Description
The relocation of more polluting industries in poorer countries due to gaps in environmental standards is known as the pollution haven effect, whereby the scale and the composition of output change across countries.
Questioni di Economia e Finanza
(Occasional Papers)
Environmental regulation and revealed comparative advantages
in Europe: is China a pollution haven?
by Daniela Marconi
N
u
m
b
e
r
67
J
u
n
e
2
0
1
0
Questioni di Economia e Finanza
(Occasional papers)
Number 67 – June 2010
Environmental regulation and revealed comparative advantages
in Europe: is China a pollution haven?
by Daniela Marconi
The series Occasional Papers presents studies and documents on issues pertaining to the
institutional tasks of the Bank of Italy and the Eurosystem. The Occasional Papers appear alongside
the Working Papers series which are specifically aimed at providing original contributions to economic
research.
The Occasional Papers include studies conducted within the Bank of Italy, sometimes in
cooperation with the Eurosystem or other institutions. The views expressed in the studies are those of the
authors and do not involve the responsibility of the institutions to which they belong.
The series is available online at www.bancaditalia.it.
ENVIRONMENTAL REGULATION AND REVEALED COMPARATIVE
ADVANTAGES IN EUROPE: IS CHINA A POLLUTION HAVEN?
by Daniela Marconi
*
Abstract
The relocation of more polluting industries in poorer countries due to gaps in
environmental standards is known as the pollution haven effect, whereby the scale and the
composition of output change across countries. Changes in the composition of the output
mix might translate into changes of comparative advantages across countries, as revealed by
trade flows. This paper focus on this issue and looks at the changes of bilateral revealed
comparative advantages (RCAs) in the last decade between China and the major fourteen EU
countries (EU14). Using industry level data on bilateral trade, air pollution, water pollution
and several measures of environmental stringency, we find that, controlling for other factors
that may have affected RCAs, such as labor costs, on average our EU14 countries have kept
or improved their advantages with respect to China in both water polluting industries (such
as paper and agro-based industries) and air polluting industries (such as basic metals and
chemicals), while they have lost competitiveness in the more clean industries (such as
machinery and fabricated metals).
JEL Classification: F14, F18.
Keywords: revealed comparative advantages, environmental regulation, industrial pollution.
Contents
I. Introduction........................................................................................................................... 5
II. Pollution abatement costs and environmental regulation ................................................... 8
III. Pollution intensities by industry and bilateral RCA......................................................... 11
IV. The empirical strategy...................................................................................................... 13
V. Results ............................................................................................................................... 15
VI. Conclusion........................................................................................................................ 20
References .............................................................................................................................. 21
Appendix ................................................................................................................................ 23
* Bank of Italy, Economics, Research and International Relations.
5
I. Introduction
1
In recent years the need to preserve and improve environmental quality has solicited
increasing efforts to abate pollution worldwide; renewed effort in rich countries has become urgent
fearing the still unknown consequences of climate change. The debate on sustainable emission
targets and required abatement trends has become particularly intense at all levels raising
coordination problems and free-riding concerns; in fact, while the quality of the environment does
not depend only on the action taken within each country’s boundaries, the burden of abatement
costs can only be effectively imposed by governments on domestic producers and consumers.
2
As long as a “healthier natural environment” is a normal good, demand for it tends to be
higher in richer countries which impose more stringent environmental regulations as compared to
poorer ones (so called “environmental Kutznets curve” hypothesis; Copeland and Taylor, 2003).
However, pollution abatement poses additional burdens on domestic firms, especially those
operating in the most polluting industries, shifting part of the inputs away from production to
pollution abatement. If the cost burden is significant enough, it might hurt the international
competitiveness of domestic firms, compared to firms located in countries with weaker
environmental standards. The relocation of more polluting industries to poorer countries due to gaps
in environmental standards is known as the pollution haven effect, whereby the scale and the
composition of output change across countries (Copeland and Taylor, 2003).
3
The existence and the
magnitude of such an effect depends on two things: (a) whether environmental regulations impose
substantial additional costs on polluting industries, and (b) whether, absent other compensative
policies, regulation differentials are large enough to impact on industry location, output
composition and trade.
4
Changes in the scale of activity affect directly profits and jobs within a country; changes in
the composition of the output mix translate into changes of comparative advantages across
1
I would like to thank Xiaolan Fu, Herman Vollebergh and Valeria Rolli for insightful comments and suggestions on
earlier versions of this paper. I am also grateful to Pietro Barone, Ivan Faiella, Marco Marinucci for valuable
discussions.
2
The possible option of imposing additional taxes on imports of goods from polluting producers encounters major legal
and practical problems. For a discussion on coordination issues see OECD (2008).
3
Copeland and Taylor (2004) distinguish between the pollution haven effect and the pollution haven hypothesis. The
first is the effect, at the margin, on trade flows and plant location of tightening up environmental regulation in richer
countries; the hypothesis instead refers to the implications for plant location and trade flows of a change in trade
regimes between countries with different environmental regulations. In our analysis both things are at play, in that in the
last ten years environmental regulation has become more stringent in richer countries and at the same time trade barriers
have been reduced. Since our focus is on environmental regulation differentials we refer to pollution haven effect.
4
We might consider the environment as an additional factor of production, together with capital, labor and land; as
environmental services in poorer countries become relatively cheaper they will be embodied in a larger share in their
exported goods. However, there is some evidence (Eliste and Fredriksson, 2002, and Grether and de Melo, 2004) that
the most polluting industries often obtain compensating transfers from governments and tend to benefit from higher
trade barriers.
6
countries. This paper focuses on this latter issue.
5
So far very few studies have tested the pollution
haven effect on revealed comparative advantages in trade (Grether and de Melo, 2004 and Cole et
al., 2005). The empirical literature has mainly looked at the effects of environmental regulation on
plant location decisions and foreign direct investment (FDI) flows, either among different regions
within the same country (Dean et al., 2005 and Zhang and Fu, 2008, for China and, among others,
Keller and Levinson, 2002, for the US) or between countries with different levels of environmental
regulation.
6
Results, in particular for the US, point to a weak relationship between plant location
decisions, or FDI flows, and environmental regulation. Exploring the link between trade flows and
environmental regulation seems more appropriate as it allows uncovering the impact of both plant
location choices and other policy-induced changes in industrial output sizes. The literature looking
at trade flows, however, also reaches mixed conclusions, with results being very sensitive to the
choice of countries, the empirical specification and the definition of environmental regulation.
7
In our analysis our measure of international specialization is an index of trade revealed
comparative advantage (RCA). Previous studies with similar dependent variables tend to find no
clear evidence of pollution haven effects (Grether and de Melo, 2004, for a set of 52 countries and
Cole et al., 2005 for the US). We consider the major fourteen EU countries (EU14) and look at the
changes in the contribution of each industry to the bilateral trade balance with China in the last
decade. The recent surge of China as world’s top exporter is often attributed not only to its low
labour costs and rapid capital accumulation but also to its “export dumping” due to weak
environmental standards compared to richer countries. On the other hand, some EU countries are
regarded as those which have committed to the most stringent environmental regulation worldwide.
Aggregate green house gas emission intensity in the EU fell in the last decade, but
comparatively less in manufacturing and construction industries than in other sectors of the
economy; the decline has been particularly intense from 1996 to 2000; a renewed effort seem to be
in place again since 2003 (Fig. 1). At the same time, environmental protection expenditures have
remained constant in terms of GDP, around 0.4 per cent.
