comparability would be to determine the proportion of start-
ups in the US that previously existed as non-employer businesses, and remove them from the
33
US start-up rates. Data are not currently available to make this adjustment, but may result
from preliminary work to link employer and non-employer universes reported in Davis et al
(2005).
Alternatively, and perhaps as an interim measure, a study of cohorts of non-employer births
could be carried out in several European countries, to see how many subsequently became
employers, and how long it was before they made this transition. The results could then be
used to model the missing data, raise the European estimates, and hence to improve
comparability with the US data. The proportion of businesses making the transition, and the
timing, are likely to vary between countries and over time, so this approach is probably most
suitable for countries that want to make one-off comparisons with the US, rather than for wider
cross-country comparisons over time.
A better solution, however, given the considerable variability of European data in Figure 4.2,
would be to define the business population for those countries so that it only included
employers, and measure entries into that population (as recommended in the OECD Business
Demography Framework). The Eurostat data do not currently support this approach, but it
would be relatively easy to adapt the current Eurostat methodology to produce the necessary
figures.
It may still be necessary to interpret any resulting figures with care, as they could be affected
by variations in the propensity to incorporate between countries. Most new businesses start
as either a corporation or a sole-proprietorship. In the case of a corporation, the entrepreneur
is normally considered to be an employee, whereas in the later, he or she is not. Thus the
choice of legal form, which could be affected by national fiscal and administrative burden
considerations, can determine whether a start-up is included in the population of employers or
not. There are very few data on this subject at present, though the overall impact on start-up
rates is thought to be quite small. The methodologically purest long-term solution, therefore,
would be to define a lower threshold in terms of total labour input (e.g. 0.5 person), which
would be independent of issues of legal form. Unfortunately this is not really feasible for the
main indicator on business start-up rates, as it would require major (and hence expensive)
changes in several countries, particularly the United States.
Another methodological difference between the US and European data is purity. The Eurostat
methodology requires extensive matching to determine which start-ups are pure births,
whereas the metadata for this US source (Armington (1998)) make clear that no attempts are
made to track the survival of individual firms. The US data will therefore include an unknown
proportion of start-ups that are not pure births in the European sense. The proportion of start-
ups that are not pure births varies from source to source, as the propensity to re-register will
be determined by legal or other requirements that are usually source specific. In Europe this
proportion is usually around 20%, though French national data suggest figures of between 30
and 40%
25
. This proportion typically increases for larger businesses. Applying this to the US
data in Figure 4.2 would reduce start-up rates to around 6-8%, well below those of the United
Kingdom and Luxembourg, the countries for which the difference in thresholds compared to
the US data is likely to be least significant.
25
See Figure 3.2 in Section 3.1 above.
34
4.2 Example 2 – Australia and Europe
The second example considers the comparability of data from Australia and Eurostat. In this
case, the Australian approach is very similar to the Eurostat methodology in terms of
thresholds and some elements of purity. Figure 4.3 shows a basic comparison of the raw
data.
Figure 4.3 – A Comparison of Australian and Eurostat Business Start-up Rates
0%
2%
4%
6%
8%
10%
12%
14%
16%
A
u
s
t
r
a
l
i
a
E
U
M
e
a
n
S
p
a
i
n
I
t
a
l
y
L
i
t
h
u
a
n
i
a
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
S
l
o
v
a
k
i
a
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
2001
2002
Sources: Australia – Experimental Estimates, Entries and Exits of Business Entities – Australian
Bureau of Statistics.
EU Mean – Mean start-up rate for the European countries shown.
Other countries – Eurostat (The full Eurostat data set includes several other countries, but
only those with data available for both years, and no known coverage issues, are shown).
Figure 4.3 appears to show that Australian start-up rates are almost 1% above the mean for
the European countries shown. However, for a true comparison, it is necessary to make
adjustments to compensate for several methodological differences.
The main difference is that the Australian population data used as the denominator for these
start-up rates are on a point in time basis, whereas those from Eurostat are “live during
period”. The Australian point in time population is defined as the population from the previous
observation, adjusted for reactivations, plus entries in the previous period, minus exits in the
previous period. This relationship is shown in Figure 4.4, taken from the Australian
publication.
Figure 4.4 - The Relationship between Populations, Entries and Exits in the Australian
Data
35
Entries and exits are measured on a monthly basis, so any short-lived businesses that enter
and then exit between the two reference points, are included in both the entry and exit figures,
and thus cancel themselves out in this model. This aids comparison with the Eurostat data,
which also include such short-lived businesses.
Building on the approach introduced in Section 3.5 above (and developed further in Annex 3),
the components of the population can be shown as in Figure 4.5.
Figure 4.5 - The Components of the Business Population
PA
t
=The opening population for period t
PB
t
=The closing population for period t
B
t
=Births in period t that survive into t+1
D
t
=Deaths in period t that were live in t-1
BD
t
=Birth and Death within period t
The Australian relationship above can therefore be re-written as PB
t
=PA
t
+(B
t
+BD
t
) – (D
t
+
BD
t
). The Eurostat population, defined on a “live during period” basis, can be expressed as
PA
t
+B
t
+BD
t
(i.e. all businesses live at the start of the period plus all births during the
period). Thus adding the birth data to the opening population data for Australia will give a
good estimate of a “live during period” population
26
.
Table 4.1 – Converting Australian Start-up Data to a Live During Period Basis
Period Opening
Population
Entries Start-up
Rate
Live During Period
Population
Live During Period
Start-up Rate
2001-2 2,935,700 334,266 11.39% 3,269,966 10.22%
2002-3 2,941,666 329,907 11.21% 3,271,573 10.08%
Source: Authors calculations using data from the Australian Bureau of Statistics.
Other differences between the Australian and European data are that the Australian data
cover all economic activities, and use a J uly to J uly reference period, whereas the European
data have a more restrictive coverage, and are on a calendar year basis. In terms of the
coverage by economic activity, it is possible to use breakdowns by industry in the Australian
publication to get a close match to the European coverage (sections C-K of the International
Standard Industrial Classification (ISIC Rev. 2)).
26
Note – this approach does not work in cases where births and deaths are not measured on a regular
basis during the period, as, although it becomes easier to measure B
t
, it becomes much more difficult to
quantify BD
t
, and hence total births in the period.
t
D
t
B
t
BD
t
PA
t
PB
t
36
Table 4.2 – Adjusting for Coverage of Economic Activity
Period ISIC
Sections
Opening
Population
Entries Start-up
Rate
Live During Period
Population
Live During Period
Start-up Rate
2001-2 C-K 2,246,229 267,601 11.91% 2,513,830 10.65%
Other 689,471 66,665 9.67% 756,136 8.82%
Total 2,935,700 334,266 11.39% 3,269,966 10.22%
2002-3 C-K 2,299,104 248,833 10.82% 2,547,937 9.77%
Other 642,562 81,074 12.62% 723,636 11.20%
Total 2,941,666 329,907 11.21% 3,271,573 10.08%
Source: Authors calculations using data from the Australian Bureau of Statistics.
The final step is to adjust for the difference in time periods. This requires certain assumptions,
which risk introducing noise into the data, but should still result in a net improvement to
comparability in this case. The first assumption is that the population of businesses (for ISIC
sections C to K) at 1 J anuary 2002 is exactly halfway between the 1 J uly populations for 2001
and 2002. This gives a value of 2,272,667. The second assumption is that the entries are
following a linear trend, thus all things being equal, the number of entries on 1 J anuary 2002,
the mid-point of the period, should be equal to the annual total for 2001-2 divided by 365, i.e.
733.153. Similarly the number of births on 1 J anuary 2003 would be 681.734, and the total for
2002 would be ((733.153 +681.734) / 2) x 365, i.e. 258,217. These figures give a “live during
period” population of 2,530,884, and hence a “live during period” start-up rate of 10.20%.
Based on the available metadata, differences due to timing, type of population and source are
likely to be negligible, as well as not being easy to quantify. There still, however, remain
questions over purity and units. In terms of purity, the Australian data seek to identify where a
re-registration is really the continuation of an existing business, but having done this, all
remaining new businesses are treated as genuine entries. In the Eurostat methodology, a
further (probably quite small) proportion of these would be removed. These are new
businesses that do not meet the requirement of being a new combination of factors of
production (e.g. a new business formed by splitting off part of the activity of an existing
business).
Regarding units, the Australian data are based on tax registrations (i.e. legal units) rather than
enterprises. This is discussed in ABS (2005), where data for large and complex business
entities (for J une 2004), show that 67,000 tax units have been combined to form 30,000 “type
of activity units”. On the assumption that the European data correspond to the Eurostat
definition of the enterprise, it would therefore be necessary to reduce the Australian
population by around 37,000 businesses, and the entries by a rather smaller proportion (as
relatively few new businesses tend to be complex from the outset). However, as discussed in
Section 3.7 above, the enterprise definition is not yet applied fully and consistently in all
European countries, therefore the value of any adjustment to the Australian data is doubtful.
It is therefore perhaps easiest to assume that the differences in purity, which would reduce the
numerator of the Australian data, and units, which would reduce the denominator, are both
relatively small, and would largely cancel each other out.
The combined impact of the adjustments made to the Australian data is shown in Figure 4.6.
As we now only have one year for which data are reasonably comparable, it would be
dangerous to draw too many conclusions, though it is interesting to see that the start-up rate
estimate for Australia is now very slightly below the mean value for the European countries
shown (10.20% and 10.36% respectively).
37
Figure 4.6 – More Comparable Start-up Rates for Europe and Australia (2002 Data)
0%
2%
4%
6%
8%
10%
12%
14%
16%
A
u
s
t
r
a
l
i
a
E
U
M
e
a
n
S
p
a
i
n
I
t
a
l
y
L
i
t
h
u
a
n
i
a
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
S
l
o
v
a
k
i
a
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
4.3 Example 3 – France and Germany Compared to Other European Countries
Data on business start-up rates for France and Germany have not yet been published by
Eurostat, though both countries are now taking an active part in the Eurostat business
demography project, and will be supplying data for future publications. This example looks at
how existing national data for France and Germany could be compared to those from
Eurostat. Figure 4.7 shows a comparison of the raw data, which does seem to show that there
are comparability issues, particularly for Germany.
Figure 4.7 – Comparing French and German Start-up Rates with Eurostat Data for Other
European Countries
0%
5%
10%
15%
20%
F
r
a
n
c
e
G
e
r
m
a
n
y
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
N
e
t
h
e
r
l
a
n
d
s
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
2001
2002
Sources: France – Créations d’Entreprises, INSEE
Germany - Start-ups and Liquidations in Germany, Institut für Mittelstandsforschung, Bonn
EU Mean – Mean start-up rate for the European countries shown.
Other countries – Eurostat (The full Eurostat data set includes several other countries, but
only OECD members with no known coverage issues, are shown).
38
The French data are produced by the national statistical institute (INSEE). They are checked
for purity (removing around one third of all entries), and measure entries to the French
business register (SIRENE), regardless of the length of survival. Thus they can be considered
as comparable to the Eurostat data in terms of purity, timing, periodicity, type of population
and source. The main differences concern coverage, and the temporal basis of the population
of businesses used as the denominator. Differences in units are likely to be negligible, as are
differences in thresholds, except when compared to the United Kingdom and, to a lesser
extent, Luxembourg.
The coverage of the French data in terms of economic activities is wider than for the Eurostat
data. Some broad economic activity breakdowns are available via the INSEE web site
(www.insee.fr), which can be re-aggregated to match the coverage of the Eurostat data. If this
is done, the start-up rates increase by around 0.4%, assuming that the ratio of pure births to
other entries is constant across activities.
The French data use a point in time (1 J anuary) business population as the denominator. To
convert this to an estimate of the live during period population, it is necessary to add total
entries (i.e. pure births and other entries) to this population. This can be done on the same
basis as for the Australian data in Section 4.2. The results of these two conversions are
shown in Table 4.3
Table 4.3 – Adjusting French Data for Temporal Basis and Coverage
Period ISIC
Sections
Opening
Population
Entries Pure
Births
Birth
Rate
Live During
Period
Population
Live During Period
Birth Rate
2001 C-K 1,927,602 226,019 147,364 7.64% 2,153,621 6.84%
Other 490,348 42,600 27,775 5.66% 532,948 5.21%
Total 2,417,950 268,619 175,140 7.24% 2,686,569 6.52%
2002 C-K 1,964,295 224,722 147,642 7.52% 2,189,017 6.74%
Other 504,491 43,737 28,735 5.70% 548,228 5.24%
Total 2,468,786 268,459 176,378 7.14% 2,737,245 6.44%
Source: Authors calculations using data from the INSEE web site.
The French data, however, also identify a proportion of entries as taking over the activities of
existing enterprises (referred to as “reprises”). These account for around 15% of entries
(15.36% in 2001 and 14.85% in 2002), and will duplicate business activity recorded in the
opening population (or possibly in other entries). Thus, in accordance with principles of
business continuity, and to avoid artificially inflating the live during period population with
duplicates, they should be removed from that population. The result of this adjustment is
shown in Table 4.4.
Table 4.4 – Removing Duplication in the Live During Period Population
Period ISIC
Sections
Opening
Population
Corrected
Entries
Pure
Births
Corrected Live During
Period Population
Corrected Live During
Period Birth Rate
2001 C-K 1,927,602 191,302 147,364 2,118,904 6.95%
2002 C-K 1,964,295 191,351 147,642 2,155,646 6.85%
Source: Authors calculations using data from the INSEE web site.
Turning to the German data, these are based on notifications of new businesses for turnover
tax purposes, supplied via the statistical business register. The register data are adjusted by
the Institut für Mittelstandsforschung (IfM), to remove new sites of existing businesses,
registrations purely for tax or administrative purposes that do not result in new business
39
activity, and registrations for activities carried out as a second job by the entrepreneur.
Business registrations due to the movement of a legal unit from one district to another or a
change in ownership or legal form are also removed. These adjustments result in
approximately 62% of notifications being considered as real births (compared to the French
figure of around 65%). Thus the data can be considered to have been corrected for purity.
In terms of timing, periodicity, type of population and source, the German data can be
considered as comparable to those from Eurostat.
The German data use a point in time population, which can be converted to a live during
period basis in a similar way to the French data above. However, the German data do not
separately identify the different categories of entries that are not real births, thus the data in
Table 4.5 below add all registrations to the opening population, which may overstate the
population, and slightly under-estimate the live during period birth rate.
Table 4.5 – Adjusting German Data for Temporal Basis
Period Opening
Population
Entries Pure
Births
Birth
Rate
Live During Period
Population
Live During Period
Birth Rate
2001 2,920,293 728,978 454,700 15.57% 3,649,271 12.46%
2002 2,926,570 723,333 451,800 15.44% 3,649,903 12.38%
Source: Authors calculations using data from IfM and the German Federal Statistical Office.
The units used are effectively the sub-set of legal units that are considered to be economically
relevant. This is unlikely to have any real impact on the number of entries, but may mean that
the population of businesses is slightly overstated, again leading to a very slight under-
estimation of start-up rates.
The data cover all economic activities except the “liberal professions”
27
, most health services
and some insurance services that are not subject to turnover tax. This is still a slightly wider
coverage than the Eurostat data. Detailed breakdowns of economic activity that would allow
more exact comparisons are not currently available, though based on the evidence for France
above, the impact is likely to be small.
The population of businesses is subject to a threshold (16,617 Euros during this period),
which means that some smaller businesses are excluded. Smaller businesses typically have
higher entry and exit rates, thus the impact of this threshold is likely to be a slight under-
estimation of start-up rates when compared to all other European countries except the United
Kingdom (where the threshold was around 90,000 Euros).
Although it is not possible to quantify the impact of these factors accurately, it is clear that the
net effect will be a slight under-estimation of business start-up rates. The rates calculated in
Table 4.5 should therefore be seen as minimum estimates, but it is unlikely that the real
values are more than 0.5% higher.
The revised estimates of business start-up rates for France and Germany are shown in Figure
4.8. These can now be seen as rather more comparable with the Eurostat data. The German
rates are still above the mean of the Eurostat rates, but are no longer the highest. This could
be expected, as the German data include the former East Germany, which may have an
upward influence on the national rate, as start-up rates published by Eurostat for the former
communist countries of Eastern and Central Europe (e.g. Hungary) are generally higher than
27
The liberal professions can generally be defined as occupations requiring special training in the arts
or sciences. These include lawyers, notaries, accountants, architects, engineers and pharmacists.
40
those for Western European countries. The adjusted data for France are slightly lower than
the raw data, but still not the lowest for the countries present.
Figure 4.8 – More Comparable Start-up Rates for European Countries
0%
5%
10%
15%
20%
F
r
a
n
c
e
G
e
r
m
a
n
y
E
U
M
e
a
n
D
e
n
m
a
r
k
S
p
a
i
n
I
t
a
l
y
L
u
x
e
m
b
o
u
r
g
H
u
n
g
a
r
y
N
e
t
h
e
r
l
a
n
d
s
F
i
n
l
a
n
d
S
w
e
d
e
n
U
n
i
t
e
d
K
i
n
g
d
o
m
N
o
r
w
a
y
2001
2002
4.4 Summary
The three examples above show that although it is not always possible to be precise, it is
clear that adjustments to compensate for differences in specific factors can sometimes help to
improve comparability. It is also clear that in compensating for one factor, it is possible to
affect others, and to introduce noise into the data, thus adjustments have to be made with
care, based on a detailed understanding of the data sources and methods.
Having said that, adjustments based on estimates of the impact of a specific factor of
comparability can still help to determine whether differences in start-up rates are likely to be
significant or not. This is no substitute for having real, comparable data, but, in the first
example above, at least it should caution an analyst against making statements that US start-
up rates are definitely higher than those in the European Union.
41
5. A Harmonised Methodological Framework and Start-up Indicators?
The discussions on factors of comparability in Section 3, and possible ways to improve the
comparability of existing data in Section 4, lead to the conclusion that the best way to get
really comparable data across countries is to harmonise as far as possible the underlying
methodology. Thus what is needed is a standard methodological framework that can be
applied in all countries, and which leads to a set of indicators of business start-ups that can be
used with confidence for cross-country comparisons. This section discusses how this might be
achieved.
5.1 Towards a Harmonised Methodological Framework
The idea of developing a harmonised methodological framework for indicators of business
demography and dynamics is not new. It has been attempted with varying degrees of success
in some of the projects outlined in Section 2, though often either the focus has been on a
limited group of countries that already share certain characteristics (as in the DOSME project),
or a common legal framework for statistics (as in the Eurostat project), or the methodology
was not detailed enough to generate real comparability.
The OECD is in the process of developing a new framework, taking into account what has
worked and not worked in the past, as well as a more detailed knowledge and understanding
of the methodological issues than in many previous projects. As business start-up rates are
an important component of business dynamics, this report will also feed into the new OECD
framework. The approach of using factors of comparability introduced in this report is being
broadened to cover business demography as a whole, and to inform the decisions on the
preferred methodology.
5.2 Different Types of Indicators
Within this harmonised methodological framework it would be possible to envisage several
different types of business start-up indicators. The traditional version, showing the number of
new businesses as a percentage of the population of businesses, is clearly the key indicator
for business start-ups, but, as this report shows, it is also not a particularly easy indicator to
define in a way that results in fully comparable data across countries. A range of other
indicators have been proposed over the years, each of these has certain merits, but none
seem to offer a full solution to the problem of international comparability.
Several studies have argued that it is better to measure start-ups by only including those
businesses that survive for a certain length of time. For the OECD Firm-level Data project, that
period was at least one year, whereas the authors of Baldwin et al (2002) recommend using
periods of up to five years. They show that start-up rates from different sources in Canada
vary more in the short-run than in the long-run, and hence recommend that a longer-run view
should be used for international comparisons.
The Eurostat business demography project goes somewhat in the opposite direction, and
seeks to include all start-ups, no matter how short-lived they are, and attempts to tackle
comparability issues through the harmonisation of sources and methods. Despite this, there is
some evidence from the limited Eurostat data currently available that would seem to support
the view of Baldwin and colleagues. If birth rates are plotted against two-year survival rates for
the European countries for which data are available, there appears to be a fairly weak
negative correlation, indicating a limited increase in convergence over time. Figure 5.1 shows
data for births in 2000, where the correlation coefficient is -0.47.
42
Figure 5.1 – Birth Rate v. 2 Year Survival Rate for Births in 2000
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
4% 6% 8% 10% 12% 14% 16% 18%
Bi rth Rate
2
Y
e
a
r
S
u
r
v
i
v
a
l
R
a
t
e
Source: Constructed from Eurostat data
There are, however, a few problems with using long-run entry rates. The first is that few policy
makers would be prepared to wait for five years for data. The second is that in countries with
genuinely dynamic business populations perhaps fuelled by very low entry and exit costs and
a strong entrepreneurial culture, it is to be expected that short-run start-up rates would be
higher and survival rates lower, reflecting an increased degree of experimentation on the part
of entrepreneurs. Long-run entry rates are less useful for identifying the extent to which this
particularly interesting phenomenon varies between countries. The real challenge is to
separate this sort of genuine variation between countries from the noise in the data due to
methodological differences.
The third problem with using long-run entry rates is that they do not affect all of the factors of
comparability equally. They should resolve most timing issues and can help to smooth the
effects of differences in periodicity, thresholds, and possibly units, though it is difficult to see
how they will have much impact on the other factors. There is even a risk that they could
actually aggravate the impact of different temporal bases for the population in that data
compiled on a live during period basis are likely to have lower birth rates (due to the higher
population), and lower survival rates (due to the inclusion of more short-lived units) than data
compiled on a point in time basis, thus the long-run birth rates could potentially be more
divergent than those for the short-run.
Despite these potential problems, there is still a role for long-run entry rates in conjunction
with short-run data. If they are sufficiently comparable between countries, they can give
another view on business dynamics, thus helping to give a better overall picture of the real
differences between countries.
A different approach to business start-up indicators is to consider the impact of new
businesses in terms of employment creation. This is of great interest to policy makers and
researchers concerned with the impact of encouraging entrepreneurship. As noted in Baldwin
et al (2002), employment based measures are less influenced by differences in thresholds,
but much more sensitive to purity. This is because thresholds tend to affect the smallest
businesses, excluding those with no or very few paid employees, whereas new businesses
that are not pure births tend to have more employees. A new business created by the merger
of two large corporations, bringing together thousands of employees, could swamp data on
employment creation.
43
Given the interest in employment creation, however, it is still useful to have a measure of the
impact of business start-ups, as long as this measure is sufficiently reliable and comparable.
