Environmental regulation and MNEs location: Does CSR matter?

E C O L O G I C A L E C O N O M I C S 6 7 ( 2 0 0 8 ) 5 5 – 6 5

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ l oca te /eco l econ

ANALYSIS

Environmental regulation and MNEs location:Does CSR matter?

Lammertjan Dam, Bert Scholtens⁎

Department of Economics, University of Groningen, PO Box 800, 9700 AV Groningen, The Netherlands

A R T I C L E I N F O

⁎ Corresponding author. Tel.: +31 50 363 7064E-mail address: [email protected] (B

0921-8009/$ – see front matter © 2007 Elsevidoi:10.1016/j.ecolecon.2007.11.007

A B S T R A C T

Article history:Received 17 August 2007Received in revised form1 November 2007Accepted 12 November 2007Available online 20 February 2008

We investigate whether firms with relatively low environmental standards are more oftenlocated in countries that are poor, corrupt or have weak environmental regulations. We findnew empirical evidence in favor of the Pollution Haven Hypothesis, which states that MNEsare transferring their dirty operations to countries with weak environmental regulation. Ourfindings suggest that these are not necessarily the poorest or most corrupt countries. Weestablish that MNEs with strong social responsibility avoid locating their operations incountries with weak environmental regulation.

© 2007 Elsevier B.V. All rights reserved.
Keywords:Corporate social responsibilityRegulationMultinational EnterpriseInternationalizationPollution Haven Hypothesis
JEL classification:D210; F230; L290; M140; O390; Q560

1. Introduction

This paper addresses the relation between the responsibility ofthe firm, its locational choices and countries' environmentalregulation. Becchetti et al. (2005) show that large and inter-nationally operating firms are more responsive to demandsfrom stakeholders to take account of many non-financial issueswhenconducting their business. It appears that theyareurged tobehave in a more socially responsible manner. Definitions forcorporate social responsibility (CSR) do vary. However, Dahlsrud(in press) finds that the existing definitions to a large degree arecongruent. CSR generally refers to actions taken by firms withrespect to their employees, communities, and the environment,which go beyond what is legally required of a firm. Recent

.. Scholtens).

er B.V. All rights reserved

theories of CSR (Baron, 2001; McWilliams and Siegel, 2000, 2001;Bagnoli andWatts, 2003; Beltratti, 2005) assert that firms engagein “profit-maximizing” CSR. That is, companies are assumed tobe socially responsible because they anticipate a benefit fromthese actions. Examples of such benefits might include reputa-tionenhancement, theability to chargeapremiumprice for theiroutput, or the use of CSR to recruit and retain high qualityemployees. These benefits are presumed to offset the highercosts associated with CSR, since resources must be allocated toallow the firm to achieve CSR status. These theoretical studiesemphasize how this activity is likely to bematrixed into a firm'sstrategies. In line with this approach, we want to find outwhether any ‘responsible’ behavior can be found in theinternationalization patterns of firms. In this respect, our

.

mailto:[email protected]

http://dx.doi.org/10.1016/j.ecolecon.2007.11.007

56 E C O L O G I C A L E C O N O M I C S 6 7 ( 2 0 0 8 ) 5 5 – 6 5

research relates to the so-called Pollution Haven Hypothesis(PHH). The PHH states that due to stronger environmentalregulations indevelopedcountries, firms indirty sectorsmigratetowards poor countries with low environmental regulation(Mani and Wheeler, 1997). Various studies test the PHH andlink foreign direct investment to environmental regulation(Sorsa, 1994; Levinson, 1996; Janicke et al., 1997; List and Co,2000; List, 2001; Antweiler et al., 2001; Talkukdar and Meisner,2001; Cole and Elliott, 2003; Damania et al., 2003; Eskeland andHarrison, 2003; Cole et al., 2006). In linewith this typeof analysis,we want to find out how the social responsibility of theinternationalizing firm interactswith environmental regulation,governance, and wealth of the countries it invests in.

The Pollution Haven Hypothesis (PHH) states that as aresult of increased regulation, multinationals shift their dirtyproduction to countries with poor environmental standards.The economic rationale for the PHH is usually explained froma comparative advantage perspective: Countries with littleregulations put fewer restrictions on a firm's operations andhence reduce non-market costs. The empirical evidence forthe PHH is at best mixed. Some studies present evidence infavor of the PHH (Low and Yeats 1992; Xing and Kolstad 2002;Mani and Wheeler 1997). There are also arguments againstthe PHH, stating that due to an increase in “global eco-consciousness”, multinationals are induced to innovate incleaner production instead of migrating towards countrieswith poor environmental standards (Letchumanan andKodama, 2000). Other studies find no evidence for or againstthe PHH (Sorsa 1994, Repetto 1995). A major obstacle in thisliterature is that data on regulation is often lacking. There-fore, proxies for environmental regulation such as corruptionindices have been used to test the PHH (Smarzynska Javorcikand Wei, 2004). However, this complicates the interpretationof such findings. On the one hand, corruption levels indicatewhether laws are enforced, but on the other hand one needssome laws and regulations to facilitate corruption.

In the last few years, we have witnessed increased attentionfor CSR (Becchetti et al., 2005; Beltratti, 2005). As suggested byHeal (2005), CSR can be defined as the extent to which firmsinternalizenon-market costs. It appears that especially largeandinternationally operating firms are responsive to demands fromvarious stakeholders to take account of other than financialindicators to conducting their business. If Multinational Enter-prises apply their domestic standards in their overseas opera-tions, we increasingly may expect that the povertycharacteristics of a country will have less impact on firms'internationalization policies. Weak environmental regulationwill not be regarded as a comparative advantage from thesocially responsible firm's perspective. Then, the presumptionin this paper is that socially responsible firmswill be less likely tolocate in countries with lax environmental regulations. We arewell aware of the fact that the presence of a firm in a countrymay be the result of investment decisionsmade long ago, underdifferent regimes. The inertia would seem to blur the effects ofregulation on the probability of firms having presence in thosecountries. However, we also witness that there is inertia inregulation and, especially, in its enforcement. Therefore, wehope that our analysis about regulatory quality and firms' socialresponsibility is informative in connection with the internatio-nalization behavior of the firm.