On the other hand, as documented in Dean and Lovely (2008), in China fast industrialization
since 1978 has led to a rapid deterioration of the air quality and of many water sources. Against this
background, in recent years there have been significant improvements in the environmental
protection legislation, although enforcement is still very weak due to diverging economic interests
5
For extensive surveys on the broader relationship between environmental regulation and international competitiveness
see SWQ (2006) and United Nations (2006).
6
A review of these studies can be found in Copeland and Taylor (2004) and Zhang and Fu (2008).
7
Grether and de Melo (2004) and Pasurka (2008) offer a summary of the various specifications and findings. Table A4
in the appendix summarize the findings of the papers mostly related to the present article.
between central and local governments (Zhang and Fu, 2008). Nonetheless, recent available data
show a certain effort in pollution abatement and treatment: in the last decade total investment in
industrial pollution treatment (PTI) has shown an increasing trend in terms of GDP, surpassing 1.4
per cent in 2008, while water and air industrial emissions (per unit of output) declined steadily (Fig.
2).
Fig. 1 EU15: Green house gas emission
intensity
(tons/thousand value added euros; 1995 prices)
Fig. 2 China: Total investment in pollution treatment
(PTI) as percentage of GDP;
Total waste water discharge (kilos/yuan output; 1995
prices) and Sulphure dioxide (SO
2
; kilos/thousand yuan
output; 1995 prices) in industry
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Manufacturing and Construction All activities
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Total industrial waste water SO2 PTI
Source: EuroStat data and author’s elaboration. Source: CEIC and author’s elaborations
In this paper we analyze the evolution of the structure of bilateral trade between EU countries
and China in relation to environmental regulation; in doing so, we suggest a new strategy to look at
this relationship, which allows to overcome the endogeneity problems associated with the measures
of environmental regulation usually adopted in the literature, such as pollution abatement costs
(PAC).
8
We use industry-level data on air-and-water-pollution intensities in China and a cross-
country index of environmental stringency first constructed by Dasgupta et al. (1995) and recently
extended by Eliste and Fredriksson (2002). Robustness checks are conducted by using: i) an
additional measure of pollution intensity by industry (the global warming potential (GWP) of
emissions per unit of output, in Europe); ii) two additional measures of environmental stringency
(GDP per capita and greenhouse gas emissions).
Preliminary findings on the evolution of the structural bilateral trade between China and
EU14 countries in 18 manufacturing industries in the period from 1996 to 2006, indicate that, after
controlling for other factors affecting trade flows (such as labour costs), there is no evidence of a
pollution haven effect. In particular we find that:
7
8
On the endogeneity issue, see Section II.
8
i) on average our EU14 countries have kept or improved their comparative advantages
with respect to China, as revealed by their bilateral trade, in both water-polluting
industries (such as paper and agro based industries) and air-polluting industries (such
as basic metals and chemicals);
ii) on average our EU14 countries have, instead, lost competitiveness in cleaner and
more internationally-mobile industries (such as communication equipment and office
and computing machinery), presumably in response to unfavourable unit-labour-cost
differentials and higher capital accumulation in China.
The paper unfolds as follows: Section II discusses endogeneity problems of pollution
abatement costs as proxy for environmental regulation; Section III presents briefly some statistical
evidence on pollution intensities and revealed comparative advantages in China and Europe;
Section IV presents our estimation strategy; Section V reports estimation results and Section VI
concludes.
II. Pollution abatement costs and environmental regulation
In principle, the pollution abatement effort in a given country reflects the stringency of its
environmental regulation. To evaluate and compare the effective cost burden of pollution abatement
and control (PAC) expenditures we would like to have a reliable measure of such expenditures by
industries and time. Unfortunately such information is not readily available for a sufficient number
of countries and time length, and comparison among countries for which data are available must be
taken with great caution, since definitions differ from country to country.
9
The most comprehensive
set of data on PAC expenditures and investments, sometimes also specified as pollution treatment
expenditures (PTE) or investments (PTI), are collected by the OECD (OECD Environmental
Compendium 2008) and the Euro Stat (2008). In the OECD data the breakdown within countries at
most distinguishes between public and business sector; in the Euro Stat data the breakdown refers to
broad industries: manufacturing, electricity, gas and water supply and mining and quarrying. The
time span covered varies from country to country, for some countries data date back to the eighties.
However, even in these favourable cases, time series are highly discontinuous and again hardly
comparable across countries. Moreover, as data are too aggregated across sectors, PAC
expenditures are likely to be endogenous, in that, as the output composition within a macro-
9
For a comprehensive survey of available industrial-level PAC data, and their measurement and comparability
problems, see Pasurka (2008).
9
industrial-branch changes, PAC expenditures change accordingly.
10
Therefore, such measures are
unsuitable to reflect the stringency of environmental regulation in a given country at a given time.
An alternative way to evaluate pollution abatement costs is by observing polluting emissions
per unit of output over time, by highly disaggregated industrial sectors. In fact, under the
assumption that the output composition within a finely defined industrial branch does not change
dramatically, the evolution of emission intensities would provide an indirect indication of the
stringency of environmental regulations. Unfortunately, however, there are no time series of
emissions per unit of output by industrial sectors and by country readily available. Emissions per
unit of output at 2-digit ISIC classification are available for China in 1995 and 2004 and for six
major European countries mainly for the year 2000.
11
For the EU and other advanced countries,
emissions per unit of output for the last fifteen years can be recovered only for the economy as a
whole or for macro-industrial branches, however, being endogenous to changes in output mix, are
not suitable to our analysis.
12
To overcome endogeneity and data shortage problems, in our empirical approach we use a
country-level index of environmental regulation to be interacted with an industry-level index of
pollution intensity; the methodology will be made clear in Section IV. As for country-level indexes
of environmental regulation, in our empirical analysis we use three different measures. The first
(STRING
i
) is a cross-country index, first constructed by Dasgupta (1995) and recently extended by
Eliste and Fredriksson (2002), based on detailed information about the environmental regulatory
framework on water, air pollution, land use and biodiversity; higher values of the index correspond
to higher level of environmental standards.
13
A second measure of environmental stringency is
given by the level of GDP per capita in 1995 (PCGDP
i
); higher levels of GDP per capita should be
associated with higher environmental standards. Finally, we consider an index of greenhouse gas
10
For an extensive discussion on the endogeneity problems that arise when using pollution abatement costs as proxy of
environmental stringency, see Levinson and Taylor (2008).
11
For Italy, detailed data for industrial emissions at 2-digit NACE-ISIC classification have been recently released by the
National Statistics Bureau, ISTAT (http://www.istat.it/dati/dataset/20070625_00/).
12
The EuroStat is developing a framework to collect industry-level time-series data on emissions linked to economic
accounts (EuroStat 2001). Up to now EuroStat has made available only data on total Carbon dioxide emissions from
1995 to 2004 classified by NAICS in manufacturing, mining and services for a large number of countries within EU27
(http://epp.eurostat.ec.europa.eu/portal/page?_pageid=0,1136239,0_45571444&_dad=portal&_schema=PORTAL).
However an inspection of the data reveal that: (a) there are many missing values over time and across industries; (b) the
number of manufacturing branches for which it is possible to express emissions in intensive form, using gross value
added classified by NACE, is at most 11; (c) intensities can be expressed only in terms of value added, because output
volumes are not available for the same classification; (d) emission intensities show several suspicious inconsistencies
which makes it hard to believe that those emissions can be confronted with national account data on value added. Data
and elaborations are available from the author upon request.