Thus the Eurostat approach of trying first to obtain harmonised data on births, then
complementing these with data on employment creation, seems worth pursuing.
In a few cases, business start-up rates have been calculated using a human population as the
denominator. This approach relies on a suitably harmonised definition of the population used,
and can be affected by various social and cultural factors as discussed in Section 3.4 above,
but produces an indicator that is perhaps better focussed on entrepreneurship propensity.
Most of the data sets studied in this report are at the level of the business, firm or enterprise,
which despite their definitional differences can generally be seen as the same unit as far as
business start-ups are concerned, given that the overwhelming majority of these have very
simple structures. The exceptions are the establishment level data available for the United
States, J apan and a few other countries.
Although establishment level data cause comparability problems, they also have clear
benefits in terms of studying business dynamics at the local or regional level, a topic which is
generating interest in various countries, notably the United States and the United Kingdom.
Thus indicators at the level of the establishment (or local unit in Europe) are worth
consideration, particularly if it is possible to determine whether new establishments are due to
pure enterprise births, other enterprise creations, or the opening of a new site by an existing
enterprise; although measures based on establishments do provide other complications
28
.
5.3 Proposed Indicators
Based on the findings of this report, it is clear that a single indicator of business start-up rates
is unlikely to meet all requirements, therefore a system based on a key indicator, supported by
a range of complementary indicators is proposed.
• The Key Start-up Indicator
The key indicator of business start-up rates should try to meet potentially conflicting
requirements. Firstly it should be meaningful and easy to interpret for non-specialists, thus it
should be based on concepts and methodologies that are as simple as possible. Secondly it
should be designed to maximise international comparability. This second requirement could
lead to a purist view that the indicator should be designed without any reference to existing
data sources, or a pragmatic view that it should be built around the data that are currently
available. The purist view is likely to delay the availability of comparable data, whereas the
pragmatic view would not necessarily lead to optimal methodological solutions. Thus the
challenge is to try to find an acceptable compromise between all of the different requirements
and views.
The best option for the numerator therefore seems to be a count of new businesses, and for
the denominator, the population of active businesses. New businesses should be split into
pure births and other creations, along the lines of the Eurostat definition as the rate of other
creations will vary between countries depending on national registration systems and
practices, whereas pure births are much more suitable for international comparability
purposes. As purity has a major impact on comparability, the key indicator should focus on
pure births, though information on other creations may have some value in terms of quality
28
For more detail see the OECD Business Demography Framework
44
assurance, and comparing the impact of national systems on the business community. The
method to determine pure births proposed by Eurostat, based on automatic matching and
limited clerical checking of large units seems suitable, and has the advantage of already being
in place in around half of the OECD member countries.
In terms of timing and thresholds, the point at which a new business takes on its first paid
employee seems to be the easiest to measure in a consistent way across countries. This does
not mean that non-employer businesses are of no interest, just that they should be considered
in a secondary indicator. This means that births are defined as entries into the population of
employers, regardless of whether the business previously existed with no employees or not.
An important issue to resolve concerns businesses that fluctuate between having employees
or not, either on a seasonal basis, or in response to market conditions. The simplest approach
is to consider a business that leaves and re-enters the population of employers within a given
period as being a reactivation, and therefore not a pure birth. Eurostat currently recommend a
two year threshold for reactivations, which has the advantages of being relatively easy to
implement, and that it provides definitive data on pure deaths more quickly than if a longer
period (or no threshold) is used.
In terms of periodicity and temporal basis, annual data seem most appropriate, based as
closely as possible on the calendar year. A comparison of point in time populations at the start
and the end of the year to determine entries and exits is the easiest approach to implement.
This allows the construction of a simple equation as used in Australia (see Section 4.2),
whereby the population at a particular point in time is defined as the population from the
previous observation, plus entries in the previous period, minus exits in the previous period.
This sort of stock and flow approach is analogous to that used for human demography, and is
easy for non-specialists to understand. The potential duplication issue raised in Section 3.5 is
also reduced, thus improving data comparability, if a point in time population is used.
The remaining question in terms of periodicity is the treatment of short-lived enterprises. It is
methodologically preferable to include all of these. This would require the use of dates to
denote when a start-up occurred, or regular (at least monthly) observations of the population.
The best source for business start-up indicators seems to be national statistical business
registers. The units and coverage of these registers are gradually becoming more
harmonised, particularly in Europe. The unit of interest is usually the “business”, but this
concept is not really defined in its own right, hence the use of the enterprise, as defined in the
System of National Accounts seems most appropriate
29
. For the purposes of this indicator, the
definitions of the enterprise in the International Standard Industrial Classification (ISIC), and
the European Union regulation on statistical units, should be regarded as sufficiently similar.
Coverage should be defined as all “market” enterprises operating in the national economy.
The term “market” should be considered as excluding the government sector and non-profit
institutions serving households. In terms of economic activity, the best solution seems to be to
request a breakdown to at least the section-level of the ISIC.
This indicator should not necessarily be seen as permanent. It is designed more from the
pragmatic than the purist point of view, based on data that are currently available. The reason
for this is to try to get a comparable dataset as quickly as possible. This sort of relatively
simple indicator may well prove to be the best approach in the longer term, but it may also be
possible to improve it based on feedback from data users and the experiences of data
providers.
29
This definition is given in Section 3.7 above.
45
• Complementary Indicators
As the key start-up indicator proposed above is unlikely to be ideal for all purposes, a number
of complementary indicators could be envisaged (see also the OECD Business Demography
Framework). These complementary indicators are presented in approximate order of priority:
o An indicator of business start-ups using the working-age population of the country
as the denominator. A consistent definition of this population would need to be
applied, but this may be a useful secondary indicator, particularly for studying
entrepreneurship. It is also, perhaps, more relevant for economies in transition,
where the population of businesses starts low, but grows rapidly. In these
circumstances, start-up rates based on the business population could give a false
impression of the volume of start-ups.
o An indicator of the start-up rate for non-employer businesses: This indicator would
be rather problematic, as it would currently be heavily affected by the wide range of
thresholds used in national sources. More methodological work would be needed
to ensure real comparability, mainly to define a suitable threshold based on some
notion of labour-input that could be applied in all countries. The interest in this type
of business is, however, probably sufficient to justify this work. In the short term,
however, the development of indicators based on information sourced from
business registers, even without threshold adjustments, should be encouraged.
o An indicator of start-ups in terms of employment created: This could be developed
alongside the key indicator proposed above, but would need to be tested for
robustness, as it could be heavily influenced by relatively small differences in
purity.
o An indicator of start-ups in terms of businesses that survive for a minimum period:
Whereas the key indicator would aim to include all start-ups, no matter how short-
lived, a comparative measure of their durability would be useful. Thus start-up
rates defined in terms of businesses that survive for at least two years, or at least
five years, could be envisaged.
o An indicator of start-ups at the site level: Establishments from North America are
probably sufficiently similar to local units in Europe to consider the possibility of a
site-level start-up indicator. Ideally this would have two components, new sites due
to pure enterprise births, and new sites created by existing enterprises. Both are of
interest for studying employment dynamics and the impact of entrepreneurship at
the regional and local levels.
o An indicator of the start-up rate of non-market businesses: Non-profit institutions
serving households are a recognised category of institutional units in national
accounts. They have a clear role in society, and their activities have been referred
to as “social entrepreneurship”, thus measures of their dynamics could be of
interest for socio-economic policy making.
46
47
6. Conclusions
The basic question underlying this project, as stated in the introduction to this report, was;
“How comparable are data on business start-up rates from different OECD countries?” The
short answer, based on the factors of comparability above is: “Not very”. This is because the
comparability factors show that simple comparisons of start-up rates from the different
countries and sources listed in Annex 2 would be misleading and of little value. The longer
answer is, however, rather more positive. Even though data are not currently very
comparable, it seems relatively easy to make a number of improvements to comparability in
the short-term, both analytically and at source.
A number of more detailed conclusions can also be drawn from this report:
• The availability of data on business start-up rates varies considerably between countries.
Some have several sources and long time series that continue up to the present, whereas
others have limited sources, data for only a few years or data series that are not being
continued. For a few OECD member countries, no data sources have been found, and the
availability of data is also very limited for non-OECD countries.
• The availability of metadata is even more variable. Even where metadata exist, they are
not always easy to find or understand, even for specialists. A common metadata template,
based on the factors of comparability above, would make a significant contribution to the
understanding of the data and the reliability of international comparisons. As a first step
towards this, Annex 1 includes proposals for harmonised terminology.
• Some previous international comparisons have suffered from a lack of detailed
understanding of comparability issues. Having said that, they have, however, provided
some useful models for assembling comparable data. The distributed data analysis
models introduced in the OECD Firm-level data study, and, more recently, the Eurostat
business demography project, seem to provide the best route to obtaining harmonised
analyses whilst retaining the detailed knowledge of the source that is necessary for
accurate interpretations of the data.
• To assess comparability of business start-up rates it is necessary to decompose them into
numerator and denominator components, and consider the factors that affect each of
these. Start-up rates that might appear comparable at first glance may be much less so
when they are decomposed in this way.
• A total of nine factors affecting the comparability of business start-up rates have been
identified. Some of these are specific to the numerator or denominator, whereas the others
affect both. The factors that have the most impact are usually the purity of the data in the
numerator, the temporal basis of the denominator, and the coverage of both, though this
varies considerably depending on the data sources being compared.
• The larger a new business is, the less likely it is to be a pure birth in the sense of being a
genuinely new combination of production factors. There is thus a direct relationship
between the amount of work done to improve purity, and the resulting observed impact of
new businesses in terms of employment creation.
• It is possible to make certain analytical adjustments to start-up data to compensate for
differences in specific comparability factors. Unfortunately when compensating for one
factor, it is possible to affect others, and to introduce noise into the data, thus adjustments
have to be made with care, based on a detailed understanding of the data sources and
methods. However, adjustments based on estimates of the impact of a specific factor of
48
comparability can still help to determine whether differences in start-up rates are likely to
be significant or not.
• It would be preferable for any adjustments to be made at source, or through discussion
with those who have a detailed understanding of the data, to reduce the risk of these
adjustments having a negative impact on comparability.
• The degree of harmonisation of data sources has a considerable impact on the
comparability of the resulting data. In this respect, statistical business registers are
perhaps the best sources for business start-up data, as they are already subject to a
degree of harmonisation, of methods, coverage and contents, particularly within Europe.
• A basic key indicator, well defined and relatively easy to implement in all countries, is
necessary to improve data comparability. This approach is in line with the forthcoming
OECD methodological framework for business demography. The key indicator should be
supplemented by a number of complementary indicators that give additional insights to
more specialist data users.
• International organisations such as the OECD need to try to influence the mind-set of
those producing national data. These data producers are often more aware of, and
influenced by national data requirements than they are of the needs for international
comparability.
It is clear that fully comparable data sets can not be produced for all OECD member countries
without more work to develop a suitable methodological framework, and considerable efforts
on the part of those countries. As discussed in Section 5, the former is already in the course
of development at the OECD, but the latter is something that would be rather unrealistic to
expect, at least in the short-term. The focus should therefore be on incremental development
towards more harmonised indicators, whilst promoting longer term convergence within an
agreed methodological framework. The immediate priority is therefore the identification of
“quick wins”, i.e. actions that could increase the international comparability of data from
individual countries for minimal cost. This could include exploring the potential for countries to
supply certain additional data that could be used to make more informed adjustments to their
start-up rates.
At the same time, it is important to increase our understanding of what the users of the data
really want, and what they will use the data for. This knowledge can then inform the future
development of indicators, ensuring that they are as relevant as possible. This sort of step-by
step approach towards a clear goal through incremental improvements may not result in fully
comparable data as quickly as some users might want, but is likely to be more acceptable to
OECD member countries, and perhaps more timely than any more radical approach. For this
reason, it is likely to provide the quickest route to more comparable data on business start-
ups.
Thus in summary, although data on business start ups are not currently very comparable
between countries, there are a number of relatively quick and easy steps that could be taken
to improve their comparability. Clear and comprehensive metadata are vital, as is a detailed
methodological framework that balances user needs and the pragmatic concerns of data
producers. The goal of internationally comparable business start-up rates is not an easy one,
but is possible.
49
7. References
The references below are to papers cited in the main body of this report. References to data
sources are included in Annexes 2 and 4.
ABS (2005), “A Statistical View of Counts of Businesses in Australia”, information paper
published by the Australian Bureau of Statistics, October 2005.http://www.abs.gov.au/Ausstats/abs@.nsf/0/97e9b20f99363f92ca2570920075085a?OpenDoc
ument
ABS (2004), “Business Entries and Exits, A Conceptual Framework”, paper produced by the
Australian Bureau of Statistics, October 2004.
Ahmad, Nadim (forthcoming), “A Framework for Business Demography Statistics”, OECD
proposals for data harmonisation.
Ahmad, Nadim and Steven Vale (2005), “Moving Towards Comparable Business
Demography Statistics”, paper presented at the OECD Structural Business Statistics Expert
Meeting, Paris, November 2005.http://www.oecd.org/dataoecd/54/13/35563231.pdf
Armington, Catherine (1998), “Statistics of US Businesses – Microdata and Tables”, paper
prepared for the US Small Business Administration.http://www.sba.gov/advo/research/rs190tot.pdf
Baldwin, J ohn R., Desmond Beckstead and Andrée Girard (2002), “The Importance of Entry
to Canadian Manufacturing with an Appendix on Measurement Issues”, Statistics Canada
research paper.http://www.statcan.ca/english/research/11F0019MIE/11F0019MIE2002189.pdf
Bartelsman, Eric, J ohn Haltiwanger and Stefano Scarpetta (2005), “Measuring and Analyzing
Cross-country Differences in Firm Dynamics”, paper presented to the 2
nd
EUKLEMS
Consortium Meeting, Helsinki, J une 2005.http://www.euklems.net/meetings/CM_Helsinki/Bartelsman,_Haltiwanger_&_Scarpetta_(2005)
Bartelsman, Eric, J ohn Haltiwanger and Stefano Scarpetta (2004), “Microeconomic Evidence
of Creative Destruction in Industrial and Developing Countries”, World Bank Policy Research
Working Paper.http://siteresources.worldbank.org/INTWDR2005/Resources/creative_destruction.pdf
Brandt, Nicola (2004), “Business Dynamics in Europe”, OECD Science, Technology and
Industry Working Paper.http://www.olis.oecd.org/olis/2004doc.nsf/43bb6130e5e86e5fc12569fa005d004c/d7330a8070
18cccbc1256e55005051c7/$FILE/J T00159809.DOC
Cella, Patrizia and Caterina Viviano (2004), “Register Entries / Exits and Demographic Flows:
Some Comparisons for Statistical Aggregates”, paper presented to the 18
th
International
Roundtable on Business Survey Frames, Beijing, October 2004.http://forum.europa.eu.int/irc/DownLoad/kfecA5J -mpGIXl0SR1OAauLw2xPI-
SvMwc82kKbj0B2SBp2UxVqIlDcEc04rLSm9uxKFCBTAaIZ3lqM0jNc9dF5L_IbkP/S8-2%20-
%20Register%20entries-exits%20and%20demographic%20flows%20-
%20some%20comparisons%20for%20statistical%20aggregates.doc
50
Choi, Bongho and Denis Ward (2004), “Analysis of Statistical Units Delineated by OECD
Member Countries”, paper presented to the 18
th
International Roundtable on Business Survey
Frames, Beijing, October 2004.http://forum.europa.eu.int/irc/DownLoad/kYedA1J Sm_GMsYREH2D5UbAjBu320H8WVHgZtF
cfKeSCBQKuSxDZA0zr4s3K81qMzfT5uGS67YAxCZLHjBgRxd9Rm0df4e1D/S3-1%20-
%20ANALYSIS%20OF%20STATISTICAL%20UNITS%20DELINEATED%20BY%20OECD%2
0MEMBER%20COUNTRIES.doc
Davis, Steven J ., J ohn Haltiwanger, Ron J armin, C. J . Krizan, J avier Miranda, Alfred Nucci
and Kristin Sandusky (2005), “Measuring the Dynamics of Young and Small Businesses:
Integrating the Employer and Nonemployer Universes”, preliminary paper presented to the
US National Bureau of Economic Research Conference on Research in Income and Wealth,
Bethesda, April 2005.http://www.nber.org/confer/2005/CRIWs05/haltiwanger.pdf
Eurostat (2003), Eurostat Manual of Recommendations for Business Registers, Chapter 13.http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/embs/registers/chapter13.doc
GEM (2004), Global Entrepreneurship Monitor 2004 Global Reporthttp://www.gemconsortium.org/document.asp?id=364
Herczog, Aimée, Hans van Hooff and Ad Willeboordse (1998), “The Impact of Diverging
Interpretations of the Enterprise Concept”, prepared for Eurostat by Statistics Netherlands.http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/embs/ent_concept/section1.html
Klapper, Leora, Luc Laeven and Raghuram Rajan (2004), “Business Environment and Firm
Entry: Evidence from International Data”, US National Bureau of Economic Research Working
Paper.http://www.nber.org/papers/w10380
Mead, Geoff (2005), “Statistics New Zealand Business Frame Strategy and Developments
Related to Statistics on SMEs and the Support of Longitudinal Business Statistics”, paper
presented at the OECD Structural Business Statistics Expert Meeting, Paris, November 2005.http://www.oecd.org/dataoecd/9/44/35485064.pdf
Mills, Duncan and J ason Timmins (2004), “Firm Dynamics in New Zealand: Comparative
Analysis with OECD Countries”, paper presented to the 2004 conference of the New Zealand
Association of Economists.http://www.nzae.org.nz/conferences/2004/88-Mills-Timmins.pdf
Pinkston, J oshua C., and J ames R. Spletzer (2004), “Annual Measures of Gross J ob Gains
and Gross J ob Losses”, US Bureau of Labor Statistics Monthly Labor Review, November
2004.http://www.bls.gov/opub/mlr/2004/11/art1full.pdf
Takahashi, Masao (2000), “Business Demography and the J apanese Business Survey
Frame”, paper presented to the 14th International Roundtable on Business Survey Frames
Auckland, November 2000.http://forum.europa.eu.int/irc/DownLoad/kveFAjJ ZmSGspYM195H5EFCl6eTNvOz6Vt5McKbY
N63r0IIuHVQp4CmHyIxc1GjlFVXmUpoo2tSfBIMGtOpIxcLHbI/Paper%20J apan%20-
%20session7.pdf
Vale, Steven (2005(a)), “International Data on Business Start-ups: Factors Affecting
Comparability”, paper presented to the 19
th
International Roundtable on Business Survey
Frames, Cardiff, October 2005.http://forum.europa.eu.int/irc/DownLoad/kXeeAJ J AmjGIcxMS0S9p_GZ2GvLRcOhH4rLSm9ux
KEH2EUGq8jov3OoGlUXySA_6pkAUF-S0DF-9gqmf-
Hj0fcg/Comparability%20RT%20Paper.doc
51
Vale, Steven (2005(b)), “The Coverage of Micro-Enterprises in Business Registers”, paper
presented at the OECD Structural Business Statistics Expert Meeting, Paris, November 2005.http://www.oecd.org/dataoecd/32/46/35506105.pdf
Vale, Steven and Claire Powell (2003), “Developments in Business Demography
- Reconciling Conflicting Demands”, paper presented to the Comparative Analysis of
Enterprise (micro) Data Conference, London, September 2003.http://www.statistics.gov.uk/events/CAED/abstracts/downloads/vale.pdf
Vale, Steven and Claire Powell (2002), “Estimating Under-coverage of Very Small
Enterprises”, paper presented to the 16
th
International Roundtable on Business Survey
Frames, Lisbon, October 2002.http://forum.europa.eu.int/irc/DownLoad/kgecA1J SmRGTex2OFEBEGfEvoFle1HrMxrNLRFur
O5CG0NNpCl6eTNvOz64c5SN2VkAUF-0RCF-9gqmf-H/ses3_UK_Paper.pdf
Volfinger, Zsolt (2004), “Coverage of the Hungarian Business Register”, paper presented to
the 18
th
International Roundtable on Business Survey Frames, Beijing, October 2004.http://forum.europa.eu.int/irc/DownLoad/kjecAJ J UmfG1uvhdvqlF0uAePVRfj3jMhqKGf0phOGF
-HOBF7z6zLRjGpRmu-AZ-um3THrGuypb4pqOIjE5Tzc1L/S5-3%20-
%20Coverage%20of%20the%20Hungarian%20Business%20Register.doc
52
53
Annex 1 – Glossary of Terms: Proposals for Harmonised
Terminology
This report notes that comparability of data on business start-ups, and business demography
more generally, is hampered by inconsistent presentation of metadata. To reduce this
problem, the creation of a standard metadata template is proposed. As a first step towards
this, a harmonised terminology is needed. This Annex proposes standard terms and
definitions, using a notation which relates events to a particular time period, t.
These terms and definitions have been used throughout this report, so this Annex also acts as
a glossary.
• Births (B
t
) – A birth is the creation of a combination of production factors with the
restriction that no other national businesses are involved in the event. Births do not include
entries into the population due to reactivations, mergers, break-ups, split-offs or other
restructuring of a group of businesses linked by ownership or control. Births also exclude
entries into a population resulting from changes to characteristics of existing businesses.
(Note – this is largely based on, and fully consistent with the Eurostat definition for
enterprise births).
• Churn – Total churn is defined as the sum of businesses joining and leaving the
population during a given period, i.e. entries plus exits (E
t
+X
t
). Pure churn excludes
entries and exits that are due to events other than births and deaths (i.e. B
t
+D
t
).
• Closing Stock (PB
t
) – The population at the end point of the period. This is usually
equivalent to the population at the start point for the following period (PA
t+1
).
• Deaths (D
t
) - A death is the dissolution of a combination of production factors with the
restriction that no other domestic businesses are involved in the event. Deaths do not
include exits from the population due to temporary inactivity, mergers, take-overs, break-
ups or other restructuring of a group of businesses linked by ownership or control. Deaths
also exclude exits from a population resulting from changes to characteristics of
businesses which remain active. (Note – this is largely based on, and fully consistent with
the Eurostat definition for enterprise deaths).
• Entries (E
t
) – All businesses that join the population during the period, regardless of
whether they are still present at the end of the period.
• Exits (X
t
) – All businesses that leave the population during the period, regardless of
whether they were present at the start of the period.
• J oiners (J
t
) – Businesses that are present in the population at the end of the period, but
were not present at the start of the period.
