Description
In this issue of this journal, Christopher Nobes comments on my classification of accounting systems [Accounting
Organizations and Society (2001) 26, 327]. He concludes that the classification is unsound because the nature of the
data made them unsuitable and they contained errors. I took as a source for my selection process a database that
reflects the decisions of the EU directives and the Transnational Accounting (TRANSACC) editors to describe
important recognition and valuation rules. This process was made explicit and followed reasonable principles.
According to the accuracy issues raised, some discrepancies could be explained by different interpretations. Additionally,
the conclusion is still valid as the results are shown to be robust to variations of weightings and methods hence
also to unsystematic failures.
Accounting classi?cation and the international harmonisation
debate: a reply to a comment
Anne d’Arcy*
Deutsche Bank AG, Taunusanlage 12, D-60325 Frankfurt am Main, Germany
Abstract
In this issue of this journal, Christopher Nobes comments on my classi?cation of accounting systems [Accounting
Organizations and Society (2001) 26, 327]. He concludes that the classi?cation is unsound because the nature of the
data made them unsuitable and they contained errors. I took as a source for my selection process a database that
re?ects the decisions of the EU directives and the Transnational Accounting (TRANSACC) editors to describe
important recognition and valuation rules. This process was made explicit and followed reasonable principles.
According to the accuracy issues raised, some discrepancies could be explained by di?erent interpretations. Addition-
ally, the conclusion is still valid as the results are shown to be robust to variations of weightings and methods hence
also to unsystematic failures.
#2003 Published by Elsevier Ltd.
Introduction: Nobes’ concerns
In this issue of this journal, Christopher Nobes
comments on my classi?cation of accounting sys-
tems (d’Arcy, 2001).
1
He concludes that the classi-
?cation is unsound because the nature of the data
made them unsuitable and they contained errors.
In particular the following concerns were raised:
1. That I did not question counter-intuitive
results for Australia but just sought to
explain them. This research method is then
associated with other classi?cation
approaches e.g. Da Costa, Bourgeois, and
Lawson (1978) or Goodrich (1982).
2. Important issues in a data set may be
swamped by trivial ones. Additionally, as
the method of data collection is sub-
jective, the impression of objectivity is
misleading.
3. Other aspects of suitability, e.g. the split
between individual and group accounting
rules and the inclusion of variables that
describe accounting rules according to the
distributability of income.
4. Errors or disputable interpretations.
As it is not possible to describe all the methods
used in and the ?ndings of a 300 page study
0361-3682/$ - see front matter # 2003 Published by Elsevier Ltd.
doi:10.1016/j.aos.2003.10.001
Accounting, Organizations and Society 29 (2004) 201–206
www.elsevier.com/locate/aos
* Tel.: +49-69-910-43035; fax: +49-69-910-47532.
E-mail address: [email protected] (A. d’Arcy).
1
Readers will note that I have commented on Nobes’ paper,
and should be aware that he also made helpful comments on
previous drafts. I also thank Christine Roßbach and David
Powell for helpful comments.
(d’Arcy, 1999), I shall brie?y discuss some aspects
of these concerns in the following analysis.
Objective of the study and interpretation of some
results
First of all, I would like to make it clear that the
purpose of my work was not to prove that my
classi?cation described accounting systems in a
better or more comprehensive way than other
approaches. I simply wanted to focus on one
aspect of accounting systems that, in my view, had
not yet been investigated in su?cient detail,
namely ?nancial reporting rules (d’Arcy, 2001, p.
332). I felt that this dimension of an accounting
system might be important in connection with
recent accounting harmonisation debates. First of
all, though, it was necessary to examine former
classi?cation approaches, suggest technical
improvements and compare the results in order to
gain a better understanding of the di?erences.
I had learned from previous attempts that
results should always be interpreted with caution
due to the limitations of databases and the
statistical techniques applied. However, I do not
accept empirical ?ndings just because they ful?l
my or others’ previous expectations. Empirical
research should allow for new insights. So, results
that seem implausible at ?rst sight do not auto-
matically indicate poor methods or data. On the
contrary, they need further investigation in order
to reveal additional aspects not re?ected in the
preceding underlying hypothesis. This is how I
tried to explain the results for Australia which
showed outsider characteristics compared with
other national systems under review. But I do not
believe that this result was counter-intuitive. When
examining the data for Australia some aspects
emerged as being special, e.g. application of the
equity method. In most systems the equity method
is required for the valuation of associated compa-
nies. In Australia, the cost method is applied and
equity method information only has to be dis-
closed. Another example can be found in the rules
for revaluation accounting, e.g. the revaluation of
intangible assets is permitted without an active
market.
