The Impact of International Institutions and Trade on Environmental Policy Convergence in Europe

1 The Impact of International Institutions and Trade on Environmental Policy Convergence in Europe Katharina Holzinger, Christoph Knill, Thomas Sommerer First Draft Paper presented at the ECPR Joint Sessions, Granada, 14-20 April 2005

2 The Impact of International Institutions and Trade on Environmental Policy Convergence in Europe Katharina Holzinger, Christoph Knill, Thomas Sommerer 1 Introduction 1 In recent years, there is a growing and renewed interest in the study of cross-national policy convergence. There is an intensified debate on the convergence and divergence of national policies, which is closely related to research on the domestic impact of European integration (Europeanization) and globalization. Does the strong growth of economic and institutional interlinkages between nation states lead to increasingly similar policies across countries? Or is the search for convergence emerging from the domestic impact of globalization and European integration "an impossible quest" (Dimitrova and Steunenberg 2000: 201), as domestic responses to global or European challenges are strongly influenced by existing domestic structures and institutions (see, for example, Caporaso et al. 2001; Héritier et al. 2001; Knill 2001)? Yet we still have a limited understanding of the phenomenon of policy convergence. What explains the adoption of similar policies across countries over time? Under which conditions can we expect that domestic policies converge or rather develop further apart? Why do countries converge on some policies, but not on others? In the literature, many factors have been suggested in order to account for the mixed empirical evidence of both convergence and divergence, including, for instance, the role of international organizations, regulatory competition between nation states, capacities for national policy adjustment, as well as parallel problem pressure (cf. Bennett 1991; Drezner 2001; Hoberg 2001: Dolowitz and Marsh 2000; Holzinger and Knill 2005). However, there is still a lack of systematic theoretical and empirical investigations about their actual explanatory relevance, implying that we still have limited knowledge about the causes and conditions of cross-national policy convergence. In this article, we aim to address this research deficit. In empirical terms, we analyze, if and to what extent we can observe a convergence of environmental policies in Europe. For this purpose, we analyze the development of 40 environmental measures across 24 countries between 1970 and 2000. 1 The article strongly builds upon empirical and theoretical work carried out in the research project "Environmental Policy Convergence in Europe (ENVIPOLCON)". Funding of this research by the European Commission within the fifth framework programme on socio-economic research is gratefully acknowledged. The ENVIPOLCON research team includes the universities of Konstanz, Nijmegen, Hamburg, Berlin (Free University) and Salzburg.

3 From a theoretical perspective, we concentrate on the explanatory power of two factors often mentioned as important causes of cross-national policy convergence, namely international economic and institutional interlinkages between nation states. To what extent can these factors account for the observed patterns of environmental policy convergence or divergence? The paper is structured as follows. In section two, we present our theoretical framework. The research design and the operationalization of variables are described in section three. In section four, we present and discuss the results of our analysis. Summary and general conclusions to be drawn from our study are part of section five. 2 Theoretical Background The literature provides us with many potential factors that might lead to cross-national policy convergence (cf. Bennett 1991; DiMaggio and Powell 1991, Dolowitz and Marsh 2000, Drezner 2001; Hoberg 2001). Basically, these factors refer to five distinct convergence mechanisms (Holzinger and Knill 2005): First, cross-national policy convergence might be simply the result of similar, but independent responses of different countries to parallel problem pressure (e.g. ageing of societies); i.e., policy convergence is caused by similar policy problems countries are reacting to (Bennett 1991: 231). Second, several studies emphasize convergence effects stemming from the imposition of policies. Imposition refers to constellations where countries or international organizations force other countries to adopt certain policies by exploiting asymmetries in political or economic power. Third, regulatory competition emerging from the increasing economic integration of European and global markets has been identified as important factor that drives the mutual adjustment of policies across countries. Fourth, emphasis is placed on the harmonization of national policies through international or supranational law. Countries are obliged to comply with international rules on which they have deliberately agreed in multilateral negotiations. Finally, cross-national policy convergence can be caused by communication and information exchange in transnational networks. As we are interested in the convergence effects caused by economic and institutional interlinkages among countries, we will only refer to those mechanisms mentioned above that are actually relevant in this respect, namely, international harmonisation, transnational communication and regulatory competition.

