Does political disintegration lead to trade disintegration?

Transcription

1 Economics of Transition Volume 15(4) 2007, Does political disintegration Blackwell Oxford, ECOT Economics Journal XXX Original Political de 2007 sousa The UK compilation Disintegration Articles and Publishing of Authors Transition lamotte Ltd 2007 and Trade The European Bank for Reconstruction Development lead to trade disintegration? Evidence from transition countries 1 José de Sousa* and Olivier Lamotte** *CES, University of Paris 1 and University of Rennes 2, Boulevard de l Hôpital, Paris, France. jdesousa@univ-paris1.fr **ESG, 25 rue Saint-Ambroise, Paris, France. olamotte@esg.fr Abstract Recent studies have found that political disintegration is a cause of severe and rapid trade disintegration in former Eastern European countries. This finding somewhat conflicts with another strand of the literature highlighting the fact that trade patterns change relatively slowly. This article aims at reconciling the apparent inconsistency between these two results. Using a theoretically grounded gravity equation, we evaluate the intensity of trade between successor states of three former countries (Czechoslovakia, the Soviet Union and Yugoslavia) in the period We find no clear evidence that political disintegration leads to systematic and severe trade disintegration. This result is consistent with the patterns displayed by using simple descriptive statistics, is robust to sensitivity checks, and supports the idea of hysteresis in trade. JEL classifications: F13, F15. Keywords: Borders, trade disintegration, political disintegration, gravity equation, transition economies. 1 We thank two anonymous referees and the editor, Wendy Carlin, for many valuable comments which significantly improved the paper. We are grateful to Céline Carrère, Patrick Guillaumont, Fabian Gouret, Julie Lochard, Mathilde Maurel, Daniel Mirza, Richard Pomfret and Gianpaolo Rossini for helpful comments. We also acknowledge earlier drafts of Baldwin (2006) for many helpful insights and comments and Jasna Dimitrijevic for providing us with exchange rate data. The usual disclaimer applies.. Published by Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA

2 826 de Sousa and Lamotte 1. Introduction The number of countries in the world increased from 74 in 1946 to 192 in 1995 (Alesina, Spolaore and Wacziarg, 2000, p. 1276). Despite such a striking stylized fact on the importance of political disintegration, trade literature focuses extensively on the effects of integration. 2 Political disintegration may have large implications for international trade, however. It leads to the creation of new states and institutions, the introduction of new currencies, the erection of new boundaries as well as new tariff and non-tariff barriers to trade. The political disintegration processes in Eastern Europe offer us a unique opportunity to examine this issue. Using the border effect approach, 3 recent papers document severe and rapid trade disintegration following political disintegration in Eastern Europe (Djankov and Freund, 2002; Fidrmuc and Fidrmuc, 2003). 4 As an illustration, around the time of the Czechoslovak disintegration, the Czech Republic and Slovakia traded 43 times more with each other than with other countries. Five years later, this magnitude is around seven (Fidrmuc and Fidrmuc, 2003). Between 1994 and 1996, Russian regions traded 60 percent more with each other than with former USSR Republics, while there was no significant difference before disintegration (Djankov and Freund, 2002). These results conflict with other important research on trade, which highlights a tendency for established bilateral trade ties to change relatively slowly (Frankel, 1997, p. 126). Several reasons may explain such hysteresis in trade patterns. First, sunk costs associated with entering a foreign market (Baldwin, 1988), building trade networks and maintaining trade relationships lower the incentive for exiting the market. Therefore, countries with a history of trading with one another whether for reasons related to politics, policies and other factors generally continue to do so (Eichengreen and Irwin, 1997). Thus, Freund (2000) documents the fact that trade links between the original members of the European Union are highly persistent despite successive enlargements. Second, informal barriers to trade may constitute a good incentive to keep trading with historic partners. The lack of information about international trading possibilities (Portes and Rey, 2005) 2 As of November 2006, a search on Google Scholar yields 1,250 answers for integration and trade and 278 for disintegration and trade, mainly related to Fidrmuc and Fidrmuc (2003). A related search on regional integration and trade yields 15,200 answers vs. 77 for regional disintegration and trade. 3 The border effect approach, initiated by McCallum (1995), documents the fact that national borders create trade barriers that go beyond the existence of explicit trade restrictions. This simple approach captures all factors that might impede trade between two countries (Feenstra, 2004). As an example, McCallum (1995) finds that, in 1988, a given Canadian province trades 20 times more with another province than with an American state of similar size and distance. The magnitude of this impact is, however, questioned and halved in Anderson and van Wincoop (2003). 4 See Pomfret (2002) for a discussion on the disintegration of the market in Central Asia.