The factual outcome of my study was that the
Australian system had an outsider position as
regards speci?c ?nancial reporting rules when
compared to the 14 other systems under review. I
did not contend that the Australian accounting
system in general has an outsider position. In fact,
when using the same database for a di?erent
research purpose, I obtained di?erent groupings
(d’Arcy, 2000). An index ranking of the level of
?nancial accounting measurement regulation—
called the degree of determination—shows a
highly-regulated group including Australia, UK
and the US. This grouping clearly re?ects the
importance of national capital markets and the
former in?uence of the British Empire on
accounting systems, as does the loose link between
tax and ?nancial accounting (Craswell, 1995, p.
100).
Nobes then links my interpretations with pre-
vious studies by Da Costa et al. (1978) or Good-
rich (1982) that also classify accounting systems
‘‘based on data that had not been prepared for
that purpose’’. The weaknesses of these studies
have been discussed in other publications (e.g.
Nair & Frank, 1980, pp. 444–445, Nobes, 1981, p.
268, Nobes, 1983, p. 2, Nobes & Matatko, 1980,
p. 70). However, these studies based their
conclusions on a database prepared by other
researchers for a di?erent purpose. This database
is not suitable for classi?cation because it mixes
the levels of accounting rules and practices and
uses ordinal scales as input into statistical methods
requiring cardinal ?gures. Moreover, it can be
shown that even if the database were suitable, the
statistical method applied—factor analysis—
would itself constitute a serious limitation (Archer
& McLeay, 1992, p. 1).
Concerning the closeness of Switzerland and
UK shown by nonmetric multidimensional scaling
(MDS), it should be noted that I used this method
to test the cluster results for the macro structure
only. The advantage of nonmetric MDS is that no
assumptions need to be made about the under-
lying transformation function. The only assump-
tion is that the data are measured at the ordinal
level. That means that the di?erences in measure-
ment rules between the national accounting sys-
tems under review are only applied in their order
202 A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206
of ranking. Consequently, this method is very
strong in ?nding macro structures, but weak in
showing adequate micro structures.
2
This was one
of the main reasons why I did not use this method
as my main research method. Furthermore, as I
am aware of this shortcoming, I did not take the
micro results as my main ?ndings and did not
attempt to explain them either.
Important versus unimportant variables: composi-
tion of the database
The suitability of a database for a speci?c pur-
pose has in my view two dimensions: ?rst, how is
the database structured and how are the data
obtained; second, how are statistical methods used
to answer the research question.
The data are based on the Transnational
Accounting (TRANSACC) reference matrix
(TRM) established by Dieter Ordelheide and
myself (Ordelheide & Semler, 1995). The structure
and the entries were based on the respective
country chapters of TRANSACC (Ordelheide &
KPMG, 1995). The goal of TRANSACC was to
describe 15 accounting systems in a systematic
way by describing their history, the regulatory and
institutional background, accounting goals as well
as principles, respective formats, recognition and
valuation rules, consolidation regulation and dis-
closure requirements. The core of the descrip-
tion—the chapters on recognition, valuation and
consolidation, had a standardised structure, which
basically re?ected the formats of the 4th and 7th
EU accounting directives. The TRM also mirrored
these formats. The TRANSACC editors felt that
the topics referred to in the directives were impor-
tant parts of an accounting system and that rela-
ted recognition and valuation rules should
therefore be described for each national account-
ing system in the country contributions. The TRM
shows these rules in more standardised form.
The idea behind the database was to accept this
link to the EU formats and hence to use all entries
from the TRM. Given that the EU and the
TRANSACC editors have already identi?ed both
these recognition and valuation rules as impor-
tant, there is absolutely no reason to ignore this
assessment. However, as the TRM itself was not
suitable for my research design, I modi?ed the
data as described in d’Arcy (2001, pp. 334–335).
Part of this modi?cation involved disregarding
some variables due to insu?cient descriptions
which did not allow proper di?erentiation between
the systems, e.g. accounting for leases. I am fully
aware that, as a result of this, the database is not
all-embracing. Though the data could be improved,
there are nevertheless important rules described that
can be used as input into cluster analysis.
As Nobes has correctly noted, I did use judge-
ment in coding the data. But using judgement did
not mean that the variables were simply selected at
random. On the contrary, I applied two principles
when making my decisions. I took all the data
from the TRM (principle of completeness) and
only disregarded items that did not di?erentiate
properly between the systems (principle of e?ec-
tive separability). For example, the separability
function is the reason why I described di?erent
rules according to the simpli?ed valuation meth-
ods for stocks in only one variable—is the last in
?rst out-method (LIFO) forbidden or not. The
?rst in ?rst out-method (FIFO) or the average cost
method are allowed in all systems under review and
therefore do not qualify to be coded as variables
that e?ectively distinguish between systems.
I agree that weighting variables may have an
important impact on results. It is also true that
assigning weights does not necessarily add sub-
jectivity. For example, giving a double weight to
those variables that have a simple coding in my
database is easily justi?ed. However, instead of
simply justifying my choice of weighting variables
and applying the cluster method, I chose to ana-
lyse the sensitivity of results based on di?erent
weightings and di?erent cluster methods (d’Arcy,
2001, pp. 341–342). The idea was to show that
under extreme conditions the macro structures
2
In fact, adding a third dimension, the multidimensional
scaling results correspond for Switzerland and the UK more
clearly with the dendrogram derived from cluster analysis. The
goodness-of-?t value according to Kruskal (1964, p. 3) increa-
ses from 0.16672 ‘‘su?cient’’ to 0.09988 ‘‘good’’. But the three-
dimensional results are not taken because they may not be
reliable as the total number of parameters being estimated is
large relative to the number of data values.