4 2.1 International Harmonization The mechanism of international harmonization leads to cross-national convergence if the involved countries comply with uniform legal obligations defined in international or supranational law. Harmonization refers to a specific outcome of international cooperation, namely to constellations in which national governments are legally required to adopt similar policies and programs as part of their obligations as members of international institutions. International harmonization and more generally international cooperation presuppose the existence of interdependencies or externalities which push governments to resolve common problems through cooperation within international institutions, hence sacrificing some independence for the good of the community (Drezner 2001: 60; Hoberg 2001: 127). Once established, institutional arrangements will constrain and shape the domestic policy choices, even as they are constantly challenged and reformed by their member states. This way, international institutions are not only the object of state choice, but at the same time consequential for subsequent governmental activities (Martin and Simmons 1998: 743). However, as member states voluntarily engage in international cooperation and actively influence corresponding decisions and arrangements, the impact of international harmonization on national policies constitutes no hierarchical process; it can rather be interpreted as "negotiated transfer" (Dolowitz and Marsh 2000: 15). The scope of cross-national convergence triggered by international harmonization is restricted to those countries that are actually committed to the same international agreements or supranational regulations; i.e., to countries that are a member of the same international or supranational institution in question. The degree of convergence caused by international harmonization is affected by two factors. First, convergence will increase with the extent to which countries are institutionally interlinked. The higher the level of common membership in international and supranational institutions among the countries, the more their environmental policies will converge. Second, convergence is affected by the obligatory potential of the international institution in question. The degree of similarity will increase with the extent to which compliance with legal obligation can actually be enforced. International institutions reveal important differences in terms of their enforcement powers. The European Union (EU) can be characterized as an institution in which such powers are comparatively well developed, given the direct effect and the supremacy of European law, the influential role of the European Court of Justice in the enforcement of Community law, the albeit restricted monitoring activities of the European Commission as well as the opportunity to financially sanction non-compliant member states. Against this background, the converging effects of European legislation can be expected to be higher than those of intergovernmental organizations or international regimes, where enforcement powers are less developed.

5 2.2 Transnational Communication International institutions might cause cross-national policy convergence not only by legally binding rules and decisions, but also by non-obligatory factors. This expectation follows from various theoretical arguments developed in the field of organisation sociology. As emphasised by DiMaggio and Powell (1991), frequently interacting organisations, such as national bureaucracies, tend to develop similar structures and concepts over time. Policy convergence results from the striving of organisations to increase their social legitimacy by embracing forms and practices that are valued within the broader institutional environment. In this context, the establishment of international institutions and networks plays an important role in facilitating the exchange of new policy models and regulatory innovations across national borders (Strang and Meyer 1993). Various ways of non-obligatory convergence can be distinguished. First, states might act mimetically to copy the successful policies of other states. The demand for similarity of structure and functioning, rather than increased efficiency drives the process of cross-national convergence (DiMaggio and Powell 1991). Second, the diffusion of professional knowledge via transnational networks or "epistemic communities", who share common beliefs and perceptions over policy problems and corresponding solutions to address these problems (Haas 1992), plays an important role in facilitating the cross-national diffusion of policy concepts by deliberation and learning (Joerges and Neyer 1997). Similar to international harmonization, convergence effects emerging from transnational communication can be expected only for those countries that are members of the same international institution or network. At the same time, the degree of convergence will increase, the higher the institutional interlinkage of countries, i.e., the more they are members of the same institutions or networks. Moreover, the degree of convergence is affected by what we refer to as "communicative potential" of the international institutions in question. This potential varies with frequency and breadth of interaction; i.e., the functional differentiation of transnational networks (Kern 2000, 267). 2.3 Regulatory Competition According to theories of regulatory competition, countries facing competitive pressures are expected to mutually adjust their policies, hence leading to cross-national convergence. Regulatory competition presupposes economic integration among countries. Especially with the increasing integration of European and global markets and the abolition of national trade barriers, the international mobility of goods, workers and capital puts competitive pressure on the nation states to redesign domestic market regulations in order to avoid regulatory burdens restricting the competitiveness of domestic industries. The pressure arises from (potential) threats of economic actors to shift their activities elsewhere, inducing governments to adjust their regulatory standards.

6 In the literature there is a broad debate about the extent to which these adjustments coincide with an upward or downward shift of regulatory levels (race to the top versus race to the bottom). While the standard expectation typically presupposes a downward shift (Hoberg 2001: 127; Simmons and Elkins 2004; Drezner 2001: 57-59), other theoretical work suggests that there are a number of conditions which may drive policy in both directions (Vogel 1995; Scharpf 1997; Kern, Jörgens and Jänicke 2000; Holzinger 2002, 2003), including, for example, the type of policy concerned (product or process standards), or the presence of other interests than business in national politics. As in this paper the analytical focus is on the degree rather than the direction of policy convergence, the decisive expectation in our context is that theories of regulatory competition predict an increase of policy similarity over time among countries exposed to competitive pressures, regardless of the regulatory level at which this convergence might occur. Having identified the general relationship between regulatory competition and policy convergence, however, we still no nothing about the conditions under which the expected convergence effects might be more or less pronounced. In this respect, various factors restrict the scope of potentially converging countries and policies. On the one hand, potential convergence effects of regulatory competition presuppose economic integration between market economies. Even in constellations of high economic integration, no competitive pressures will emerge in and between non-market economies. This scenario applies in particular to the communist countries before 1990. On the other hand, a qualification applies to the policies for which convergence effects are predicted. Adjustments are only expected for trade-related policies, such as product or process standards. No convergence will occur for policies, which are subject to low competitive pressures from international markets. This holds true for all policies that are not directly related to products or to production processes, such as nature protection, or bird protection. It holds also true if trade-related policies are concerned, but the effects of the regulation on production costs are low. In this case they do not affect competition between industries in different countries. As regards the degree of convergence, theories of regulatory competition generally predict that countries adjust policy instruments and regulatory standards in order to cope with competitive pressures emerging from international economic integration. The more exposed a country is to competitive pressures following from high economic integration (emerging from its dependence on trade of goods, capital and services with other countries); the more likely it is that its policies will converge to other states with international exposure. In other words, the degree of convergence depends on the level of competitive pressures to which countries are exposed.

