An Assessment of the Europe Agreements Effects on Bilateral Trade, GDP, and Welfare

An Assessment of the Europe Agreements Effects on Bilateral Trade, GDP, and Welfare Peter Egger and Mario Larch 2nd July 2007 Abstract The so-called Europe Agreements had been enacted more than a decade ago to initiate the integration of goods markets between the 15 EU incumbent economies as of 1995 and 10 potential entrants located in Central and Eastern Europe. This paper evaluates the trade, GDP, and welfare effects of these agreements by means of structural estimation of a bilateral trade flow model. The results support four conclusions. First, the agreements stimulated significant convergence in multilateral resistance terms between the EU15 and the 10 Central and Eastern European countries (CEEC) brought about by bilateral tariff reductions. Second, they exerted significant positive effects on goods trade between the EU15 incumbents and the CEEC. Third, EU15 GDP responded by an increase of about one percent while that in the 10 CEEC increased by approximately 1.5 percent in response to the agreements. Fourth, the effects on welfare were moderate in the EU15 but amounted to more than three percent in the involved CEEC. Key words: Gravity model; Europe Agreements; Structural estimation; GDP and welfare effects of trade agreements JEL classification: F14; F15 Acknowledgements: To be added. Affiliation: Ifo Institute for Economic Research, Ludwig-Maximilian University of Munich, CESifo, and Centre for Globalization and Economic Policy, University of Nottingham. Address: Ifo Institute for Economic Research, Poschingerstr. 5, 81679 Munich, Germany. E-mail: egger@ifo.de. Affiliation: Ifo Institute for Economic Research and CESifo. Address: Ifo Institute for Economic Research, Poschingerstr. 5, 81679 Munich, Germany. E-mail: larch@ifo.de.

1 Introduction Europe Agreements with ten Central and Eastern European countries (CEEC) had been enacted in 1994 (Hungary and Poland), 1995 (Bulgaria, Czech Republic, Romania, and Slovak Republic), 1998 (Estonia, Latvia, and Lithuania), and 1999 (Slovenia) with the European Union member states (EU15) to prepare these economies for a possible enlargement later on. The main goal of the agreements was the (partial) elimination of tariffs among the EU15 countries and the involved CEEC on a bilateral level to facilitate goods market integration in advance of political integration (see the original texts at the webpage of the European Commission and Inotai, 1995, for an early reflection on them). 1 Apart from their effects on bilateral trade flows, what was the effect of these agreements on the involved countries multilateral trade? Moreover, how did GDP and welfare in these economies respond to the bilateral liberalization of political trade barriers? These questions obviously demand for an econometric framework that lies beyond reduced form estimation. This paper strives for an answer along the lines of a structural model of bilateral trade, following Anderson and van Wincoop (2003). The framework exploits information contained in zero as well as non-zero trade flows along the lines of Santos Silva and Tenreyro (2006). Our main goal with that approach at hand is to evaluate the effects of the Europe Agreements on three economic aggregates: bilateral goods trade flows within and across country blocs in Europe, GDP, and welfare. 2 Obviously, reduced form models would not support such an analysis, since it would not be possible to estimate multilateral effects of bilateral variables in a way consistent with general equilibrium. Our findings can be summarized as follows. Both the Interim Agreements and the 1 The Europe Agreements have been preceded by so-called Interim Agreements, signed in the years 1992 (Czech Republic, Hungary, Poland, and Slovak Republic), 1993 (Bulgaria and Romania), 1995 (Estonia, Latvia, and Lithuania), and 1997 (Slovenia). The purpose of the Interim Agreements was similar to that of the Europe Agreements: alleviating trade policy barriers to goods trade among the EU15 and the mentioned CEEC. With Hungary (in 1988), Poland (1989), and Czechoslovakia (1990) the EU signed trade agreements even before the Interim Agreements. However, these agreements were less compulsory than the Interim and Europe Agreements (see Winters, 1992; Sapir, 1994; Ramsey, 1995; and Beach, 1997). Richard Baldwin (1994, p. 125) refers to the Interim Agreements as Europe Agreements minus political provisions. 2 See Greenaway (1998) and Greenaway and Milner (2002) for surveys on the possible effects of bilateral trade policy on trade, GDP, and welfare. 2

