Disintegration, Transition and Tradeâ
Disintegration, Transition and Trade∗ Jos´e de Sousa†
Olivier Lamotte‡
September 10, 2004 - Very preliminary draft
Abstract This paper studies the scope of trade disintegration in a context of transition and political separatism. We use the gravity equation and the border effect methodology to evaluate the home bias of Soviet Union, former Yugoslavia and Czechoslovakia from 1993 to 2001. We find high home biases for each federation, even if they significantly decrease during our sample period. We also find strong evidence that transition had a positive impact on trade and on the reduction of the home bias. Finally we point out some regional differences, as transition process explains a smaller part of the home bias in the case of Czechoslovakia.
JEL Classification: F13, F15 Keywords: disintegration, borders, transition
∗
We are grateful to Fabian Gouret, Julie Lochard, Mathilde Maurel and Daniel Mirza for providing valuable comments. † LESSOR, University of Rennes 2, ROSES-CNRS, University of Paris 1, 106-112 boulevard de l’Hˆopital, 75647 Paris cedex 13, E-mail: [email protected] ‡ University of Cergy-Pontoise and ROSES-CNRS, University of Paris 1, 106-112 boulevard de l’Hˆopital, 75647 Paris cedex 13, E-mail: [email protected]
1
1
Introduction
Political disintegration leads to the creation of new boundaries and the erection of new barriers to trade (tariff and non-tariff barriers). As a result, trade intensity between successor states tends to decrease. Prior studies found evidence that political disintegration is a source of trade disintegration in Eastern European countries (Djankov and Freund, 2002; Fidrmuc and Fidrmuc, 2003) and Central Asia (Pomfret, 2002). These results are challenging for two reasons. First, other things being equal, sucessor states continue to trade more between themselves than with other partners. Second, the concurrent transition process has often been ignored. This paper addresses the question of trade disintegration in a setting of political separatism and transition. We wonder whether the decrease of trade intensity is the sole result of political disintegration. Our main argument is that a part of the trade disintegration can be explained by the transition process. The literature documents that transition countries in Eastern Europe have experienced large reorientation of their trade flows towards western economies (Andreff and Andreff, 1995; Maurel and Cheikbossian, 1998) and 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). The transition process seems to favor trade disintegration. The analysis of political disintegration is quite rare in the economic literature. Almost all studies focus on the motives and the conditions for disintegration1 and do not study the consequences of such a process. The border effect methodology opens a new field of research for the study of disintegration experiences. Following the influential work of McCallum (1995), it is now widely documented that national borders matter for trade. As an example, a Canadian province trades 20 times more with other Canadian provinces than with American states of a given size and distance (McCallum, 1995). 1
See Alesina and Spolaore, 1997; Alesina et al., 2000; Bolton and Roland, 1997.
2
The application of the border effect methodology to the analysis of disintegration permits to determine the level of trade barriers between successor states compared to the external barriers. Wei (1996) defines the border effect as “an all-inclusive summary of barriers to trade”. Consequently, the border effect of a politically disintegrated region is defined as an all-inclusive summary of remaining ties that make intra-successor states trade relatively larger than the trade with other countries. In case of a political disintegration, we expect a decrease of the former national home bias, as a substantial part of the former intra-national trade flows decrease. We test this idea and the influence of the transition process by re-integrating virtually the former country, i.e. by assuming that political disintegration did not take place. For this issue, Soviet Union, former Yugoslavia and Czechoslovakia constitute relevant examples. They enable to measure the speed and the scope of trade disintegration in a context of transition and political separatism. We use a gravity equation and a sample of 52 countries (cf. infra for details) to evaluate the home bias of these former Federations from 1993 to 2001 and provide some explanations to their evolution. This paper makes several contributions to the existing literature. We find a high home bias in the former Federations during the nineties. This bias decrease significantly during our sample period. We also find strong evidence that transition explained a large part of the home bias decrease. Finally we show that transition had a different impact on the economic disintegration of the countries under focus. We proceed as follows. In section 2, we discuss the econometric specification of the model and the dataset. Results are laid in section 3. In section 4, we provide some robustness checks and we conclude in section 5.
3
2 2.1
Model set-up and data description The gravity model
The empirics are based on the gravity model, which capacity to explain trade flows has been successful empirically. The traditional gravity equation takes the following form:
ln(Xij ) = β0 +β1 ln(GN Pi )+β2 ln(GN Pj )+β3 ln(Distanceij )+β4 (HOM Eij )+εij (1) where (Xij ) is merchandise exports from country i to country j, (GN Pi ) and (GN Pj ) are national incomes of the exporter and importer, (Distanceij ) the distance between i and j, and following McCallum (1995) (HOM Eij ) is a dummy variable equal to 1 for intra-national trade and 0 for international trade. More precisely, it is equal to 1 for trade flows between successor states of Soviet Union, Former Yugoslavia and Czechoslovakia and 0 otherwise. The (HOM Eij ) variable measures the size of the border effect and allows us to determine the magnitude of the disintegration process. The coefficients on national incomes are expected to be positive and the distance coefficient negative. The coefficient of (HOM Eij ) will be positive if intra-national trade flows are larger than international ones, i.e. if trade flows among successor states of former Federations are larger than flows between the successor states and their other trading partners. The antilog of the coefficient indicates the size of the border effect. In the core of the paper, we divide the (HOM Eij ) variable into three dummies, one for each former Federation. As an example, the dummy “Soviet Union” is equal to 1 for inter-(former) Soviet Union trade and 0 otherwise. In this way, we estimate the average border effect of each former federation. εij is an error term. In order to determine the effect of transition on the disintegration process we introduce in equation 1 an index of progress in transition for each partner.