7 3 Research Design and Variables In view of the research questions addressed in this paper, the degree of convergence of national environmental policies is conceived as the dependent variable. In this context, convergence is measured and identified by increasing policy similarity over time. Based on the previous theoretical considerations, we focus on three factors as the main independent variables that are expected to account for differences in environmental policy convergence: (1) the degree of interlinkage of countries in international institutions with obligatory potential, (2) the degree of interlinkage of countries in international institutions with communicative potential, and (3) the degree of economic interlinkage; i.e., the extent to which a country is linked to other countries by its trade relations, and hence subject to competitive pressures emerging from regulatory competition. Figure 1: Dependent and Independent Variables Independent Variables Dependent Variable International Harmonization obligatory potential of institutional interlinkage Transnational Communication communicative potential of institutional interlinkage Economic Interlinkage regulatory competition Degree of Policy Convergence policy adoption policy instruments policy settings 3.1 Dependent Variable Selection of Policies Items and Countries In order to assess the degree of environmental policy convergence we rely on the dataset developed in the EU-sponsored project on Environmental Policy Convergence in Europe (ENVIPOLCON). From this dataset, we selected 40 policy items in accordance with the following criteria. First, the policies under investigation constitute a balanced sample covering all environmental media. Second, the

8 policies cover the whole range of different policy types that are of particular theoretical interest. To assess the effects of regulatory competition, we included both trade-related policies (product and process standards) and policies that are not subject to competitive pressures through economic integration. To compare effects of the obligatory and communicative potential of international institutions, our sample covers not only obligatory items (for which a legally-binding standard at the international level exists), but also non-obligatory measures (where internationally harmonized rules are lacking). Third, with selection of policies, we are able to analyze convergence effects with regard to different policy dimensions, namely, the extent to which countries have developed a specific policy or not with regard to a certain problem (policy presence), the instruments applied (e.g. command-andcontrol regulation versus policies based on economic incentives) as well as the settings of the instruments (i.e. levels of environmental standards). The policies under investigation are listed in Annex 1. To measure convergence, we compare the degrees of cross-national policy similarity at four points of time (1970, 1980, 1990, and 2000). The year 1970 represents the starting point of national and international environmental policy. In 1980, the first wave of laws had been passed in the more environmentally advanced countries. Moreover, various international organisations had launched environmental programmes and policies for the first time. In 1990 more or less comprehensive environmental policies had been put in place in almost all industrialised countries as well as international organisations. The year 2000, finally, represents the situation after the fall of the Berlin Wall, with growing trade relations between Western and Eastern European states, and with a number of Central and Eastern European (CEE) countries seeking membership in the EU. Therefore, according to the theories outlined above, over the last decade an overall approximation of environmental policies in Europe can be expected. The 24 countries under investigation include the member states of the EU-15 (except Luxembourg), Norway, Switzerland, Poland, Slovakia, Hungary, Bulgaria, Romania as well as the United States, Mexico and Japan. The country selection thus includes different groups of countries with regard to the influence of EU-membership: the founding members of the EU, countries joining the EU in 1973, 1981/1986, 1995 and 2004, countries currently negotiating EU-accession, countries that are not members but closely affiliated with the EU (Norway and Switzerland) and countries not institutionally integrated or affiliated with the EU. For the collection of the data, a questionnaire was developed that was completed by environmental policy experts for each country 2. The information provided by the country experts was crosschecked and compared with regard to completeness and correctness. 2 The workload as well as the conditions of payment for the country experts was defined in specific working contracts.

9 Classical Approaches to Convergence With regard to the measurement of convergence, the classical approach is to focus on decreases of the variation coefficient (σ convergence). This approach, however, bears certain disadvantages with regard to a sound assessment of convergence. First, as the variation coefficient is calculated on the basis of the whole sample of countries, similarity increases among country subgroups might be overlooked. It is even conceivable that one or more outliers lead to an increase in the variation coefficient, although the other countries are converging. Second, and related to this point, this approach does not allow for statements on the extent to which single countries are converging to the whole group of countries. The variation coefficient constitutes no reference point against which developments in single countries can be assessed. Third, on the basis of the variation coefficient, convergence can only be assessed for metrical, but not for nominal data. Fourth, the variation coefficient does not measure convergence effects that might result from the fact that countries adopt a certain policy. Figure 2: National Policy Development for Limit Values for CO-Emissions from Passenger Cars, 1970-2000 Limit values: Passenger cars (CO emissions in g/km) 1970-2000 45 FRA GER GRE HUN NOR POR ROM SWE SWI USA 40 35 30 25 20 15 10 5 0 1970 1980 1990 2000 Coeff. Var. 0,17 0,44 1,03 0,32 Mean (g/km) 38,65 28,12 6,64 3,0 N 2 7 8 10 The following two examples taken from our dataset illustrate these points. They show the development of national environmental policies for a representative sample of ten out of our 24 countries under investigation. Figure 2 shows national changes in the regulation of limit values for CO