Europe Agreements fostered trade between the EU15 and the CEEC. While the agreements reduced trade within the EU15 area and within the bloc of CEEC, according to our estimates, they raised trade by more than 100 percent between the EU15 and the CEEC. The latter triggered estimated effects on GDP of about one percent within the EU15 area and ones of more than 1.5 percent for the involved CEEC. The estimated cumulative welfare effects measured by the equivalent variation were even larger and amounted to about two percent for the EU15 and to more than three percent for the involved CEEC. The remainder of the paper is organized as follows. The next section introduces the empirical model. Section 3 describes the data, reports on the estimation results, and summarizes the quantification of the trade, GDP, and welfare effects associated with the Europe Agreements and the Interim Agreements. The last section contains concluding remarks. 2 The empirical model Anderson and van Wincoop (2003) derive a bilateral new trade theory gravity model which takes into account general equilibrium effects of trade barriers (see also Feenstra, 2004, for a discussion of that model). In particular, that model allows for the evaluation of trade barrier changes by considering both direct effects on bilateral consumer prices and indirect ones on the overall consumption bundle of consumers in that market. With symmetric trade barriers between countries i and j (t ij = t ji ), aggregate nominal exports (X ij ) may be expressed in the following way: 3 X ij = y iy j t 1 σ ij P σ 1 i P σ 1 j, (1) y W where y i N j (X ij),y j N i (X ji) denotes nominal GDP of countries i and j, and y W N N i j (X ij) is world GDP, σ is the constant elasticity of substitution among products/varieties, and P 1 σ i,p 1 σ j 3 X ii refers to aggregate goods delivery to domestic consumers. are so-called multilateral resistance terms, which are 3

proportional to the consumer price indices obtained in models with love-of-variety preferences as derived in Dixit and Stiglitz (1977) and used in Anderson (1979), Krugman (1980), Bergstrand (1985), Eaton and Kortum (2002), and Anderson and van Wincoop (2003). Even though multilateral resistance terms are unobserved, they can be obtained as solutions to the system of nonlinear equations of the form 4 P 1 σ i = N j=1 θ j = y j y W ( P σ 1 j j. θ j t 1 σ ji ) i (2) Assume that trade costs are defined as t ij e z ijβ, where zij is a K 1 vector of trade barriers variables and β is K 1 vector of parameters relating the elements of z ij to t ij. In particular, let us assume that z k,ij corresponds to the Europe Agreement membership indicator for country-pair ij, which is set to unity if such an agreement is in place and to zero, otherwise. What is the impact of the Europe Agreements on bilateral normalized trade, nominal GDP, and welfare? For this, we need to compare the de facto situation cum Europe Agreements reflected by z k,ij to an unobserved counterfactual situation with z c,k,ij = 0. For this, let us define the counterfactual trade friction vector z ij = [z 1,ij,...,z c,1,ij,...,z K,ij ] corresponding to a ceteris paribus change in the indicator variable of interest. Note that, in the counterfactual, we can not simply solve for counterfactual multilateral resistance terms P 1 σ c,i from (2). The reason is that counterfactual 4 Recently, Baier and Bergstrand (2006) proposed two versions of a linear approximation of the nonlinear model of Anderson and van Wincoop (2003). For instance, the linearized model is applied by Carrère (2006) in an interesting paper on the effects of regional trade agreements in general on bilateral trade. It turns out that the linearized model can not be applied to answer the questions of interest here for two main reasons. First, it does not support estimation of GDP and welfare effects of (neither natural nor political) trade barrier variables. Second, the Europe Agreements entail a fairly large change in trade policy, while the linearized model only works well for marginal changes. Apart from that, a necessary condition for the linearized model to be applicable in our context would be that the standard deviation in bilateral trade costs is negligible. In our data-set, the mean of log bilateral distance is 6.97, while the standard deviation is 0.78 (the minimum is 2.95 and the maximum amounts to 8.33). Under such conditions and a reasonable elasticity of substitution parameter of about σ = 10 (see Feenstra, 1994; Anderson and van Wincoop, 2003, 2004), the linearized model leads to a bias of the effects on trade flows of more than hundred percentage points (this conclusion is based on the results from a large-scale Monte Carlo study conducted by the authors, which is available upon request). 4