13
The index is based on Country Reports prepared for the 1992 United Nations Conference on Environmental and
Development.
emissions per unit of output in manufacturing and construction, averaged over the period 1995-2005
(GHGE
i
); countries with more stringent regulation should score lower values of this index
14
Figure 3 below shows the dispersion of our proxies for environmental stringency across the
European countries in our sample; Table 1 sows how the three measures are correlated with each
other.
15
Even though we concentrate on STRING
i
, PCGDP
i
and GHGE
i
in our analysis, for
completeness purposes, we also report two additional indexes, the Environmental Protection Index
(EPI
i
) and the Sustainable Society Index (SSI
i
); countries with better environmental standards
should score higher values for these indexes.
16
It is worth noting that the dispersion across EU
countries is appreciable; in general correlations between these indexes are quite high, with the
exception of those between PCGDP
i
and EPI
i
and PCGDP
i
and SSI
i
.
Fig. 3 Environmental regulation proxies in fourteen EU countries
Source: author’s elaborations on data from Eliste and Fredricson (2002), Euro Stat and World Bank.
14
This latter measure is from Euro Stat and reported in Table A3 in the Appendix.
15
The index PCGDP is constructed by taking the ratio of each country per capita GDP to the average of the sample
countries, multiplied by 100; the index GHGE is given by the ratio of the average emission intensities in the 1995-2005
period in each country to the sample average, multiplied by 100. As shown in Table 1 the index by Eliste and
Fredriksson (STRING) shows a correlation coefficient equal to 0.53 with PCGDP in 1995, such a correlation grow to
0.8 in 2005; the correlation coefficient with GHGE is equal to -0.74; the correlation coefficient between PCGDP and
GHGE is equal to -0.56.
10
16
Regression results with these two indexes are qualitatively similar to those found in the main regressions; they
available from the author upon request.
Table 1. Correlations between measures of environmental stringency
STRING
i
PCGDP
i
GHGE
i
EPI
i
SSI
i
STRING
i
1.00
PCGDP
i
0.53 1.00
GHGE
i
-0.74 -0.56 1.00
EPI
i
0.50 -0.03 -0.62 1.00
SSI
i
0.53 0.29 -0.66 0.74 1.00
III. Pollution intensities by industry and bilateral RCA
The 18 manufacturing industries considered are ranked by pollution intensity according to
sulphur dioxide (SO
2
) air emissions and chemical oxygen demand (COD) of water discharge in
China; an additional measure is provided by the global warming potential (GWP) emissions per
unit of output in Italy, which is chosen as representative for Europe (see columns 1 to 6 in Table
2).
17
Industries are further classified as either resource-based (RB) or non-resource-based (NRB).
This distinction is important, as RB industries are characterized by a very low degree of
international mobility compared to NRB ones. Consequently, they may react differently to changes
in environmental regulation. We will elaborate further on this issue in Section VI.
For each industry we also compute an index of bilateral trade-revealed comparative advantage
(RCA) with respect to China. Our measure is not straightforward and deserves some explanations.
We compute sectoral RCAs for each EU14 country as follows:
100 * *
18
1
18
1
18
1
18
1
18
1
18
1
? ? ? ?
? ?
= = = =
= =
+
+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
+
?
?
+
?
=
j j
ij ij
ij ij
j j
ij ij
j j
ij ij
ij ij
ij ij
ij
M X
M X
M X
M X
M X
M X
RCA
. (1)
Where: i is the EU reporting country (i=1,...14); j is the industrial sector (j=1,...,18); ( )
are country i’s bilateral exports (imports) to (from) China in sector j; values are expressed in current
US dollars. Equation (1) measures directly the contribution of each sector to the bilateral trade
ij
X
ij
M
11
17
We concentrate on SO
2j
and COD
j
emissions because the other two emissions available for China, SMOKE and
DUST (particulate), are strongly correlated with SO
2j.
The Global Warming Potential (GWP) emission index, instead,
aggregates three greenhouse gases CO
2
, N
2
O and CH
4
(with weight 1, 310 and 21 respectively). Moll et al. (2007) have
calculated GWP per unit of output by 2-digit ISIC classification in 1995 or 2000 for seven European countries
(Denmark, Germany, Italy, Netherlands, Spain, Sweden and United Kingdom). Italian GWP emissions show the highest
correlation with any other European country, on average the correlation coefficient is equal to 0.91.
balance; such a measure does not depend on the size of the overall balance, but only on its
composition.
18
An inspection of Table 2 reveals that RCA changes over the period 1996-2006 for the average
of EU14 countries (column (7)) tend to be negatively correlated with the pollution intensity
measures, that implies that Europe’s comparative advantages with respect to China have actually
tended to worsen in the cleanest industries and to improve in the dirtiest ones.
19
Table 2. S02 and COD emissions in China (kilos per thousand yuan output, 1995 yuan) and changes of RCAs in
EU14 by industrial sector.
(1) (2) (3) (4) (5) (6) (7)
Sectors
COD
emissions
per unit of
output in
China
(2004)
SO2
emissions
per unit of
output in
China
(2004)
GWP
emissions
per unit of
output in
Italy
(2000)
COD
rank
SO2
rank
GWP
rank
EU14
Average
RCA changes
(1996-2006)
Resource-based (RB) industries
Coke and Petroleum 0.08 0.85 5.11 10 6 3 -0.01
Pulp, paper, paper products, printing and publishing 5.21 1.41 1.46 1 2 5 0.01
Food products, beverages and tobacco 1.16 0.44 0.34 2 8 8 0.16
Wood 0.92 1.15 0.33 3 4 9 -0.15
Non-resource-based (NRB) industries
Non-metallic minerals 0.14 4.26 9.76 6 1 1 -0.18
Basic metals 0.12 1.26 5.96 7 3 2 0.99
Chemicals 0.67 1.13 1.80 4 5 4 0.39
Rubber and Plastics 0.10 0.26 0.49 9 11 6 0.55
Motor vehicles 0.06 0.06 0.44 12 15 7 1.58
Textiles, textile products, leather and footwear 0.66 0.54 0.31 5 7 10 2.05
Fabricated metals 0.08 0.32 0.30 11 9 11 -0.19
Machinery 0.05 0.18 0.23 14 12 12 -0.02
Transport equipment 0.06 0.06 0.22 13 16 13 -0.42
Furniture and Other Mfg. 0.12 0.28 0.20 8 10 14 1.10
Medical, Precision and Optical Instruments 0.05 0.08 0.19 15 14 15 1.01
Electrical Machinery 0.02 0.16 0.16 18 13 16 0.62
Communications Equipment 0.03 0.03 0.16 17 18 17 -5.10
Office and Computing Machinery 0.03 0.03 0.06 16 17 18 -2.40
Source: Dean and Lovely (2008); Moll et al. (2007); EUKLEMS database; OECD-STAN bilateral trade database and author’s
elaborations.
Note: Sectors are classified according to the 2-digit ISIC rev. 3 nomenclature. RB indicates resource-based industries, NRB non-
resource-based industries based on UNIDO definitions reported in Malatu et al. (2004). RCA are based on equation (1), changes are
computed on the difference between the average value of index over the period 2001-2005 and the average value of the index in the
period 1996-2000.
18
The index varies between -50 and +50 and the sum across all the j sectors is equal to zero. A positive (negative) sign
of the index in sector j indicates that the reporting EU country has a comparative advantage (disadvantage) in that
sector, relative to all other sectors. This index allows to rank the products according to their importance, takes into
account intra-industry flows and allows for international comparisons (Marconi and Rolli, 2008). Detailed data for each
country in the sample are reported in Table A2 in the Appendix.