3 Political Disintegration and Trade 827 or a weak contract enforcement (Anderson and Marcouiller, 2002) increases the cost of trading with new partners. Such factors explain the crucial role of business and social networks in international trade (Rauch, 2001; Rauch and Trindade, 2002) and the persistence of trade relationships. In this article, we attempt to reconcile the apparent inconsistency between severe trade disintegration, following political separatism, and hysteresis in trade patterns. Using the three former countries of Eastern Europe (Czechoslovakia, the Soviet Union and Yugoslavia) as an experiment, we find no clear evidence that political disintegration leads to systematic and severe trade disintegration. Our finding is consistent with the patterns displayed by using simple descriptive statistics and supports the idea of hysteresis in trade. We provide evidence that the literature on the effect of political separatism on trade is incomplete for two reasons. First, it does not fully take account of Anderson and van Wincoop s (2003) lessons, that is, appropriate control for the multilateral resistance terms. Actually, the more resistant to trade with all others a region is, the more it is pushed to trade with a given bilateral partner (Anderson and van Wincoop, 2003). Second, the literature only relates to a limited sample of successor states in the Eastern European countries. For instance, Fidrmuc and Fidrmuc (2003), our main empirical reference, only study two countries out of five for the former Yugoslavia and 6 out of 11 for the former Soviet Union (hereafter FSU). We find that controlling for the multilateral resistance terms and using a more comprehensive sample is required for a better assessment of the extent of trade disintegration. The remainder of the article is organized as follows. In Section 2, we highlight the inconsistency between descriptive statistics and literature findings on the effect of political disintegration on trade. In Section 3, we discuss the empirical specifications and present the dataset. Then, in Section 4, we report and discuss the results on the evolution of trade intensity over time. In Section 5, we check the robustness of our results to Fidrmuc and Fidrmuc s (2003) specification and sample. Finally, we draw the conclusions in Section Political disintegration and trade: Revisiting the evidence Table 1 presents some summary statistics about trade patterns of the former Central and Eastern European countries (hereafter CEECs): Czechoslovakia, the FSU and the former Yugoslavia. The intra-national/international trade ratio indicates how intense trade relationships are between successor states (intra-national trade) relative to their trade with third partners (international trade). We observe differences both in the level of the ratio and its evolution between 1993 and In 1993, the highest ratio is found for the FSU (43.47 percent) and its evolution is rather stable until 1998 (43.54 percent). A sharp decline begins in 1998, seven years after the

4 828 de Sousa and Lamotte Table 1. Trade evolution of former countries (ratio of intra-national over international trade 1 ), in %, Czechoslovakia Soviet Union Yugoslavia All All Baltic States BLR UKR RUS All HRV SVN Notes: 1 The ratio is defined as trade within the former country (intra-national) over trade of the successor states of the former country with third partners (international trade). All means that all successor states are considered to compute the above ratio. For ease of comparison with the literature, we also compute the ratio on subsets of the former countries, such as Croatia Slovenia. However, to avoid any bias on the size of the trade reported on the denominator, we drop the remaining successor states of the former country considered. Note that BLR, Belarus; UKR, Ukraine; RUS, Russia; HRV, Croatia and SVN, Slovenia. Source: Authors calculation from the Chelem Cepii dataset (cf. infra for details). Soviet Union s political disintegration. Czechoslovakia is also quite integrated in 1993 but experiences a continuous decrease of its ratio from 33.5 to 11.2 percent in In contrast, the evolution of the former Yugoslavia s ratio is rather stable, with percent in 2001 vs percent in This ratio constitutes a rough and crude indicator of the evolution of trade patterns within each former country. They, however, simply illustrate that former countries do not experience similar patterns of trade, and political disintegration does not lead to systematic trade disintegration. If these statistics are suggestive, trade shares alone are not sufficient to get reliable conclusions about trade disintegration. Controlling for the standard determinants of trade, Fidrmuc and Fidrmuc (2003) (hereafter FF) assess the impact of political disintegration on trade among the former CEECs. They find that the intensity of trade relations fell sharply after disintegration (p. 820). Figure 1 plots their year-by-year estimates of the above normal level of trade between the successor states during the period FF s (2003) results confirm a trade disintegration pattern in the case of the former Czechoslovakia. They also find a decreasing trend in the case of Croatia and

5 Political Disintegration and Trade 829 Figure 1. Disintegration episodes in Central and Eastern Europe, , Fidrmuc and Fidrmuc (2003) Slovenia for the period The other two cases, plotted in the lower part of Figure 1, do not suggest a trade disintegration story, however. In contrast, our statistics indicate continuous trade integration for the Baltic States since As for the Belarus Ukraine Russia relationships, trade disintegration seems to take place only after FF s (2003) period of investigation. A more detailed comparison of FF s (2003) results with the descriptive statistics on raw data and our results is made in Section Empirical specification 3.1 Benchmark specification The empirics are based on the border effect methodology. To evaluate the impact of the political border, McCallum (1995) estimates a simple empirical log-linear specification: ln( X ) = a + a ln( Y ) + a ln( Y ) + a ln( Dist ) + a ( Home ) + ε, ijt 0 1 it 2 jt 3 ij 4 ij ijt (1) where i and j denote countries, t denotes time, and the variables are defined as: (X ij ) is exports from country i to country j; (Y i ) and (Y j ) are the gross domestic products of countries i and j; (Dist ij ) is the distance from i to j; (Home ij ) is a dummy variable equal to 1 for trade between successor states of former countries and 0 otherwise; 5 and (ε ijt ) is the usual error term controlling for the possible errors in measuring trade flows.

