international trade and environmental regulation: time ... · international trade and environmental...

Environmental and Resource Economics17: 233–257, 2000.© 2000Kluwer Academic Publishers. Printed in the Netherlands.

233

International Trade and Environmental Regulation:Time Series Evidence and Cross Section Test

XINPENG XUDepartment of Business Studies, Hong Kong Polytechnic University, Hong Kong and AustraliaJapan Research Centre, The Australian National University (E-mail: [email protected])

Accepted 30 September 1999

Abstract. This paper examines empirically whether more stringent domestic environmental policiesreduce the international competitiveness of environmentally sensitive goods (ESGs). Our time seriesevidence indicates that there are no systematic changes in trade patterns of ESGs in the lastthree decades, despite the introduction of more stringent environmental regulations in most of thedeveloped countries in the 1970s and 1980s. This observed phenomenon is then subjected to amulti-country econometric test using an extended gravity-equation framework. The test suggeststhat, overall, more stringent environmental regulations do not reduce total exports, exports of ESGsand exports of non-resource-based ESGs. Neither was there any evidence to support the hypothesisthat new trade barriers emerge to offset the effects of more stringent environmental regulations.

Key words: environmental regulation, international competitiveness, trade and the environment

JEL classification: F14, Q28, L50

1. Introduction

The intellectual history of trade and the environment has evolved in two waves(Levinson 1996). The first wave of research peaked in the late 1970s and seemsto have been inspired by the introduction of stringent environmental regulationsin developed countries from the early 1970s. The second wave occurred in the1990s, mainly motivated by the debate over international trade agreements such asthe North American Free Trade Agreement (NAFTA) and the Uruguay Round ofthe General Agreement on Tariffs and Trade (GATT). The current debate involvessignificant issues relating to environmental protection, export competitiveness,industry migration and the use of environmental regulations such as non-tariff tradebarriers including eco-labeling (Bhagwati and Hudec 1996; Dean 1992). Althoughthe policy debates relating to NAFTA and GATT may have subsided, fundamentallong-term issues of trade and the environment remain.

Concerns about the effects of domestic environmental policies on trade havebeen expressed in the following ways. The first is the so-called ‘race to the bottom’effect. Will trade with foreign countries with lower environmental standards forcea domestic country to lower its own standards as a result of political pressures

234 XINPENG XU

brought to bear on governments to ensure the survival of domestic industries? Willthere be a tendency towards a ‘race to the bottom’ when trade among countries withdifferent environmental standards is liberalised? This concern mainly emanatesfrom countries with higher environmental standards.

The second concern is the so-called ‘pollution haven’ hypothesis (Walter andUgelow 1979). If free trade occurs between countries with different environmentalstandards, will countries with lower environmental standards tend, over time, todevelop a comparative advantage in environmentally sensitive industries with theresult that ‘havens’ for the world’s dirty industries emerge?1 (Cropper and Oates1992). The flip side of this concern is the fear that capital, and associated jobs, willmove out of countries with high standards, a tendency termed ‘industry flight’.

The third concern centres on export competitiveness. This is the debate overwhether increasingly stringent domestic environmental regulations will reduce theinternational competitiveness of environmentally sensitive industries. The ‘exportcompetitiveness’ line receives considerable attention whenever countries are in theprocess of passing new pollution control measures. This is not just an anxietyexpressed by developed countries where environmental regulations are supposedto be more stringent, but is also an issue of importance in developing countries, andone that affects significantly the development strategies of developing countries.

Fourth, there are concerns about unfair trade, ‘eco-dumping’ and calls for ‘levelplaying fields’ and harmonisation of environmental standards across countries.Trade with countries with lower environmental standards is regarded as ‘unfairtrade’ because of the absence of a ‘level playing field’. Such fears prompt calls forharmonisation of environmental standards across countries.2

Fifth and finally, there are concerns expressed by developing countries, about a‘new protectionism’ that uses trade measures to achieve environmental goals suchas import bans,3 and uses environmental measures to achieve economic goals suchas eco-labeling. Under this scenario, developing countries seeking to gain accessto developed countries’ markets view higher product or process standards, counter-vailing duties, or import restrictions on goods and services produced in countrieswith lower environmental standards as a new form of protectionism.

At the heart of all these concerns is the impact of environmental standards onindustrial competitiveness. If the impact of environmental standards on industrialcompetitiveness is negligible, there will be no need to fear a ‘race to the bottom’.If this impact is trivial, ‘dirty industries’ will not migrate to locations with lowerenvironmental standards and the ‘pollution haven’ hypothesis would be proved tobe groundless. If it is negligible, there would be no basis for arguments about‘level playing fields’ and calls for harmonisation of international environmentalstandards.

This paper seeks to examine this issue from two perspectives: time seriesevidence and cross section test. We seek first to explore time series evidence todetermine whether the pattern of export performance of environmentally sensi-tive goods (ESGs) has undergone systematical changes due to the introduction

INTERNATIONAL TRADE AND ENVIRONMENTAL REGULATION 235

of stringent environmental standards in most developed countries in the 1970sand the 1980s. The observed phenomenon is then subjected to a multi-countryeconometric test using an extended gravity-equation framework. We test thefollowing two hypotheses: first, do more stringent environmental regulations lowertotal exports, and/or ESGs, and/or non-resource-based ESGs exports? Next, donew trade barriers emerge to offset the effects of more stringent environmentalregulations?

The rest of the paper is organized as follows. The following section provides abrief review of the received empirical and theoretical literature. Section 3 providestime series evidence of the changing trade pattern of ESGs. Section 4 implements across section test using an extended gravity equation. We provide a brief discussionon theoretical issues in Section 5. The final section concludes.

2. Literature

Studies examining directly the effects of changing abatement costs on trade flowsof ESGs include Walter (1973), Robison (1988), Kalt (1988), Leonard (1988),Tobey (1990), Grossman and Krueger (1993), Van Beers and van den Bergh (1997),among others. Low and Yeats (1992) on the other hand, examine the changing tradepattern of ESGs directly.

Walter (1973) looks into the pollution content of US trade using input–output analysis. In that study, direct and overall environmental control chargesfor 83 goods and services categories contained in the 1966 US input–output table(including those from intermediate goods) are calculated. The pollution content ofUS exports is found to be 1.75 per cent of total exports while the pollution contentof US imports is found to be 1.52 per cent of total imports. Walter considers thisdifference to be insignificant, and concludes that ESGs are trade-neutral at best andmarginally biased against US export industries at worst.

Robison (1988) follows the line of Walter (1973) using 1973 and 1977 input–output tables. This discrete time series result indicates that the ratio of theabatement content of US imports to US exports has risen from 1.151 in 1973, to1.167 in 1977 and 1.389 in 1982. On the basis of this result, Robison concludesthat there is some evidence that US pollution control programs have changed UScomparative advantage such that more high-abatement-cost goods will be importedand more low-abatement-cost goods exported.

Using a conventional Heckscher–Ohlin framework at a disaggregated industrylevel in the United States, Kalt (1988) regresses net exports of theith industryon that industry’s use of the national endowment of factors of production. Kalt(1988) shows that domestic environmental regulation appears to have a negativeeffect on industries’ trade performance. By contrast, Leonard (1988) found littleevidence that pollution control measures have exerted a systematic effect on inter-national trade and investment by conducting a large case study of trade and foreigninvestment flows for several key industries and countries.

