ENVIRONMENTAL REGULATIONS 1163

Do Environmental Regulations
Influence Trade Patterns? Testing
Old and New Trade Theories
Matthew A. Cole and Robert J. R. Elliott

1. INTRODUCTION

HE relationship between trade liberalisation and the environment has received

T a great deal of attention in recent years, amongst both academics and policy
makers. The last thirty years have been characterised by both a steady decrease in
global trade barriers and a steady increase in environmental regulation, particu-
larly in the developed world. During this time, a large literature examining differ-
ent aspects of the trade-environment relationship has developed (see e.g. Siebert
et al., 1980; Anderson and Blackhurst, 1992; Chichilnisky, 1994; Copeland and
Taylor, 1994 and 1995; Antweiler et al., 2001; and Cole and Elliott 2003). One
particular focus of attention has been on the possible influence of environmental
regulations on global trade patterns.
It has been claimed, for example, that trade between two countries with dif-
ferent levels of environmental regulations will lead to the low regulation country
specialising in pollution-intensive production (Baumol and Oates, 1988). In the
developed world the cost of complying with environmental regulations appears
to be steadily increasing over time and, for the USA alone, was estimated to be
$184 billion in 2000, equivalent to 2.6 per cent of US GNP.1 Since the string-
ency of environmental regulations increases with income (Dasgupta et al., 1995),
this line of reasoning suggests that developing countries possess a comparative

MATTHEW COLE is from the Department of Economics, University of Birmingham. ROBERT

ELLIOTT is from the School of Economic Studies, University of Manchester. The authors would
like to thank Marius Brulhart, Nick Horsewood, Toby Kendall, John Bates and two anonymous
referees for their helpful comments and suggestions on an earlier draft, but retain responsibility for
all remaining errors.
1
US Environmental Protection Agency (1990) estimated in 1992 US dollars. This is an estimate of
private sector compliance costs and therefore omits personal consumption abatement, government
abatement and government regulation and monitoring.

© Blackwell Publishing Ltd 2003, 9600 Garsington Road, Oxford, OX4 2DQ, UK
and 350 Main Street, Malden, MA 02148, USA 1163
© Blackwell Publishing Ltd 2003
1164 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

advantage in pollution-intensive production. If so, then we may see dirty industries

relocating from the North to the South (foreign direct investment), or simply
dirty industries from the developed world becoming displaced from the world
market by similar industries in developing countries. This phenomenon, known
commonly as the pollution haven hypothesis, has been cited as one explanation
for the inverted-U relationship often estimated between per capita income and
emissions of local air pollution (e.g. Grossman and Krueger, 1995; and Cole
et al., 1997). Theoretical models of pollution havens include Pethig (1976),
McGuire (1982) and Baumol and Oates (1988) who conclude that those countries
that do not control pollution emissions, whilst others do, will ‘voluntarily become
the repository of the world’s dirtiest industries’ (Baumol and Oates, 1988, p. 265).
A number of authors have empirically tested whether environmental regula-
tions affect trade patterns, although results have been inconclusive. Lucas et al.
(1992) and Birdsall and Wheeler (1992) find that the growth in pollution intensity
in developing countries was highest in periods when OECD environmental regu-
lations were strengthened. Mani and Wheeler (1998) examine the import-export
ratio for dirty industries and find evidence consistent with the pollution haven
hypothesis, although they claim that such havens appear to have been temporary.
Similarly, Antweiler et al. (2001) examine the impact of trade liberalisation on
city-level sulphur dioxide concentrations and also claim to find some evidence
of pollution haven pressures. Van Beers and Van den Bergh (1997) find some
evidence to suggest that regulations are influencing trade patterns, although Harris
et al. (2002) claim that no such influence is found if fixed effects are included in
the model. In a notable change of direction, recent papers by Levinson and
Taylor (2001) and Ederington and Minier (2001) claim that environmental regu-
lations should be treated as a secondary trade barrier, i.e. a means of protecting
domestic industry. If this is the case, then the stringency of regulations may be
a function of trade as well as trade being a function of regulations. When treated
as an endogenous variable, both Levinson and Taylor (2001) and Ederington and
Minier (2001) find that US environmental regulations do influence US trade
patterns.
In contrast, Tobey (1990) and Janicke et al. (1997) find no evidence to sug-
gest that the stringency of a country’s environmental regulations is a determinant
of its net exports of dirty products. Similarly, Xu and Song (2000) find that
environmental regulations do not appear to influence trade in embodied environ-
mental factor services. The OECD (1997), in a review of the literature on FDI
and the environment, state that ‘fears of a “race to the bottom” in environmental
standards, based on the idea of “pollution havens”, may be generally unfounded’
(OECD, 1997, p. 13). Also in a review of the literature, Jaffe et al. (1995)
conclude that there is little evidence to suggest that stringent environmental
regulations have a significant effect on industrial competitiveness in developed
countries. Finally, in an overview of the recent empirical literature, Ferrantino

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1165

and Linkins (1999) conclude that the effects of trade liberalisation on the level of
global pollution are ambiguous.
One approach to examining the impact of environmental regulations on trade
patterns is via the standard Heckscher-Ohlin-Samuelson (HOS) framework where
comparative advantage is determined by factor endowment differentials. In this
approach, net exports are expressed as a function of factor endowments, includ-
ing environmental regulations. An often-cited study by Tobey (1990) uses this
methodology. However, the empirical observation that much of the post-war
expansion of trade was between countries of similar size and relative factor
endowments has raised questions regarding the HOS framework’s ability to ex-
plain actual trade patterns. A preliminary investigation of the trade patterns of
‘dirty’ industries also reveals a significant level of two-way trade in products
from the same product grouping, commonly known as intra-industry trade (IIT).
The existence of IIT led to the development of ‘new’ trade theories that were able
to explain the co-existence of inter- and intra-industry trade (see e.g. Lancaster,
1980; Dixit and Norman, 1980; Krugman, 1980 and 1981; Helpman, 1981; Falvey,
1981; and Helpman and Krugman, 1985). These models usually rely on differen-
tiated products and an element of imperfect competition with increasing returns
to scale.
The aim of this paper is to examine the impact of environmental regulations on
trade patterns within the traditional comparative advantage based model and
within the ‘new’ trade-theoretic framework. In the former we will test whether
the stringency of a country’s environmental regulations influences its net exports
of pollution-intensive output. In the ‘new’ trade model we are asking a slightly
different question. Since this approach is concerned with bilateral trade and the
share of intra- and inter-industry trade within total trade, we are testing whether
environmental regulations, like other factor endowments, influence the composi-
tion of trade, i.e. the extent to which countries trade within the same, or different,
industries.2
With regard to the HOS framework, we extend Tobey’s (1990) analysis in
a number of ways: (i) we use a larger and more up to date dataset that allows
us to assess whether the impact of regulations on trade patterns has changed
since the mid-1970s; (ii) we test two alternative measures of environmental
regulations; (iii) where possible, we include industry dummies to control for
unobserved industry characteristics that may affect the relationship between regu-
lations and net exports; (iv) we control for the potential endogeneity of environ-
mental regulations.
Turning to the ‘new’ trade model, we are unaware of any previous study that
tests the effect of environmental regulations within a framework of this type.

2
Environmental regulations may be interpreted as a measure of a country’s ‘environmental’
endowment.

© Blackwell Publishing Ltd 2003

1166 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

More specifically, we include environmental regulation differentials alongside

other factor endowment differentials as a possible explanation of the share of
inter-industry trade within total trade, with determinants of the share of intra-
industry trade also included. We, again, control for possible endogeneity thereby
providing the first cross-country trade analysis to incorporate the possible
endogeneity of environmental regulations.3
The remainder of the paper is organised as follows: Section 2 provides the
econometric analysis based on a model of comparative advantage, Section 3
estimates the ‘new’ trade model and Section 4 provides an interpretation of the
results. Section 5 summarises and concludes.