6 830 de Sousa and Lamotte According to the gravity Equation (1), trade between a pair of countries is proportional to their respective incomes, and inversely proportional to their bilateral distance. Thus, estimated coefficients on national incomes are expected to be positive and coefficients on distance negative. The estimate of (Home ij ) is expected to be positive if, other things being equal, trade flows among successor states are larger than trade flows between successor states and their outside partners. The antilog of the estimated coefficient of (Home ij ) gives the size of the trade intensity within the former country, called the former country bias. Equation (1) is simple and easily interpretable but suffers from some methodological problems. First, a crucial variable is omitted. Actually, after controlling for incomes and bilateral distance, country pairs with remote alternative partners trade more than country pairs with close alternative partners. This relative distance effect, also called relative price effect or multilateral resistance effect (Anderson and van Wincoop, 2003), is due to price differences across countries. A country with a central location receives higher aggregate demand for its output than a peripheral country, resulting in higher export prices and lower trade with a given partner (Harrigan, 2003). This relative distance effect can be approximated by the use of a-theoretic remoteness variables (Wei, 1996), complex non-linear least square estimations (Anderson and van Wincoop, 2003), price indices (Baier and Bergstrand, 2001; Bergstrand, 1985, 1989) which do not capture all non-pecuniary trade costs, and country-specific dummies (Anderson and van Wincoop, 2003; Eaton and Kortum, 2002; Feenstra, 2004). The latter is the most simple and the most used solution among the efficient ones. Moreover, country-fixed effects may capture other important country characteristics. 6 As already pointed out, the literature on the effect of political separatism on trade does not fully account for the importance of the multilateral resistance. Fidrmuc and Fidrmuc s (2003) results are robust to the introduction of remoteness variables, which may capture multilateral resistance effects. Their functional form is however entirely disconnected from the theory (Anderson and van Wincoop, 2003). Djankov and Freund (2002) estimate, as a benchmark specification, a traditional gravity equation. This equation differs from the theoretical one in that the multilateral resistance terms are missing. 7 5 McCallum (1995) evaluates the Canadian domestic bias and defines Home ij as 1 for Canadian interprovincial trade and 0 for trade between a Canadian province and an American State. 6 The introduction of country specific dummies partly controls for the prevalence of informal trade in Eastern Europe. For instance, wars and sanctions in the former Yugoslavia disrupted standard trade routes and created incentives and means for the development of traffic which is not included in the official statistics. 7 In their robustness checks, Djankov and Freund (2002) allow for heterogeneity across countries. They introduce a country fixed effect defined as one if the given country is either an exporter or an importer. This method is however inappropriate to capture the multilateral resistance terms.

7 Political Disintegration and Trade 831 The joint determination of output and trade in equilibrium is a second concern (see Harrigan, 1996). To deal with this potential endogeneity, we follow the theoretical prediction of Anderson and van Wincoop (2003) and constrain at unity the coefficients of country income terms. This also mitigates income measurement errors which is an acute problem in transition economies (see below). Applying this constraint and taking into account the relative distance effect lead to a gravity equation consistent with theory: X ijt ln b b ln( ij) b ( ij) i j ijt, (2) YY it jt = Dist 2 Home γ λ ε where the dependent variable is now the ratio of bilateral trade to the product of national incomes. γ i and λ j are country fixed effects for the exporter and the importer, respectively. To stay as close as possible to McCallum (1995) and Anderson and van Wincoop (2003), Equation (2) retains a trade costs specification related to bilateral distance and whether countries are separated by a border. However, such a trade costs specification may bias the estimates of (Home ij ). For example, the estimate will be downward biased if we omit an important trade cost variable negatively correlated with (Home ij ). The problem is that trade costs are not directly observable and our main objective is to identify them. Consequently, we control for additional factors influencing trade patterns. First, geographic peculiarities such as the existence of a land border between trading partners usually affect bilateral trade. Second, linguistic ties are also trade-enhancing. However, such ties are empirically difficult to evaluate and are highly correlated with the land border. Consequently, we follow FF (2003) and only retain the influence of English speaking countries, since most of the other countries sharing a language, in our dataset, also share a border. 8 Third, we control for the influence of a myriad of regional trade agreements (RTAs). Moreover, since we aim to estimate the former country bias of each former country, we replace the Home ij variable with three dummies: one for Czechoslovakia, one for Yugoslavia and one for the FSU. Then, the theoretically consistent equation to be estimated is: X ijt ln ln( ij) ( ij) ( ij) YY it jt = β0 + β1 Dist + β2 Czechoslovakia + β3 Yugoslavia + β4( Soviet Unionij) + β5( Land Borderij) + β ( English Speaking ) + β ( RTA ) + γ + λ + ε, 6 ij k kijt i j ijt k (3) where Land Border ij = 1 if i and j share a land border and 0 otherwise, English Speaking ij = 1 if i and j share the English language and 0 otherwise, RTA is a set of 8 As an example, France and Belgium share a border and a language.