236 XINPENG XU

Tobey (1990) sets up a Heckscher–Ohlin–Vanek (HOV) multi-factor, multi-commodity model. Using 1975 data for 23 countries, Tobey regresses the netexports of five different industries that are classified as pollution intensive on thestocks of productive factors including the environment. He finds no evidence thatthe introduction of environmental control measures has caused trade patterns todeviate from the HOV predictions.

Grossman and Krueger (1993) investigate empirically the environmentalimpacts of NAFTA. They regress 1987 US imports from Mexico (relative to totalUS shipments) in 135 industries on factor shares which reflect the factor intensityof each industry. Environmental intensity is approximated by the ratio of pollutionabatement costs to total value added in that US industry. Grossman and Kruegerfind that the traditional determinants of trade and investment patterns are signi-ficant, but that the alleged competitive advantages created by lax pollution controlsin Mexico play no substantial role in motivating trade and investment flows.

A recent study by Van Beers and van den Bergh (1997) applies a gravity modelto examine the impact of relatively strict environmental regulations on a country’sexports and imports. Their cross-section study for 1992 shows that the broadlydefined environmental stringency variable does not exert significant effects onbilateral trade flows while the narrow one (more directly linked to the PolluterPays Principle) reveals a significant negative impact on exports. In the case of tradeflows of pollution-intensive goods used as an independent variable, no effect of arelatively stringent environmental policy on exports of pollution-intensive goodsis found. They distinguish between resource-based and non-resource-based (orfootloose) bilateral trade flows, and find that a stringent environmental policy doeshave a significant negative impact on non-resource-based exports.

In general, one shortcoming of the existing literature is that the changing patternof export performance of ESGs over time is seldom explored and the characteristicsof dynamic comparative advantage are missing. This leads to an incomplete pictureof the impact of environmental standards on industrial competitiveness. Low andYeats (1992) first explored this issue by analysing sectoral changes that occurredin developed and developing countries’ actual trade and revealed comparativeadvantage in heavily polluting industries. Their result indicates that developingcountries gained a comparative advantage in pollution-intensive products at agreater rate than developed countries. The limitation of this study is that it focuseson one particular industry (Iron and Steel Pipes and Tubes, SITC 678) and, whenconsidering the ESG groups, it only looks at the overall performance of two groupsof countries, namely developed countries and developing countries. The study alsoonly looks at the beginning (late 1960s) and end years (late 1980s). All this mightresult in an incomplete picture of the changing pattern of export performance ofESGs over time.

As Levinson (1996: 450) concludes, ‘[t]he literature surveyed is almost unan-imous in its conclusion that environmental regulations have not affected inter-jurisdictional trade or the location decisions of manufacturers. Where studies have


found statistically significant effects of these regulations, the effects are alwaysquite small.’

However, the empirical evidence on the dynamic effects of environmental regu-lations offers a different picture. Lanjouw and Mody (1993) analyse the impactof stringent environmental regulations on the patenting of environmental techno-logies, using international data on expenditure for compliance with environmentalregulation and environmental patents. Their results indicate that increases in envi-ronmental compliance costs lead to increases in the patenting of new environmentaltechnologies, after a one- to two-year lag. Jaffe and Palmer (1997) analyse the sameissue in an econometric study using data on pollution control expenditure (PACE),value added data and R&D by industry over time (1977–1991). Although they findno statistically significant relationships between regulatory compliance expendi-tures and patenting activity, they do find that there exists a significant positiverelationship between regulatory compliance expenditures and R&D expendituresby the regulated industry, which is broadly consistent with the findings of Lanjouwand Mody (1993). A recent study by Berman and Bui (1998), using firm leveldata, shows that more stringent air quality regulation results in an increase inproductivity levels in petroleum refining in the United States. They conclude thatenvironmental regulations are productivity enhancing.

The theoretical literature on trade effects of domestic environmental regula-tions generally follows one of four approaches: (1) Partial equilibrium approach:static, without innovation (see, for example, Anderson 1992). One advantage ofthis modeling strategy is that the welfare effect of environmental and trade policiescan easily be analysed. Although most partial equilibrium models on trade andenvironment focus on the effects on the environment of trade and/or environ-mental policy changes, trade effects of environmental policy can only be drawnindirectly. (2) Partial equilibrium approach: static, with innovation. Comparativestatic partial equilibrium analysis sheds considerable light on the welfare effects oftrade and/or environmental policy and provides some insights into the trade effectsof domestic environmental policy. Notwithstanding, it is less helpful in under-standing the environment–competitiveness debate, notably the ‘new paradigm’suggested by Porter and van der Linde (1995) in which innovation plays a crucialrole. According to Porter and van der Linde (1995: 97–98), ‘[t]he new paradigmof international competitiveness is a dynamic one, based on innovation.’ In orderto capture the effect of innovation, Palmer, Oates and Portney (1995), drawingon some of the earlier literature on innovation in abatement technology, present amodel in which even incentive-based environmental regulation results in reducedprofits for the regulated firm. They demonstrate that abatement cost is simplyunproductive. (3) A large body of theoretical models developed along the linesof the Heckscher–Ohlin (HO) theorem. Since the 1970s most of the studies haveanalysed the theoretical impact of domestic environmental policy on trade patternsand predict that a country will export ESGs if it has less stringent environmentalregulations and hence a relatively abundant environmental factor service. See, for

238 XINPENG XU

example, Siebert, Eichberger, Gronych and Pethig (1980) and Baumol and Oates(1988). With recent papers by Copeland and Taylor (1994) using a demand andsupply general equilibrium device, and Chichilnisky (1994) looking at North–South trade from the perspective of property rights, their results are also mainlyin the spirit of H–O theorem. The common theme behind both partial and generalH–O theorems is that an increase in the stringency of environmental regulation willlead to a loss of competitiveness in ESGs. This result lends support to concerns inaffected industries but in general fails to explain the existing empirical evidence.(4) Strategic trade models include Markusen (1997), Barret (1994), Ulph and Ulph(1996) and Rauscher (1994). These studies consider related questions of strategyand provide various explanations of trade effects of environmental policy. Bartik(1988) and Motta and Thisse (1994) consider effects of environmental regulationson location choice.

3. Time Series Evidence

To provide time series evidence of the changing trade pattern of ESGs acrosscountries over time, we use a comprehensive dataset of annual trade flows of ESGsdisaggregated at the four-digit level of the SITC from 1965 to 1995 for 34 reportercountries.4 These 34 reporter countries accounted for nearly 80 per cent of worldexports (and trade) of ESGs in 1995. They include 25 of the 29 OECD countries5

as of May 1997, and major East Asian developing economies. There are 134 ESGcommodities at the four-digit level and 286,905 observations in total.

As is well known and discussed by Gagnon and Rose (1995),6 the value ofinternational trade flows has increased substantially in the last 40 years. This ispartly a result of inflation, partly a result of real economic growth and partly aresult of the increasing importance of trade relative to total output. In particular, amacroeconomic imbalance may result in substantial changes in net exports.