Our study uses firm-level data by Ethical InvestmentResearch Services (EIRIS) on CSR and by AMADEUS onsubsidiary location of 540 large European MNEs. We consider44,149 subsidiaries located in 188 different countries. Apartfrom using the traditional proxies, such as corruption indicesor wealth measures, we also use more direct measures ofcountry environmental regulation from two sources pro-vided by the World Bank, namely its World BusinessEnvironment Survey (WBES) and the World DevelopmentIndicators (WDI). We estimate a binary choice model to testwhether firms that adopt a less stringent environmentalstandard are relatively more present in poor countries orcountries with weak environmental regulation. As such, byfocusing on the international location behavior of largemultinationals, we investigate the interactions among CSR,wealth and environmental quality. Themain question in thispaper is: Can PHH behavior be observed when we takeaccount of CSR? We find that firms with low socialresponsibility locate their operations more often in countrieswith weak environmental regulation. The PHH-literature alsofocuses on unobserved heterogeneity and endogeneity ofpollution regulations. For example, countries that receivelots of investment in polluting industries may impose strictregulations as a consequence. Countries that become richeras a consequence of investment may in turn levy stricterregulations. These causal relations are of importance inconnection with the PHH. But as our data about CSR are onlyavailable on a cross-section basis, we are unable to investi-gate them. Therefore, we provide preliminary evidence aboutthe interaction between corporate social responsibility andthe international location behavior of firms.

This paper adds to the existing literature in various ways.First of all, we test the PHH using actual location data ofMultinational Enterprises. By using firm-level data we adopt amore direct analysis compared to other studies that test thePHH. Furthermore, we use country-level data that are a betterindication of environmental regulation compared to thePollution Abatement and Control Expenditures surveys ofthe OECD or other proxies used in testing the PHH. We alsoshed some new light on whether poverty and poor environ-mental regulation are related. We find that firms that set highinternal environmental standards will not experience asupposedly comparative advantage when locating in coun-tries with low environmental regulation. On the other hand,firms with low environmental responsibility might haveincentives to engage in PHH behavior. Using firm-level dataand direct measures of a country's environmental regulation,we also find new evidence that for firms with low environ-mental standards PHH behavior can be observed. However, ifwe consider countries with high corruption or high poverty asbeing “havens” we do not find similar evidence. Thus, usingcorruption indices and income as proxies for the quality ofenvironmental regulation is not helpful as it leads to thewrong conclusions. The major finding of this study is thatfirms with good social responsibility tend to avoid locatingtheir operations in countries where environmental regulationis weak. However, firms with poor social responsibility appearto move their operations to these countries.

The structure of the remainder of this paper is as follows. InSection 2, we present our model and introduce our data. The

Table 1 – Descriptive statistics environmental responsibility of Multinational Enterprises

Correlations

Variable Min Max Mean Standarddeviation

Environmentalpolicy

Environmentalmanagement

Environmentalreporting

Environmentalperformance

impactimprovement

Environmental Policy −1 3 0.26 1.41 1Environmental Management −1 3 0.41 1.71 0.79 1Environmental reporting −1 3 −0.39 1.08 0.70 0.67 1Environmental Performance

Impact Improvement−1 3 −0.16 1.18 0.72 0.68 0.69 1

Source: EIRIS.For variable definitions see Appendix A.

57E C O L O G I C A L E C O N O M I C S 6 7 ( 2 0 0 8 ) 5 5 – 6 5

results of our analysis and the discussion are in Section 3. Abrief conclusion is in Section 4.

2. Data and methodology

The data on firms' CSR is provided by the Ethical InvestmentResearch Service (EIRIS). EIRIS has composed a cross-sectionaldatasetwhich covers 2685MNEs, locatedworld-wide andwhichcontains informationoncompanypolicy, reporting aswell asonbreaches by or convictions of theMNE. The topics that are dealtwith are environmental issues, stakeholder issues, businessethics and genetic engineering. Accordingly, ratings between −2and 3 are assigned. The details on CSR scoring are in AppendixA. We used four indicators of environmental quality: “Environ-mental Policy”, “Environmental Management”, “EnvironmentalReporting” and “Environmental Performance Impact Improve-ment”. The descriptive statistics of these variables are reportedin Table 1. We consider these indicators to have an underlyinglatent variable whichmeasures a firm's environmental respon-sibility. We applied factor analysis on these four indicators togenerate a single factor. Accordingly, we named the resultingfactor scores “Environmental Responsibility” and used this

Table 2 – Number of MNEs by industry and home country

Country Basicmaterials

Consumergoods

Consumerservices

Healthcare

In

Austria 1 0 0 0Belgium 2 1 2 1Switzerland 5 4 1 8Germany 5 9 7 7Denmark 0 2 0 5Spain 0 2 10 1Finland 3 2 2 0France 2 12 14 2United Kingdom 15 32 73 9Greece 0 1 2 0Italy 0 4 7 0Luxembourg 0 0 0 0Netherlands 2 5 4 1Norway 2 1 0 0Portugal 0 0 1 0Sweden 3 4 2 3All 40 79 125 37

variable in our econometric analysis. Table 1 also shows ahigh correlation between the four indicators of corporate en-vironmental responsibility.

From thedataset, we selected companies that are in theDowJones Stoxx 600 selection list, a list of the largest publicly quotedEuropean companies. We disregard financials such as banks orinsurance companies in our analysis. Table 2 gives an overviewof the number of Multinational Enterprises, classified by thecountry in which the company is based and by the industry thecompany is in.Wealsovisualize the global presenceof the firmsin Fig. 1. Overall, it appears thatmost MNEs are based in the UK.We see that the number ofMNEs per home countrymore or lessreflect the relative population sizes of the countries (i.e. largercountries havemore MNEs). An exception, however, is Switzer-land, which is relatively often represented. We observe thatsome countries are dominantly active in certain industries. Forexample, Spain and Italy have a relatively high share of com-panies in utilities, the Netherlands in the oil and gas industryand the UK dominates in consumer services. A simple Chi-square test of independence indeed rejects the hypothesis thatindustry and home country are independent variables.

