M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Industrial Migration in the Chemical Sector: Do Countries with Lax
Environmental Regulations Specialize in Polluting Industries?
Martin Persson
Department of Physical Resource Theory, Chalmers University of Technology / Goteborg University
412 96 Goteborg, Sweden (email: frtmp@fy.chalmers.se)
January 2003
Abstract
This study sets out to capture the effect that environmental policy has on trade and specifically then
examining the notion that lax environmental regulations will constitute a basis of comparative advantage,
causing the least regulated jurisdictions to specialize in polluting industries. This is done partly by a
survey of the empirical literature on this subject and partly by performing an econometrical test of this,
the so-called pollution haven hypothesis. The literature survey presents evidence of a relocation of
production capacity in polluting industries towards less developed countries, with generally less stringent
environmental regulations, during the last decades. Although, there is little evidence that this shift has
primarily been driven by differences in environmental regulations, it is more likely to be explained by
other factors such as increasing domestic demand and more abundant natural resource endowments in the
industrialising countries. The econometrical test adopts a Heckscher-Ohlin-Vanek (HOV) model of trade,
regressing net exports from 54 countries in 23 sectors of the chemical industry on endowments in seven
factors of production and two measures of stringency of environmental regulations based on emissions of
sulphur dioxide emissions (SO2) on a national level and emissions organic matter in wastewater (BOD) in
the chemical sector. Though results indicate that environmental regulations have an effect on trade in a
few sectors, most notably in the inorganic chemical sector, the effect is quantitatively weak and the
results are sensitive to changes in the model. Thus, the weight of evidence suggests that environmental
regulations have had a limited impact on global trade flows and production patterns to date, even in the
most polluting industry sectors.
1 Introduction
The last two decades have seen the emergence of the environmental issue as an
aspect of trade liberalization. A debate that has been spurred by the fear that trade
liberalisations will lead to environmental degradation. The relation between trade, the
environment and the policies governing the two cover a vast area that can roughly be
divided into two main topics:
i. the effect of trade on the environment and
ii. the effect of environmental policy on trade.
The former regards the scale effect of trade, i.e. the fear that liberalized trade will lead
to economic growth damaging the environment through increasing pollution and
unsustainable use of natural resources. This effect might be counterbalanced, however,
as the economic growth, and subsequent higher incomes, raises the demand for
environmental protection, bringing forth cleaner production techniques. The latter topic
covers the compositional effect of trade, i.e. the relation between environmental policy
and global production and trade patterns in pollution intensive goods. 1 It is the
1
The decomposition of the environmental effect of trade into three interacting elements, a scale
effect, a technique effect and a composition effect, was first introduced as a theoretical concept by
Grossman & Krueger (1991).
1
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
combined effect of these three components, the scale, technique and compositional
effect, that will determine whether increased trade will be detrimental to the
environment or not.
This paper deals with the latter topic and specifically the question to which extent
differences in domestic environmental policies is a source of comparative advantage
among nations, leading to a shift in global trade patterns. The debate on the composition
effect has given rise to two related hypotheses. First it is argued that unilateral adoption
of strict environmental policies will cause heavily polluting industries to relocate due to
lost competitiveness from the higher cost of compelling with stringent standards. This is
the industrial flight-hypothesis, often referred to as a push effect of strict environmental
regulations.
The second hypothesis implies that lax environmental standards will become the
source of comparative advantage among nations, turning the least regulated areas into
pollution havens. This then represents the pull effect of lax environmental regulations,
attracting pollution intensive production. Although these two hypotheses are
interconnected, the pollution haven hypothesis does not presuppose the industrial flight
hypothesis. Nations with lax environmental regulations might specialize in pollutionintensive sectors, in line with the pollution haven-hypothesis, without the actual
relocation of industries between countries. Thus one might have industrial migration, to
pollution havens, through shifts in the global production patterns of polluting industries,
but without industrial flight.
There is evidence that, during the last decades, a migration of production capacity
in polluting industries to less developed countries, with generally less stringent
environmental regulations, has taken place. The purpose of this study is to investigate to
which extent this shift in production patterns is a result of internationally differing
environmental standards. Empirical studies so far do not strongly support the pollution
haven hypothesis, stating that other factors are more important than environmental
regulations in determining competitiveness among nations, even in the most polluting
sectors. On the other hand we have, more anecdotal, evidence of industrial flight from
the OECD due to environmental regulations in sectors such as textile, tanning, wood
furniture finishing, mineral processing, commodity chemicals and phosphate fertilizers.
The question is whether these examples are just rare exceptions or if empirical studies
so far have not been able to unravel the full impact of differing environmental regimes
upon global production patterns?
This study sets out to answer this question, partly by examining the shortcomings
and caveats, as well as the strengths, of the empirical evidence presented so far, and
partly by performing an econometric test of the pollution haven hypothesis in the
chemical sector, based on trade data from 54 industrialized and developing countries in
1992. The latter is done building on the knowledge from the studied empirical work,
utilizing a Heckscher-Ohlin- Vanek (HOV) model of trade and incorporating the
strictness of environmental regulations as a variable. Especially this study sets out to
improve the latter, the measurement of environmental stringency in a country.
The paper is structured as follows: Section 2 gives a background to the pollution
haven hypothesis, providing the theoretical foundation of the issue and by investigating
the importance of this question in the process of environmental policy- making. In
section 3 a survey of the empirical evidence to date is presented, analyzing strengths
and weaknesses of these studies. This section ends with a discussion on what
conclusions can be drawn from the presented studies and how one can improve the
2
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
empirical testing of the pollution haven hypothesis. Some of these conclusions are then
incorporated in the model presented in section 4, aiming at a test of the pollution haven
hypothesis in the chemical sector. A large part of this section is devoted to developing
an adequate measure of the stringency of a country’s environmental regulations. The
results from this model are presented and discussed in sections 5 and 6 respectively. The
conclusions of this discussion and the previo us sections are then presented in section 7,
ending the paper.
2 Background
This section offers a background to the pollution haven hypothesis from two
different perspectives. First the theoretical foundation of the pollution haven hypothesis,
in a trade theory-context, is presented. Then, in section 2.2, we will have a look at how
the notion that strict environmental regulations affect competitiveness across nations
influence the environmental regulatory process. This serves as a backdrop against which
one should relate the discussions in the following sections.
2.1 Pollution Havens - The Theoretical Background
International trade theory draws upon differences among nations to explain the
patterns of world trade. In the Heckscher-Ohlin model, the dominating trade theory of
the 20th century, comparative advantages and trade among nations are determined by
differences in endowments in the factors of production. In a model with two trading
countries, two traded goods and two factors of production, labor and capital, it can be
shown that a country will export the good that uses intensively its relatively abundant
factor and import the good that uses intensively its relatively scarce factor. This,
referred to as the Heckscher-Ohlin theorem, implies that a country will specialize in the
production that utilizes factors in which it is abundant but, contrary to the Ricardian
model, will generally not specialize completely under free trade.
Expanding this model to an arbitrary number of goods and factors means that the
Heckscher-Ohlin theorem has to be reinterpreted. Differences in factor endowments will
still determine comparative advantage but will refer to trade in factor services rather
than in goods. That is, a country will export the services of its abundant factors and
import the services of its scarce factors. This is the essence of the Heckscher-OhlinVanek(HOV) theorem. Especially when the number of goods exceeds the number of
factors, trade as net exports of goods is indeterminate and cannot be derived from factor
endowments. Thus goods serve as a bundle through which factor services are exchanged
and the same factor content can be achieved through many bundles of goods.
Although differences in factor endowments are not the sole determinants of trade it
provides one of the most important explanations for observed international trade. Also
other determinants of trade such as differences in technology and economics of scale
may be incorporated in the Heckscher-Ohlin framework by relaxing some of its
underlying assumptions. This model can also be extended to include stringency of
environmental regulation as a factor of production. Starting with the environment itself
it can be seen as a factor of production used directly as an input in agricultural or
industrial production or indirectly as a receptor of waste, air and water pollution that are
byproducts of production. Thus high assimilative capacity of air, water and land in a
country implies a high environmental endowment.
3
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
The environmental endowment can then of course be altered by regulations,
limiting the use of the resource. Thus a country can move from being relatively
abundant in environmental endowments to being relatively scarce, when implementing
tougher environmental standards. In summary then, the environmental endowment in a
country is determined by two factors: the assimilative capacity of the environment and
the social preferences regarding environmental quality. Following this, international
trade can, in line with the HOV theorem, in part be seen as an implicit trade in
environmental services through embodied pollution.
This means that in a world with no impediments to trade, countries that are well
endowed in environmental factors will specialize in the more polluting industries, even
if we have optimal environmental policies in place. Thus, it is not the existence of
pollution havens per se that is the problem but rather that countries do not enforce
optimal environmental regulations for fear of losing comparative advantages. But in a
world where environmental policies in general are sub-optimal pollution havens may of
course increase pollution further, increasing the welfare losses of the sub-optimal
policies.
Thus, the theoretical notion that differences in environmental standards constitute a
basis for comparative advantage among nations, and that this will lead to countries with
weak regulation specializing in polluting industries, is not challenged. But more
ambiguous are the theoretical predictions concerning the environmental and welfare
consequences of this effect, especially when combining this, the composition effect,
with the scale and the technique effect.
2.2 Pollution Havens and the Race to the Bottom
Regardless of empirical evidence in favor of the industrial flight- and/or the
pollution haven- hypotheses, the argument that environmental regulation affect
competitiveness is certainly one that is present in the environmental policy- making
process. It is an argument frequently used by industrial lobbying groups opposing
stricter environmental standards as well as politicians fearing the loss of jobs and
diminishing investments in domestic industries (Lomas, 1988; Esty & Geradin, 1998).
This has lead to the fear from environmental groups that differences in the stringency of
environmental regulation might trigger a race to the bottom2 as nations competing for
jobs and investments try to undercut the environmental standards of their competitors.
Some authors argue that it would be economically irrational for jurisdictions to
enter a race to the bottom (Nordhaus, 1994; Revesz 1992), asserting that regulatory
competition will lead to adoption of efficient environmental regulations (Revesz, 1992).
Although this view has been criticized as unrealistic (Porter, 1999; Esty & Geradin,
1998). Also, more importantly, even though entering a race to the bottom may be
economically irrational, the fears that stringent environmental regulations pose a threat
to competitiveness, and thus employment, economic growth etcetera, may in a political
context prove strong enough to initiate a regulatory race, regardless of the legitimacy of
2
The theory of the race to the bottom was originally developed in a context of US federalism,
building on the logic of the prisoners dilemma -situation (for a good presentation of these issues, see
Revesz, 1992), and is thus used not exclusively in an environmental regulatory context. Also the term
does not imply that regulation will actually hit bottom but rather that the (environmental) standards
resulting from the process are suboptimal, thus the term ‘race towards the bottom’ would be more
appropriate (Porter, 1999; Esty & Geradin, 1998).
4
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
these claims. Especially if the environmental benefits, weighed against the potential
losses in welfare in other areas, are generally under-priced.
Although there are some, anecdotal, evidence of lowering of environmental
standards or the lax enforcement of existing standards (Vogel, 2001; Esty & Geradin,
1998) there seems to have been no general lowering of environmental regulations due to
competitiveness concerns (Porter, 1999). Porter (1999) argues that, at least in the OECD
countries, this is due to the strong support for stringent environmental regulations
making it politically impossible to lower already implemented standards. On the other
hand there is ample evidence showing that although standards are not relaxed to any
great extent, competitive concerns strongly hamper the process of strengthening
environmental regulations. This has been referred to as the regulatory chill or political
drag effect (Porter, 1999; Esty & Geradin, 1998).
Esty & Geradin (1998) draws upon several examples where initiatives to implement
energy taxes, as a response to climate change, in the EU, US, Australia and Japan has
failed due to industrial lobbying and competitive concerns. A recent Swedish example is
that of the green certificates-reform, subsidizing renewable electricity production, where
the electricity intensive industries were excluded due to the cost buying green
certificates would imply. Porter (1999) finds an even stronger political drag effect in the
rapidly industrializing countries, with generally low environmental standards and a big
industrial influence on the regulatory process, due to the high economical dependence
on the export-industries. This gives rise to what Porter calls a stuck at the bottom-effect
regarding environmental policies for these countries.
In an excellent study, Owen Lomas (1988) shows how competitiveness concerns
has shaped the environmental policy- making within the European Community. The
work towards a uniform environmental policy in the Community was essentially
justified in economic terms, harmonized standards seen as a prerequisite for a
functioning common market where industries in different countries could compete on
equal terms. Thus there is a common interest in economic integration policy and
environmental policy, both which may be opposed by the self- interest of individual
member-states seeking to keep their comparative advantages in certain areas.