To abstract these effects from our data, the export revealed comparativeadvantage (XRCA) index is used. This XRCA index, introduced by Balassa(1989) in 1965, is defined as a country’s share in the exports of a particularcommodity divided by the share of that particular commodity in the world exportsof manufactured goods, as follows

XRCAki =(Xkiw

Xliw

)÷(Xkww

Xlww

)(1)

whereXRCAki gives countryi’s export revealed comparative advantage in industryk, X stands for exports, subscriptw stands for world, subscriptk represents industryk, and subscriptl stands for total exports. This XRCA index works reasonably wellin terms of the above-mentioned data issue. Since it is an index, the inflation effectcan be removed if it is an across-the-board increase in the prices of all commodities.By dividing exports of a particular commodity category by total manufactured


exports, this index also takes into account macroeconomic trade balance effects.For instance, a 1 per cent growth in exports spread uniformly across all goods (forexample, when domestic savings are greater than domestic investment) will notaffect the level of this index. Furthermore, by dividing a country’s export sectoralshare of a particular commodity category by the same sectoral share in the worldexports of manufactured goods, a general increase or decrease in world exports of aparticular commodity (growth effect) will not change the level of this index either.

For reasons that will shortly become clear, a normalization is used for thecommodity trade share. This measures the relative importance of a particularcommodity trade share in the trade of total ESGs at a particular point in time,as follows:

Sit = 1

2·(Xit

Xet+ Mit

Met

)· 100 (2)

whereX andM refer to exports and imports respectively,i refers to a particularcommodity category within ESGs,t refers to a point in time ande indicates totalESGs. The sum of any time period over all ESGs is 100, andSit is a percentagemeasure.

This dataset is analysed from the following three perspectives. We first comparepercentage changes of the export performance of ESGs in 1995 with that in 1965for each of the reporter countries. Table I shows the weighted version of the break-down of dichotomous measures of the XRCA index between the beginning (1965)and the end of the period (1995). These tables are reported using the followingmatrix:

1995 N 1995 S Total

1965 N A11 A12 A1

1965 S A21 A22 A2

Total A3 A4 100

Notes: A1 and A3 are percentages of normalised trade flows of ESGs that were not specialized in1965 and 1995 respectively. A2 and A4 are percentages of normalised trade flows of ESGs thatwere specialized in 1965 and 1995 respectively. A22 and A11 are percentages of normalised tradeflows of ESGs that were and were not specialized both in 1965 and in 1995, respectively. A21refers to percentage of normalised trade flows of ESGs that were specialized in 1965 but becamenon-specialized in 1995. A12 refers to percentage of normalised trade flows of ESGs that were notspecialized in 1965 but became specialized in 1995. ‘N’ stands for ‘non-specialization’ with theXRCA index less than one while ‘S’ refers to ‘specialization’ with the XRCA index greater thanone. It is a dichotomous measure in the sense that each commodity at a particular point in timeis either in the position of ‘S’ or ‘N’. ‘1965 N’ therefore represents those commodities that didnot have ‘revealed comparative advantages’ in 1965 while ‘1965 S’ represents those commoditiesthat had ‘revealed comparative advantages’ in 1965. The same logic applies to both ‘1995 N’ and‘1995 S’.

240 XINPENG XU

Since this is the weighted version (using commodity shares in the ESG group in1990 as the weight) of theXRCAdichotomy, the number in the tables represents thepercentage of trade flows rather than the percentage of the number of commodities.These trade flow percentages of ESGs should sum to 100 at any given point intime. The fourth column of each of the two-way tables is the breakdown of the1965 ESGs trade flows while the fourth row of each two-way table is the 1995breakdown.

For example, in the case of Australia, 49.2 per cent of the normalized tradeflows of ESGs were in a position of ‘non-specialization’ while 50.8 per cent of thenormalized trade flows of ESGs were in a position of ‘specialization’ in 1965 (seeTable I). Among the 49.2 per cent of the normalized trade flows of ESGs whichwere in a position of ‘non-specialization’ in 1965, 38.1 per cent remained in aposition of ‘non-specialization’ in 1995 while 11.1 per cent switched to a positionof ‘specialization’. The same logic applies to the column-wise explanation.

If stringent environmental standards reduce the international competitivenessof ESGs in a significant way, one would expect that the export performance ofthe ESGs for developing countries would increase while that for developed coun-tries would decrease. However, one striking feature revealed in Table I is thattrade volumes that move from a ‘specialization’ position to a ‘non-specialization’position account for no more than 15 per cent of the ESGs trade volumes for themajority countries.

Further, if taking into account those trade volumes that move from a posi-tion of ‘non-specialization’ to a position of ‘specialization’, one can see that theyalways exceed trade volumes that move from a ‘specialization’ position to a ‘non-specialization’ position with the exceptions of Japan and Norway.7 Even in thecase of Japan and Norway, this difference is very small, 9.57 per cent and 5.95per respectively. It becomes clear that the pattern of export performance of ESGsis quite persistent in the sample period. Those commodities which did not displaymuch ‘revealed comparative advantage’ at the beginning of the sample period tendto remain in a position of ‘non-specialization’ while those commodities which didhave a ‘revealed comparative advantage’ at the beginning of the sample periodremain in a position of ‘specialization’.

A second, and more rigorous statistical test of the association, Kendall’s tau-btest statistic, also conveys the economic message that there is a strong associationbetween the export performance of ESGs between 1965 and 1995. Thep-valueshows that the null hypothesis that the two series are distributed independently canbe rejected at a significant level of 1 per cent for most of the countries.

Finally, to look at the export performance of ESGs in the intervening years,a time series pattern of export performance of ESGs is calculated. We use as acompetitiveness indicator the percentage trade share of those ESGs that indicateda ‘specialization’ in total ESGs trade for each year and each country. Since adichotomous measure can be assigned to each commodity at a particular point intime, the normalized trade share of those commodities (within the ESGs group)


Table I. Breakdown of two way tables: selected countries

Australia Austria

1995 N 1995 S Total 1995 N 1995 S Total

1965 N 38.1 11.1 49.2 1965 N 20.0 17.1 37.1

1965 S 10.1 40.7 50.8 1965 S 12.2 50.7 63.0

Total 48.2 51.9 100 Total 32.2 67.8 100

Kendall’s tau-b: 0.27 Kendall’s tau-b: 0.34

p-value: 0.0001 No. of ESGs: 133 p-value: 0.0001 No. of ESGs: 133

Canada Switzerland

1995 N 1995 S Total 1995 N 1995 S Total

1965 N 31.2 23.3 54.5 1965 N 37.9 21.5 59.3

1965 S 5.8 39.8 45.6 1965 S 0.5 40.2 40.7

Total 37.0 63.1 100 Total 38.4 61.6 100


p-value: 0.0001 No. of ESGs: 125 p-value: 0.0001 No. of ESGs:133

Denmark Finland

1995 N 1995 S Total 1995 N 1995 S Total

1965 N 38.6 19.4 58.0 1965 N 31.2 20.6 51.8

1965 S 7.5 34.5 42.0 1965 S 2.7 45.6 48.2

Total 46.1 53.9 100 Total 33.8 66.2 100



Japan USA

1995 N 1995 S Total 1995 N 1995 S Total

1965 N 52.4 8.1 60.5 1965 N 41.4 14.2 55.6

1965 S 17.7 21.8 39.5 1965 S 14.3 30.1 44.4

Total 70.1 29.9 100 Total 55.8 44.2 100



Netherlands Norway

1995 N 1995 S Total 1995 N 1995 S Total

1965 N 22.1 16.4 38.4 1965 N 26.5 15.3 41.8

1965 S 0.9 60.7 61.6 1965 S 21.2 37.0 58.2

Total 23.0 77.0 100 Total 47.7 52.3 100



Source: Author’s calculations on the basis of UN Commodity Trade database from InternationalEconomic Databank, The Australian National University, Canberra.