12
19
The evolution of bilateral trade RCAs with respect to China clearly does not describe the entire sectoral evolution of
production in EU14; indeed, if we look at the evolution of value added by industry we find a positive correlation with
the rankings by pollution intensity (0.5), meaning that in the last decade the cleanest industries have benn also the most
dynamic ones in Europe (see also Appendix, fig. A1a and A1b).
Also, it is interesting to note that, despite possible differences in pollution intensities between
EU countries and China due to different output mix or eco-efficiency, the ranking of industries by
SO
2
emissions in China (column (5)) and by GWP emissions in Europe (column (6)) are quite
similar (correlation of 0.8).
IV. The empirical strategy
The model - In order to measure the impact of environmental regulation on the evolution of the
structure of bilateral trade between EU14 countries and China, we want a dependent variable as
much as possible independent of macroeconomic effects. In much of the existing empirical
literature, the dependent variable is specified as industry’s net exports normalized by industry’s
value added, we use, instead, our index of RCA described in equation (1) departing from the
literature in two ways: (a) we take bilateral, instead of total trade; (b) we normalize net exports by
gross flows (a normalization better suited for international comparisons).
20
We estimate the following equations:
) 2 ( . *
* 2
4
3 ) 1995 ( 2 ) 1995 ( 1 ) 1999 2006 (
ij i j
i j ij ij i j ij
string COD
string SO ULC RVASH RCA
? ?
? ? ? ? ?
+ +
+ + + + = ?
?
ij i j ij ij i j ij
string GWP ULC RVASH RCA ? ? ? ? ? ? + + + + + = ?
?
*
3 ) 1995 ( 2 ) 1995 ( 1 ) 1999 2006 (
. (3)
Where our dependent variable is the change of bilateral RCA
ij
between country i
and China in sector j, in the 1999-2006 period; RVASH
ij
is the share of value added of sector j in
country i, in real terms, relative to the average sector share across EU14 in 1995.
) 1999 2006 ( ?
?
ij
RCA
21
ULC
ij
is the unit
labour cost in sector j in country i; it is measured as nominal wage over value added at 1995 prices.
The next three variables are our variables of interest, i.e., the pollution haven variables. These
variables are constructed interacting industry-specific pollution intensities (as measured by SO
2j
emissions, CODj emissions, or, alternatively in equation (3), GWPj emissions) with country-
specific indexes of environmental regulation (string
i
). The methodology of interacting an industry
characteristic (in our case pollution intensity) with a country characteristic (in our case, the
stringency of environmental regulation) was proposed by Rajan and Zingales (1998). These
interacted variables have the advantage of varying across industries and countries and should
capture the pollution haven effect. We expect that a more stringent environmental regulation should
20
See Bugamelli, 2001; OECD, 2005 and Marconi and Rolli, 2008.
13
21
The value added share of sector j in country i is normalized with respect to the EU14 average share of sector j. A
similar role would be played by trade specialization by sector at the beginning of the period. Indeed, running
regressions replacing RVASH with the index of “Export specialisation relative to OECD23 and total manufacturing
(XSPEC23M)” taken from OECD STAN, leaves results qualitatively unchanged. Results are available from the author
upon request.
induce a worsening of RCAs in more polluting industries and, conversely, an improvement in less
polluting ones.
RVASH
ij
and ULC
ij
are intended to capture, respectively, Hecksher-Ohlin-Samuelson
(henceforth H-O-S) and Ricardian determinants of RCAs. According to the H-O-S theory, which
assumes same technologies across countries, countries specialize according to their relative factor
endowments; this means that relative industrial composition, as captured by the relative shares of
value added in industry j of country i, RVASH
ij
, should reflect relative endowments in European
countries. For the Ricardian argument, instead, countries tend to specialize according to their
relative productivity, summarized by unit labor costs (ULC
ij
). As it is plausible that both
endowments and technology play a role in shaping RCAs with respect to China, we include both
variables in our regression. Negative (positive) coefficients for initial specialization differentials
should capture convergence (divergence) of endowments, while unit labor costs differentials should
capture Ricardian comparative advantages.
22
In order to correct for high volatility in trade data first we take 3-year moving averages of
RCAs, and subsequently we take the average over the latest ten-years available. Finally,
j
? and
i
?
capture industry and country-specific fixed effects, reducing the omitted variable bias. The choice to
concentrate on bilateral trade between EU countries and China eliminates the need to control for
tariffs and other common variables.
As the pollution intensity of industries might differ between China and Europe in equation (3)
we use the Global Warming Potential of industries in Europe (GWP
j
), as alternative industry-
specific pollution variable. In Table 3 we report the correlations between industrial-pollutant
intensities; it is worth nothing that with GWP
j
and SO
2j
are highly correlated (0.87).
Finally, string
i
(string
i
= STRING
i
, PCGDP
i
, GHGE
i
) is the index of the stringency of
environmental regulation in country i. PCGDP
i
and STRING
i
should act in the same direction, that
is higher levels of the index should imply higher demand for environmental protection; on the
? +
22
RVASH should be interpreted as a comprehensive measure of relative factor endowments, possibly reflecting
both physical and human capital. For robustness check we construct also a measure of physical and human capital
intensities (k
ij
and h
ij
) given by the interaction of industry-specific intensities ( , ) with country-specific
endowments of physical and human capital ( , ), as suggested in Nunn (2007). In this case the equation becomes:
j
k
j
h
i
k
i
h
ij i j ij i j i j i j
C
ij
string GWP ULC h h B k k RCA ? ? ? ? ? + + + + + = ?
?
) * ( ) * ( ) * (
3 ) 1995 ( 2 ) 1995 ( ) 1995 ( 1 ) 1995 ( ) 1995 ( 1 ) 1999 2006 (
14
Industry-specific human and physical capital intensities are derived from US data. As results are essentially unchanged,
we do not report them; they are available from the author upon request.
contrary, GHGE
i
works in the opposite direction (higher values of GHGE indicate more polluting
emissions per unit of output and, therefore, lower environmental standards).
Table 3. Correlations between industrial pollutant
SO
2j
SO
2j
*RB SO
2j
*NRB COD
j
COD
j
*RB COD
j
*NRB
COD
j
0.21 1
COD
j
*RB 0.68 1
COD
j
*NRB 0.28 1
GWP
j
0.87 -0.04
GWP
j
*RB 0.15 -0.19
GWP
j
*NRB 0.95 0.18
The data - Trade data, classified by 2-digit NACE-ISIC rev. 3 nomenclature, are from the OCSE-
STAN Bilateral Trade database. Data on value added and labour compensations are from the
EUKLEMS database (available athttp://www.euklems.net/), these data are classified by 2-digit
NACE –ISIC rev.3 nomenclature. Data on SO
2j
and COD
j
emissions (kilos per thousand of 1995
Yuan in 2004), classified by 2-digit ISIC rev.3 nomenclature, are from Dean and Lovely (2008) and
are reported in Table 2 above and Table A1 in the Appendix. Industry-level data on “Global
warming potential” are from Moll et al (2007). The variable STRING is from Eliste and Fredricson
(2002). Greenhouse gas emissions in manufacturing and construction and value added for EU14 are
from Euro Stat. Per capita GDP at 1995 prices and 2005 purchasing power parities are from the
World Bank. The Environmental Performance Index (EPI) is fromhttp://epi.yale.edu/Home; the
Sustainable Society Index (SSI) is available athttp://www.sustainablesocietyindex.com/ssi-
2008.htm.
V. Results
As anticipated in the previous section, the expected signs of our coefficients are the following:
?
1
0) indicates convergence (divergence) of relative endowments;
?