Data on the international location of firms is extracted fromreported subsidiaries of firms. To this end, we use AMADEUS, a

dustrials Oilandgas

Technology Telecom-munications

Utilities All

2 0 0 1 1 51 0 0 2 1 10

10 0 2 1 0 3114 0 3 1 2 483 0 0 1 0 118 3 2 2 6 342 1 2 1 0 13

13 2 8 1 2 5661 7 19 6 12 2341 1 0 2 1 84 2 1 3 6 271 0 0 0 0 14 4 5 1 0 261 1 1 1 0 71 0 0 1 1 49 1 1 2 0 25

135 22 44 26 32 540

Fig. 1 –Global presence of Multinational Enterprises.


largedatabase that containsaccounting informationonEuropeanfirms. Note that a subsidiary can have subsidiaries itself. Ac-cordingly, AMADEUS classifies subsidiaries at different account-ing levels, where each subsidiary level is divided into sublevels.We look at the subsidiaries at the highest reported level and useinformation on the country location of the subsidiary and themost recent information on sales and assets of the subsidiary(2004–2005). We created a pooled and balanced cross-sectiondataset of 540 companies. For each company we have informa-tion on presence in 233 countries (a list of which is available fromthe authors), yielding vectors of 125,820 observations. Surely, notevery individual firm has operations in every country. Impress-ively however, in 188 of the 233 countries at least one multi-national is present. Table 3 gives an overview of the averagenumber of countries an MNE is operating in by region andindustry. It shows that, on average, an MNE is active in 17countries. Firms that produce basic materials are active in morecountries than firms from other industries and conduct most oftheir activities in Europe. It appears that theoil andgas industry ismost evenly scattered over the globe. Theutilities industry scoresthe loweston internationalpresence.Moreover,mostMNEs in thedataset are active in the US and Canada, which explains theaverage of around two for the region Central and North America.The EasternAsian, European andNorthAmericanmarkets are byfar themost attractive in absolute as well as in relative numbers.

We also extracted firm specific control variables from theAMADEUSdatabase.Theseareageof theMNEinyears,numberofemployees, leverage as measured by debt divided by total assetsand liquidity. Furthermore, we extractedmarket capitalization inbillionsof euros fromtheDow JonesStoxx 600 selection list. Olderand larger firms,measured by age, size, and employees, are likely

to be located in a larger number of countries. The cash position ofthe firm and the borrowing capacity, proxied by liquidity andleverage respectively, both can have an impact on locationdecisions. An overview of the descriptive statistics is in Table 4. Ifonecompares themedianvalues to themeanof thevariablesage,employees, leverage and liquidity inTable 4, it becomes clear thatthese variables have a heavily skewed distribution. For example,theaverageMNEhas35,048employees,whereas themedianMNEhas 12,854 employees. To account for these skewed distributionswe calculated the natural logarithm of these variables. Asexpected, employees and market capitalization are highlycorrelated (correlation coefficient equals 0.61) as larger firmsrequire both more capital and more labor. Age shows somecorrelation with employees and market capitalization. One canargue that the growthofMNEs is initially high, butwhena certainlevel of size is reached, the additional yearswill notmattermuchfor size. The liquidity ratio and leverage showno correlationwiththeother firmcharacteristics. Interestingly, the largerMNEsshowbehavior which is more environmentally responsible, as thevariable environmental responsibility has a positive and rela-tively largecorrelationwithemployeesandmarket capitalization.This is not very remarkable as a lot of literature also comes upwith this result (amongothers, seeLockett etal., 2006). Inourcase,the main reason appears to be that our indicators of environ-mental conduct on the firm's ability to provide externalstakeholders with information. In general, it is the larger firmsthat can best afford such ‘overhead’ costs.

We use three pairs of variables on the country level. We usetwodistinct sources each tomeasure environmental regulation,corruption, and poverty. Tomeasure a country's environmentalregulation standard, we use the World Business Environment

Table 3 – Average number of countries in which MNEs are operating by industry and region

Region (total #countries)

Industry

Basicmaterials

Consumergoods

Consumerservices

Healthcare

Industrials Oilandgas

Technology Telecom-munications

Utilities AverageMNE

Africa (58) 2.1 2.1 0.7 0.8 1.5 3.4 0.5 0.5 0.3 1.3Antarctica (4) 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0Caribbean andBahamas (21)

0.4 0.5 0.1 0.2 0.3 1.0 0.2 0.7 0.1 0.3

Central and NorthAmerica (13)

2.6 2.2 0.9 2.4 1.8 2.5 1.7 1.2 1.0 1.7

Eastern Asia (25) 4.3 3.6 1.2 3.9 2.4 2.1 2.6 0.8 0.4 2.4Europe (45) 12.0 11.2 6.2 12.0 9.3 8.0 8.3 7.4 4.6 8.7Middle East (15) 0.6 0.8 0.3 0.5 0.6 1.0 0.3 0.3 0.2 0.5Oceania (29) 1.0 0.8 0.3 1.0 0.6 0.6 0.4 0.1 0.2 0.6SouthAmerica (13)

2.8 2.1 0.7 1.7 1.6 2.3 1.1 0.9 1.3 1.5

WesternAsia (10)

0.2 0.2 0.1 0.1 0.1 0.3 0.1 0.0 0.0 0.1

World (233) 25.8 23.4 10.5 22.6 18.3 21.1 15.3 12.0 8.0 17.0

Table entries are industry averages of the number of countries an MNE is operating in per region. Total number of countries per region is inparentheses. A list of the countries included is available with the authors. The column Average MNE is a sample average irrespective of industryand the row World is a sample average irrespective of Region.


Survey (WBES) 2000 by theWorldBankGroup,which is based ona questionnaire contains information on financial and legalconstraints. This database has firm-level data for almost tenthousand firms in80 countries. The survey in the first placegoesinto the assessment of firmmanagers about the external factorsthat impact upon the way in which they operate theirbusinesses. This database is also used in the study by Becket al. (2005). The advantage of this survey is that itmeasures thestringencyof regulations that businesses experience inpractice.A country canhavevery strict environmental laws, but thesearenot effective if they are not enforced. We also extracted theperceived corruption in a country from this dataset. Anothersource we use is the World Development Indicators (WDI)dataset, also provided by the World Bank Group. This datasethas all types of information aboutmacroeconomic performanceand institutional arrangements. From the WDI dataset, wecounted in how many international environmental treaties

Table 4 – Descriptive statistics for Multinational Enterprises

Min Max Mean Standarddeviation

Median Age Em

Age in years 0 171 46 40 31 1Employees 35 419,200 35,048 60,985 12,854 0.26Marketcapitalization

0.13 155.89 6.37 14.75 1.95 0.15

Liquidity 0.08 16.72 1.30 1.22 1.01 −0.04Leverage 0.05 1.51 0.62 0.18 0.63 0.03Environmentalresponsibilityb

−1.50 1.82 0.00 0.97 0.20 0.12

For variable definitions see Appendix A.Sources: AMADEUS, EIRIS.a In these correlations and all subsequent calculations, natural logarithmLiquidity to account for the skewed distribution.b Factor scores of the four environmental responsibility indicators listed

such as the Kyoto protocol a country is participating and howmany plans or strategies a country adopted. We also used theWDI dataset to get information on poverty. We first used anational poverty measure, namely the percentage of thepopulation that is below the national poverty line. We alsoused an international poverty measure, namely the percentageof the population that has an income of less than 2 US$ a day.The datawe use on corruption is by Transparency International(TI). TI is a global civil society organisation trying to createchange towards a world free of corruption. TI constructs a so-called Corruption Perceptions Index (CPI) 2004, rating 146countries from 1 to 10 measuring corruption, with 1 being themost corrupt and 10 the least. In this index, Finland is the leastcorrupt country and Haiti and Bangladesh are themost corrupt.Alternatively, we also use the Kaufmann et al. (2005) dataset.This set presents estimates of six dimensions of governancecovering 209 countries and territories for five time periods: 1996,

Correlationsa

ployees Marketcapitalization

Liquidity Leverage Environmentalresponsibility

10.61 1.00

−0.08 −0.02 10.26 0.03 −0.23 10.36 0.47 −0.04 0.04 1

s have been taken of Age, Employees, Market capitalization and

in Table 1.