10 emissions from passenger cars. The figure illustrates not only a clear move towards increasingly stricter standards (i.e. evidence of a race to the top rather than a race to the bottom), but also at least by graphical impression an increase in the similarity of national standards between 1970 and 2000. This impression, however, is partially contradicted when looking at changes in the variation coefficient. This indicator suggests a diverging development between 1970 and 1990. It is only for the period between 1990 and 2000, that we can observe a decrease in variation. Moreover, it is obvious that statements of convergence or divergence are only possible for the whole sample, but not for the performance of single countries. A similar judgement applies for our second example that refers to limit values for industrial discharges of zinc into the surface water. Also in this case, the graphical illustration not only indicates an increase in regulatory strictness, but also in the similarity of national standards over time. Again, however, the impression of policy convergence is not confirmed by the variation coefficient that increases throughout the whole period of investigation, hence suggesting a diverging rather than converging development. Figure 3: National Policy Development for Industrial Discharges of Zinc into Surface Water, 1970-2000 Limit values: Surface water (industrial discharges of Zinc in mg/l) 1970-2000 AUT 10 BUL FRA 8 GER HUN JAP 6 MEX POL 4 SWI UKD 2 0 1970 1980 1990 2000 Coeff. Var. 0,00 0,43 0,71 1,20 Mean (mg/l) 5,00 4,00 2,75 2,47 N 1 3 6 10

11 The Pair Approach To overcome the above-mentioned weaknesses in the analysis of convergence, we rely on an alternative indicator that is defined on the comparison of country pairs. According to this approach, the basic unit of analysis are not countries, but country pairs. Consequently, convergence implies an increase of policy similarity between a certain pair of countries over time. The hypotheses formulated above and specified below hence refer to country pairs; i.e., the more two countries are interlinked institutionally and economically, the more their policies will converge. To calculate pair similarity and pair convergence, for each policy, all country pairs are compared. This extends the number of cases from 24 countries to 276 country pairs (with each pair being considered only once, e.g. Austria-Belgium, but not Belgium-Austria). When comparing the presence of policies and similarities of instruments, score "1" means that two countries A and B have the same policy or the same instrument, whereas "0" means that they are dissimilar 3. For setting items, by contrast, we apply a normalized metrical score based on differences between limit values, leading to a similarity scale between "1" (limit values are identical) and "0" (country pair with the most dissimilar setting values). When summing up the similarity scores across all 40 policies, we apply a weighted scale, differentiating between the similarity scores for policy presence, instrument and setting items. The maximum similarity score is "1" for policy presence items, "2" for instrument items and "3" for similarity of settings. This weighting is based on the fact that similarity of policy settings necessarily implies that countries apply similar instruments. At the same time, similarity of instruments presumes the presence of a respective policy. Given our sample of 40 policy items, we thus arrive at a weighted similarity scale from "0 to 89" points. For a better interpretation and for reasons of comparability with other scales, these scores are transformed to percentage scales, with the maximum of 89 points corresponding to 100% similarity of environmental policies. These similarity scores can easily be transformed to convergence scores. Convergence is measured by changes in percentage points of absolute similarity between t 1 and t 2. In a later step, we analyze not only the whole sample of 40 items, but also the subgroups "policy presence" and "settings". For policy presence, the unweighted score ("0" or "1") is used, while for settings, we rely on the unweighted metrical score between "0 and 1". No further attention will be paid to convergence on policy instruments, as for the latter, the limited number of cases does not provide a sufficient basis for statistical analysis. This means that by the pair-approach, we assess a type of convergence that is similar to σ convergence, as far as we focus on the growing together of countries, regardless of the direction of 3 If both countries have no policy, they are also dissimilar by definition, as we do not have 0 on both sides, but an "empty set" that cannot be compared. If a policy would be abolished in both countries, than nonexistence of policy would be interpreted as similarity. Empirically however, this is not the case in our data set.

12 the policy changes. In contrast to σ convergence, however, the pair approach has the advantage of having a reference point for analyzing the convergence of single country pairs to the whole group. In addition, by the use of country pairs, the sample size becomes large enough for running regression models. Therefore, the data is not pooled, and difficulties commonly associated with pooling can be avoided. Moreover, an assessment of convergence is possible on the basis of nominal data and convergence effects emerging from the adoption of policies across countries are taken into account. This way, the pair approach offers an innovative and direct access to the study of sigma-convergence, measuring the increase or decrease of policy similarity between countries on the bilateral level that is the most basic one, without recourse to any aggregation. Descriptive Analysis The descriptive application of the pair approach is illustrated in figures 4 and 5. Figure 4 shows convergence trends for three items, namely limit values for lead in petrol, CO emissions of passenger cars and industrial discharges of zinc into surface water. While we can observe similarity increases for all of these items during the observation period, both similarity levels and similarity changes differ across policies and over time. Figure 4: Similarity of Average Country Pairs for Selected Policies, 1970-2000 100,00% 90,00% 80,00% 70,00% 60,00% 50,00% 40,00% 30,00% Lead in petrol C0 emissions Zink discharges 20,00% 10,00% 0,00% 1970 1980 1990 2000