4
The new specification is the following: ln(Xij ) = β0 + β1 ln(GN Pi ) + β2 ln(GN Pj ) + β3 ln(Distanceij ) + β4 (HOM Eij ) + β5 ln(IRPi ) + β6 ln(IRPj ) + εij
(2)
where (IRPi ) and (IRPj ) stands for Index of Reform Progress. Estimated coefficients of progress in transition are expected to be positive. According to the theory, the traditional gravity equation 2 is misspecified while omitting “multilateral trade resistance” variables (Anderson and Van Wincoop, 2003). Such variables “summarize the average trade resistance between a country and its trading partners” (Anderson and Van Wincoop, 2004: 28). Given a bilateral trade relationship between i and j, an increase of the multilateral resistance of j raises its trade with i, other things being equal. Put differently, multilateral resistance is related to the distance of country pairs relative to their trading partners. “Trade will be higher between country pairs that are far from the rest of the world than between country pairs that are close to the rest of the world” (Harrigan, 2002). This relative distance is difficult to capture but can be approximated by several ways: the use of a complex non-linear least square estimation (Anderson and van Wincoop, 2003), the use of prices indexes (Bergstrand, 1985, 1989; Baier and Bergstrand, 2001) which do not capture all non-pecuniary trade costs, and country-specific dummies (Anderson and van Wincoop, 2003; Eaton and Kortum, 2002 and Feenstra, 2004). The latter is the most simple solution among the efficient ones. However, the use of country-specific dummies does not fit with the aim of the paper for two reasons. First, the country-specific dummies are time-invariant but multilateral resistances variables on a several years sample are not. Second, introducing jointly country-specific dummies and country variables (rather than country pair variables) induces a serious problem of multicolinearity, since explanatory variables are seriously correlated. In fact, country progress in transition, which hardly varies over time, is captured by fixed effects (see Melitz 2004 for similar problems). It would therefore be impossible to estimate the effect of transition. As a result, we 5
follow Harrigan (2002) in deriving an ad-hoc proxy of the relative distance as a remoteness index for each partner. The estimated equations take the following form:
ln(Xij ) = β0 + β1 ln(GN Pi ) + β2 ln(GN Pj ) + β3 ln(Distanceij ) + β4 (HOM Eij ) + β5 ln(IRPi ) + β6 ln(IRPj ) + β7 ln(Remotenessi ) + β8 ln(Remotenessj ) + εij
(3)
where (Remotenessi ) and (Remotenessi ) are the index of remoteness of countries i and j. Estimated coefficients of remoteness are expected to be positive. Methodological problems, as well as definitions of variables and presentation of data are discussed in the following sub-section.
2.2
The database
Our sample consists of 52 countries-22 EU members2 , 11 non-EU OECD countries3 , 12 CIS countries4 and 7 Southeastern European countries5 . The time span covers the period from 1993 to 2001. We therefore have a potential of 52x51x9=23868 observations. Bilateral trade exports (Xij) come from the Chelem-Cepii database. To our knowledge this is the only database that allows working on such a long time span for all Southeastern countries. Missing values come from the IMF DOTS (Direction Of Trade Statistics database). Flows are measured in millions of current dollars. GNP and distance also come from the Chelem-Cepii database. GNP is measured in millions of current dollars. Distance is the 2
Austria, Belgium and Luxemburg (together), Czech Republic, Denmark, Estonia, Finland, France, Germany, Great Britain, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain and Sweden. 3 Australia, Canada, Iceland, Japan, Korea, Mexico, New Zealand, Norway, Switzerland, Turkey, United States 4 Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine, Uzbekistan 5 Albania, Bosnia-Herzegovina, Bulgaria, Croatia, Macedonia, Romania and SerbiaMontenegro.
6
shortest distance “as the crow flies” between capital cities in kilometers. Remoteness takes account of the fact that “what matters for bilateral export volume is not just the absolute distance between the two countries, but their geographic positions relative to all other countries”(Wei, 1996: 5). It can be defined as a weighted average of distances between a country i and all its partners. Several measures of remoteness have been proposed in the literature6 We tried all of them but since they give very similar results we follow Nitsch (2000). He proposes the following formula: Remotenessi =
1 Pk Σk ( GN ) Dik
The introduction of an index of transition progress required finding some data on transition for the nine years of the sample and for all countries. The European Bank for Reconstruction and Development (EBRD) provides the only complete database on such an issue7 . The EBRD measures, on an annual basis, the progress of transition countries towards a market economy along several dimensions. The dimensions are the following8 : (1) price liberalization, (2) foreign exchange and trade liberalization, (3) small-scale privatization, (4) large-scale privatization, (5) enterprise reform, (6) competition policy, (7) infrastructure reform, (8) banking sector reform and (9) reform of non-banking financial institutions. The progress in each dimension is scored on a scale from 0.7 (low level) to 4.3 (high level). Non transition economies are considered as benchmarks and were given the score of 4.3 in all dimensions. In order to get a unique index of transition progress we computed the so-called Index of Reform Progress (IRP), constructed by Sachs (1996). It is obtained by simply summing the 6
Wei (1996) measures the remoteness of a country k as: Remotenessi = Σi (wk Di k) GN Pk where wk = GN Pworld and k = i, j. Helliwell (1998) proposed another measure of remoteD
ness: Remotenessi = Σk ( GNkjPk ) 7 Other databases (Heritage Foundation, Fraser Institute, Penn World Table, IMF Country Reports, World Bank development indicators) do not include all successor states of Soviet Union and former Yugoslavia for the nineties. 8 For availability reasons we considered nine of the eleven dimensions, excluding “legal extensiveness” and “legal effectiveness”.
7
EBRD’s subindexes9 . Some limits have to be highlighted before proceeding to the results. First, one has to be cautious regarding South East European countries: Kaminski and De la Rocha (2003) highlight huge gaps when comparing mirror statistics, i.e. the same flow declared by the exporter and the importer. Second, wars and embargoes on the territories of former Yugoslavia in the nineties disrupted standard trade routes and created incentives and means for the development of traffics and more generally of the unofficial economy. Third, measures of the reform process are highly imperfect because the EBRD “have relied on subjective indices rather than directly observable variables” (Falcetti, Raiser and Sanfey, 2002).
3
Results
We now proceed to the estimations of the specification 3. The results are presented in table 1 and 2. We first estimate a single border effect (Estimation 1) and then break it out to measure the border effect for each dismissed Federation (Estimation 2). Finally we study the impact of transition on the border effect and we divide the sample into three time-periods (Table 2). Estimation (1) exhibits positive coefficients of supply (GN Pit ) and demand (GN Pjt ) forces on the volume of trade as well as a negative coefficient of distance (Distanceij ). This last result means that, other factors fixed, increasing bilateral distance impedes trade. The coefficient on (HOM Eij ) is positive and significant. The border effect is obtained by computing the antilog of the coefficient. On average, between 1993 and 2001, trade between successors states of the three former Federations was more than 12 (exp(2.53)=12.55) times larger than their international trade. As expected this bias is large and is consistent with what is found in the literature for the “real” intra-national trade. The positive coefficient on the Index of Reform Progress variables indicates that reforms favors trade. Moreover it seems that reforms in the export9
The values of the IRP index for each country are available in the appendix (Table 7).