8 832 de Sousa and Lamotte k dummies capturing the influence of regional trade agreements, so RTA = {EU ijt, EFTA ijt, EU EFTA ijt, CEFTA ijt, ASSOC ijt, EFTA ASSOC ijt, CUFSU ijt, NAFTA ijt, CER ijt, EU TUR ijt }. 9 Additionally, the antilog of the estimated coefficient of each successor state s relationship (Czechoslovakia ij, Yugoslavia ij and the Soviet Union ij ) indicates the size of each former country bias. 3.2 Data Our sample consists of 48 countries: 22 EU members, 10 8 OECD non-eu members, CIS countries 12 and 6 Southeastern European countries. 13 The time span covers the period from 1993 to We therefore have a potential of = 2,257 annual observations. Some observations are missing due to the lack of trade data for FSU countries, mainly among Central Asian countries. Trade is measured as bilateral exports (X ij ) from country i to country j, in millions of current dollars, and comes from the CHELEM CEPII database. To our knowledge this is the only database that allows for working on a long time span for transition countries. Using the United Nations COMTRADE database, the CEPII builds the CHELEM dataset by harmonizing the mirror trade statistics, based on the reliability of the reporter. 14 It therefore takes account of the potential drawbacks of trade statistics in transition countries (see Filer and Hanousek, 2002; Kaminski and de la Rocha, 2003). Bilateral distance is also from the CEPII and is computed as the shortest distance between capital cities in kilometres. GDP and population come from the World Development Indicators of the World Bank. GDP 9 EU ijt = 1 if i and j are European Union (EU) members, 0 otherwise; EFTA ijt = 1 if i and j are members of the European Free Trade Agreement, 0 otherwise; EU EFTA ijt = 1 if i is an EU member and j an EFTA member or vice versa, 0 otherwise; CEFTA ijt = 1 if i and j are members of the Central European Free Trade Agreement or vice versa, 0 otherwise; ASSOC ijt = 1 if a non-eu member i has signed an association agreement with an EU-member, 0 otherwise; EFTA ASSOC ijt = 1 if i is an EFTA member and j has signed an association agreement with the EU or vice versa, 0 otherwise; CUFSU ijt = 1 if i and j are members of the customs union between Russia, Belarus, Kazakhstan, Tajikistan and Kyrgyzstan, 0 otherwise; NAFTA ijt = 1 if i and j are members of the North American free trade agreement, 0 otherwise; CER ijt = 1 i and j are members of the Closer Economic Relations free trade agreement, 0 otherwise; EU TUR ijt = 1 if i and j are members of the customs union between the EU and Turkey. 10 Austria, Belgium and Luxembourg (together), the Czech Republic, Denmark, Estonia, Finland, France, Germany, Great Britain, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, the Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain and Sweden. 11 Australia, Canada, Japan, New Zealand, Norway, Switzerland, Turkey and the United States. 12 Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine and Uzbekistan. 13 Bosnia Herzegovina, Bulgaria, Croatia, Macedonia, Romania and Serbia Montenegro. 14 The CEPII aims to provide a coherent and exhaustive overview of world trade. The harmonizing process depends on the extent of inconsistencies between the partners declarations for a given bilateral trade flow. This process relies on an established hierarchy of the declaring countries. This hierarchy ranks the countries according to their reliability and the regularity of publication of their statistics. For further details see <

9 Political Disintegration and Trade 833 is in millions of current dollars and population in millions of inhabitants. Details about the sample and the correlation matrix are available in the Appendix Tables A1 and A2. 4. Results We now estimate Equation (3) year by year, from 1993 to 2001, to study the evolution of the former countries biases over time. Due to missing observations the panel is unbalanced and results may be affected by the annual variation of observations. To eliminate this composition effect, which is a concern in FF (2003), we balance the panel and make the yearly estimations comparable. 15 The results of the yearly estimations are shown in Table 2. Beyond the estimates of the former country bias, several results are noteworthy. 16 First, distance elasticities relative to trade are negative in all estimations. Thus, other things being equal, an increase of bilateral distance decreases trade. The magnitudes of the estimated coefficients are in line with the literature. Second, a common land border encourages trade. This result may be explained by the importance of unobservable affinities between neighbouring countries, such as social or historical links, as well as large volumes of border trade. Third, English speaking countries trade more with each other. All these estimates are rather stable over time. 17 The coefficients of interest to us are the estimates of the former country bias levels (% 2, % 3, and % 4). All these estimated coefficients are highly significant and positive, indicating a high trade intensity between successor states of former countries. Figure 2 plots our year-by-year estimates of the above normal level of trade between the successor states during the period (the vertical line indicates when FF s (2003) data end). Some results are worth interpreting. First, at the beginning of the period, the home bias level of each former country is relatively high. For instance, in 1993, trade between successor states of former Yugoslavia is a factor 29 times [= exp (3.35)] higher than trade with outside partners. The results show a factor 20 for Czechoslovakia and 13 for the FSU. These results are not significantly different from each other and are consistent with the literature. Small and/or developing economies tend to have higher domestic trade intensities than large and/or developed economies (see Anderson and van Wincoop, 2003; De Sousa and Lochard, 2005; Helliwell, 1998). 15 This approach may introduce a bias due to truncated data, but our results are robust with an unbalanced panel. 16 Note that the adjusted R 2 s are higher when the income elasticities are not constrained to unity (between 0.65 and 0.75) since the explanatory power of the national income variables contributes to their calculation. The root mean squared error (RMSE) of the regressions are, however, little changed, with or without this constraint. See Head and Mayer (2000) on this point. 17 The estimated coefficient of distance decreases over time but this decrease is not statistically significant.