Table 5 – Correlations of country characteristics

EnvironmentalRegulation

Environmental Plansand Treaties

Control ofCorruption

CorruptionPerception

InternationalPoverty

NationalPoverty


1.00

Environmental Plansand Treaties

0.22 1.00

Control of Corruption 0.11 0.05 1.00Corruption Perception 0.04 −0.01 0.97 1.00International Poverty 0.04 0.13 −0.50 −0.49 1.00National Poverty 0.05 −0.03 −0.41 −0.44 0.54 1.00

Sources: Transparancy International, World Development Indicators, World Business Environment Survey, Kaufmann et al. (2005). For variabledefinitions see Appendix A.


1998, 2000, 2002 and 2004. The dimensions are Voice andAccountability, Political Stability and Absence of Violence,Government Effectiveness, Regulatory Quality, Rule of Law,andControl ofCorruption.Weusedthemeasureof corruptionofthis dataset. We provide detailed variable definitions inAppendix A. An overview of the correlations of these countryindicators is in Table 5.

In Table 5 we see that the three pairs of measures about,corruption, environmental regulation, and of poverty all arepositively correlated. Higher values of Environmental Reg-ulation and Corruption indicate better regulation and/ormore favorable conditions. Higher levels of poverty indicate ahigher percentage of the population being poor. The correla-tion of poverty with control of corruption has the expectednegative sign. However, the correlation with EnvironmentalRegulation is weak for both Corruption and Poverty anddiffers in sign for the various combinations. To this extent, itappears that it matters which variable one uses to test thePHH. Is the pollution haven a haven of high corruption (astested by Smarzynska and Wei, 2004), a haven of lowregulation, or is the haven to be associated with poverty?We will consider all three possibilities in our analysis in thenext section.

Weuse the comparative advantage approach to testwhethersome countries are more attractive with respect to theirenvironmental standards. The economic argument is thatcountries with little regulation have lower non-market costs.As we view CSR as the extent to which a firm internalizesmarket costs (see Heal, 2005), we expect that firms with lowerenvironmental responsibility will choose to locate their opera-tions in countries with poor environmental standards. Moreresponsible firms do not consider low regulation as being acomparative advantage and, ceteris paribus, they have lessincentives to operate in these countries.

The model we estimate is a binary location choice model.For a more detailed discussion on these models we refer toMcFadden (1974) or Greene (2000). We assume that the choiceof the subsidiary location is the dependent variable. For eachfirm, we try to explain the choice of whether or not to bepresent in a country. We constructed a binary variable Yij

which is equal to 1 if company i has subsidiaries in country j.We stay close to the existing empirical literature on binarychoice models and briefly discuss the background of themodel we estimate. We assume that the benefits Bij to firm

i (i=1,…, N) of locating in country j (i=1,…, J) is a latentvariable:

Bij ¼ Dij þ eij:

Here, Dij is the deterministic part and ɛij an error term. Dij isrelated to subsidiary country characteristics zj and parent-level firm group characteristics xkj in the following way:

Dij ¼ xkjbþ zjg:

Here the subscript j in the term xkj reflects that we do not apriori exclude possible interaction between parent-level firmgroup characteristics and subsidiary country characteristics.The firm chooses the location if the benefits are large enough,say larger than B⁎, and we only observe this outcome. Theprobability of observing firm i choosing location j is:

Pij ¼ P Yij ¼ 1� � ¼ P BijNB4

� �:

The actual outcome given Dij eventually depends on thedistribution of the error terms εij.

We specifically want to test whether there is a significantinteraction effect between a firm's CSR score and a country'senvironmental regulation. We add the following control vari-ables: age in years, number of employees, leverage asmeasuredby debt divided by total assets, liquidity and market capitaliza-tion. We took the logarithm of all of these variables, except forleverage. We did this as the distribution of these variables ismuch skewed. Furthermore, we created a “home” dummy,which is equal to one if we consider subsidiaries located in thesame country as where the MNE is based. We omitted theobservations forwhich this dummywas equal to one. There hasbeen some debate whether cultural distance is an importantdeterminant in international diversification, e.g. a meta-analy-sis byTihanyi et al. (2005) indicates that these differencesdonotseem to matter, particularly for firms based outside the US.Nonetheless, we add a colonial dummy variable, which is equalto one if the country where the subsidiary is located is a formercolony of the countrywhere theMNE is headquartered. As such,we control for common language advantages, historical ties, oradvantages of similarities of regulatory systems. Usually, ameasure of distance to the home country is also added as acontrol variable in spatial models. However, since all our firmsare European-based the distances on a global scalewill not varythat much and the country dummywill also account for spatial


effects. We therefore estimate the following relation for threepairs of different models:

E Yij� � ¼ Prob: Yij ¼ 1

� �

¼ KðajCountryj þ bkIndustryþ gFormerColonyþ giFirmi

þþdðEnvironmental Responsibilityi � CountryHavenjÞÞ:

Here, Country and Industry are the country and industryfixed effects. Again, we omit the observation if firm i isbased in country j, since in this case Yij is always equal toone. Environmental Responsibilityi is company i's environ-mental responsibility score. The models we test differ withrespect to the operationalization of CountryHavenj. Formodels 1 and 2, CountryHavenj is an indicator of countryj's Environmental Regulation. For models 3 and 4 we used ameasure of Corruption in country j. For models 5 and 6CountryHavenj is an indicator of country j's Poverty. Λ is the

Table 6 – Country presence of MNEs and possible havens

Coefficient (P-value) Mod

E

Control variablesLog market capitalization 0.32

(15.0Log age 0.34

(17.2Log liquidity 0.24

(6.60Leverage −0.3

(−2.Log employees 0.45

(18.7Former colony 0.67

(4.18Environmental responsibility −0.2

(−9.