trade costs t c,ij e z c,ijβ are known while counterfactual GDP-shares θc,j are not. However, under the assumption of an endowment economy we can obtain a solution for both P 1 σ c,i and θ c,i for all i and compute x,ij 100 t1 σ ij P σ 1 i P σ 1 j t 1 σ c,ij P σ 1 c,i P σ 1 c,j t 1 σ c,ij P σ 1 c,i P σ 1 c,j (3) as the percentage-response of GDP-normalized trade X ijy W y i y j to having t ij instead of t c,ij. Furthermore, we obtain the change on nominal GDP for given P 1 σ c,i y,i 100 θ 1 i 1 σ P 1 1 σ i θ 1 c,i θ 1 1 σ c,i P 1 c,i P 1 c,i and θ c,i as. (4) Finally, the associated change in welfare in terms of the equivalent variation is then determined as (see also Anderson and van Wincoop, 2001) e,i 100 [ ( θi θ c,i ) 1 1 σ ( P c,i P i ) 2 1]. (5) The subsequent analysis will focus on averages of these estimates particularly associated with the Europe Agreements for different groups of country-pairs/countries on the continent. 3 Empirical analysis 3.1 Data We use data on nominal bilateral trade flows in U.S. dollars from the United Nations World Trade Database among 30 European economies (see the Appendix for a list of countries) over the years 1990-2001. This covers all periods during which Interim or Europe Agreements had been enacted, including two years before the first Interim Agreement and two years after the last Europe Agreement was concluded. For both Interim and Europe 5

Agreements, we use a binary indicator variable to capture the respective direct effect on trade. In terms of the model outlined in Section 2, we should think of the corresponding parameter estimates as corresponding to β IA and β EA. 5 The other trade friction variables (i.e., elements of z ij in Section 2) are the log of bilateral distance and a set of indicator variables: a regional trade agreement dummy variable (unity if tariffs are preferentially eliminated among two trading partners and zero else); a common border indicator (unity in case of a common border and zero else); and a common language indicator (unity in case of a common official language and zero else). In this data-set, about 28 percent of the cells of the trade flow matrix contain zeros. We estimate the model (primarily) by Poisson pseudo-maximum likelihood for two reasons. First, it naturally copes with zero trade flows 6 and, second, one may account for heteroskedasticity. The latter is particularly important if trade flows in levels are measured with error. 3.2 Estimation We report three alternative sets of estimates: one is based on Poisson pseudo-maximum likelihood estimation (PMLE); a second one uses Negative-binomial PMLE; and a third one relies on OLS. All models include fixed country-by-time effects. Of course, this wipes out the multilateral resistance terms due to perfect collinearity, but we can retrieve these terms subsequently, since they only involve estimates of β k and the trade friction variables z k,ij for all k and ij along with GDP-shares θ i for all i. While the two PMLE models use the left-hand-side variable in levels-form, the OLS model uses the log thereof. The latter excludes all zero trade flows from the model. In any case, the parameters of the indicator variables are semi-elasticities and that one of the log distance variable is an elasticity (see Cameron and Trivedi, 1998). 5 There, we use subscripts IA and EA to refer to Interim Agreements and Europe Agreements, respectively. The two parameters should be interpreted as β IA = (1 σ)lnb IA and β EA = (1 σ)lnb EA, respectively, where b IA 1 and b EA 1 are the tariff/trade cost equivalents of the Interim and Europe Agreements. 6 Alternatively, Eaton and Tamura (1994) estimated a non-linear trade flow model in levels. See Santos Silva and Tenreyro (2006) about the pros and cons of non-linear estimation of gravity models in levels. 6