242 XINPENG XU

Source: Author’s calculations on the basis of UN Commodity Trade database fromInternational Economic Databank, The Australian National University, Canberra.

Figure 1. Time series pattern of the overall competitiveness of ESGs for selected countries.

is summed to provide a percentage share of the normalized trade of all ESGscommodities. This indicator therefore can be expressed as

Cit ={

k∑Skit | XRCAkit ≥ 1

}(3)

where Cit is the competitive indicator for countryi at timet. k denotes commodity(ESGs).Sit denotes shares of commodityk in countryi’s total trade at timet.

Figure 1 shows the time series pattern of this competitiveness indicator forselected countries that claim to have higher environmental standards. This simplefigure reveals a more striking result. The share of the normalized trade volume ofthose ESGs with an XRCA greater than one as to total ESGs trade did not decreaseover time for most of the countries, except Japan. Even in the case of Japan, therewas a stark increase in competitiveness of ESGs after the end of the 1980s.

4. Cross Section Test

The above analysis suggests that the export performance of ESGs is persistentthroughout the sample period despite the introduction of stringent environ-mental standards by the industrialised countries two decades ago. This observedphenomenon is now subjected to a multi-country test using an extended gravityequation model.


The empirical literature on testing the trade effects of domestic environmentalpolicy has mainly followed the conventional Heckscher–Ohlin approach, with netexports of disaggregated industries as the explanatory variable in a multilateraltrade flow context (for example, Kalt 1988; Tobey 1990). Van Beers and van denBergh (1997), however, argue that a disadvantage of this approach is that theeffects of differences in strict environmental regulations on trade flows betweencountries may cancel out as multilateral trade flows are an aggregate of bilateraltrade. Furthermore, changes in the stringency of environmental regulations mayhave different impacts on resource-based and non-resource-based (footloose) envi-ronmentally sensitive industries. Non-resource-based ESGs can respond to changesof environmental costs by migrating to countries with less stringent environmentalregulations (Markusen 1997). On the other hand, resource-based ESGs may notbe sensitive to alterations in the stringency of environmental regulations. Jaffe etal. (1995), for example, argue that natural resource endowments may partly orlargely explain the pattern of pollution-intensive exports. Therefore, testing theeffects on exports of non-resource-based environmentally sensitive goods may bea more sensible approach than testing total exports or total ESGs.

It is well known that importing countries’ trade tariffs exert negative effectson their trading partners’ export performance.8 It has also been argued that tradebarriers may emerge to offset the trade effects of more stringent environmentalregulations. For example, Leidy and Hoekman (1994) discuss the possibility thatnew trade barriers may have emerged to offset the effects of more stringent envi-ronmental regulations. It is therefore necessary to disentangle the trade effectsthat arise from reporter countries’ relatively stringent environmental regulationsfrom their own and their trading partners’ high import tariffs. We shall includeindicators that measure the degree of trade protection and test the effects explicitly.We therefore set out the hypothesis of this study as follows.

Hypothesis I Higher stringency of environmental regulations lowers total bilat-eral exports, bilateral exports of ESGs and bilateral exports ofnon-resource-based ESGs.

Hypothesis II New trade barriers emerge to offset the trade effects of morestringent environmental regulations.

Although Van Beers and van den Bergh (1997) employ a similar framework to testthe trade effects of environmental policy, this study differs from theirs in threemajor ways. First, we make use of a recently available, unique environmentalperformance index developed by a World Bank research team after two years’work based on 31 countries’ environmental reports presented to the United NationsConference on Environment and Development in 1992 by 145 countries. TheUNCED reports are similar in form as well as coverage, and permit cross-countrycomparisons. This is expected to minimise the possible measurement errors arisingfrom inter-country environmental regulations significantly.

244 XINPENG XU

Second, our sample avoids possible sample selection bias, which is a crucialissue in drawing inferences from the sample regression results. ‘The 31 UNCEDreport countries were selected randomly from the total of 145 by the Work Bankresearch team. These 31 countries range from highly industrialized to extremelypoor, they are drawn from every world region, and they range in size anddiversity from China to Jamaica’ (Dasgupta et al. 1995: 3). Van Beers and vanden Bergh’s (1997) sample consists of 21 OECD-countries only. Since the tradeeffects of domestic environmental policy are mainly a North–South rather than aNorth–North problem, there is an obvious sample selection bias problem in theirstudy.

Third, a technical limitation relating to their measurement of export flows isthat macro fluctuation is not properly taken into account. Choosing exports in aparticular year without removing the effects of cyclical and macro disturbancesmight bias the result significantly.

EMPIRICAL MODEL OF TRADE WITH ENVIRONMENTAL REGULATIONS

The gravity model was developed independently, by Dutch economists Tinbergen(1962) and his collaborator Linnemann (1966) and Finnish economists Pöyhönen(1963) and Pulliainen (1963). It has been used extensively in empirical studies ofinternational trade since then. As Anderson put it, gravity equation is ‘[p]robablythe most successful empirical trade device of the last twenty-five years’ and‘usually produces a good fit’ (1979: 106). The theoretical foundations of thegravity model can be found in Anderson (1979), Helpman and Krugman (1985)and Bergstrand (1985).

The basic log-linear version of gravity equation is given by

ln(Xij ) = α0+ β1ln(Yi)+ β2ln(Ni)+ β3ln(Yj )+ β4ln(Nj)+ (4)

β5ln(Dij )+ β6ln(ENVi)+ β7ln(ENVi)+ εijwhereXij is exports from countryi to country j; (Yi) and (Yj ) are GDP forexporting countryi and importing countryj; (Ni) and (Nj ) are populations of coun-triesi andj;Dij is the geographical distance between countriesi andj, respectively;(ENVi) and (ENVj ) are environmental stringency indices developed by the WorldBank and defined as trade ‘distortions’;α0 is a constant term that accounts forthe effects of unmeasured trade distortions on exports and we leave the error termεij to take care of all the possible measurement errors andεij is assumed to beindependently and identically distributed.

Since import tariffs are frequently argued to be either an ‘artificial’ obstacle totrade or new trade barriers that have emerged to offset the effects of more stringentenvironmental regulations, we extend the above model further to include variablesof import tariffs both for reporting and partner countries. Equation (4) therefore ismodified as


ln(Xij ) = α0+ β1ln(Yi)+ β2ln(Ni)+ β3ln(Yj )+ β4ln(Nj)+β5ln(Dij )+ β6ln(ENVi)+ β7ln(ENVi)+ β8ln(DTi)+ (5)

β9ln(DTj)+ εij

where (DTi) and (DTj ) are import tariffs of countriesi andj, respectively.Different versions of this gravity equation have been used in the empirical

studies. However, equation (4) is the most general version in the sense that itencompasses other more restricted gravity models. For example, Kalirajan andShand (1998) do not include population in their equation but instead use GDPper capita. This is equivalent to placing the restriction that the coefficients on GDPand population are of equal value but the opposite sign. Tamirisa (1998) uses GDPper capita rather than GDP in his gravity equation analysis. Mathematically, this isessentially equation (4) after simple reparameterisation.