A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206 203
remained stable so that neither di?erent weight-
ings nor di?erent cluster methods would sig-
ni?cantly change the results. I therefore varied the
similarity coe?cients as input into cluster analysis
to simulate weightings for the extreme condition
that all matches or all non-matches of variables had
a double weighting. It has been shown that in any
constellation double weightings would lead to the
same macro structure of clusters (d’Arcy, 1999, pp.
169–170, d’Arcy 2001, p. 342). Additionally, vari-
ation of the cluster methods—I compared the results
produced by the single linkage, complete linkage
and average linkage within groups methods—also
shows great robustness in the cluster results. Hence,
the ?ndings were proved to be robust against var-
iations of weightings and cluster methods.
Other aspects of suitability
I agree that there are some problems in distin-
guishing between individual and group accounts.
However, the EU and the TRANSACC editors
decided to make a distinction and I simply applied
this concept. It may be a very continental Eur-
opean view to distinguish the dataset for these two
types of accounting rules. However this di?er-
entiation may help to avoid a typical ‘‘Anglo-
American bias’’ where the dataset ignores di?erent
rules in both types of accounts. In most cases, the
rules for individual accounts are also valid in
the consolidation process for the group accounts.
But some accounting problems are treated
di?erently and are re?ected accordingly in the
database, e.g. the valuation of associated compa-
nies or the valuation of goodwill arising from
asset deals. The dataset is not all-embracing as it
does not include all accounting problems. How-
ever, concentration on one set of accounts, e.g.
group accounts, would ignore these di?erences.
Moreover, this di?erentiation also allows further
analysis. For example, clustering data for indivi-
dual and group accounts separately shows that the
Australian outsider position can be traced back
primarily to individual accounting rules. Based on
the second part of the database, which deals with
additional rules arising from the preparation of
group accounts, Japan comes out as an outsider.
Turning to the second concern, that variables
describing accounting rules according to the dis-
tributability of income should not be included, I
strongly recommend that those rules be con-
sidered. As distributable pro?t is equated in this
case with pro?t in general, these rules are also
relevant for consolidated statements and are thus
important characteristics of a national accounting
system. I have to admit that the labels of these
variables do not adequately convey their purpose
and could therefore be misunderstood. The issue is
whether the recognition of an asset or liability
a?ects pro?t or not. In some accounting systems,
the recognition of certain assets is only allowed
if the di?erence is not accounted for as pro?t. In
this case an equity item has to be recognised.
The pro?t and loss account is not a?ected by
this transaction. Accordingly, group pro?t is not
in?uenced by recognition of this asset. For
example, TRM di?erentiates between systems
that allow or require the recognition of develop-
ment costs. Additionally, a variable indicates
whether the capitalisation of development costs
in?uences pro?t by recognising a respective
equity item (including group pro?t) in the
respective periods or not.
I know that the concept of the determination
function of distributable pro?t is unknown in the
US or under International Accounting Standards
(IAS). For the related accounting method, however,
we ?nd comparable e?ects from an analysis of
certain IAS or US GAAP rules. For example IAS
39 or FAS 133 require the valuation of securities
classi?ed as available for sale at fair or market
value. The value di?erence only a?ects the respective
equity item and not pro?t. If a valuation di?erence
a?ects pro?t, this can be seen as an important
characteristic of a national accounting system.
Errors and disputable interpretations
Due to the limits on space I would like to focus
on two examples only. More generally, I still
believe that some of the ‘‘Error’’ issues listed
above are caused by di?erent interpretations.
Nobes states that the exception to the obligation
to prepare group accounts for small groups does
204 A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206
not exist in the UK for listed companies. As I
generally described rules applicable to corpor-
ations only and—if there is a di?erence—to listed
companies, this exception was not expected to be
included in the database. However, in Germany
and some other continental European countries
there are no speci?c rules for listed companies (or,
there were none in 1995). So in Germany we have
the general rule that small groups—de?ned by
certain criteria—do not have to publish group
accounts. Listed companies—as set out in com-
pany law—have to follow the rules for large cor-
porations. Following the systematics of law, we
?nd the description of the general rule for all
companies, but listed companies are character-
istically not small. Thus, following the principle
of completeness in data selection, this variable is
still useful for di?erentiating between national
accounting systems and should be included.
Another issue is the interpretation of the mean-
ing of a rule. As an outcome of the 7th EU direc-
tive, all EU member states had to transpose into
national law the rule that subsidiaries with dis-
similar activities should not be consolidated. Con-
sistently, we ?nd this prohibition for all EU
member states, including the UK, in TRANSACC.