2. THE HECKSCHER-OHLIN-VANEK (HOV) APPROACH

In this section we provide a detailed cross-sectional analysis of the role-played

by factor endowments and environmental regulations in determining trade patterns.
The Heckscher-Ohlin-Samuelson (HOS) framework originates from the notion
that different commodities use factors in different proportions and that countries
are endowed with factors of production in different proportions. The Heckscher-
Ohlin-Vanek (HOV) model is the ‘factor content’ version of the HOS model and
allows us to consider the N-good S-factor case (N > 2 and S > 2), since it avoids the
problem of defining factor intensities in the presence of more than two factors.4
To empirically investigate the impact of environmental regulations on trade
flows within the HOV model, we estimate equation (1) which expresses a coun-
try’s net exports as a function of its factor endowments. This equation is derived
from the outline of the HOV model provided in Appendix A.
S
Wij = ∑ bikVkj i = 1, . . . , N j = 1, . . . , T (1)
k =1

where Wij are net exports from sector i by country j, Vkj are endowments of
resource k in country j, and bik are the coefficients to be estimated. Equation (1)
is estimated using 14 factor endowments together with two alternative measures
of the stringency of environmental regulations. The data cover 60 developed
and developing countries for 1995. The dependent variable is each country’s net

3
The previous studies to have incorporated endogeneity (Levinson and Taylor, 2001; and Ederington
and Minier, 2001) focus purely on US trade.
4
Previous empirical tests of the HOV include Leamer (1980) and Bowen et al. (1987). Trefler
(1993a and 1995) emphasises the HOV’s reliance on the factor price equalisation theorem and
internationally identical technologies. He includes a variable of productivity differences that is
found to significantly improve empirical results.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1167

exports in one of four dirty sectors. The sectors are Iron and Steel, Chemicals,
Pulp and Paper, and Non-Ferrous Metals. The explanatory variables, which
cover a wide range of factor endowments, are as follows: the capital stock, three
measures of labour endowment (professional and technical workers, literate
non-professional workers and illiterate workers), two measures of environmental
regulations (discussed below), mineral endowments (lead, zinc, iron and copper),
oil, gas and coal endowments, tropical forest area, non-tropical forest area and
area of cropland. Appendix B defines these variables, provides details on the data
sources and lists the countries included in our sample.
We include two measures of the stringency of environmental regulations,
ENVREG and ENVPOL. The former is provided by Eliste and Fredriksson (2001)
who built on the work of Dasgupta et al. (1995). Dagupta et al. gathered informa-
tion from individual country reports compiled under United Nations Conference
on Environment and Development (UNCED) guidelines. Each report is based
on identical survey questions and provides detailed information on the state of
environmental policies, legislation and enforcement within each country. Using
this information, Dasgupta et al. (1995) developed an index of the stringency of
environmental regulations for 31 countries. Eliste and Fredriksson (2001) then
used the same methodology to extend the index to 60 countries. ENVPOL is a
proxy for the stringency of environmental regulations based on each country’s
change in energy intensity (energy use/GDP) over the period 1980–95, together
with the level of energy intensity in 1980. Van Beers and Van den Bergh (1997)
use a similar measure of environmental regulations. For the 60 countries in our
sample, ENVREG and ENVPOL have a correlation coefficient of 0.77. Appendix B
provides more information on the calculation of ENVPOL.
Tobey (1990) incorporates a measure of environmental regulations in an HOV
estimation from the mid-1970s and uses a sample of 23 countries (and 12 degrees
of freedom). He does not find a statistically significant relationship between
environmental regulations and net exports, although given the number of degrees
of freedom this is not entirely surprising. We extend Tobey’s estimations in a
number of ways: (i) we have 60 countries in our sample rather than 23; (ii) we
use data for 1995 rather than the mid-1970s allowing us to test the possibility that
the increased stringency of environmental regulations during the intervening
period will have changed the relationship between regulations and net exports;
(iii) we test two alternative measures of environmental regulations; (iv) whilst
we undertake industry-specific estimations, we also pool all dirty industries and
include industry dummies. This allows us to control for unobserved industry
characteristics that may affect the relationship between regulations and net ex-
ports; (v) related to point (iv) is the issue of endogeneity. If environmental
regulations are themselves a function of trade flows, rather than the other way
around as has been assumed, then the estimated results will be spurious. We
therefore estimate the impact of regulations on trade flows assuming firstly that

© Blackwell Publishing Ltd 2003

1168 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

TABLE 1
HOV Estimation Results (Dependent variable: net exports in 1995 US$)

Non-ferrous Paper and Iron and

Variable Panela Metals Pulp Steel Chemicals

LAB1 −10.6 73.4 −119.8 −109.8 113.6

LAB2 −21.1** −23.8*** −23.0** −20.4* −17.4***
LAB3 40.07** 21.0 55.8** 57.1** 26.2***
CAPITAL 1.5* −3.0*** 0.45 7.4*** 1.27***
ENVREG 34.4 −8.7 102.3 38.3 57.7
LEAD 4.1 −17.8** 37.2*** −4.1 1.5
ZINC −2.5 5.8** −17.4*** 4.7 −0.017
IRON 0.52 1.2** 0.55 0.87 −0.52
COPPER 0.16 0.57*** 0.10 −0.05 0.038
OIL 3.5 9.02*** −2.3 6.6 0.70
GAS −20.9** −16.9*** −37.8*** −31.0*** 2.2
COAL 10.2 11.7 7.5 12.0 9.5*
TROPFOR 8.5 −6.79 22.09* 8.28 10.5**
NONTROP 28.1*** 6.1 97.8*** 8.65 −0.22
CROPLAND −36.7** −31.8*** −30.3 −54.2** −30.5***
R2 0.801 0.893 0.852 0.782 0.643
n 240 60 60 60 60

Notes:
For reasons of space, t-statistics have not been reported. Instead, *, ** and *** denote significance at 90, 95 and
99 per cent, respectively.
a
Where ‘panel’ refers to the inclusion of all four sectors in the same regression. This estimation includes
industry dummies, but for reasons of space these are not reported.

such regulations are exogenous, but then allow for the fact that they may be
endogenous.
Table 1 provides the results estimated individually for each sector, together
with those from a ‘panel’ estimation in which all four sectors are included together.5
These estimations stem from equation (1) and hence environmental regulations
are here taken to be exogenous. Note that in all estimations, environmental regu-
lations are not significantly correlated with net exports from dirty sectors. When
we replaced ENVREG with ENVPOL in equation (1) the results were almost
identical with ENVPOL remaining non-significant across estimations albeit with
varying signs. The evidence in Table 1 seems to confirm Tobey’s (1990) findings.
Turning to the other results, many variables are statistically significant and the
R2s are suggestive of a generally good fit to the model. For example, for two
sectors, iron and steel and chemicals, we find capital stock to be positively and
significantly related to net exports of dirty products. In addition, we find zinc, iron
and copper endowments to be highly correlated with net exports of non-ferrous
metals; forests (particularly non-tropical) to be positively and significantly related