2
The relocation of more polluting industries in poorer countries due to gaps in environmental standards is known as the pollution haven effect, whereby the scale and the composition of output change across countries.
Questioni di Economia e Finanza
(Occasional Papers)
Environmental regulation and revealed comparative advantages
in Europe: is China a pollution haven?
by Daniela Marconi
N
u
m
b
e
r
67
J
u
n
e
2
0
1
0
Questioni di Economia e Finanza
(Occasional papers)
Number 67 – June 2010
Environmental regulation and revealed comparative advantages
in Europe: is China a pollution haven?
by Daniela Marconi
The series Occasional Papers presents studies and documents on issues pertaining to the
institutional tasks of the Bank of Italy and the Eurosystem. The Occasional Papers appear alongside
the Working Papers series which are specifically aimed at providing original contributions to economic
research.
The Occasional Papers include studies conducted within the Bank of Italy, sometimes in
cooperation with the Eurosystem or other institutions. The views expressed in the studies are those of the
authors and do not involve the responsibility of the institutions to which they belong.
The series is available online at www.bancaditalia.it.
ENVIRONMENTAL REGULATION AND REVEALED COMPARATIVE
ADVANTAGES IN EUROPE: IS CHINA A POLLUTION HAVEN?
by Daniela Marconi
*
Abstract
The relocation of more polluting industries in poorer countries due to gaps in
environmental standards is known as the pollution haven effect, whereby the scale and the
composition of output change across countries. Changes in the composition of the output
mix might translate into changes of comparative advantages across countries, as revealed by
trade flows. This paper focus on this issue and looks at the changes of bilateral revealed
comparative advantages (RCAs) in the last decade between China and the major fourteen EU
countries (EU14). Using industry level data on bilateral trade, air pollution, water pollution
and several measures of environmental stringency, we find that, controlling for other factors
that may have affected RCAs, such as labor costs, on average our EU14 countries have kept
or improved their advantages with respect to China in both water polluting industries (such
as paper and agro-based industries) and air polluting industries (such as basic metals and
chemicals), while they have lost competitiveness in the more clean industries (such as
machinery and fabricated metals).
JEL Classification: F14, F18.
Keywords: revealed comparative advantages, environmental regulation, industrial pollution.
Contents
I. Introduction........................................................................................................................... 5
II. Pollution abatement costs and environmental regulation ................................................... 8
III. Pollution intensities by industry and bilateral RCA......................................................... 11
IV. The empirical strategy...................................................................................................... 13
V. Results ............................................................................................................................... 15
VI. Conclusion........................................................................................................................ 20
References .............................................................................................................................. 21
Appendix ................................................................................................................................ 23
* Bank of Italy, Economics, Research and International Relations.
5
I. Introduction
1
In recent years the need to preserve and improve environmental quality has solicited
increasing efforts to abate pollution worldwide; renewed effort in rich countries has become urgent
fearing the still unknown consequences of climate change. The debate on sustainable emission
targets and required abatement trends has become particularly intense at all levels raising
coordination problems and free-riding concerns; in fact, while the quality of the environment does
not depend only on the action taken within each country’s boundaries, the burden of abatement
costs can only be effectively imposed by governments on domestic producers and consumers.
2
As long as a “healthier natural environment” is a normal good, demand for it tends to be
higher in richer countries which impose more stringent environmental regulations as compared to
poorer ones (so called “environmental Kutznets curve” hypothesis; Copeland and Taylor, 2003).
However, pollution abatement poses additional burdens on domestic firms, especially those
operating in the most polluting industries, shifting part of the inputs away from production to
pollution abatement. If the cost burden is significant enough, it might hurt the international
competitiveness of domestic firms, compared to firms located in countries with weaker
environmental standards. The relocation of more polluting industries to poorer countries due to gaps
in environmental standards is known as the pollution haven effect, whereby the scale and the
composition of output change across countries (Copeland and Taylor, 2003).
3
The existence and the
magnitude of such an effect depends on two things: (a) whether environmental regulations impose
substantial additional costs on polluting industries, and (b) whether, absent other compensative
policies, regulation differentials are large enough to impact on industry location, output
composition and trade.
4
Changes in the scale of activity affect directly profits and jobs within a country; changes in
the composition of the output mix translate into changes of comparative advantages across
1
I would like to thank Xiaolan Fu, Herman Vollebergh and Valeria Rolli for insightful comments and suggestions on
earlier versions of this paper. I am also grateful to Pietro Barone, Ivan Faiella, Marco Marinucci for valuable
discussions.
2
The possible option of imposing additional taxes on imports of goods from polluting producers encounters major legal
and practical problems. For a discussion on coordination issues see OECD (2008).
3
Copeland and Taylor (2004) distinguish between the pollution haven effect and the pollution haven hypothesis. The
first is the effect, at the margin, on trade flows and plant location of tightening up environmental regulation in richer
countries; the hypothesis instead refers to the implications for plant location and trade flows of a change in trade
regimes between countries with different environmental regulations. In our analysis both things are at play, in that in the
last ten years environmental regulation has become more stringent in richer countries and at the same time trade barriers
have been reduced. Since our focus is on environmental regulation differentials we refer to pollution haven effect.
4
We might consider the environment as an additional factor of production, together with capital, labor and land; as
environmental services in poorer countries become relatively cheaper they will be embodied in a larger share in their
exported goods. However, there is some evidence (Eliste and Fredriksson, 2002, and Grether and de Melo, 2004) that
the most polluting industries often obtain compensating transfers from governments and tend to benefit from higher
trade barriers.
6
countries. This paper focuses on this latter issue.
5
So far very few studies have tested the pollution
haven effect on revealed comparative advantages in trade (Grether and de Melo, 2004 and Cole et
al., 2005). The empirical literature has mainly looked at the effects of environmental regulation on
plant location decisions and foreign direct investment (FDI) flows, either among different regions
within the same country (Dean et al., 2005 and Zhang and Fu, 2008, for China and, among others,
Keller and Levinson, 2002, for the US) or between countries with different levels of environmental
regulation.
6
Results, in particular for the US, point to a weak relationship between plant location
decisions, or FDI flows, and environmental regulation. Exploring the link between trade flows and
environmental regulation seems more appropriate as it allows uncovering the impact of both plant
location choices and other policy-induced changes in industrial output sizes. The literature looking
at trade flows, however, also reaches mixed conclusions, with results being very sensitive to the
choice of countries, the empirical specification and the definition of environmental regulation.
7
In our analysis our measure of international specialization is an index of trade revealed
comparative advantage (RCA). Previous studies with similar dependent variables tend to find no
clear evidence of pollution haven effects (Grether and de Melo, 2004, for a set of 52 countries and
Cole et al., 2005 for the US). We consider the major fourteen EU countries (EU14) and look at the
changes in the contribution of each industry to the bilateral trade balance with China in the last
decade. The recent surge of China as world’s top exporter is often attributed not only to its low
labour costs and rapid capital accumulation but also to its “export dumping” due to weak
environmental standards compared to richer countries. On the other hand, some EU countries are
regarded as those which have committed to the most stringent environmental regulation worldwide.
Aggregate green house gas emission intensity in the EU fell in the last decade, but
comparatively less in manufacturing and construction industries than in other sectors of the
economy; the decline has been particularly intense from 1996 to 2000; a renewed effort seem to be
in place again since 2003 (Fig. 1). At the same time, environmental protection expenditures have
remained constant in terms of GDP, around 0.4 per cent.