Even if in practice all subsidiaries have to be con-
solidated because dissimilarity is de?ned very
stringently, the code ‘‘exclusion required’’ re?ects
the rule properly.
These and other examples show the great di?-
culties in comparing national ?nancial reporting
rules. Some confusion is partly caused by di?erent
translations. Additionally, in some cases the con-
cepts do not ?t, as Nobes has demonstrated using
the example of the terms provisions and liabilities.
However, other di?erences, like the two examples
described above, arise because of di?erent inter-
pretations and are neither right nor wrong.
Conclusion
According to Nobes, the dataset is neither sui-
table nor accurate. Surprisingly, he states that the
subjectivity of the data selection process is a
weakness of this approach. In my view empirical
research is always subjective, which means that
this shortcoming cannot be resolved in general.
But I took as a source for my selection process a
database that re?ects the decisions of the EU
directives and the TRANSACC editors to describe
important recognition and valuation rules. Addi-
tionally, I tested the robustness of the results
rather than introduce weightings or change the
selection. This process was made explicit as sug-
gested by Roberts (1995, p. 660). As the goal of
TRANSACC and the TRM is to describe and
compare national accounting systems, the same is
the case for the modi?ed database. Furthermore, I
am not aware of any ‘‘objective’’ empirical method
to ?nd clusters of national accounting systems. But
I believe, that the consideration of the judgement
of the EU as well as the TRANSACC editors
enhances reliability.
As no database can be all-embracing, the di?er-
entiation between individual and group account-
ing rules does not impair suitability of the
database just because of the fact that some rules
are not described. This di?erentiation re?ects the
structure of the 4th and 7th EU Directives and
allows to gain further insights. I also place
emphasis on the inclusion of variables re?ecting
accounting rules regarding pro?t distributability.
These rules a?ect the calculation of accounting
pro?t—even in the group accounts—and are in
this regard an important part of a national
accounting system.
According to the accuracy issues raised, some
discrepancies could be explained by di?erent
interpretations. However, unsystematic failures
are not unavoidable. But taking the robustness of
the results to variations of weightings and methods
into account, it is unlikely that those failures
would harm the ?ndings of this study. So my
conclusions are still valid.
References
Archer, S., & McLeay, S. (1992). European ?nancial reporting:
issues of de?nition and classi?cation of national systems of
?nancial reporting. Paper presented at the Annual Congress
of the European Accounting Association. Madrid, Spain.
Crasswell, A. (1995). Australia individual accounts. In D.
Ordelheide, & KPMG (Eds.), Transnational accounting
(TRANSACC) (pp. 69–167). Basingstoke: McMillan.
A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206 205
d’Arcy, A. (1999). Gibt es eine anglo-amerikanische oder
eine kontinentaleuropa¨ische Rechnungslegung?—Klassen
nationaler Rechnungslegungssysteme zwischen Politik und
empirischer U
¨
berpru¨fbarkeit. Dissertation. Frankfurt: Peter
Lang.
d’Arcy, A. (2000). The degree of determination of national
accounting systems—an empirical investigation. Schma-
lenbach Business Review, 53, 45–67.
d’Arcy, A. (2001). Accounting classi?cation and the interna-
tional harmonisation debate—an empirical investigation.
Accounting Organizations and Society, 26, 327–349.
Da Costa, R. C., Bourgeois, J. C., & Lawson, W. M. (1978). A
classi?cation of international ?nancial accounting practices.
International Journal of Accounting, 13, 73–85.
Goodrich, P. S. (1982). A typology of international accounting
principles and policies. AUTA Review, 14, 37–61.
Kruskal, J. B. (1964). Multidimensional scaling by optimizing
goodness-of-?t to a nonmetric hypothesis. Psychometrika,
29, 1–27.
Nair, R. D., & Frank, W. G. (1980). The impact of disclosure
and measurement practices on international accounting clas-
si?cation. Accounting Review, 55, 426–450.
Nobes, C. W. (1981). An empirical analysis of international
accounting principles: a comment. Journal of Accounting
Research, 19, 268–270.
Nobes, C. W. (1983). A judgemental international classi?cation
of ?nancial reporting practices. Journal of Business, Finance
and Accounting, 10, 1–19.
Nobes, C. W., & Matatko, J. (1980). Classi?cation of national
systems of ?nancial accounting. AUTA Review, 12, 57–78.
Ordelheide, D., & Semler, A. (1995). A reference matrix. In D.
Ordelheide, & KPMG (Eds.), Transnational accounting
(TRANSACC) (pp. 1–67). Basingstoke: McMillan.
Ordelheide, D., KPMG. Transnational accounting (TRANS-
ACC). Basingstoke: McMillan.
Roberts, A. D. (1995). The very idea of classi?cation in inter-
national accounting. Accounting Organizations and Society,
20, 639–664.