• J oiners and leavers within period (J L
t
) – Businesses that are not present in the population
at the start or the end of the period, but are present in at least one observation of the
population between these two points (or would be if such observations were made).
• Leavers (L
t
) – Businesses that are present in the population at the start of the period, but
are not present at the end of the period.
54
• Opening Stock (PA
t
) – The population at the start point of the period. This is usually
equivalent to the population at the end point for the previous period (PB
t-1
).
• Other Entries (OE
t
) – All entries that are not births
• Other Exits (OX
t
) – All exits that are not deaths
• Population – All businesses that meet certain predefined criteria.
o Live during Period (P
t
) - All businesses that meet certain predefined criteria at any
time during a specified time period.
o Point in Time – All businesses that meet certain predefined criteria at a specific
temporal reference point.
• Purity – The degree to which pure births and deaths are distinguished from other
demographic events.
• Survivors (S
t
) All businesses that are in the population at both the start and the end of the
period
55
Annex 2 - Inventory of Data on Business Start-ups by Country
Introduction
This Annex provides an inventory of the available data and metadata found for each OECD
country. It also includes information from other countries for which data have been found in
the course of this work. In some cases, data are also available for groups of countries (e.g.
the members of the European Union), via international agencies.
The information contained in this Annex has been compiled based on searches of the Internet
in Autumn 2005, and the author’s knowledge of sources. The focus is on official data sources,
usually from National Statistical Institutes, though other sources are considered for some
countries.
The following pages list the sources found for each country, using a standard template for
each source. They include links to the data where possible, but do not attempt to explicitly
assess the comparability or any of the other dimensions of the quality of the data. Summary
metadata, including coverage and definitions, are included, as well as information on the
availability of more detailed metadata.
Several countries have participated in the Eurostat business demography data collections, the
DOSME (Demography of Small and Medium-sized Enterprises) project, and/or the OECD
firm-level data project, so data are available via those routes. These sources are included for
each country if appropriate, but to avoid repetition, information on the coverage and definitions
used are provided separately at the end.
1. Australia
One data source available
Title – Web publication “Experimental Estimates, Entries and Exits of Business Entities”
Source – Australian Bureau of Statistics, 2005
Internet address –http://www.ausstats.abs.gov.au/Ausstats/subscriber.nsf/Lookup/2EB3AE08FFBC9AD4CA257
0280078B69E/$File/8160055001_2001-02,%202002-03%20and%202003-04.pdf
Contents – Register-based population, entry, survival and exit estimates.
Breakdowns – The data are broken down by economic activity, size, geography and “type of
business entity” (legal form). Size breakdowns are not given for survival data.
Metadata – Some metadata are included in the publication, further information is contained in
an additional paper, “Business Entries and Exits – A Conceptual Framework”, available from
the Australian Bureau of Statistics on request. See also the paper “Development of Statistics
on Business Demography and Continuity in Australia” -http://forum.europa.eu.int/irc/DownLoad/kxeFAAJ UmZGMYjKH--
EUNFa3pFKPjEWCqF4EiCwmAUM8GSHYRf6dTHzryIxJ 9UZ-bYR3R1H-
BbbkSskDDjYv4G8BZM/ses3_Australia_Paper1.pdf
56
Period covered – The population estimates are as at 1 J uly for 2001 to 2003. Entries and
exits are available for the periods 1 J uly to 30 J une 2001/2 to 2003/4. One and two year
survival rates are available for entries in 2001/2 and one year survival rates for entries in
2002/3.
Coverage – ANZSIC division M (government) and non-market government entities are
excluded. The threshold for registration is generally 50,000 Australian Dollars (approx.
€31,000), with some exceptions, and some voluntary registrations.
Definitions
• The population estimates are on a point in time basis.
• The unit used is the legal unit, i.e. entities registered for an Australian Business
number (ABN).
• Entries are defined as the allocation of a tax role within the Australian Business
Register (ABR). This excludes inactive businesses, changes in legal form, and
reactivations. A check is also made for newly created businesses that take over the
activity of one or more existing businesses. These are excluded where identified.
• Exits are defined as the cancellation of all tax roles within the ABR, with similar
inclusions and exclusions to those for entries.
2. Austria
One data source available
Title – Unternehmensneugründungen in Österreich 1993-2004
Source – Wirtschaftskammern Österreich (WKO)
Internet address –http://portal.wko.at/wk/dok_detail_file.wk?AngID=1&DocID=344536&DstID=1721&StID=17871
2
Contents – Start year stock and new registrations for 1993 to 2004
Breakdowns – New registrations are broken down by economic activity, legal form,
geography and (for natural persons) sex and age of the entrepreneur.
Metadata – There are some descriptive metadata in the introduction to the publication.
Period covered – 1993 to 2004
Coverage – Registrations with the WKO
Definitions
• The population estimates are on a point in time basis (start of the year), and consist of
active WKO registrations. The population data are not subject to the same adjustments
as those for new registrations.
• The basic unit is the registration at the WKO, which can be considered as a legal unit.
However, the corrections made bring the unit used for analysis much closer to the
enterprise.
• New registrations are adjusted to remove re-registrations, dormant units, and multiple
registrations for the same enterprise. Adjustments are also made for registration lags.
57
3. Belgium
Two data sources available
a) Title – Démographie des entreprises (1998-2004)
Source – Statistics Belgium
Internet address –http://statbel.fgov.be/figures/d422_fr.asp
Contents – Value-added tax (VAT) stock, new registrations, de-registrations and liquidations
Breakdowns – The data are broken down by legal form, economic activity and geography.
Metadata – Limited to table headings and notes. More detailed metadata exist in the
publication “Démographie des entreprises 2002”, which also contains more comprehensive
data breakdowns, but just for 2002.
Period covered – The population estimates are as at 31 December for 1998 to 2003.
Registrations and de-registrations are available for the years 1998 to 2003. Liquidations are
available for the years 1998 to 2004.
Coverage – Some specific legal, medical, financial, social and personal services are exempt
from value-added tax. There is no VAT registration threshold. Public sector entities are
included if they are registered for VAT.
Definitions
• The stock data are on a point in time basis, and include those VAT registrations
marked as active at the end of the year.
• The unit used is the VAT registration, which is broadly equivalent to the legal unit.
• Entries are defined as VAT registrations whose year of creation is the reference year,
including those that have been removed from the VAT register before the end of that
year.
• Exits are defined as VAT registrations whose year of removal from the VAT register is
the reference year.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2000. Data
for births exist for 1998 and 2000 (not 1999), and data for deaths exist for 1998 and 1999.
58
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
4. Canada
Three data sources available
a) Title – Business Dynamics in Canada, 2001 (supplemented for 2002 by data prepared for
FORA).
Source – Statistics Canada – Longitudinal Employment Analysis Program (LEAP) Database.
Internet address –http://www.statcan.ca:8096/bsolc/english/bsolc?catno=61-534-X (note:
$25 charge)
Contents – This publication includes data on business populations, birth and death rates, and
survival.
Breakdowns – Business populations, births and deaths are broken down by size and
economic activity categories based on knowledge intensity. Business populations are also
broken down by geography.
Metadata – The publication contains a chapter on methodology.
Period covered – The population estimates are available for 1991 to 2001, births are
available from 1992 to 2001, and deaths from 1991 to 2000. Additional data on population and
births for 2002, and deaths for 2001 is taken from a short report prepared by Statistics
Canada for FORA.
Coverage – All employers in Canada, public and private (i.e. data do not include businesses
with no employees). Non employing businesses are included in the report prepared for FORA,
but only for three years, and have considerable variation in the rates, which could call into
question the validity of these data.
Definitions
• The unit used is the firm, which, at the national level is equivalent to the legal unit.
• Births are defined as firms that are not present on the LEAP database in year t, but are
present in year t+1. The birth rate is the number of new enterprises in t+1 divided by
the total number of firms observed in year t+1.
• Deaths are defined as firms that are present on the LEAP database in year t, but are
not present in year t+1. The death rate is the number of enterprises operating in t, but
not in t+1, divided by the total number of firms observed in year t+1. Note this is
different to the rates calculated in many other countries where the population in year t
is the denominator.
b) Title – Self-Employment Entry and Exit Flows
Source – Statistics Canada – Paper by Zhengxi Lin, Garnett Picot and J anice Yates
Internet address –http://www.statcan.ca/english/research/11F0019MIE/11F0019MIE1999134.pdf
59
Contents – This paper contains a table with counts of self-employed persons, and rates for
entry and exit, as well as other related data and analyses.
Breakdowns – No breakdowns of the entry and exit data are given in the paper.
Metadata – The paper contains descriptive metadata on sources and definitions.
Period covered – The population estimates are available for 1981 to 1995, entries are
available from 1982 to 1995, and exits from 1981 to 1994.
Coverage – All persons declaring income from self employment in their annual tax returns to
revenue Canada.
Definitions
• The unit used is the person completing a tax return.
• Self-employment entries are income-tax filers who report earnings from self-
employment in one year but not the previous year
• Self-employment exits are income-tax filers who report earnings from self-employment
in one year but not the next.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Canada are available from 1984 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
5. Czech Republic
Two data sources available
a) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
60
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2001 and 2002, and data for deaths exist for 2000 and 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
6. Denmark
Three data sources available
a) Title – Statistical Yearbook
Source – Statistics Denmark
Internet address
2005 (data for 2001) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2005.aspx
2003 (data for 2000) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2003.aspx
2001 (data for 1999) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2001.aspx
61
2000 (data for 1998) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2000.aspx
Contents – Counts of new enterprises
Breakdowns – Data are broken down by economic activity
Metadata – Very limited
Period covered – Birth counts are available for 1998 to 2001
Coverage – Data exclude agriculture and public administration
Definitions
• The unit used is the enterprise
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2001. Data
for births exist for 1998 to 2001, and data for deaths exist for 1997 to 2000.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Denmark are available from 1981 to 1994.
62
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
7. Finland
Three data sources available
a) Title – Enterprise openings and closures
Source – Statistics Finland
Internet address –http://www.stat.fi/til/aly/index_en.html Data only accessible via Finnish
version -http://www.stat.fi/til/aly/index.html
Contents – Counts of enterprise openings and closures. Stock figures are available
separately from the StatFin database, but may not have the same coverage.
Breakdowns – Data are broken down by economic activity, legal form and geography.
Metadata – Mostly in Finnish
Period covered – Data are available from 1999 to 2004
Coverage – The openings and closures data are derived from Statistics Finland’s business
register. They only cover those enterprises engaged in business activity that are liable to pay
value-added tax or act as employers. Foundations, housing companies, voluntary
associations, public authorities and religious communities are excluded. The data cover state-
owned enterprises, but not those owned by municipalities.
Definitions
• The unit used is the enterprise
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
63
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Finland are available from 1989 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
8. France
Three data sources available
a) Title – Créations d'entreprises
Source – INSEE
Internet address –http://www.insee.fr/fr/ffc/chifcle_liste.asp?theme=9&soustheme=1&souspop=
Contents – Counts of enterprise creations, split into new creations, resumptions and re-
activations
Breakdowns – Breakdowns by economic activity, size and legal form are available.
Metadata – Limited, e.g. some key definitions.
Period covered – Data are available from 1993 to 2004
Coverage – The data cover all of France, including the overseas départements.
Definitions
• The unit is assumed to be the enterprise.
• Three categories of enterprise creation are identified:
o Pure creations (creations “ex nihilo”) where the new enterprise does not take
over the activities of a previously existing enterprise.
o Reactivations, where a person who has previously been self-employed re-
starts a self-employed activity.
o Resumptions, where a new business takes over an activity previously carried
out by another enterprise.
64
b) Title – La Création en Chiffres
Source – Agence Pour la Création d’Entreprises (APCE)
Internet address –http://www.apce.com/index.php?rubrique_id=261&type_page=I
Contents – Counts of enterprise creations, split into new creations (“ex nihilo”), resumptions
and re-activations
Breakdowns – None
Metadata – Very limited
Period covered – Data are available from 1993 to 2004
Coverage – No information
Definitions – None given – the data are very similar to, but not the same as, those from
INSEE.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for France are available from 1990 to 1996.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
9. Germany
Three data sources available
a) Title – Business Notifications
Source – Federal Statistics Office, Germany
Internet address –http://www.destatis.de/themen/e/thm_unternehmen.htm
65
Contents – Business registrations, modifications and de-registrations, and counts of
businesses liable to pay turnover tax.
Breakdowns – The data are broken down by economic activity.
Metadata – Descriptive metadata are available via the website
Period covered – Registration and de-registration data are available for 2001 to 2003. Data
on the population of businesses liable for tax are available for 2002 and 2003.
Coverage – The registration and de-registration data are assumed to cover the whole
economy. The population of businesses liable to pay turnover tax covers businesses with a
turnover of at least €16,620 per year. It covers most economic activities, with exceptions for
certain health, public administration, insurance and agricultural activities.
Definitions
• The unit for registrations and de-registrations is effectively the local unit as “the
obligation to report business registrations and de-registrations applies to enterprises,
branch offices and dependent sub-offices”.
• Registration is required when a new activity is started or a business is taken over, be it
through purchase or succession, a partner entering the business, a change in legal
form, or a relocation of the business to a different registration district.
• De-registration is required when a business is shut down completely or in part, or is
sold, a partner withdraws from the business, the legal form is changed, or the business
is relocated to a different registration district.
b) Title – Start-ups and Liquidations in Germany 1991 - 2004
Source – Institut für Mittelstandsforschung Bonn
Internet address –http://www.ifm-bonn.org/dienste/gruendungen-engl.htm
Contents – Counts of business start-ups and liquidations. Some enterprise population totals
in Table 1 of:http://www.ifm-bonn.org/dienste/kap-2.pdf
Breakdowns – The start-up and liquidation data are broken down into the former East and
West Germany.
Metadata – Limited metadata available via the web site.
Period covered – Start-up and liquidation data are available for 1991 to 2004. Data on the
population of businesses are available for 1994 and 1996 to 1999 (IfM have provided
estimates for the missing years).
Coverage – The population of businesses is subject to a threshold (€17,500 since 2003), and
covers all economic activities except the “liberal professions”, most health services and some
insurance services that are not subject to value added tax.
Definitions
• The units used are effectively the sub-set of legal units that are considered to be
economically relevant.
66
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Germany are available from 1978 to 1998.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex. Note – data for Germany only cover the former West Germany.
10. Greece
No data sources found
11. Hungary
Three data sources are available
a) Title – Enterprises and Non-profit organisations
Source – Hungarian Central Statistical Office
Internet address –http://portal.ksh.hu/portal/page?_pageid=38,341368&_dad=portal&_schema=PORTAL
Contents – Annual counts of registered economic corporations and unincorporated
enterprises, as well as quarterly counts of new registrations
Breakdowns – Both are broken down by legal form, the population data are also broken
down by economic activity.
Metadata – Information on definitions and sources is available on the web site. See also the
paper “Coverage of the Hungarian Business Register” at:http://forum.europa.eu.int/irc/DownLoad/kjecAJ J UmfG1uvhdvqlF0uAePVRfj3jMhqKGf0phOGF
-HOBF7z6zLRjGpRmu-AZ-um3THrGuypb4pqOIjE5Tzc1L/S5-3%20-
%20Coverage%20of%20the%20Hungarian%20Business%20Register.doc
Period covered – Population are available from 1994 to 2004. Annual data on new
registrations can be constructed by adding the four quarterly totals for 2001 to 2004.
67
Coverage – The data cover all businesses that hold an active registration and tax number in
the administrative register, including most government bodies. There is no registration
threshold in Hungary, so part-time businesses are included. Approximately 75% of
registrations are considered to be economically active by the Hungarian statistical office. All
economic activities are covered, though NACE division L (public administration) is excluded
from the counts broken down by activity.
Definitions
• The unit is referred to as the enterprise, but the definition is closer to that of a legal
unit.
• The population data are point in time estimates for the end of the year.
b) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
c) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2000 to 2002, and data for deaths exist for 1999 to 2001.
68
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
12. Iceland
Two data sources available
a) Title – Registered Enterprises and Organisations 1995-2001
Source – Statistics Iceland
Internet address –http://www.statice.is/?pageid=1198&src=/temp_en/fyrirtaeki/fyrirtaeki.asp
Contents – Counts of registered enterprises, new registrations and “new depreciation”
Breakdowns – The data are broken down by legal form.
Metadata – Only brief footnotes are available.
Period covered – The data are available for 1995 to 2001.
Coverage – The data seem to cover all legal forms.
Definitions
• The population figures appear to be point in time estimates, as at the end of the year.
• The unit used appears to be the legal unit.
b) Title – Enterprises / New Registrations by Economic Activity
Source – Statistics Iceland
Internet address –http://www.statice.is/?pageid=1198&src=/temp_en/fyrirtaeki/fyrirtaeki.asp
Contents – Counts of enterprises and new registrations
Breakdowns – The data are broken down by economic activity (NACE section).
Metadata – Only brief footnotes are available.
Period covered – The enterprise population data are available for 1999 to 2004, the new
registrations data are available from 1995 to 2004.
Coverage – The data seem to cover all economic activities.
Definitions
• Non available
13. Ireland
No data sources found, though Ireland are starting to supply data for the Eurostat business
demography project
69
14. Italy
Three data sources available
a) Title – Movimprese
Source – InfoCamere
Internet address –http://www.infocamere.it/movi_search.htm
Contents – Counts of total registrations and active registrations, new registrations, cessations
and changed registrations at the Italian chamber of commerce.
Breakdowns – The data are broken down by economic activity and geography.
Metadata – A glossary and other metadata are available on the web site (in Italian)
Period covered – The data are available for 1995 to 2004.
Coverage – The data do not cover NACE section L (public administration), and presumably
do not cover government units.
Definitions
• The unit used is the legal unit
• The population data are point in time, and appear to relate to the end of the year
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
c) Title – OECD Firm-Level Data Project
Source – OECD
70
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Italy are available from 1987 to 1993.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
15. Japan
One data source found
Title – Establishment and Enterprise Census
Source – Statistics bureau of J apan
Internet address –http://forum.europa.eu.int/irc/DownLoad/kveFAjJ ZmSGspYM195H5EFCl6eTNvOz6Vt5McKbY
N63r0IIuHVQp4CmHyIxc1GjlFVXmUpoo2tSfBIMGtOpIxcLHbI/Paper%20J apan%20-
%20session7.pdf
There are also some data on establishment and enterprise populations athttp://www.stat.go.jp/english/data/jigyou/kekka.htm.
Contents – Table 3 of the paper at the first address above includes counts of existing
establishments (i.e. survivors), “newly-organised establishments” and “abolished
establishments” based on data from the 1989, 1994 and 1999 establishment and enterprise
censuses.
Breakdowns – The data for 1999 are broken down by economic activity and employment size
band.
Metadata – The paper contains definitions and descriptive metadata.
Period covered – The data are available for 1989, 1994 and 1999. Annualised “opening” and
“abolishment” rates are also given for the periods between censuses.
Coverage – The data do not cover sole-proprietor businesses in agriculture, forestry and
fishing activities, or any businesses classified to domestic services, foreign governments or
international agencies. Several other specific exclusions are listed in the paper.
Definitions
• The unit used is the establishment
• The population data are on a point in time basis
• A newly-organized establishment is defined as an establishment that had been newly-
organized or had moved into the present place since the date of the preceding census.
71
• An abolished establishment is defined as an establishment that had moved to a
different place or had been closed since the date of the preceding census.
16. Korea
No data sources found, though there are some counts of establishments at:http://kosis.nso.go.kr/cgi-
bin/SWS_1021.cgi?KorEng=2&A_UNFOLD=1&TableID=MT_ETITLE&TitleID=HA&FPub=4&U
serID=
17. Luxembourg
Two data sources available
a) Title – Démographie des Entreprises
Source – STATEC
Internet addresses –http://www.statistiques.public.lu/stat/TableViewer/tableView.aspx?ReportId=258http://www.statistiques.public.lu/stat/TableViewer/tableView.aspx?ReportId=259http://www.statistiques.public.lu/stat/tableviewer/document.aspx?FileId=209
Contents – Counts of the stock of enterprises at the start of the year, new enterprises created
during the year in the framework of the policy of economic diversification, and requests for
authorisation for establishments. Note – from the numbers given, the new enterprise data
would only seem to account for a small proportion of all enterprise births.
Breakdowns – The stock and new enterprise data are broken down by very broad categories
of economic activity. The requests for authorisation for establishments are broken down by
nationality of the requestor (Luxembourgish or foreign).
Metadata – Some metadata are available by following the information links within the tables
on the web site.
Period covered – Data on the stock of enterprises are available for 2002 to 2004. Data on
new enterprises are available for 1990, and 2000 to 2004. Data on requests for authorisation
for establishments are available for 1990 and 1995 to 2004.
Coverage – There is no specific information on coverage.
Definitions
• The data on the stock of enterprises are on a point in time basis (1 J anuary of the
reference year).
• The unit used is assumed to be the enterprise for the stock and new enterprise data,
and the local unit for the data on requests for authorisation for establishments.
b) Title – Business demography indicators
Source – Eurostat
72
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
18. Mexico
No data sources found
19. Netherlands
Three data sources available
a) Title – Establishment and Closure of Businesses
Source – Statistics Netherlands
Internet address –http://statline.cbs.nl/StatWeb/table.asp?PA=07223eng&D1=a&D2=0&D3=(l-11)-
l&DM=SLEN&LA=en&TT=2
Contents – Counts (and employment) of the stock of businesses (as at 1 J anuary),
businesses opening, and businesses closing.
Breakdowns – The data are available broken down by economic activity.
Metadata – Metadata are available by clicking on the table headings on the web site.
Period covered – The data are available from 1993 to 2002 (closures only to 1996).
Coverage – The data exclude certain NACE categories (Sections A, B, E, L, M and N, and
divisions 70, 73, 91 and 92). On this basis it is assumed that most government activity is also
excluded.
Definitions
• The stock of businesses is a point in time estimate
• The unit used is the “business” which is assumed to be close to the enterprise, as the
terms are both used in the metadata.
• Establishment of a business is the formation of a new enterprise. This implies that the
statistical criteria for enterprises (autonomy, description and external orientation) have
to be met. Moreover, the enterprise has to be economically active, i.e. at least one
73
person works in the enterprise for at least 15 hours a week. The enterprise has to be a
new one, i.e. not the continuation of one or more existing enterprises.