DATA AND MEASUREMENT

Empirical implementation of the model described in the previous section requiresdata on bilateral total exports, bilateral exports of ESGs and bilateral exportsof non-resource-based ESGs, GDP, population, bilateral geographic distance,environmental stringency and import tariffs.

The major data constraint concerns the environmental stringency variable. Datafor environmental stringency are generally not available, especially comparablecross-sectional data. Fortunately, we are able to make use of a set of unique envi-ronmental stringency indices recently developed by the World Bank (see Dasguptaet al. 1995). This set of indices encompasses many aspects of environmental policyand therefore can be used as the ‘proxy variable’ for environmental stringency.Since the survey considers the state of policy and performance in four environ-mental dimensions, namely, Air, Water, Land and Living Resources, the resultingenvironmental stringency index is a composite index of these four environmentaldimension indices. The data belong to the family of index numbers, with the largernumber indicating high stringency of environmental policy.

Unfortunately, some countries do not engage in bilateral trade with other samplecountries. We therefore have to eliminate them from the sample. The resultingnumber of sample countries is 20. They are Bangladesh, Brazil, Bulgaria, China,Egypt, Finland, Germany, India, Ireland, Kenya, Korea, the Netherlands, Nigeria,Pakistan, the Philippines, South Africa, Switzerland, Thailand, Trinidad/Tobagoand Tunisia. The model is estimated using cross-sectional data in 1990 and there isa total of 361 observations.

Non-resource-based ESGs include ‘Iron and Steel’ (SITC 67), ‘Metal Manufac-tures, NES’ (SITC 69), ‘Cement’ (SITC 661) and ‘Agricultural Chemicals’ (SITC599). This definition is based on Low and Yeats (1992) and is discussed by VanBeers and van den Bergh (1997).

246 XINPENG XU

Data on bilateral exports are obtained from the United Nation’s COMTRADEdatabase. Since current bilateral export data are subject to various distortions suchas macroeconomic imbalance and inflation, we deflate export data in the period1988 to 1992 using an export price index for each country obtained from the WorldBank’sWorld Development Indicator(1997). We then try to remove possible macrodistortion effects by simply averaging bilateral exports from 1988 to 1992 to obtainthe ‘true’ bilateral export data in 1990. Bilateral trade data are in US$ ’000.

Data on real GDP and population are obtained from Penn-World Tables, Mark5.6, by Summers and Heston (1991). Real GDP is in US$ billion while populationis expressed in millions. Geographic distance is the direct-line distance betweenthe capital cities of countries. It is measured in kilometers.

Since measurement of the intensity of non-tariff barriers is challenging, andthe available measures are inadequate, we use explicit import tariffs to capturethe border distortions. Therefore, in this study, the effects of non-tariff barriers(other than environmental stringency) are not measured separately but accountedfor in the intercept (Tamirisa 1998 makes the same assumption). Import tariffs aremeasured as tariff revenue as a share of total imports and are obtained from theWorld Bank’sWorld Development Indicator(1997).

A list of summary statistics of the above variables, including mean, standarddeviation, minimum, maximum and number of observations, is given in Table II.Table III provides the correlation coefficient between variables. Most of the vari-ables are less correlated, with the exception of GDP and population. Highlycorrelated coefficients between GDP and population may not be a serious problemsince we can test the restriction that the coefficients of GDP and population are thesame but of opposite sign. If this restriction is valid, we have to use GDP per capitarather than GDP and population separately. Furthermore, what we are interested inis the sign and coefficient of the environmental regulation variable and those of thetariff variable, rather than those of GDP or population. Since these variables areless correlated, we can exclude the possibility of multicollinearity.

ESTIMATION RESULTS

The estimation of equation (4) in general will violate the statistical hypothesis ofhomoscedasticity since our sample countries differ greatly in terms of income andcountry size. The error terms associated with large countries might have variancesmuch larger than the error terms associated with smaller countries. This may alsobe the case with income. We conduct various heteroscedasticity tests9 and are ableto reject the statistical assumption of homoscedasticity. In the light of this, we firstmake use of the White (1980) heteroscedasticity-consistent covariance matrix andcalculate thet-value based on the corrected standard errors. The result is reportedin Table IV. We also implement the maximum likelihood estimation of equation (4)and (5) to correct possible ‘dependent variable heteroscedasticity’.10

INT

ER

NA

TIO

NA

LT

RA

DE

AN

DE

NV

IRO

NM

EN

TAL

RE

GU

LAT

ION

247

Table II. Summary statistics

Xij Xdij Xnij Yi ENVi DTi Dij DYj Yj Nj ENVj

Mean 348183.7 82945.9 31671.3 301.4 630.2 13.9 6530.5 14.0 290.6 141.6 618.0

Standard deviation 1676619.2 431684.4 158670.7 406.3 190.4 11.7 3682.6 11.4 398.1 289.8 192.3

Minimum 1.0 1.0 1.0 10.0 366.0 1.0 600.0 1.0 10.0 1.3 366.0

Maximum 19686350.0 5387058.0 2221034.0 1501.0 951.0 50.0 17520.0 50.0 1501.0 1134.0 951.0

Count 361 361 361 361 361 361 361 361 361 361 361

Source: Author’s calculations.

248X

INP

EN

GX

U

Table III. Correlations

Xij Xdij Xnij DTi Yi Ni ENVi Dij DTj Yj Nj ENVj

Xij 1.00

Xdij 0.98 1.00

Xnij 0.96 0.94 1.00

DTI −0.15 −0.15 −0.15 1.00

YI 0.12 0.11 0.16 0.21 1.00

NI −0.03 −0.03 −0.01 0.38 0.88 1.00

ENVI 0.23 0.22 0.22 −0.62 −0.06 −0.27 1.00

Dij −0.20 −0.18 −0.17 0.03 0.03 0.02 −0.20 1.00

DTj −0.15 −0.13 −0.12 −0.05 −0.01 −0.02 0.03 0.03 1.00

Yj 0.09 0.09 0.08 −0.01 −0.05 −0.05 0.00 0.01 0.21 1.00

Nj −0.03 −0.02 0.00 −0.02 −0.05 −0.05 0.01 0.00 0.37 0.88 1.00

ENVj 0.22 0.20 0.18 0.03 0.00 0.01 −0.05 −0.19 −0.60 −0.02 −0.25 1.00

Source: Author’s calculations.


Table IV. Estimation results of extended gravity model (Equation 4)

Trade flows ln(Xij ) ln(Xdij ) ln(Xnij )

ln(Yi ) 1.61 2.38 3.09

(6.32)** (8.54)** (10.58)**

ln(Ni ) −0.43 −1.13 −1.31

(−1.89)* (−4.47)** (−5.10)**

ln(Yj ) 1.00 1.11 1.16

(4.82)** (3.78)** (3.94)**

ln(Nj ) −0.09 −0.07 −0.38

(−0.45) (−0.26) (−1.47)

ln(Dij ) −0.90 −0.92 −0.81

(−7.86)** (−6.39)** (−4.66)**

ln(ENVi ) 2.32 2.86 2.60

(3.42)** (3.51)** (3.05)**

ln(ENVj )1 67.0 87.0 61.