On the other hand, as documented in Dean and Lovely (2008), in China fast industrialization
since 1978 has led to a rapid deterioration of the air quality and of many water sources. Against this
background, in recent years there have been significant improvements in the environmental
protection legislation, although enforcement is still very weak due to diverging economic interests
5
For extensive surveys on the broader relationship between environmental regulation and international competitiveness
see SWQ (2006) and United Nations (2006).
6
A review of these studies can be found in Copeland and Taylor (2004) and Zhang and Fu (2008).
7
Grether and de Melo (2004) and Pasurka (2008) offer a summary of the various specifications and findings. Table A4
in the appendix summarize the findings of the papers mostly related to the present article.
between central and local governments (Zhang and Fu, 2008). Nonetheless, recent available data
show a certain effort in pollution abatement and treatment: in the last decade total investment in
industrial pollution treatment (PTI) has shown an increasing trend in terms of GDP, surpassing 1.4
per cent in 2008, while water and air industrial emissions (per unit of output) declined steadily (Fig.
2).
Fig. 1 EU15: Green house gas emission
intensity
(tons/thousand value added euros; 1995 prices)
Fig. 2 China: Total investment in pollution treatment
(PTI) as percentage of GDP;
Total waste water discharge (kilos/yuan output; 1995
prices) and Sulphure dioxide (SO
2
; kilos/thousand yuan
output; 1995 prices) in industry
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
Manufacturing and Construction All activities
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
Total industrial waste water SO2 PTI
Source: EuroStat data and author’s elaboration. Source: CEIC and author’s elaborations
In this paper we analyze the evolution of the structure of bilateral trade between EU countries
and China in relation to environmental regulation; in doing so, we suggest a new strategy to look at
this relationship, which allows to overcome the endogeneity problems associated with the measures
of environmental regulation usually adopted in the literature, such as pollution abatement costs
(PAC).
8
We use industry-level data on air-and-water-pollution intensities in China and a cross-
country index of environmental stringency first constructed by Dasgupta et al. (1995) and recently
extended by Eliste and Fredriksson (2002). Robustness checks are conducted by using: i) an
additional measure of pollution intensity by industry (the global warming potential (GWP) of
emissions per unit of output, in Europe); ii) two additional measures of environmental stringency
(GDP per capita and greenhouse gas emissions).
Preliminary findings on the evolution of the structural bilateral trade between China and
EU14 countries in 18 manufacturing industries in the period from 1996 to 2006, indicate that, after
controlling for other factors affecting trade flows (such as labour costs), there is no evidence of a
pollution haven effect. In particular we find that:
7
8
On the endogeneity issue, see Section II.
8
i) on average our EU14 countries have kept or improved their comparative advantages
with respect to China, as revealed by their bilateral trade, in both water-polluting
industries (such as paper and agro based industries) and air-polluting industries (such
as basic metals and chemicals);
ii) on average our EU14 countries have, instead, lost competitiveness in cleaner and
more internationally-mobile industries (such as communication equipment and office
and computing machinery), presumably in response to unfavourable unit-labour-cost
differentials and higher capital accumulation in China.
The paper unfolds as follows: Section II discusses endogeneity problems of pollution
abatement costs as proxy for environmental regulation; Section III presents briefly some statistical
evidence on pollution intensities and revealed comparative advantages in China and Europe;
Section IV presents our estimation strategy; Section V reports estimation results and Section VI
concludes.
II. Pollution abatement costs and environmental regulation
In principle, the pollution abatement effort in a given country reflects the stringency of its
environmental regulation. To evaluate and compare the effective cost burden of pollution abatement
and control (PAC) expenditures we would like to have a reliable measure of such expenditures by
industries and time. Unfortunately such information is not readily available for a sufficient number
of countries and time length, and comparison among countries for which data are available must be
taken with great caution, since definitions differ from country to country.
9
The most comprehensive
set of data on PAC expenditures and investments, sometimes also specified as pollution treatment
expenditures (PTE) or investments (PTI), are collected by the OECD (OECD Environmental
Compendium 2008) and the Euro Stat (2008). In the OECD data the breakdown within countries at
most distinguishes between public and business sector; in the Euro Stat data the breakdown refers to
broad industries: manufacturing, electricity, gas and water supply and mining and quarrying. The
time span covered varies from country to country, for some countries data date back to the eighties.
However, even in these favourable cases, time series are highly discontinuous and again hardly
comparable across countries. Moreover, as data are too aggregated across sectors, PAC
expenditures are likely to be endogenous, in that, as the output composition within a macro-
9
For a comprehensive survey of available industrial-level PAC data, and their measurement and comparability
problems, see Pasurka (2008).
9
industrial-branch changes, PAC expenditures change accordingly.
10
Therefore, such measures are
unsuitable to reflect the stringency of environmental regulation in a given country at a given time.
An alternative way to evaluate pollution abatement costs is by observing polluting emissions
per unit of output over time, by highly disaggregated industrial sectors. In fact, under the
assumption that the output composition within a finely defined industrial branch does not change
dramatically, the evolution of emission intensities would provide an indirect indication of the
stringency of environmental regulations. Unfortunately, however, there are no time series of
emissions per unit of output by industrial sectors and by country readily available. Emissions per
unit of output at 2-digit ISIC classification are available for China in 1995 and 2004 and for six
major European countries mainly for the year 2000.
11
For the EU and other advanced countries,
emissions per unit of output for the last fifteen years can be recovered only for the economy as a
whole or for macro-industrial branches, however, being endogenous to changes in output mix, are
not suitable to our analysis.
12
To overcome endogeneity and data shortage problems, in our empirical approach we use a
country-level index of environmental regulation to be interacted with an industry-level index of
pollution intensity; the methodology will be made clear in Section IV. As for country-level indexes
of environmental regulation, in our empirical analysis we use three different measures. The first
(STRING
i
) is a cross-country index, first constructed by Dasgupta (1995) and recently extended by
Eliste and Fredriksson (2002), based on detailed information about the environmental regulatory
framework on water, air pollution, land use and biodiversity; higher values of the index correspond
to higher level of environmental standards.
13
A second measure of environmental stringency is
given by the level of GDP per capita in 1995 (PCGDP
i
); higher levels of GDP per capita should be
associated with higher environmental standards. Finally, we consider an index of greenhouse gas
10
For an extensive discussion on the endogeneity problems that arise when using pollution abatement costs as proxy of
environmental stringency, see Levinson and Taylor (2008).
11
For Italy, detailed data for industrial emissions at 2-digit NACE-ISIC classification have been recently released by the
National Statistics Bureau, ISTAT (http://www.istat.it/dati/dataset/20070625_00/).
12
The EuroStat is developing a framework to collect industry-level time-series data on emissions linked to economic
accounts (EuroStat 2001). Up to now EuroStat has made available only data on total Carbon dioxide emissions from
1995 to 2004 classified by NAICS in manufacturing, mining and services for a large number of countries within EU27
(http://epp.eurostat.ec.europa.eu/portal/page?_pageid=0,1136239,0_45571444&_dad=portal&_schema=PORTAL).
However an inspection of the data reveal that: (a) there are many missing values over time and across industries; (b) the
number of manufacturing branches for which it is possible to express emissions in intensive form, using gross value
added classified by NACE, is at most 11; (c) intensities can be expressed only in terms of value added, because output
volumes are not available for the same classification; (d) emission intensities show several suspicious inconsistencies
which makes it hard to believe that those emissions can be confronted with national account data on value added. Data
and elaborations are available from the author upon request.