qHavenq interaction termEnvironmental Responsibility×Environmental Regulation 0.04

(1.94Environmental Responsibility×Environmental Plans and Treaties

Environmental Responsibility×Control of Corruption

Environmental Responsibility×Corruption Perception

Environmental Responsibility×International Poverty

Environmental Responsibility×National Poverty

Number of observations 36,9

The estimated logit regression is: Presence=E[Yij]=Λ(αj Countryj+βk IndusCountryHavenj). Yij=1 if MNE i is present in country j. Λ is the logistic funcountry is a former colony of the countrywhere theMNE is based. Firmi areCountryHavenj we used Environmental Regulation, Environmental PlansInternational Poverty, and National Poverty respectively. Higher values ohigher values of CountryHaven indicate better regulation or higher levelsnot reported. Definitions of the variables are in Appendix A. The t-values⁎ indicates significance at 10, ⁎⁎ at 5, and ⁎⁎⁎ at 1%, respectively.

logistic distribution. Higher values of EnvironmentalResponsibilityi indicate higher social responsibility, highervalues of CountryHavenj indicate better perceived environ-mental regulation, lower levels of corruption, and higherlevels of poverty. We are specifically interested in the signand magnitude of the coefficient δ of the interacted term,Environmental Responsibilityi×CountryHavenj.

3. Results

The estimation results with respect to different specificationsof the model that estimates Presence=E[Yij] are in Table 6.Note that differences in sample sizes due to data availability ofsome countries can explain the differences of the coefficientsof the various models. For brevity's sake, we do not report the

el 1 Model 2 Model 3 Model 4 Model 5 Model 6

The qHavenq in this model is:

nvironmentalRegulation

Corruption Poverty

9⁎⁎⁎ 0.307⁎⁎⁎ 0.291⁎⁎⁎ 0.299⁎⁎⁎ 0.317⁎⁎⁎ 0.324⁎⁎⁎7) (17.64) (17.69) (17.67) (14.22) (12.36)0⁎⁎⁎ 0.316⁎⁎⁎ 0.308⁎⁎⁎ 0.310⁎⁎⁎ 0.329⁎⁎⁎ 0.365⁎⁎⁎2) (20.00) (20.23) (20.16) (15.65) (15.47)4⁎⁎⁎ 0.242⁎⁎⁎ 0.237⁎⁎⁎ 0.248⁎⁎⁎ 0.159⁎⁎⁎ 0.139⁎⁎⁎) (8.31) (8.76) (8.95) (4.85) (3.56)26⁎⁎ −0.333⁎⁎⁎ −0.320⁎⁎⁎ −0.279⁎⁎ −0.832⁎⁎⁎ −0.891⁎⁎⁎16) (−2.76) (−2.83) (−2.46) (−5.33) (−4.92)7⁎⁎⁎ 0.444⁎⁎⁎ 0.436⁎⁎⁎ 0.438⁎⁎⁎ 0.517⁎⁎⁎ 0.507⁎⁎⁎0) (22.65) (23.57) (23.01) (24.31) (18.53)9⁎⁎⁎ 0.661⁎⁎⁎ 0.690⁎⁎⁎ 0.651⁎⁎⁎ 0.790⁎⁎⁎ 0.815⁎⁎⁎) (5.69) (5.77) (5.59) (4.30) (4.16)29⁎⁎⁎ −0.231⁎⁎⁎ −0.170⁎⁎⁎ −0.179⁎⁎⁎ −0.177⁎⁎⁎ −0.172⁎⁎⁎54) (−11.19) (−7.76) (−8.08) (−6.46) (−5.69)

5⁎)

0.066⁎⁎⁎(4.30)

−0.080⁎⁎⁎(−4.81)

−0.079⁎⁎⁎(−4.64)

0.047⁎(1.73)

0.000(0.26)

49 65,923 81,597 65,001 69,611 37,802

try+η FormerColony+γiFirmi+δ(Environmental Responsibilityi×ction, conditional on Country fixed effects. FormerColony=1 if thethe reported firm characteristics. Industry are industry dummies. Forand Treaties, Control of Corruption, Corruption Perception,f Environmental Responsibility indicate higher social responsibility,of poverty. For brevity sake, the country and industry fixed effects areare calculated using the Huber–White robust standard errors.


country and industry fixed effects. Models 1 to 6 test for threepossible “Havens”. Models 1 and 2 directly test the classicalPollution Haven Hypothesis, namely that low country envir-onmental regulation is a haven for “dirty” industries. We finda positive and significant interaction between EnvironmentalResponsibility and Environmental Regulation, and Environ-mental Responsibility and Environmental Plans and Treaties,supporting the PHH. So, for both measures, we find evidencein favor of the PHH. It could be that perhaps it is in fact thecorrupt countries to which firms are transferring theiroperations. In this perspective, models 3 and 4 test whetherthere is an interaction effect between Corruption and Envir-onmental Responsibility. Here, we do find a negative sig-nificant interaction effect for both corruption measures,indicating that less responsible firms are less present incorrupt countries. For models 5 and 6, in which the supposedhaven is one of poverty, we findweak positive to no significantinteraction. Nonetheless, the main message of our analysis is

Table 7 – Country presence of MNEs in dirty industries and pos

Coefficient (P-value) Mod

E

Control variablesLog market capitalization 0.37

(11.3Log age 0.35

(14.1Log liquidity 0.28

(6.35Leverage −0.2

(−1.Log employees 0.38

(13.8Former colony 0.28

(2.49Environmental Responsibility 0.01

(0.36

qHavenq interaction termEnvironmental Responsibility×Environmental Regulation 0.03

(1.23Environmental Responsibility×Environmental Plans and Treaties

Environmental Responsibility×Control of Corruption

Environmental Responsibility×Corruption Perception

Environmental Responsibility×International Poverty

Environmental Responsibility×National Poverty

Number of observations 19,1

The estimated logit regression is: Presence=E[Yij]=Λ(αj Countryj+βk IndusCountryHavenj). Yij=1 if MNE i is present in country j. The regression is cofunction, conditional on Country fixed effects. FormerColony=1 if the couare the reported firm characteristics. Industry are industry dummies. ForPlans and Treaties, Control of Corruption, Corruption Perception, InternaEnvironmental Responsibility indicate higher social responsibility, higherpoverty. For brevity sake, the country and industry fixed effects are not repcalculated using the Huber–White robust standard errors.⁎ indicates sign

that we find support for the PHH in connection with firms'CSR. Furthermore, we establish that using proxies for envir-onmental regulation such as corruption and poverty instead ofdirectmeasures of environmental regulation can possibly leadto the wrong conclusions.