The problem when environmental policy is warranted mainly as to serve the
objective of economic integration is for example that the harmonization of
environmental standards becomes more important than what the policy actually
achieves on the environmental area. Also if the environmental policy is subordinated to
economic integration, when these two interests do not coincide, the environmental
policy will suffer. Lomas presents several cases concerning air and water pollution
policy within the Community where this has been legible and where the economic selfinterests of individual member states have severely reduced the stringency of the
resulting policies. Thus, although in the beginning competitiveness concerns provided
the impetus in developing common environmental standards within the Community, for
the same reasons the quality of the resulting policies have been far from satisfying from
an environmental perspective.
One should also note that a mirror image of the race to the bottom problem also
exist in the environmental area, for example regarding the treatment of hazardous waste.
Here the negative environmental consequences by far outweigh the economical benefits,
which may trigger a race to the top as countries seek to drive out certain activities from
their jurisdiction by strengthening environmental standards (Esty & Geradin, 1998;
Revesz, 1992). This problem, often referred to as the NIMBY (not- in- my-back-yard)
5
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
syndrome, may thus also lead to the adoption of non-optimal standards and location of
the concerned activities.
As has been shown above the political drag effect is a reality, leading to a
regulatory chill in the environmental area. We will now turn to the underlying
assumption behind this effect, that lax environmental regulation actually is a source of
comparative advantage among nations, leading to industrial flight and the creation of
pollution havens.
3 Pollution Havens - The Empirical Evidence
This section examines the empirical work concerning the pollution haven
hypothesis to date. Although quite extensive, the vast amount of articles regarding this
subject makes this survey far from complete, though it covers some of the major works
in the area. It takes as its starting point the increased costs inflicted upon the industry by
strengthened environmental standards and the effect this has on competitiveness. The
empirical work on the pollution haven hypothesis is then presented, divided into three
different categories depending on the method used in the study. The section ends with a
discussion on what conclusions can be drawn from the presented empirical evidence and
how to improve the methods by which they are performed.
3.1 Environmental Control Costs and Competitiveness
At the heart of the industrial flight and the pollution haven hypotheses lies the
notion that stringent environmental regulations affect competitiveness through increased
production costs. If stringent environmental standards do not levy higher cost on the
companies there will be no loss in comparative advantages and consequently no
changes in production or trade patterns in polluting goods. Thus the natural starting
point when trying to empirically verify the pollution haven hypothesis is to examine the
environmental control costs (ECC).
There are many ways in which environmental regulations may affect production
costs, directly and indirectly. The direct costs may include expenses for new
technology, acquiring special know-how, additional labor costs etc (Sterner, 1996).
These costs are also dependent on the manner in which the environmental regulations
are imposed, through incentive-based policy instruments or through command-andcontrol regulation. Thus not only the stringency but also the form of environmental
standards may affect industrial location (Jaffe et al., 1995). The indirect effect of
environmental regulations, increasing costs for polluting industries, might for example
arise from limiting technical options, waste disposal options and the availability of
certain inputs (Sterner, 1996; Xing & Kolstad, 2002) or through higher legal and other
transactional costs (Jaffe et al., 1995). Then there is the argument that harder
environmental regulations might actually trigger innovations and improvements in
efficiency and thus improved competitiveness, the so called Porter hypothesis (Porter &
van der Linde, 1995; Jaffe et al., 1995; Sterner, 1996; Ekins & Speck, 1998), although
this notion has been heavily challenged.
Due to this wide array of direct and indirect effects 3 , estimating real world ECC is
not an easy task. The first problem is that defining which expenditures should be
3
For a comprehensive list, or ‘taxonomy’, of societal costs that arise from environmental
regulations, see Jaffe et al. (1995)
6
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
included in ECC is rather arbitrary (Jaffe et al., 1995; Sterner, 1996). The cost for endof-the-pipe solutions is quite easily obtained, but what about capital investments that
improve efficiency as well as environmental performance. Whether or not a cost is
classified as an ECC may also vary with time (Sterner, 1996). Finally should costs for
ECC above the regulation level, for a company that wishes to maintain an
environmental image, be included? (Jaffe et al., 1995)
Still, a number of studies on ECC have been undertaken, concluding that ECC
constitute a small part of total production costs, typically 1-3 per cent in the most
polluting sectors (Tobey, 1990; Medhurst, 1993; Jaffe et al., 1995; Sterner, 1996).
Although, these costs might be underestimated because of the difficulties in measuring
them. Also, while ECC in running costs might not be that high, regarding investments
environmental expenditures in some sectors constitute 18-20 per cent of total annual
investments (Stevens, 1993). Moreover, marginal ECC is likely to be higher than
average ECC, meaning that in the future ECC might rise substantially with stricter
environmental policies and become an increasingly important determinant of
competitiveness and trade (Stevens, 1993; Copeland & Taylor, 1994). Recent case
studies also show that in certain sectors ECC are in fact high enough to have an impact
on trade patterns (Larson et al., 2002).
To conclude, one should also in this setting distinguish between the concept of
competitiveness concerning companies on the one hand and nations on the other (Ekins
& Speck, 1998). The discussion above regarded the former and latter was at part
covered by the discussion at the end of section 2.1. While a loss of competitiveness in a
certain sector in a country may be devastating to industries in that sector the economy as
a whole may restructure so that other, more competitive sectors grow, leaving the
overall competitiveness of the country unaffected. Although, this restructuring may be
costly, especially if the industries experiencing a loss in competitiveness are in
economically important, exporting, sectors (Ekins & Speck, 1998). The empirical
studies concerning the effect of environmental regulations on macroeconomic level are
ambiguous, but, just as on the firm level, the overall impact is most likely to be
relatively small, reducing economic growth in the order of a few tenth of per cent per
annum (Ekins & Speck, 1998).
3.2 Testing the Pollution Haven Hypothesis examining Foreign
Direct Investment Flows
Moving on to the objective of testing the pollution haven hypothesis, there are two
main indicators that can be used to verify the theoretical predictions, capital flows
through foreign direct investments (FDI) and changes in the composition of production,
directly or through trade flows. Changes in FDI ought to be the fastest indicator of the
impact of differences in environmental regulation if the pollution haven hypothesis
holds true, production patterns and trade flows only reflecting long run shifts in a
country’s production composition (Xing & Kolstad, 2002; Sterner, 1996). On the other
hand changes in production patterns in line with the pollution haven hypothesis may
occur without the actual migration of industries or capital. Production in polluting
industries in highly regulated countries may decline while the industries in less
regulated countries start to expand merely through reinvested profits or domestic capital
(Sterne r, 1996). Thus, studies on FDI may test the industrial flight hypothesis, or the
push effect of more stringent environmental regulations, but will not capture the full
7
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
impact of the pull effect and thereby to which extent differing environmental regimes
leads to the creation of pollution havens.
A couple of early studies on US FDI (Walter, 1982; Leonard, 1988), presented in
Sterner (1996) and Jaffe et al. (1995), finds no evidence that US polluting industries are
relocating to areas with laxer environmental standards. Contrary to these studies
Smarzynska & Wei (2001) and Xing & Kolstad (2002) find a correlation, although
weak, between environmental stringency in the host country and FDI. Smarzynska &
Wei look at the investment decisions of 534 major multinational firms in 24 countries in
east/central Europe and the former Soviet Union. Adopting a model where FDI is
dependent on market size, labour cost, corporate tax level, corruption level and strength
of environmental standards in the host country they find that polluting, but not other,
industries are more apt to invest in countries with less stringent environmental
regulations. The relation is weak however and do not survive the numerous robustness
checks, especially it is sensitive to how environmental stringency is measured, the
strongest support gained when it is judged by the host country’s participation in
international environmental treaties.
Xing & Kolstad study US FDI in six different sectors, non-polluting and polluting,
in 22 different countries, 7 being developing countries, for the years 1985/1990. Their
model is similar to that of Smarzynska & Wei, with FDI depending on market size, tax
rate, industrial profitability and environmental regulations. The stringency of the latter
is estimated from the level of SO2 emissions in the host country, a measure that has
been criticised as it depends on industrial activity level and is therefore not properly
exogenous (Sterner, 1996). They conclude that US FDI in sectors with high ECC,
chemicals and primary metals, has a significant negative relationship with the
stringency of environmental regulations in the host country, while this is not the case in
the less polluting sectors. Again caution about the conclusions should be taken, as the
number of observations in the study is quite low.
3.3 Testing the Pollution Haven Hypothesis examining Changes
in Production Patterns
In the studies that search for shifts in production patterns to verify the pollution
haven hypothesis two different approaches can be distinguished. The first is to simply
look at how the overall pattern of production and trade in pollution intensive goods has
changed over time. The second approach, presented in the next section, draws on
theoretical trade models to explain current patterns of production and trade. In a study
employing the former approach Lucas et al. (1992) examine production patterns of 37
sectors in 80 different countries over the period 1960 to 1988. They find that in the
1960’s toxic intensity, defined as toxic releases in pounds per dollar of output, grew at a
faster rate in the developed countries while in the later decades, when strict
environmental policies where implemented in the OECD countries, the toxic intensity
of the developing countries grew at the quickest rate. But the study also shows that
growth in toxic intensity has been most rapid in the developing countries with closed
economies, implying that differences in environmental regulation was not the main
reason for the shift in production patterns. There is in fact a wide range of reasons more
likely than differences in environmental standards, to why the observed shift occurred,
the most natural being that the changes in production simply reflect changes in domestic
demand.
8
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
One way to try to overcome the latter problem is to instead focus on trade flows.
Low & Yeates (1992) do this by looking at the so called revealed comparative
advantage (RCA) of the 40 most polluting industries in 109 different countries for the
period 1965 to 1988. A country’s RCA in a certain industry is defined as the ratio
between the domestic industry’s share in the country’s total exports and that industry’s
share in total world exports. If this index exceeds unity it is interpreted as a revealed
comparative advantage for that country in the specific sector. Low & Yeats find that
while the share of exports from highly polluting industries in the developed world is
declining throughout the period, as is the share of these products in total world trade, the
polluting industries account for an increasing share of the developing countries exports.
As a result, the number of developing countries with a RCA above unity in the polluting
industries rises between 1965 to 1988 with 40 per cent, an increase fivefold that in nonpolluting industries. The number of industrial countries with a RCA above unity in
polluting industries also rose, albeit at a lower rate, 14 per cent. Thus, there was a
general dispersion of polluting industries throughout the world, although this trend was
clearly strongest towards the developing countries.
Although this may be seen as evidence supporting the pollution haven hypothesis
there are again a number of factors, other than environmental policy, that may explain
these results. It may still reflect an increased demand for pollution- intensive goods
within the developing world, resulting in more production facilities being built there
(Jaffe et al., 1995). This would also explain the decline in world exports in these goods
as more developing countries become self reliant in production. Finally the results may
to a large extent be explained by differences in natural resource endowments.
Recognising this, Low & Yeats conclude that the trends in production patterns they find
are “unlikely to be adequately explained by environmental policy”.
Following the methodology of Low & Yeats, Sterner (1996) examines the RCA of
the chemical sector between 1983 and 1992. He finds some shifts in production patterns
of polluting industries towards developing countries but nothing that lends strong
support to the pollution haven hypothesis. Jänicke et al. (1996) examine changes of net
exports as a share of domestic production in 11 natural resource based, major polluting
industries in the highly industrialised countries between 1960 and the beginning of the
1990’s. They find no strong tendency towards relocation of ‘dirty’ industries to
developing countries and explain this with low costs for ECC even in the most polluting
sectors, the fact that these industries are capital intensive and require professional skills,
two factors generally lacking in the developing world, and finally that the developed
world will not implement environmental policies that seriously threaten employment.
When adding the findings of Jänicke et al. to those of Lucas et al. and Low & Yeats,
the conclusion of the latter two, that the shift in production patterns in polluting
industries is not a result of industrial flight and pollution havens, is reinforced. As the
net export-consumption ratios in the highly industrialized countries examined in the
study by Jänicke et al. have been fairly constant, the growth in production capacity in
the developing countries is most likely to be explained by raised demand for these
products within this group of countries. Still, not all industrial sectors in the other
studies are included in that of Jänicke et al., for example the chemical sector, part of
which shows very strong tendency of migration towards developing countries in the
Low & Yeats’ study. Also the level of aggregation is higher in the study by Jänicke et
al. and thus shifts within different industrial sectors will be undetectable. The latter is an
important aspect that we will have reason to return to later.