(2.63)** (1.04) (0.70)

Constant −18.72 −21.0 −22.22

(−2.93)** (−2.77)** (−2.62)**

Log-likelihood −695.09 −789.23 −795.43

R2 adjusted 0.665 0.629 0.67

F-statistic 103.24 88.13 105.41

Number of observations 361 361 361

Notes: ** denotes significant at the 1% level; * denotes significant at the 5% level.Figures in parentheses aret-ratios.Source: Author’s calculations.

Table IV provides three versions of regressions for equation (4) withdependent variables using total bilateral exports (ln(Xij )), bilateral exports of ESGs((ln(Xdij )) and bilateral exports of non-resource-based ESGs (ln(Xnij )), respec-tively. The results suggest that all coefficients have the expected signs in all threeregressions. GDP in countryi indicating the potential export supply, and GDP incountryj indicating the potential import demand are both found to be positive, andhighly significant factors (at the 1 per cent level) in determining bilateral exportsfor the full sample. Parameters on population variables for the exporting countriesare negative and significant, indicating that the greater the population the higherthe domestic market–foreign market ratio. Coefficients of population variables for

250 XINPENG XU

importing countries are negative but not significant for all three regressions. Coef-ficients of distance variables are negative and statistically significant at the 1 percent level for all three equations, suggesting that the higher the transportation coststhe lower the bilateral exports.

The coefficients of ln(ENVi ) are all positive and statistically significant atleast at the 1 per cent level. This suggests that the hypothesis that more stringentenvironmental regulations lower total bilateral exports, exports of ESGs and non-resource-based ESGs exports can be rejected. There is no evidence to support thehypothesis that higher environmental regulations lower exports of non-resource-based ESGs, an argument that has been frequently put forward on the basis thatnon-resource-based ESGs are footloose industries and respond significantly tochanges in the stringency of environmental regulations.

To investigate whether new trade barriers emerge to offset the trade effects ofmore stringent environmental regulations and to disentangle the effects of envi-ronmental regulations and those of border distortions, we include import tariffvariables and carry out a regression on equation (5). The results, with White’s(1980) heteroscedasticity-correction, are reported in Table V. The overall coeffi-cient significant tests using F-statistics are significant, ranging from 71.84 to 90.51with (9, 351) degree freedom. Adjusted R-square values range from 0.64 to 0.69.Both GDP variables for exporting and importing countries and distance variableshave the expected sign and are at least statistically significant at the 1 per cent level.Even then, high environmental stringency in exporting countries is not associatedwith lower total bilateral exports, lower ESGs exports or lower non-resource-basedESGs exports.

We find that partner countries’ high import tariffs do reduce reporter countries’export performance in total bilateral exports, ESGs exports and non-resource-basedESGs’ exports, significant at least at the 1 per cent level. However, the hypothesisthat new trade barriers emerge to offset the trade effects of higher stringencyof environmental regulations cannot be accepted in the light of this test. Coeffi-cients of import tariffs (ln(DTi)) for exporting countries appear to be positive butstatistically insignificant.

Results from the dependent variable heteroscedasticity model using themaximum likelihood estimation technique reveal similar patterns. This suggeststhat our findings are rather robust.

We not only make use of the composite environmental stringency index in ourregressions, but also carry out separate tests on equation (4) and (5) using eachof the four dimensional environmental stringency variables, namely, Air, Water,Land and Living Resource indices. Results from these tests generally conform tothe above findings. This is not surprising since there is a high correlation betweenthese variables.


Table V. Estimation results of extended gravity model (Equation 5)

Trade flows ln(Xij ) ln(Xdij ) ln(Xnij )

ln(Yi ) 1.55 2.32 3.04

(5.88)** (8.01)** (10.33)**

ln(Ni ) −0.38 −1.08 −1.27

(−1.63) (−4.19)** (−4.96)**

ln(Yj ) 0.85 0.90 0.85

(4.07)** (3.05)** (2.98)**

ln(Nj ) 0.03 0.10 −0.13

(0.18) (0.37) (−0.53)

ln(Dij ) −0.91 −0.94 −0.83

(−8.29)** (−6.76)** (−5.07)**

ln(ENVi ) 1.98 2.48 2.28

(2.84)** (2.92)** (2.52)**

ln(ENVj ) 0.77 −0.30 −1.15

(1.14) (−0.35) (−1.27)

ln(DTi ) −0.17 −0.18 −0.15

(−1.39) (−1.21) (−0.88)

ln(DTj ) −0.46 −0.60 −0.91

(−3.57)** (−3.38)** (−5.22)**

Constant −8.81 −8.74 −5.68

(−1.30) (−1.05) (−0.60)

Log-likelihood −688.83 −783.14 −782.44

R2 adjusted 0.67 0.64 0.69

F-statistic 84.04 71.84 90.51

Number of observations 361 361 361

Notes: ** denotes significant at the 1% level; * denotes significant at the 5% level.Figures in parentheses aret-ratios.Source: Author’s calculations.

5. Theoretical Discussion

The limitation of those theoretical models following the HOV theorem, for exampleBaumol and Oates (1988), lies in the fact that it explains little about the trade effectsof domestic environmental policy. Existing empirical work, including the abovetime series evidence and cross section test, consistently rejects the hypothesis that

252 XINPENG XU

there is a significant negative impact of domestic environmental policy on trade.11

This is not surprising given the fact that, ‘[e]mpirically, the HOV theorem has beenrepeatedly rejected over the years and rightfully so: it performs horribly. Factorendowments correctly predict the direction of factor service trade about 50 percent of the time, a success rate that is matched by a coin toss’ (Trefler 1995: 1029).However, there are two ways that could incorporate the technology factor intothe conventional HOV model and these extensions improve the explanations ofchanging trade patterns of ESGs significantly.

First, the assumption of identical technology can be relaxed in the basic H–O world. Under Ricardo’s framework, international differences in technology arenot only allowed, but serve as the basis for explaining positions of compar-ative advantage. Empirically, researchers have consistently found relative labourproductivity to be a powerful explanatory variable for trade flows.12 Since the1960s trade theorists have been interested in modifying the HO model to allowcountries to differ in their prevailing states of technology. Jones (1970) providesa comprehensive theoretical account of Ricardian technology under the H–Oframework.

As suveryed by Falvey (1994), technology differences can be divided into twomajor categories, namely, product augmenting and factor augmenting. Product-augmenting technology differences imply that one country can produce a largeroutput from the same factor inputs in a particular sector (or sectors) and the tech-nology therefore acts very much like product price changes. Factor-augmentingtechnology differences imply that a factor (or factors) in one country is uniformlymore productive than the same factor in the other, independent of the sector inwhich the factor is employed and this type of technology acts very much like factorendowment changes. If technology differences are purely factor augmenting, thetrade pattern might be explained in terms of ‘effective’ factor endowments. But tothe extent that technology differences are product augmenting, there are now twopotential determinants of the pattern of trade. It is straightforward to apply thisto analyse the relationship between environmental regulation and trade. If thereis a factor-augmenting technology difference between the North and the South inwhich the North possesses advanced technology, this may lead to production andexport of more pollution-intensive goods in the North.