The analysis so far does not acknowledge the situation thatfor so-called dirty industries the notion of a pollution havenmight bemore relevant than for clean industries.We thereforeconducted the same regressions for a subset of firms, namelythose that operate in dirty industries, which are BasicMaterials, Consumer Goods, Oil and Gas, and Industrials. Theregression results are in Table 7. We find a positive andsignificant interaction effect for Environmental Plans andTreaties, supporting the PHH. We do not find a significanteffect for the interaction with Environmental Regulation, butthis couldbedue to theheavily reducednumberofobservations.The regression results of the subset of firms in clean industries(not reported), show no significant interaction effect with

sible havens

el 1 Model 2 Model 3 Model 4 Model 5 Model 6


nvironmentalRegulation

Corruption Poverty

9⁎⁎⁎ 0.349⁎⁎⁎ 0.328⁎⁎⁎ 0.333⁎⁎⁎ 0.406⁎⁎⁎ 0.434⁎⁎⁎0) (13.22) (13.12) (12.83) (12.85) (12.31)1⁎⁎⁎ 0.331⁎⁎⁎ 0.322⁎⁎⁎ 0.326⁎⁎⁎ 0.340⁎⁎⁎ 0.376⁎⁎⁎5) (17.34) (17.51) (17.41) (11.87) (11.67)8⁎⁎⁎ 0.297⁎⁎⁎ 0.291⁎⁎⁎ 0.305⁎⁎⁎ 0.182⁎⁎⁎ 0.160⁎⁎⁎) (7.67) (8.02) (8.17) (4.05) (3.15)95⁎⁎ −0.390⁎⁎⁎ −0.390⁎⁎⁎ −0.311⁎⁎⁎ −0.684⁎⁎⁎ −0.813⁎⁎⁎98) (−3.16) (−3.25) (−2.66) (−3.83) (−3.82)7⁎⁎⁎ 0.381⁎⁎⁎ 0.379⁎⁎⁎ 0.376⁎⁎⁎ 0.421⁎⁎⁎ 0.401⁎⁎⁎5) (16.22) (16.70) (16.18) (14.63) (11.80)6⁎⁎ 0.359⁎⁎⁎ 0.403⁎⁎⁎ 0.366⁎⁎⁎ 0.333⁎⁎ 0.421⁎⁎⁎) (3.55) (3.86) (3.60) (2.17) (2.66)0 −0.005 0.046⁎ 0.049⁎⁎ 0.067⁎⁎ 0.065⁎) (−0.22) (1.90) (2.01) (2.24) (1.83)

1)

0.065⁎⁎⁎(4.22)

−0.083⁎⁎⁎(−4.98)

−0.081⁎⁎⁎(−4.92)

0.040(1.42)

−0.000(−0.09)

61 33,938 41,108 33,460 22,226 19,598

try+η FormerColony+γiFirmi+δ(Environmental Responsibilityi×nducted for a sub-sample of MNEs in dirty industries. Λ is the logisticntry is a former colony of the country where the MNE is based. Firmi

CountryHavenj we used Environmental Regulation, Environmentaltional Poverty, and National Poverty respectively. Higher values ofvalues of CountryHaven indicate better regulation or higher levels oforted. Definitions of the variables are in Appendix A. The t-values areificance at 10, ⁎⁎ at 5, and ⁎⁎⁎ at 1%, respectively.

Table 8 – Country presence of MNE's and possible havens in poor countries

Coefficient (P-value) Model 1 Model 2 Model 3 Model 4 Model 5 Model 6



Corruption Poverty

Control variablesLog market capitalization 0.338⁎⁎⁎ 0.325⁎⁎⁎ 0.318⁎⁎⁎ 0.328⁎⁎⁎ 0.326⁎⁎⁎ 0.324⁎⁎⁎

(13.31) (14.24) (14.17) (14.34) (13.87) (12.36)Log age 0.335⁎⁎⁎ 0.336⁎⁎⁎ 0.339⁎⁎⁎ 0.335⁎⁎⁎ 0.332⁎⁎⁎ 0.365⁎⁎⁎

(14.41) (15.83) (16.43) (16.08) (14.78) (15.47)Log liquidity 0.169⁎⁎⁎ 0.149⁎⁎⁎ 0.160⁎⁎⁎ 0.162⁎⁎⁎ 0.157⁎⁎⁎ 0.139⁎⁎⁎

(4.58) (4.47) (4.96) (4.97) (4.54) (3.56)Leverage −0.692⁎⁎⁎ −0.866⁎⁎⁎ −0.829⁎⁎⁎ −0.764⁎⁎⁎ −0.814⁎⁎⁎ −0.891⁎⁎⁎

(−4.22) (−5.28) (−5.19) (−4.90) (−4.96) (−4.92)Log employees 0.507⁎⁎⁎ 0.517⁎⁎⁎ 0.513⁎⁎⁎ 0.507⁎⁎⁎ 0.514⁎⁎⁎ 0.507⁎⁎⁎

(20.54) (23.59) (24.03) (23.47) (22.54) (18.53)Former colony 0.808⁎⁎⁎ 0.788⁎⁎⁎ 0.811⁎⁎⁎ 0.801⁎⁎⁎ 0.793⁎⁎⁎ 0.815⁎⁎⁎

(3.80) (4.53) (4.72) (4.58) (4.30) (4.16)Environmental Responsibility −0.207⁎⁎⁎ −0.227⁎⁎⁎ −0.192⁎⁎⁎ −0.199⁎⁎⁎ −0.179⁎⁎⁎ −0.172⁎⁎⁎

(−7.13) (−7.38) (−7.01) (−6.96) (−6.59) (−5.69)

qHavenq interaction termEnvironmental Responsibility×Environmental Regulation 0.071⁎⁎⁎

(3.18)Environmental Responsibility×Environmental Plans and Treaties 0.057⁎

(1.83)Environmental Responsibility×Control of Corruption 0.024

(0.59)Environmental Responsibility×Corruption Perception 0.008

(0.18)Environmental Responsibility×International Poverty 0.033

(1.21)Environmental Responsibility×National Poverty 0.000

(0.26)Number of observations 32,270 53,015 61,313 49,788 41,951 37,802

The estimated logit regression is: Presence=E[Yij]=Λ(αj Countryj+βk Industry+η FormerColony+γiFirmi+δ(Environmental Responsibilityi×CountryHavenj). Yij=1 if MNE i is present in country j. The regression is conducted for a sub-sample of non-OECD countries and othernon-high-income countries. Λ is the logistic function, conditional on Country fixed effects. FormerColony=1 if the country is a former colony ofthe country where the MNE is based. Firmi are the reported firm characteristics. Industry are industry dummies. For CountryHavenj we usedEnvironmental Regulation, Environmental Plans and Treaties, Control of Corruption, Corruption Perception, International Poverty, and NationalPoverty respectively. Higher values of Environmental Responsibility indicate higher social responsibility, higher values of CountryHavenindicate better regulation or higher levels of poverty. For brevity sake, the country and industry fixed effects are not reported. Definitions of thevariables are in Appendix A. The t-values are calculated using the Huber–White robust standard errors. ⁎ indicates significance at 10 and ⁎⁎⁎ at 1%,respectively.