10 834 de Sousa and Lamotte Table 2. Former country bias in Central and Eastern Europe, Dependant variable ln(exports ij /GDP i GDP j ) ln(gdp it ) ln(gdp jt ) ln(distance ij ) 0.72* 0.75* 0.71* 0.73* 0.67* 0.66* 0.61* 0.59* 0.64* (0.10) (0.10) (0.10) (0.10) (0.09) (0.10) (0.09) (0.09) (0.10) Land Border ij 0.71* 0.74* 0.84* 0.69* 0.74* 0.65* 0.75* 0.75* 0.72* (0.17) (0.15) (0.14) (0.15) (0.14) (0.15) (0.14) (0.15) (0.16) English Speaking ij 1.75* 1.62* 1.58* 1.48* 1.50* 1.58* 1.63* 1.58* 1.61* (0.21) (0.19) (0.19) (0.18) (0.17) (0.17) (0.17) (0.17) (0.16) Czechoslovakia ijt 3.09* 2.57* 2.37* 2.36* 2.22* 2.01* 1.69* 1.56* 1.47* (0.29) (0.21) (0.22) (0.20) (0.19) (0.19) (0.19) (0.20) (0.20) Soviet Union ijt 2.53* 2.47* 2.15* 2.37* 2.05* 2.03* 1.85* 1.66* 1.56* (0.29) (0.27) (0.25) (0.26) (0.27) (0.26) (0.26) (0.25) (0.26) Yugoslavia ijt 3.35* 3.52* 3.51* 3.10* 3.14* 3.05* 3.01* 3.05* 3.12* (0.42) (0.46) (0.46) (0.43) (0.31) (0.32) (0.28) (0.27) (0.26) RTA fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Exporter fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Importer fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes R Number of obs. 1,737 1,737 1,737 1,737 1,737 1,737 1,737 1,737 1,737 RMSE Notes: * indicates significance at the 1% level. Heteroskedasticity-robust standard errors are in parentheses. The estimated coefficients of exporter, importer and regional trade agreements (RTA) fixed effects as well as intercept are not shown. They are available upon request. Second, the observation of the home bias evolution reveals two different trends, in line with the descriptive statistics of Table 1. On the one hand, we observe a stable progression of the home bias for the former Yugoslavia and the FSU up to 1998 (the last year of FF s sample). Several non-exclusive factors may explain such a trend. Trade reorientation to the West, which may decrease home bias, is limited by the EU trade restrictions at the beginning of the period. Trade preferences with the former Yugoslavia entered into force only after Moreover, the political disintegration may have not broken up business and social networks operating across former countries. Former networks persist, overlap new borders and encourage trade between successor states. 18 A final and related explanation is the 18 See Rauch and Trindade (2002) for evidence of trade-creating effects of the largest trans-national network (or set of interlinked national networks): the overseas Chinese.

11 Political Disintegration and Trade 835 Figure 2. Evolution of the home bias in Central and Eastern Europe, Note: Dotted lines represent confidence intervals. role of affinities between successor states, resulting from linguistic, historical and cultural links. This effect on trade is only partly captured by the introduction of observable characteristics, such as a land border variable. On the other hand, we observe a severe decrease of the home bias for the FSU since and Czechoslovakia over the whole period. The latter result agrees with FF (2003). Czechoslovakia experienced continuous trade disintegration despite the creation of a customs union and a clearing-account payment mechanism between the successor states following political disintegration. In 2001, the Czech Republic and Slovakia traded only four times more with each other than with other partners. The FSU showed a similar home bias in However, trade disintegration began in 1998, only several years after political disintegration. These results show that political disintegration has different consequences for trade. In other words, political disintegration does not systematically lead to trade disintegration. Several other factors may affect the evolution of trade patterns of successor states. This conclusion contrasts with FF (2003). In the next section, we check the robustness of this conclusion and investigate the source of the differences with FF (2003). 5. Fidrmuc and Fidrmuc s (2003) sample and specification We now turn to the comparison of our results with FF (2003). We replicate their estimation, using their specification and their sample. This strategy requires some 19 From 1999, the value of the FSU s home bias becomes statistically different from the 1993 value. It is worth noting that when we estimate Equation (3) with an unbalanced panel, like FF (2003), we do not find any trade disintegration for the FSU over the period. This may be the result of a composition effect due to sample changes.