The second approach is attributed to Davis (1995). Davis develops a model thatallows for both Ricardian and Heckscher–Ohlin influences (termed the Heckscher–Ohlin–Ricardian model or HOR). The model assumes there are two countries,One and Two, two factors, K (capital) and L (labour), three goods, X1, X2 and Y.The first two goods are perfectly competitive, intra-industry goods and are capital-intensive relative to Y. Preferences are identical and homothetic. Technologies areassumed to be identical across countries in X2 and Y, with a small Hick-neutralproductivity difference in X1. Since there is an absolute technical advantage in theproduction of good X1, only the technology of country One (which is assumed to


have the superior technology in the production of X1) can be used in production ofthis good.

The results are in sharp contrast with the H–O model. First, the H–O modelsimply predicts that a country exports the good that uses intensively its rela-tively abundant factor of production. The HOR model predicts that there are twopossibilities. If country One is the labour-abundant country, but has an absoluteadvantage in the production of good X1, it produces the world supply of thatcapital-intensive good, so must export it. Even if One is the capital-abundantcountry, since its factor endowment net of this may make it the ‘labour-abundant’country in the residual, factor endowments are employed to produce the goods withcommon technologies. Labour-abundant country Two would then export capital-intensive good X2. Second, in strong contrast to the H–O model, HOR predicts thateven countries with identical endowment ratios may engage in substantial trade,namely, intra-industry trade.

Since a Ricardian element is brought into the HO framework, technologybecomes an important factor along with relative factor endowment in deter-mining the pattern of trade. The HOR model has received strong empiricalsupport recently. For example, Harrigan (1997), in an econometric study, findsthat relative productivity which is in the spirit of the Ricardian model, whencombined with factor endowments, significantly improves the explanations forchanging trade patterns. Trefler’s (1995: 1029) empirical test also reveals that‘HOV is rejected empirically in favor of a modification that allows for homebias in consumption and international technology differences’. This modifica-tion can also be easily extended to explain why there is a persistence of tradepatterns of ESGs. The reason is simply being that environment regulations areproductivity enhancing.13 Therefore, stringent environmental regulations do notreduce international competitiveness of ESGs.

6. Conclusion

In this study, I provide time series evidence over whether the pattern of exportperformance of ESGs has undergone systematic changes in the period between the1960s and the 1990s. The important empirical finding is that the export perform-ance of ESGs for most of the countries remained unchanged between the 1960s andthe 1990s despite the introduction of stringent environmental standards in most ofthe developed countries in the 1970s and the 1980s.

This observed phenomenon is further put into a multi-country econometric test,using an extended gravity equation model. We test two hypotheses explicitly. Thefirst is whether countries with higher stringency of environmental regulations lowertheir exports of ESGs and/or non-resource-based ESGs. The second is whether newtrade barriers emerge to offset the trade effects of more stringent environmentalpolicy. We are able to make use of a unique set of comparative environmentalstringency indices recently developed by the World Bank.

254 XINPENG XU

Our results reject the above two hypotheses, suggesting that countries with morestringent environmental regulations do not reduce their exports of ESGs and/ornon-resource-based ESGs, and that new trade barriers do not emerge to offset thetrade effects of more stringent environmental policy in any statistically significantway. Our findings are robust to alternative versions of heteroscedasticity correctionestimations and alternative environmental dimension indices such as Air, Water,Land and Living Resource indices also developed by the World Bank.

An implication of this study is that environmental regulations may beproductivity enhancing. Dynamic technological innovation may be more importantin determining the export performance of ESGs. A recent study by Berman andBui (1998), using firm level data, shows that more stringent air quality regulationresults in an increase in productivity levels in petroleum refining in the UnitedStates. They conclude that abatement investments are productive. Our results areconsistent with their findings and should help refocus the debate on the rela-tionship between environmental regulation and competitiveness in internationaltrade.

Acknowledgement

I wish to thank two anonymous referees and the Editor of this journal for theirconstructive comments and suggestions. All remaining errors are mine.

Notes

1. Dirty industries here refer to environmentally sensitive industries.2. For detailed discussions, see Verbruggen and Kuik (1997).3. The US ban on imports of Mexican tuna caught using purse seine nets, which also kill dolphins,

stands out as a case in point. Two GATT dispute resolution determinations concluded that theUS measure violated trade rules. For details, see Esty (1994).

4. The definition of ESGs follows Low and Yeats (1992).5. Except Hungary, Czech Republic, Turkey and Iceland.6. Gagnon and Rose (1995) used similar methodology to test product cycle theory.7. The United States is on the margin with 14.34 per cent to 14.16 per cent.8. See Linnemann and Verbruggen (1991).9. All results that are not reported in this paper are available on request.

10. See Endnote 9.11. As surveyed by Dean (1992), ‘there is little evidence of any significant impact of EEC

(Environment Control Cost) on the pattern of trade’.12. See Obstfeld & Rogoff (1996).13. For a detailed discussion, see Berman and Bui (1998).


References

Anderson, J. E. (1979), ‘A Theoretical Foundation for the Gravity Equation’,American EconomicReview69(1), 106–116.

Anderson, K. (1992), ‘The Standard Welfare Economics of Policies Affecting Trade and the Environ-ment’, in K. Anderson and R. Blackhurst, eds.,The Greening of World Trade Issues. Ann Arbor:University of Michigan Press.

Balassa, B. (1989), ‘Comparative Advantage, Trade Policy and Economic Development: Introduction– An Autobiographical Essay’, in B. Balassa, ed.,Comparative Advantage, Trade Policy andEconomic Development. New York: New York University Press.

Barrett, S. (1994), ‘Strategic Environmental Policy and International Trade’,Journal of PublicEconomics54(3), 325–338.

Bartik, T. J. (1988), ‘The Effects of Environmental Regulation on Business Location in the UnitedStates’,Growth and Change19(3), 22–44.

Baumol, W. J. and W. E. Oates (1988),The Theory of Environmental Policy. Cambridge: CambridgeUniversity Press.

Bergstrand, J. H. (1985), ‘The Gravity Equation in International Trade: Some MicroeconomicFoundations and Empirical Evidence’,Review of Economics and Statistics67(3), 474–481.

Berman, E. and T. M. Bui (1998), ‘Environmental Regulation and Productivity: Evidence from OilRefineries’,NBER Working Paper6776, 1–35.

Bhagwati, J. and R. E. Hudec (1996),Fair Trade and Harmonization: Prerequisites for Free Trade?Cambridge and London: MIT Press.

Chichilnisky, G. (1994), ‘North-South Trade and the Global Environment’,American EconomicReview84(4), 851–874.

Copeland, B. R. and M. S. Talyor (1994), ‘North-South Trade and the Environment’,QuarterlyJournal of Economics109(3), 755–787.

Cropper, M. L. and W. E. Oates (1992), ‘Environmental Economics: A Survey’,Journal of EconomicLiterature30(2), 675–740.

Dasgupta, S., A. Mody, S. Roy and D. Wheeler (1995), ‘Environmental Regulation and Development:A Cross-Country Empirical Analysis’,World Bank Policy Research working paperNo. 1448,1–24.

Davis, D. R. (1995), ‘Intra-industry Trade: A Heckscher–Ohlin–Ricardo Approach’,Journal ofInternational Economics39(3–4), 201–226.