206 A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206
doc_921231309.pdf
In this issue of this journal, Christopher Nobes comments on my classification of accounting systems [Accounting
Organizations and Society (2001) 26, 327]. He concludes that the classification is unsound because the nature of the
data made them unsuitable and they contained errors. I took as a source for my selection process a database that
reflects the decisions of the EU directives and the Transnational Accounting (TRANSACC) editors to describe
important recognition and valuation rules. This process was made explicit and followed reasonable principles.
According to the accuracy issues raised, some discrepancies could be explained by different interpretations. Additionally,
the conclusion is still valid as the results are shown to be robust to variations of weightings and methods hence
also to unsystematic failures.
Accounting classi?cation and the international harmonisation
debate: a reply to a comment
Anne d’Arcy*
Deutsche Bank AG, Taunusanlage 12, D-60325 Frankfurt am Main, Germany
Abstract
In this issue of this journal, Christopher Nobes comments on my classi?cation of accounting systems [Accounting
Organizations and Society (2001) 26, 327]. He concludes that the classi?cation is unsound because the nature of the
data made them unsuitable and they contained errors. I took as a source for my selection process a database that
re?ects the decisions of the EU directives and the Transnational Accounting (TRANSACC) editors to describe
important recognition and valuation rules. This process was made explicit and followed reasonable principles.
According to the accuracy issues raised, some discrepancies could be explained by di?erent interpretations. Addition-
ally, the conclusion is still valid as the results are shown to be robust to variations of weightings and methods hence
also to unsystematic failures.
#2003 Published by Elsevier Ltd.
Introduction: Nobes’ concerns
In this issue of this journal, Christopher Nobes
comments on my classi?cation of accounting sys-
tems (d’Arcy, 2001).
1
He concludes that the classi-
?cation is unsound because the nature of the data
made them unsuitable and they contained errors.
In particular the following concerns were raised:
1. That I did not question counter-intuitive
results for Australia but just sought to
explain them. This research method is then
associated with other classi?cation
approaches e.g. Da Costa, Bourgeois, and
Lawson (1978) or Goodrich (1982).
2. Important issues in a data set may be
swamped by trivial ones. Additionally, as
the method of data collection is sub-
jective, the impression of objectivity is
misleading.
3. Other aspects of suitability, e.g. the split
between individual and group accounting
rules and the inclusion of variables that
describe accounting rules according to the
distributability of income.
4. Errors or disputable interpretations.
As it is not possible to describe all the methods
used in and the ?ndings of a 300 page study
0361-3682/$ - see front matter # 2003 Published by Elsevier Ltd.
doi:10.1016/j.aos.2003.10.001
Accounting, Organizations and Society 29 (2004) 201–206
www.elsevier.com/locate/aos
* Tel.: +49-69-910-43035; fax: +49-69-910-47532.
E-mail address: [email protected] (A. d’Arcy).
1
Readers will note that I have commented on Nobes’ paper,
and should be aware that he also made helpful comments on
previous drafts. I also thank Christine Roßbach and David
Powell for helpful comments.
(d’Arcy, 1999), I shall brie?y discuss some aspects
of these concerns in the following analysis.
Objective of the study and interpretation of some
results
First of all, I would like to make it clear that the
purpose of my work was not to prove that my
classi?cation described accounting systems in a
better or more comprehensive way than other
approaches. I simply wanted to focus on one
aspect of accounting systems that, in my view, had
not yet been investigated in su?cient detail,
namely ?nancial reporting rules (d’Arcy, 2001, p.
332). I felt that this dimension of an accounting
system might be important in connection with
recent accounting harmonisation debates. First of
all, though, it was necessary to examine former
classi?cation approaches, suggest technical
improvements and compare the results in order to
gain a better understanding of the di?erences.
I had learned from previous attempts that
results should always be interpreted with caution
due to the limitations of databases and the
statistical techniques applied. However, I do not
accept empirical ?ndings just because they ful?l
my or others’ previous expectations. Empirical
research should allow for new insights. So, results
that seem implausible at ?rst sight do not auto-
matically indicate poor methods or data. On the
contrary, they need further investigation in order
to reveal additional aspects not re?ected in the
preceding underlying hypothesis. This is how I
tried to explain the results for Australia which
showed outsider characteristics compared with
other national systems under review. But I do not
believe that this result was counter-intuitive. When
examining the data for Australia some aspects
emerged as being special, e.g. application of the
equity method. In most systems the equity method
is required for the valuation of associated compa-
nies. In Australia, the cost method is applied and
equity method information only has to be dis-
closed. Another example can be found in the rules
for revaluation accounting, e.g. the revaluation of
intangible assets is permitted without an active
market.
The factual outcome of my study was that the
Australian system had an outsider position as
regards speci?c ?nancial reporting rules when
compared to the 14 other systems under review. I
did not contend that the Australian accounting
system in general has an outsider position. In fact,
when using the same database for a di?erent
research purpose, I obtained di?erent groupings
(d’Arcy, 2000). An index ranking of the level of
?nancial accounting measurement regulation—
called the degree of determination—shows a
highly-regulated group including Australia, UK
and the US. This grouping clearly re?ects the
importance of national capital markets and the
former in?uence of the British Empire on
accounting systems, as does the loose link between
tax and ?nancial accounting (Craswell, 1995, p.