Table 1 Table 1 summarizes our findings. According to the pseudo-r 2 based on the loglikelihood ratio of the constant-only model and the preferred alternative, the Poisson- PMLE model performs best among the chosen ones. Yet, most of the coefficients of interest are very similar when comparing the Poisson PMLE with the Negative-binomial PMLE model. Even with OLS, most of the coefficients are even quantitatively similar. 7 Note that the Europe Agreement indicator variable is constructed such that the Europe Agreement effect is captured by the sum of the Interim and Europe Agreement parameters. The relative magnitude of these two parameters changes when using OLS as compared to one of the PMLE models. Hence, leaving out about a third of the observations may indeed be harmful for the quantification, here. It seems relatively unimportant of whether the Poisson or the Negative-binomial PMLE model is used for inference, instead. We base our subsequent results on the Poisson PMLE model. 8 Table 2 A crucial difference between simple OLS or fixed country-by-time effects models and the structural approach adopted here is that log bilateral trade flows are non-log-linear functions (rather than log-linear ones) of bilateral trade frictions by virtue of the multilateral resistance terms given in (2). Hence, the multilateral resistance terms will be of crucial importance for inference of the Interim and Europe Agreement effects on trade, GDP, and welfare. We summarize the mean and standard deviation of the multilateral resistance terms for different country groups and years in Table 2. The table reveals that 7 Obviously, the coefficient of common language takes on a negative sign in both the Negative-binomial and the OLS models. Also the magnitude of the log distance coefficient varies significantly across the estimated models. However, we illustrate in Section 3.4 that the order of magnitude of the overall effect of the Interim and Europe Agreements is not as large as some of the estimated parameters suggest. 8 In an extensive sensitivity analysis, we also pursued less parsimonious models by using further indicator variables such as colonial ties (various types thereof), a same country dummy, etc. (see Glick and Rose, 2002, for a set of such indicators). However, it turns out that with the chosen country sample and period coverage, the mentioned results are robust and the parsimonious models perform as well as the less parsimonious ones. Moreover, the parameter estimates for the agreement dummies of interest are virtually unaffected by the inclusion of the mentioned variables. 7

the difference in multilateral resistance terms between the EU15 and the ten CEEC that became new members of the EU in 2005/2007 (and were involved in Europe Agreements in advance) declined substantially over the reference period. In contrast, the difference between the EU15 and the three European Free Trade Area (EFTA) economies increased as did that between the EU15 and two other South-Eastern European economies. Apart from bloc-specific economic growth, the reduction of trade barriers associated with the Interim and Europe Agreements seems to be a prime candidate for that change. We will shed light on the economic relevance of these agreements in the following subsection. 3.3 Simulating the Europe Agreement effects on bilateral trade, GDP, and welfare In Table 3, we report means of the percentage changes of normalized trade as in (3), GDP as in (4), and welfare as in (5) for different country blocs. In the table, we focus on the Europe Agreements, taking the Interim Agreements as given. We will talk about a joint cumulative effect of the Interim and Europe Agreements later on. Table 3 The results support the following conclusions. As for the trade effects, the Europe Agreements between the EU15 and the ten CEEC reduced trade within the EU15 and among the CEEC that enacted a Europe Agreement in a particular period (trade diversion). At the same time, these agreements fostered trade between the EU15 and the involved countries substantially (trade creation). Overall, this led to positive effects on these countries trade, showing up in fairly small positive effects on GDP of the EU15 economies but more sizeable ones (of about one percent) for the involved CEEC. The welfare effects of these agreements are estimated to exceed the GDP effects by a factor of two (net trade creation). 9 9 The latter is consistent with endowment economy models of monopolistic competition in international trade. In these models, trade costs affect producer prices but not real production. Since they affect consumption, the welfare effects of trade frictions will be larger than the GDP effects. 8