Environmental Plans and Treaties, making a stronger case forthe PHH in relation to firms' responsibility. Thus, again,we see anegative effect for the corruption measures, indicating thatresponsible firms are relatively more often located in corruptcountries. Finally, it appears that poverty has no significanteffect on location behavior conditional on corporate environ-mental responsibility.

As a last robustness check,we estimate themodel for a subsetof non-OECD and other non-high income countries. The estima-tion results are in Table 8. Again we find support for the PHH, asreflected by the positive and significant interaction of environ-mental responsibilities with both measures of environmentalstandards. For the corruption measures, we do not find asignificant interaction effect, as opposed to the previous analysis.Finally, the fact thatpoverty isnot apotentialhavenseems tobearobust result aswedonot find a significant interactioneffectwitheither measure.

Thus, although the effectwe find about firms' responsibilityin relation to their presence in countries with particularcharacteristics is statistically significant, one might questionwhether the effect is large enough to speak of true pollutionhavens. Nonetheless, we find some evidence in favor of thePHHbasedmainlyon firmpolicy, not on firmperformance. Theactual differences in levels of pollution associated withdifferences in environmental responsibility scores might beconsiderable. So, on the one hand, there might even be astronger “haven” effect thanweobserve. Thenagain, details onhow “irresponsible” firms diversify their operations in everycountry are not in our dataset. We can only note that certaintypes of firms are relativelymore present in countrieswith lowregulation; we do not know the exact nature of their activities.It could be that these ‘dirty’ MNEs are abusing low regulationcountries. On the other hand, they might produce their goodsin the home country and simply distribute through the


subsidiaries in the other countries. One has to be careful toconclude that firms with lower responsibility are actuallyconducting PHH behavior without knowing the exact levelsand whereabouts of pollution. Furthermore, our analysis ismerelydescriptive, aswith a cross-sectional datasetwe cannotaccount for endogeneity problemsor causal relations onwhichthe recent literature (e.g. Cole et al., 2006) has focused.Elaborating on the dataset is a major challenge in our futureresearch. Nevertheless, we would like to classify our findingsabout the locational behavior ofMNEs in connectionwith theircorporate social responsibility as “strong circumstantial evi-dence” in favor of the PHH.

4. Conclusion

This paper investigates the relationship between the corporatesocial responsibility and location choices of MultinationalEnterprises (MNEs) and environmental regulation, governance,and wealth of countries. More specifically, we address thequestion whether MNEs transfer their “dirty” operations towardspoor countries, corrupt countries or countries with low environ-mental regulation; the so-called Pollution Haven Hypothesis(PHH).We relate this issue to the responsibility of the firm. In thisperspective, we regard corporate social responsibility as theextent to which a firm internalizes market costs. Firms that sethigh internal environmental standards will not experience asupposedly comparative advantage when locating in countrieswith poor environmental regulation. On the other hand, firmswith little environmental responsibility might have incentives toengage in PHH behavior. Using firm-level data and directmeasures of a country's environmental regulation, we find newevidence that for firms with weak environmental standards PHHbehavior can be observed. However, if we consider countrieswithhigh corruption or high poverty as being “havens”we do not findsimilar evidence. Thus, we establish that it is predominantlyfirms with poor social responsibility that appear to move theiroperations to countrieswithweak regulation. The ‘good’, i.e.mostresponsible, firms tend to avoid locating their operations in thesecountries. As such, we conclude that CSR does matter withrespect toMNEs locational behavior.At the same timewesee thatit is also the larger firms that are socially responsible. It could bethat larger firms that channel their growth by diversifying theiroperations in several countries experience a considerable varietyof stakeholder demands due to, for instance, differences inregulationor cultural differences.We leave thisnexusof size, CSRand international locationdecisions as a topic for future research.

Appendix A

Variable
Definition Source
EnvironmentalPerformanceImpactImprovement

“What level ofimprovements inenvironmental impact canthe Companydemonstrate?” (No data or

EIRIS

Appendix A (continued )

Variable
Definition Source
inadequate data=−1 Noimprovement=0, Minorimprovement=1,Significantimprovement=2, Majorimprovement=3)

EnvironmentalReporting

“How does EIRIS rate thecompany's environmentalreporting?” (Inadequate=−1, Weak=0 Moderate=1,Good=2, Exceptional=3)

EIRIS

EnvironmentalManagement

“How does EIRIS rate thecompany's environmentalmanagement system?”(Inadequate=−1, Weak=0,Moderate=1, Good=2,Exceptional=3)

EIRIS

EnvironmentalPolicy

“How does EIRIS rate thecompany's environmentalpolicy and commitment?”(Inadequate=−1, Weak=0,Moderate=1, Good=2,Exceptional=3)

EIRIS

EnvironmentalResponsibility

Factor scores based on afactor analysis of the abovefour CorporateEnvironmentalresponsibility indicators.

Owncalculations

Total Assets
Reported total assets as of2004 in thousands of U.S.dollars.
Amadeus

Leverage
Ratio of: (current liabilities+non-current liabilities) /total assets as of 2004.
Amadeus

Age
Age in years of the companyas of 2004, based on thereported date ofincorporation
Amadeus

Employees
Number of reportedemployees as of 2004
Amadeus

Liquidity
Reported liquidity ratio (%)as of 2004
Amadeus

Market Cap
Free Float MCap (in BEUR) asof 03.01.2005
Dow Jones StoxxSel. List

InternationalPoverty

Percentage of thepopulation below $2 a day

WorldDevelopmentIndicators

NationalPoverty

Percentage of thepopulation below thenational poverty line


Environmentalplans and Treaties

Standardized values of thecount of “Participation intreaties (Climate change,Ozone Layer, CFC control,Law of the Sea, Biologicaldiversity, Kyoto protocol)”and “Environmentalstrategies or action plans”and “Biodiversityassessments, strategies oraction plans”


CorruptionPerception Index(CPI)

Standardized values of 2004Corruption PerceptionsIndex (CPI) (CPI Scorerelates to perceptions of the

TransparencyInternational


Appendix A (continued )

Variable
Definition Source
degree of corruption as seenby business people andcountry analysts andranges between 10 (highlyclean) and 0 (highlycorrupt).)