• Closure of enterprises implies discontinuation of all activities, hence
no continuation of activities by other enterprises.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1999 to 2001. Data
for births exist for 1998 to 2002, and data for deaths exist for 1998 to 2000.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
c) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for the Netherlands are available from 1987 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
20. New Zealand
One data source available
Title – SMEs in New Zealand: Structure and Dynamics - 2005
74
Source – New Zealand Ministry of Economic Development
Internet address –http://www.med.govt.nz/irdev/ind_dev/smes/2005/index.html
Contents – Counts of the stock of enterprises in February of each year (Figure 1 of the
publications for 2001 to 2005), and enterprise births and deaths during the year (table
underlying Figure 15 of the 2005 publication).
Breakdowns – Stock data are broken down by employee size band. There are no
breakdowns of the birth and death data.
Metadata – The 2005 publication contains extensive metadata, including a glossary.
Period covered – The stock data are available for 2000 to 2004. The data on births and
deaths are available for 1998 to 2004.
Coverage – The data exclude agriculture production (ANZSIC subdivision A01). They
also exclude businesses of “little economic significance”, i.e. those that fail to meet at least
one of the following criteria:
• greater than $30,000 (approx €17,500) annual taxable expenses or sales
• rolling mean employee count of greater than three
• in a tax-exempt industry (except for residential property leasing and rental)
• part of a group of enterprises
• registered for tax and involved in agriculture or forestry.
Definitions
• The stock data are on a point in time basis, with a February reference date.
• The unit used appears to be the legal unit, though the term ‘enterprise’ is used.
• Data on the entry and exit of firms include administrative changes such as
restructuring and changes of ownership, as well as genuine business start-ups and
closures.
21. Norway
Two data sources available
a) Title – Statbank Norway / Enterprises
Source – Statistics Norway
Internet address –http://statbank.ssb.no/statistikkbanken/default_fr.asp?PLanguage=1
Contents – Count and employment data on the population of enterprises, new enterprises
and enterprise “drop-outs”. (Limited data on survival are also available at -http://www.ssb.no/english/subjects/10/01/fordem_en/tab-2004-12-01-01-en.html).
Breakdowns – The data are broken down by geography, economic activity, legal form and
size band.
Metadata – Detailed methodological notes and definitions are available athttp://www.ssb.no/vis/foretak_en/about.html
75
Period covered – The data on the population of enterprises are available for 2001 to 2005.
The data on new enterprises and enterprise drop-outs are available for 2001 to 2004.
Coverage – Enterprises classified to public administration, agriculture, forestry and fishing are
excluded, as are central and local government units.
Definitions
• A new enterprise in a given period is an enterprise registered with dates that indicate
start-up in this period.
• The number of new established enterprises is the number of a new enterprises
corrected for the change of ownership. That is - new enterprises that take over existing
activity are not counted as new established enterprise, but only as a new enterprise.
• A discontinuance of an activity is counted as a drop-out. If all of the establishment is
closed down, and is not taken over by another enterprise, the drop-out is also
classified as a closure.
• The population of enterprises is a point in time estimate as of 1st J anuary.
• The unit used is the enterprise
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2001. Data
for births exist for 1998 to 2001, and data for deaths exist for 1999 and 2000.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
22. Poland
Two data sources available
a) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
76
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
b) Title – Entry and Exit Rates in the Polish manufacturing
Source – National Bank of Poland
Internet address –http://www.fcee.urv.es/departaments/economia/recerca/grit/Catala/web/papers/Rogowski-
Socha.pdf
Contents – This paper compares data from several national sources on business entry and
exit. Most sources concentrate only on manufacturing, but whole economy entry and exit rates
from the REGON register are included in Table 10.
Breakdowns – The data are broken down by broad economic activity.
Metadata – The paper contains some limited metadata, mainly describing the source.
Period covered – Entry and exit rates are available for 1998 to 2003
Coverage – The data are claimed to cover the whole economy, but the breakdown by broad
economic activity does not include data for agriculture, business services, public
administration, health, education or personal services, so these activities may be excluded. A
warning is given that around a half of the businesses included in REGON were inactive in
1999, dropping to 30-40% in 2003. This could imply that entry rates as a proportion of active
businesses should be much higher, but it is likely that a proportion of entries are themselves
inactive.
Definitions
• The data appear to use a point in time population.
• The unit used is referred to as the enterprise, but is not defined.
23. Portugal
Two data sources available
Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
77
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
b) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for Portugal are available from 1983 to 1997.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
24. Slovakia
Two data sources available
a) Title – Demography of Small and Medium-sized Enterprises (DOSME) Study
Source – Eurostat
Internet address –http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
Contents – Data on enterprise populations, births, deaths, survival and factors of success.
Breakdowns – The data are broken down by country, economic activity, size, legal form and
characteristics of the entrepreneur.
78
Metadata – The final report describes the methodology used to produce the data it contains.
Other descriptive metadata is available on the project web site.
Period covered – Data are available for births and deaths from 1994 to 2001
Coverage and Definitions – See information on DOSME standard coverage and definitions
at the end of this Annex.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2000 to 2002, and data for deaths exist for 2000 and 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
25. Spain
Two data sources available
a) Title – Demografía de las Empresas
Source – Instituto Nacional de Estadística
Internet address –http://www.ine.es/inebase/cgi/um?M=/t37/p201&O=inebase&N=&L=0
Contents – Counts of enterprises, creations (split between pure births and reactivations), and
cessations.
Breakdowns – The population, creation and cessation counts are broken down by economic
activity, legal form and size. The INEbase data warehouse allows more detailed breakdowns
of total creations and cessations by the same variables.
Metadata – Some descriptive metadata are available via the web site above.
Period covered – The population counts are available for 1999 to 2005. Data on creations
and cessations are available for 1998 to 2004, but the split of creations into pure births and
reactivations is only present for 2001 to 2004.
79
Coverage – The data appear to exclude agriculture, forestry, fishing and public administration
activities, as well as central and local government units.
Definitions
• The population counts are on a point in time basis (1 J anuary).
• The unit used is the enterprise.
• Creations (altas) are defined as new registrations in DIRCE (the statistical business
register) that start their activities in the reference year.
• Cessations (bajas) are defined as units that cease activity in DIRCE in the reference
year.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 2000 to 2002. Data
for births exist for 2000 to 2002, and data for deaths exist for 2000 and 2001.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
26. Sweden
Two data sources available
a) Title – Nystartade företag (New enterprise starts)
Source – Statistics Sweden
Internet address –http://www.scb.se/templates/tableOrChart____27185.asphttp://www.scb.se/templates/Standard____36176.asp
Contents – Counts of new enterprises, and the stock of enterprises (and local units) at 1
J anuary each year.
Breakdowns – The data on new enterprises are broken down by broad economic activity
categories. The series on the stock of enterprises is not broken down, though annual
publications provide detailed breakdowns by size (employment and turnover), legal form,
economic activity and geography for individual years.
Metadata – Descriptive metadata about the Swedish statistical business register are available
athttp://www.scb.se/templates/Listning2____31034.asp , including some definitions.
80
Period covered – The data on new enterprises are available for 1996 to 2004. The data on
the stock of enterprises are available for 1971 to 2004, though include some discontinuities
(e.g. in 1996) due to changes in the scope of the administrative sources used for the statistical
business register.
Coverage – The data on the stock of enterprises cover all economic activities and legal
forms. It is not clear whether the data on new enterprises have the same coverage.
Definitions
• The stock of enterprises is on a point in time basis (1 J anuary).
• The unit used is the enterprise, though enterprises are defined as active legal units.
b) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2002. Data
for births exist for 1998 to 2002, and data for deaths exist for 1997 to 2002.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
27. Switzerland
One data source available
Title – Démographie des entreprises
Source – Office Fédéral de la Statistique, Switzerland
Internet address –http://www.bfs.admin.ch/bfs/portal/fr/index/themen/industrie_und_dienstleistungen/unternehm
en/blank/medienmitteilungen.htmlhttp://www.bfs.admin.ch/bfs/portal/fr/index/themen/industrie_und_dienstleistungen/uebersicht/
blank/publikationen.html?publicationID=853
Contents – Count and employment data for new enterprises. The second internet address
above gives counts of the stock of enterprises for years in which a census of enterprises has
been carried out.
81
Breakdowns – The data on new enterprises are broken down by economic activity. There are
various breakdowns available for data from censuses of enterprises.
Metadata – Methodological notes are included in the press release on new enterprises.
Period covered – Counts of new enterprises are available for 1999 to 2003. Data on the
stock of enterprises are available for 1985, 1995, 1998 and 2001.
Coverage – Economic activities in agriculture, forestry, fishing and public administration
(NACE sections A, B and L) are not covered. National and local government units are also
excluded.
Definitions
• The stock data are on a point in time basis
• The unit used is the enterprise
• New enterprises are defined as “ex nihilo” creations, i.e. pure births, with reactivations
and take-overs excluded.
28. Turkey
One data source available
Title – Newly Established and Liquidated Companies and Firms
Source – State Institute of Statistics
Internet address –http://www.die.gov.tr/english/SONIST/SIRKET/sirket.html, orhttp://www.die.gov.tr/TURKISH/SONIST/SIRKET/sirket.html (Turkish version)
Contents – Counts of newly established companies and co-operatives and newly established
firms, as well as liquidations of both.
Breakdowns – Some breakdowns by economic activity and geography are available for more
recent data.
Metadata – Some definitions are available in the SIS Data Dictionary
(http://www.die.gov.tr/TURKISH/SOZLUK/dataa.html).
Period covered – Data are available for 1995, and 1997 to 2004.
Coverage – All economic activities seem to be covered, though the counts for agriculture look
rather low. Central and local government units do not seem to be covered.
Definitions
• Firms are defined as “business establishments excluding companies and
cooperatives.”
• Companies are defined as “a number of persons forming an establishment for
commercial purposes as a result of economic and social joining.”
• Co-operatives are defined as “legal entity operating without fixed capital that may be
established by public institutions, provincial special administrations, municipalities,
associations or societies, whose aim is to provide certain economic benefits to
shareholders, especially in relation to their occupation and livelihood through aid and
solidarity.”
82
• Newly established companies and co-operatives, and liquidations, are those
announced in the Turkish Trade Register Gazette.
29. United Kingdom
Four data sources available
a) Title – Value-Added Tax Registrations and De-registrations
Source – Department for Trade and Industry – Small Business Service
Internet address –http://www.sbs.gov.uk/sbsgov/action/layer?r.l2=7000000243&r.l1=7000000229&r.s=tl&topicId
=7000011757
Contents – Counts of the stock of value-added tax (VAT) registered businesses, new
registrations and de-registrations.
Breakdowns – Data are broken down by economic activity and geography.
Metadata – A paper on the methodology used is available via the web site.
Period covered – The data are available for 1994 to 2003. A previous series from 1980 to
1993 is also available, but the data are not directly comparable due to large changes in the
VAT threshold.
Coverage – The data cover all economic activities and legal forms, though coverage is limited
for certain activities that are exempt from VAT, particularly in the education and health
sectors.
Definitions
• The stock data are on a point in time basis (1 J anuary).
• The unit used is the VAT registration, which approximates to the legal unit.
b) Title – Barclays Small Business Surveys
Source – Barclays
Internet address –http://www.business.barclays.co.uk/BRC1/jsp/brccontrol?task=articleFWgroup&value=6502&t
arget=_self&site=bbb
Contents – Data on business start-ups and closures, as well as the total population of
businesses are contained in a series of quarterly reports.
Breakdowns – The data are broken down in different ways each year, including by
geography, economic activity, and sex of the entrepreneur.
Metadata – Limited metadata are available within the reports.
Period covered – Data are available from 1995 to 2004, though for latter years they are
increasingly broadly rounded estimates.
83
Coverage – The data only cover England and Wales. They are based on business current
account openings and closures at Barclays, multiplied by estimates of their share of the
business banking market. This makes it unlikely that central and local government activities
will be included. Businesses that do not operate via business current accounts are also
excluded.
Definitions
• The unit used is the business account, which will be close to the definition of the
enterprise.
• The population data are on a point in time basis.
c) Title – Business demography indicators
Source – Eurostat
Internet address –http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
Contents – Population of active enterprises, births, deaths, survival and growth
Breakdowns – The data are broken down by country, economic activity, size and legal form
Metadata – A methodological manual exists, but is not yet published.
Period covered – Data for the population of active enterprises exist for 1997 to 2003. Data
for births exist for 1998 to 2003, and data for deaths exist for 1997 to 2003.
Coverage and definitions – See information on Eurostat standard coverage and definitions
at the end of this Annex.
d) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for the United Kingdom are available from 1986 to 1997, except 1992.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
84
30. United States
Six data sources available
a) Title – Statistics of US Businesses / Dynamic Data
Source – US Census Bureau
Internet address –http://www.census.gov/csd/susb/susbdyn.htm
Contents – Counts of establishment stock, births, deaths, expansions and contractions, and
associated employment changes.
Breakdowns – The data are broken down by economic activity, size band (based on
employment) and geography.
Metadata – Papers with descriptive metadata and definitions are available via the web site.
Period covered – Data are available for 1995 to 2001.
Coverage – Businesses without employees are excluded. All economic activities are covered
except crop and animal production (NAICS 111,112), rail transportation (NAICS 482), National
Postal Service (NAICS 491), pension, health, welfare, and vacation funds (NAICS 525110,
525120, 525190), trusts, estates, and agency accounts (NAICS 525920), private households
(NAICS 814), and public administration (NAICS 92). Governmental establishments are
excluded except for wholesale liquor establishments (NAICS 4228), retail liquor stores (NAICS
44531), Federally-chartered savings institutions (NAICS 522120), Federally-chartered credit
unions (NAICS 522130), and hospitals (NAICS 622).
Definitions
• The stock data are on a point in time basis (first quarter).
• The unit used is the establishment, defined as “a single physical location where
business is conducted or where services or industrial operations are performed.” This
is broadly equivalent to the local unit.
• Other units referred to are:
o Enterprise - A business organization consisting of one or more domestic
establishments that were specified under common ownership or control. The
enterprise and the establishment are the same for single-establishment firms.
o Firm - A business organization consisting of one or more domestic
establishments in the same state and industry that were specified under
common ownership or control. The firm and the establishment are the same for
single-establishment firms. For each multi-establishment firm, establishments
in the same industry within a state will be counted as one firm.
• Establishment births are establishments that have zero employment in the first quarter
of the initial year and positive employment in the first quarter of the subsequent year.
• Establishment deaths are establishments that have positive employment in the first
quarter of the initial year and zero employment in the first quarter of the subsequent
year.
b) Title – Firm Size Data
Source – US Small Business Administration
85
Internet address –http://www.sba.gov/advo/research/data.html
Contents – Counts of the population of firms, births and deaths. Employment data are also
available.
Breakdowns – Data on the population of firms are broken down by size band (employment)
and economic activity. There are no breakdowns of the data on firm births and deaths.
Metadata – Extensive metadata are available in the paper “Statistics of U.S. Businesses –
Microdata and Tables”, available on the website.
Period covered – Data on the population or firms are available for 1988 to 2002. Data on
births and deaths are available for 1989 to 2001.
Coverage – The coverage is as for source 1 above, as the firm level data are derived from
the US Census Bureau establishment-level Statistics of US Businesses.
Definitions
• The population counts cover all businesses that had an active payroll at any point
during the year, so can be considered as “live during period” data.
• The unit used is the firm, which is defined as “the largest aggregation of business legal
entities under common ownership or control”, so corresponds most closely to the
European definition of the Enterprise Group (truncated or all-residential rather than
global).
• Firm birth and death definitions correspond to those for establishments in source 1
above.
c) Title – Business Employment Dynamics, Quarterly Data
Source – Bureau of Labor Statistics
Internet address –http://www.bls.gov/bdm/home.htm
Contents – Counts and rates for establishment openings and closures each quarter.
Breakdowns – The data can be broken down by economic activity
Metadata – Descriptive metadata and definitions are available via the web site.
Period covered – Data are currently available from quarter 3 of 1992 to quarter 4 of 2004
inclusive.
Coverage – The data exclude business with no employees, central and local government
units, and some non-profit organizations. Certain economic activities are also excluded
(religious organizations, some small farms, the Armed Forces and railways).
Definitions
• No stock data are given, but these can be estimated from birth counts and rates (or
death counts and rates) for the same quarter. These can then be used to calculate
annual birth and death rates. Note; Birth and death data give slightly different stock
figures, but these are all within the margins of error associated with the use of rounded
86
data, and are unlikely to affect the annual birth and death rate estimates by more than
0.2%.
• The unit used is the establishment, which is broadly equivalent to the local unit.
• Openings are either establishments with positive third month employment for the first
time in the current quarter, with no links to the prior quarter, or with positive third month
employment in the current quarter following zero employment in the previous quarter.
• Closings are either establishments with positive third month employment in the
previous quarter, with no positive employment reported in the current quarter, or with
positive third month employment in the previous quarter followed by zero employment
in the current quarter.
d) Title – Business Employment Dynamics, Annualised Data
Source – Bureau of Labor Statistics
Internet address –http://www.bls.gov/opub/mlr/2004/11/art1full.pdf
Contents – This paper gives annualised versions of the data in source c) above, by removing
businesses that enter and exit within the year, and those entries that are really the
continuation of a previous registration.
Breakdowns – The data are not broken down in any way.
Metadata – The metadata for source c) mostly apply. The paper contains information on the
method for annualising the data.
Period covered – Data are available for 1998 to 2001 inclusive.
Coverage and Definitions – As for source c).
e) Title – Longitudinal Business Database
Source – US Census Bureau
Internet address –http://www.ces.census.gov/ces.php/abstract?paper=101647
Contents – The database contains linked records of establishments and firms over time. It
can be used to produce data on business dynamics. The internet address above is that of a
paper describing the database, which includes data on births and deaths. A second paper is
available with more detailed analyses for the retail sector – see:http://www.ces.census.gov/ces.php/abstract?paper=101704
Breakdowns – The data in the paper are not broken down in any way, but the database
would allow a range of detailed breakdowns.
Metadata – The paper contains descriptive metadata.
Period covered – The paper presents stock, birth and death data for 1976 to 1999.
Coverage – The source data cover establishments with paid employees. Economic
activity coverage is the same as for source number 1 above.
87
Definitions
• The stock data are on a point in time basis
• The unit used is the establishment
• Births are records that were active in one year, but not the previous year, adjusted for
reactivations
• Deaths are records that were active in one year, but not the next, adjusted for
reactivations
f) Title – OECD Firm-Level Data Project
Source – OECD
Internet address –http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
Contents – This source contains count and employment data for continuing businesses,
entries, exits and “one year” businesses
Breakdowns – Data are broken down by economic activity (ISIC 2-digit level).
Metadata – The website contains links to various papers containing descriptive metadata on
sources, methods and definitions.
Period covered – Data for the United States are available from 1989 to 1996.
Coverage and Definitions – See the general information on the OECD firm-level data project
at the end of this Annex.
31. Brazil
One data source available
Title – Estatísticas do Cadastro Central de Empresas - 2001
Source – Instituto Brasileiro de Geografia e Estatística (IBGE)
Internet address –http://www.ibge.gov.br/home/estatistica/economia/cadastroempresa/2000/Publicacao_comple
ta.pdfhttp://www.ibge.gov.br/home/estatistica/economia/cadastroempresa/2001/cempre2001.pdfhttp://www.ibge.gov.br/home/estatistica/economia/cadastroempresa/2002/cempre2002.pdf
Contents – Counts of enterprises and local units. The 2001 and 2002 publications also
contain data on births and deaths in specific sections on business demography.
Breakdowns – Data are broken down by economic activity, size and geography.
Metadata – Some descriptive metadata are available within the publications (in Portuguese).
See also the paper “Brazilian Enterprise Birth and Death rates by economic activity from 1997
to 2001” at:http://forum.europa.eu.int/Public/irc/dsis/businessurvey/library?l=/2003_rome/sessions7simpro
vingsbrsc&vm=detailed&sb=Title
88
Period covered – Data on the stock of enterprises are available for 1999 to 2002. Data on
births and deaths are available from 1997 to 2002.
Coverage – All economic activities and legal forms (including public administration) are
covered in the stock figures. The births and deaths for 1997 to 2001 cover “manufacturing”
(ISIC sections C +D), “trade” (ISIC section G), and services (ISIC sections H +I +J ). Birth
and death data for 2002 also include a category of “others” (ISIC sections A +B +E +Q)
Definitions
• The stock of enterprises is a point in time estimate at 31 December.
• The unit used is referred to as the enterprise, though is equivalent to the legal unit
• A birth in a given year is defined by the existence of an enterprise identification
number in the Business Register that was not found in the preceding year
• A death is the absence of an enterprise identification number that was found in the
previous year
• Birth and death rates were calculated dividing the number of births and deaths in each
year by the population of enterprises of the previous year.
• The birth and death study has not considered mergers and acquisitions as separate
demographic events. Also, as the business register is mainly based on administrative
records, if an enterprise fails to submit an administrative form in a certain year, this can
result in a false death followed by a false birth.
32. China
Data are being prepared from the 2004 economic census. Start-up rates are provisionally
estimated to be between 20% and 30%
Some data for corporate registrations in Hong Kong are available athttp://www.info.gov.hk/cr/key/index.htm
Eurostat Business Demography Indicators
Introduction
Eurostat have started a project to collect harmonised data on business demography from the
Member States of the European Union (EU). Romania and Norway have also participated on
a voluntary basis. The first data collections were in 2001, initially on a pilot basis. Data are
now available from 1997 to 2003, though not all EU countries have participated, and those
that have, have not provided data for all periods. Current data availability is shown in the table
below:
1997 1998 1999 2000 2001 2002
P B D P B D P B D P B D P B D P B D
Belgium X X X X X X X X
Czech
Republic X X X X X X X
Denmark X X X X X X X X X X X X X
Estonia X X X X X X X X X
Spain X X X X X X X X X X X X X X X X
Italy X X X X X X X X X X X X X X X X
Latvia X X X X X X X X X
89
Lithuania X X X X X X X X X
Luxembourg X X X X X X X X X X X X X X X X
Hungary X X X X X X X X X
Netherlands X X X X X X X X X X
Portugal X X X X X X X X X X X X X X X X
Slovenia X X X X X X X X X
Slovakia X X X X X X X X X
Finland X X X X X X X X X X X X X X X X
Sweden X X X X X X X X X X X X X X X X X
United
Kingdom X X X X X X X X X X X X X X X X X
Romania X X X X X X X X X
Norway X X X X X X X X X X X
Key:
P =Population data
B =Births data
D =Deaths data
X =Data available
Some data on survival and growth are also available.