100).
Nobes then links my interpretations with pre-
vious studies by Da Costa et al. (1978) or Good-
rich (1982) that also classify accounting systems
‘‘based on data that had not been prepared for
that purpose’’. The weaknesses of these studies
have been discussed in other publications (e.g.
Nair & Frank, 1980, pp. 444–445, Nobes, 1981, p.
268, Nobes, 1983, p. 2, Nobes & Matatko, 1980,
p. 70). However, these studies based their
conclusions on a database prepared by other
researchers for a di?erent purpose. This database
is not suitable for classi?cation because it mixes
the levels of accounting rules and practices and
uses ordinal scales as input into statistical methods
requiring cardinal ?gures. Moreover, it can be
shown that even if the database were suitable, the
statistical method applied—factor analysis—
would itself constitute a serious limitation (Archer
& McLeay, 1992, p. 1).
Concerning the closeness of Switzerland and
UK shown by nonmetric multidimensional scaling
(MDS), it should be noted that I used this method
to test the cluster results for the macro structure
only. The advantage of nonmetric MDS is that no
assumptions need to be made about the under-
lying transformation function. The only assump-
tion is that the data are measured at the ordinal
level. That means that the di?erences in measure-
ment rules between the national accounting sys-
tems under review are only applied in their order
202 A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206
of ranking. Consequently, this method is very
strong in ?nding macro structures, but weak in
showing adequate micro structures.
2
This was one
of the main reasons why I did not use this method
as my main research method. Furthermore, as I
am aware of this shortcoming, I did not take the
micro results as my main ?ndings and did not
attempt to explain them either.
Important versus unimportant variables: composi-
tion of the database
The suitability of a database for a speci?c pur-
pose has in my view two dimensions: ?rst, how is
the database structured and how are the data
obtained; second, how are statistical methods used
to answer the research question.
The data are based on the Transnational
Accounting (TRANSACC) reference matrix
(TRM) established by Dieter Ordelheide and
myself (Ordelheide & Semler, 1995). The structure
and the entries were based on the respective
country chapters of TRANSACC (Ordelheide &
KPMG, 1995). The goal of TRANSACC was to
describe 15 accounting systems in a systematic
way by describing their history, the regulatory and
institutional background, accounting goals as well
as principles, respective formats, recognition and
valuation rules, consolidation regulation and dis-
closure requirements. The core of the descrip-
tion—the chapters on recognition, valuation and
consolidation, had a standardised structure, which
basically re?ected the formats of the 4th and 7th
EU accounting directives. The TRM also mirrored
these formats. The TRANSACC editors felt that
the topics referred to in the directives were impor-
tant parts of an accounting system and that rela-
ted recognition and valuation rules should
therefore be described for each national account-
ing system in the country contributions. The TRM
shows these rules in more standardised form.
The idea behind the database was to accept this
link to the EU formats and hence to use all entries
from the TRM. Given that the EU and the
TRANSACC editors have already identi?ed both
these recognition and valuation rules as impor-
tant, there is absolutely no reason to ignore this
assessment. However, as the TRM itself was not
suitable for my research design, I modi?ed the
data as described in d’Arcy (2001, pp. 334–335).
Part of this modi?cation involved disregarding
some variables due to insu?cient descriptions
which did not allow proper di?erentiation between
the systems, e.g. accounting for leases. I am fully
aware that, as a result of this, the database is not
all-embracing. Though the data could be improved,
there are nevertheless important rules described that
can be used as input into cluster analysis.
As Nobes has correctly noted, I did use judge-
ment in coding the data. But using judgement did
not mean that the variables were simply selected at
random. On the contrary, I applied two principles
when making my decisions. I took all the data
from the TRM (principle of completeness) and
only disregarded items that did not di?erentiate
properly between the systems (principle of e?ec-
tive separability). For example, the separability
function is the reason why I described di?erent
rules according to the simpli?ed valuation meth-
ods for stocks in only one variable—is the last in
?rst out-method (LIFO) forbidden or not. The
?rst in ?rst out-method (FIFO) or the average cost
method are allowed in all systems under review and
therefore do not qualify to be coded as variables
that e?ectively distinguish between systems.
I agree that weighting variables may have an
important impact on results. It is also true that
assigning weights does not necessarily add sub-
jectivity. For example, giving a double weight to
those variables that have a simple coding in my
database is easily justi?ed. However, instead of
simply justifying my choice of weighting variables
and applying the cluster method, I chose to ana-
lyse the sensitivity of results based on di?erent
weightings and di?erent cluster methods (d’Arcy,
2001, pp. 341–342). The idea was to show that
under extreme conditions the macro structures
2
In fact, adding a third dimension, the multidimensional
scaling results correspond for Switzerland and the UK more
clearly with the dendrogram derived from cluster analysis. The
goodness-of-?t value according to Kruskal (1964, p. 3) increa-
ses from 0.16672 ‘‘su?cient’’ to 0.09988 ‘‘good’’. But the three-
dimensional results are not taken because they may not be
reliable as the total number of parameters being estimated is
large relative to the number of data values.