5
In all cases a Breusch-Pagan test did not reject the null of homoscedastic variances.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1169

to net exports of paper and pulp; and that countries with large endowments of
fertile land (cropland) do not tend to specialise in these four heavy industrial
sectors. The dependence of these sectors on capital and natural resource endow-
ments may explain why empirical evidence for any pollution haven effect to date
is generally weak.
We do, however, find one or two slightly puzzling results. For instance, the
capital stock is estimated as being negatively (and significantly) related to net
exports of non-ferrous metals. Turning to the labour endowments, we find LAB1
(professional and technical workers) to be a non-significant determinant of net
exports. In contrast, LAB2 (literate non-professional workers) is negatively and
significantly related to net exports, with LAB3 (illiterate workers) being posi-
tively related to net exports. The LAB3 finding at least would seem to suggest
that these are low skill sectors, yet we know that this is not entirely true. Finally,
we find gas extraction to be negatively, and highly significantly, related to net
exports in four out of the five estimations. Again, this result is difficult to explain.
However, it is questionable whether environmental regulations should be con-
sidered to be exogenous, as has been the case so far. If trade considerations play
a role in the setting of environmental regulations, as is assumed by second-best
trade models (see e.g. Trefler, 1993b), then regulations should clearly be treated
as endogenous. It is feasible, for instance, that if net exports were declining in
pollution-intensive industries, the reaction of government may be to reduce the
stringency of regulations to boost the competitiveness of these industries. Such a
positive relationship could therefore offset any negative impact of regulations on
net exports, and therefore must be controlled for. In such a situation it is neces-
sary to estimate simultaneous equations whereby the impact of regulations on net
exports is estimated in a manner that controls for simultaneity between these two
variables.
In addition to equation (1), which expresses net exports as a function of factor
endowments, including environmental regulations, it is necessary to introduce a
second equation that identifies the determinants of environmental regulations. We
believe the key determinant of the stringency of a nation’s environmental regula-
tions is per capita income and therefore include this, along with net exports,
in equation (2). Since we are estimating the relationship between net exports in
a single dirty industry against national environmental regulations, it could be
argued that endogeneity is unlikely to be found. Nevertheless, it is still possible
that simultaneity exists, particularly when we combine four industries into a
single panel.

ENVREGj = α + β1Yj + β2Wij + ei (2)

where Wij refers to net exports in dirty industry i, country j and Yj denotes per
capita income in country j. Equations (1) and (2) are estimated simultaneously

© Blackwell Publishing Ltd 2003

1170 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

using two-stage least squares, with ENVREGj and Wij treated as endogenous
variables. All other variables are treated as exogenous, instrumental variables.
Our results are provided in Appendix C. The sign and significance of the
determinants of net exports can be seen to be almost identical to those provided
in Table 1, in which regulations were treated as exogenous. The ENVREG
variable is not a statistically significant determinant of net exports in any of the
estimations. Furthermore, in only one instance (iron and steel) are net exports a
significant determinant of environmental regulations.
In sum, whether environmental regulations are treated as exogenous or endo-
genous and whether they are measured as ENVREG or ENVPOL, they are not
statistically significant determinants of dirty net exports, within an HOV frame-
work. What these HOV results do suggest, however, is that iron and steel and
chemicals are both highly capital intensive, whilst non-ferrous metals and paper
and pulp are both natural resource intensive. The HOV model does, however,
appear to explain trade patterns with some, if not total, success.

3. THE IMPERFECT COMPETITION APPROACH

A shortcoming of the HOV model, as defined, is that it is unable to explain

trade between two countries within the same industry, that is, it cannot explain
the phenomenon of intra-industry trade. However, an empirical feature of inter-
national trade is the co-existence of inter- and intra-industry trade. Appendix A
provides an overview of a model of monopolistic competition with differentiated
products (see Helpman, 1987). Within this model, inter-industry trade will be
motivated by relative factor abundance (and perhaps environmental regulations),
whilst intra-industry trade will be motivated by the exchange of varieties of dif-
ferentiated products.
The Grubel and Lloyd (GL) index measures the share of trade that is intra-
industry in nature and was first presented in Grubel and Lloyd (1975). The GL
index provides a common measure of IIT between countries j and k, over all
industries.6

2 Σ i min( Xijk , Xikj )

IITjk = (3)
Σ i ( Xijk + Xikj )

where Xijk are exports of industry i from country j to country k. Following Hummels
and Levinsohn (1995), equation (3) can be thought of in the following way:

6
IIT within a specific industry can also be measured by equation (3) if the ‘aggregation over all
industries’ (Σi) terms are removed.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1171

INTRA
IITjk = . (4)
INTRA + INTER

It can therefore be seen that, controlling for the size of the countries, if both
countries have identical capital-labour ratios, no trade will be motivated by rela-
tive factor endowments and hence inter-industry trade (INTER) will be zero and
the share of trade that is intra-industry (IITjk) will equal 1. Conversely, if there
are differences between capital-labour ratios then INTER will increase and INTRA
will decrease. By allowing relative factor endowments to affect trade patterns,
this approach still draws heavily on the HOV model, and therefore allows us to
test for factor endowment (and environmental regulation) effects. In any trade-
pair, the greater the difference between two countries’ capital-labour ratios, the
greater will be the share of inter-industry trade and the lesser will be the share of
intra-industry trade. Similarly, the greater the difference in environmental regula-
tions between two countries, the greater will be their share of inter-industry trade
and, again, the lesser will be their share of intra-industry trade.
Drawing on Helpman (1987) and Hummels and Levinsohn (1995) we would
like to estimate equation (5):

K j Kk T j Tk
IITjki = β 0 + β1 ln − + β 2 ln − k + β3 ln|ENVj − ENV k |
Lj Lk Lj L
+ β4 ln|PcY j − PcY k | + β5 min(ln GDP j, ln GDP k )
+ β6 max(ln GDP j, ln GDP k ) + β 7 BORDER + εjk (5)

where IITjki is the GL index measuring the share of IIT between countries j and
k in dirty sector i. K denotes the country’s capital stock, L the labour force, and
T the endowment of fertile land. ENV represents the stringency of environmental
regulations (we test both ENVREG and ENVPOL), whilst PcY denotes per capita
income. MinGDP and MaxGDP are included to control for relative size effects
and BORDER is a common border dummy. Note that in our North-South estima-
tions the common border dummy is replaced by a dummy for trade-pairs with
colonial links, given the fact that very few developed countries share borders
with developing countries.
Our dependent variable, however, is bounded between 0 and 1. This means
that a linear or log linear estimation of equation (5) may generate predicted
values for IITjki that are outside the range 0 to 1. A logistic function does not
have this particular problem but its logit transformation is unable to cope with
exact values of 0 or 1. In this study we do not have any observations where
IITjki = 1, but we do have a significant number where IITjki = 0. Therefore, follow-
ing Balassa and Bauwens (1987) we use non-linear least squares of the logistic
function:

© Blackwell Publishing Ltd 2003

1172 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

1
IITjki = + ε jk (6)
1 + exp(− β ′ x jk )

where β is the vector of regression coefficients, x is the vector of explanatory

variables (as defined in equation (5)) and εjk is the random disturbance term. Pro-
vided that the disturbances of the regression are normally distributed, non-linear
least squares are maximum likelihood estimators and are therefore consistent and
asymptotically efficient.
Referring back to the explanatory variables identified in equation (5), expected
signs are β1 < 0, β2 < 0, β3 < 0, β4 < 0, reflecting the fact that the smaller the
difference in capital, land, environmental regulations and per capita income
between countries, the greater will be the share of intra-industry trade between
those countries. It is also predicted that two countries will have a higher share of
IIT the closer their levels of GDP. Thus, the greater the minimum level of GDP
and the smaller the maximum level of GDP, within a trade-pair, the greater will
be the IIT share. Thus we expect β5 > 0 and β6 < 0. Finally, having a common-
border is expected to increase the IIT share and hence we expect β7 > 0. As with
our HOV model, estimations are made for individual dirty sectors and for a
‘panel’ of all four dirty sectors. In the latter estimation, industry-specific dum-
mies are included to control for industry-specific effects. We also initially assume
environmental regulations to be exogenous and then allow for possible endogeneity.
Other factor endowment differentials were also initially included in equation (6)
(e.g. minerals, forest cover) but, in contrast to the HOV model, were not found to
be robust across all specifications. We therefore focus on the variables listed in
equation (5).
Note that per capita income differentials (PcYdiff )7 are included to capture
demand-side influences on IIT, namely the effect of preferences. In line with the
assumption of identical homothetic preferences from Appendix A, our initial runs
of equation (6) did not include PcYdiff. However, our results suggested that the
estimated coefficient on the capital-labour differential was picking up the effects
of per capita income (see discussion of results below). Following Linder (1961)
and Bergstrand (1990), we therefore allow for the possibility that a divergence of
per capita incomes will represent a divergence of tastes, thereby reducing the
share of trade that is intra-industry in nature. Since PcYdiff, in principle, will
capture both demand- and supply-side influences we have controlled for the
latter by including capital-labour differentials. In contrast, Helpman (1981 and
1987) assumes that tastes are homothetic, and includes per capita income differ-
entials simply as a proxy for factor endowment differentials. However, it appears
likely that Helpman’s estimated coefficients on PcYdiff also capture demand-side