Standardized countryaverages of: “Please judgeon a four point scale howproblematic are thesedifferent regulatory areasfor the operation andgrowth of your business;Environmental regulations:1=No Obstacle, 2=MinorObstacle, 3=ModerateObstacle, 4=MajorObstacle”

World BusinessEnvironmentSurvey (WBES)

Control ofCorruption

Based on several hundredindicators, drawn from 37separate data sourcesconstructed by 31 differentorganisations. Compiledusing an unobservedcomponent technique byKaufmann et al. (2005).

Worldbank, seeKaufmann et al.(2005)

R E F E R E N C E S

Antweiler, W., Copeland, B.R., Taylor, M.S., 2001. Is free trade goodfor the environment? American Economic Review 91, 877–908.

Bagnoli, M., Watts, S., 2003. Selling to socially responsibleconsumers: competition and the private provision of publicgoods. Journal of Economics and Management Strategy 12,419–445.

Baron, D., 2001. Private politics, corporate social responsibility andintegrated strategy. Journal of Economics and ManagementStrategy 10, 7–45.

Becchetti, L., di Giacomo, S., Pinnachio, D., 2005. Corporate SocialResponsibilityandCorporatePerformance:Evidence fromaPanelof US Listed Companies CEIS Research Paper Series, No. 26-78.

Beck, T., Demirgüç-Kunt, A., Maksimovic, V., 2005. Financial andlegal constraints to growth: does size matter? Journal ofFinance 60, 137–177.

Beltratti, A., 2005. The complementarity between corporategovernance and corporate social responsibility. Geneva Paperson Risk & Insurance: Issues & Practice 30, 373–386.

Cole, M.A., Elliott, R.J.R., 2003. Determining the trade-environmentcomposition effect: the role of capital, labour andenvironmental regulations. Journal of EnvironmentalEconomics and Management 46, 363–383.

Cole, M.A., Elliott, R.J.R., Fredriksson, P.G., 2006. Endogenouspollution havens: does FDI influence environmental regulations?Scandinavian Journal of Economics 108, 157–178.

Dahlsrud, A., in press. How corporate social responsibility isdefined: an analysis of 37 definitions. Corporate SocialResponsibility and Environmental Management (DOI: 10.1002/csr.132).

Damania, R., Fredriksson, P.G., List, J.A., 2003. Trade liberalization,corruption, and environmental policy formation: theory andevidence. Journal ofEnvironmental EconomicsandManagement46, 490–512.

Eskeland, G.S., Harrison, A.E., 2003. Moving to greener pastures?Multinationals and the Pollution Haven Hypothesis. Journal ofDevelopment Economics 70, 1–23.

Greene, W.H., 2000. Econometric Analysis. Prentice-HallInternational, London.

Heal, G.M., 2005, Corporate social responsibility? An economic andfinancial framework, geneva papers on risk and insurance:issues and practice, 30, 387–409.

Janicke, M., Binder, M., Monch, H., 1997. Dirty industries': patternsof change in industrial countries. Environmental and ResourceEconomics 9, 467–491.

Kaufmann, D., Kraay, A., Mastruzzi, M., 2005. Governance MattersIV: Governance Indicators for 1996–2004. World Bank,Washington DC.

Letchumanan, R., Kodama, F., 2000. Reconciling the conflictbetween the ‘Pollution-Haven’ hypothesis and an emergingtrajectory of international technology transfer. Research Policy29, 59–79.

Levinson, A., 1996. Environmental regulations andmanufacturer'slocation choices: evidence from the census of manufacturing.Journal of Public Economics 61, 5–29.

List, J.A., 2001. U.S. country-level determinants of inbound FDI:evidence from a two-step modified count data model.International Journal of Industrial Organization 19, 953–973.

List, J.A., Co, C.Y., 2000. The effects of environmental regulationson foreign direct investment. Journal of EnvironmentalEconomics and Management 40, 1–20.

Lockett, A., Moon, J., Visser, W., 2006. Corporate socialresponsibility in management research: focus, nature, salienceand sources of influence. Journal of Management Studies 43,115–136.

Low, P., Yeats, A., 1992. Do dirty industriesmigrate. In: Low, P. (Ed.),International Trade and the Environment, World BankDiscussion Paper 159. World Bank, Washington D.C.

Mani, M., Wheeler, D., 1997. In search of ‘Pollution-Havens’? Dirtyindustry in the world economy, 1960–1995. Working PaperSeries. Research Project on Social and EnvironmentalConsequences of Growth-Oriented Policies, Policy ResearchDepartment. World Bank, Washington, DC.

McFadden, D., 1974. Conditional logit analysis of qualitative choicebehaviour. In: Zarembka, P. (Ed.), Frontiers in Econometrics.Academic Press, New York.

McWilliams, A., Siegel, D., 2000. Corporate social responsibilityand financial performance: correlation or misspecification?Strategic Management Journal 21, 603–609.

McWilliams, A., Siegel, D., 2001. Corporate social responsibility: atheory of the firm perspective. Academy of ManagementReview 26, 117–127.

Repetto, R., 1995. Jobs, Competitiveness and EnvironmentalRegulation: What are the Real Issues. World ResourcesInstitute, Washington, D.C.

Smarzynska Javorcik, B.K., Wei, S., 2004. Pollution havens andforeign direct investment: dirty secret or popular myth?Contributions to Economic Analysis & Policy 3 (2) Article 8.http://www.bepress.com/bejeap/contributions/vol3/iss2/art8.

Sorsa, P., 1994. Competitiveness and environmental standards:some exploratory results. Policy Research Working Paper 1249.World Bank, Washington, DC.

Talkukdar, D., Meisner, C., 2001. Does the private sector help orhurt the environment: evidence from CO2 pollution indeveloping countries. World Development 29, 827–840.

Tihanyi, L., Griffith, D.A., Russell, C.J., 2005. The effect of culturaldistance on entry mode choice, international diversification,and MNE performance: a meta-analysis. Journal ofInternational Business Studies 36, 270–283.

Xing, Y., Kolstad, C.D., 2002. Do lax environmental regulationsattract foreign investment? Environmental and ResourceEconomics 21, 1–22.

http://www.bepress.com/bejeap/contributions/vol3/iss2/art8

Documents

Environmental regulation and MNEs location: Does CSR matter?