8
Table 1: Average border effects of former Federations Period 93-01 93-01 Dependent variable: ln exports (1) (2) a ln(GNPi ) 0.87 0.88a 0.01
ln(GNPj )
0.75
0.01 a
0.01
ln(Distanceij )
0.01
-0.93
a
0.02
Homeij
2.53
0.75a -0.93a 0.02
a
0.07
2.48a
Soviet Unionij
0.07
2.92a
Former Yugoslaviaij
0.14
2.67a
Czechoslovakiaij
0.07
ln(Index of Reform Progressi )
1.38
a
1.38a
0.05
0.05
ln(Index of Reform Progressj )
0.87a
0.87a
0.05
0.05
ln(Remotenessi )
0.08a
0.08a
0.03
0.03
0.01
0.01
0.03
0.03
Yes 21409 0.63
Yes 21409 0.63
ln(Remotenessj ) Year fixed effects # of observations R-squared Notes:
a, b
and
c
define 1%, 5% and 10% significance level respectively. Standard Errors (heteroskedasticity robust) are presented in italic.
9
ing country have a higher stimulating effect than reforms in the importing country. The coefficients of remoteness (Remotenessi and Remotenessj ) are positive. In estimation 2 the (HOM Eij ) variable is divided into three dummies: (SovietU nionij ), (f ormerY ugoslaviaij ) and (Czechoslovakiaij ). Unsurprisingly, trade intensity between successor states of the three former Federations have been higher than trade intensity with outside partners. For 1993-2001 the border effect is 12 (=exp(2.48)) for Soviet Union, more than 18 (=exp(2.92)) for former Yugoslavia, and 14 (=exp(2.67)) for Czechoslovakia. The other variables are not affected by the new specification. The fact that the coefficient on HOM Ei j is the lowest for the Soviet Union is probably explained by the various degrees of integration that can be found in the area. The picture would probably be different if we had considered sub-regions of Soviet Union. For example, the would certainly be higher for each Republic with Russia or between the Baltic States and lower between more remote Republics, for example between the Baltic Republics and the Central Asian Republics. The same remark applies for former Yugoslavia. Lamotte (2003) points out very different degrees of integration between the Republics of former Yugoslavia on a bilateral basis. In table 2 we present the impact of the progress in reforms ij on the border effect and its evolution. The estimations are first run without and then with the Index of Reform Progress variables (hereafter called the benchmark estimation). Table 2: Evolution of the average border effectsa Period Average Border Effect Border Effect Soviet Union Border Effect Former Yugoslavia Border Effect Czechoslovakia a The
Without IRP 93-95 96-98 99-01 11.59 9.02 7.39 11.36 8.76 7.10 11.82 10.80 10.07 22.87 13.87 10.38
With IRP 93-95 96-98 99-01 19.49 11.59 9.78 19.49 10.91 9.21 20.91 19.49 15.96 20.91 13.87 10.38
border effect is computed as the antilog of the coefficient of each variable.
The first striking result is that the coefficient on (HOM Eij ) is always 10
higher when the progress in reform is controlled for. This result could be surprising. Actually the interpretation is that if transitional reforms had not taken place the border effect would have been higher. Therefore transition had an impact on the border effect. More precisely reforms have reduced the gap between the relative sizes of the inside and the outside borders of the former Federations. A second interesting result emerges from the observation of the evolution of the border effect. Figure 1 plots the evolution of the average border effect over time. The gap between the two curves, i.e. the border effect in presence or not of the reform index (IRPi and IRPj ), is decreasing. The explanatory power of the transition as a source of trade disintegration is therefore decreasing over time. Other interesting remarks can be drawn from the observation of the evolution of the border effects. In figures 2 to 4 we present the evolution of the border effects of the three dismissed federations. The first interesting point is that the trade disintegration took different intensity in the three considered areas. On the sample period the border effect was divided by two for Czechoslovakia whereas it decreased less than 15% for former Yugoslavia. The second interesting point is that the impact of reforms on the border effect has been different for each former Federation. In the three cases the gap between the two curves representing the border effects tends to decrease. But the part of the trade disintegration explained by the transition process is quite different between the three cases. The most surprising result occurs for Czechoslovakia (figure 4). For the first period the curves are inverted compared to the two previous figures, meaning that reforms had a positive impact on the mutual integration of the two successor states. From the mideighties it seems that the reforms did not have an impact on trade integration between the Czech Republic and Slovakia any more. In the following section we will test whether these results are robust to other specifications.
11
4
Robustness checks
In order to check the robustness of our results we augment the basic specification with two sets of additional variables that concern geography and regional integration agreements. We proceed as in the previous section; we measure a single border effect, then we break out the effect by zone and finally we determine its evolution on three sub-periods. We first control for geographical characteristics by augmenting equation (3) with dummy variables that are usually included in gravity equations. (Common languageij ) and (Common borderij ) take the value 1 if the two partners share respectively the same language or a common border. They are both expected to have a positive sign. (Landlockenessi ) and (Landlockenessj ) take the value 1 if the exporter or the importer is a landlocked country, i. e. with no direct access to the sea. It is expected with a negative sign. Results are presented in table 3 and 4. The only significant change between estimations (1) and (2) (table 1) and estimations (3) and (4) (table 3) concerns the level of the border effect. In all cases the coefficient is lower. It is not surprising since a characteristic of member states of a Federation is to have a common border and/or a common language. A part of the effect is therefore captured by the common language and border dummies. From table 4 and figures 5 to 8 we can see that the trends and the extent of the gaps remain almost the same. The second set of robustness checks concerns regional integration agreements. We add the following dummy variables 10 : (EUij ), (CEFTAij ), (Baltic FTAij ), (Association Agreementsij ) and (Custom Union in FSUij ). Each variable takes the value 1 if trade occurs between two members of the agreement. Results are laid in tables 5 and 6 and figures 9 to 12. The previous results are very slightly affected by the introduction of new variables. The coefficients on the trade agreements are significant and consistent with the 10
(EU ) stands for the European Union, (CEFTA) for the Central European Free Trade Agreement, (Baltic FTA) for the free trade agreement between Baltic states, (Association Agreements) for the trade agreements between the EU and Central and Eastern European Countries and (Custom Union in FSU ) for the custom union between Russia, Belarus, Kazakhstan, Tajikistan and Kirghizstan.
12
literature, except for the custom union in Soviet Union. From the comparison from tables 2 and 6 we can observe that the coefficients of the border effects dummies are a bit higher in the specification without IRP. Concerning the trend of the curve, it is similar to the previous specifications except for former Yugoslavia. The introduction of additional variables reduce slightly the decrease of the border effect.