9
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Summing up, the weight of evidence suggests that a shift in production patterns in
pollution intensive production towards the less developed countries has occurred during
the last decades. Yet, the picture that emerges from these studies is that this shift
primarily have been driven by other factors than differences in environmental standards.
3.4 Testing the Pollution Haven Hypothesis utilising Trade Theory
Models
The second approach when looking at shifts in production patterns and trade in
relation to the pollution haven hypothesis is, as mentioned in the previous section,
testing models of trade theory, such as the Heckscher-Ohlin model presented above,
including stringency of environmental regulations as an explanatory variable. This
means that the effect of environmental regulations can be disentangled from other
factors, suc h as the effect of differing natural resource or capital endowments discussed
above, and thus this kind of studies has a great advantage over the ones presented in the
previous section when testing the pollution haven hypothesis. Another advantage when
incorporating the stringency of environmental regulations explicitly is that shifts in
production patterns between developed countries with differing standards is also
accounted for. Not, as above, assuming that the world consists of two types of countries,
developed ones with uniform, stringent, standards and developing ones with weak
standards.
Probably the first study utilising this approach was that of Kalt (1988). Regressing
net exports of 78 US industries for the years 1967 and 1977 on a set of variables
including estimations of ECC he found a significant negative relationship between
exports and ECC in the manufacturing sector. A bit troublesome though is the fact that
this effect was even stronger when excluding the chemical industry from the sample.
Tobey (1990) examines the exports from five pollution intensive industries, mining,
paper, chemical, steel and metals, in 23 different countries, 10 being developing, in
1975. In addition to the endowments provided by Leamer (1984), composed of capital,
different types of labour, land and natural resources, Tobey includes a variable of
environmental endowment, measured by the stringency of environmental policy based
on an UNCTAD survey (Walter & Ugelow, 1979). In this cross-country analysis Tobey
conducted several statistical tests of the hypothesis that environmental policy affect net
exports but finds no significant evidence for this.
Van Beers & van den Bergh (1997) adopt a somewhat different approach where
bilateral trade flows are studied, in a so-called gravity model. They also set out to
improve the measurement of environmental stringency, better reflecting the costs for
ECC in line with the polluter pay principle. They arrive at two measures, one wide and
one narrow, the latter based on change in and le vel of energy intensity. They also make
a distinction in the study between resource based and non-resource based (footloose)
industries as the location of main production factor of the former, the resource, should
be expected to be far more important than differences in environmental regulation.
Examining the trade flows between 21 OECD-countries they find that stringent
environmental regulations, as measured by the narrow indicator, exert a negative
influence on total exports as well as exports from the most polluting footloose
industries. This relation is not found for aggregated exports from all polluting
industries, probably due to the fact that the majority of them are resource based.
Wilson et al. (2002) follows the HOV- methodology of Tobey (1990) when
examining net exports in five pollution intensive sectors, the same as in Tobey, from 24
10
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
countries between 1994 and 1998. They aim at improving the method used in earlier
studies by incorporating qualitative aspects of factor endowments and by including not
only the stringency of environmental regulation but also the effectiveness in enforcing
it. The former is done by including secondary school enrolment rate as a variable, as a
proxy for human capital and technological level of the country. The measure on
strictness of environmental policy as well as enforcement is constructed using data from
a study by Dasgupta et al. (2001), where the state of environmental policy and
performance in 31 countries were assessed through surveys in different industries. The y
find that the effect of environmental regulation and enforcement is negative and
significant for all industries except pulp and paper.
3.5 Can We Close the Book on Pollution Havens?
The findings of the literature survey on the empirical literature on the ind ustrial
flight and pollution haven hypothesis are summarized in Table 1. As have been clear
from the presentation of these studies in the last three sections, the empirical evidence is
mixed. The two studies presented using FDI as an indicator find some evidence of
environmental regulations in the host country affecting investment decisions. Although,
in both cases the results are weak and the authors of both studies caution against
drawing too strong conclusion from them. The studies examining trade and production
patters directly suggest that the growth of production capacity in polluting industries has
been higher in the developing countries than in the developed since 1960. However
there is little evidence that this process been driven by strengthened environmental
regulations in the developed world. The strongest evidence in favor of the pollution
haven hypothesis comes from the studies examining trade patterns, utilizing theoretical
trade models.
Thus, judging by the mixed empirical evidence, the push and pull effects of
environmental regulations seem to be weak, not causing the formation pollution havens
to any great extent. There may be two reasons for this:
i. Pollution havens simply do not exist, there is no strong effect of environmental
regulations upon trade and production patterns causing jurisdictions with lax
environmental standards do not specialize in polluting industries.
ii. Pollution havens do exist, but the empirical studies trying to validate the
hypothesis have for some reason been unsuccessful.
We will now investigate both these possibilities in some detail, beginning with the
former.
There are several reasons why we should not expect to find a strong relationship
between environmental regulation and industrial location. To begin with, production
and trade patterns are determined by a wide array of factors. As have been mentioned
earlier, the cost for complying with environmental regulations is quite low, meaning
that it is not likely to be the determining factor of competitiveness.
Capital abundance is probably an important factor as the pollution intensive
industries generally are capital intensive (Cole, 2000; Antweiler et al., 1998; Jänicke et
al., 1997). These two effects generally counteract as capital abundance and
environmental stringency tends to be correlated, the developed, capital abundant,
countries having strict environmental standards while the opposite holds for developing
countries (Antweiler et al., 1998). Natural resource endowment may in some industries
be another important factor (Low & Yeats, 1992; Jaffe et al., 1995), environmental
stringency having an effect on trade and location only for non-resource based,
11
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Table 1: Summary of the survey of empirical literature on the pollution haven hypothes is. In the results
column, significant indicates that the study finds evidence verifying the hypothesis. For a presentation of
each study, see the text.
Study
Indicator
Time
period
Industrial scope
Geographical scope
Results
Smarzynska &
Wei, 2001
FDI
-
534 multinational
firms
24 transition
economies
Significant /
Insignificant
Xing &
Kolstad, 2002
FDI
1985/90
6 manufacturing
industries
15 developed and 7
developing countries
Significant,
but weak
Manufacturing
output
1960-88
37 manufacturing
industries
80 countries
Insignificant
Low & Yeats,
1992
RCA of trade
1965-88
40 manufacturing
industries
109 countries
Insignificant
Sterner, 1996
RCA of trade
1983-92
49 sectors in 3
manufacturing
industries
50-66 countries in
each sector
Insignificant
Jänicke et al.,
1997
Net exportsconsumption
ratio
1960-92
11 basic materials
industries
32 industrial countries
Insignificant
Kalt, 1988
Net exports
1967/77
78 manufacturing
industries
United States
Significant
Tobey, 1990
Net exports
1975
5 manufacturing
industries
13 developed and 10
developing countries
Insignificant
van Beers &
van den
Bergh, 1997
Bilateral trade
flows
1992
14 manufacturing
industries
21 OECD-countries
Significant for
non resource
based
industries
Wilson et al.,
2002
Net exports
1994-98
5 manufacturing
industries
6 developed and 18
developing countries
Significant
Lucas et al.,
1992
‘footloose’, industries (van Beers & van den Bergh, 1997). Other factors that also affect
competitiveness, and thereby trade and locational decisions, are labour costs, tax rates,
energy costs, infrastructure, market access etcetera (Esty & Geradin, 1998; Jaffe et al.,
1995).
Also the actual expenditures on ECC may be offset by domestic subsidies to
polluting industries, which can take different forms, direct or indirect (Eliste &
Fredriksson, 2002; van Beers & van den Bergh, 2001; van Beers & van den Bergh,
1997). The effect of these subsidies on trade is estimated to be considerable, 87,3 per
cent of annual global subsidies, amounting to at least $950 billion, affecting 96,7 per
cent of world trade (van Beers & van den Bergh, 2001). Another way by which
domestic policy might affect the predictions of the pollution haven hypothesis is import
restrictions, through environmental standards, that accompany tougher domestic
environmental standards (van Beers & van den Bergh, 1997).
12
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Finally, when it comes to the industrial flight hypothesis the costs for the actual
relocation of industries might be considerable, for instance through the presence of large
sunk costs (Esty & Geradin, 1998). Also many multinational firms may chose to apply
the environmental standards of their home country in all their plants for efficiency
reasons and to avoid environmental liability, the so called Bhopal-effect (Smarzynska &
Wei, 2001; Vogel, 2001; Esty & Geradin, 1998; Jaffe et al., 1995). On the other hand,
multinational firms account for a relatively small share of the production in most
developing countries (Vogel, 2001) and thus, again, one might have a strong pollution
haven effect not coupled to industrial flight.
The conclusion of the arguments above is not that competitiveness is not affected
by environmental regulation but rather that we should expect the effect to be rather
small, thus having limited impact on trade flows and production patterns. On the other
hand there are examples of industries where evidence points at strict environmental
regulations as a reason for relocation of production capacity, such as the textile, tanning
(Esty & Geradin, 1998), wood furniture finishing, mineral processing, commodity
chemicals (Stevens, 1993) and phosphate fertiliser industries (Heerings, 1993).
In the case of the fertilizers industry in Western Europe, studied by Heerings
(1993), tightening of environmental standards in the sector, in the late 80’s and early
90’s, lead to relocation of production capacity, mainly to producers in Northern Africa
with less stringent environmental regulations to comply with. Differing environmental
standards do not alone explain this shift however. Shrinking European demand for
fertilizers together with hardening international competition from producers in Northern
Africa and USA with direct access to raw material started this process that was
accelerated by the implementation of stricter environmental standards.
In the tanning industry, examined in a series of studies from the FIL Research
Program at Oslo University (Gjerdåker, 1999; Hesselberg, 1999; Knutsen, 1999;
Odegard, 1999), there is also evidence of locational shift both on a regional and on a
global scale, driven by strengthened environmental regulations within Western Europe.
Faced with increasing international competition, following the liberalisation of the
economies in the Czech Republic, Poland and Brazil, tanners in these countries were
forced to specialise in the most polluting, low quality, activities in order to survive.
Tanners in Western Europe on the other hand, confronted with more stringent
environmental regimes, has responded by outsourcing the most polluting activities to
these countries, while specializing in design and less polluting processes, requiring high
quality raw materials. This aspect is important as it represents a division of labour
within the industry that might not be evident when studying it on an aggregated level as
in the studies presented above.
As in the case of the fertilizers industries the studies of the tanning industry in
Germany, Italy and Portugal show that it was not the strengthened environmental
regulations alone that caused the locational shift. These industries where already
weakened by hardened international competition and the loss of domestic client
industries, such as shoe manufacturers. Yet in Germany, which was the first country to
impose stricter environmental regulations, this is said to have been decisive in the
process of relocation. Also important was the limited time for adjustment to the new
standards, which proved as challenging as the actual strictness of them.
In all, the case studies in certain polluting industries reinforce the statement that
location of polluting industries is governed by a complex structure of factors, strictness
of environmental regulations being one. The question still remains however: on a global
13
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
market with increasing competitive pressure, are these, more anecdotal, examples just
rare exceptions or have empirical studies so far not been able to isolate and unravel the
overall impact of environmental regulations on trade and production patterns?
Some of the shortcomings of the empirical studies to date have been outlined in the
presentations above. FDI as an indicator can only be expected to cover part of the
change in production patterns, if there is such a change, as this will only partly occur
through the actual relocation of industries. The studies looking only at the changes in
production and trade patterns have the major drawback of not working with a
measurement of environmental stringency. Thus, from these results alone, it is
impossible to conclude whether the changes found reflect differences in environmental
standards or depend on some of the many other factors mentioned above. The most
promising approach seems to be examining trade patterns utilising trade theory models.
One problem this kind of studies face is the validity of the underlying trade models.
Empirical studies have been sadly underrepresented in the field of trade theory (Davies
& Weinstein, 2001b; Leamer, 1984) and some of the predictions of the HOV- model
have been heavily challenged (Davis & Weinstein, 2001a; Bowen et al., 1987). This
may partly be accounted to the poor availability of data (Maskus, 1991; Bowen et al.,
1987). Also, relaxing some of the standard assumptions, like indifferences in
technology, yields far better results, lending strong support to the HOV- model (Davis &
Weinstein, 2001a; Davis & Weinsten, 2001b; Trefler, 1993). The important point in this
context however, is that although the HOV- model is challenged at some points it is
agreed that differences in factor endowments are important determinants of trade,
though they are not the only determinants (Markussen et al., 1995, pp. 228, Maskus,
1991; Leamer, 1984).