7
For simplicity, we now denote differences in capital/labour ratios, land per head, environmental
regulations and per capita income as K/Ldiff, T/Ldiff, ENVREGdiff and PcYdiff, respectively.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1173

influences. In order to explore these relationships, we estimate equation (6) with

both capital-labour differentials and per capita income differentials, and then
omit each of these variables individually. All results are discussed below.
Equation (6) is estimated for two samples of trade-pairs, for 1995. The first
sample contains 630 trade-pairs and includes both developed and developing eco-
nomies. However, to examine whether the impact of environmental regulations
on trade flows is stronger between North-South countries than between the coun-
tries within the full sample, the second sample contains only North-South trade-
pairs (i.e. trade between developed and developing countries), with 406 trade-pairs
considered in total. Although the assumption of product differentiation and ident-
ical homothetic preferences within many of the trade-pairs, particularly in the
North-South sample, may be questionable, the North-South IIT indices are not
as low as may have been expected.8 We therefore believe that the estimation
of cross-section variations in IIT is appropriate within both the full sample and
the North-South sample.9
We believe this model structure provides a more appropriate framework to
estimate trade flows, since it allows us to separate the potential determinants of
intra-industry trade (country size and preferences) from the potential determin-
ants of inter-industry trade (factor endowments and environmental regulations).
However, a drawback is that by identifying only the shares of intra- and inter-
industry trade, this approach does not allow us to identify the direction of any
change in inter-industry trade (net trade). Thus, we can identify whether the
difference between two countries’ environmental regulations increases the share
of net trade, but we cannot say whether this represents an increase in the share
of net exports or net imports. This issue is returned to in Section 4. We believe
nevertheless, that this approach is a useful way of assessing whether environ-
mental regulations, like other factor endowments, influence trade patterns and
specifically the proportions of total trade that are intra- and inter-industry in
nature. Estimation results are provided in Tables 2 and 3.
The primary concern of this study is the role played by environmental regula-
tions in determining trade patterns. In both parts of Table 2 we find our environ-
mental regulation differential variable to be a negative determinant of the share
of IIT across all sectors. This indicates that the greater the differences in environ-
mental regulations between two countries, the smaller will be their share of intra-
industry trade within total trade and the greater will be their share of inter-industry
trade in these pollution-intensive sectors. It is notable, however, that in the North-
South sample, ENVREG is not statistically significant as a determinant of IIT, in
contrast to the full sample. We also estimate the above regressions using ENVPOL,

8
IIT data information: Full panel sample; mean IIT = 0.21, per cent of zeros = 26 per cent, n = 2520.
North-South panel sample; mean IIT = 0.14, per cent of zeros = 33 per cent, n = 1620.
9
We also estimated a North-North sample with results almost identical to those from the full sample.

© Blackwell Publishing Ltd 2003

1174 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

TABLE 2
Estimation Results

Panel Non-ferr. Paper and Iron and

Variable (Basic) Panel Metals Pulp Steel Chemicals

Panel A: IIT Estimation Results Using the Full Sample (1995)

K/Ldiff 0.095*** 0.68*** 0.78*** 0.61*** 0.74*** 0.62***
T/Ldiff 0.0906*** 0.023 −0.037 0.066 −0.0023 0.048
PcYdiff – −0.42*** −0.45*** −0.40*** −0.44*** −0.41***
ENVREGdiff – −0.10*** −0.11*** −0.10*** −0.086*** −0.11***
MinGDP 0.14*** 0.22*** 0.23*** 0.22*** 0.22*** 0.22***
MaxGDP −0.14*** −0.032 −0.048 −0.027 −0.041 −0.015
BORDER – 1.67*** 1.79*** 1.60*** 1.97*** 1.40***
R2 0.40 0.52 0.48 0.49 0.50 0.65
n 2520 2520 630 630 630 630

Panel B: IIT Estimation Results Using the North-South Sample (1995)

K/Ldiff 0.28*** 0.81*** 0.92*** 0.50*** 1.16*** 0.78***
T/Ldiff 0.10*** 0.101*** 0.0028 0.13** 0.15** 0.11**
PcYdiff – −0.46*** −0.51*** −0.33*** −0.58*** −0.49***
ENVREGdiff – −0.023 −0.022 −0.0045 −0.058 −0.036
MinGDP 0.12*** 0.22*** 0.18** 0.18** 0.28*** 0.27***
MaxGDP −0.12*** −0.0021 −0.012 −0.025 −0.024 −0.023
COLONY – 0.46** 0.30* 0.19 1.02*** 0.66**
R2 0.32 0.35 0.28 0.34 0.32 0.50
n 1624 1624 406 406 406 406

Notes:
Where *, ** and *** indicate significance at 90, 95 and 99 per cent respectively.
The panel estimations include industry dummies, but for reasons of space these are not reported.

TABLE 3
Estimated Coefficients for ENVPOL

Non-ferr. Paper and Iron and

Sample Variable Panel Metals Pulp Steel Chemicals

Full ENVPOL −0.064*** −0.059* −0.064* −0.073** −0.055**

North-South ENVPOL −0.420 −0.057 −0.140 −0.154* −0.300

Notes:
Where *, ** and *** indicate significance at 90, 95 and 99 per cent respectively.

our alternative measure of environmental regulations which stems from changes

in energy intensity. The estimated coefficients for the variables are almost ident-
ical to those in Table 2 and hence Table 3 simply reports the estimated coeffi-
cients for ENVPOL. Again in all cases ENVPOL is negative although it is less
significant than ENVREG, particularly in the North-South sample. The generally
lower significance of ENVPOL may reflect the fact that it is only a proxy for the
stringency of environmental regulations.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1175

The panel (basic) estimation from Table 2 checks for consistency with the
results of previous studies (e.g. Hummels and Levinsohn, 1995) by examining
whether basic factor differentials, together with GDP size, are significant deter-
minants of the share of IIT. Our variables are all highly significant and have signs
as predicted by theory, with the exception of the capital-labour and, for some
estimations, the land-labour differentials (K/Ldiff and T/Ldiff ) which we estimate
to be positive determinants of IIT. For K/Ldiff, in particular, this finding is robust
across all of our estimations, both panel- and sector-specific. Our results therefore
suggest that the greater the capital-labour (and perhaps land-labour) differentials
between two countries, the lower their inter-industry trade share and hence the
greater their intra-industry trade share. This clearly does not support the theory
developed in Appendix A. Hummels and Levinsohn also test capital-labour and
land-labour differentials as determinants of IIT, for a smaller sample of 91 OECD
trade-pairs for individual years covering the period 1962–1983. They find the
land-labour differential to be negative throughout, although, interestingly, whilst
K/Ldiff is negative and significant for the early years in their sample, for the later
years it becomes positive and significant. Furthermore, Greenaway et al. (1999),
in a study of UK IIT with the EU, also find capital-labour differentials to be a
positive, significant determinant of IIT. Our result is therefore consistent with the
notion that capital-labour (and perhaps land-labour) differences are no longer
positive determinants of net trade.10
Whilst K/Ldiff is a positive determinant of IIT, we find PcY to be a statistically
significant negative determinant, suggesting that our separation of demand and
supply influences is appropriate. Note that we also estimate equation (6) in the
absence of PcY and, alternatively, in the absence of K/Ldiff, although for reasons of
space we have not reported these results. In these estimations, there is strong evi-
dence that the included variable is picking up the effects of the omitted variable.
For instance, when we include K/Ldiff but omit PcY, the coefficients on K/Ldiff
and ENVREG are smaller than those in Table 2, suggesting that they are partially
capturing the (negative) effects of per capita income. Since both K/Ldiff and
ENVREG are highly correlated with PcY this is not surprising. The inclusion of
both K/Ldiff and PcY therefore appears appropriate in order to capture both the
demand and supply influences on IIT, a conclusion also reached by Bergstrand