13
Table 3: Estimations with geographic variables Period 93-01 93-01 Dependent variable: ln exports (3) (4) a ln(GNPi ) 0.88 0.88a 0.01
0.01
ln(GNPj )
0.75a
0.75a
0.01
0.009
ln(Distanceij )
-0.82a
-0.82a
0.02
0.02
Homeij
2.27
a
0.07
2.25a
Soviet Unionij
0.07
Former Yugoslaviaij
2.61a
Czechoslovakiaij
1.49a
0.16
0.08
ln(Index of Reform Progressi )
1.36
a
0.05
ln(Index of Reform Progressj )
0.82
0.05 a
0.05
ln(Remotenessi )
0.03
0.04
0.03
0.03
ln(Remotenessj )
-0.03
-0.03
0.03
0.03
Common languageij
1.17a
1.15a
0.04
0.04
0.70
a
0.05
Landlocki
0.56
0.70a 0.05
a
0.08
a, b
0.83a
0.05
Common borderij
Notes:
1.36a
0.57a 0.03
Landlockj
0.35
a
0.36a
0.03
0.03
Year fixed effects # of observations R-squared
Yes 21409 0.64
Yes 21409 0.64
and
c
define 1%, 5% and 10% significance level respectively. Standard Errors (heteroskedasticity robust) are presented in italic.
14
Table 4: Border effect with geographic variablesa Period Average Border Effect Border Effect Soviet Union Border Effect Former Yugoslavia Border Effect Czechoslovakia a The
Without IRP 93-95 96-98 99-01 8.41 7.17 6.05 8.58 7.24 5.93 7.54 7.17 7.09 6.75 5.53 3.97
93-95 14.01 14.15 15.49 5
With IRP 96-98 99-01 9.12 7.92 8.76 7.61 13.73 12.18 4.85 3.42
border effect is computed as the antilog of the coefficient of each variable.
15
Table 5: Estimations with regional integration Period 93-01 Dependent variable: ln exports (5) ln(GNPit ) 0.86a ln(GNPjt )
0.01
0.01
0.74a
0.74a
0.009 a
0.009 a
-0.81
ln(Distanceij )
0.019 a
Homejt
agreements 93-01 (6) 0.86a
-0.81 0.02
2.45 0.07
Soviet Uniont
2.37a
Former Yugoslaviat
3.01a
0.07
0.13
2.68a
Czechoslovakiat
0.07
ln(Index of Reform Progressi )
1.16
a
0.05
0.05
ln(Index of Reform Progressj )
0.63
a
0.63a
0.05
0.05
ln(Remotenessi )
0.10a
0.10a
0.03
0.03
ln(Remotenessj )
0.05
0.06c
0.03
0.03
EUij
1.36
a
0.03
CEFTAij
0.43
0.43a 0.05
Baltic Free Trade Areaij
0.47
a
0.55a
0.08
0.09
Association Agreementsij
0.45a
0.45a
0.03
0.03
-0.03
0.04
0.21
0.22
Yes 21409 0.64
Yes 21409 0.64
Custom Union in FSUij Year fixed effects # of observations R-squared a, b
1.37a 0.03
a
0.04
Notes:
1.16a
and
c
define 1%, 5% and 10% significance level respectively. Standard Errors (heteroskedasticity robust) are presented in italic.
16
Table 6: Border effect with regional integration agreementsa Period Average Border Effect Border Effect Soviet Union Border Effect Former Yugoslavia Border Effect Czechoslovakia a The
Without IRP 93-95 96-98 99-01 10.49 9.12 7.54 9.97 8.50 6.89 13.06 14.73 14.73 27.66 17.29 9.87
93-95 17.12 16.61 20.90 20.90
With IRP 96-98 99-01 10.91 9.02 9.87 8.25 21.11 18.91 15.18 9.21
border effect is computed as the antilog of the coefficient on each variable.
17
5
Conclusion
The disintegration of Soviet Union, former Yugoslavia and Czechoslovakia resulted in the creation of new boundaries between the constituent Republics. In this paper we show that the political disintegration itself is not the only cause of trade disintegration and that a more plausible explanation is the transition process. The transition of the newly independent states towards the market economy created the conditions for redeployment of their trade and the trade disintegration of the area. Even if Successor States were still together in the same country they would surely trade relatively less between themselves than in the eighties. However the impact of transitional reforms have been different for the three considered areas. They affected the most trade between successor states of former Yugoslavia whereas they had almost no impact on trade integration between the Czech Republic and Slovakia.
18
References Alesina A. and E. Spolaore (1997), “On the number and size of nation”, Quarterly Journal of Economics, 112(4), 1027-1056. Alesina A., E. Spolaore and R. Wacziarg (2000), “Economic integration and political disintegration”, American Economic Review, 90(5), 1276-1296. Anderson J.E. and E. Van Wincoop (2004), “Trade costs”, Mimeo. Anderson J.E. and E. Van Wincoop (2003), “Gravity with gravitas: a solution to the border effect puzzle”, American Economic Review, 93 (1), 170-192. Andreff M. and W. Andreff (1995), “Economic disintegration in Eastern Europe: towards a new integration?” in B. Dallago and G. Pegoretti, eds., Integration and disintegration in European Economies, Dartmouth Publishing, 113-141. Baier S. and Bergstrand J.H. (2001), “The Growth of World Trade: Tariffs, Transport Costs, and Income Similarity”, Journal of International Economics, 53(1), 1-27. Bergstrand J.H. (1985), “The gravity equation in international trade: some microeconomic foundations and empirical evidence”, Review of Economics and Statistics, 67 (3), 474-481. Bergstrand J.H. (1989), “The generalised gravity equation, monopolistic competition and the factor-propositions theory of international trade”, Review of Economics and Statistics, 71(1), 143-153. Bolton P. and G. Roland (1997), “The break-up of nations: a political economy analysis”, Quarterly Journal of Economics, 112(4), 1057-1090. Disdier A.-C. and J.-L. Mucchielli (2002), “Biais domestique et concurrence des processus d’int´egration dans les ´echanges de l’Europe du Sud-Est”, Economie Internationale, 89-90, 209-225. Djankov S. and C. Freund (2002), “Trade flows in the Soviet Union - 1987 to 1996”, Journal of Comparative Economics, 30(1), 76-90. Eaton, J. and S. Kortum (2002), “Technology, Geography and Trade”, Econometrica, 70(5), 1741-1779. EBRD (2002), Transition Report, London. 19
Falcetti E., M. Raiser and P. Sanfey (2002), “Defying the odds: initial conditions, reforms and growth in the first decade of transition”, Journal of Comparative Economics, 30(2), 229-250. Feenstra, R. (2004), Advanced International Trade: Theory and Evidence, Princeton, Princeton University Press. Fidrmuc J. and J. Fidrmuc (2003), “Disintegration and trade”, Review of International Economics, 11(5), 811-829. Harrigan, J. (2002), “Specialization and the volume of trade: do the data obey the laws?”, in Choi, K. and J. Harrigan, eds., Handbook of international trade, Basil Blackwell, Oxford, UK. Havrylyshyn O. and H. Al-Atrash (1998), ”Opening up and geographic diversification of trade in transition economies”, IMF Working Paper, 22. Helliwell J. F. (1998), How much do national borders matter?, The Brookings Institution Press. Kaminski, B. and M. de la Rocha (2003), “Stabilization and association process in the Balkans: integration options and their assessment”, World Bank Policy Research Working Paper, 3108. Lamotte (2004), “Disintegration and trade in Southeastern Europe”, Les cahiers de la MSE, S´erie Jaune, University of Paris 1. Maurel M. and G. Cheikbossian (1998), “The new geography of Eastern European Trade”, Kyklos, 51(1), 45-71. McCallum, J. (1995), “National borders matter: Canada-U.S. regional trade patterns”, American Economic Review, 85, 615-623. Melitz, J. (2004), “Distance, trade and political association”, Mimeo. Nitsch, V.(2000), “National Borders and International Trade: Evidence from the EU”, Canadian Journal of Economics, 33(4), 1091-1105. Pomfret, R. (2002), “National borders and disintegration of market area in central Asia after 1991”, in U. M¨ uller and H. Schultz, eds., National borders and economic disintegration in Modern East Central Europe, 261-79, Frankfurter Studien zur Grenzregion, vol. 8, Berlin Verlag arno Spritw Gmbh, Berlin Germany. 20
Sachs, J. D. (1996), “The transition at mid-decade”, American Economic Review, 86(2), 128-133. Wei, S.-J. (1996),“Intra-national versus international trade: how stubborn are nations in global integration?”, NBER Working Paper, 5531.