A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206 203
remained stable so that neither di?erent weight-
ings nor di?erent cluster methods would sig-
ni?cantly change the results. I therefore varied the
similarity coe?cients as input into cluster analysis
to simulate weightings for the extreme condition
that all matches or all non-matches of variables had
a double weighting. It has been shown that in any
constellation double weightings would lead to the
same macro structure of clusters (d’Arcy, 1999, pp.
169–170, d’Arcy 2001, p. 342). Additionally, vari-
ation of the cluster methods—I compared the results
produced by the single linkage, complete linkage
and average linkage within groups methods—also
shows great robustness in the cluster results. Hence,
the ?ndings were proved to be robust against var-
iations of weightings and cluster methods.
Other aspects of suitability
I agree that there are some problems in distin-
guishing between individual and group accounts.
However, the EU and the TRANSACC editors
decided to make a distinction and I simply applied
this concept. It may be a very continental Eur-
opean view to distinguish the dataset for these two
types of accounting rules. However this di?er-
entiation may help to avoid a typical ‘‘Anglo-
American bias’’ where the dataset ignores di?erent
rules in both types of accounts. In most cases, the
rules for individual accounts are also valid in
the consolidation process for the group accounts.
But some accounting problems are treated
di?erently and are re?ected accordingly in the
database, e.g. the valuation of associated compa-
nies or the valuation of goodwill arising from
asset deals. The dataset is not all-embracing as it
does not include all accounting problems. How-
ever, concentration on one set of accounts, e.g.
group accounts, would ignore these di?erences.
Moreover, this di?erentiation also allows further
analysis. For example, clustering data for indivi-
dual and group accounts separately shows that the
Australian outsider position can be traced back
primarily to individual accounting rules. Based on
the second part of the database, which deals with
additional rules arising from the preparation of
group accounts, Japan comes out as an outsider.
Turning to the second concern, that variables
describing accounting rules according to the dis-
tributability of income should not be included, I
strongly recommend that those rules be con-
sidered. As distributable pro?t is equated in this
case with pro?t in general, these rules are also
relevant for consolidated statements and are thus
important characteristics of a national accounting
system. I have to admit that the labels of these
variables do not adequately convey their purpose
and could therefore be misunderstood. The issue is
whether the recognition of an asset or liability
a?ects pro?t or not. In some accounting systems,
the recognition of certain assets is only allowed
if the di?erence is not accounted for as pro?t. In
this case an equity item has to be recognised.
The pro?t and loss account is not a?ected by
this transaction. Accordingly, group pro?t is not
in?uenced by recognition of this asset. For
example, TRM di?erentiates between systems
that allow or require the recognition of develop-
ment costs. Additionally, a variable indicates
whether the capitalisation of development costs
in?uences pro?t by recognising a respective
equity item (including group pro?t) in the
respective periods or not.
I know that the concept of the determination
function of distributable pro?t is unknown in the
US or under International Accounting Standards
(IAS). For the related accounting method, however,
we ?nd comparable e?ects from an analysis of
certain IAS or US GAAP rules. For example IAS
39 or FAS 133 require the valuation of securities
classi?ed as available for sale at fair or market
value. The value di?erence only a?ects the respective
equity item and not pro?t. If a valuation di?erence
a?ects pro?t, this can be seen as an important
characteristic of a national accounting system.
Errors and disputable interpretations
Due to the limits on space I would like to focus
on two examples only. More generally, I still
believe that some of the ‘‘Error’’ issues listed
above are caused by di?erent interpretations.
Nobes states that the exception to the obligation
to prepare group accounts for small groups does
204 A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206
not exist in the UK for listed companies. As I
generally described rules applicable to corpor-
ations only and—if there is a di?erence—to listed
companies, this exception was not expected to be
included in the database. However, in Germany
and some other continental European countries
there are no speci?c rules for listed companies (or,
there were none in 1995). So in Germany we have
the general rule that small groups—de?ned by
certain criteria—do not have to publish group
accounts. Listed companies—as set out in com-
pany law—have to follow the rules for large cor-
porations. Following the systematics of law, we
?nd the description of the general rule for all
companies, but listed companies are character-
istically not small. Thus, following the principle
of completeness in data selection, this variable is
still useful for di?erentiating between national
accounting systems and should be included.
Another issue is the interpretation of the mean-
ing of a rule. As an outcome of the 7th EU direc-
tive, all EU member states had to transpose into
national law the rule that subsidiaries with dis-
similar activities should not be consolidated. Con-
sistently, we ?nd this prohibition for all EU
member states, including the UK, in TRANSACC.