13 Moving from single policies to the aggregate level of all 40 policies under investigation, the overall impression of a manifest convergence trend is quite similar. Figure 5 shows a permanent increase of average similarity of all country pairs over time, with slight convergence during the 1970s and strong convergence during the last decade, implying that in 2000, the average country pair reaches the level of 57% similarity. Moreover, figure 5 illustrates convergence trends for five country pairs. Having a reputation for early and strict environmental regulation, Japan/Sweden has been the most similar pair in 1970 as well as in 1980. In 1990 and 2000, by contrast, the pairs with the highest similarity are Germany/Switzerland and Denmark/Netherlands. Figure 5: Similarity of Selected Country Pairs and Average Pair, 1970-2000 100,00% 90,00% Similiarity Percentage 80,00% 70,00% 60,00% 50,00% 40,00% 30,00% 20,00% AVE JAP SWE UKD USA DEN NED GER SWI HUN POL 10,00% 0,00% 1970 1980 1990 2000 An overall summary of our empirical findings is provided in table 1. The table gives information about policy similarity for 276 country pairs with respect to four points of time (1970, 1980, 1990 and 2000). Moreover, similarity increases between these points are listed (indicating policy convergence). With respect to the policies, the table not only lists developments for the whole sample, but also provides results for different policy subgroups. In this respect, several findings seem of particular interest. First, convergence effects are most pronounced with regard to policy presence items followed by convergence on instruments, with convergence on settings being least developed. Second, when looking at policy types, it becomes apparent that similarity increases for trade-related policies are more pronounced than for policies not related to trade. This gap, however, is much stronger for setting items, while for policy presence items differences are very small. A similar statement applies for the

14 distinction between obligatory and non-obligatory items, with similarity increases being generally stronger in the former case. Table 1: Policy Similarity and Policy Convergence: Mean Values in % for 276 Country Pairs Policy Similarity Policy Convergence 1970 1980 1990 2000 1970s 1980s 1990s Pair-approach all policies (40 items) 0,02 0,11 0,27 0,57 0,09 0,16 0,30 Settings (21) 0,00 0,07 0,18 0,39 0,07 0,11 0,21 Policy instruments (12) 0,05 0,19 0,32 0,49 0,13 0,13 0,16 Presence of policy (40) 0,03 0,12 0,30 0,65 0,09 0,17 0,35 Setting trade-related (16) 0,01 0,06 0,16 0,46 0,05 0,11 0,29 Setting non trade-related (5) 0,00 0,00 0,06 0,17 0,00 0,05 0,12 Presence of policy trade-related (26) 0,02 0,14 0,34 0,72 0,12 0,20 0,38 Presence of policy non trade-related (14) 0,05 0,09 0,22 0,52 0,04 0,13 0,31 Setting obligatory (3/8/11/12) 0,02 0,13 0,27 0,54 0,12 0,15 0,26 Setting non-obligatory (18/13/10/9) 0,00 0,01 0,05 0,18 0,01 0,04 0,13 Presence of policy obligatory (3/8/13/19) 0,05 0,32 0,55 0,75 0,27 0,23 0,20 Presence of policy non-obligatory (37/32/27/21) 0,06 0,07 0,18 0,56 0,00 0,11 0,38 3.2 Independent Variables In the following, we discuss the operationalization of independent variables related to the mechanisms identified above (international harmonization, transnational communication and regulatory competition). The set of independent variables is composed of two types of variables. Changes in similarity of policies over time are not only explained by change rates, such as the accession to international institutions or changes in the income level but also by variables that refer to absolute figures at the beginning of the decade, conceived as "potentials" that are supposed to cause convergence in the following years, such as communicative potential through membership in international institutions or the volume of bilateral trade. Both harmonization and communication are based on the effects of international institutions. A sample of 35 institutions (see the list provided in annex 2) is used for the institutional indicators.