21
Appendix
Figure 1: Average border effect
Figure 2: Border effect of Soviet Union
22
Figure 3: Border effect of former Yugoslavia
Figure 4: Border effect of Czechoslovakia
23
Figure 5: Average border effect with geographic dummies
Figure 6: Border effect of Soviet Union with geographic dummies
24
Figure 7: Border effect of former Yugoslavia with geographic dummies
Figure 8: Border effect of Czechoslovakia with geographic dummies
25
Figure 9: Average border effect with regional trade agreements
Figure 10: Border effect of Soviet Union with regional trade agreements
26
Figure 11: Border effect of former Yugoslavia with regional trade agreements
Figure 12: Border effect of Czechoslovakia with regional trade agreements
27
Table 7: Index of reform progress (IRP) 1993 1994 Non transition countries 38,7 38,7 Albania 16,3 18 Armenia 13 13,6 Azerbaidjan 11 11 Belarus 13,7 13,7 Bosnia-Herzegovina 11 10 Bulgaria 17 19,3 Croatia 18,3 22,4 Czech Republic 26,7 28,7 Estonia 22,7 25,7 Georgia 12 12 Hungary 25,3 27,7 Kazakhstan 13 15 Kyrgyzstan 15 21,3 Latvia 18 23 Lituania 21 23 Macedonia 15,6 19,3 Moldova 14,7 17,7 Poland 26 27 Romania 16 19,3 Russia 19 21,1 Serbia-Montenegro 14,3 12,3 Slovakia 24,7 26,4 Slovenia 23,7 25,1 Tajikistan 12,7 12,7 Turkmenistan 9 9,7 Ukraine 11,7 12,7 Uzbekistan 12 17
1995 1996 1997 38,7 38,7 38,7 20,3 21,7 21,7 18,4 20,7 21,3 14,7 16 18,3 17,4 15,7 14 10 11,3 13,3 20,3 20,7 24,3 23,4 25,4 26 28,7 29,6 30 27 27,6 28,9 17 21 23,3 30 30,3 32,2 19,4 23 24 23 24 24 24 25,3 26,9 24,7 25,7 26,3 20,3 20,7 20,7 22 22,3 23 28,7 29,3 30,2 20,7 20,7 23,7 22,7 25 25,9 12,3 12,3 12,7 27,1 27,7 27,7 25,8 26,7 27,3 14,7 15 15,6 10,7 10,7 12,7 19 20 21,2 19,4 19,4 18,8
Sources: EBRD Transition Report 2002 & author’s calculation.
28
1998 38,7 21,7 22,6 20 13 17,3 24,3 26,3 30,3 29,6 23,6 32,8 24,3 23,6 26,7 26,6 22,4 23,6 30,8 24,3 22,3 12,4 28 28,3 17,7 12,4 20,9 18,4
1999 38,7 21,7 22,9 20 13 17,3 25,2 26,9 30,6 30,7 23,9 33,2 23,3 23,6 27 27,3 22,4 23,6 30,8 25,1 21,7 12,4 28,6 28,9 18 12,4 21,2 17,7
2000 38,7 23 22,9 20,7 13,7 18,3 27 27,5 30,9 31,3 25,2 33,6 23,9 23,6 27,6 28,4 23,7 23,9 31,5 25,7 22,3 12,7 28,9 29,3 18,6 11,7 22,3 17,4
2001 38,7 23,3 24,3 21,3 14,3 19 27 27,9 31,6 31,6 25,2 33,6 24,3 23,6 27,9 29,3 24,3 24,2 31,8 26,3 23 16 29,5 29,3 19,1 11 22,6 18,5
Olivier Lamotte‡
September 10, 2004 - Very preliminary draft
Abstract This paper studies the scope of trade disintegration in a context of transition and political separatism. We use the gravity equation and the border effect methodology to evaluate the home bias of Soviet Union, former Yugoslavia and Czechoslovakia from 1993 to 2001. We find high home biases for each federation, even if they significantly decrease during our sample period. We also find strong evidence that transition had a positive impact on trade and on the reduction of the home bias. Finally we point out some regional differences, as transition process explains a smaller part of the home bias in the case of Czechoslovakia.