The provision of data on business demography is likely to become compulsory for EU Member
States when the current draft revision of the EU Structural Business Statistics Regulation
comes into force, probably in 2006.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. Countries supplying data are also requested to provide
information on how closely they have followed the recommended methodology.
Coverage
The published data cover all economic activities in NACE sections C to K, except
management activities of holding companies (class 74.15). This means that agriculture,
forestry, fishing, public administration, health, education, other community, social and
personal service activities, activities of households, and extra-territorial organizations and
bodies are excluded. All legal forms are covered except central and local government, and
non-profit organisations serving households.
The coverage of data from individual countries is also influenced by the coverage of their
business registers, particularly in terms of size thresholds. These are in turn influenced by
national administrative sources, which vary considerably from country to country. For
example, the current value-added tax threshold in the United Kingdom is around €85,000,
whereas it is zero, or close to zero in most other countries. These differences can be partly
offset (as in the UK) by using a range of sources to improve coverage, but they still lead to
noticeable differences in data for the smallest size classes.
Definitions
• The statistical unit is the enterprise, however, the methodology used recognises that
some countries only hold data at the level of the legal unit, and attempts to
compensate for this through matching routines.
• The population of active enterprises consists of all enterprises that had either turnover
or employment at any time during the reference period, i.e. it is on a “live during
period” basis.
90
• Enterprise births are defined as the creation of a combination of production factors
with the restriction that no other enterprises are involved in the event. They include
enterprises started by a person who previously performed the same activity, but as an
employee, and newly born national or foreign subsidiaries that are real enterprises
(legal units rather than just local units or branches) with autonomy of decision making,
where new production factors are created, rather than transferred from another unit.
The following categories are excluded:
o Enterprises that are created by merging production factors or by splitting them
into two (or more) enterprises (break-ups, mergers, split-offs, restructuring)
o Newly created enterprises that simply take over the activity of a previously
created enterprise (take-over)
o Any creations of additional legal units/enterprises solely for the purpose of
providing a single production factor (e.g. the real estate or personnel) or an
ancillary activity (see note below) for an existing enterprise.
o An enterprise that is registered when an existing enterprise changes legal
form. E.g. a successful sole proprietor moves operations from his home to
another location and at the same time changes the legal form of the enterprise
to a limited liability company.
o Reactivated enterprises if they restart activity within 2 calendar years.
o Temporary associations and joint ventures that do not involve the creation of
new factors of production.
• Enterprise deaths are defined as the dissolution of a combination of production factors
with the restriction that no other enterprises are involved in the event. Events leading
to the closure of an enterprise that are not considered to be deaths are:
o Enterprises that close down due to merging or breaking-up of production
factors (break-ups, mergers, restructuring)
o Enterprises whose activity is taken over by another enterprise (take-over)
o Enterprises that are deleted due to a change of legal form, e.g. a successful
sole proprietor moving operations from his home to another location and at the
same time changing the legal form of the enterprise to a limited liability
company is a case that should be excluded.
o Reactivated enterprises if they restart activity within 2 calendar years.
The DOSME Project
Introduction
DOSME (Demography of Small and Medium-sized Enterprises) was the name given to a
series of projects to develop statistical business registers and data on business demography
and factors of success in a group of Central and Eastern European countries. The DOSME
projects were financed by the European Union, through Eurostat. Data were collected from
samples of new and existing enterprises between 1994 and 2000, and have been
consolidated into a single database held by Eurostat.
Details of the data collections, methodology and publications from this series of projects are
all available on the DOSME web site
30
. Some of the countries involved have continued the
data collections since the end of the project, but most have now started to participate in the
Eurostat Business Demography Project. Data are available for Bulgaria, Czech Republic,
Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia and Slovenia for 1995 to
2000. Some data are also available for Albania and Macedonia for earlier periods.
30
http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
91
Coverage – NACE Rev. 1 sections A, B and L (agriculture, forestry, fishing and public
administration) are excluded, as are central and local government units. The data are based
on survey results weighted using business register counts to give whole population estimates.
This limits the coverage of the final data sets in terms of the availability and quality of detailed
breakdowns.
Definitions
• The stock data used are point in time estimates, though approximations to “live during
period” populations have been possible by adding births during a particular year to the
stock at the start of that year.
• The unit used is theoretically the enterprise, but in practice it is usually the legal unit.
As the surveys focussed mainly on small units, the impact of this is probably
negligible.
• Births and deaths are defined as registrations and de-registrations with the relevant
administrative sources. It is recognised that these may be subject to lags, and that
some types of businesses were not required to register in certain countries.
The OECD Firm-Level Data Project
Introduction
The OECD firm-level project involved bringing together data from ten OECD countries (United
States, Germany, France, Italy, United Kingdom, Canada, Denmark, Finland, the Netherlands
and Portugal). It aimed to use on a common analytical framework, including the
harmonisation, to the extent possible, of key concepts (e.g. entry, exit, or the definition of the
unit of measurement) as well as the definition of common methodologies for studying firm-
level data.
The data were used to analyse firm demographics, resulting in a number of papers available
via the project web site:http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
The data were derived from business registers (Canada, Denmark, France, Finland,
Netherlands, United Kingdom and United States) or social security databases (Germany and
Italy). Data for Portugal were drawn from an employment-based register containing
information on both establishments and firms. These databases allow firms to be tracked
through time because addition or removal of firms from the registers (at least in principle)
reflects the actual entry and exit of firms.
Coverage
While some data sources included even single-person businesses, others omitted firms
smaller than a certain size, usually in terms of the number of employees, but sometimes in
terms of other measures such as sales (as is the case in the data for France and Italy).
Analyses based on the data from this project typically exclude single-person businesses.
However, because smaller firms tend to have more volatile firm dynamics, remaining
differences in the threshold across different country datasets should be taken into account in
the international comparison.
The data were compiled on an annual basis, covering varying time spans. The German,
Danish and Finnish register data cover the longest time periods, while data for the other
92
countries are available for shorter periods of time or, although available for longer periods,
include significant breaks in definitions or coverage.
Special efforts were made to organise the data along a common industry classification (ISIC
Rev.3). In countries where the data collection by the statistical agency varied across major
sector (e.g., construction, industry, services), a firm that switched between major sectors
could not be tracked as a continuing firm but ended up creating an exit in one sector and an
entry in another. Most countries have been able to provide firm demographic data across
most sectors of the economy, with the exception that public services are often not included
(the United Kingdom is a special case where data only refer only to manufacturing).
Definitions
• Unit of observation: Data used in the study refer to the firm as the unit of reference,
with the exception of Germany where data are only available with reference to
establishments, and Finland where data are reported with reference to both firms and
establishments.
• Firm entry: The number of firms entering a given industry in a given year. It comprises
firms observed as (out, in, in) the register in time (t – 1, t, t +1).
• Firm exit: The number of firms that leave the register. It comprises firms observed as
(in, in, out) the register in time (t – 1, t, t +1).
• Continuing firms: The number of firms that were in the register in a given year, as well
as in the previous and subsequent year. It comprises firms observed as (in, in, in) the
register in time (t – 1, t, t +1).
• One-year firms: The number of firms that were present in the register for only one
year. It comprises firms observed as (out, in, out) the register in time (t – 1, t, t +1).
93
Annex 3 - Defining Business Populations: Comparing Point in Time
and Live During Period Estimates
1. Introduction
The definition of a population of businesses can have a significant impact on any data derived
from it. This Annex looks at different ways in which business populations have been defined in
terms of the time dimension. It focuses specifically on the role of populations in business start-
up rates, but obviously has the potential for wider application. It concludes by proposing a
model to help improve the international comparability of business data. The terminology used
in this Annex is consistent with that proposed in Annex 1.
There are two main approaches to defining business populations with respect to time. A “point
in time” population is a relatively simple concept, and consists of all businesses deemed to be
in scope at a given point in time, usually on a specific reference day. A “live during period”
population, however, consists of all businesses that were in scope at any point during a given
reference period. This Annex aims to explain the nature of the differences between these two
approaches and look at ways to estimate the impact of using different populations. The focus
is on deriving methods to convert data compiled on one basis to provide more comparable
estimates.
It is clear that a live during period population will be larger than one constructed on a point in
time basis. The extent of the difference will depend on various factors, but mainly on the
length of the period, and the degree of churn (i.e. joiners and leavers) in the business
population. As a result, business start-up data compiled using a point in time population are
not likely to be comparable with those based on a live during period approach.
Differences in populations compiled using the two approaches are typically in the range of
10% to 15% for annual data. The choice of population can therefore affect start-up rates by up
to 2%, with the impact highest when start-up rates themselves are high. This issue is of
particular relevance for international comparability purposes, as business demography data
from Eurostat are compiled using live during period populations, whereas almost all other
sources favour the point in time approach.
2. Components of the population
Point in time and live during period populations of businesses can be broken down into a
number of components, which can then be re-aggregated in different ways to give different
types of population estimates. The basic components are shown in Figure 1 below.
94
Figure 1: A Simple Model for Business Populations
In this model:
• PA
t
=The population at the start of period t
• PB
t
=The population at the end of period t
• S
t
=businesses present in both populations (i.e. “survivors”)
• L
t
=businesses that are in population PA
t
, but not PB
t
(i.e. “leavers”)
• J
t
=businesses that are not in population PA
t
, but are in B
t
(i.e. “joiners”)
• J L
t
=businesses that are not present in PA
t
or PB
t
, but would be present in an
intermediate population (i.e. they join and leave within period t)
The population of businesses considered in scope at the start of the period (PA
t
), sometimes
referred to as the opening stock, can be defined as: PA
t
=S
t
+L
t
. Similarly the population at
the end of the period (PB
t
), or closing stock, can be defined as: PB
t
=S
t
+J
t
. As PB
t
=PA
t+1
, it
follows that: PA
t
=S
t-1
+J
t-1
, PB
t
=S
t+1
+L
t+1
, and that PB
t
=PA
t
+J
t
– L
t.
Businesses in the
sub-set J L
t
do not appear in either population.
Total entries to the population are defined as E
t
=J
t
+J L
t
, and total exits as X
t
=L
t
+J L
t
, so P
t
=PA
t
+E
t
=PB
t
+X
t
. Thus to convert from a point in time to a live during period population, it
is necessary to know, or have reasonable estimates for E
t
or X
t
, or J L
t
and one of L
t
or J
t
. In
practice, J L
t
is rarely available from published data sources, and such businesses are usually
ignored as they are not present in PA
t
or PB
t
, thus there is a risk that P
t
could be
underestimated. If they are included, data usually take the form of E
t
and / or X
t
so they
appear in both, rather than as a separate group. The latter is usually the case in data sets
based on population observations at a series of intermediate points between PA
t
and PB
t
. The
size, and hence the importance of J L
t
will depend on the length of period t. If t is one month, it
is relatively safe to assume that J L
t
is very small. If PA
t
and PB
t
are derived from economic
censuses with a five year interval, however, J L
t
will be much larger.
Eurostat “live during period” enterprise survival data covering 48 observations for 18 countries
over 4 years show that on average 87.23% of births in a given year are also active in the
following year. This indicates that the size of J L
t
is typically 14.64% of that of J
t
, where t is one
year. Thus for annual estimates, where data on J
t
are available, but data on J L
t
are not, it
would be reasonable to define entries as E
t
=J
t
(1 +0.1464) and estimate P
t
or PA
t
using: P
t
=
PA
t
+J
t
(1 +0.1464). The value of 0.1464 is the best current estimate of the ratio of J L
t
to J
t
,
for European Union countries, and will obviously vary over time and space, so this value is
replaced by c in the remainder of this Annex.
t
S
t
L
t
J
t
J L
t
PA
t
PB
t
95
3. Other Models
3.1 OECD Firm-level Data Project
Figure 2: The OECD Firm-level Data project Model
Note – different notation is used in this model:
S
t
=Survivors for year t (defined as in / in / in, for t-1 / t / t+1)
X
t
=Exits for year t (defined as in / in / out, for t-1 / t / t+1)
E
t
=Entries for year t (defined as out / in / in, for t-1 / t / t+1)
O
t
=One year firms (defined as out / in / out, for t-1 / t / t+1)
A point in time approach was used for this data collection project. Firms that entered and
exited between observations are not recorded, and “one year firms” are also excluded from
most analyses due to data quality concerns. This effectively defines firm entries and exits only
in terms of firms that existed for at least one year.
This project also took the relatively unusual approach of identifying entries for period t as firms
that appeared during the period t-1 to t, whereas exits are defined as those that disappeared
between t and t+1.
t -1 t +1 t
S
t
O
t
X
t-1
E
t
X
t
O
t-1
E
t+1
O
t+1
96
3.2 Eurostat Business Demography
Figure 3: The Eurostat Business Demography Model
Notation:
t =Reference year
N
t
=Population of enterprises active at any point during t
R
t
=Real births in t
X
t
=Other entries in t
D
t
=Real deaths in t
Y
t
=Other exits in t
The Eurostat model in Figure 3 is based on the “live during period approach”. This project is
ongoing, and will be extended to all European Union (EU) countries following a revision to the
EU Structural Business Statistics Regulation. The definitions of the various populations do not
take into account the length of survival. A business can be a birth and a death in the same
period, and will also be counted in the population of active enterprises for that period. A
methodology based on checking for reactivations (within two years), then matching, then
manual inspection of large units, is used to separate “real” (i.e. pure) births and deaths from
other entries and exits.
3.3 National Data Models
A wide variety exists in the models used to define the populations used for national business
demography data. Some of the differences are purely down to terminology or notation, and
most models can be seen as derivatives of one or more of those presented above. For
example, most sources in the United States use models similar to that shown in Figure 1,
whereas the Australian model mixes elements from Figure 1 and Figure 3.
4. Purity of Entries and Exits
A complicating factor for many population models is that for business demography purposes,
it is often of interest to split total entries (E
t
) and exits (X
t
) into a number of sub-components
linked to different types of demographic events. A basic distinction for many data users, as
t
N
t
D
t
X
t
R
t
Y
t
97
seen in the Eurostat model, is to split entries into pure births
31
(also referred to as creations
ex-nihilo), and other entries (e.g. due to restructuring, re-registration, reactivation, merger,
break-up or split-off), with a similar split for exits into pure deaths and other exits. If more
detail is required, and available, each different demographic event can form a separate sub-
component of entries and exits.
32
The model in Figure 4 assumes a simple split of entries and exits on the basis of purity, i.e.
separating pure births (B
t
) from other entries (OE
t
), and pure deaths (D
t
) from other exits
(OX
t
). It also assumes that the length of survival of the business is not relevant, in that
businesses are included whether or not they survive into the subsequent period.
Figure 4: Developing the Concept of Purity
Unfortunately, one problem with the live during period approach is that a proportion of other
entries (OE
t
) will be due to new businesses taking over the activities of businesses recorded
as other exits (OX
t
). Technically, many of these cases should be considered as the continuity
of a business, and should not be recorded as entries and exits. However, as most data
sources are based either directly or indirectly on registrations and de-registrations with
administrative or tax sources, it is unlikely that all such take-over cases are recorded as
business continuity, particularly for small businesses. This will vary from country to country
and between sources, depending on the nature of the source and register maintenance
procedures. The way in which business continuity is treated in the source will therefore affect
the degree of duplication in a live during period population. This, in turn, will affect the
comparability of indicators based on live during period populations.
31
Defined in the Eurostat Business Demography Methodological Manual as: “the creation of a
combination of production factors with the restriction that no other enterprises are involved in the event.
Births do not include entries into the population due to: mergers, break-ups, split-off or restructuring of a
set of enterprises. It does not include entries into a sub-population resulting only from a change of
activity”.
32
A typology of demographic events is proposed in Chapter 13 of the Eurostat Business Registers
Manual of Recommendations, see:http://forum.europa.eu.int/irc/dsis/bmethods/info/data/new/embs/registers/chapter13.doc
t
E
t
OX
t
OE
t
B
t
D
t
X
t
=X
t
E
t
=
98
5. Towards a Standard Model for Business Populations
So far, this Annex has concentrated on understanding and explaining the different models
used to define populations for business demography purposes. The next logical step is to look
for ways to move towards common standards, with the aim of improving international data
comparability.
A major constraint is that any changes to current methods are likely to have costs both
financially and in terms of comparability of data series over time. Thus an approach is
required that minimises the impact of any change, whilst maximising comparability and
standardising terminology. It also needs to recognise the requirements of splitting entries
based on both purity and survival through an observation point.
Figure 5: A Standard Model?
t
S
t
D
t
OE
t
BS
t
OX
t
PA
t
PB
t
BD
t
BOX
t
OED
t
OEOX
t
B
t
OES
t
SD
t
SOX
t
B
t
=
OE
t
=
OED
t
BD
t
=D
t
BOX
t
OEOX
t
=OX
t
99
The point in time population at the start of period t (PA
t
) can be defined by rearranging the
equation P
t
=PA
t
+E
t
from Section 2, to give PA
t
=P
t
– E
t
. The live during period population
(P
t
) is defined in the same way as in the Eurostat model (referred to there as N
t
). E
t
gives the
total number of new businesses in period t, and is equivalent to B
t
+OE
t
(pure births plus
other entries) in Figure 4. Therefore PA
t
can be defined as PA
t
=P
t
– (B
t
+OE
t
), or using the
Eurostat notation, PA
t
=N
t
– (R
t
+X
t
).
33
Eurostat do not currently publish data on X
t
, but as
this is a by-product of the data production process it is relatively easy to obtain. In this way,
consistent measures of PA
t
can be defined based on data from a wide range of models.
PA
t
can also be defined as the survivors throughout period t plus the leavers in period t that
were live in the previous period. From Figure 1; PA
t
=S
t
+L
t
, however, in Figure 5, L
t
is split
into SD
t
(survivors from the previous period that were pure deaths in t) and SOX
t
(survivors
from the previous period that were other exits in t), so in terms of Figure 5, PA
t
=S
t
+SD
t
+
SOX
t
.
Either P
t
or PA
t
can be used as the basis for business demography statistics. The point in time
approach (PA
t
) is currently the more popular of the two, and has the advantages of being a
relatively simple concept to explain, and being analogous to the population basis used for
human demography statistics. It also largely avoids the potential duplication issues that can
affect the comparability of P
t
(see Section 4 of this Annex). For these reasons, the use of PA
t
is recommended, though it should be remembered that it is relatively simple to substitute P
t
if
required.
Having defined a stock population, the next step is to determine the dynamic populations of
entries and exits. In Figure 1, the sum of the entries in period t is defined as J
t
+J L
t
(or
J
t
(1+c)). The total number of entries in period t from Figure 5 is B
t
+OE
t
, which is equivalent to
R
t
+X
t
in the Eurostat notation (Figure 3). Thus a standard measure of entries (E
t
) can be
calculated reasonably easily from all sources.
Unfortunately, national factors (e.g. units, sources, coverage, definitions, thresholds etc.), and
other movements into and out of scope, can influence total entries, reducing the value of this
variable for international comparison purposes. A more comparable variable should be the
number of pure births, but this assumes perfect knowledge to separate pure births from other
entries. Much progress in developing a standard methodology for this process has been made
through the Eurostat business demography project, such that reasonable estimates of B
t
, and
OE
t
are now possible for a number of European countries.
Figure 5 breaks down B
t
and OE
t
(and the corresponding variables for leavers, D
t
and OX
t
)
into three components. B
t
can be seen as consisting of businesses that survive into t+1 (BS
t
),
and those that do not. The latter category can be broken down into pure deaths (BD
t
) and
other exits (BOX
t
). If similar components are derived for OE
t
, D
t
and OX
t
(note: some
components are shared), these components can be re-grouped to form populations J
t
(=BS
t
+
OES
t
), L
t
(=SD
t
+SOX
t
), and J L
t
(=BD
t
+BOX
t
+OED
t
+OEOX
t
).
Similar issues apply to the way the population of leavers is defined. D
t
offers purity, and the
potential for greater comparability, whereas L
t
may be easier to measure in practice. To
facilitate data conversion, it will also be necessary to calculate an equivalent to the value c to
express J L
t
as a proportion of L
t
.
One issue not covered so far is how to deal with reactivations, i.e. businesses that leave a
population (by closing temporarily) then re-join it. Recording these as a death followed by a
33
The robustness of the relationship described by these equations depends on the extent of duplication
in P
t
(see the last paragraph of Section 4). The higher the degree of duplication, the less robust the
relationships.
100
birth does not fit well with the purity approach, particularly if the period of closure is short. For
Eurostat business demography purposes, a two year threshold is applied, so that periods of
closure of less than two years do not result in deaths and births.
If the period of temporary closure does not include a point at which the population is
observed, i.e. the temporary closure starts and ends between the dates of PA
t
and PB
t
, it may
not be recorded. This means that, if t is one year, in theory a business can be inactive for over
11 months, yet still be recorded as having survived throughout the period, whereas a business
that closes for a few days either side of the PB
t
reference date would be recorded as a leaver
in t, and a joiner in t+1.
To improve consistency, a rule that a business has to be out of the population for at least two
consecutive observations to be considered a pure death and birth seems reasonable where t
is one year. Thus a reactivation that was out of the population for just one period (e.g. PA
t
)
would be included in the populations for other events (SOX
t-1
, and OES
t
) rather than those for
pure births and deaths. The only real problem with this approach is that it introduces a lag for
data on deaths, though this can be at least partially overcome by estimation.
For completeness, in addition to the total population, entries and exits, it is useful to determine
the population of businesses that survive throughout the period (or are at least present at the
start and the end of the period), S
t
. S
t
can be defined simply (from Figure 1) as PB
t
– J
t
. To
relate this to the Eurostat populations in Figure 3, it is necessary to refer to PB
t
as PA
t+1
,
which has been shown above to correspond to N
t+1
– (R
t+1
+X
t+1
). J
t
has been shown to equal
to (R
t
+X
t
) / (1+c). Thus, by substitution, S
t
=N
t+1
– (R
t+1
+X
t+1
) – ((R
t
+X
t
) / (1+c)).
Having derived the various populations of interest, it is useful to note certain logical
relationships based on a stock and flow basis, which can be used as a quality check, or to
derive a missing population. The basic equation is that opening stock, plus entries, minus
exits should equal closing stock: PA
t
+E
t
– X
t
=PB
t
(it also follows that PA
t
+J
t
– L
t
=PB
t
). In
terms of Figure 5, this equation can also be expressed as PA
t
+(BS
t
+OES
t
) – (SD
t
+SOX
t
) =
PB
t
.