15 Basically, there are two types of international institutions relevant in the field of environmental policy. The first type represents international environmental treaties, protocols or regimes. The harmonization effects of these institutions arise from accession: Members ratify the treaty and have to comply immediately. This implies that convergence effects from accession occur only once. The second type has these "harmonization effects through accession " as well; however, it also has "harmonization effects through membership": The institution continuously produces regulatory output that leads to enduring and steadily renewed harmonization effects. Hence, in contrast with accession effects membership in t 1 will imply convergence in subsequent periods. International Harmonization To measure the extent to which countries are interlinked with international institutions with obligatory potential, we distinguish between two variables in order to cover both "membership effects" and "accession effects". The first variable refers to convergence effects of membership in international institutions with obligatory potential. From the international institutions relevant in the environmental field, only the EU has this opportunity of internal harmonization by adopting rules that are obligatory for its members. Hence, we use a dummy variable of EU membership to test the following hypothesis: (H 1.1): If two countries are member of the EU in t 1, the environmental policies of both countries will convergence in the following period The second variable takes account of the "accession effects". To measure these effects, we collected data on the membership of countries in the above-mentioned 35 international organizations and regimes. To grasp differences in their potential convergence effects, the institutions are weighted in two ways. First, we consider differences in their encompassingness, i.e., the number of policy areas, the scope of environmental problems, as well as the relative importance of environmental issues compared to other problems covered by the institution in question. Second, we differentiate between the institutions' obligatory potential. We distinguish three types of international law (supranational law, international hard law and international soft law) and four levels of monitoring (existence of courts, specialized monitoring agencies, reporting duties of member states, traditional instruments of diplomacy) in order take account of the fact that institutions might differ with respect to their capacity to enforce the compliance with their rules. The changes in this weighted membership score between t 1 and t 2 are taken as a proxy for convergence effects through accession to international institutions. The exact calculation of the indices can be found in annex 3. (H 1.2): The higher the score of common institutional accession (weighed by obligatory) potential of a country pair from t 1 to t 2, the more the environmental policies of both countries will converge towards each other during the same period

16 Transnational Communication Potential effects of international institutions on the convergence of national policies are not restricted to the adoption of legally binding rules. Rather these institutions represent an arena for communication and information exchange among their members an aspect that can trigger crossnational convergence by itself and independent of an institution's regulatory output. These institutional effects that exceed effects of harmonization are included in the variable "communication effects through membership". With this variable, we grasp effects of membership in international institutions in t 1 on policy convergence between t 1 and t 2. In contrast to harmonization, accession effects are not considered in a separate variable. This is based on the assumption that convergence effects emerging from transnational communication will not unfold immediately with accession, but become relevant only after a certain period of membership. The variable is based on membership data of the international institutions listed in annex 2. Similar to the operationalization of accession effects in the case of international harmonization, we rely on weighted institutional membership data. First, we again differentiate between levels of institutional encompassingness as outlined above. Second, we consider the length of a country pair's membership, assuming that convergence effects of transnational communication increase with membership duration. Third, we include four additional indicators to grasp the communicative potential of an institution, namely the frequency of annual meetings of national representatives, the number of organisational bodies where national representatives meet, the existence or non-existence of permanent national representatives and the size of permanent staff of the institution in question 4 (see annex 3). (H 2): The higher the score of common institutional membership (weighed by communicative potential) of a country pair in t 1, the more the environmental policies of both countries will converge towards each other in the following period Regulatory Competition Theories of regulatory competition commonly refer to the trade figures as a proxy for the economic interlinkage between countries that is supposed to lead towards an increase in policy similarity over time. Trade flows are usually captured by an index of trade openness. The index based on a country's import and export figures divided by the size of its internal market (size of GDP). This approach, however, has often been criticized, as capturing nothing more than the difference between large and small states. 4 Due to constraints in data accessibility, these indicators are based on the current state of the institutions in question; i.e. we do not consider potential changes, for instance in staff size or frequency in meetings, over time.

17 For the pair-approach, a different type of trade data is used, namely bilateral trade flows. We therefore are able to better capture the competitive situation between two national economies. In so doing, we add the absolute figures for import and export flows for each country pair, relying on trade data from the OECD statistical compendium. We include only trade flows between market economies, as trade between non-market economies (e.g. CEE countries before the 1989) is not expected to unfold effects focused upon by theories of regulatory competition. 5 (H 3): The higher the trade flows between a country pair in t 1, the more the environmental policies of both countries will converge towards each other in the period between t 1 and t 2 Control Variables It is the central objective of this article to analyze the extent to which international economic and institutional interlinkages among countries lead to similarity increases in their environmental policies. In focusing on these factors, however, we do not imply that other convergence causes might not be relevant. In the following, we therefore discuss other explanatory factors of convergence mentioned in the literature. This way, we are able to make a more comprehensive judgment about the relative explanatory power of the international convergence causes in which we are primarily interested. A first variable can be derived from the theoretical literature on policy diffusion, policy transfer and policy learning (cf. Holzinger and Knill 2005, Simmons and Elkins 2004). In these theories, emphasis is placed on two factors that facilitate the transfer and emulation of policies, hence leading to increases in cross-national policy similarity over time, namely cultural ties between countries (e.g. in terms of language or religion) and geographical proximity. These factors are expected to work even in the absence of strong international institutional interlinkages between countries. One important advantage of the pair approach is the possibility to directly include aspects of cultural similarity and geographical proximity in the quantitative analysis. Both aspects are assessed in a variable that is based on the sum of three different indicators: (1) the existence of a common border between a country pair, (2) the sharing of a common language and (3) the existence of common historical and religious tradition, expressed by the similarity of religion for each country pair. We conceive this variable as static, assuming that cultural ties and geographic proximity constitute highly stable factors that are not subject to far-reaching changes over time (see annex 3). (H 4): The higher cultural similarity between a country pair the more the environmental policies of both countries will convergence towards each other 5 Instead of focusing on absolute values of trade flows, one could also think of relying upon changes in trade figures over time. However, as the empirical correlation between both types of data is very high (r around.90), we only consider absolute data.