JEL Classification: F13, F15 Keywords: disintegration, borders, transition
∗
We are grateful to Fabian Gouret, Julie Lochard, Mathilde Maurel and Daniel Mirza for providing valuable comments. † LESSOR, University of Rennes 2, ROSES-CNRS, University of Paris 1, 106-112 boulevard de l’Hˆopital, 75647 Paris cedex 13, E-mail: [email protected] ‡ University of Cergy-Pontoise and ROSES-CNRS, University of Paris 1, 106-112 boulevard de l’Hˆopital, 75647 Paris cedex 13, E-mail: [email protected]
1
1
Introduction
Political disintegration leads to the creation of new boundaries and the erection of new barriers to trade (tariff and non-tariff barriers). As a result, trade intensity between successor states tends to decrease. Prior studies found evidence that political disintegration is a source of trade disintegration in Eastern European countries (Djankov and Freund, 2002; Fidrmuc and Fidrmuc, 2003) and Central Asia (Pomfret, 2002). These results are challenging for two reasons. First, other things being equal, sucessor states continue to trade more between themselves than with other partners. Second, the concurrent transition process has often been ignored. This paper addresses the question of trade disintegration in a setting of political separatism and transition. We wonder whether the decrease of trade intensity is the sole result of political disintegration. Our main argument is that a part of the trade disintegration can be explained by the transition process. The literature documents that transition countries in Eastern Europe have experienced large reorientation of their trade flows towards western economies (Andreff and Andreff, 1995; Maurel and Cheikbossian, 1998) and 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). The transition process seems to favor trade disintegration. The analysis of political disintegration is quite rare in the economic literature. Almost all studies focus on the motives and the conditions for disintegration1 and do not study the consequences of such a process. The border effect methodology opens a new field of research for the study of disintegration experiences. Following the influential work of McCallum (1995), it is now widely documented that national borders matter for trade. As an example, a Canadian province trades 20 times more with other Canadian provinces than with American states of a given size and distance (McCallum, 1995). 1
See Alesina and Spolaore, 1997; Alesina et al., 2000; Bolton and Roland, 1997.
2
The application of the border effect methodology to the analysis of disintegration permits to determine the level of trade barriers between successor states compared to the external barriers. Wei (1996) defines the border effect as “an all-inclusive summary of barriers to trade”. Consequently, the border effect of a politically disintegrated region is defined as an all-inclusive summary of remaining ties that make intra-successor states trade relatively larger than the trade with other countries. In case of a political disintegration, we expect a decrease of the former national home bias, as a substantial part of the former intra-national trade flows decrease. We test this idea and the influence of the transition process by re-integrating virtually the former country, i.e. by assuming that political disintegration did not take place. For this issue, Soviet Union, former Yugoslavia and Czechoslovakia constitute relevant examples. They enable to measure the speed and the scope of trade disintegration in a context of transition and political separatism. We use a gravity equation and a sample of 52 countries (cf. infra for details) to evaluate the home bias of these former Federations from 1993 to 2001 and provide some explanations to their evolution. This paper makes several contributions to the existing literature. We find a high home bias in the former Federations during the nineties. This bias decrease significantly during our sample period. We also find strong evidence that transition explained a large part of the home bias decrease. Finally we show that transition had a different impact on the economic disintegration of the countries under focus. We proceed as follows. In section 2, we discuss the econometric specification of the model and the dataset. Results are laid in section 3. In section 4, we provide some robustness checks and we conclude in section 5.
3
2 2.1
Model set-up and data description The gravity model
The empirics are based on the gravity model, which capacity to explain trade flows has been successful empirically. The traditional gravity equation takes the following form:
ln(Xij ) = β0 +β1 ln(GN Pi )+β2 ln(GN Pj )+β3 ln(Distanceij )+β4 (HOM Eij )+εij (1) where (Xij ) is merchandise exports from country i to country j, (GN Pi ) and (GN Pj ) are national incomes of the exporter and importer, (Distanceij ) the distance between i and j, and following McCallum (1995) (HOM Eij ) is a dummy variable equal to 1 for intra-national trade and 0 for international trade. More precisely, it is equal to 1 for trade flows between successor states of Soviet Union, Former Yugoslavia and Czechoslovakia and 0 otherwise. The (HOM Eij ) variable measures the size of the border effect and allows us to determine the magnitude of the disintegration process. The coefficients on national incomes are expected to be positive and the distance coefficient negative. The coefficient of (HOM Eij ) will be positive if intra-national trade flows are larger than international ones, i.e. if trade flows among successor states of former Federations are larger than flows between the successor states and their other trading partners. The antilog of the coefficient indicates the size of the border effect. In the core of the paper, we divide the (HOM Eij ) variable into three dummies, one for each former Federation. As an example, the dummy “Soviet Union” is equal to 1 for inter-(former) Soviet Union trade and 0 otherwise. In this way, we estimate the average border effect of each former federation. εij is an error term. In order to determine the effect of transition on the disintegration process we introduce in equation 1 an index of progress in transition for each partner.
4
The new specification is the following: ln(Xij ) = β0 + β1 ln(GN Pi ) + β2 ln(GN Pj ) + β3 ln(Distanceij ) + β4 (HOM Eij ) + β5 ln(IRPi ) + β6 ln(IRPj ) + εij
(2)
where (IRPi ) and (IRPj ) stands for Index of Reform Progress. Estimated coefficients of progress in transition are expected to be positive. According to the theory, the traditional gravity equation 2 is misspecified while omitting “multilateral trade resistance” variables (Anderson and Van Wincoop, 2003). Such variables “summarize the average trade resistance between a country and its trading partners” (Anderson and Van Wincoop, 2004: 28). Given a bilateral trade relationship between i and j, an increase of the multilateral resistance of j raises its trade with i, other things being equal. Put differently, multilateral resistance is related to the distance of country pairs relative to their trading partners. “Trade will be higher between country pairs that are far from the rest of the world than between country pairs that are close to the rest of the world” (Harrigan, 2002). This relative distance is difficult to capture but can be approximated by several ways: the use of a complex non-linear least square estimation (Anderson and van Wincoop, 2003), the use of prices indexes (Bergstrand, 1985, 1989; Baier and Bergstrand, 2001) which do not capture all non-pecuniary trade costs, and country-specific dummies (Anderson and van Wincoop, 2003; Eaton and Kortum, 2002 and Feenstra, 2004). The latter is the most simple solution among the efficient ones. However, the use of country-specific dummies does not fit with the aim of the paper for two reasons. First, the country-specific dummies are time-invariant but multilateral resistances variables on a several years sample are not. Second, introducing jointly country-specific dummies and country variables (rather than country pair variables) induces a serious problem of multicolinearity, since explanatory variables are seriously correlated. In fact, country progress in transition, which hardly varies over time, is captured by fixed effects (see Melitz 2004 for similar problems). It would therefore be impossible to estimate the effect of transition. As a result, we 5
follow Harrigan (2002) in deriving an ad-hoc proxy of the relative distance as a remoteness index for each partner. The estimated equations take the following form:
ln(Xij ) = β0 + β1 ln(GN Pi ) + β2 ln(GN Pj ) + β3 ln(Distanceij ) + β4 (HOM Eij ) + β5 ln(IRPi ) + β6 ln(IRPj ) + β7 ln(Remotenessi ) + β8 ln(Remotenessj ) + εij
(3)
where (Remotenessi ) and (Remotenessi ) are the index of remoteness of countries i and j. Estimated coefficients of remoteness are expected to be positive. Methodological problems, as well as definitions of variables and presentation of data are discussed in the following sub-section.