12 836 de Sousa and Lamotte modifications. Consequently, we only retain Croatia and Slovenia within the former Yugoslavia, and Belarus, Russia, Ukraine and the Baltic States within the FSU. We keep the Czech Republic and Slovakia, and drop the remaining successor states. To adopt FF s (2003) specification, we also redefine our set of home variable dummies. We replace the Soviet Union ijt and the Yugoslavia ijt variables with three new dummies: Croatia Slovenia ijt (equal to 1 for trade between Croatia and Slovenia and 0 otherwise); Belarus Russia Ukraine ijt (equal to 1 for trade between Belarus, Russia and Ukraine and 0 otherwise); Baltic States ijt (equal to 1 for trade between Baltic States and 0 otherwise). Thus, the only difference between FF s (2003) empirics and ours is the source of the dataset, which allows us to extend their results beyond The results are presented in Table 3. Using FF s (2003) specification and sample, we confirm three of their results. We find severe and continuous trade disintegration in the cases of Czechoslovakia and Croatia Slovenia. Moreover, we do not find any evidence of trade disintegration in the case of the Baltic States. These results hold after Surprisingly, Belarus, Russia and Ukraine do not tend to trade more with each other than with outside partners over the period. Estimates of their home bias are not significant due to large standard errors. Note that FF (2003) finds significant positive estimates but with puzzling reversals (see below) and also large variation of the standard errors. We now attempt to replicate their estimates with our specification (Equation 3) on their sample. We show the results in Table 4 and plot the year-by-year estimates of the above normal level of trade between the successor states during the period in Figure 3. We confirm trade disintegration for Czechoslovakia and no trade disintegration for the Baltic States. However, we find conflicting results for Croatia Slovenia and Belarus Russia Ukraine. In both cases we do not find any evidence of trade disintegration. The home bias level of Croatia Slovenia is lower in 2001 compared to 1993 but the difference is not statistically significant. The case of Belarus Russia Ukraine is rather puzzling. As in Table 4, these successor states did not trade more with each other than with other partners in 1993 and The subsequent results indicate rather high trade intensity between them from 1994 to In contrast, FF (2003) finds somewhat puzzling reversals of the home bias level. It decreases from 32 to 11 between 1994 and 1995, and increases from 8 to 31 between 1997 and The latter result is explained by the Russian crisis in However, the crisis occurred in August 1998 and one may be doubtful about such a magnitude of change in trade patterns. Using our specification on their sample, we also find an increase of trade intensity between Belarus, Russia and Ukraine around this time, but it occurs between 1998 and 1999, and its magnitude is much more plausible. To sum up, we find that the use of a different dataset does not explain the differences between our results and FF s (2003) results. The main differences are due to the use of a theoretically founded gravity equation and a more comprehensive sample of successor states (see Section 4).

13 Table 3. Home bias of former countries, Fidrmuc and Fidrmuc s (2003) sample and specification, Dependent variable ln(exports ij ) ln(gdp it ) 1.06* 1.02* 1.00* 1.01* 1.03* 1.04* 1.02* 1.01* 1.00* (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) ln(gdp jt ) 0.99* 0.96* 0.94* 0.92* 0.92* 0.92* 0.94* 0.95* 0.92* (0.03) (0.03) (0.03) (0.03) (0.02) (0.02) (0.03) (0.03) (0.03) ln(distance ij ) 0.83* 0.82* 0.81* 0.80* 0.81* 0.84* 0.85* 0.85* 0.84* (0.06) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) (0.05) Land Border ij 0.81* 0.85* 0.90* 0.89* 0.87* 0.77* 0.71* 0.72* 0.73* (0.15) (0.15) (0.15) (0.14) (0.14) (0.14) (0.14) (0.14) (0.14) English Speaking ij 1.78* 1.67* 1.58* 1.56* 1.52* 1.56* 1.59* 1.53* 1.57* (0.20) (0.18) (0.18) (0.19) (0.18) (0.18) (0.18) (0.17) (0.17) EC-12 ij 0.97* 0.84* 0.83* 0.83* 0.74* 0.76* 0.81* 0.70* 0.70* (0.11) (0.10) (0.10) (0.10) (0.10) (0.10) (0.10) (0.10) (0.10) CEFTA ij *** 0.27*** 0.28*** *** (0.16) (0.16) (0.16) (0.15) (0.15) (0.15) (0.15) (0.14) (0.13) EFTA ij 1.50* 1.37* 1.32* 1.29* 1.15* 1.22* 1.24* 1.10* 1.06* (0.17) (0.17) (0.17) (0.17) (0.17) (0.17) (0.18) (0.17) (0.16) Europe Agreements ij 0.18** 0.15*** ** 0.22* 0.26* 0.32* (0.09) (0.09) (0.08) (0.08) (0.08) (0.08) (0.08) (0.08) (0.07) EC-12 EFTA3 ij 1.21* 1.15* 1.15* 1.15* 1.08* 1.13* 1.14* 1.01* 0.94* (0.13) (0.12) (0.13) (0.12) (0.12) (0.12) (0.13) (0.12) (0.12) Political Disintegration and Trade 837

14 Table 3. (cont) Home bias of former countries, Fidrmuc and Fidrmuc s (2003) sample and specification, Dependent variable ln(exports ij ) EFTA3 Assoc. ij 0.48* 0.53* 0.45* 0.46* 0.49* 0.57* 0.59* 0.48* 0.44* (0.13) (0.14) (0.14) (0.13) (0.13) (0.12) (0.13) (0.13) (0.12) Czechoslovakia ijt 2.85* 2.68* 2.56* 2.44* 2.20* 1.99* 1.81* 1.71* 1.65* (0.16) (0.17) (0.17) (0.15) (0.15) (0.15) (0.15) (0.14) (0.13) Croatia Slovenia ijt 3.00* 2.33* 2.18* 2.07* 1.86* 1.78* 1.79* 1.68* 1.58* (0.18) (0.18) (0.21) (0.20) (0.24) (0.26) (0.20) (0.20) (0.26) Baltic States ijt 2.75* 2.65* 2.49* 2.69* 2.75* 2.92* 2.93* 2.88* 2.88* (0.21) (0.22) (0.26) (0.27) (0.29) (0.26) (0.23) (0.21) (0.24) Belarus Russia Ukraine ijt (2.68) (1.71) (0.96) (2.23) (1.36) (2.31) (1.33) (2.05) (2.75) Exporter fixed effects No No No No No No No No No Importer fixed effects No No No No No No No No No R Number of observations 1,238 1,238 1,238 1,238 1,238 1,238 1,238 1,238 1,238 RMSE Notes: *, ** and *** indicate significance at the 1%, 5% and 10% level, respectively. Heteroskedasticity-robust standard errors are in parentheses. 838 de Sousa and Lamotte