18 A second control variable refers to the well-documented relationship between the level of economic development of a country and the comprehensiveness and strictness of its environmental policy. In theoretical terms, this relationship is often emphasized by reference to the Kuznets curve. It is expected that there is a u-shaped relationship between per capita income (as a measure of economic development) and environmental quality. For the pair approach, we take the income figures of the poorer country in order to measure the economic development of a country pair. Data on GDP per capita is taken from the OECD statistical compendium. In addition, we consider changes in income levels over time. (H 5.1): The higher the income level of the poorer country in a country pair in t 1, the more the environmental policies of both countries will convergence towards each other in the following period (H 5.2) The higher changes in income level of the poorer country in a country pair between t 1 and t 2, the more the environmental policies of both countries will convergence towards each during the same period A final control variable to be considered refers to the existence of domestic political pressures towards a comprehensive and stringent environmental policy. To measure this pressure, one can conceive of many indicators, including, for instance, the number of environmental organizations or the public opinion data on environmental awareness. In view of their important role in bringing about political changes in domestic environmental policy, we focus on the political influence of green parties. Political influence of green parties can also be seen as a very general proxy for environmental problem pressure that is otherwise difficult to grasp for environmental policy in general. The influence of green parties is measured by three aspects, electoral success, membership in parliament, and participation in government (see annex 3). A high score is attributed to a country pair, if green parties are equally highly influential in both countries. We consider not only absolute levels of influence, but also changes in green party influence over time. (H 6.1): The higher the influence of green parties in a country pair in t 1, the more the environmental policies of both countries will converge towards each other in the following period (H 6.2): The higher the changes in the influence of green parties in a country pair between t 1 and t 2, the more the environmental policies of both countries will convergence towards each other during the same period Finally, the effects of pre-existing similarity of polices on convergence in later periods will be investigated. A list of all independent variables and their descriptive statistics is given in table 2.

19 Table 2: Independent Variables Mechanism Variable Model Mean Std. dev Min Max Corr. International Harmonization Transnational Communication Common EU-Membership 1970s 0,04 0,19 0,00 1,00-1980s 0,10 0,30 0,00 1,00-1990s 0,20 0,40 0,00 1,00 - Common Accession to 1970s 9,93 4,56 0,00 25,43 0,81 international institutions, 1980s 13,60 7,22 0,00 32,05 0,81 weighted by Obligatory Potential 1990s 21,82 9,87 2,25 41,65 0,96 Common Membership in 1970 17,19 7,31 6,53 34,12 0,96 International Institutions, 1980 44,16 15,48 13,05 75,69 0,90 weighted by Communicative Potential 1990 78,07 26,19 22,93 124,82 0,84 Regulatory Competition Economic Development Bilateral Trade Data Income level GDP/capita Similarity Change in Income Level GDP/capita Similarity 1970 470323,10 1264901,87 0,00 10583918 0,80 1980 2561096,36 6663802,46 0,00 52105439 0,87 1990 5783849,95 14614787,54 0,00 137000000 0,95 1970 7453,63 6053,53 1158,00 23446,00 0,74 1980 9883,70 7893,28 1329,00 27672,00 0,71 1990 12089,06 9764,62 1531,00 38713,00 0,70 1970s 2430,07 1988,54 171,00 7926,00 0,43 1980s 2205,36 2018,22-95,00 11041,00 0,42 1990s 2177,49 2449,80-302,00 10073,00 0,73 Social Emulation Cultural Similarity All 0,64 0,76 0,00 3,00 - Political Pressure Pre-existing Similarity Influence of Green Parties Change Influence of Green Parties Policy Similarity of All Policy Items 1970 0,00 0,00 0,00 0,00-1980 0,02 0,15 0,00 1,00-1990 0,41 0,71 0,00 2,00-1970s 0,02 0,15 0,00 1,00-1980s 0,39 0,71-1,00 2,00-1990s 0,10 0,87-2,00 3,00-1970 2,28 3,58 0,00 21,35-1980 11,39 9,91 0,00 37,89-1990 27,21 14,67 2,25 60,60-4 Model and Results 4.1 Model and Estimation Three cross-sectional models structure the following analysis of policy convergence, one for each decade of the observation period from 1970 to 2000. Each cross-section includes observations on 276 country pairs. We apply simple regression models with standard OLS estimations. As introduced above, the dependent variable for all models is the change rate of policy similarity from the beginning of the decade to its end, indicated in percentage points. Starting with the most general dependent