13
The index is based on Country Reports prepared for the 1992 United Nations Conference on Environmental and
Development.
emissions per unit of output in manufacturing and construction, averaged over the period 1995-2005
(GHGE
i
); countries with more stringent regulation should score lower values of this index
14
Figure 3 below shows the dispersion of our proxies for environmental stringency across the
European countries in our sample; Table 1 sows how the three measures are correlated with each
other.
15
Even though we concentrate on STRING
i
, PCGDP
i
and GHGE
i
in our analysis, for
completeness purposes, we also report two additional indexes, the Environmental Protection Index
(EPI
i
) and the Sustainable Society Index (SSI
i
); countries with better environmental standards
should score higher values for these indexes.
16
It is worth noting that the dispersion across EU
countries is appreciable; in general correlations between these indexes are quite high, with the
exception of those between PCGDP
i
and EPI
i
and PCGDP
i
and SSI
i
.
Fig. 3 Environmental regulation proxies in fourteen EU countries
Source: author’s elaborations on data from Eliste and Fredricson (2002), Euro Stat and World Bank.
14
This latter measure is from Euro Stat and reported in Table A3 in the Appendix.
15
The index PCGDP is constructed by taking the ratio of each country per capita GDP to the average of the sample
countries, multiplied by 100; the index GHGE is given by the ratio of the average emission intensities in the 1995-2005
period in each country to the sample average, multiplied by 100. As shown in Table 1 the index by Eliste and
Fredriksson (STRING) shows a correlation coefficient equal to 0.53 with PCGDP in 1995, such a correlation grow to
0.8 in 2005; the correlation coefficient with GHGE is equal to -0.74; the correlation coefficient between PCGDP and
GHGE is equal to -0.56.
10
16
Regression results with these two indexes are qualitatively similar to those found in the main regressions; they
available from the author upon request.
Table 1. Correlations between measures of environmental stringency
STRING
i
PCGDP
i
GHGE
i
EPI
i
SSI
i
STRING
i
1.00
PCGDP
i
0.53 1.00
GHGE
i
-0.74 -0.56 1.00
EPI
i
0.50 -0.03 -0.62 1.00
SSI
i
0.53 0.29 -0.66 0.74 1.00
III. Pollution intensities by industry and bilateral RCA
The 18 manufacturing industries considered are ranked by pollution intensity according to
sulphur dioxide (SO
2
) air emissions and chemical oxygen demand (COD) of water discharge in
China; an additional measure is provided by the global warming potential (GWP) emissions per
unit of output in Italy, which is chosen as representative for Europe (see columns 1 to 6 in Table
2).
17
Industries are further classified as either resource-based (RB) or non-resource-based (NRB).
This distinction is important, as RB industries are characterized by a very low degree of
international mobility compared to NRB ones. Consequently, they may react differently to changes
in environmental regulation. We will elaborate further on this issue in Section VI.
For each industry we also compute an index of bilateral trade-revealed comparative advantage
(RCA) with respect to China. Our measure is not straightforward and deserves some explanations.
We compute sectoral RCAs for each EU14 country as follows:
100 * *
18
1
18
1
18
1
18
1
18
1
18
1
? ? ? ?
? ?
= = = =
= =
+
+
?
?
?
?
?
?
?
?
?
?
?
?
?
?
+
?
?
+
?
=
j j
ij ij
ij ij
j j
ij ij
j j
ij ij
ij ij
ij ij
ij
M X
M X
M X
M X
M X
M X
RCA
. (1)
Where: i is the EU reporting country (i=1,...14); j is the industrial sector (j=1,...,18); ( )
are country i’s bilateral exports (imports) to (from) China in sector j; values are expressed in current
US dollars. Equation (1) measures directly the contribution of each sector to the bilateral trade
ij
X
ij
M
11
17
We concentrate on SO
2j
and COD
j
emissions because the other two emissions available for China, SMOKE and
DUST (particulate), are strongly correlated with SO
2j.
The Global Warming Potential (GWP) emission index, instead,
aggregates three greenhouse gases CO
2
, N
2
O and CH
4
(with weight 1, 310 and 21 respectively). Moll et al. (2007) have
calculated GWP per unit of output by 2-digit ISIC classification in 1995 or 2000 for seven European countries
(Denmark, Germany, Italy, Netherlands, Spain, Sweden and United Kingdom). Italian GWP emissions show the highest
correlation with any other European country, on average the correlation coefficient is equal to 0.91.
balance; such a measure does not depend on the size of the overall balance, but only on its
composition.
18
An inspection of Table 2 reveals that RCA changes over the period 1996-2006 for the average
of EU14 countries (column (7)) tend to be negatively correlated with the pollution intensity
measures, that implies that Europe’s comparative advantages with respect to China have actually
tended to worsen in the cleanest industries and to improve in the dirtiest ones.
19
Table 2. S02 and COD emissions in China (kilos per thousand yuan output, 1995 yuan) and changes of RCAs in
EU14 by industrial sector.
(1) (2) (3) (4) (5) (6) (7)
Sectors
COD
emissions
per unit of
output in
China
(2004)
SO2
emissions
per unit of
output in
China
(2004)
GWP
emissions
per unit of
output in
Italy
(2000)
COD
rank
SO2
rank
GWP
rank
EU14
Average
RCA changes
(1996-2006)
Resource-based (RB) industries
Coke and Petroleum 0.08 0.85 5.11 10 6 3 -0.01
Pulp, paper, paper products, printing and publishing 5.21 1.41 1.46 1 2 5 0.01
Food products, beverages and tobacco 1.16 0.44 0.34 2 8 8 0.16
Wood 0.92 1.15 0.33 3 4 9 -0.15
Non-resource-based (NRB) industries
Non-metallic minerals 0.14 4.26 9.76 6 1 1 -0.18
Basic metals 0.12 1.26 5.96 7 3 2 0.99
Chemicals 0.67 1.13 1.80 4 5 4 0.39
Rubber and Plastics 0.10 0.26 0.49 9 11 6 0.55
Motor vehicles 0.06 0.06 0.44 12 15 7 1.58
Textiles, textile products, leather and footwear 0.66 0.54 0.31 5 7 10 2.05
Fabricated metals 0.08 0.32 0.30 11 9 11 -0.19
Machinery 0.05 0.18 0.23 14 12 12 -0.02
Transport equipment 0.06 0.06 0.22 13 16 13 -0.42
Furniture and Other Mfg. 0.12 0.28 0.20 8 10 14 1.10
Medical, Precision and Optical Instruments 0.05 0.08 0.19 15 14 15 1.01
Electrical Machinery 0.02 0.16 0.16 18 13 16 0.62
Communications Equipment 0.03 0.03 0.16 17 18 17 -5.10
Office and Computing Machinery 0.03 0.03 0.06 16 17 18 -2.40
Source: Dean and Lovely (2008); Moll et al. (2007); EUKLEMS database; OECD-STAN bilateral trade database and author’s
elaborations.
Note: Sectors are classified according to the 2-digit ISIC rev. 3 nomenclature. RB indicates resource-based industries, NRB non-
resource-based industries based on UNIDO definitions reported in Malatu et al. (2004). RCA are based on equation (1), changes are
computed on the difference between the average value of index over the period 2001-2005 and the average value of the index in the
period 1996-2000.
18
The index varies between -50 and +50 and the sum across all the j sectors is equal to zero. A positive (negative) sign
of the index in sector j indicates that the reporting EU country has a comparative advantage (disadvantage) in that
sector, relative to all other sectors. This index allows to rank the products according to their importance, takes into
account intra-industry flows and allows for international comparisons (Marconi and Rolli, 2008). Detailed data for each
country in the sample are reported in Table A2 in the Appendix.