10
Hummels and Levinsohn (1995) undertake a comprehensive sensitivity analysis to investigate
how robust their results are to alternative specifications. One point they raise is that the relationship
between IIT and K/Ldiff may be non-linear so they include a quadratic term for K/Ldiff. They find
the OLS coefficient on the linear (quadratic) term to be negative (positive). Fixed effects estimates,
however, show that both terms are statistically insignificant. The removal of the quadratic term
(again in a fixed effects framework) leads to a positive, significant coefficient on the linear K/Ldiff
term. Hummels and Levinsohn offer a number of explanations for this positive result including a
lack of time series variation in K/Ldiff, the problem of categorical aggregation and the role of
geography (via cross-border trade). However, they still end up with a series of ‘inconclusions’.
See Hummels and Levinsohn (1995) for further details.

© Blackwell Publishing Ltd 2003

1176 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

TABLE 4A
Simultaneous Equations Results Using the Full Sample (1995)

Non-ferr. Paper and Iron and

Variable Panel Metals Pulp Steel Chemicals

Panel A: Dependent Variable: IIT (GL index)

K/Ldiff 0.77*** 0.83*** 0.65*** 0.76*** 0.69***
T/Ldiff −0.0093 −0.085* 0.051 −0.020 0.033
PcYdiff −0.30*** −0.28*** −0.27*** −0.36*** −0.29***
ENVREGdiff −0.17*** −0.23*** −0.17*** −0.11*** −0.16***
MinGDP 0.22*** 0.21*** 0.19*** 0.21*** 0.20***
MaxGDP −0.053** −0.072 −0.034 −0.047 −0.029
BORDER 1.66*** 1.81*** 1.52*** 1.88*** 1.45***
R2 0.53 0.52 0.49 0.50 0.66

Panel B: Dependent Variable: ENVREGdiff

PcYdiff 0.20*** 0.20*** 0.20*** 0.20*** 0.20***
IIT −41.6*** −43.7*** −42.5*** −38.6*** −38.8***
R2 0.47 0.47 0.47 0.47 0.47
n 2520 2520 630 630 630

(1990). Furthermore, including PcY also reduces the possibility of ENVREG

picking up demand effects.
Turning to the other results in Table 2, in line with Helpman (1987) and
Hummels and Levinsohn (1995) we find GDP differentials (which they call ‘size’)
to be a negative and partially significant determinant of the share of IIT. In addi-
tion, we find that two countries that share borders will typically have a greater
share of IIT than two countries that do not. Finally, in the North-South sample we
also find that two countries with colonial links (e.g. UK and India) will generally
have a higher share of IIT than two countries without such historical links.
As discussed previously, however, it may not be appropriate to treat environ-
mental regulations and trade flows (IIT) as exogenous variables. In common with
the HOV section, we therefore again estimate a simultaneous equations model
with our first equation estimating IIT as a function of factor endowment differen-
tials (equation (6)) and our second equation estimating environmental regulations
as a function of per capita income differentials and IIT. Tables 4A and 4B report
the results for the full sample and the North-South samples, respectively.
The results in Tables 4A and 4B are fully supportive of those in Table 2, with
virtually all variables statistically significant, again, with the exception of land
differentials (T/Ldiff ). We now estimate the environmental regulation variables
play an even greater role in determining IIT. All ENVREG coefficients are larger
than those estimated in Table 2 and it is notable that these coefficients are now
statistically significant for the North-South sample, whereas they were not when
treated as exogenous variables. We also find the share of IIT to be a negative,
statistically significant determinant of environmental regulations in virtually all

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1177

TABLE 4B
Simultaneous Equations Results Using the North-South Sample (1995)

Non-ferr. Paper and Iron and

Variable Panel Metals Pulp Steel Chemicals

Panel A: Dependent Variable: IIT (GL index)

K/Ldiff 1.12*** 1.25*** 0.74*** 1.53*** 1.10***
T/Ldiff 0.010 −0.082 −0.029 −0.0015 0.17***
PcYdiff −0.44*** −0.48*** −0.29** −0.57*** −0.49***
ENVREGdiff −0.18*** −0.18** −0.16** −0.21** −0.18***
MinGDP 0.19*** 0.22** 0.15** 0.26*** 0.19***
MaxGDP 0.047 0.074 0.014 0.070 0.022
COLONY 0.45** 0.29* 0.20 1.01*** 0.64**
R2 0.42 0.33 0.35 0.33 0.51

Panel B: Dependent Variable: ENVREGdiff

PcYdiff 0.19*** 0.20*** 0.19*** 0.19*** 0.18***
IIT −34.7*** −54.9** −60.4** −28.34 7.4
R2 0.50 0.50 0.50 0.50 0.50
n 1624 1624 406 406 406

Notes for Tables 4A and 4B:

Estimated using 2SLS. Where *, ** and *** indicate significance at 90, 95 and 99 per cent respectively. Note
that almost identical results were estimated when ENVREG was replaced with ENVPOL. These latter results
are available upon request from the authors.
The panel estimations include industry dummies, but for reasons of space these are not reported.

estimations. This suggests that falling net, or inter-industry, trade (i.e. rising IIT)
lowers environmental regulation differences.
In sum, we have found evidence to suggest that environmental regulations are
statistically significant determinants of the share of inter-industry trade. Further-
more, we also find evidence to suggest that both environmental regulations and
IIT should be treated as endogenous variables.

4. INTERPRETATION OF THE ECONOMETRIC RESULTS

At face value, our results from the ‘new’ trade model may appear to contradict
those from the HOV model by finding a significant relationship between environ-
mental regulations and trade patterns. However, it is important to be clear how
these two models differ. The HOV model found no statistically significant relation-
ship between an individual country’s environmental regulations and that country’s
volume of net exports in a pollution-intensive industry. In contrast, the ‘new’ trade
model concentrates on bilateral trade and the shares of intra- and inter-industry
trade in total trade.
Equation (3), which defined the GL index and which formed our dependent
variable in the ‘new’ trade model, can also be defined in the following way:

© Blackwell Publishing Ltd 2003

1178 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

| Xijk − Mijk |
GLi = 1 − (7)
( Xijk + Mijk )

where | Xijk − Mijk | denotes the absolute value of net trade.11 Thus, our results
indicate that the smaller the differential between two countries’ environmental
regulations the smaller will be the share of the absolute value of their net trade,
in total trade. To put it another way, the larger the differential between two
countries’ environmental regulations the larger the share of net trade in total
trade. Since the GL index incorporates the absolute value of net trade, we are
saying nothing here about the direction of any change in net trade (i.e. whether
it represents an increase in net exports or net imports). Furthermore, since the
absolute value of net trade is expressed as a share of total trade, we are also
saying nothing about the level, or volume, of net trade. When considered in this
way, there is no reason to expect the same relationship between environmental
regulations and the dependent variable within the two trade models.
Thus, whilst our HOV results provide no evidence to suggest that environ-
mental regulations are reducing net exports of dirty output, our ‘new’ trade results
do suggest that environmental regulation differentials are influencing trade patterns.
This influence is more subtle than that tested for in the HOV model and suggests
that differences in the stringency of regulations between two countries influence
the composition of trade between those countries, i.e. whether two countries trade
within the same, or different, industries. Furthermore, this finding is made whether
we use a full sample, a North-South sample or a North-North sample.
In terms of the pollution haven hypothesis, whilst the ‘new’ trade model
cannot provide definitive results, a rising share of net trade in total trade, associ-
ated with bilateral regulation differentials, is consistent with the existence of
pollution haven effects. It suggests that countries with relatively lax environ-
mental regulations may possess a comparative advantage in pollution-intensive
output. Although the HOV model provided no direct evidence of this, the ‘new’
trade model does focus on bilateral trade and does control for the determinants of
intra-industry trade. As such, it may be a more appropriate model in which to
model issues such as this.