Having defined the populations referred to above, it is then relatively straightforward to apply
them to study business survival and growth rates, though this is beyond the scope of this
Annex. The proposed standard model also implies the introduction of harmonised
terminology, the various elements of which are defined in Annex 1.
6. Conclusions
It seems feasible to apply a standard model for defining business populations that can accept
data from a variety of sources using different methodologies. As a result, it should be possible
to improve the international comparability of data on business populations, business
demography, and small and medium–sized enterprises (SMEs), whilst not imposing significant
additional burdens on data suppliers in national statistical institutes.
This approach will remove, or at least reduce the impact of a number of the different factors
affecting comparability identified in this report. The next logical step is therefore to test the
proposed model, and refine it where necessary, using data from as many countries as
possible.
101
Annex 4 – Business Start-up Data for Selected Countries:
Comparisons of National Sources
Introduction
This Annex looks at intra-country comparability of business start-up data for 10 countries for
which three or more sources allowing the construction of start-up rates have been identified in
the inventory in Annex 2. The purpose of this exercise is to try to explain the differences
between sources in terms of the factors of comparability identified in the main body of this
report. This work is based on the assumption that any differences in data relating to the same
country and the same time period must be purely methodological in nature.
To facilitate comparability the data shown have been summarised, and converted to a
standard format. This has, in some cases, included the calculation of birth and death rates.
1. Canada
Three sources of data on business start-ups have been identified for Canada.
a) Business Dynamics in Canada
Statistics Canada – Longitudinal Employment Analysis Program (LEAP) Database
(supplemented for 2002 by data prepared for FORA).http://www.statcan.ca:8096/bsolc/english/bsolc?catno=61-534-X
This publication includes data on business populations, birth and death rates, and survival. It
also contains a chapter on methodology. Business populations, births and deaths are broken
down by size and economic activity categories based on knowledge intensity. Business
populations are also broken down by geography.
Population estimates are available for 1991 to 2001 and counts of births are available from
1992 to 2001. Additional data on population and births for 2002, and deaths for 2001 is taken
from a short report prepared by Statistics Canada for FORA.
The data cover all employers in Canada, public and private (i.e. data do not include
businesses with no employees). The unit used is the firm, which, at the national level is
equivalent to the legal unit. Births are defined as firms that are not present on the LEAP
database in year t-1, but are present in year t. The birth rate is the number of new enterprises
in t divided by the total number of firms observed in year t.
The data in the publication are protected by copyright, so are not reproduced here.
b) Self-Employment Entry and Exit Flows
Statistics Canada – Paper by Zhengxi Lin, Garnett Picot and J anice Yateshttp://www.statcan.ca/english/research/11F0019MIE/11F0019MIE1999134.pdf
This paper contains a table with counts of self-employed persons, and rates for entry and exit,
as well as other related data, analyses, and descriptive metadata on sources and definitions.
No breakdowns of the entry data are given in the paper. The population estimates are
available for 1981 to 1995, and entries are available from 1982 to 1995.
102
The data cover all persons whose self-employment earnings are the dominant source of
earnings in the year according to their annual tax returns to revenue Canada, but are based
on a 10% sample. The unit is therefore the person completing the tax return. Self-employment
entries are income-tax filers who report earnings from self-employment in one year but not the
previous year.
Population Entries Birth Rate
1981 915,140
1982 931,240 194,750 20.91%
1983 953,350 197,700 20.74%
1984 988,590 208,030 21.04%
1985 990,980 197,280 19.91%
1986 1,019,390 221,760 21.75%
1987 1,069,690 248,630 23.24%
1988 1,099,470 248,370 22.59%
1989 1,125,410 253,710 22.54%
1990 1,159,370 269,500 23.25%
1991 1,191,930 273,190 22.92%
1992 1,253,290 293,330 23.40%
1993 1,334,050 312,620 23.43%
1994 1,400,760 330,810 23.62%
1995 1,471,800 355,940 24.18%
c) OECD Firm-level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
Canada are available for 1984 to 1997.
Annual point in time populations were used, based on the Canadian statistical business
register. New businesses had to be present in both the reference year and the following year
to be counted as a birth in published analyses of the data. “One-year” businesses were
identified separately, but have been included as births in the table below to try to improve
comparability with other sources. Businesses without employees were excluded
Year Population Total Entries Start-up Rate
1984 701,115 128,837 18.38%
1985 729,929 119,424 16.36%
1986 757,980 117,843 15.55%
1987 779,956 116,916 14.99%
1988 781,594 103,998 13.31%
1989 783,415 103,096 13.16%
1990 798,855 121,773 15.24%
1991 796,223 115,662 14.53%
1992 798,215 113,740 14.25%
1993 801,127 114,308 14.27%
1994 808,849 117,512 14.53%
1995 810,336 88,301 14.46%
1996 806,777 92,570 14.72%
1997 825,389 92,686 15.07%
103
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
Self-employment Business Dynamics in Canada OECD Firm-level Data Project
The chart shows an apparently high degree of convergence for data from the OECD Firm-
level project and Business Dynamics in Canada for the years where the data overlap. As
would be expected, birth rates are much higher for self-employment businesses. Data before
1990 show rather more variability, which may be genuine, or may indicate that these data are
less reliable.
b) New Businesses
0
50
100
150
200
250
300
350
400
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
T
h
o
u
s
a
n
d
s
Self-employment Business Dynamics in Canada OECD Firm-level Data Project
104
This chart shows that whilst start-up rates were more or less equivalent for data from the
OECD firm-level project, and Business Dynamics in Canada, the levels of new businesses are
not so close, though both follow a similar, rather stable trend during the period of overlap.
There is no readily apparent explanation for the difference between these sources in the
metadata, though it is almost certainly due to different coverage, possibly of the public sector.
The number of new self-employed businesses shows a rapid growth during the 1990’s,
suggesting that self-employed businesses are becoming increasingly important in Canada.
c) Business Populations
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
M
i
l
l
i
o
n
s
Self-employment Business Dynamics in Canada OECD Firm-level Data Project
The patterns here are quite similar to those for new businesses, with a rapid growth of self-
employment businesses, and steady, parallel trends for the employer businesses covered by
the other two sources.
105
2. Denmark
Three sources of data on business start-ups have been identified for Denmark.
a) Statistical Yearbook
Statistics Denmark:
2005 (data for 2001) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2005.aspx
2003 (data for 2000) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2003.aspx
2001 (data for 1999) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2001.aspx
2000 (data for 1998) -http://www.dst.dk/HomeUK/Statistics/ofs/Publications/Yearbook/2000.aspx
The yearbooks contain counts of new enterprises broken down by economic activity, for 1998
to 2001, as well as the end-year population (used as the start population for the following year
in the table below). They contain very little metadata. Data exclude agriculture and public
administration, and the unit used is the enterprise.
Year Population New Enterprises Start-up Rate
1998 16,063
1999 279,037 17,734 6.36%
2000 284,446 18,640 6.55%
2001 284,166 16,447 5.79%
2002 281,653
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Danish data for the population of active enterprises are available for 1997 to 2001, and data
on births cover 1998 to 2001.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Danish statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
106
Year Population Births Birth Rate
1997 243,946
1998 245,762 24,755 10.07%
1999 253,887 27,562 10.86%
2000 261,911 26,137 9.98%
2001 261,926 24,275 9.27%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
Denmark are available for 1981 to 1994.
Annual point in time populations were used, taken at the end of November each year from the
Danish pay and performance database. New businesses had to be present in both the
reference year and the following year to be counted as a birth in published analyses of the
data. “One-year” businesses were identified separately, but have been included as births in
the table below to try to improve comparability with other sources. Businesses without
employees were excluded
Year Population Total Entries Start-up Rate
1981 136512 21334 15.63%
1982 136373 18148 13.31%
1983 137585 18162 13.20%
1984 139760 18414 13.18%
1985 140914 17961 12.75%
1986 138496 13609 9.83%
1987 140826 16054 11.40%
1988 138772 15491 11.16%
1989 136380 14099 10.34%
1990 135448 14857 10.97%
1991 133565 15420 11.54%
1992 132187 14778 11.18%
1993 126070 13282 10.54%
1994 126113 12567 9.96%
107
Graphical Comparisons
a) Birth Rates
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
OECD Firm-level Data Statistical Yearbook Eurostat Business Demography
All sources seem to indicate a downwards trend in birth rates over time. The difference
between the data from the Statistical Yearbook and those from Eurostat is not easy to explain
based on the available metadata, but may be related to coverage.
b) New Businesses
0
5
10
15
20
25
30
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Statistical Yearbook Eurostat Business Demography
The number of new businesses in the OECD firm-level data set is generally lower than the
other sources due to the exclusion of non-employer businesses, and of new businesses that
did not survive for at least a year. Again the data from Eurostat are higher than those from the
statistical Yearbook, which is not easy to explain given the greater restrictions of the Eurostat
data in terms of coverage of economic activity, as well as the requirements for a relatively high
level of purity in this source.
108
c) Business Populations
0
50
100
150
200
250
300
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Statistical Yearbook Eurostat Business Demography
The one employee threshold used for the OECD firm-level data clearly has an impact on the
population, appearing to remove around half of the units included in the other sources.
Interestingly, the Statistical Yearbook data show higher population levels than those from
Eurostat. This appears to be rather counter-intuitive compared to the number of new
businesses. The Eurostat population would normally be expected to be larger as it is on a live
during period basis, but the impact of this seems to be outweighed by the relatively limited
coverage of economic activities.
109
3. Finland
Three sources of data on business start-ups have been identified for Finland.
a) Enterprise Openings and Closures
Statistics Finland –http://www.stat.fi/til/aly/index_en.html, Data only accessible via Finnish
version -http://www.stat.fi/til/aly/index.html
This source contains counts of enterprise openings and closures. Stock figures are available
separately from the StatFin database, but may not have the same coverage. The data are
broken down by economic activity, legal form and geography, and are available for 1999 to
2004. Some metadata are available, mostly in Finnish.
The openings data are derived from Statistics Finland’s business register. They only cover
those enterprises engaged in business activity that are liable to pay value-added tax or act as
employers. Foundations, housing companies, voluntary associations, public authorities and
religious communities are excluded. The data cover state-owned enterprises, but not those
owned by municipalities. The unit used is the enterprise
Year Stock Openings Birth Rate
1999 219,516 21,460 9.78%
2000 222,817 22,361 10.04%
2001 224,847 21,942 9.76%
2002 226,593 22,190 9.79%
2003 228,422 23,886 10.46%
2004 253,617 24,756 9.76%
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Finnish data for the population of active enterprises exist for 1997 to 2002 and data on births
cover 1998 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Finnish statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
110
Year Population Births Birth Rate
1997 229,786
1998 234,521 19,659 8.38%
1999 233,380 17,581 7.53%
2000 233,451 16,614 7.12%
2001 235,746 16,841 7.14%
2002 237,065 17,174 7.24%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
Finland are available for 1989 to 1997.
Annual point in time populations were used, taken from the Finnish statistical business
register. The coverage of the register improved in 1994 for smaller enterprises, which may
account for at least part of the peak in birth rates in 1994/5. New businesses had to be
present in both the reference year and the following year to be counted as a birth in published
analyses of the data. “One-year” businesses were identified separately, but have been
included as births in the table below to try to improve comparability with other sources.
Businesses without employees were excluded.
Year Population Total Entry Start-up Rate
1989 231,311 46,791 20.23%
1990 252,426 30,207 11.97%
1991 210,501 17,271 8.20%
1992 209,982 25,745 12.26%
1993 179,549 17,961 10.00%
1994 176,804 21,830 12.35%
1995 194,092 25,018 12.89%
1996 202,085 18,883 9.34%
1997 207,008 14,991 7.24%
111
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
Eurostat Business Demography Enterprise Openings and Closures OECD Firm-level Data
The Enterprise Openings and Closures data show a fairly similar trend to those from Eurostat
during the period of overlap. The higher levels of the former are likely to be due mainly to the
use of a point in time population, and the greater purity of the Eurostat data. The fluctuations
in the OECD firm-level data look odd compared to the relative smoothness of the other two
series. This may be partly due to the turbulence in the Finnish economy in the early 1990’s,
but it also seems likely that there is rather a lot of “noise” in the data.
b) New Businesses
0
5
10
15
20
25
30
35
40
45
50
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
Eurostat Business Demography Enterprise Openings and Closures OECD Firm-level Data
Again the OECD firm-level data show much more variability, with the figure for 1989 looking
particularly implausible. The levels also seem rather high given that the metadata state that
businesses without employees were excluded, particularly as that the Eurostat data indicate
that around 60% of births had no employees. As mentioned above, purity is likely to account
for much of the difference in levels between the Enterprise Openings and Closures and the
112
Eurostat data, though the limited coverage of economic activities in the latter will also play a
part.
c) Business Populations
0
50
100
150
200
250
300
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
Eurostat Business Demography Enterprise Openings and Closures OECD Firm-level Data
In terms of the business population, the OECD firm-level data show much more stability,
particularly for latter years, though the level still looks too high if businesses with no
employees are really excluded. The other two sources are close in both trend and level,
suggesting that the limited coverage of the Eurostat data is cancelled out by the effects of
using a live during period population. The 2004 figure for the Enterprise Openings and
Closures data looks rather high, and represents a significant deviation from the trend.
Unfortunately it is too early to tell whether or not it will be confirmed by the Eurostat data.
113
4. France
Three sources of data on business start-ups have been identified for France
a) Créations d'entreprises
INSEE:http://www.insee.fr/fr/ffc/chifcle_liste.asp?theme=9&soustheme=1&souspop=
This source contains counts of enterprise creations, split into new creations, resumptions and
re-activations, as well as some population data. Data are broken down by economic activity,
size and legal form. Some metadata are available, for example key definitions. Data are
available from 1993 to 2004, and cover all of France, including the overseas départements.
Three categories of enterprise creation are identified:
• Pure creations (creations “ex nihilo”) where the new enterprise does not take over the
activities of a previously existing enterprise.
• Reactivations, where a person who has previously been self-employed re-starts a self-
employed activity.
• Resumptions, where a new business takes over an activity previously carried out by
another enterprise.
Year Stock Creations Creation Rate Pure Creations Pure Creation Rate
1993 2,307,638 272,264 11.80% 169,620 7.35%
1994 292,847
1995 283,608
1996 273,811 170,233
1997 269,430 165,277
1998 264,601
1999 266,919
2000 270,043 174,718
2001 2,417,950 268,619 11.11% 175,140 7.24%
2002 2,468,786 268,459 10.87% 176,378 7.14%
2003 2,498,082 291,986 11.69% 197,675 7.91%
2004 2,568,647 318,757 12.41% 222,747 8.67%
b) La Création en Chiffres
Agence Pour la Création d’Entreprises (APCE):http://www.apce.com/index.php?rubrique_id=261&type_page=I
This source contains counts of enterprise creations, split into new creations (“ex nihilo”),
resumptions and re-activations. No breakdowns are given, and metadata are very limited.
Data are available from 1993 to 2004, and are very similar to, but slightly higher than those
from INSEE, perhaps suggesting a slight timing or coverage difference.
114
Year Total Creations Creations Ex Nihilo
1993 273,462 170,919
1994 294,131 183,764
1995 284,853 178,923
1996 275,275 171,628
1997 271,088 166,850
1998 266,446 166,190
1999 268,919 169,674
2000 272,072 176,754
2001 270,564 177,015
2002 270,206 178,008
2003 293,840 199,399
2004 320,265 223,995
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for
France are available for 1990 to 1996.
Annual point in time populations were used, taken at the end of the year from a fiscal
database and an enterprise survey. New businesses had to be present in both the reference
year and the following year to be counted as a birth in published analyses of the data. “One-
year” businesses were identified separately, but have been included as births in the table
below to try to improve comparability with other sources. Manufacturing businesses with an
annual turnover of less than 3.8 million French Francs, and service businesses with a turnover
of less than 1.1 million French Francs are excluded.
Year Population Total Entry Start-up Rate
1990 474,118 100,596 21.22%
1991 477,666 66,814 13.99%
1992 505,580 77,098 15.25%
1993 488,757 68,424 14.00%
1994 516,730 90,544 17.52%
1995 505,871 58,727 11.61%
1996 493,432 50,560 10.25%
115
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
OECD Firm-level Data SIRENE Creations SIRENE Pure Creations
The SIRENE data clearly show the impact of correcting for purity when the other comparability
factors are held constant. The pure creation rate is consistently almost 4% lower than the total
creation rate. As with other countries there is considerable variability in the start-up rate from
the OECD firm-level data, and for the one year of overlap, 1993, the rate is higher than that
for the SIRENE data. This could be a result of the high turnover threshold used for French
data in this source.
b) New Businesses
0
50
100
150
200
250
300
350
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
OECD Firm-level Data SIRENE Creations SIRENE Pure Creations
APCE Creations APCE Creations "ex nihilo"
For this chart, it is possible to add the APCE Creations data. As can be seen, the APCE data
very closely follow those from SIRENE, such that the APCE creations “ex nihilo” seems to be
116
a good indicator for the missing variables in the SIRENE pure creations series. For the four
years of overlap, the trend followed by the OECD firm-level data seems reasonably correlated
to those of the other sources, with the difference in levels clearly attributable to the threshold
in the OECD data.
c) Business Populations
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data SIRENE
This chart is less interesting given the lack of overlap. Both sources follow a stable to slightly
rising trend, and the clear impact of the threshold in the OECD data is again visible.
117
5. Germany
Three sources of data on business start-ups have been identified for Germany
a) Business Notifications
Federal Statistics Office, Germany:http://www.destatis.de/themen/e/thm_unternehmen.htm
This source contains data on business registrations, modifications and de-registrations, as
well as counts of the stock of businesses liable to pay turnover tax. Registration data are
available for 2001 to 2003. Data on the population of businesses liable for tax are available for
2002 and 2003. The data are broken down by economic activity.
Some descriptive metadata are available via the website. The data on new registrations are
assumed to cover the whole economy. The population of businesses liable to pay turnover tax
covers businesses with a turnover of at least €16,620 per year. Most economic activities are
covered, with the exception of certain health, public administration, insurance, and agricultural
activities.
The unit for registrations and de-registrations is effectively the local unit as “the obligation to
report business registrations and de-registrations applies to enterprises, branch offices and
dependent sub-offices”. Registration is required when a new activity is started or a business is
taken over, be it through purchase or succession, a partner entering the business, a change in
legal form, or a relocation of the business to a different registration district. Thus quite a high
proportion of registrations will not be pure births.
Year Population Business Registrations Birth Rate
2001 728,978
2002 2,926,570 723,333 24.72%
2003 2,915,482 810,706 27.81%
b) Start-ups and Liquidations in Germany
Institut für Mittelstandsforschung (IfM), Bonn:http://www.ifm-bonn.org/dienste/gruendungen-
engl.htm
This source contains counts of business start-ups and liquidations based on notifications of
new businesses. The data are adjusted to remove new sites of existing businesses,
registrations purely for tax or administrative purposes that do not result in new business
activity, and registrations for activities carried out as a second job by the entrepreneur.
Enterprise population totals for some years are available in Table 1 of:http://www.ifm-
bonn.org/dienste/kap-2.pdf. These do not provide a full series, but IfM have been able to
provide additional data (1992, 2000 – 2003) or suggest appropriate approximations (1991,
1993 and 1995) to fill the gaps.
The start-up data are broken down into the former East and West Germany, and are available
for 1991 to 2004. The population of businesses is subject to a threshold (€17,500 since 2003),
and covers all economic activities except the “liberal professions”
34
, most health services and
some insurance services that are not subject to VAT. The units used are effectively the sub-
set of legal unit that are considered to be economically relevant.
34
The liberal professions can generally be defined as occupations requiring special training in the arts
or sciences. These include lawyers, notaries, accountants, architects, engineers and pharmacists.
118
Year Population Start-ups Birth Rate
1991 2,572,202 531,000 20.64%
1992 2,631,812 494,000 18.77%
1993 2,709,443 486,000 17.94%
1994 2,787,074 493,000 17.69%
1995 2,775,000 528,000 19.03%
1996 2,762,925 507,000 18.35%
1997 2,797,759 507,100 18.13%
1998 2,859,983 512,800 17.93%
1999 2,886,268 493,100 17.08%
2000 2,909,150 471,700 16.21%
2001 2,920,293 454,700 15.57%
2002 2,926,570 451,800 15.44%
2003 2,915,482 507,900 15.54%
2004 572,600
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data, which
cover only the former West Germany, are available for 1978 to 1998, though data on entries
for 1978 are clearly incomplete.
Annual point in time populations were used, taken from a social security database. The unit
used is referred to as the “plant”, thus is likely to be closer to the definition of the local unit
than to that of the enterprise. Only businesses with one or more employees are included, and
certain public sector units are considered out of scope. New businesses had to be present in
both the reference year and the following year to be counted as a birth in published analyses
of the data. “One-year” businesses were identified separately, but have been included within
“Total Entry” in the table below to try to improve comparability with other sources.
Year Population Total Entry Start-up Rate
1978 1,320,297 2,292 0.17%
1979 1,339,660 160,550 11.98%
1980 1,369,687 161,363 11.78%
1981 1,384,396 159,594 11.53%
1982 1,399,054 162,219 11.59%
1983 1,405,631 165,856 11.80%
1984 1,418,422 153,286 10.81%
1985 1,436,305 168,097 11.70%
1986 1,448,569 166,603 11.50%
1987 1,449,059 169,702 11.71%
1988 1,470,122 177,681 12.09%
1989 1,503,586 172,989 11.51%
1990 1,503,757 162,153 10.78%
1991 1,539,597 183,197 11.90%
1992 1,572,557 183,180 11.65%
1993 1,589,724 180,054 11.33%
119
1994 1,602,366 174,098 10.87%
1995 1,618,343 175,963 10.87%
1996 1,626,563 172,455 10.60%
1997 1,632,956 175,693 10.76%
1998 1,638,470 182,290 11.13%
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
OECD Firm-level Data Start-ups and Liquidations in Germany Business Notifications
There is clearly a problem with the OECD Firm-level Data for 1978, otherwise this series
seems quite stable over time. The IfM Start-ups and Liquidations data show consistently
higher start-up rates. This is to be expected as they include non-employer businesses (subject
to a turnover threshold), which generally have higher entry and exit rates than employer
businesses. It is also possible that the coverage of the former East Germany in the IfM data
may also contribute to higher start-up rates, as it has been observed in the Eurostat business
demography project that start-up rates are slightly higher in the former communist countries of
Eastern and Central Europe than they are in Western Europe.