One caveat of the studies utilising the HOV approach to date, i.e. the studies of
Tobey (1990) and Wilson et al. (2002), as well as other studies, is the level of
aggregation. Both these studies examine trade in five aggregated commodity groups
covering the industries with the highest ECC. As these groups include a wide array of
industries, that might have highly differing pollution intensities and levels of ECC, a
strong effect of environmental policy on trade in certain industries might not be
detected. Also, as mentioned in the case of the tanning industry, the level of aggregation
may mask shifts in the division of labour between polluting and non-polluting activities
within an industry sector when examining FDI or trade data. At least concerning large
and diverse sectors such as the chemical industry, shifts of this type within the industry
is much more likely to occur than a total migration of all production capacity to
countries with lax environmental regulation.
Moreover, a major problem in all studies outlined above is the measurement of the
stringency of environmental policy. Bearing in mind that it is not only the stringency of
environmental regulations “on the book” that matters, but also the effectiveness by
which they are enforced and possible offsetting effects of subsidies or other domestic
policies, supporting pollution intensive industries.
Still, later studies show a higher tendency towards accepting the pollution haven
hypothesis. This may be a result of better understanding of the forces governing these
processes thus resulting in better methods for testing the hypothesis. It may also be
result of an increasing effect as environmental regulation in the developed world is
strengthened, resulting in higher marginal ECC, while this process is slower in the
developing world, due to the stuck at the bottom effect.
14
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
4 Method
In this section the method used to test the hypothesis that countries with lax
environmental regulation tend to specialize in polluting industries is presented. The
model adopted builds upon the conclusions drawn in the previous section and thus is an
attempt to improve the models used in former studies. This is done partly by including a
measure of differences in production functions between countries in the model, but most
importantly by developing the measurement of stringency in environmental regulation,
an aspect that is thoroughly investigated in this section. The section continues with a
discussion on how to interpret the results of the final model and what results we are to
expect. Also this section includes description of the data and ends with a discussion on
some econometric aspects of the model.
4.1 The Model
The model adopted in this study builds upon the Heckscher-Ohlin- Vanek (HOV)
model empirically verified by Leamer (1984), utilised in the studies by Tobey (1990)
and Wilson et al. (2002). Leamer showed that net exports in a commodity group
approximately is a linear function of factor endowments so that
Tj = β0 + β1 V1j + ... + βnVnj + εj,
where Tj is net exports from country j, β0 a constant term, βi are factor coefficients
specific for the commodity group in question, Vij is the endowment of factor i in country
j and ε j is an error term. Utilizing this model, trade in a specific commodity can be
regressed on resource endowments across countries to reveal the specific influence of
the different endowments on the trade in that commodity. This study seeks to do that in
the chemical sector, encompassing environmental regulation as an explanatory variable.
This study adopts a model with endowments in capital, K, two types of labour,
arable land, A, and three types of natural resource endowments. Labour endowment is
split up in two groups in order to include knowledge capital. The groups are based on
the degree of education, the first, Lp, without and the other, Ls, with secondary
education. The natural resource endowments are production of solid fuels, i.e. coal, S,
and production of liquid and gaseous fuels, i.e. oil, P, and fossil gas, N.
Also, the model in this study includes differences in the production functions of
capital and labour between countries. This is done in the manner developed by Trefler
(1993), where the endowments in the model are adjusted to productivity-equivalent
units with a factor πij, estimated from the difference between factor prices across
countries. Thus, setting the factor productivity of one country to unity, given
international data on factor prices, the relative factor productivity of the other countries
can be calculated and we have
Tj = β0 + βKπKjKj + βLp πLjLpj + βLsπLjLsj + βAAj + βS Sj + βPPj + βNNj + ε j.
As mentioned earlier, the environment can be seen as a factor of production whose
endowment, i.e. use, is limited by environmental regulations. It is the aim of this study
to see how the stringency in environmental regulation, and consequently the size of the
environmental endowment, affects trade in the chemical sector. We will now turn to the
subject of how to make an estimation of the size of this endowment.
15
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
4.2 Measuring the Stringency of Environmental Regulation
As has already been mentioned, measuring the stringency of environmental
regulations is a task faced with many difficulties. This study aims at measuring the
effect of environmental regulation on competitiveness, an effect that will only occur if
ECC are increased by stricter regulations. But the level of ECC are not only dependent
on the stringency of regulations but also on the form of the regulations, the nature of the
environmental problem subject to regulation, any resulting gains in efficiency from the
regulation, any offsetting subsidies, etcetera. Thus, as there is no linear relationship
between ECC and stringency of environmental regulations, even if there were accurate
estimations of ECC in different countries available, this alone could not be used as a
measure of environmental stringency. An adequate measure of stringency of
environmental regulation should, directly or indirectly, capture: (i) the strictness of the
environmental legislation ‘on the book’; (ii) the enforcement of the regulations; and (iii)
the effect of any offsetting subsidies to polluting industries.
In the literature one can distinguish two main categories of indicators on strictness,
input- and output-oriented (van Beers & van den Bergh, 1997). Input-oriented indicators
relate to a country’s efforts in environmental protection, for example estimates of the
strictness of the legislation, expenditures on environmental research and development or
on pollution abatement and control. This was the approach adopted in the UNCTAD
study (Walter & Ugelow, 1979) used in the empirical analysis of Tobey (1990). Also
the index developed by Dasgupta et al. (2001), used in the empirical work of Wilson et
al. (2002), is based almost exclusively on input-oriented indicators.
But a country that devotes much effort to environmental regulation might still
support polluting industries sensitive to increased ECC through direct and indirect
subsidies. This will then result in less effective environmental regulation that will not be
detected by input-oriented indicators. Also, a country may have strict environmental
policies on the book but as long as they are not strongly enforced they will not affect the
competitiveness of the industry. An example of this is the European Union where
individual member states, at least on an early stage, have sought to delay or only
partially implement common environmental policies (Lomas, 1988).
Output-oriented indicators on the other hand, set out to capture the effect of
environmental regulations. Under the assumption that better environmental performance
is due to stricter regulations, output-oriented indicators are a better proxy of
environmental stringency because it will be an indirect measure of not only the
stringency but also the enforcement of environmental regulations. Also, output-oriented
indicators capture the effect of subsidies offsetting some of the effect of strict
regulations. Thus it seems as if output-oriented indicators may be adequate proxies for
strictness of environmental regulation, satisfying all the criteria posed above.
A problem with output-oriented indicators is that they may be endogenous, meaning
that they will be dependent on what we are supposed to be testing, i.e. the pollution
intensity of a country’s production mix. For example the study of Smarzynska & Wei
(2001), that applies input- as well as output-oriented indicators to assess environmental
stringency, uses as one indicator reductions in lead and CO2 emissions over a time
period. Although these figures are adjusted with the changes in GDP for the same
period, to compensate for changes in total output, they still might reflect changes in the
composition of production in the country towards a more or less polluting mix and are
thus not properly exogenous.
16
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Also, using changes in pollution, or as van Beers & van den Bergh (1997) change in
energy consumption, might also be misleading as countries that implemented tough
environmental standards in an early state will then seem to have weak standards.
Although, as the case study of the tanning industry in Germany, Italy and Portugal
showed (Knutsen, 1999), a limited time to adjust to tougher standards may have adverse
effects on competitiveness and thus this could be interesting to include in a model, but
then one has to distinguish between this effect and that of the absolute stringency of
regulations.
Another important aspect of environmental indicators is the level of aggregation.
When testing the pollution haven hypothesis in a certain industrial sector it seems
reasonable to try to estimate the environmental regulation in that specific sector, as a
country with high ambitions in the environmental area might still make legislative
exemptions or give subsidies to polluting industries it is economically dependent on,
especially then exporting industries competing on an international market. A more
general measure of stringency of environmental regulations might on the other hand
also measure indirect effects of regulations, on prices of energy, transports, raw
materials etcetera, that affect the competitiveness in all industrial sectors. Thus it seems
as if a good estimation of environmental regulation should encompass societal as well
as industrial specific indicators of environmental performance.
What kind of output-oriented indicators are then suitable for estimating stringency
of environmental regulations? Restricting the discussion to pollutants, a suitable
indicator in this setting should at least fulfil the following base criteria : (i) be emitted as
a result of production; (ii) be subject to regulation due to its direct effects on humans or
the environment; (iii) have well known abatement technologies available for
implementation, with moderate or high abatement costs; and (iv) for econometric
purposes have data available for a wide set of developed and developing economies.
Criterion (iii) follows the logic that if a country will not implement environmental
regulations that seriously threaten competitiveness, it still might employ strict
environmental regulations in areas where marginal costs are low but not where they are
high. Two pollutants that satisfy these criteria are atmospheric sulphur dioxide, SO2 ,
and organic matter in wastewater, measured as biological oxygen demand, BOD.
4.2.1 Emissions of Sulphur Dioxide - A Societal Indicator of Environmental
Regulations
Sulphur dioxide is a pollutant both responsible for the acidification of grounds and
waters and it has direct noxious effects on humans, being an important component in
urban smog. Annual anthropogenic emissions amount to about 70 million tons SO2 ,
about three times the natural emissions emanating from volcanoes, decaying organic
matter and sea spray. The main source of SO2 emissions is combustion of fossil fuels,
i.e. coal and oil, fossil gas having negligible sulphur content, accounting for about 85
per cent of global anthropogenic emissions (UNDP, 2000, pp. 64). The second largest
human source is smelting of ores.
Reduced SO2 emissions can be accomplished through shifting to fuels with lower
sulphur content, through desulphurisation of the fuel prior to combustion, through the
use of appropriate combustion techniques such as fluidized bed combustion or through
emission control technologies such as sorbent injection in the combustion process or
flue gas scrubbing. For both the pre- and post-combustion desulphurisation
technologies, efficient but costly techniques are available. Also, shifting to fuels with
17
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
lower sulphur content, be it between coals with differing sulphur contents or from oil to
fossil gas, generally increases costs.
The emissions of SO2 in a country therefore mainly depend on three variables, the
amount of fossil fuels consumed, the sulphur content of those fuels and the use of
abatement technologies. Thus, when estimating the stringency of environmental
regulation by looking at SO2 emissions one should try to take into account all measures
taken in a country to reduce emissions, i.e. use of abatement technologies as well as
choice of fuels. This study aims at doing this by adopting a measure defined as the ratio
between a country’s SO2 emissions from fossil fuel use and the total amount of coal and
oil consumed. The measure, Ej(SO2 ), is defined as follows,
Ej(SO2 ) = Sem j (GgS/yr) / FFCj (PJ/yr),
where Sem j is the annual SO2 emissions in country j and FFCj is the annual sum of coal
and oil consumed in country j. The effect of this is that both the use of sulphur
abatement technologies and the use of low sulphur content fuels are included in the
measure. The latter because if a country chooses to consume fuels with low sulphur
content, this will result in a lower ratio in the same way as will the use of abatement
techniques. Thus this measure will reflect choices between fuels with different sulphur
content in the same category, i.e. coal and oil as well as between them.
One could argue that the consumption of fossil gas also should be included in the
measure, in order to take into account shifts from oil to gas as a mean to lower sulphur
emissions. The crucial assumption in this measure however, is that countries make an
active choice between fuels with differing sulphur content. This is probably accurate for
oil, which to a large extent is traded on a world market. The world markets for coal and
gas on the other hand are much more limited, international trade in coal and fossil gas
accounting for only 13 and 19 per cent of world production respectively (UNDP, 2000,
pp. 124-126). Thus countries consuming large quantities of coal and gas do so to a
larger extent from domestic resources. Yet, the world market for coal has seen a
dramatic change over the last years as an effect of rising demand for low sulphur coal
(IEA, 1997). Also the large reserves of coal available ensure that many countries can
choose between qualities with differing sulphur content. Fossil gas reserves on the other
hand are much more sparsely distributed than coal and thus the consumption of gas to a
greater extent reflect access to gas resources rather than an active choice aimed at
reducing SO2 emissions.
Following the argumentation above, the proposed measure will be accurate in
countries that have adopted measures to reduce SO2 emissions and where the use of
fuels reflects an active choice. Only in countries that have taken no action to mitigate
sulphur emissions and where low sulphur content fuels are used ‘unintentionally’ will
the Ej(SO2 ) measure be misinterpreted as stemming from tough regulations. If a country
that has adopted no sulphur control regulations uses high sulphur content fuels this quite
rightly results in a high value of Ej(SO2 ).