As said before, we took the Interim Agreement effects as given in Table 3. Hence, the reported effects could be interpreted as period-specific ceteris paribus contributions of the Europe Agreements. Of course, it seems interesting to estimate the joint effect of the Interim and the Europe Agreements. Yet, a table similar to Table 3 would be hard to read, since Interim Agreements had been concluded in 1992, 1993, 1995, and 1997 while Europe Agreements had been enacted in 1994, 1995, 1998, and 1999. 10 Therefore, we focus on the cumulative effect of the Interim and Europe Agreements evaluated in the year 1999. Table 4 provides the corresponding estimates. Table 4 Overall, we estimate GDP effects that are about 50 percent higher for the involved country groups than the cumulative effects of the Europe Agreements alone (the latter would approximately correspond to the row sum in Table 3). Similar conclusions apply for the welfare effects. The effects on GDP are estimated at about one percent on the EU15 members and at about 1.5 percent on the involved CEEC. Our estimates of a moderate positive effect of the Interim and Europe Agreements together on EU15 GDP support early predictions of the CEEC s integration in the EU by Cadot, de Melo, and Messerlin (1995). 11 The estimated joint impact of the two types of agreements is less than twice that of the Europe Agreements, even though the parameter estimate of the Interim Agreements was larger than that of the Europe Agreements in Table 1. This is a consequence of the non-linearity of the model working through the multilateral resistance terms. Let us compare these figures with others work along the lines of calibrated general equilibrium (CGE) modeling. For instance, the magnitude of the impact on the CEEC s welfare relative to the EU15 is smaller than the one projected in Baldwin, Francois, and Portes (1997) for Eastern Enlargement of the European Union by means of a CGE model 10 Since both Interim and Europe Agreements were concluded in 1995, it is difficult to disentangle their effects in such a year. Therefore, it seems preferable to look at their joint impact. 11 However, their focus was on the net creation of jobs in selected countries. Instead, ours is on the net creation of trade as reflected in GDP and welfare effects, here. 9

(they did not consider the Interim and Europe Agreements in particular). However, the size of our estimated welfare effect on the CEEC is fairly close to their simulation of what they call the conservative (i.e., trade-liberalization-based) effect on real GDP, there. 12 Heijdra, Keuschnigg, and Kohler (2002) computed long-run welfare effects of the EU15 trade liberalization with the CEEC for Germany of less than one percent of GDP (in their Table 2). They projected long-run welfare effects for the EU15 economies that were positive for most of the countries and up to two percent for Austria. In line with those results, Kohler (2004) reports welfare effects of EU Eastern Enlargement on the EU15 economies of less than one percent of GDP for the EU15 economies. Accordingly, we may conclude that the pattern projected here is similar to CGE-based work on Eastern Enlargement of the EU: the CEEC gain relatively more than the EU15 incumbents from trade liberalization. However, there is a quantitative difference regarding the welfare effects estimated here as compared to earlier work. The cumulative joint welfare effects of the Interim and Europe Agreements are somewhat larger with our estimates than in previous CGE-based work. The reason for the latter lies in differences regarding the trade effects of bilateral trade liberalization between the EU15 and the CEEC. And it roots in the consideration of multilateral resistance in our analysis in both parameter estimation and the calculation of trade, GDP, and welfare effects. 3.4 Sensitivity analysis We undertook a sensitivity analysis to check the robustness of our findings along several lines. We summarize the findings in Table 5. Table 5 The first column of results in the table repeats the findings from Table 4 to facilitate the comparison. The remaining columns refer to sensitivity checks, numbered (i)-(v). In 12 In their Table 3 (p. 138), they report an effect of about 1.5 percent on real GDP of the CEEC, which is fairly close to our welfare estimate, but only one of 0.2 percent on EU15 GDP which is much less than the effects estimated in our ex-post analysis. 10