Dean, J. M. (1992), ‘Trade and the Environment: A Survey of the Literature’, in P. Low, ed.,International Trade and the Environment. World Bank Discussion Papers, no. 159.

Esty, D. C. (1994),Greening the GATT: Trade, Environment, and the Future. Washington: Institutefor International Economics.

Falvey, R. E. (1994), ‘The Theory of International Trade’, in D. Greenaway and A. Winters eds.,Surveys in International Trade. Oxford and Cambridge, Massachussetts: B. Blackwell.

Gagnon, J. E. and A. K. Rose (1995), ‘Dynamic Persistence of Industry Trade Balances: HowPervasive Is It’,Oxford Economic Papers47(2), 229–248.

Grossman, G. M. and A. B. Krueger (1993), ‘Environmental Impacts of a North American FreeTrade Agreement’, in P. Garber, ed.,The U.S.–Mexico Free Trade Aggrement. Cambridge,Massachusetts, MIT Press.

Harrigan, J. (1997), ‘Technology, Factor Supplies, and International Specialization: Estimating theNeoclassical Model’,American Economic Review87(4), 475–494.

Helpman, E. and P. Krugman (1985),Market Structure and Foreign Trade. Cambridge, Massachu-setts: MIT Press.

Jaffe, A. B. and K. Palmer (1997), ‘Environmental Regulation and Innovation: A Panel Data Study’,Review of Economics and Statistics79(4), 610–619.

256 XINPENG XU

Jaffe, A. B., S. R. Peterson, P. R. Portney and R. N. Stavins (1995), ‘Environmental Regulationand the Competitiveness of U.S. Manufacturing: What Does the Evidence Tell Us?’,Journal ofEconomic Liternature33, 132–163.

Jones, R. (1970), ‘The Role of Technology in the Theory of International Trade’, in R. Vernon, ed.New York: NBER.

Kalirajan, K. P. and R. Shand (1998), ‘Economic Links Within the IOR-ARC: an EmpiricalAnalysis’, mimeo, The Australian National University, Canberra.

Kalt, J. P. (1988), ‘The Impact of Domestic Environmental Regulatory Policies on U.S. Interna-tional Competitiveness’, in M. A. Spence and H. A. Hazard, eds.,International Competitiveness.Cambridge, Massachussetts: Harper and Row.

Lanjouw, J. O. and A. Mody (1993), ‘Stimulating Innovation and the International Diffusion ofEnvironmentally Responsive Technology: The Role of Expenditures and Institutions’, mimeo,World Bank.

Leidy, M. P. and B. M. Hoekman (1994), ‘ “Cleaning Up” while Cleaning Up? Pollution Abatement,Interest Groups and Contingent Trade Policies’,Public Choice78(3–4), 241–258.

Leonard, H. J. (1988),Pollution and the Struggle for the World Product. Cambridge: CambridgeUniversity Press.

Levinson, A. (1996), ‘Environmental Regulations and Industry Location: International and DomesticEvidence’, in J. Bhagwati and R. E. Hudec, eds.,Fair Trade and Harmonization: Prerequisitesfor Free Trade?Cambridge, Massachussetts and London: MIT Press.

Linnemann, H. (1966),An Econometric Study of International Trade Flows. Netherlands: North-Holland Publishing Company.

Linnemann, H. and H. Verbruggen (1991), ‘GSTP Tariff Reduction and Its Effects on South-SouthTrade in Manufactures’,World-Development19(5), 539–551.

Low, P. and A. Yeats (1992), ‘Do “Dirty” Industries Migrate?’, in P. Low, ed.,International Tradeand the Environment. World Bank Discussion Papers, no. 159.

Markusen, J. R. (1997), ‘Costly Pollution Abatement, Competitiveness, and Plant LocationDecisions’,Resource and Energy Economics19, 299–320.

Motta, M. and J. F. Thisse (1994), ‘Does Environmental Dumping Lead to Delocation?’European-Economic-Review38(3–4), 563–576.

Obstfeld, M. and R. Kenneth (1996),Foundations of International Macroeconomics. Cambridge,Massachussetts and London: MIT Press.

Palmer, K., W. E. Oates and P. R. Portney (1995), ‘Tightening Environmental Standards: The Benefit-Cost or the No-Cost Paradigm?’,Journal of Economic Perspectives9(4), 119–132.

Porter, M. E. and C. van der Linde (1995), ‘Toward a New Conception of the Environment-Competitiveness Relationship’,Journal of Economic Perspectives9(4), 97–118.

Poyhonen, P. (1963), ‘A Tentative Model for the Volume of Trade between Countries’,Weltwirtschaftliches ArchivBand 90 (Heft 1), 93–99.

Pulliainen, K. (1963), ‘A World Trade Study: an Econometric Model of the Pattern of the CommodityFlows in International Trade in 1948–1960’,Ekonomiska Samfundets Tidskrif(2), 78–91.

Rauscher, M. (1994), ‘On Ecological Dumping’,Oxford Economic Papers46(5), 822–840.Robison, H. D. (1988), ‘Industrial Pollution Abatement: The Impact on Balance of Trade’,Canadian

Journal of Economics21(1), 187–199.Siebert, H., J. Eichberger, R. Gronych and R. Pethig (1980),Trade and Environment: A Theoretical

Enquiry. Amsterdam: Elsevier/North-Holland.Summers, R. and A. Heston, ‘The Penn World Table (Mark 5): An Expanded Set of International

Comparisons, 1950–1988’,Quarterly Journal of Economics106(9), 327–368.Tamirisa, N. T. (1998), ‘Exchange and Capital Controls as Barriers to Trade’,IMF Working Paper.Tinbergen, J. (1962),Shaping the World Economy: Suggestions for an International Economic

Policy. New York: The Twentieth Century Fund.


Tobey, J. A. (1990), ‘The Effects of Domestic Environmental Policies on Patterns of World Trade:An Empirical Test’,Kyklos43(2), 191–209.

Trefler, D. (1995), ‘The Case of the Missing Trade and Other Mysteries’,American Economic Review85(5), 1029–1046.

Ulph, A. and D. Ulph (1996), ‘Trade, Strategic Innovation and Strategic Environmental Policy – AGeneral Analysis’, in C. Carraro, Y. Katsoulacos and A. Xepapadeas, eds.,Environmental Policyand Market Structure. Dordrecht: Kluwer Academic.

Van Beers, C. and J. C. J. M. van den Bergh (1997), ‘An Empirical Multi-country Analysis of theImpact of Environmental Regulations on Foreign Trade Flows’,Kyklos50(1), 29–46.

Verbruggen, H. and O. Kuik (1997), ‘WTO Ministerial Conference in Singapore: EnvironmentalDiversity versus Harmonization’,Environmental-and-Resource-Economics10(4), 405–413.

Walter, I. (1973), ‘The Pollution Content of American Trade’,Western Economic Journal11(1),61–70.

Walter, I. and J. Ugelow (1979), ‘Environmental Policies in Developing Countries’,Ambio8(2, 3),102–109.

White, H. (1980), ‘A Heteroscedasticity-Consistent Covariance Matrix Estimator and a Direct Testfor Heteroscedasticity’,Econometrica48, 817–838.

international trade and environmental regulation: time ... · international trade and environmental...

Documents