The Business Notifications data have even higher rates. The reasons for this are likely to be
due to the units used (local units rather than enterprises), and that many of the apparent start-
ups are really re-registrations of existing business activities.
120
b) New Businesses
0
100
200
300
400
500
600
700
800
900
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Start-ups and Liquidations in Germany Business Notifications
The exclusion of non-employer businesses, and businesses in the former East Germany, is
clearly apparent in the OECD Firm-level data in this chart. The IfM data show a little more
variability over time. This could be due to non-employer business patterns showing a greater
responsiveness to the economic cycle, though there is no hard evidence for this theory. Again
the Business Notifications data show the highest levels reflecting the units used and the
inclusion of many re-registrations.
c) Business Populations
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data Start-ups and Liquidations in Germany Business Notifications
The OECD Firm-level and IfM data sets show a very consistent and slowly rising trend, though
the level of the former is low due to the coverage and threshold limitations. Unfortunately the
shortness of the Business Notifications series limits the conclusions that can be drawn about
this source, though the population estimates are the same as those provided by IfM for the
two years available.
121
6. Hungary
Three sources of data on business start-ups have been identified for Hungary.
a) Enterprises and Non-profit Organisations
Hungarian Central Statistical Office:http://portal.ksh.hu/portal/page?_pageid=38,341368&_dad=portal&_schema=PORTAL
This source contains annual point in time (end year) counts of registered economic
corporations and unincorporated enterprises, as well as quarterly counts of new registrations.
Both are broken down by legal form, the population data are also broken down by economic
activity. Data on new registrations are available from 2001 to 2004.
The data cover all businesses that hold an active registration and tax number in the
administrative register, including most government bodies. There is no registration threshold
in Hungary, so part-time businesses are included. Approximately 75% of registrations are
considered to be economically active by the Hungarian statistical office. All economic activities
are covered, though NACE division L (public administration) is excluded from the counts
broken down by activity. The unit is referred to as the enterprise, but the definition is currently
closer to that of a legal unit.
Year Population New Enterprises Birth Rate
2000 1,175,480
2001 1,207,831 125,233 10.37%
2002 1,236,890 115,878 9.37%
2003 1,263,990 106,471 8.42%
2004 1,286,993 103,271 8.02%
b) Demography of Small and Medium-sized Enterprises (DOSME) Study
http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/pages/publications/DOSME Extensi
on%20Final%20Report.doc. For more information about the DOSME project, see also -http://forum.europa.eu.int/irc/dsis/dosme/info/data/en/index.htm
This source contains data on enterprise populations, births, deaths, survival and factors of
success. The data are broken down by economic activity, size, legal form and characteristics
of the entrepreneur. The final report describes the methodology used to produce the data it
contains. Other descriptive metadata is available on the project web site. The key feature of
this source is that data are based on surveys of businesses rather than directly on the
business register. This will introduce certain survey errors in addition to other methodological
differences.
Data on births are available for Hungary from 1994 to 2001. Known problems with the
observation of business closures in this project led to the construction of trend adjusted
closure data, which has resulted in a certain smoothing of the population data, as shown in
the chart below.
122
The impact on the population of enterprises of using trend-adjusted deaths data
350
400
450
500
550
1994 1995 1996 1997 1998 1999 2000
T
h
o
u
s
a
n
d
s
Raw Population Population Based on Trend-adjusted Deaths
The data below cover NACE sections C to K, to improve comparability with the Eurostat data,
though data for other economic activities are also available.
Year Population Births Birth Rate
1994 456,376 96,654 21.18%
1995 486,975 87,193 17.91%
1996 495,722 63,805 12.87%
1997 488,747 46,993 9.61%
1998 497,401 61,777 12.42%
1999 504,922 59,953 11.87%
2000 520,505 67,432 12.96%
c) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Hungarian data are available for 2000 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Hungarian statistical business
register with economic activities in NACE sections C to K (production, construction, trade and
most services), except class 74.15, management activities of holding companies. All legal
forms are covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
Year Population Births Birth Rate
2000 526,553 71,395 13.56%
2001 542,288 68,963 12.72%
2002 576,609 83,817 14.54%
123
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
DOSME Eurostat Business Demography Central Statistical Office (KSH)
The data on birth rates seem to show an overall downward trend from the immediate post-
communist period in the mid-1990’s. This is fairly typical of other European countries making
the transition to market economies at this time. The trough in 1997 in the DOSME data may
well be genuine, but there is a risk that it may also be partially due to survey errors, or over-
smoothing of the population for inconsistencies in closure data. It is interesting to note,
however, that the Eurostat data series appears to take over where the DOSME data ended.
The divergence between the Eurostat and KSH data between 2001 and 2002 looks
suspicious, but could be due to differences in coverage and purity.
b) New Businesses
0
20
40
60
80
100
120
140
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
DOSME Eurostat Business Demography Central Statistical Office (KSH)
124
Again the Eurostat data seem to take over where the DOSME data end in 2000. The main
difference in this chart, however, is that the KSH data are rather higher than those from the
other sources. This is to be expected, however, as the KSH data have a wider coverage of
legal forms and economic activities, as well as a lower degree of purity than the Eurostat data.
c) Business Populations
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
DOSME Eurostat Business Demography Central Statistical Office (KSH)
Here, the apparent continuation from the DOSME to the Eurostat data is most striking,
particularly given that the DOSME populations are on a point in time basis, whereas the
Eurostat ones are live during period, and could be expected to be around 10% higher. One
clue might be in the fact that the KSH populations are so much higher. This will be partly due
to the wider coverage noted above, but also to the inclusion of a relatively large proportion of
inactive units (possibly up to 25%). This may also be a problem, on a lesser scale, in the
DOSME data.
125
7. Italy
Three sources of data on business start-ups have been identified for Italy
a) Movimprese
InfoCamere:http://www.infocamere.it/movi_search.htm
This source contains counts of total registrations, active registrations, new registrations,
cessations and changed registrations at the Italian chamber of commerce. The data are
broken down by economic activity and geography, and are available for 1995 to 2004.
A glossary and other metadata are available on the web site (in Italian). The data do not cover
NACE section L (public administration), and presumably do not cover government units. The
unit used is the legal unit. The population data are point in time, and appear to relate to the
end of the year, so have been carried over as start-year populations for the following year inn
the table below.
Year
Active Registrations
at 1 J anuary
New Registrations Birth Rate
1995 350,498
1996 3,578,931 505,354 14.12%
1997 3,806,838 1,260,364 33.11%
1998 4,704,107 408,475 8.68%
1999 4,727,504 390,074 8.25%
2000 4,774,264 403,408 8.45%
2001 4,840,366 421,451 8.71%
2002 4,897,933 417,204 8.52%
2003 4,952,053 389,342 7.86%
2004 4,995,738 425,510 8.52%
2005 5,061,859
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Italian data are available for 1998 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Italian statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
126
Year Population Births Birth Rate
1998 3,596,450 409,272 11.38%
1999 3,677,890 278,104 7.56%
2000 3,760,098 291,856 7.76%
2001 3,833,049 294,866 7.69%
2002 3,853,598 283,463 7.36%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for Italy
are available for 1987 to 1993.
Annual point in time populations were used, taken from a social security database. Only
businesses with one or more employees are included. New businesses had to be present in
both the reference year and the following year to be counted as a birth in published analyses
of the data. “One-year” businesses were identified separately, but have been included as
births in the table below to try to improve comparability with other sources.
Year Population Total Entries Start-up Rate
1987 1,115,036 118,676 10.64%
1988 1,150,278 123,394 10.73%
1989 1,177,162 141,112 11.99%
1990 1,191,290 116,359 9.77%
1991 1,191,651 105,252 8.83%
1992 1,195,573 105,222 8.80%
1993 1,151,733 92,444 8.03%
127
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
35%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
OECD Firm-level Data InfoCamere Eurostat Business Demography
There is clearly a problem for the InfoCamere data for 1997, and possibly the Eurostat data
for 1998. The most likely cause is a large increase in the scope of the source. From 1999
onwards these two sources seem much more stable, producing closely comparable rates,
though with possibly slightly diverging trends.
b) New Businesses
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data InfoCamere Eurostat Business Demography
Again the InfoCamere and Eurostat data seem to follow similar, if slightly diverging trends
after 1999. The difference in levels will be mainly due to a mixture of coverage and purity,
though the difference in units (legal unit for InfoCamere, enterprise for Eurostat) may also play
128
a small part. The OECD firm-level data show a lower level again, which is due to the exclusion
of businesses without employees in this source.
c) Business Populations
0
1
2
3
4
5
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
OECD Firm-level Data InfoCamere Eurostat Business Demography
The higher levels in the InfoCamere population data will be due mainly to wider coverage,
and, to a much lesser extent, to the difference in units when compared to the Eurostat data,
as noted above. The difference between these sources would be even greater if the Eurostat
data were on a point in time basis like those from InfoCamere. The rapid growth in the
InfoCamere population between 1997 and 1998 lends weight to the theory that there was a
significant increase in the coverage of this source around that time, as suggested above. The
OECD firm-level data are relatively low again due to the exclusion of non-employer
businesses.
129
8. The Netherlands
Three sources of data on business start-ups have been identified for the Netherlands.
a) Establishment and Closure of Businesses
Statistics Netherlands:http://statline.cbs.nl/StatWeb/table.asp?PA=07223eng&D1=a&D2=0&D3=(l-11)-
l&DM=SLEN&LA=en&TT=2
This source contains counts (and employment) of the stock of businesses (as at 1 J anuary),
businesses opening, and businesses closing. The data are available broken down by
economic activity, and are available for 1993 to 2002. The data exclude certain NACE
categories (Sections A, B, E, L, M and N, and divisions 70, 73, 91 and 92). On this basis it is
assumed that most government activity is also excluded.
Metadata are available by clicking on the table headings on the web site. The unit used is the
“business” which is assumed to be close to the enterprise, as the terms are both used in the
metadata. Establishment of a business is the formation of a new enterprise. This implies that
the statistical criteria for enterprises (autonomy and external orientation) have to be met.
Moreover, the enterprise has to be economically active, i.e. at least one person works in the
enterprise for at least 15 hours a week. The enterprise has to be a new one, i.e. not the
continuation of one or more existing enterprises.
Year Population New Businesses Birth Rate
1993 376,300 29,000 7.71%
1994 382,080 26,000 6.80%
1995 386,360 33,000 8.54%
1996 406,585 34,000 8.36%
1997 425,780 31,000 7.28%
1998 452,450 30,000 6.63%
1999 464,620 31,000 6.67%
2000 473,095 39,000 8.24%
2001 482,295 40,000 8.29%
2002 486,575 38,000 7.81%
b) Eurostat Business Demography Indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form.
Dutch data are available for 1998 to 2002.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the Dutch statistical business register
with economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. All legal forms are
covered except central and local government, and non-profit organisations serving
households.
130
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
Year Population Births Birth Rate
1999 523,243 49,999 9.56%
2000 534,339 50,475 9.45%
2001 541,538 52,053 9.61%
c) OECD Firm-Level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for the
Netherlands are available for 1987 to 1997.
Annual point in time populations were used, taken from the general business register. Only
businesses with one or more employees are included. New businesses had to be present in
both the reference year and the following year to be counted as a birth in published analyses
of the data. “One-year” businesses were identified separately, but have been included as
births in the table below to try to improve comparability with other sources.
Year Population Total Entry Start-up Rate
1987 589,220 48,250 8.19%
1988 640,787 74,612 11.64%
1989 689,359 56,486 8.19%
1990 732,253 82,990 11.33%
1991 759,372 78,554 10.34%
1992 799,563 89,530 11.20%
1993 852,417 107,235 12.58%
1994 915,444 102,139 11.16%
1995 944,909 98,678 10.44%
1996 909,841 101,896 11.20%
1997 932,260 107,013 11.48%
131
Graphical Comparisons
a) Birth Rates
0%
2%
4%
6%
8%
10%
12%
14%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
OECD Firm-level Data Establishment and Closure of Businesses Eurostat Business Demography
The OECD firm-level data show rather a lot of variation, particularly in the earlier years, but
tend to stabilise at around 11% towards the end of the series. This rate is slightly higher than
that from the Eurostat data, due to the interaction of greater purity and the use of a live during
period population reducing the Eurostat
rates, whilst the threshold of one employee
would be expected to reduce the firm-level
rates. The data from the Establishment and
Closure of Business source seem to exhibit
a cyclical pattern, which interestingly
shows a strong negative correlation to the
real GDP growth rate data for the
Netherlands (sourced from Eurostat). This
may be coincidence, but could be worth
further investigation.
b) New Businesses
0
20
40
60
80
100
120
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
T
h
o
u
s
a
n
d
s
OECD Firm-level Data Establishment and Closure of Businesses Eurostat Business Demography
0%
2%
4%
6%
8%
10%
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
GDP Growth Start-up Rate
132
The number of new businesses in the OECD firm-level data set is surprisingly high compared
to the other sources, particularly as the metadata state that only businesses with employees
are included. The coverage in terms of economic activity of this source is higher, but it is likely
that purity and the unit of observation are more important factor in explaining the difference.
The other two sources seem more comparable, with the differences probably due to the more
restricted coverage and the threshold of 15 hours labour input per week in the Establishment
and Closure of Businesses source. The apparent cyclical pattern in the data from this source
is once again evident in this chart.
c) Business Populations
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
M
i
l
l
i
o
n
s
OECD Firm-level Data Establishment and Closure of Businesses Eurostat Business Demography
As for new businesses, the population data from the OECD firm-level project is rather high,
again casting doubt on the statement in the metadata that non-employer businesses are
excluded. Units and coverage, particularly the inclusion of non-active units, are likely to
account for a significant part of the difference. The Eurostat population at least 10% higher
than that from the Establishment and Closure of Businesses source due mainly to it being on
a live during period basis. Relatively small differences in coverage and threshold will also play
a minor role.
133
9. United Kingdom
Four sources of data on business start-ups have been identified for the United Kingdom.
a) Value-Added Tax Registrations and De-registrations
UK Department for Trade and Industry – Small Business Servicehttp://www.sbs.gov.uk/sbsgov/action/layer?r.l2=7000000243&r.l1=7000000229&r.s=tl&topicId
=7000011757
This source contains counts of the stock of businesses registered for value-added tax (VAT),
as well as new registrations and de-registrations. The data are broken down by economic
activity and geography, and are available on a calendar year basis for 1994 to 2003. A
previous series from 1980 to 1993 is also available, but the data are not directly comparable
due to large changes in the VAT threshold.
A paper on the methodology used is available via the web site. The data cover all economic
activities and legal forms, though coverage is limited for certain activities that are exempt from
VAT, particularly in the education and health sectors. The stock data are on a point in time
basis (1 J anuary). The unit used is the VAT registration, which approximates to the legal unit.
The data are sourced from the UK statistical business register, so a proportion of registrations
are statistical rather than purely administrative, particularly for larger businesses.
Year Stock at 1/1 Registrations Birth Rate
1994 1,629,120 169,210 10.39%
1995 1,623,575 164,910 10.16%
1996 1,627,905 169,590 10.42%
1997 1,645,950 185,950 11.30%
1998 1,683,675 184,770 10.97%
1999 1,719,330 178,450 10.38%
2000 1,744,380 179,585 10.30%
2001 1,767,530 168,445 9.53%
2002 1,783,135 175,700 9.85%
2003 1,794,920 189,890 10.58%
2004 1,810,460
b) Barclays Small Business Surveys
http://www.business.barclays.co.uk/BRC1/jsp/brccontrol?task=articleFWgroup&value=6502&t
arget=_self&site=bbb
This commercial source contains data on business start-ups and closures, as well as the total
population of businesses in a series of quarterly reports. The data are broken down in
different ways each year, including by geography, economic activity, and sex of the
entrepreneur, and are available from 1995 to 2004, though for latter years they are
increasingly broadly rounded estimates.
Limited metadata are available within the reports. The data only cover England and Wales,
and are based on business current account openings and closures at Barclays, multiplied by
estimates of their share of the business banking market. This makes it unlikely that central
and local government activities will be included, and the extent of coverage of non-profit
organisations is unclear. Businesses that do not operate via business current accounts are
134
also excluded by definition. The population data are on a point in time basis. The unit used is
the business account, which is expected to be fairly close to the definition of the enterprise,
particularly for new businesses.
Year Stock at 1/1 Births Birth Rate
1995 2,656,570 471,406 17.74%
1996 2,680,924 477,630 17.82%
1997 2,621,702 476,690 18.18%
1998 2,655,889 454,628 17.12%
1999 2,721,198 438,727 16.12%
2000 2,773,646 438,745 15.82%
2001 2,770,000 342,000 12.35%
2002 2,720,000 373,500 13.73%
2003 2,687,000 446,300 16.61%
2004 2,800,000 452,800 16.17%
c) Eurostat business demography indicators
http://epp.eurostat.cec.eu.int/portal/page?_pageid=0,1136195,0_45572097&_dad=portal&_sc
hema=PORTAL
This source contains estimates of the population of active enterprises, births, deaths, survival
and growth. The data are broken down by country, economic activity, size and legal form. UK
data for the population of active enterprises are available for 1997 to 2003, and data on births
cover 1998 to 2003.
The methodological basis for the EU data collection has been set out in a manual, though this
has not yet been published. The data cover units on the UK statistical business register with
economic activities in NACE sections C to K (production, construction, trade and most
services), except class 74.15, management activities of holding companies. Separate data for
some countries (including the UK) are also available for NACE sections M, N and O (health,
education, community, social and personal services). These have been added in to the table
below to improve coverage. This means that agriculture, forestry, fishing, public
administration, activities of households, and extra-territorial organizations and bodies remain
excluded. All legal forms are covered except central and local government, and non-profit
organisations serving households.
The unit used is the enterprise. The population of active enterprises consists of all enterprises
that had either turnover or employment at any time during the reference period, i.e. it is on a
“live during period” basis. Enterprise births are defined as the creation of a combination of
production factors with the restriction that no other enterprises are involved in the event.
Year
Population of
Active Enterprises Births Birth Rate
1997 1,898,810
1998 1,958,750 256,285 13.08%
1999 2,016,395 257,840 12.79%
2000 2,041,685 242,595 11.88%
2001 2,084,540 244,105 11.71%
2002 2,115,325 242,945 11.48%
2003 2,183,125 281,460 12.89%
135
d) OECD Firm-level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for most
countries were drawn from either statistical or administrative business registers, usually at the
level of the enterprise or firm. The UK data, however, were taken from a series of frames for
an annual survey of production businesses. The units used (“reporting units”) were designed
for data collection purposes, and tended to change as business structures evolved, making
them less stable over time than enterprises, and the coverage was determined by survey
requirements, which varied over time, so was rather less comprehensive than that of the
business register, particularly for smaller businesses.
Annual point in time populations were used, based on survey frames usually drawn around
October of each year (variations in the frame date may cause some minor comparability
issues). New businesses had to be present in both the reference year and the following year
to be counted as a birth in published analyses of the data. “One-year” businesses were
identified separately, but have been included as births in the table below to try to improve
comparability with other sources.
Year Stock Entries During Year Entry Rate
1986 148741 23872 16.05%
1987 150778 24114 15.99%
1988 154956 27315 17.63%
1989 158131 38646 24.44%
1990 151945 20775 13.67%
1991 147984 14952 10.10%
1992 x x x
1993 148057 50897 34.38%
1994 157975 31526 19.96%
1995 174825 47639 27.25%
1996 166981 21316 12.77%
1997 169826 21218 12.49%
Note: Data for 1992 are missing, but it looks likely that births for this year are included in 1993 figures.
136
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
30%
35%
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
VAT Registrations Barclays Eurostat OECD Firm-level Data
The OECD Firm-level data show considerably more variability than those from the other
sources. This is likely to be due to changes in coverage between years, as well as the
instability over time of the unit used (the reporting unit). The break in 1992 is likely to be linked
to the introduction of a new statistical business register around this time, which led to a new
definition and numbering system for reporting units. The data only really seem to show
plausible rates at the start and the end of the period covered.
The Barclays data also show more variability than the remaining two sources. This could
reflect the greater coverage of very small businesses, which are known to be more volatile
than their larger counterparts. The pronounced trough in 2001 could be at least partly due to
the “foot and mouth disease” epidemic amongst farm animals in the UK in that year. This had
a particularly strong effect on rural businesses.
VAT Registration data show a much more stable trend, but interestingly do seem to follow a
similar pattern to the Barclays data, looking rather like a smoothed version, albeit at a lower
absolute level. This could be taken as a positive indication of the quality of the two data sets.
Start-up rates for the Eurostat data again follow similar trends, with the increased coverage of
small businesses more than compensating for the use of live during period population figures
when compared to the VAT Registrations series.
137
b) New Businesses
0
100
200
300
400
500
600
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
T
h
o
u
s
a
n
d
s
VAT Registrations Barclays Eurostat OECD Firm-level Data
The OECD Firm-level data show much lower levels because they only include manufacturers,
and exclude many smaller businesses. The other sources have a much more complete
coverage of economic activities. Both the Barclays, and to a lesser extent, the Eurostat
sources, have a higher coverage of small businesses than the VAT registrations data, due to
the high VAT registration threshold in the UK. As is the case for birth rates, the Barclays,
Eurostat and VAT registrations data show similar patterns, though the fluctuations in the
Barclays data are much more exaggerated.
c) Business Populations
0.0
0.5
1.0
1.5
2.0
2.5
3.0
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
VAT Registrations Barclays Eurostat OECD Firm-level Data
Barclays data show the highest population, despite only covering England and Wales. The
data from this source would be around 11 - 12% higher if they covered all of the UK (i.e.
including Scotland and Northern Ireland). The higher population is due to a much greater
coverage of very small businesses (low turnover, no employees) than the other sources, and
no restrictions in terms of economic activities.
VAT Registrations data are lower than those from Eurostat, despite originating from the same
business register, due to the interaction of three coverage related issues, and one basic
138
difference in methodology. VAT registrations data have more comprehensive coverage in
terms of economic activities (particularly agriculture) and legal forms (non-profit institutions),
but this is more than compensated for by a lower coverage of very small businesses, and the
fact that the Eurostat population data are on a “live during period” basis, whereas the VAT
registrations population includes only those registrations live on a specific date (1 J anuary).