experiment (i), we compute the results based on the Poisson PMLE model in Table 1 and assume an elasticity of substitution of σ = 15 rather than σ = 10 as in the benchmark case. This leads to GDP and welfare effects that are smaller by about one third than the original ones. Anderson and van Wincoop (2003) refer to values between σ = 5 and σ = 10 as plausible ones. In our case, the quantitative impact of small changes in the elasticity of substitution around a value of σ = 10 leads to small deviations from the original results. In (ii), we include fixed country-pair effects in addition to exporter-time and importer-time effects in a Poisson PMLE model. 13 Again, the associated GDP and welfare effects are somewhat smaller than in the original model but the relative magnitude of the impact on the EU15 versus the 10 CEEC remains about the same. The results in (iii) rest upon the Negative-binomial PMLE parameters in Table 1 and obtains GDP and welfare effects that are comparable to those with Poisson PMLE when assuming that σ = 15 as in (i). Using the least-squares dummy variable model parameters of Table 1 instead of Poisson PMLE in (iv), the estimated GDP and welfare effects are somewhat larger on average than in the benchmark case. Finally, when adding dummy variables with Poisson PMLE in experiment (v) for colonial relationships between exporter and importer, for the exporter and the importer to have had a common colonizer, and for them to have once formed part of the same country, we obtain estimates of the GDP and welfare effects which are fairly close to the original ones. Overall, we may conclude that the findings are qualitatively insensitive to these changes and they are even fairly robust in quantitative terms. 4 Concluding remarks This paper delivers an assessment of the trade, GDP, and welfare effects of the Interim and Europe Agreements on the involved Western and Central and Eastern European economies. Econometrically, such an investigation has to rely on structural model estimation, which obeys general equilibrium effects. The paper, therefore, uses a framework 13 Econometrically, the model outlined by Mátyás (1997) is nested in the one proposed here. 11

recently proposed by Anderson and van Wincoop to approach the goal. We estimate a positive effect of both the Interim and the Europe Agreements on bilateral trade in a panel of 30 European economies and the period 1990-2001. We identify a significant convergence of multilateral resistance terms as a measure of a country s weighted multilateral trade costs between the EU15 and the 10 Central and Eastern European countries (CEEC) that had been involved in the agreements (and became EU members in 2005/07). One reason for this result lies in the elimination of political trade frictions (tariffs) between the EU15 and the CEEC over the mentioned period. The change in multilateral resistance terms together with the direct effects of the agreements triggered net trade creation effects on the involved economies. The latter shows up in cumulative GDP effect estimates on the CEEC in response to the agreements of more than 1.5 percent, under reasonable assumptions about the elasticity of substitution across produced varieties. Even the effects on the EU15 countries amounted to about one percent on average. The adopted approach supports inference of the welfare effects associated with the Interim and Europe Agreements. The estimated cumulative welfare effects of these agreements amount to more than three percent of GDP on the involved CEEC and to about two percent in the EU15 countries. Qualitatively, these results are insensitive to alternative functional form assumptions about the data generating process of the disturbances or to the inclusion and exclusion of trade friction and facilitation variables. References Anderson, James E. (1979), A theoretical foundation for the gravity equation, American Economic Review 69, 106-16. Anderson, James E. and Eric van Wincoop (2001), Borders, trade, and welfare, Brookings Trade Forum, 207-30. Anderson, James E. and Eric van Wincoop (2003), Gravity with gravitas: a solution to the border puzzle, American Economic Review 93, 170-92. 12