Consequently, it seems as if the proposed measure, Ej(SO2 ), could serve as an
adequate measure of the stringency of the environmental policy in a country on the
societal scale. This as the costs for complying with emission regulations affect all
sectors of the society, either directly, through investments in abatement technologies or
raised fuel costs, or indirectly, through for example increased energy prices. Still,
assuming that the real stringency of environmental regulations is directly proportional to
Ej(SO2 ) implies that a given cut in SO2 emission, per consumption of fossil fuels,
reflects the same strengthening of environmental regulations independently of how large
18
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
the initial emissions where. This also means that, measured this way, environmental
regulations can only be strengthened to a certain point when emissions equals zero, an
event unlikely to occur in reality.
A more realistic assumption is that the strengthening of environmental regulation
reflects a certain reduction in SO2 emissions, in per cent of initial emissions. This would
represent a lin- log relationship between the real stringency of environmental regulations
and the Ej(SO2 ) measure. Also when interpreting the trade model this implies that a
certain reduction in SO2 emissions, measured in per cent, will give a fixed effect on net
exports, which better reflects the increasing marginal cost of emission abatement, and
thus the effect on ECC and competitiveness, than would a linear relationship.
Thus, including this measure in the model as proposed above we have
Tj = β0 + βKπKjKj + βLp πLjLpj + βLsπLjLsj + βAAj + βS Sj + βPPj + βNNj +
βSO2 log(Ej(SO2 )) + ε j.
4.2.2 Emissions of Organic Matter in Wastewater - An Industry Specific Indicator of
Environmental Regulation
Organic matter and chemicals, emitted through wastewater, are by-products of
various industrial activities and is a major source of surface water pollution. The
released organic material is consumed by naturally occurring bacteria, using up the
oxygen dissolved in the water. With high enough releases of organic matter the oxygen
levels in the waters may drop to levels so low that fish and other aquatic organisms
perish. Also, this process leads to the release of ammonium, which, when converted to
ammonia, is poisonous to fish. A low oxygen level is often considered the single most
important factor when determining the extent of pollution in a water body (Nemerow &
Dasgupta, 1991, pp. 4). The rate at which the oxygen is consumed is referred to as the
biological oxygen demand (BOD), measuring the amount of organic nutrients in the
wastewater.
The emissions of organic matter can quite easily be reduced through end of pipetreatment of the wastewater. The pollution control of wastewater is generally performed
in two steps; primary, or mechanical treatment and secondary, or biological treatment 4 .
In primary treatment first large objects are physically removed, then sand, dirt, other
solids and finally the solid organic matter, or sludge, is settled out in subsequent steps.
The purpose of the secondary treatment is to decompose the organic nutrients and this
can be achieved in two main ways; either by passing the wastewater through a trickling
filter, consisting of a rock bed as the site for biological breakdown of the organic matter
by microorganisms, or by a so called activated sludge system, where microorganisms
are dissolved in the wastewater and air is bubbled through in an aeration tank, providing
the oxygen needed for the breakdown.
Primary and secondary treatment removes more than 90 per cent of original BOD in
the wastewater. The technologies for this type of treatment are well developed and
readily available, although at noticeable costs. For example, the World Bank estimates
that installing BOD treatment plants in Thailand for the beverage and textile industry,
contributing to 37 per cent of Thailand’s BOD load, would entail investment costs in the
order of 5.4 billion US dollars and annual operating and maintenance costs of 600
million US dollars (World Bank, 2001, pp. 28).
4
For an extensive presentation of these processes, see for example Peirce et al. (1998), pp.105-123.
19
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
From the above, emissions of organic water pollutants, measured as BOD, also
seem to meet the requirements of an output-oriented indicator as argued for earlier. One
reason for choosing BOD as an indicator is also that, through the World Banks annual
dataset World Development Indicators (WDI), statistical data on emissions specifically
in the chemical industry for a wide range of countries is available. Dividing these
emissions with annual output in the chemical sector in the corresponding country, yields
a measure of to which extent organic wastes from the chemical sector are controlled.
The proposed measure of environmental stringency based on emissions of BOD is thus
Ej(BOD) = BODem j (kg/yr) / Vch j (’000 US$/yr),
where BODem j is the annual BOD emissions in country j and Vch j is the output in the
chemical sector the same year in country j.
The main caveat with this measure is that emissions and output is measured in the
chemical sector as a whole, but emissions of BOD varies between industries. Sectors
with generally high BOD emissions are for example industries producing detergents,
pesticides, plastics, resins or formaldehyde, while the acids and fertilizer industries have
low BOD emissions. Thus, the industrial mix of the country may influence the value of
Ej(BOD) and the measure may not be properly exogenous, invalidating cross country
comparisons. One way to overcome this would be to assign different weights for the
outputs in different sectors according to BOD ‘intensity’. The problem is that the output
data is aggregated in only six groups, dividing for example the chemicals into
‘industrial’ and ‘others’, for which no such emission factor can be determined.
As it is not possible to determine whether the proposed measure meets the
requirements of exogeneity or not, we will have to examine the results of the regression
analysis to see if this might be the case. If the results show a strong statistical
significance for the BOD measure mainly in sectors with generally high BOD
emissions, the measure most certainly is endogenous and will have to be left out.
Otherwise the effect of different production mixes across countries may not be large
enough to cause problems and the proposed measure will be a good indicator of laxity
of environmental regulations in the chemical sector. Also, as argued for regarding the
E(SO2 ) measure, a lin- log relationship between the actual strictness of environmental
regulations and the proposed measure is assumed and thus we have the final model
Tj = β0 + βKπKjKj + βLp πLjLpj + βLsπLjLsj + βAAj + βS Sj + βPPj + βNNj +
βSO2 log(Ej(SO2 )) + βBOD log(Ej(BOD)) + ε j.
4.3 Interpreting the model
Given the model above, one can regress net exports in a commodity group on the
endowments in the factors of production included, plus the measures of stringency of
environmental regulations, across a set of countries. This will then yield estimates of the
factor specific coefficients, βi, for all variables. The method by which these multiple
regressions are performed in this study, is that of ordinary least squares (OLS). We will
now have a brief look on how to interpret the results of these regressions and what
results we are to expect.
In a model where all variables are taken as logarithms, the coefficients can be
interpreted as elasticity’s. As some of the variables in this model, e.g. net exports and
fossil fuel production, assume zero or negative values a logarithmic model cannot be
adopted, meaning that the values of the coefficients cannot be interpreted this way. In
the case of the environmental stringency-variables, a positive value of the coefficient in
20
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
this model implies that a strengthening of regulations, corresponding to a certain
reduction in emissions of SO2 or BOD in per cent, will result in a certain cut in net
exports in dollars. This cut will however be independent on the size of the initial
exports, which is clearly counterintuitive, meaning that one should not draw too strong
conclusions from a quantitative analysis of the results, although a quantitative analysis
may still give an indication as to whether the effect is strong or weak.
Following the logic of the pollution haven hypothesis, we should expect the values
of the coefficients βSO2 and βBOD to be positive and significant in the regressions.
Regarding the other variables, we should expect the value of the coefficient for capital
endowment to be positive as the chemical industry is capital intensive. The coefficient
of unskilled labour should be negative and that of skilled labour should be positive.
Regarding the natural resource variables, both Tobey (1990) and Wilson et al. (2002), as
well as in the original work by Leamer (1984), includes the sum of production of liquid
and gaseous fuels as a variable. The argument for including them as separate variables
here is that as oil, contrary to fossil gas, to a large extent is traded on a world market
and thus domestic production does not present a large advantage with respect to access
to the resource. This means that we should expect the coefficient of oil production to be
insignificant while the coefficients of coal and fossil gas production should be positive
in commodity groups where these resources are used as raw materials.
4.4 The Data
The trade data originally comes from the International Trade Centre (ITC) at
UNCTAD and was kindly provided by Thomas Sterner, taken from his earlier study
(1996). The dataset is assembled by SITC (Standard International Trade Classification,
revision 3) codes at the three digits level, consisting of 23 commodity aggregates in the
chemical sector (SITC codes 511-598)5 in the year 1992, given in millions of US
dollars, 1992 years value. The original dataset comprises trade data for 90 countries, 21
developed countries, 5 transition economies in eastern Europe and 64 developing
countries, although for many of the developing countries trade data was only available
for a few of the trade aggregates. Also, as the data presented below on endowments in
the factors of production were not obtainable for all these countries, the number of
countries included in the final regressions varied between 36 and 48 in the different
aggregates 6 . Still, this greatly exceeds the number observations in earlier studies based
on trade data, none of which is based on more than 24 country observations.
The data on capital was calculated from statistics on Gross fixed capital formation
in constant 1995 US dollars between 1978-1992, taken from the World Bank’s World
Development Indicators, 2002, and assuming an average life of 15 years and a constant
depreciation rate of 13.3 per cent 7 . As in Trefler (1993), the price on capital used to
calculate the productivity factor for capital is the PPP-adjusted Price level on investment
taken from the Penn World Tables, mark 6.1, for the year 1992. Capital is measured in
billions of US dollars. Data on endowment of arable land was also taken from World
Development Indicators, 2002 and is given in million hectares.
5
For a complete list of the 23 SITC codes and their classifications, see Appendix A: SITCclassifications.
6
For a list of all countries that is included in this study, see Appendix B: Country List and the
Measures of Stringency of Environmental Regulations.
7
This approach was adopted from the study by Wilson et al., 2002.
21
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
The data on total labour force was taken from WDI, 2002. In order to divide this
group into two, with and without secondary education, data on gross secondary school
enrolment rates were used. This value is defined as the ratio of total enrolment,
regardless of age, to the population of the age group that officially corresponds to this
level of education. The data are taken from WDI, 2001 and 2002. The average gross
enrolment rate was then calculated for the years 1960-1990, using data in five year
intervals, as this ought to be a good estimate of the share of the active population, in
1992, with secondary education. In order to get a more precise estimate one should of
course account for the changes in population in each country in this time period.
Both labour groups were adjusted to productivity-equivalent units in the manner
described earlier, using wages in manufacturing taken from ILO, Yearbook of Labour
Statistics, 1995, and UN’s Statistical Yearbook, 44th issue. Wages were converted into
PPP adjusted US dollars, 1992 years value, using average annual exchange rates from
UN’s Statistical Yearbook, 44th issue, and consumption based PPP adjustments for
exchange rates taken from Penn World Tables. Labour endowments are given in
millions of workers.
Data on production of fossil fuels were taken from UN’s Energy Statistics
Yearbook, with solid fuels production measured as millions of tons, coal equivalents,
then converted into petajoules (PJ). Liquid and gaseous fuel production is the crude
petroleum production, given in tons and converted into PJ, and the natural gas
production, given in PJ, respectively.
Finding reliable data on SO2 emissions is not an easy task. One caveat is that most
estimates of SO2 emissions available are constructed from data on fossil fuel
consumption. If these estimations are thoroughly preformed, using accurate data on
sulphur contents in fuels and emission factors, encompassing combustion and abatement
techniques used, for each country in question, this poses no problem. But if standard
values on these variables are used across countries and regions, the resulting data in
most cases will be inconsistent with actual emissions.
The SO2 emission estimates in the EDGAR database, version 3.2, suitable for this
study as it contains data for about 230 countries in 1990/95, are constructed this way.
Comparing these estimates with the national emission inventories reported to the
Cooperative Programme for Monitoring and Evaluation of the Long-range
Transmission of Air Pollutants in Europe (EMEP) by the signatories to the Convention
on Long-range Transboundary Air Pollution (CLRTAP) shows large discrepancies, with
EMEP data in the range of 18 to 150 per cent, and on average 74 per cent, of EDGAR
emission estimates. As the EMEP data has gone through extensive investigations on
accuracy it is most likely to be highly consistent with real emissions.
Therefore, only where no other data sources were available, notably in the Latin
American countries, were the EDGAR data used. For the parties of the CLRTAP data
on SO2 emissions originating from fossil fuel use was taken from the EMEP database.
For the Asian countries total SO2 emissions were taken from the study by Streets et al.
(2000), based on the Asian RAINS-model as well as on a series of other studies. As this
study is based on a detailed inventory for 1990 and data for many other countries were
also available only for 1990, all emission data was taken for this year. Where available,
the share of the Asian emissions originating from fossil fuel use was taken from Arndt
et al. (1997), estimating Asian SO2 emissions for 1987-88.
For most of the remaining countries data was taken from the UN Convention on
Climate Change (UNFCCC) country reports. Where data for 1990 was not available,
22
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
data for the year closest was used. Also, in a few countries where data on emissions
from fossil fuel combustion was not available this was proxied with the global share of
total SO2 emissions emanating from fossil fuel use, 85 per cent. The data on national
consumption of coal and oil in 1990 used for calculating the values of Ej(SO2 ) are taken
from the Energy Information Administration (EIA) database World Energy.