12
19
The evolution of bilateral trade RCAs with respect to China clearly does not describe the entire sectoral evolution of
production in EU14; indeed, if we look at the evolution of value added by industry we find a positive correlation with
the rankings by pollution intensity (0.5), meaning that in the last decade the cleanest industries have benn also the most
dynamic ones in Europe (see also Appendix, fig. A1a and A1b).
Also, it is interesting to note that, despite possible differences in pollution intensities between
EU countries and China due to different output mix or eco-efficiency, the ranking of industries by
SO
2
emissions in China (column (5)) and by GWP emissions in Europe (column (6)) are quite
similar (correlation of 0.8).
IV. The empirical strategy
The model - In order to measure the impact of environmental regulation on the evolution of the
structure of bilateral trade between EU14 countries and China, we want a dependent variable as
much as possible independent of macroeconomic effects. In much of the existing empirical
literature, the dependent variable is specified as industry’s net exports normalized by industry’s
value added, we use, instead, our index of RCA described in equation (1) departing from the
literature in two ways: (a) we take bilateral, instead of total trade; (b) we normalize net exports by
gross flows (a normalization better suited for international comparisons).
20
We estimate the following equations:
) 2 ( . *
* 2
4
3 ) 1995 ( 2 ) 1995 ( 1 ) 1999 2006 (
ij i j
i j ij ij i j ij
string COD
string SO ULC RVASH RCA
? ?
? ? ? ? ?
+ +
+ + + + = ?
?
ij i j ij ij i j ij
string GWP ULC RVASH RCA ? ? ? ? ? ? + + + + + = ?
?
*
3 ) 1995 ( 2 ) 1995 ( 1 ) 1999 2006 (
. (3)
Where our dependent variable is the change of bilateral RCA
ij
between country i
and China in sector j, in the 1999-2006 period; RVASH
ij
is the share of value added of sector j in
country i, in real terms, relative to the average sector share across EU14 in 1995.
) 1999 2006 ( ?
?
ij
RCA
21
ULC
ij
is the unit
labour cost in sector j in country i; it is measured as nominal wage over value added at 1995 prices.
The next three variables are our variables of interest, i.e., the pollution haven variables. These
variables are constructed interacting industry-specific pollution intensities (as measured by SO
2j
emissions, CODj emissions, or, alternatively in equation (3), GWPj emissions) with country-
specific indexes of environmental regulation (string
i
). The methodology of interacting an industry
characteristic (in our case pollution intensity) with a country characteristic (in our case, the
stringency of environmental regulation) was proposed by Rajan and Zingales (1998). These
interacted variables have the advantage of varying across industries and countries and should
capture the pollution haven effect. We expect that a more stringent environmental regulation should
20
See Bugamelli, 2001; OECD, 2005 and Marconi and Rolli, 2008.
13
21
The value added share of sector j in country i is normalized with respect to the EU14 average share of sector j. A
similar role would be played by trade specialization by sector at the beginning of the period. Indeed, running
regressions replacing RVASH with the index of “Export specialisation relative to OECD23 and total manufacturing
(XSPEC23M)” taken from OECD STAN, leaves results qualitatively unchanged. Results are available from the author
upon request.
induce a worsening of RCAs in more polluting industries and, conversely, an improvement in less
polluting ones.
RVASH
ij
and ULC
ij
are intended to capture, respectively, Hecksher-Ohlin-Samuelson
(henceforth H-O-S) and Ricardian determinants of RCAs. According to the H-O-S theory, which
assumes same technologies across countries, countries specialize according to their relative factor
endowments; this means that relative industrial composition, as captured by the relative shares of
value added in industry j of country i, RVASH
ij
, should reflect relative endowments in European
countries. For the Ricardian argument, instead, countries tend to specialize according to their
relative productivity, summarized by unit labor costs (ULC
ij
). As it is plausible that both
endowments and technology play a role in shaping RCAs with respect to China, we include both
variables in our regression. Negative (positive) coefficients for initial specialization differentials
should capture convergence (divergence) of endowments, while unit labor costs differentials should
capture Ricardian comparative advantages.
22
In order to correct for high volatility in trade data first we take 3-year moving averages of
RCAs, and subsequently we take the average over the latest ten-years available. Finally,
j
? and
i
?
capture industry and country-specific fixed effects, reducing the omitted variable bias. The choice to
concentrate on bilateral trade between EU countries and China eliminates the need to control for
tariffs and other common variables.
As the pollution intensity of industries might differ between China and Europe in equation (3)
we use the Global Warming Potential of industries in Europe (GWP
j
), as alternative industry-
specific pollution variable. In Table 3 we report the correlations between industrial-pollutant
intensities; it is worth nothing that with GWP
j
and SO
2j
are highly correlated (0.87).
Finally, string
i
(string
i
= STRING
i
, PCGDP
i
, GHGE
i
) is the index of the stringency of
environmental regulation in country i. PCGDP
i
and STRING
i
should act in the same direction, that
is higher levels of the index should imply higher demand for environmental protection; on the
? +
22
RVASH should be interpreted as a comprehensive measure of relative factor endowments, possibly reflecting
both physical and human capital. For robustness check we construct also a measure of physical and human capital
intensities (k
ij
and h
ij
) given by the interaction of industry-specific intensities ( , ) with country-specific
endowments of physical and human capital ( , ), as suggested in Nunn (2007). In this case the equation becomes:
j
k
j
h
i
k
i
h
ij i j ij i j i j i j
C
ij
string GWP ULC h h B k k RCA ? ? ? ? ? + + + + + = ?
?
) * ( ) * ( ) * (
3 ) 1995 ( 2 ) 1995 ( ) 1995 ( 1 ) 1995 ( ) 1995 ( 1 ) 1999 2006 (
14
Industry-specific human and physical capital intensities are derived from US data. As results are essentially unchanged,
we do not report them; they are available from the author upon request.
contrary, GHGE
i
works in the opposite direction (higher values of GHGE indicate more polluting
emissions per unit of output and, therefore, lower environmental standards).
Table 3. Correlations between industrial pollutant
SO
2j
SO
2j
*RB SO
2j
*NRB COD
j
COD
j
*RB COD
j
*NRB
COD
j
0.21 1
COD
j
*RB 0.68 1
COD
j
*NRB 0.28 1
GWP
j
0.87 -0.04
GWP
j
*RB 0.15 -0.19
GWP
j
*NRB 0.95 0.18
The data - Trade data, classified by 2-digit NACE-ISIC rev. 3 nomenclature, are from the OCSE-
STAN Bilateral Trade database. Data on value added and labour compensations are from the
EUKLEMS database (available athttp://www.euklems.net/), these data are classified by 2-digit
NACE –ISIC rev.3 nomenclature. Data on SO
2j
and COD
j
emissions (kilos per thousand of 1995
Yuan in 2004), classified by 2-digit ISIC rev.3 nomenclature, are from Dean and Lovely (2008) and
are reported in Table 2 above and Table A1 in the Appendix. Industry-level data on “Global
warming potential” are from Moll et al (2007). The variable STRING is from Eliste and Fredricson
(2002). Greenhouse gas emissions in manufacturing and construction and value added for EU14 are
from Euro Stat. Per capita GDP at 1995 prices and 2005 purchasing power parities are from the
World Bank. The Environmental Performance Index (EPI) is fromhttp://epi.yale.edu/Home; the
Sustainable Society Index (SSI) is available athttp://www.sustainablesocietyindex.com/ssi-
2008.htm.
V. Results
As anticipated in the previous section, the expected signs of our coefficients are the following:
?
1
0) indicates convergence (divergence) of relative endowments;
?
2