20 variable that includes a representative sample of 40 environmental policy items, this model is complemented by ten other dependents for diverse subgroups of policies. Through the inclusion of nine independent variables, we hope to cover all major explanatory mechanisms. A more serious problem for this type of analysis is found in multicollinearity of independent variables that may occur in various contexts. Trade figures, for example, correlate with EU-membership, and wealthier countries are assumed to be institutionally interlinked to a higher degree. Although correlations in the dataset are not particularly high (from.01 to.6), the variance inflation factors indicate significant problems for a stable estimation of coefficients. This problem is solved by the orthogonalization of those independent variables that are correlated to a significantly high degree. Based on a two-step regression approach and in accordance with theoretical considerations, explanatory variables are brought into a hierarchical order (figure 6). For subordinate variables, only the unstandardized residual from bivariate regression with their respective higherranking variable is included in the equation. The common effect is entirely attributed to the variable higher in ranks. The orthogonalization has to be taken into account for the interpretation of coefficients; for example, only those effects are captured by the residualized trade variable that go beyond the effect of common EU-membership. The correlation of the original variables with their residuals can be seen in the last column of table 2. Figure 5: Orthogonalization of Independent Variables

Table 3: Regression Results 21 1970s Mechanism Variable All Policies Presence of policies All Trade- Non-trade Related related Obligatory Non- obligatory Settings All Trade- Non-trade Related related Obligatory Constant -** -** -** -** -** -** -** -** -** -** -** Harmonization EU-Membership,251**,272**,287** -,045,253**,134**,248**,243** -,185**,167 Accession to Institutions,255**,224**,245** -,080,289**,062,254**,257** -,275** -,009 Transnational Communication Institutional Membership,372**,409**,386**,151**,386**,209**,288**,294** -,310**,004 Regulatory Competition Bilateral Trade -,052 -,017 -,024,040,013 -,004 -,096* -,100* - -,079 -,065 Social Emulation Cultural Similarity,067 -,015,001 -,094,086* -,104**,255**,298** -,259**,058 Income Level GDP per Capita,436**,436**,432**,042,364**,306**,169**,070 -,092*,251** Changes in GDP,086*,058,012,225**,116**,038,148**,155** -,101**,177** Political Pressure Existence of Green Parties - - - - - - - - - - - Changes Green Parties -,055 -,094* -,101*,013 -,084* -,112* -,005 -,067 -,049 -,148** R² (corr),46,45,44,07,47,12,36,35,00,34,09 N 276 276 276 276 -** -** 276 276 276 276 276 Non- obligatory 1980s Constant -** -** -** -** -** -** -** -** -** -** -** Harmonization EU-Membership,094**,077*,008,238**,099*,035,076,104**,070,117** -,086 Accession to Institutions,320**,364**,357**,097*,345**,260**,232**,220**,253**,133**,267** Transnational Communication Institutional Membership,515**,536**,522**,159**,378**,492**,349**,359** -,021,429**,004 Regulatory Competition Bilateral Trade -,021 -,091* -,081 -,058 -,248**,057,112**,128**,094,104**,065 Social Emulation Cultural Similarity,158**,110**,104**,043,190**,013,254**,237**,061,169**,266** Income Level GDP per Capita,013,037,117** -,236** -,001,067 -,018 -,036 -,150** -,014 -,025 Changes in GDP -,120** -,116** -,072 -,168**,011 -,177**,011,030 -,114** -,029,059 Political Pressure Existence of Green Parties,090*,034,080* -,149**,032,026,158*,215** -,035,069,229** Changes Green Parties -,064 -,108** -,062 -,174** -,038 -,126** 0,95*,176** -,127**,137** -,042 R² (corr),46,47,45,19,36,36,39,46,10,33,22 N 276 276 276 276 276 276 276 276 276 276 276 1990s Constant -** -** -** -** -** -** -** -** -** -** -** Harmonization EU-Membership -,099*,017 -,050,142**,120** -,092 -,116** -,126**,076 -,014 -,191** Accession to Institutions,508**,432**,378**,311**,380**,360**,414**,404**,214**,373**,240** Transnational Communication Institutional Membership -,569** -,458** -,458** -,062 -,363** -,324** -,518** -,528**,296** -,510** -,235** Regulatory Competition Bilateral Trade -,078* -,099* -,048 -,113* -,144** -,005 -,012,011 -,124** -,123**,139** Social Emulation Cultural Similarity -,076 -,091* -,066 -,063 -,131** -,008 -,049,065 -,066 -,115**,069 Income Level GDP per Capita -,006,004 -,195**,410** -,212**,221**,159**,129*,317**,191**,034 Changes in GDP,052,011,050 -,082,094* -,080,163**,128**,005,271** -,069 Political Pressure Existence of Green Parties,060,117**,097**,056,047,129** -,129 -,137**,088 -,145** -,026 Changes Green Parties,099**,184**,090**,210**,181**,093 -,042 -,057,109* -,005 -,071 R² (corr),55,43,44,20,45,24,35,34,23,38,16 N 276 276 276 276 276 276 276 276 276 276 276 ** = significant at the level of p< 0,05 * =significant at the level of p< 0,10. Standardized coefficients for OLS-estimation, with partially orthogonalized independent variables