5. SUMMARY AND CONCLUSIONS

The complex interrelationships between trade, environmental regulations and

the composition of the global economy have become a focal point for international

11
Total trade is equal to the sum of intra-industry trade plus the absolute value of net trade, i.e.
Total tradeijk = 2min(Xijk, Xikj) + | Xijk − Mijk |.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1179

policy makers. With this in mind, this paper has examined trade patterns, in the
context of two trade models, to ascertain whether the influence of environmental
regulations is discernible.
Within the HOV model, we found no evidence to suggest that either of our
two measures of environmental regulations were statistically significant deter-
minants of ‘dirty’ net exports. However, we did find that net exports from iron
and steel and chemical industries were highest in capital-abundant countries,
whilst net exports of non-ferrous metals and paper and pulp were highest in
countries endowed with minerals and forests, respectively. Both of these findings
may explain why environmental regulations are not influencing trade patterns by
a greater amount. In the case of capital, since it is the developed world that is
capital abundant, this may explain why Northern iron and steel and chemical
industries are not relocating to the developing world, even in the face of stringent
environmental regulations. Similarly, in the case of other natural resource endow-
ments, the reliance of non-ferrous metals and paper industries on such locally
sourced resources may again explain why they are not relocating to take advant-
age of lower regulations. The estimation of a simultaneous equations HOV model,
which allowed for the possible endogeneity of environmental regulations and net
exports, did not change any of these findings.
In contrast to the HOV model, the ‘new’ trade model does not estimate net
exports but rather the share of total trade that is intra- and inter-industry. Thus,
whilst it still explains inter-industry trade (i.e. net trade) it now does so in a
manner that simultaneously explains intra-industry trade. As has been noted, this
approach does not allow us to identify the direction of any change in net trade
resulting from a change in regulations. Thus, we are essentially asking a different
question to that asked within the HOV section. We are asking whether environ-
mental regulations, like other factor endowments, influence the composition of
trade, i.e. the extent to which two countries trade within the same, or different,
industries. Our results suggest that the shares of trade that are intra- and inter-
industry are indeed influenced by environmental regulation differentials between
two countries. If regulations are treated as exogenous variables, we find them
to be negative and statistically significant determinants of IIT shares within the
full sample, and negative and non-significant determinants in the North-South
sample. In common with the IIT literature (e.g. Hummels and Levinsohn, 1995)
we also find country size, preferences and a common border dummy to be signi-
ficant determinants of IIT shares. Contrary to expectations, we find capital-labour
differentials to be a positive determinant of IIT shares. Once environmental regu-
lations and IIT shares are treated as endogenous variables we find the coefficients
on the ENVREG variable increase in size and significance and also find this
variable to be now significant within the North-South sample. Whilst we are not
directly modelling the direction of net trade, we have noted that an increased share
of net trade in total trade, resulting from an increase in bilateral environmental

© Blackwell Publishing Ltd 2003

1180 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

regulations differentials, is consistent with the pollution haven hypothesis.

Finally, IIT shares are also found to be a negative determinant of environmental
regulation differentials, suggesting that falling inter-industry trade shares (e.g. a
falling share of net exports) are associated with falling environmental regulation
differentials.
We should finish on a note of caution. Although the analysis draws on a
reasonably large number of cross-sections (60 countries in the HOV model, 630
trade-pairs in the IIT model), we have data for only one year (1995). This reflects
the fact that our preferred environmental regulations variable (ENVREG) is only
available for 1995.

APPENDIX A

Theoretical Background

a. The Heckscher-Ohlin-Vanek Model

This section follows Murrell (1990) and is constructed to derive equation (1)
in Section 2. We do not include all of the intermediate steps but simply those that
are pertinent to the derivation of our equation (1).
The standard HOV model assumes (1) many goods (i = 1 . . . N), many endow-
ments (k = 1 . . . S) and many countries ( j = 1 . . . T) where S = N,12 (2) identical
linearly homogeneous production functions for homogeneous products with given
technology, (3) identical homothetic preferences, (4) immobile factors of produc-
tion between countries but mobile within a country, (5) no transport costs or trade
barriers. To derive equation (1) we also assume sufficient factor endowment
similarities so all countries are within the same ‘cone of diversification’ and that
perfect competition in factor and product markets and constant returns to scale
results in factor price equalisation.
Let Qij be the amount of good i produced by country j where Qj is the vector of
N outputs and Vkj be the jth country’s endowment of factor k where Vj is the
vector of S factor endowments. The input-output coefficients make up the factor
intensity matrix A with elements aki representing the quantities of factor k used in
producing a unit of output of good i. Let pi be the price of good i, γ k be the price
of factor k and Gj be the national income (GDP) of country j.
If A is invertible:

Qj = A−1Vj (A1)

12
In the general case N ≥ S. See Leamer (1984, pp. 16–18) for ways in which models with N > S
can be converted into models where N = S.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1181

Exports, Wij are then defined as the difference between production and consump-
tion: Wj = Qj − Cj, where Cj is the vector of consumption for country j and in
addition, ci represents the proportion of income spent on good i and c is the
vector of expenditure shares across all goods. From assumption (3) consumption
of any good, at given prices, is an equal proportion of national income in all
countries. We can therefore describe the cross-country pattern of consumption as:

Cj = cGj. (A2)

Denoting world values with a w subscript (because world production must

equal consumption):

A −1Vw
c= (A3)
Gw

therefore, from (A1), (A2) and (A3):

Wj = A−1Vj − A−1Vw(Gj /Gw). (A4)

Denoting the elements of A−1 by aij we can arrive at:

S  γk  S 
Wij = ∑ ik G  ∑ is sw   Vkj ,
a − a V (A5)
k =1  w  s =1  


where the term in the square brackets is independent of j and therefore constant
across countries, our final equation system is simply:
S
Wij = ∑ bikVkj i = 1, . . . , N j = 1, . . . , T (A6)
k =1

where bik represents the term in the square brackets in equation (A5). This
means we are able to predict a country’s net exports of each of N traded goods in
the world economy from data on its resource endowments in conjunction with
parameters that are constant across countries.

b. A Monopolistic Competition Model of Trade

Consider an economy with two countries (Home and Foreign where * indicates
the foreign country), two factors (K and L) and two sectors. Given assumptions
(2)–(6) of the HOV model, now assume X is a differentiated product subject to
increasing returns to scale and Y is a homogeneous product subject to constant

© Blackwell Publishing Ltd 2003

1182 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

returns to scale. Assuming free entry and monopolistic competition, equilibrium

is characterised by a large number of firms each producing a unique variety of X
and making zero profits. Assume X is the capital-intensive product, the home
country is capital abundant and the number of firms is given by n = X/x, where x
is also the number of varieties.
With zero transport costs and a utility function that rewards variety, all varieties
of X will be demanded in both countries. Moreover, each country will consume
an amount of each variety in proportion to its world share of GDP, G where:

s = G/G and s* = (1 − s) (A7)

and G + G* = G. With balanced trade, the Home country consumes spn*x*

(−spX*) of the Foreign X good, and the Foreign country consumes s*pnx (=s*pX)
of the Home country’s X, and p is the price of all varieties of good X (where the
price of Y is normalised to 1).
The standard result is that there will be two-way trade and that the Home
country will be a net exporter of X and a net importer of Y. The total volume of
trade is given by:

VT = s*pX + spX* + sY − Y (A8)

and the share of trade that is intra-industry is given by:

2 min (s*pX, spX*)

IIT = (A9)
s*pX + spX* + (sY − Y )

where X and X* denote the production of X in the Home and Foreign country
respectively and Y is the total production of Y.
When factor endowments are identical, all trade is intra-industry and no trade
is motivated by relative factor abundance. If a reallocation of factors widens the
capital-labour ratio and the relative size of the country remains unchanged, then
IIT will decrease and inter-industry trade will increase.