15 Political Disintegration and Trade 839 Table 4. Home bias of former countries, Fidrmuc and Fidrmuc s (2003) sample and our specification, Dependent variable ln(exports ij /GDP i GDP j ) ln(gdp it ) ln(gdp jt ) ln(distance ij ) 0.82* 0.83* 0.78* 0.74* 0.76* 0.71* 0.69* 0.70* 0.72* (0.11) (0.12) (0.12) (0.13) (0.11) (0.12) (0.12) (0.11) (0.11) Land Border ij 0.94* 0.97* 1.01* 0.97* 0.94* 0.91* 0.94* 0.93* 0.90* (0.18) (0.17) (0.17) (0.17) (0.16) (0.17) (0.17) (0.16) (0.16) English Speaking ij 1.62* 1.60* 1.49* 1.49* 1.45* 1.53* 1.56* 1.50* 1.53* (0.21) (0.19) (0.19) (0.20) (0.18) (0.18) (0.18) (0.18) (0.17) Czechoslovakia ijt 2.91* 2.52* 2.40* 2.22* 2.10* 1.91* 1.57* 1.38* 1.34* (0.22) (0.19) (0.21) (0.19) (0.19) (0.19) (0.18) (0.18) (0.21) Croatia Slovenia ijt 2.92* 2.32* 2.42* 2.63* 2.33* 2.36* 2.35* 2.36* 2.27* (0.32) (0.32) (0.31) (0.34) (0.30) (0.23) (0.26) (0.25) (0.23) Baltic States ijt 3.07* 2.88* 2.60* 2.76* 2.59* 2.79* 2.60* 2.54* 2.57* (0.34) (0.32) (0.29) (0.32) (0.30) (0.27) (0.25) (0.23) (0.24) Belarus Russia Ukraine ijt * 2.82* 2.72* 2.69* 2.99* 2.65* 2.54* (2.59) (1.66) (0.63) (0.50) (0.72) (0.68) (0.57) (0.45) (0.42) RTA fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Exporter fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes Importer fixed effects Yes Yes Yes Yes Yes Yes Yes Yes Yes R Number of observations 1,238 1,238 1,238 1,238 1,238 1,238 1,238 1,238 1,238 RMSE Notes: * indicates significance at the 1% level. Heteroskedasticity-robust standard errors are in parentheses. Estimated coefficients of exporter, importer and RTA fixed effects and intercept are not shown. They are available upon request. 6. Concluding remarks The empirical evidence shows that political disintegration did not lead systemically to severe and rapid trade disintegration. In Eastern Europe, the only case of substantial trade disintegration following political separatism was found for the former Czechoslovakia. Concerning the other former countries, the results are far from showing evidence of a severe disintegration. Thus, it seems that the link

16 840 de Sousa and Lamotte Figure 3. Home bias of former countries, , Fidrmuc and Fidrmuc s (2003) sample and our specification Note: Dotted lines represent confidence intervals. between political and trade disintegration is not as clear as previously stated and that lots of other complicated stuff matters (Baldwin, 2006, p. 23). Two factors may explain the difference between the former Czechoslovakia and both the former Yugoslavia and the former Soviet Union. A first explanation is found in the transition process. The institutional reforms may favour trade redeployment to outside partners. Thus, those which have made the most progress in structural reforms have also gone farthest in diversifying their exports to new destinations (Havrylyshyn and Al-Atrash, 1998). Actually, transition reforms took place faster in the former Czechoslovakia and played an obvious role in the decline of bilateral trade between the Czech Republic and Slovakia. Thus, even if political disintegration had not taken place, the two successor states would have traded less with each other. A second explanation is the extent of trade liberalization. Since the mid-1990s, the trade of the former Czechoslovakia has been embedded in the framework of various regional free trade agreements with the EU and Central European countries and the membership of the World Trade Organization. This liberalization process eases trade reorientation and integration into the world economy. By contrast, except for Slovenia, trade liberalization of the former Yugoslavia and the former Soviet Union is far behind. References Alesina, A., Spolaore, E. and Wacziarg, R. (2000). Economic integration and political disintegration, American Economic Review, 90, pp Anderson, J. E. and Marcouiller, D. (2002). Insecurity and the pattern of trade, Review of Economics and Statistics, 84, pp