As mentioned earlier the data on emissions of organic water pollutants, measured as
BOD, is taken from WDI, 2002. The emissions are given for the chemical sector as a
whole, defined as manufacturing activities in ISIC group 35, chemicals. Annual outputs
in the chemical sector, in thousands of US dollars, are taken from the United Nations
Industrial Development Organisation (UNIDO) and refer to the same ISIC
classification. The estimated values of both measures of stringency of environmental
regulations for all countries in the analysis can be found in Appendix B: Country List
and the Measures of Stringency of Environmental Regulations.
4.5 Econometric Aspects
As mentioned above, the coefficients, βi, are estimated by multiple regressions of
net exports on the explanatory variables, using the ordinary least squares (OLS) method.
In order assure that the OLS estimators are efficient and unbiased, the results have to be
tested for heteroskedasticity. A formal test for this is the White test. As the statistical
tool used in this study has no built- in test for heteroskedasticity a simpler version of the
White test is preformed by testing the correlation between the squares of the residuals
and the explanatory variables and their squares respectively. If this test shows any signs
of heteroskedasticity then a more thorough test should be performed.
A second statistical problem that might arise is that of multicollinearity, that may
occur when two or more explanatory variables are highly intercorrelated. The effect of
this is that the variances of the individual parameters, βi, for these variables are so high
that the individual effects of each of the explanatory variables on the dependent variable
cannot be separated. Two things are worth noting here. First of all high intercorrelations
between variables alone is not sufficient to cause the multicollinearity problem. The
variance of βi depends also on the variance of the error terms, σ2 , and the variance of
the explanatory variable in question, so that intercorrelations between the variables may
be counterbalanced by small σ2 and large variances in the explanatory variables. Thus,
high intercorrelations together with large standard errors in the parameters may be a
sign of multicollinearity.
Secondly, the multicollinearity problem does not violate any of the basic
assumptions regarding the OLS method and all estimators therefore remain unbiased
and efficient and it is only the parameters of the explanatory variables that are
intercorrelated that are affected by large standard errors. This means that problem with
multicollinearity is larger if it affects the variables from which we wish to draw policy
implications, but is less severe if it only affects the other explanatory variables.
Table 2 shows the values of R2 , the coefficient of multiple determination, when
regressing a variable on the other explanatory variables. The high intercorrelations
between the non-environmental variables suggests that multicollinearity might become
a problem, but as argued above, only if the standard errors of their parameters become
so large that no statistical significance can be obtained. For the variables measuring
stringency of environmental regulation on the other hand, there seems to be no risk for
multicollinearity.
23
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Table 2: Intercorrelations between the explanatory variables when regressing one of them on the others.
The high R2 values for the non-environmental variables suggests that multicollinearity might become a
problem although this is not the case for the variables measuring stringency of environmental regulations.
Variable:
Capital
Labour, with primary education or less
Labour, with secondary education
Arable land
Solid fuel production
Liquid fuel production
Gaseous fuel production
Environmental stringency – BOD measure
Environmental stringency – SO2 measure
R2
0.92
0.88
0.98
0.84
0.90
0.82
0.96
0.31
0.11
5 Results
The results from the regressions on the 23 trade aggregates in the chemical sector
are shown in Table 3-6. It confirms the notion that the chemical industry is capital
intensive, the coefficient for capital, βK, being positive and highly significant for most
sectors. The negative impact of unskilled, as well as educated, labor on trade in the
chemical sector is also relatively consistent across the commodity groups. Although this
finding is consistent with previous studies adopting the same approach (Leamer, 1984;
Tobey, 1990) it indicates that total labor force with secondary education might not be a
good measure of knowledge capital, as one should expect that endowment to be
positively correlated with exports in, at least parts of, the chemical sector. The suspicion
that knowledge capital is not well represented in the model is reinforced by the poor fit
of the model in the medical and pharmaceutical group, where endowment in this factor
of production ought to be decisive.
Endowment in arable land shows no consistent impact on net exports in the
chemical sector. The impact of solid fuel production on trade is positive and significant
for nearly all sectors, the only exception being fertilizers where it is negative and
significant. As argued for earlier, the influence of oil production on trade is insignificant
for most sectors while production of fossil gas has a positive impact on net exports in
the chemical sectors, notably in sectors where it is used as a raw material, such as the
fertilizer industry.
The high statistical significance for the non-environmental variables presented
above suggests that multicollinearity is not a serious problem. Also, the correlations
between the residuals and the explanatory variables and their squares, respectively, are
generally low, in absolute value less than 0.10, and never exceeding 0.35. Thus, there
seems to be no risk for biased results due to heteroskedasticity.
The coefficient of E(BOD) is negative and significant in four cases, two of the
inorganic chemicals sectors, the soap, cleansing and polishing preparations sector and
the styrene polymers sector. The negative relationship in the inorganic sectors may be
seen as an indication that the BOD measure is in fact endogenous as argued for before,
as these sectors are low in BOD emissions. The soap and cleansing sector, as well as the
plastics industry, on the other hand is generally very high in BOD emissions. Still, these
results, together with the doubts raised earlier, indicates that E(BOD) may not be an
adequate measure of stringency of environmental regulations.
24
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
The effect of the stringency of environmental regulation on the societal scale,
proxied by the measure E(SO2 ), is positive and significant in five sectors. The most
considerable impact is in the inorganic chemicals industry, where the subsectors SITC
522 and 523 with positive and significant values on βSO2, represent about 82 per cent of
total exports. The accuracy of these results is endorsed by the Low & Yeats study,
where the inorganic chemical sector was one of the industries exhibiting the strongest
trend in relocation of production capacity from the developed to the developing
countries.
The quantitative interpretation of the positive coefficient βSO2 is that if a country
unilaterally strengthens its environmental regulation regime so that the SO2 emissions
are reduced by 10 per cent, then total net exports in the chemical sector will decrease
with about 34 million US dollars annually. About 7 million of this decrease will occur
in the organic chemicals sector (SITC 512), 16 million in the inorganic chemicals sector
(SITC 522 and 523), 10 millions in the soap and cleansing products sector (SITC 554)
and 1 million in the primary plastics sector (SITC 572). This represents a median of
only 0.6 per cent of total exports in the chemical sector in the developed countries in
this study, while it represents a median of 11.9 per cent of total exports in the
developing countries in this study. Although the latter figure is most likely an
overestimate, as data on exports for some sectors in the chemical industry is lacking for
many of the developing countries in this study. Still, as the developed countries are the
main exporters in the chemical sector, developing countries will see its export reduced
more relative to initial exports than will developed countries.
As discussed in section 4.3, one should not draw to strong conclusions from this
quantitative analysis, but one can conclude that although these results present evidence
in favor of the pollution haven hypothesis, the effect is weak.
5.1 Sensitivity analysis
Since the model is based on absolute values on net exports and endowments, a
single country that differ significantly from the overall pattern may determine the results
of the regression analysis to a great extent. Therefore, two alternations of the analysis
will be presented here, in order to test the robustness of the results. The first way to
avoid large impact from single countries in the sample is by removing outliers by
examining the residuals, εi, of the regressions. Doing this, the results of the regressions
are not altered to any great extent regarding the non-environmental variables. In most
sectors the only difference is in the level of statistical significance for these variables.
Thus these results are fairly robust.
When it comes to the environmental regulations variables on the other hand, the
results are much more sensitive. In the regressions with outliers removed βBOD is
positive and significant in four sectors (SITC 516, 531, 583 and 591) while it is negative
and significant in three (SITC 522, 554 and 572). Again these results may partly be due
to endogenity of the variable. The statistical significance of βSO2 is also altered so that it
is positive and significant only in SITC 522 and 572 while it is negative and significant
in SITC 515 and 591. Quantitatively the effect is still weak. Thus, the sensitivity of the
results regarding the environmental regulations variables only strengthens the view that
environmental regulations have a limited impact on trade and production patterns.
25
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Table 3: The OLS estimators of the endowment coefficients, βi , in the organic chemical sector. Statistical
significance of the different factors of production is indicated by *, ** and *** at the 90, 95 and 99 per
cent level respectively. Standard errors are in parentheses. The number of observations is denoted by n.
511
β0
512
-80.6
105.4 **
(112.4)
βK
(53.5)
0.126 **
0.140 ***
(0.073)
βLp
3.09
(17.87)
βLs
-22.49
βA
-0.74
βS
βP
βN
βBOD
βSO2
Adj. R2
n
(1.27)
0.0110
65.6
23.0
(114.2)
(38.6)
0.077 ***
-30.38 ***
2.25 **
(2.66)
34.5
(145.0)
-28.75 ***
(8.15)
0.0645 ***
(0.03)
(0.01)
-0.0237
0.0021
(0.038)
(0.018)
516
-3.9
(0.028)
-32.82 ***
515
(40.9)
(0.034)
(8.49)
(17.40)
Organic chemicals
513
514
0.176 **
(0.091)
(6.87)
3.11
(6.69)
1.87 **
(1.04)
0.0216 **
(0.01)
0.156 **
0.061 **
(0.079)
(0.026)
-34.21 *
-48.68 ***
-23.22 ***
(22.45)
(19.26)
(6.43)
-32.95 *
-32.13 **
-8.45 *
(21.64)
(18.76)
(6.15)
-0.10
3.08
1.07
(3.35)
(2.93)
(0.99)
0.0939 ***
0.0974 ***
0.0422 ***
(0.04)
(0.03)
(0.01)
0.0496 ***
0.0319
0.0046
-0.0030
(0.015)
(0.048)
(0.041)
(0.014)
0.169 **
0.085 ***
0.033
-0.007
(0.07)
(0.03)
(0.03)
(0.09)
(0.08)
-13.3
15.6
28.0
-170.8
-68.2
47.6
(218.0)
(98.8)
(80.8)
(273.1)
(227.4)
(61.4)
28.9
145.4 *
-0.088 ***
0.055 **
(0.03)
7.6
39.9
-43.4
32.1
(191.1)
(89.0)
(71.6)
(234.6)
(200.9)
(65.7)
0.38
39
0.45
40
0.75
42
0.06
41
0.18
44
0.79
42
Table 4: The OLS estimators of the endowment coefficients, βi , in the inorganic chemical and the dyeing,
tanning and coloring materials sectors. Notation as in Table 3.
Inorganic chemicals
522
β0
72.7 *
(53.2)
βK
0.191 ***
(0.035)
βLp
-9.74
(8.50)
βLs
-43.83 ***
(8.25)
βA
-4.15 ***
(1.27)
βS
βP
βN
βBOD
0.0996 ***
Adj. R2
n
524
531
8.6
-5.8
74.9
(37.3)
(32.0)
(109.4)
0.078 ***
(0.026)
-19.47 ***
(6.43)
-15.55 ***
(6.26)
1.08
(0.98)
0.0604 ***
-0.102 ***
(0.021)
0.284 ***
(0.076)
532
10.8 *
(7.1)
0.019 ***
(0.005)
533
97.2
(114.8)
0.250 ***
(0.074)
12.45 **
-32.67 **
-2.48 **
-22.98
(6.44)
(18.44)
(1.25)
(18.04)
-2.58
-54.89 ***
-4.73 ***
-40.47 **
(5.00)
(17.96)
(1.17)
(17.57)
-1.74 **
(0.84)
0.0003
2.59
0.16
-0.43
(2.80)
(0.19)
(2.71)
0.1115 ***
0.0085 ***
0.0859 ***
(0.01)
(0.01)
(0.01)
(0.03)
(0.00)
(0.03)
0.0025
0.0123
-0.0097
-0.0017
-0.0001
-0.0076
(0.018)
(0.014)
(0.011)
(0.040)
(0.003)
(0.039)
0.116 ***
0.021
0.083 ***
0.072
0.009 **
0.114 *
(0.03)
(0.03)
(0.02)
(0.07)
(0.01)
(0.07)
-164.3 *
-121.6 *
-69.1
-93.0
-6.9
-141.3
(74.9)
(68.2)
(217.7)
(9.5)
(210.6)
(102.2)
βSO2
523
Dyeing, tanning and coloring mat.
237.7 ***
116.7 **
44.0
73.4
13.2
253.1
(89.8)
(66.3)
(57.7)
(192.3)
(10.7)
(193.2)
0.60
41
0.60
45
0.92
37
0.30
44
0.28
48
0.37
39
26
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Table 5: The OLS estimators of the endowment coefficients, βi , in the medical and pharmaceutical,
perfume and cleansing product, fertilizer and unrefined plastic sectors. Notation as in Table 3.
Medical & pharmaceutical prod.