APPENDIX B

Data Information

Net exports United Nations (1996), International Trade Statistics Yearbook

IIT National Asia Pacific Economic and Scientific Database (NAPES)
LAB Economically active population, from World Bank (1999) World
Development Indicators 1999 CDROM

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1183

LAB1 Professional and technical workers (thousands). International Labour Office

(various years) Yearbook of Labour Statistics
LAB2 Literate non-professional workers (thousands). Calculated as LAB-LAB1-
LAB3
LAB3 Illiterate workers. Calculated as LAB*illiteracy rate. The latter is from World
Bank (1999) World Development Indicators 1999 CDROM
CAPITAL Physical capital stock. The sum of annual Gross Domestic Investment
assuming an average life of 15 years. GDI data from World Bank (1999).
World Development Indicators 1999
ENVREGS Eliste and Fredriksson (2001) (based on Dasgupta et al., 1995)
ENVPOL Calculated using the change in energy intensity between 1980 and 1995 and
the level of energy intensity in 1980. The former was calculated using the
averages of years 1980 and 1981 and 1994 and 1995, to reduce the effect
of the end-years. The two variables were ranked, these ranks were summed
and then ranked again. These values were then divided by 60 (the number
of countries in the sample). Subtracting the result from 1 then provides a
measure between 0 and 1, with 1 = high regulations and 0 = low regulations.
Energy intensity is defined as total energy use divided by GDP. From World
Bank (1999) World Development Indicators 1999 CDROM.
LEAD, ZINC, Value of extraction (thousand 1995 US$). US Geological Survey (1997
IRON, COPPER and 1998). Minerals Information 1997 and 1998
OIL Value of oil extraction (millions of 1995 US$). International Energy Agency
(1996). Oil and Gas Information 1996
GAS Value of gas extraction (millions of 1995 US$). International Energy Agency
(1997). Natural Gas Information 1997
COAL Value of coal extraction (millions of 1995 US$). International Energy Agency
(1997). Coal Information 1997
TROPFOR Thousand hectares of tropical forest. World Resources Institute (1998). World
Resources 1998/99
NONTROP Thousand hectares of non-tropical forest. World Resources Institute (1998).
World Resources 1998/99
CROPLAND Thousand hectares of cropland. World Resources Institute (1998). World
Resources 1998/99
GDP World Bank (1999) World Development Indicators 1999 CDROM

Countries included in the HOV sample:

1. Argentina 13. Denmark 25. Ireland 37. N. Zealand 49. Sweden
2. Australia 14. Dom. Rep. 26. Italy 38. Nigeria 50. Switzerland
3. Austria 15. Ecuador 27. Jamaica 39. Norway 51. Tanzania
4. Bangladesh 16. Egypt 28. Japan 40. Pakistan 52. Thailand
5. Belgium 17. Ethiopia 29. Jordan 41. P.N.Guinea 53. Trin.&Tob.
6. Brazil 18. Finland 30. Kenya 42. Paraguay 54. Tunisia
7. Bulgaria 19. France 31. Korea Rep. 43. Philippines 55. Turkey
8. Canada 20. Germany 32. Malawi 44. Poland 56. UK
9. Chile 21. Greece 33. Mexico 45. Portugal 57. USA
10. China 22. Hungary 34. Morocco 46. Senegal 58. Venezuela
11. Colombia 23. Iceland 35. Mozambique 47. S. Africa 59. Zambia
12. Czech Rep. 24. India 36. Netherlands 48. Spain 60. Zimbabwe

© Blackwell Publishing Ltd 2003

1184 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

The IIT sample used a subset of 36 of these countries (to produce 630 trade-pairs), since the
NAPES dataset does not report data for all 60 countries. They are:
1. Australia 11. France 21. Mexico 31. Sweden
2. Austria 12. Germany 22. Netherlands 32. Switzerland
3. Bangladesh 13. Greece 23. New Zealand 33. Thailand
4. Belgium 14. Hungary 24. Norway 34. Turkey
5. Canada 15. Iceland 25. Pakistan 35. UK
6. Chile 16. India 26. P.N.Guinea 36. USA
7. China 17. Ireland 27. Philippines
8. Czech Rep. 18. Italy 28. Poland
9. Denmark 19. Japan 29. Portugal
10. Finland 20. Korea, Rep. 30. Spain

APPENDIX C

HOV Estimations with Environmental Regulations and Net Exports

Treated as Endogenous Variables

Non-ferrous Paper and Iron and

Variable Panela Metals Pulp Steel Chemicals

Dependent Variable: Net Exports

LAB1 −13.7 62.2 −126.4 −87.6 96.7
LAB2 −21.1** −23.6*** −22.9** −20.8* −17.1***
LAB3 40.25** 21.7* 56.2*** 55.7** 27.2***
CAPITAL 1.5* −3.0*** 4.5 7.4*** 1.2***
ENVREG 38.5 5.9 11.1 9.2 2.7
LEAD 4.2 −17.5** 37.4*** −4.7 2.0
ZINC −2.5 5.7** −17.4*** 1.4 −0.04
IRON 0.50 1.1** 0.51 1.0 −0.6*
COPPER 0.17 0.58*** 0.11 −0.06 0.04
OIL 3.4 9.0*** −2.3 6.6 0.6
GAS −21.0*** −17.2*** −38.1*** −30.3*** 1.6
COAL 10.3 12.1* 7.7 11.1 10.1*
TROPFOR 8.8 −5.7 22.7* 6.1 12.1**
NONTROP 28.0*** 6.0 97.8*** 8.8 −0.38
CROPLAND −36.5** −31.0*** −29.8 −55.7*** −29.4***
R2 0.79 0.89 0.85 0.78 0.63

Dependent Variable: Environmental Regulations

Per Capita Y 2.9*** 3.0*** 2.9*** 3.0*** 2.9***
Net Exports 7.0 1.9 1.4 −2.9** 1.8
R2 0.77 0.77 0.78 0.77 0.77
n 240 60 60 60 60

Notes:
Estimated using 2SLS. For reasons of space, t-statistics have not been reported. Instead, *, ** and *** denote
significance at 90, 95 and 99 per cent, respectively. Note that almost identical results were estimated when
ENVREG was replaced with ENVPOL. These latter results are available upon request from the authors.
a
Where ‘panel’ refers to the inclusion of all four sectors in the same regression. This estimation includes
industry dummies, but for reasons of space these are not reported.