17 Political Disintegration and Trade 841 Anderson, J. E. and van Wincoop, E. (2003). Gravity with gravitas: A solution to the border effect puzzle, American Economic Review, 93, pp Baier, S. and Bergstrand, J. H. (2001). The growth of world trade: Tariffs, transport costs, and income similarity, Journal of International Economics, 53, pp Baldwin, R. (1988). Hysteresis in import prices: The beachhead effect, American Economic Review, 78, pp Baldwin, R. (2006). In or Out: Does It Matter? An Evidence-based Analysis of the Euro s Trade Effects, London: CEPR. Bergstrand, J. H. (1985). The gravity equation in international trade: Some microeconomic foundations and empirical evidence, Review of Economics and Statistics, 67, pp Bergstrand, J. H. (1989). The generalized gravity equation, monopolistic competition and the factor-propositions theory of international trade, Review of Economics and Statistics, 71, pp De Sousa, J. and Lochard, J. (2005). Do currency barriers solve the border effect puzzle? Evidence from the CFA Franc zone, Review of World Economics/Weltwirtschaftliches Archiv, 141, pp Djankov, S. and Freund, C. (2002). Trade flows in the Soviet Union 1987 to 1996, Journal of Comparative Economics, 30, pp Eaton, J. and Kortum, S. (2002). Technology, geography and trade, Econometrica, 70, pp Eichengreen, B. and Irwin, D. A. (1997). The role of history in bilateral trade flows, in Frankel, J. A. (ed.), The Regionalisation of the World Economy, Chicago, IL: University of Chicago Press, pp Feenstra, R. (2004). Advanced International Trade: Theory and Evidence, Princeton, NJ: Princeton University Press. Fidrmuc, J. and Fidrmuc, J. (2003). Disintegration and trade, Review of International Economics, 11, pp Filer, R. K. and Hanousek, J. (2002). Data watch: Research data from transition economics, Journal of Economic Perspectives, 16, pp Frankel, J. (1997). Regional Trading Blocs, Washington D.C.: Institute for International Economics. Freund, C. (2000). Different paths to free trade: The gains from regionalism, Quarterly Journal of Economics, 115, pp Harrigan, J. (1996). Technology, factor supplies, and international specialization: Estimating the neoclassical model, American Economic Review, 87, pp Harrigan, J. (2003). Specialization and the volume of trade: Do the data obey the laws? in Choi, K. and Harrigan, J. (eds), Handbook of International Trade, Oxford: Blackwell Publishing, pp Havrylyshyn, O. and Al-Atrash, H. (1998). Opening up and geographic diversification of trade in transition economies, IMF Working Paper No. 22, Washington D.C.: IMF. Head, K. and Mayer, T. (2000). Non Europe: The causes and magnitude of market fragmentation in the EU, Review of World Economics/Weltwirtschaftliches Archiv, 136, pp Helliwell, J. F. (1998). How Much Do National Borders Matter?, Washington, D.C.: The Brookings Institution Press. Kaminski, B. and de la Rocha, M. (2003). Stabilization and association process in the Balkans: Integration options and their assessment, World Bank Policy Research Working Paper No. 3108, Washington D.C.: World Bank,

18 842 de Sousa and Lamotte McCallum, J. (1995). National borders matter: Canada US regional trade patterns, American Economic Review, 85, pp Pomfret, R. (2002). National borders and disintegration of market area in central Asia after 1991, in Müller, U. and Schultz, H. (eds), National Borders and Economic Disintegration in Modern East Central Europe, Berlin: Frankfurter Studien zur Grenzregion, 8, Berlin Verlag arno Spritw Gmbh, pp Portes, R. and Rey, H. (2005). The determinants of cross-border equity flows, Journal of International Economics, 65, pp Rauch, J. (2001). Business and social networks in international trade, Journal of Economic Literature, 39, pp Rauch, J. E. and Trindade, V. (2002). Ethnic Chinese networks in international trade, Review of Economics and Statistics, 84, pp Wei, S. -J. (1996). Intra-national versus international trade: How stubborn are nations in global integration? NBER Working Paper No. 5531, Cambridge, MA: NBER, Appendix Table A1. Descriptive statistics, Number of Mean Standard deviation Min Max observations ln(x ij /Y i Y j ) 15, ln(dist ij ) 15, Border ij 15, English ij 15, EU ijt 15, EFTA ijt 15, EUEFTA ijt 15, CEFTA ijt 15, CUFSU ijt 15, ASSOC ijt 15, EFTA ASSOC ijt 15, NAFTA ijt 15, EU TUR ijt 15, CER ijt 15, Yugoslavia 15, Baltic States 15, Soviet Union 15, Czechoslovakia 15,

19 Table A2. Correlation matrix of the variables, ln(x ij /Y i Y j ) ln (Dist ij ) Border ij English ij EU ijt EFTA ijt EU EFTA ijt CEFTA ijt CUFSU ijt ln(x ij /Y i Y j ) ln(dist ij ) Border ij English ij EU ijt EFTA ijt EU EFTA ijt CEFTA ijt CUFSU ijt ASSOC ijt EFTA ASSOC ijt NAFTA ijt EU TUR ijt CER ijt Czechoslovakia Yugoslavia Baltic Soviet Union ASSOC ijt EFTA ASSOC ijt NAFTA ijt EU TURKEY ijt CER ijt Czechoslovakia Yugoslavia Baltic States Soviet Union ASSOC ijt EFTA ASSOC ijt NAFTA ijt EU TURKEY ijt CER ijt Czechoslovakia Yugoslavia Baltic States Soviet Union Note: Coefficients with significance levels lower than 5% are not shown. Political Disintegration and Trade 843