2.2
The database
Our sample consists of 52 countries-22 EU members2 , 11 non-EU OECD countries3 , 12 CIS countries4 and 7 Southeastern European countries5 . The time span covers the period from 1993 to 2001. We therefore have a potential of 52x51x9=23868 observations. Bilateral trade exports (Xij) come from the Chelem-Cepii database. To our knowledge this is the only database that allows working on such a long time span for all Southeastern countries. Missing values come from the IMF DOTS (Direction Of Trade Statistics database). Flows are measured in millions of current dollars. GNP and distance also come from the Chelem-Cepii database. GNP is measured in millions of current dollars. Distance is the 2
Austria, Belgium and Luxemburg (together), Czech Republic, Denmark, Estonia, Finland, France, Germany, Great Britain, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Netherlands, Poland, Portugal, Slovakia, Slovenia, Spain and Sweden. 3 Australia, Canada, Iceland, Japan, Korea, Mexico, New Zealand, Norway, Switzerland, Turkey, United States 4 Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia, Tajikistan, Turkmenistan, Ukraine, Uzbekistan 5 Albania, Bosnia-Herzegovina, Bulgaria, Croatia, Macedonia, Romania and SerbiaMontenegro.
6
shortest distance “as the crow flies” between capital cities in kilometers. Remoteness takes account of the fact that “what matters for bilateral export volume is not just the absolute distance between the two countries, but their geographic positions relative to all other countries”(Wei, 1996: 5). It can be defined as a weighted average of distances between a country i and all its partners. Several measures of remoteness have been proposed in the literature6 We tried all of them but since they give very similar results we follow Nitsch (2000). He proposes the following formula: Remotenessi =
1 Pk Σk ( GN ) Dik
The introduction of an index of transition progress required finding some data on transition for the nine years of the sample and for all countries. The European Bank for Reconstruction and Development (EBRD) provides the only complete database on such an issue7 . The EBRD measures, on an annual basis, the progress of transition countries towards a market economy along several dimensions. The dimensions are the following8 : (1) price liberalization, (2) foreign exchange and trade liberalization, (3) small-scale privatization, (4) large-scale privatization, (5) enterprise reform, (6) competition policy, (7) infrastructure reform, (8) banking sector reform and (9) reform of non-banking financial institutions. The progress in each dimension is scored on a scale from 0.7 (low level) to 4.3 (high level). Non transition economies are considered as benchmarks and were given the score of 4.3 in all dimensions. In order to get a unique index of transition progress we computed the so-called Index of Reform Progress (IRP), constructed by Sachs (1996). It is obtained by simply summing the 6
Wei (1996) measures the remoteness of a country k as: Remotenessi = Σi (wk Di k) GN Pk where wk = GN Pworld and k = i, j. Helliwell (1998) proposed another measure of remoteD
ness: Remotenessi = Σk ( GNkjPk ) 7 Other databases (Heritage Foundation, Fraser Institute, Penn World Table, IMF Country Reports, World Bank development indicators) do not include all successor states of Soviet Union and former Yugoslavia for the nineties. 8 For availability reasons we considered nine of the eleven dimensions, excluding “legal extensiveness” and “legal effectiveness”.
7
EBRD’s subindexes9 . Some limits have to be highlighted before proceeding to the results. First, one has to be cautious regarding South East European countries: Kaminski and De la Rocha (2003) highlight huge gaps when comparing mirror statistics, i.e. the same flow declared by the exporter and the importer. Second, wars and embargoes on the territories of former Yugoslavia in the nineties disrupted standard trade routes and created incentives and means for the development of traffics and more generally of the unofficial economy. Third, measures of the reform process are highly imperfect because the EBRD “have relied on subjective indices rather than directly observable variables” (Falcetti, Raiser and Sanfey, 2002).
3
Results
We now proceed to the estimations of the specification 3. The results are presented in table 1 and 2. We first estimate a single border effect (Estimation 1) and then break it out to measure the border effect for each dismissed Federation (Estimation 2). Finally we study the impact of transition on the border effect and we divide the sample into three time-periods (Table 2). Estimation (1) exhibits positive coefficients of supply (GN Pit ) and demand (GN Pjt ) forces on the volume of trade as well as a negative coefficient of distance (Distanceij ). This last result means that, other factors fixed, increasing bilateral distance impedes trade. The coefficient on (HOM Eij ) is positive and significant. The border effect is obtained by computing the antilog of the coefficient. On average, between 1993 and 2001, trade between successors states of the three former Federations was more than 12 (exp(2.53)=12.55) times larger than their international trade. As expected this bias is large and is consistent with what is found in the literature for the “real” intra-national trade. The positive coefficient on the Index of Reform Progress variables indicates that reforms favors trade. Moreover it seems that reforms in the export9
The values of the IRP index for each country are available in the appendix (Table 7).
8
Table 1: Average border effects of former Federations Period 93-01 93-01 Dependent variable: ln exports (1) (2) a ln(GNPi ) 0.87 0.88a 0.01
ln(GNPj )
0.75
0.01 a
0.01
ln(Distanceij )
0.01
-0.93
a
0.02
Homeij
2.53
0.75a -0.93a 0.02
a
0.07
2.48a
Soviet Unionij
0.07
2.92a
Former Yugoslaviaij
0.14
2.67a
Czechoslovakiaij
0.07
ln(Index of Reform Progressi )
1.38
a
1.38a
0.05
0.05
ln(Index of Reform Progressj )
0.87a
0.87a
0.05
0.05
ln(Remotenessi )
0.08a
0.08a
0.03
0.03
0.01
0.01
0.03
0.03
Yes 21409 0.63
Yes 21409 0.63
ln(Remotenessj ) Year fixed effects # of observations R-squared Notes:
a, b
and
c
define 1%, 5% and 10% significance level respectively. Standard Errors (heteroskedasticity robust) are presented in italic.