© Blackwell Publishing Ltd 2003

 ENVIRONMENTAL REGULATIONS 1185

REFERENCES

Anderson, K. and R. Blackhurst (1992), The Greening of World Trade Issues (Harvester Wheatsheaf,
New York).
Antweiler, W., B. R. Copeland and M. S. Taylor (2001), ‘Is Free Trade Good for the Environ-
ment?’ American Economic Review, 91, 4, 877–908.
Balassa, B. and L. Bauwens (1987), ‘Intra-Industry Specialization in a Multi-Country and Multi-
Industry Framework,’ Economic Journal, 97, 923–39.
Baumol, W. J. and W. E. Oates (1988), The Theory of Environmental Policy (Cambridge Univer-
sity Press, Cambridge).
Bergstrand, J. H. (1990), ‘The Heckscher-Ohlin-Samuelson Model, the Linder Hypothesis and the
Determinants of Bilateral Intra-Industry Trade,’ The Economic Journal, 100, 1216–29.
Birdsall, N. and D. Wheeler (1993), ‘Trade Policy and Industrial Pollution in Latin America:
Where are the Pollution Havens?’ Journal of Environment and Development, 2, 1, 137–49.
Bowen, H. P., E. E. Leamer and L. Sveikaukas (1987), ‘Multi-Country, Multi-Factor Tests of the
Factor Abundance Theory’, American Economic Review, 77, 791–809.
Chichilnisky, C. (1994), ‘North-South Trade and the Global Environment,’ American Economic
Review, 84, 851–74.
Cole, M. A. and R. J. R. Elliott (2003), ‘Determining the Trade-Environment Composition Effect:
The Role of Capital, Labour and Environmental Regulations,’ Journal of Environmental
Economics and Management (forthcoming).
Cole, M. A., A. J. Rayner and J. M. Bates (1997), ‘The Environmental Kuznets Curve: An
Empirical Analysis,’ Environment and Development Economics, 2, 4, 401–16.
Copeland, B. R. and M. S. Taylor (1994), ‘North-South Trade and the Environment’, Quarterly
Journal of Economics, 109, 755–87.
Copeland, B. R. and M. S. Taylor (1995), ‘Trade and Transboundary Pollution’, American
Economic Review, 85, 716–37.
Dasgupta, S., A. Mody, S. Roy and D. Wheeler (1995), ‘Environmental Regulation and Develop-
ment: A Cross-Country Empirical Analysis,’ Working Paper No. 1448 (World Bank, Policy
Research Department).
Dixit, A. V. and V. Norman (1980), The Theory of International Trade (Cambridge University
Press, Cambridge).
Ederington, J. and J. Minier (2001), ‘Is Environmental Policy a Secondary Trade Barrier? An
Empirical Analysis’ (mimeo).
Eliste, P. and P. G. Fredriksson (2001), ‘Does Trade Liberalisation Cause a Race to the Bottom
in Environmental Policies? A Spatial Econometric Analysis’ (forthcoming in L. Anselin and
R. Florax (eds.), New Advances in Spatial Econometrics).
Falvey, R. (1981), ‘Commercial Policy and Intra-Industry Trade,’ Journal of International
Economics, 11, 495–511.
Ferrantino, M. J. and L. A. Linkins (1999), ‘The Effect of Global Trade Liberalisation on Toxic
Emissions in Industry’, Weltwirtschaftliches Archiv, 135, 1, 128–55.
Greenaway, D., C. Milner and R. J. R. Elliott (1999), ‘UK Intra-Industry Trade with the EU: A
Multi-Country and Multi-Industry Analysis,’ Oxford Bulletin of Economics and Statistics, 61, 3.
Grossman, G. M. and A. B. Krueger (1995), ‘Economic Growth and the Environment,’ Quarterly
Journal of Economics, 110, 2, 353–57.
Grubel, H. and P. Lloyd (1975), Intra-Industry Trade: The Theory and Measurement of Inter-
national Trade in Differentiated Products (Macmillan, London).
Harris, M. N., L. Konya and L. Matyas (2002), ‘Modelling the Impact of Environmental Regula-
tions on Bilateral Trade Flows: OECD, 1990–96,’ World Economy, 25, 3, 387–405.
Helpman, E. (1981), ‘International Trade in the Presence of Product Differentiation, Economies
of Scale and Monopolistic Competition: A Chamberlin-Heckscher-Ohlin Approach,’ Journal of
International Economics, 11, 305–40.
Helpman, E. (1987), ‘Imperfect Competition and International Trade: Evidence from Fourteen
Industrialised Countries,’ Journal of the Japanese and International Economies, 1, 62–81.

© Blackwell Publishing Ltd 2003

1186 MATTHEW A. COLE AND ROBERT J. R. ELLIOTT

Helpman, E. and P. Krugman (1985), Market Structure and Foreign Trade (MIT Press, Cambridge).
Hummels, D. and J. Levinsohn (1995), ‘Monopolistic Competition and International Trade: Recon-
sidering the Evidence,’ Quarterly Journal of Economics, 110, 3, 799–836.
Jaffe, A. B., S. R. Peterson, P. R. Portney and R. N. Stavins (1995), ‘Environmental Regulation
and the Competitiveness of US Manufacturing: What Does the Evidence Tell Us?’ Journal of
Economic Literature, 33, 132–63.
Janicke, M., M. Binder and H. Monch (1997), ‘ “Dirty Industries”: Patterns of Change in Industrial
Countries,’ Environmental and Resource Economics, 9, 467–91.
Krugman, P. (1980), ‘Scale Economies, Product Differentiation and the Pattern of Trade,’ American
Economic Review, 70, 950–59.
Krugman, P. (1981), ‘Intra-Industry Specialization and the Gains from Trade,’ Journal of Political
Economy, 89, 959–73.
Lancaster, K. (1980), ‘Intra-Industry Trade under Perfect Monopolistic Competition,’ Journal of
International Economics, 10, 151–75.
Leamer, E. (1980), ‘The Leontief Paradox, Reconsidered,’ Journal of Political Economy, 88, 495–
503.
Leamer, E. (1984), Sources of International Comparative Advantage: Theory and Evidence
(MIT Press, Cambridge).
Levinson, A. and M. S. Taylor (2001), ‘Trade and the Environment: Unmasking the Pollution
Haven Hypothesis’ (mimeo).
Linder, S. (1961), An Essay on Trade and Transformation (Wiley, New York).
Lucas, R. E. B., D. Wheeler and H. Hettige (1992), ‘Economic Development, Environmental
Regulation and the International Migration of Toxic Industrial Pollution: 1960–88,’ in P. Low
(ed.), International Trade and the Environment. World Bank Discussion Paper No. 159.
Mani, M. and D. Wheeler (1998), ‘In Search of Pollution Havens? Dirty Industry in the World
Economy, 1960–1995,’ Journal of Environment and Development, 7, 3, 215–47.
McGuire, M. C. (1982), ‘Regulation, Factor Rewards and International Trade,’ Journal of Public
Economics, 17, 3, 335–54.
Murrell, P. (1990), The Nature of Socialist Economies: Lessons from East European Foreign Trade
(Princeton University Press).
OECD (1997), ‘Foreign Direct Investment and the Environment: An Overview of the Literature’
(December, http://www.oecd.org/daf/env/index.htm).
Pethig, R. (1976), ‘Pollution, Welfare, and Environmental Policy in the Theory of Comparative
Advantage,’ Journal of Environmental Economics and Management, 2, 160–69.
Siebert, H., J. Eichberger, R. Gronych and R. Pethig (1980), Trade and the Environment: A
Theoretical Enquiry (Elsevier/North Holland, Amsterdam).
Tobey, J. (1990), ‘The Effects of Domestic Environmental Policies on Patterns of World Trade: An
Empirical Test,’ Kyklos, 43, 2, 191–209.
Trefler, D. (1993a), ‘International Factor Price Differentials: Leontief was Right!’ Journal of
Political Economy, 101, 961–87.
Trefler, D. (1993b), ‘Trade Liberalization and the Theory of Endogenous Protection: An
Econometric Study of US Import Policy,’ Journal of Political Economy, 11, 138–60.
Trefler, D. (1995), ‘The Case of Missing Trade and Other Mysteries’, American Economic Review,
85, 1029–46.
Van Beers, C. and J. C. J. M. Van den Bergh (1997), ‘An Empirical Multi-Country Analysis of the
Impact of Environmental Regulations on Foreign Trade Flows,’ Kyklos, 50, 1, 29–46.
Xu, X. and L. Song (2000), ‘Regional Cooperation and the Environment: Do “Dirty” Industries
Migrate?’ Weltwirtschaftliches Archiv, 136, 1, 137–57.

© Blackwell Publishing Ltd 2003