Even if in practice all subsidiaries have to be con-
solidated because dissimilarity is de?ned very
stringently, the code ‘‘exclusion required’’ re?ects
the rule properly.
These and other examples show the great di?-
culties in comparing national ?nancial reporting
rules. Some confusion is partly caused by di?erent
translations. Additionally, in some cases the con-
cepts do not ?t, as Nobes has demonstrated using
the example of the terms provisions and liabilities.
However, other di?erences, like the two examples
described above, arise because of di?erent inter-
pretations and are neither right nor wrong.
Conclusion
According to Nobes, the dataset is neither sui-
table nor accurate. Surprisingly, he states that the
subjectivity of the data selection process is a
weakness of this approach. In my view empirical
research is always subjective, which means that
this shortcoming cannot be resolved in general.
But I took as a source for my selection process a
database that re?ects the decisions of the EU
directives and the TRANSACC editors to describe
important recognition and valuation rules. Addi-
tionally, I tested the robustness of the results
rather than introduce weightings or change the
selection. This process was made explicit as sug-
gested by Roberts (1995, p. 660). As the goal of
TRANSACC and the TRM is to describe and
compare national accounting systems, the same is
the case for the modi?ed database. Furthermore, I
am not aware of any ‘‘objective’’ empirical method
to ?nd clusters of national accounting systems. But
I believe, that the consideration of the judgement
of the EU as well as the TRANSACC editors
enhances reliability.
As no database can be all-embracing, the di?er-
entiation between individual and group account-
ing rules does not impair suitability of the
database just because of the fact that some rules
are not described. This di?erentiation re?ects the
structure of the 4th and 7th EU Directives and
allows to gain further insights. I also place
emphasis on the inclusion of variables re?ecting
accounting rules regarding pro?t distributability.
These rules a?ect the calculation of accounting
pro?t—even in the group accounts—and are in
this regard an important part of a national
accounting system.
According to the accuracy issues raised, some
discrepancies could be explained by di?erent
interpretations. However, unsystematic failures
are not unavoidable. But taking the robustness of
the results to variations of weightings and methods
into account, it is unlikely that those failures
would harm the ?ndings of this study. So my
conclusions are still valid.
References
Archer, S., & McLeay, S. (1992). European ?nancial reporting:
issues of de?nition and classi?cation of national systems of
?nancial reporting. Paper presented at the Annual Congress
of the European Accounting Association. Madrid, Spain.
Crasswell, A. (1995). Australia individual accounts. In D.
Ordelheide, & KPMG (Eds.), Transnational accounting
(TRANSACC) (pp. 69–167). Basingstoke: McMillan.
A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206 205
d’Arcy, A. (1999). Gibt es eine anglo-amerikanische oder
eine kontinentaleuropa¨ische Rechnungslegung?—Klassen
nationaler Rechnungslegungssysteme zwischen Politik und
empirischer U
¨
berpru¨fbarkeit. Dissertation. Frankfurt: Peter
Lang.
d’Arcy, A. (2000). The degree of determination of national
accounting systems—an empirical investigation. Schma-
lenbach Business Review, 53, 45–67.
d’Arcy, A. (2001). Accounting classi?cation and the interna-
tional harmonisation debate—an empirical investigation.
Accounting Organizations and Society, 26, 327–349.
Da Costa, R. C., Bourgeois, J. C., & Lawson, W. M. (1978). A
classi?cation of international ?nancial accounting practices.
International Journal of Accounting, 13, 73–85.
Goodrich, P. S. (1982). A typology of international accounting
principles and policies. AUTA Review, 14, 37–61.
Kruskal, J. B. (1964). Multidimensional scaling by optimizing
goodness-of-?t to a nonmetric hypothesis. Psychometrika,
29, 1–27.
Nair, R. D., & Frank, W. G. (1980). The impact of disclosure
and measurement practices on international accounting clas-
si?cation. Accounting Review, 55, 426–450.
Nobes, C. W. (1981). An empirical analysis of international
accounting principles: a comment. Journal of Accounting
Research, 19, 268–270.
Nobes, C. W. (1983). A judgemental international classi?cation
of ?nancial reporting practices. Journal of Business, Finance
and Accounting, 10, 1–19.
Nobes, C. W., & Matatko, J. (1980). Classi?cation of national
systems of ?nancial accounting. AUTA Review, 12, 57–78.
Ordelheide, D., & Semler, A. (1995). A reference matrix. In D.
Ordelheide, & KPMG (Eds.), Transnational accounting
(TRANSACC) (pp. 1–67). Basingstoke: McMillan.
Ordelheide, D., KPMG. Transnational accounting (TRANS-
ACC). Basingstoke: McMillan.
Roberts, A. D. (1995). The very idea of classi?cation in inter-
national accounting. Accounting Organizations and Society,
20, 639–664.
206 A. d’Arcy / Accounting, Organizations and Society 29 (2004) 201–206
doc_921231309.pdf