Anderson, James E. and Eric van Wincoop (2004), Trade costs, Journal of Economic Literature 42, 691-751. Baier, Scott L. and Jeffrey H. Bergstrand (2006), Bonus vetus OLS: a simple approach for addressing the Border Puzzle and other gravity-equation issues, unpublished manuscript, University of Notre Dame. Baldwin, Richard (1994), Towards an Integrated Europe, Centre for Economic Policy Research, London. Baldwin, Richard, Joseph Francois, and Richard Portes (1997), The costs and benefits of eastern enlargement: the impact on the EU and central Europe, Economic Policy 12, 127-76. Beach, Derek (1997), The negotiation of the Europe Agreements with Poland, Hungary, and Czechoslovakia - A three-level game, paper presented at the work shop on East-West Regimes After the Cold War at the Thorkil Kristensen Institute, Centre for East-West Research, May 15-16, 1997, South Jutland University Centre, Denmark. Bergstrand, Jeffrey H. (1985), The gravity equation in international trade: some microeconomic foundations and empirical evidence, Review of Economics and Statistics 67, 474-81. Cadot, Olivier, Riccardo Faini, and Jaime de Melo (1995), Early trade patterns under the Europe Agreements, European Economic Review 39, 601-10. Cameron, A. Colin and Pravin K. Trivedi (1998), Regression Analysis of Count Data, University Press: Cambridge. Carrère, Céline (2006), Revisiting the effects of regional trade agreements on trade flows with proper specification of the gravity model, European Economic Review 50, 223-47. Dixit, Avinash K. and Joseph E. Stiglitz (1977), Monopolistic competition and optimum product diversity, American Economic Review, 67, 297-308. Eaton, Jonathan and Samuel Kortum (2002), Technology, geography, and trade, Econometrica 70, 1741-79. Eaton, Jonathan and Akiko Tamura (1994), Bilateralism and regionalism in Japanese and U.S. trade and foreign direct investment patterns, Journal of the Japanese and International Economies 8, 478-510. Feenstra, Robert (1994), New product varieties and the measurement of international prices, American Economic Review, 84, 157-77. 13

Feenstra, Robert (2004), Advanced International Trade: Theory and Evidence, Princeton University Press, Princeton, NJ. Glick, Reuven and Andrew K. Rose (2002), Does a currency union affect trade? The time series evidence, European Economic Review 46, 1125-51. Greenaway, David (1998), Does trade liberalization promote economic development?, Scottish Journal of Political Economy 45, 491-511. Greenaway, David and Chris Milner (2002), Regionalism and gravity, Scottish Journal of Political Economy 49, 574-85. Heijdra, Ben, Christian Keuschnigg, and Wilhelm Kohler (2002), Easern enlargement of the EU: jobs, investment and welfare in present member countries, CESifo Working Paper No. 718. Inotai, András (1995), From the association agreements to full membership? The dynamics of relations between the Central and Eastern European countries and the European Union, paper presented at the Fourth Biennial International Conference of the European Community Studies Association, May 11-14, 1995, Charleston, SC. Kohler, Wilhelm (2004), Eastern enlargement of the EU: a comprehensive welfare assessment, Journal of Policy Modeling 26, 865-88. Krugman, Paul R. (1980), Scale economies, product differentiation, and the pattern of trade, American Economic Review 70, 950-59. Mátyás, László (1997), Proper econometric specification of the gravity model, World Economy 20, 363-9. Ramsey, Lynn E. (1995), The implications of the Europe Agreements for an expanded European Union, The International and Comparative Law Quarterly 44, 161-71. Santos Silva, João M.C. and Silvana Tenreyro (2006), The log of gravity, Review of Economics and Statistics 88, 641-58. Sapir, André (1994), The Europe Agreements: implications for trade laws and institutions. Lessons from Hungary, in L. Alan Winters (ed.), Foundations of an Open Economy: Trade Laws and Institutions for Eastern Europe, Centre for Economic Policy Research, London, 89-107. Winters, L. Alan (1992), The Europe Agreements: with a little help from our friends, in The Association Process: Making it Work, Centre for Economic Policy Research, 17-33. 14

Appendix Country coverage EU15: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, United Kingdom. 10 CEEC: Bulgaria, Czech Republic, Estonia, Hungary, Lithuania, Latvia, Poland, Romania, Slovak Republic, Slovenia. Other Western European countries (EFTA): Iceland, Norway, Switzerland. South-Eastern European countries: Croatia, Makedonia. 15

16

17

18

19