Ess. oils, perfumes & cleaning prod.
541
551
553
554
β0
75.3
-5.0
24.9
97.1
57.5
-1.0
(387.8)
(30.1)
(68.5)
(97.7)
(66.3)
(6.5)
βK
βLp
0.043
0.012
-0.002
(0.267)
(0.021)
(0.046)
-100.59 *
-9.53 **
(64.39)
βLs
(5.10)
-7.96
3.27
(15.63)
-33.72
-4.91
1.56
(4.85)
(10.93)
-20.80 **
-0.12
(11.07)
(1.02)
0.62
-0.43
-2.36
1.19
(0.79)
(1.63)
(2.38)
(1.83)
0.0188
0.0607 **
(0.02)
(0.03)
0.0522
-0.0059
0.0211
(0.161)
(0.011)
(0.027)
-0.008
0.013
-0.025
(0.26)
(0.02)
(0.04)
-386.2
-9.8
-25.5
(701.7)
(40.0)
(127.8)
βSO2
0.88
(1.05)
-32.05 **
(0.01)
βBOD
-32.53 ***
(11.68)
0.33
0.0233 ***
-0.003
(0.004)
(15.42)
(0.11)
βN
0.076 *
(0.047)
(9.82)
0.2473 **
βP
0.165 ***
(0.065)
(10.87)
(63.07)
βA
βS
Polymers of styrene
Fertilizers
562
572
-0.27 **
(0.16)
-0.0520 ***
0.0061 ***
(0.02)
(0.00)
-0.0300
-0.0360 *
0.0022
(0.034)
(0.025)
(0.002)
0.129 **
0.262 ***
(0.06)
(0.05)
-335.5 **
-0.005
(0.00)
4.2
(184.5)
-15.9 *
(95.1)
-186.4
-35.0
134.0
(663.0)
(48.6)
(112.8)
(167.3)
(103.5)
(11.1)
0.02
41
0.14
46
-0.07
36
0.23
39
0.88
45
0.65
41
Adj. R2
n
227.8 *
(12.1)
-15.5
15.0 *
Table 6: The OLS estimators of the endowment coefficients, βi , in the refined plastics sector as well as
for chemical materials and products not else specified. Notation as in Table 3.
Plastics, refined
582
583
β0
βK
-34.7
101.5
14.2
3.4
121.6
(127.9)
(247.6)
(70.4)
(30.5)
(170.2)
0.216 ***
(0.085)
βLp
-34.04 *
(20.81)
βLs
βA
βS
βP
βN
βBOD
βSO2
Adj. R2
n
Materials & products n.e.s.
591
592
598
0.474 ***
(0.168)
-105.05 ***
(41.15)
-34.26 *
-54.82 *
(20.23)
(39.96)
0.052
(0.049)
0.057 ***
(0.021)
0.262 **
(0.115)
-16.24 *
-5.10
-76.67 ***
(11.91)
(5.21)
(28.34)
-5.48
(11.58)
-19.95 ***
(5.10)
-41.28 *
(27.07)
1.18
6.95
-0.22
0.43
1.51
(3.12)
(6.21)
(1.80)
(0.80)
(4.33)
0.0839 **
0.0846
0.0342 **
(0.03)
(0.07)
(0.02)
-0.0724 *
-0.1232 *
0.0067
(0.045)
(0.088)
(0.030)
0.0284 ***
(0.01)
-0.0325 ***
(0.011)
0.1653 ***
(0.05)
0.0086
(0.061)
0.201 **
0.314 **
0.025
0.076 ***
0.162 *
(0.08)
(0.17)
(0.05)
(0.02)
(0.12)
125.5
93.2
17.0
-11.1
5.0
(254.8)
(483.2)
(128.3)
(60.2)
(274.5)
-184.8
-63.8
-14.0
-40.3
271.9
(216.0)
(429.4)
(121.9)
(53.5)
(292.3)
0.56
41
0.64
41
0.26
43
0.49
45
0.58
41
27
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Another way to solve the possible problem of single countries, with large absolute
net exports or endowments, affecting the results, is by normalizing the model. Here this
is done by performing the regression with net exports divided by the mean of the
country’s total exports and imports as the dependant variable, thus reflecting the
comparative advantage of a certain sector, and the non-environmental variables taken
per capita. This model is also more intuitive as the environmental regulations variables,
as well as the others, will have an effect on the comparative advantage of a certain
sector, not giving a fixed effect on net exports regardless of the size of the initial exports
as in the original model.
The regressions using this model yield small R2 -values and statistical significance in
fewer cases. In seven of the sectors none of the coefficients are statistically significant.
The most consistent results are those of capital, positive and significant in six sectors,
and βSO2, also positive and significant in six sectors (SITC-511, 514, 522, 533 583 and
598). Again, a quantitative analysis indicates that if the results are correct, the impact of
environmental regulations on exports still is weak. A ten per cent reduction of SO2
emissions will in the concerned sectors result in change in comparative advantage
equivalent to a cut in net exports as a share of average exports and imports of 0.02-0.07
per cent.
6 Discussion
Although this study set out to correct some of the caveats in the earlier studies, it is
far from complete in this aspect. The largest caveat probably still is the measure of
environmental regulations. A more accurate measure of this should probably be
obtained by a multi-dimensional analysis, for example more in line with that of
Dasgupta et al. (2001), not relying on a single indicator. Yet this is a daunting task if
one aims at developing a measure for a large number of countries, especially then
developing countries, due to the lack of data. Although the problem of measurement
errors is likely to be largest concerning the environmental variables, the accuracy of the
data on the other variables may also affect the results to some extent.
Also, when studying such complex issues as this, one should not draw too strong
conclusions solely from aggregated statistical analysis’. Case studies, as those outlined
in section 3.5, may be able to better grasp the intricate system of causes governing the
location of production of polluting industries than this type of analysis. On the other
hand, case studies have to be complemented by aggregated studies in order to get an
overall picture of, in this case, the extent to which environmental regulations affect
production and trade patterns.
Still, even regarding the provisions discussed above, the results are sufficiently
sound to present some evidence of countries with lax environmental regulations
specializing in parts of the chemicals sector. The effect is weak though and the results
are sensitive to changes in the model. Thereby the results strengthen the view that
environmental regulations have a limited impact on competitiveness, trade flows and
production patterns, even in the most polluting industry sectors. As argued for in section
3.5, this does not imply that the theoretical foundation for the pollution haven
hypothesis is incorrect, it merely indicates that so far this effect is relatively weak.
Several reasons why that should not surprise us have also been presented.
28
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
7 Conclusions and policy implications
This study aimed at testing the pollution haven hypothesis in one of the most
pollution intensive industries, the chemical sector. The results indicate that
environmental regulations affect trade in a few of the che mical sectors, most notably in
the inorganic chemical sector, still even here the effect quantitatively is weak. Also,
these results are not robust to the sensitivity checks preformed.
The study also set out to improve some aspects of the methodology used in previous
studies, which was accomplished by including differences in production functions, such
as differing technology, in the model, by testing the hypothesis on a more disaggregated
level in order to reveal shifts in production patterns within the chemical industry
between more or less polluting sub-sectors and by performing the analysis on a larger
set of countries than in previous studies.
Still, the method can be further improved, not least when it comes to the variable
measuring the stringency of environmental regulations. Some of the weaknesses, such
as the difficulty of obtaining accurate cross-country data, are inherit in the model
however and can never be totally avoided. But while some of the complex of causes
governing the location of polluting industries may be better understood by performing
case studies in certain sectors, such analysis’ have to be complemented by more
aggregated studies of this kind in order to get the overall picture on the interaction
between environmental policy and trade.
From the survey of empirical literature on the pollution haven hypothesis the
following conclusions are drawn. Evidence strongly suggests that during the last
decades a relocation of production capacity in polluting industry sectors towards less
developed countries, with generally less stringent environmental regimes, has taken
place, but there is little evidence that this shift has primarily been driven by differences
in environmental regulations. The observed process is more likely to be explained by
economic growth in the developing world, increasing domestic demand and production
capacity in polluting industries. Access to natural resources also seems to be a factor
driving the locational shift in these industries.
Although the theoretical foundation for the fear that differing environmental
standards across countries represent a comparative advantage that will turn the least
regulated jurisdictions into pollution havens is sound, there is no lack of explanations as
to why the observed effect is relatively weak. First of all, the costs inflicted upon firms
due to environmental regulations generally are quite low, amounting to a few per cent of
total expenditures. Thus, factors such as capital abundance and the aforementioned
natural resource endowments are more important in determining industrial location.
Environmental control costs may also be offset by domestic subsidies to vulnerable
industrial sectors or through efficiency gains resulting from the regulations, the latter
often referred to as the Porter hypothesis.
Although the overall effect of environmental regulations on trade and production
patterns to date is small, in some sectors, such as the leather tanning and the fertilizers
industries, it has had considerable impact. Still, case studies of these industries show
that differences in environmental regulations do not alone account for the shifts in
production patterns. To sum up, the survey of the empirical literature endorses the
results of the econometric analysis, leading to the conclusion that the influence of
environmental regulations on competitiveness to date has been relatively small, having
a considerable effect only in sectors already economically vulnerable.
29
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
It seems as if the most important effect so far regarding this issue has not been that
of environmental regulations upon trade but that of competitiveness concerns regarding
environmental regulations on the environmental policy-making process. There is ample
evidence that the fear of loss of competitiveness and thus employment and economic
growth has caused a political drag, hampering the process of strengthening
environmental regulations. This effect seems to be most pronounced in the rapidly
industrializing countries whose economies to a larger extent rely on exporting
industries, creating a stuck at the bottom-effect.
Although the empirical evidence indicates that today these fears are largely
unwarranted, one cannot argue that differences in environmental regimes will not affect
trade and production patterns in the future and tha t countries can take the action to curb
environmental degradation they see fit, disregarding competitive concerns. On the
contrary, if the political drag effect is strongest in the developing countries, then the
disparity between the environmental regulations in the developed and the developing
world may increase in the future, strengthening the pull effect on polluting industries
towards countries with lax environmental regulations.
Therefore, given the possibility of an increased pull effect in the future and that the
regulatory chill effect already exert a great pressure on the environmental regulatory
process, competitive concerns regarding environmental regulations has to be tackled. A
variety of responses have been suggested by economists, environmentalists and policymakers, ranging from the approach of regulatory competition to that of complete
harmonization of standards across countries 8 .
Important to remember in this context is that differences in environmental standards
between countries may be jus tified on the basis of varying preferences and assimilative
capacity of the environment. Although determining whether real world differences in
regulations are based on legitimate advantages or reflect market failures is extremely
difficult, this has to be taken into account, implying for example that in many cases will
the complete harmonization of standards most likely be economically inefficient. Still,
harmonization may be performed in such a way that differences in standards due to
local or regional circumstances and preferences are allowed for, for example by the
adoption of minimum (or maximum) standards or the harmonization of environmental
targets, e.g. ambient air or water standards.
8
For an excellent introduction to the different types of responses proposed, see Esty & Geradin
(1998). These issues are also covered by Nordhaus (1994) and Revesz (1992).
30
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
References
Arndt, R.L., Carmichael, G.R., Streets, D.G. and Bhatti, N. (1997), “Sulfur Dioxide Emissions and
Sectorial Contributions to Sulfur Deposition in Asia”, Atmospheric Environment, 31(10), pp. 15531572
Antweiler, W., Copeland, B.R. and Taylor, M.S. (1998), “Is Free Trade Good for the Environment?”,
NBER Working Papers Series, Working Paper 6707, National Bureau of Economic Research,
Cambridge
van Beers, C. and van den Bergh, J.C.J.M. (1997), ”An Empirical Multi-Country Analysis of the Impact
of Environmental Regulations on Foreign Trade Flows”, Kyklos, 50(1), pp. 29-46
van Beers, C. and van den Bergh, J.C.J.M. (2001), ”Preseverance of Perverse Subsidies and their Impact
on Trade and Environment”, Ecological Economics, 36(3), pp. 475-486
Bowen, H.P., Leamer, E.E. and Sveikauskas, L. (1987), “Multicountry, Multifactor Tests of the Factor
Abundance Theory”, The American Economic Review, 77(5), pp.791-809
Cole, M.A. (2000), “Air Pollution and ‘Dirty’ Industries: How and Why Does the Composition of
Manufacturing Output Change with Economic Development?”, Environmental and Resource
Economics, 17(1), pp. 109-123
Dasgupta, S., Mody, A., Roy, S. and Wheeler, D. (2001) “Environmental Regulation and Development: A
Cross-country Empirical Analysis”, Oxford Development Studies, 29(2), pp. 173-187
Davis, D.R. and Weinstein, D.E. (2001a), “An Account of Global Factor Trade”, The American Economic
Review, 91(5), pp.1423-1453
Davis, D.R. and Weinstein, D.E. (2001b), “What Role for Empirics in International Trade?”, NBER
Working Papers Series, Working Paper 8543, National Bureau of Economic Research, Cambridge
Ekins, P. and Speck, S (1998), “The Impacts of Environmental Policy on Competitiveness: Theory and
Evidence” in International Competitiveness and Environmental Policies , Barker, T. and Köhler, J.