139
10. United States
Five sources of data on business start-ups have been identified for the United States.
a) Statistics of US Businesses / Dynamic Data
US Census Bureau -http://www.census.gov/csd/susb/susbdyn.htm
This source contains counts of the stock of establishments, births, deaths, expansions and
contractions, and associated employment changes. The data are broken down by economic
activity, size band (based on employment) and state, and are available for 1995 to 2001.
Papers with descriptive metadata and definitions are available via the web site. Businesses
without employees are excluded. All economic activities are covered except crop and animal
production (NAICS 111,112), rail transportation (NAICS 482), National Postal Service (NAICS
491), pension, health, welfare, and vacation funds (NAICS 525110, 525120, 525190), trusts,
estates, and agency accounts (NAICS 525920), private households (NAICS 814), and public
administration (NAICS 92). Governmental establishments are excluded except for wholesale
liquor establishments (NAICS 4228), retail liquor stores (NAICS 44531), Federally-chartered
savings institutions (NAICS 522120), Federally-chartered credit unions (NAICS 522130), and
hospitals (NAICS 622).
The stock data are on a point in time basis (businesses with employees in the first quarter),
though counts of establishments that had employees in any quarter of the year are also
available. The unit used is the establishment, which is defined as “a single physical location
where business is conducted or where services or industrial operations are performed.” This is
broadly equivalent to the European definition of the local unit. Establishment births are defined
as establishments that have zero employment in the first quarter of the initial year and positive
employment in the first quarter of the subsequent year. Establishment deaths are
establishments that have positive employment in the first quarter of the initial year and zero
employment in the first quarter of the subsequent year. The definitions of births and deaths
are thus quite broad, and correspond to all recorded creations and closures respectively.
Year Population Establishment Births Birth Rate
1995 5,878,957 697,457 11.86%
1996 5,970,420 822,582 13.78%
1997 6,120,714 719,616 11.76%
1998 6,187,599 713,002 11.52%
1999 6,248,411 709,079 11.35%
2000 6,297,423 727,320 11.55%
2001 6,345,890 787,309 12.41%
b) Firm Size Data
US Small Business Administration –http://www.sba.gov/advo/research/data.html
This source contains counts of the population of firms, births and deaths. Employment data
are also available. The data on the population of firms are broken down by size band
(employment) and economic activity. There are no breakdowns of the data on firm births and
deaths. Data on the population or firms are available for 1988 to 2002. Data on births and
deaths are available for 1989 to 2001.
140
Extensive metadata are available in the paper “Statistics of U.S. Businesses – Microdata and
Tables”, available on the website. The coverage is basically the same as source 1 above, as
the firm level data are derived from the US Census Bureau establishment statistics. The
population counts cover all businesses that had an active payroll at any point during the year,
so can be considered as “live during period” data. The unit used is the firm, which is defined
as “the largest aggregation of business legal entities under common ownership or control”, so
corresponds most closely to the European definition of the Enterprise Group (truncated or all-
residential rather than global). Firm birth and death definitions correspond to those for
establishments in source 1 above.
Year Employer Firms Firm Births Birth Rate
1988 4,954,645
1989 5,021,315 584,892 11.65%
1990 5,073,795 541,141 10.67%
1991 5,051,025 544,596 10.78%
1992 5,095,356 564,504 11.08%
1993 5,193,642 570,587 10.99%
1994 5,276,964 594,369 11.26%
1995 5,369,068 597,792 11.13%
1996 5,478,047 590,644 10.78%
1997 5,541,918 589,982 10.65%
1998 5,579,177 579,609 10.39%
1999 5,607,743 574,300 10.24%
2000 5,652,544 585,140 10.35%
2001 5,657,774 569,750 10.07%
2002 5,697,759
Note: The population data have been taken from a different table to the data on births and deaths. The
assumption (in the absence of any evidence to the contrary) is that they are on a comparable basis.
c) Business Employment Dynamics
Bureau of Labor Statistics –http://www.bls.gov/bdm/home.htm, and Pinkston and Spletzer
(2004) -http://www.bls.gov/opub/mlr/2004/11/art1full.pdf
This source contains counts and rates for establishment openings and closures each quarter.
The data can be broken down by economic activity, and are currently available from quarter 3
of 1992 to quarter 4 of 2004 inclusive.
Descriptive metadata and definitions are available via the web site. The data exclude
business with no employees, central and local government units, and some non-profit
organizations. Certain economic activities are also excluded (religious organizations, some
small farms, the Armed Forces and railways). The unit used is the establishment, which is
broadly equivalent to the European definition of the local unit. Openings are either
establishments with positive third month employment for the first time in the current quarter,
with no links to the prior quarter, or with positive third month employment in the current quarter
following zero employment in the previous quarter. Closings are either establishments with
positive third month employment in the previous quarter, with no positive employment
reported in the current quarter, or with positive third month employment in the previous
quarter followed by zero employment in the current quarter.
141
No stock data are given, but they can be estimated from openings counts and rates (or
closures counts and rates) on a quarterly basis. These can then be used to calculate annual
birth and death rates. The data on openings and closures give slightly different stock figures.
This is due to rounding of the counts of openings and closures (to the nearest thousand), and
the rates (to one decimal place). The derived stock figures based on openings and closures
for each year are within the margins of error associated this level of rounding. The impact on
the annual opening and closure rate estimates is less than 0.2%.
The paper by Pinkston and Spletzer explores the impact of short-lived businesses on the
data, and gives annualised data for 1998 to 2001. Their method removes very short-lived
businesses, and false start-ups due to businesses that have previously been in the population
of employers, but were temporarily absent. The effect on start-up rates is dramatic.
Year Population Births Birth Rate Annualised Births Annualised Birth Rate
1993 5,419,807 1,171,000 21.61%
1994 5,544,268 1,223,000 22.06%
1995 5,738,196 1,242,000 21.64%
1996 5,828,816 1,306,000 22.41%
1997 5,902,142 1,326,000 22.47%
1998 6,045,896 1,344,000 22.23% 778,826 12.99%
1999 6,096,898 1,409,000 23.11% 804,022 13.19%
2000 6,200,692 1,405,000 22.66% 809,301 13.09%
2001 6,268,227 1,363,000 21.74% 790,237 12.67%
2002 6,344,799 1,374,000 21.66%
2003 6,378,568 1,355,000 21.24%
2004 6,535,698
Note: The population is calculated as the median value of the intersection between the ranges of
possible values based on the births and deaths data for the second quarter of each year (i.e. it is the
estimated population as at 1 April). Births and deaths are for the period 1 April year t to 31 March year
t+1. 1 April is used as a reference date in an attempt to improve comparability with source 1, which is
calculated on a March to March basis.
d) Longitudinal Business Database
US Census Bureau -http://www.ces.census.gov/ces.php/abstract?paper=101647
This database contains linked records of establishments and firms over time. It can be used to
produce data on business dynamics. The internet address above is that of a paper describing
the database, which includes data on births and deaths. A second paper is available with
more detailed analyses for the retail sector – see:http://www.ces.census.gov/ces.php/abstract?paper=101704
The data in the paper are not broken down in any way, but the database would allow a range
of detailed breakdowns. The paper presents stock, birth and death data for 1976 to 1999.
The paper contains descriptive metadata. The source data cover establishments with paid
employees. Economic activity coverage is the same as for source 1 above. The stock data are
on a point in time basis, and the unit used is the establishment. Births are records that were
active in one year, but not the previous year, adjusted for reactivations. Deaths are records
that were active in one year, but not the next, adjusted for reactivations.
Year Population Births Birth Rate
1976 4,945,528 824,563 16.67%
142
1977 5,125,942 844,422 16.47%
1978 5,152,243 683,598 13.27%
1979 5,330,266 681,813 12.79%
1980 5,283,897 610,991 11.56%
1981 5,244,139 649,292 12.38%
1982 5,294,765 702,036 13.26%
1983 5,586,606 755,528 13.52%
1984 5,833,945 779,039 13.35%
1985 5,981,692 771,830 12.90%
1986 6,098,536 763,103 12.51%
1987 6,174,220 851,033 13.78%
1988 6,228,218 717,030 11.51%
1989 6,388,877 797,117 12.48%
1990 6,645,560 933,622 14.05%
1991 6,729,082 799,454 11.88%
1992 6,759,906 787,850 11.65%
1993 6,860,000 746,635 10.88%
1994 6,973,457 760,594 10.91%
1995 7,077,456 754,795 10.66%
1996 7,167,943 766,265 10.69%
1997 7,305,127 894,978 12.25%
1998 7,351,196 754,708 10.27%
1999 7,405,245 828,164 11.18%
Note: The population data are those establishments considered to be active in the longitudinal
database. Births and deaths have been adjusted to remove all reactivations.
e) OECD Firm-level Data Project
http://www.oecd.org/document/4/0,2340,fr_2649_37451_1962948_1_1_1_37451,00.html
This project brought together data from ten OECD member countries, using a common
analytical framework, based on the harmonisation, as far as possible, of key concepts
(e.g. entry, exit, or the definition of the unit of measurement) and methodology. Data for the
United States are available for 1989 to 1996.
Annual point in time populations were used, taken from the prototype longitudinal business
database. Only businesses with one or more employees are included. New businesses had to
be present in both the reference year and the following year to be counted as a birth in
published analyses of the data. “One-year” businesses were identified separately, but have
been included as births in the table below to try to improve comparability with other sources.
Year Population Total Entry Start-up Rate
1989 4648625 528711 11.37%
1990 4798181 526578 10.97%
1991 4867411 503077 10.34%
1992 4981011 550934 11.06%
1993 5051860 505943 10.01%
1994 5137618 515718 10.04%
1995 5224433 519906 9.95%
1996 5311984 530919 9.99%
143
Graphical Comparisons
a) Birth Rates
0%
5%
10%
15%
20%
25%
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
Longitudinal Business Database Statistics of US Businesses
Firm Size Data Business Employment Dynamics - Summed Quarterly Data
OECD Firm-level Data Business Employment Dynamics - Annualised Data
The Business Employment Dynamics quarterly data set is a clear outlier in terms of birth
rates. The annualised data set clearly show that this is almost entirely due to periodicity and
data purity. The remaining data sources appear to give fairly comparable measures of start-up
rates, typically between 10% and 13%.
Based on the table under the population chart below, if the population of firms active in March
was used as the denominator for the Firm Size data set, this would have the effect of
increasing the birth rate for this source by around 1.5%, taking it to a similar, or very slightly
higher level than that for Statistics of US Businesses.
This suggests the interesting conclusion that where data are otherwise comparable, the
choice of firm or establishment as the unit of observation makes little difference to business
start-up rates. The increased volatility usually associated with establishments is, in this case,
cancelled out by the higher population of establishments, i.e. both the numerator and the
denominator are higher for establishments, but the rate is almost identical. Taking this a step
further, this suggests that, for international comparison purposes, differences in units may not
be a major obstacle. Unfortunately these conclusions are only based on data for one country
for a limited period of time, so it remains to be seen how safe they are in a wider context.
144
b) New Businesses
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Longitudinal Business Database Statistics of US Businesses
Firm Size Data Business Employment Dynamics - Summed Quarterly Data
OECD Firm-level Data Business Employment Dynamics - Annualised Data
For births, the clear outlier is again the Business Employment Dynamics quarterly data. This is
due to issues of purity (i.e. including virtually all establishment creations), and a much greater
chance of including short-lived businesses.
The number of births from the Longitudinal Business Database seems close to those from
Statistics of US Businesses, but this masks two differences between the data that seem to
largely cancel each other out. All things being equal, the number of births from the
Longitudinal Business Database should be higher, simply because the population for that
source is more comprehensive, however, this seems to be balanced by the greater extent of
data matching within this source, to remove all reactivations and other “false” births.
The spike in the Longitudinal Business Database series in 1997 is acknowledged as
suspicious, and could perhaps be linked to the very similar spike in the data from Statistics of
US Businesses the year before. If this is the case, it could indicate the presence of a lag
between these two sources.
The gap between the Statistics of US Businesses series and that from the Firm Size Data
source indicates the proportion of new establishments created by existing firms, assuming
that very few new firms have more than one site when they are created. It is interesting to
note that the spike in the Statistics of US Businesses data for 1996 is not present in the Firm
Size Data series. This might indicate that it was more of a source processing issue than a real
increase. Similarly the increasing divergence between these sources for 2001, combined with
the fact that Statistics of US Businesses is the only source to show an increase between 2000
and 2001, might suggest similar processing issues.
The data for the OECD Firm-level series are taken from a prototype of the Longitudinal
Business Database, which explains the similar trends, however there is clearly a difference in
coverage.
145
c) Business Populations
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
2
0
0
1
2
0
0
2
2
0
0
3
2
0
0
4
M
i
l
l
i
o
n
s
Business Employment Dynamics Statistics of US Businesses OECD Firm-level Data
Longitudinal Business Database Firm Size Data
All five sources show a similar trend, but there is considerable variation in the levels. The
series from the Longitudinal Business Database is a clear outlier. This is likely to be due to
coverage, as this source includes certain economic activities that are excluded in the other
sources, particularly farms, public administration and state education.
Statistics of US Businesses and Business Employment Dynamics have very similar coverage,
but are based on two different business frames. Frame maintenance and timing of updates is
likely to account for the slight differences between them.
The Firm Size Data source is derived from the same register as Statistics of US Businesses,
with identical coverage, hence the similar trend. It shows a lower level, which is to be
expected as this is the only source based on the firm, which can be an aggregate of the
individual establishments used in the other sources. This effect would be greater if it was not
partly cancelled out by the inclusion of all firms that had an active payroll during the year,
rather than just on the March reference date used for Statistics of US Businesses. The table
below is derived from US Census Bureau data, and shows that the population of firms would
be about 13% lower if it only included those with a payroll in March.
Year Employer Firms
(whole year)
Firms with no
employees in March
Employer Firms
(March)
% Difference
1997 5,541,918 719,978 4,821,940
-12.99%
1998 5,579,177 711,899 4,867,278
-12.76%
1999 5,607,743 709,074 4,898,669
-12.64%
2000 5,652,544 726,862 4,925,682
-12.86%
2001 5,657,774 703,837 4,953,937
-12.44%
2002 5,697,759 770,041 4,927,718
-13.51%
146
Conclusions
In one sense it is easier to compare different data sources from one country, than data from
different countries, as, at least in theory, they should give the same answer when all of the
differences in methodology have been removed. However, this exercise proves the benefit of
highlighting those differences in methodology which might otherwise have been put down to
genuine variations if data from different countries were being compared. This, in turn, gives a
better understanding of the factors affecting comparability, which can then be used to improve
international comparisons of data by separating out genuine variations from those caused by
methodological differences.
It is possible to explain most of the apparent discrepancies between the data from the
different sources considered for each country above by a close study of the metadata
available for each source, and by making assumptions (of varying degrees of robustness) of
the impact of the main differences in methodology. The lack of standardisation of metadata, in
terms of content, terminology and presentation, sometimes combined with a certain lack of
clarity, particularly for non-specialists, makes this task rather more difficult than it should be.
For several of the countries above the explanations and assumptions have been verified by,
those responsible for the source, or are partly based on additional information from national
experts. This shows that the metadata necessary for better informed international
comparisons could be made available relatively easily.
147
Annex 5 – Business Closures
Introduction
The focus of this report has been on the comparability of data on business start-up rates, as
this is perhaps the key indicator for studies of business demography, and one of the most
important for entrepreneurship. Business start-ups, however, only give part of the picture. To
properly understand and interpret the data, it is necessary to know the extent to which new
businesses are adding to economic activity or replacing businesses that have closed.
Measures of business closure rates are therefore a very useful complement to start-up
indicators.
This interdependency between start-up and closure rate data has been recognised in the
Eurostat business demography project, where there has been a specific effort to develop
methodologies for these indicators that closely mirror each other. This approach seems both
logical and successful, and implies that the factors of comparability affecting business start-up
rates proposed in the main body of this report are also likely to be relevant for closure rates.
This Annex explores this hypothesis further, looking at similarities and differences in the ways
the factors can be applied.
The chart below shows business closure rate data for a number of countries, including two
sources for the United States, as published by those countries or Eurostat (for more
information on the sources, see Annex 2). It complements that on start-up rates in the
introduction to the main body of this report (Figure 1.1). It shows a similar degree of variability,
though there are also more gaps in the data.
Business Closure Rate Data for Selected Countries
0%
5%
10%
15%
20%
25%
A
u
s
t
r
a
l
i
a
C
a
n
a
d
a
D
e
n
m
a
r
k
F
i
n
l
a
n
d
G
e
r
m
a
n
y
I
t
a
l
y
L
a
t
v
i
a
N
e
t
h
e
r
l
a
n
d
s
N
e
w
Z
e
a
l
a
n
d
N
o
r
w
a
y
P
o
r
t
u
g
a
l
S
l
o
v
a
k
i
a
S
p
a
i
n
S
w
e
d
e
n
U
K
U
S
(
S
U
S
B
)
U
S
(
B
L
S
)
2000 2001 2002
Sources: National statistical office and Eurostat publications and internet sites
As is the case for Figure 1.1, the chart above is not a particularly meaningful or valid
comparison, as the variability shown is due rather more to methodological differences than to
real variations between countries.
148
Factors Affecting the Comparability of Business Closure Rates
This section considers how far the factors of comparability for business start-up rates
identified in Section 3 of this report can be applied to business closure rate data.
• Purity
It is clear from the discussions on the purity of start-up data that the separation of “pure
deaths” from other exits will have a significant impact on data from many sources. It is
logically easier to consider purity of start-ups and closures together rather than separately,
because in many cases, apparent closures and start-ups can be linked, proving that these
businesses have in fact continued to operate, despite appearances to the contrary. There is
thus likely to be a strong correlation between the ratios of pure births to total entries, and pure
deaths to total exits, for any given source.
As with start-ups, reactivations can be difficult to deal with conceptually. A business that is
dormant for a few months before re-starting would not normally be considered to be a pure
death, however longer periods are not so easy to deal with. If a threshold is applied for start-
ups, it is logical that the same threshold should be applied for closures, otherwise businesses
will not be treated in a consistent way in the two data sets.
There is a specific problem for closures, however, in that the longer the period of time allowed
for potential reactivations, the greater the lag in the production of definitive closure rate data.
Eurostat apply the same two-year reactivation threshold for closures as they do for start-ups,
so the lag for definitive closure data is two years longer than for start-up data relating to the
same period. This is partly resolved by the release of provisional estimates until the definitive
closure rates can be calculated.
• Timing
This issue is often more significant for business closures than for start-ups, as the closure
process can take many years in some cases, and reporting of closures to administrative and
fiscal bodies tends to be rather slower than for start-ups. An entrepreneur might consider a
business to be closed from the day he or she stops taking on new work or ceases trading. For
accounting purposes there is likely to be a further period during which payments are sought
from debtors and made to creditors, until the business accounts can be finalised. There may
then be a further period for administrative or fiscal purposes during which any outstanding
obligations are either fulfilled or written-off. Finally there may also be some sort of legal
procedure, which may take place before, during or after the above. The point at which a
closure is recorded will therefore be determined by the nature of the data source.
As for start-ups, a closely related issue is that of lags, the different events above may not be
notified immediately. Work on lags in the value-added tax data used for the British statistical
business register, described in Section 3.2 of this report, revealed whilst registration lags had
a median value close to forty days, that for de-registrations was nearly seventy days.
• Periodicity
The issues affecting closure rate data are similar to those described for start-up data.
149
• Type of Population
Virtually all business closure rate data currently available use businesses rather than people
as the population, so this factor is unlikely to be important for the comparability of existing
closure rate data.
• Temporal Basis
The differences between point in time and live during period populations are described in
Section 3.5 of this report, and, in more detail, in Annex 3. The use of a live during period
population will result in a higher denominator and lower closure rates. The issues affecting
closure rates are the same as those affecting start-up rates.
• Source
Where the source of closure data is a statistical or administrative business register, the issues
are largely the same as for start-up data, however it is important to know how closures are
defined in these sources, at what point in the process closures are identified, and with what
lags (see the comments on timing above). This is often more complicated for statistical
business registers, as these tend to be updated from a number of statistical and
administrative sources, all of which may have different definitions and lags for closures.
There is a fundamental problem with survey data on closures, in that if a business has closed,
it is often difficult to make contact to confirm this, so it can be difficult to differentiate between
closures, businesses that choose not to respond, and those that can not be contacted
because they have moved to an unknown address. The wider coverage of a census can help
to reduce this problem, but is unlikely to eliminate it.
• Units
The issues affecting closure rates are similar to those affecting start-up rates. It should be
noted that establishment or legal unit closures do not necessarily equate to enterprise
closures.
• Coverage
In some cases, closures may be indicated when a business is still active, but has moved out
of scope of the source. Sometimes this may be entirely due to a change in the source rather
than any change on the part of the business. It is also possible that businesses that do not
respond to, or comply with the requirements of a particular source, may be treated as closed
by that source, usually after a certain number of periods of non-compliance or non-response.
• Thresholds
Similarly, a business can appear to close if it moves out of scope of a particular source by
dropping below a certain threshold. Such businesses, however, often continue to operate,
albeit at a lower level. An example may be an artisan who reaches retirement age and stops
full-time activities, but continues his or her business on a part-time basis, perhaps just for a
limited range of customers.
Businesses that operate close to the threshold used for a particular source, e.g. sole
proprietors that take on employees only when market conditions allow, are likely to move in
and out of scope of that source, possibly quite frequently. They would normally be considered
to be reactivations, as discussed in the section on purity above, rather than pure deaths
150
followed by pure births. As for start-ups, the international application of a standard threshold
would considerably improve the comparability of business closure rates.
• Other Factors
As with business start-ups, various other factors can affect the international comparability of
closure rates, including the complexity of administrative procedures, the impact of tax, subsidy
and other policies, the nature of the political system, and a wide range of other economic,
political, social and cultural factors. These factors relate more the sort of variation in data that
users are really interested in, than to data production methodology. Thus, as for start-ups, this
Annex only focuses on methodological factors of comparability. If these can be better
understood, and eliminated where possible, data users have a much better chance to observe
the non-methodological factors in a less biased way.
Conclusion
This annex demonstrates that the factors of comparability derived for business start-up rates
in the main body of this report, can also be applied relatively easily to data on business
closures. There are some differences in terms of the relative importance of the different
factors, for example identifying a business closure, and attributing it to a specific point in time
is often more of a problem than identifying when start-ups occur. However, these differences
are relatively minor, so it is recommended to treat closures as complementary to start-ups in
terms of developing data collection and comparison methodology.
doc_408540456.pdf