(eds.), International Studies in Environmental Policy Making Series, Edward Elgar, Cheltenham, UK
Eliste, P. and Fredriksson, P.G. (2002), ”Environmental Regulations, Transfers and Trade: Theory and
Evidence”, Journal of Environmental Economics and Management, 43(2), pp. 234-250
Esty, D.C. and Geradin, D. (1998), “Environmental Protection and International Competitiveness: A
Conceptual Framework”, Journal of World Trade, 32(3), pp. 5-46
Gjerdåker, A. (1999), “Leather tanning in Scandinavia”, FIL Working Papers, No. 18, Department of
Sociology and Human Geography, University of Oslo
Grossman, G.M. and Kruger, A.B. (1991), “Environmental Impacts of a North American Free Trade
Agreement”, NBER Working Papers Series, Working Paper 3914, National Bureau of Economic
Research, Cambridge
Heerings, H. (1993), “The Role of Environmental policies in Influencing Patterns of Investments of
Transnational Corporations: Case Study of the Phosphate Fertilizer Industry”, in Environmental
Policies and Industrial Competitiveness, OECD Documents, OECD, Paris
Hesselberg, J. ed. (1999), ”International competitiveness: The tanning industry in Poland, the Czech
Republic, Brazil and Mexico”, FIL Working Papers, No. 15, Department of Sociology and Human
Geography, University of Oslo
IEA (1997), International Coal Trade – The Evolution of a Global Market, International Energy Agency,
OECD, Paris
Jaffe, A.B., Peterson, S.R., Portney, P.R. and Stavins, R.N. (1995), “Environmental Regulation and
International Competitiveness: What Does the Evidence Tell US?”, Journal of Economic Literature,
33(1), pp. 132-163
Jänicke, M., Binder, M. And Mönch, H. (1997), ”’Dirty Industries’: Patterns of Change in Industrial
Countries”, Environmental and Resource Economics, 9(4), pp. 467-491
Kalt, J.P. (1988), ”The Impact of Domestic Environmental Regulatory Policies on U.S. International
Competitiveness”, in International Competitiveness, Spence, A.M. and Hazard, H.A. (eds.),
Cambridge, Harper and Row, Ballinger, pp. 221-262
Knutsen, H.M. (1999), ”Leather tanning, environmental regulations and competitiveness in Europe: A
comparative study of Germany, Italy and Portugal”, FIL Working Papers, No. 17, Department of
Sociology and Human Geography, University of Oslo
31
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Larson, B.A., Nicolaides, E., Al Zu’bi, B., Sukkar, N., Laraki, K., Matoussi, M.S., Zaim, K. and
Chouchani, C. (2002), “The Impact of Environmental Regulations on Exports: Case Study from
Cyprus, Jordan, Morocco, Syria, Tunisia and Turkey”, World Development, 30(6), pp. 1057-1072
Leamer, E. (1984), Sources of International Comparative Advantage: Theory and Evidence, MIT Press,
Cambridge
Lomas, O. (1988), “Environmental Protection, Economic Conflict and the European Community”, McGill
Law Journal, 33, pp.506-539
Low, P. and Yeats, A. (1992), "Do ‘Dirty’ Industries migrate?" in International Trade and the
Environment, Low, P. (ed.), World Bank Discussion Papers, Paper No. 159, The World Bank,
Washington D.C., pp. 89-103
Lucas, R, Wheeler, D. and Hettige, H. (1992), "Economic Development, Environmental Regulation and
the International Migration of Toxic Industrial Pollution: 1960-88”, in International Trade and the
Environment, Low, P. (ed.), World Bank Discussion Papers, Paper No. 159, The World Bank,
Washington D.C., pp. 67-86
Markusen, J.R., Melvin, J.R., Kaempfer, W.H. and Maskus, K.E. (1995), International trade: theory and
evidence, New York, McGraw-Hill
Maskus, K.E. (1991), “Comparing International Trade Data and Product and National Characteristics for
the Analysis of Trade Models”, in International economic transactions: issues in measurement and
empirical research, University of Chicago Press, Chicago
Nordhaus, W.D. (1994), “Locational Competition and the Environment: Should Countries Harmonize
Their Environmental Policies?”, Cowles Foundation Discussion Paper 1079, Cowles Foundation,
Yale University
Nemerow, N.L. and Dasgupta, A.(1991), Industrial and hazardous Waste Treatment, Van Nordstrand
Reinhold, New York
Odegard, J.T. (1999), “Leather tanning in Brazil”, FIL Working Papers, No. 19, Department of Sociology
and Human Geography, University of Oslo
Peirce, J.J., Weiner, R.F. and Veslind, P.A. (1998), Environmental Pollution and Control, ButterworthHeinemann
Porter, g. (1999), “Trade Competition and Pollution Standards: ‘Race to the Bottom’ or ‘Stuck at the
Bottom’?”, Journal of Environment and Development, 8(2), pp. 133-151
Porter, M.E. and van der Linde, C. (1995), ”Toward a New Conception of the EnvironmentCompetitiveness Relationship”, Journal of Economic Perspectives, 9(4), pp. 97-118
Revesz, R.L. (1992), “Rehabilitating Interstate Competition: Rethinking the ‘Race-to-the-Bottom’
Rationale for Federal Environmental Regulation”, New York University Law Review 67(6), pp. 12101254
Smarzynska, B.K. and Wei, S-J. (2001), “Pollution Havens and Foreign Direct Investment: Dirty Secret
or Popular Myth?”, World Bank Working Papers, Paper No. 2673, The World Bank, Washington
D.C.
Sterner, T. (1996), “Competitiveness, Trade and Environment: Revealed Comparative Advantage in
Chemical Products”, unpublished working paper, Department of Environmental Economics,
Göteborg University
Stevens, C. (1993), “Synthesis Report: Environmental Policies and Industrial Competitiveness”, in
Environmental Policies and Industrial Competitiveness, OECD Documents, OECD, Paris
Streets, D.G., Tsai, N.Y., Akimoto, H. and Oka, K (2000), “Sulfur Dioxide Emissions in Asia in the
Period 1985-1997”, Atmospheric Environment, 34(26), pp. 4413-4424
Tobey, J., (1990) “The Effects of Domestic Environmental Policies on Patterns of World Trade:
an
Empirical Test.” Kyklos, 43(2), pp. 191-209
Trefler, D. (1993), ”International Factor Price Differences: Leontief Was Right!”, The Journal of Political
Economy, 101(6), pp.961-987
Ugelow, J. (1982), “A Survey of Recent Studies on Costs of Pollution Control and the Effects on Trade.”
in S. Rubin (ed.) Environment and Trade, New Jersey, Allanheld, Osmun and Co.
UNDP (2000), World Energy assessment: energy and the challenge of sustainability, New York
Vogel, D. (2001), “Environmental Regulation and Economic Integration” in Esty, D.C. and Geradin, D.
(eds.), Regulatory competition and economic integration: comparative perspectives , International
economic law series, Oxford University Press, Oxford
32
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Wilson, J.S., Tsunehiro, O. and Sewadeh, M. (2002), “Dirty Exports and Environmental regulation: Do
Standards Matter to Trade?”, World Bank Working Papers, Paper No. 2806, The World Bank,
Washington D.C.
Worldbank (2001), Thailand Environment Monitor 2000, World Bank Office, Bangkok
Xing, Y. and Kolstad, C.D. (2002), “Do Lax Environmental Regulations Attract Foreign
Investment?”, Environmental an Resource Economics, 21, pp. 1-22
Appendix A: SITC-classifications, Rev. 3
51 - Organic chemicals
511 - Hydrocarbons, n.e.s., and their halogenated, sulphonated, nitrated or nitrosated derivatives
512 - Alcohols, phenols, phenol-alcohols, and their halogenated, sulphonated, nitrated or nitrosated
derivatives
513 - Carboxylic acids and their anhydrides, halides, peroxides and peroxyacids; their halogenated,
sulphonated, nitrated or nitrosated derivatives
514 - Nitrogen-function compounds
515 - Organo-inorganic compounds, heterocyclic compounds, nucleic acids and their salts, and
sulphonamides
516 - Other organic chemicals
52 - Inorganic chemicals
522 - Inorganic chemical elements, oxides and halogen salts
523 - Metal salts and peroxysalts, of inorganic acids
524 - Other inorganic chemicals; organic and inorganic compounds of precious metals
53 - Dyeing, tanning and colouring materials
531 - Synthetic organic colouring matter and colour lakes, and preparations based thereon
532 - Dyeing and tanning extracts, and synthetic tanning materials
533 - Pigments, paints, varnishes and related materials
54 - Medicinal and pharmaceutical products
541 - Medicinal and pharmaceutical products, other than medicaments of group 542
55 - Essential oils and resinoids and perfume materials; toilet, polishing and cleansing preparations
551 - Essential oils, perfume and flavour materials
553 - Perfumery, cosmetic or toilet preparations (excluding soaps)
554 - Soap, cleansing and polishing preparations
56 - Fertilizers (other than those of group 272)
562 - Fertilizers (other than those of group 272)
57 - Plastics in primary forms
572 - Polymers of styrene, in primary forms
58 - Plastics in non-primary forms
582 - Plates, sheets, film, foil and strip, of plastics
583 - Monofilament of which any cross-sectional dimension exceeds 1 mm, rods, sticks and profile
shapes, whether or not surface-worked but not otherwise worked, of plastics
59 - Chemical materials and products, n.e.s.
591 - Insecticides, rodenticides, fungicides, herbicides, anti-sprouting products and plant-growth
regulators, disinfectants and similar products, put up in forms or packings for retail sale or as
preparations or articles (e.g., sulphur-treated bands, wicks and candles, and fly-papers)
592 - Starches, inulin and wheat gluten; albuminoidal substances; glues
598 - Miscellaneous chemical products, n.e.s.
33
M. Persson
Industrial Migration in the Chemical Sector
January, 2003
Appendix B: Country List and the Measures of Stringency
of Environmental Regulations
This table includes all countries for which data was used in the regression analysis’s and their
stringency of environmental regulations, measured in the manner described in section 4.2 (i.e. lower
values means more stringent regulations, higher values laxer).
Country
E(SO2 ) E(BOD)
Country
E(SO2 ) E(BOD)
Country
E(SO2 ) E(BOD)
Austria
0.11
0.21
USA
0.32
0.16
Korea
0.41
0.25
Belgium
0.21
0.19
Bulgaria
1.16
2.29
Malaysia
0.33
0.74
Canada
0.33
0.20
Hungary
0.99
0.89
Mexico
0.43
0.57
Denmark
0.24
0.24
Argentina
0.24
0.26
Morocco
0.76
0.62
Finland
0.30
0.22
Bangladesh
0.10
3.34
Pakistan
1.01
0.85
Germany
0.26
0.44
Chile
0.31
0.37
Panama
0.22
0.34
Greece
0.21
0.27
China
0.54
4.39
Paraguay
0.17
1.30
Ireland
0.61
0.18
Colombia
0.28
0.90
Peru
0.37
0.39
Italy
0.32
0.23
Costa Rica
0.35
1.26
Philippines
0.58
0.62
Japan
0.06
0.13
Egypt
0.49
1.32
Singapore
0.19
0.37
Netherlands
0.08
0.16
Ethiopia
0.25
4.15
South Africa
0.30
0.58
New Zealand
0.16
0.33
Guatemala
1.28
0.68
Sri Lanka
0.45
4.23
Norway
0.03
0.14
Hong Kong
0.23
0.59
Thailand
0.68
0.30
Portugal
0.47
0.52
India
0.46
1.36
Tunisia
0.46
0.25
Spain
0.60
0.21
Indonesia
0.25
1.99
Turkey
0.26
0.22
Sweden
0.08
0.16
Israel
0.37
0.28
Uruguay
0.61
0.54
Switzerland
0.06
0.19
Jordan
0.51
0.51
Venezuela
0.33
0.46
United Kingdom
0.52
0.20
Kenya
0.31
0.41
Zimbabwe
0.50
1.29
34