9
ing country have a higher stimulating effect than reforms in the importing country. The coefficients of remoteness (Remotenessi and Remotenessj ) are positive. In estimation 2 the (HOM Eij ) variable is divided into three dummies: (SovietU nionij ), (f ormerY ugoslaviaij ) and (Czechoslovakiaij ). Unsurprisingly, trade intensity between successor states of the three former Federations have been higher than trade intensity with outside partners. For 1993-2001 the border effect is 12 (=exp(2.48)) for Soviet Union, more than 18 (=exp(2.92)) for former Yugoslavia, and 14 (=exp(2.67)) for Czechoslovakia. The other variables are not affected by the new specification. The fact that the coefficient on HOM Ei j is the lowest for the Soviet Union is probably explained by the various degrees of integration that can be found in the area. The picture would probably be different if we had considered sub-regions of Soviet Union. For example, the would certainly be higher for each Republic with Russia or between the Baltic States and lower between more remote Republics, for example between the Baltic Republics and the Central Asian Republics. The same remark applies for former Yugoslavia. Lamotte (2003) points out very different degrees of integration between the Republics of former Yugoslavia on a bilateral basis. In table 2 we present the impact of the progress in reforms ij on the border effect and its evolution. The estimations are first run without and then with the Index of Reform Progress variables (hereafter called the benchmark estimation). Table 2: Evolution of the average border effectsa Period Average Border Effect Border Effect Soviet Union Border Effect Former Yugoslavia Border Effect Czechoslovakia a The
Without IRP 93-95 96-98 99-01 11.59 9.02 7.39 11.36 8.76 7.10 11.82 10.80 10.07 22.87 13.87 10.38
With IRP 93-95 96-98 99-01 19.49 11.59 9.78 19.49 10.91 9.21 20.91 19.49 15.96 20.91 13.87 10.38
border effect is computed as the antilog of the coefficient of each variable.
The first striking result is that the coefficient on (HOM Eij ) is always 10
higher when the progress in reform is controlled for. This result could be surprising. Actually the interpretation is that if transitional reforms had not taken place the border effect would have been higher. Therefore transition had an impact on the border effect. More precisely reforms have reduced the gap between the relative sizes of the inside and the outside borders of the former Federations. A second interesting result emerges from the observation of the evolution of the border effect. Figure 1 plots the evolution of the average border effect over time. The gap between the two curves, i.e. the border effect in presence or not of the reform index (IRPi and IRPj ), is decreasing. The explanatory power of the transition as a source of trade disintegration is therefore decreasing over time. Other interesting remarks can be drawn from the observation of the evolution of the border effects. In figures 2 to 4 we present the evolution of the border effects of the three dismissed federations. The first interesting point is that the trade disintegration took different intensity in the three considered areas. On the sample period the border effect was divided by two for Czechoslovakia whereas it decreased less than 15% for former Yugoslavia. The second interesting point is that the impact of reforms on the border effect has been different for each former Federation. In the three cases the gap between the two curves representing the border effects tends to decrease. But the part of the trade disintegration explained by the transition process is quite different between the three cases. The most surprising result occurs for Czechoslovakia (figure 4). For the first period the curves are inverted compared to the two previous figures, meaning that reforms had a positive impact on the mutual integration of the two successor states. From the mideighties it seems that the reforms did not have an impact on trade integration between the Czech Republic and Slovakia any more. In the following section we will test whether these results are robust to other specifications.
11
4
Robustness checks
In order to check the robustness of our results we augment the basic specification with two sets of additional variables that concern geography and regional integration agreements. We proceed as in the previous section; we measure a single border effect, then we break out the effect by zone and finally we determine its evolution on three sub-periods. We first control for geographical characteristics by augmenting equation (3) with dummy variables that are usually included in gravity equations. (Common languageij ) and (Common borderij ) take the value 1 if the two partners share respectively the same language or a common border. They are both expected to have a positive sign. (Landlockenessi ) and (Landlockenessj ) take the value 1 if the exporter or the importer is a landlocked country, i. e. with no direct access to the sea. It is expected with a negative sign. Results are presented in table 3 and 4. The only significant change between estimations (1) and (2) (table 1) and estimations (3) and (4) (table 3) concerns the level of the border effect. In all cases the coefficient is lower. It is not surprising since a characteristic of member states of a Federation is to have a common border and/or a common language. A part of the effect is therefore captured by the common language and border dummies. From table 4 and figures 5 to 8 we can see that the trends and the extent of the gaps remain almost the same. The second set of robustness checks concerns regional integration agreements. We add the following dummy variables 10 : (EUij ), (CEFTAij ), (Baltic FTAij ), (Association Agreementsij ) and (Custom Union in FSUij ). Each variable takes the value 1 if trade occurs between two members of the agreement. Results are laid in tables 5 and 6 and figures 9 to 12. The previous results are very slightly affected by the introduction of new variables. The coefficients on the trade agreements are significant and consistent with the 10
(EU ) stands for the European Union, (CEFTA) for the Central European Free Trade Agreement, (Baltic FTA) for the free trade agreement between Baltic states, (Association Agreements) for the trade agreements between the EU and Central and Eastern European Countries and (Custom Union in FSU ) for the custom union between Russia, Belarus, Kazakhstan, Tajikistan and Kirghizstan.
12
literature, except for the custom union in Soviet Union. From the comparison from tables 2 and 6 we can observe that the coefficients of the border effects dummies are a bit higher in the specification without IRP. Concerning the trend of the curve, it is similar to the previous specifications except for former Yugoslavia. The introduction of additional variables reduce slightly the decrease of the border effect.
13
Table 3: Estimations with geographic variables Period 93-01 93-01 Dependent variable: ln exports (3) (4) a ln(GNPi ) 0.88 0.88a 0.01
0.01
ln(GNPj )
0.75a
0.75a
0.01
0.009
ln(Distanceij )
-0.82a
-0.82a
0.02
0.02
Homeij
2.27
a
0.07
2.25a
Soviet Unionij
0.07
Former Yugoslaviaij
2.61a
Czechoslovakiaij
1.49a
0.16
0.08
ln(Index of Reform Progressi )
1.36
a
0.05
ln(Index of Reform Progressj )
0.82
0.05 a
0.05
ln(Remotenessi )
0.03
0.04
0.03
0.03
ln(Remotenessj )
-0.03
-0.03
0.03
0.03
Common languageij
1.17a
1.15a
0.04
0.04
0.70
a
0.05
Landlocki
0.56
0.70a 0.05
a
0.08
a, b
0.83a
0.05
Common borderij
Notes:
1.36a
0.57a 0.03
Landlockj
0.35
a
0.36a
0.03
0.03
Year fixed effects # of observations R-squared
Yes 21409 0.64
Yes 21409 0.64
and
c
define 1%, 5% and 10% significance level respectively. Standard Errors (heteroskedasticity robust) are presented in italic.
14
Table 4: Border effect with geographic variablesa Period Average Border Effect Border Effect Soviet Union Border Effect Former Yugoslavia Border Effect Czechoslovakia a The
Without IRP 93-95 96-98 99-01 8.41 7.17 6.05 8.58 7.24 5.93 7.54 7.17 7.09 6.75 5.53 3.97
93-95 14.01 14.15 15.49 5
With IRP 96-98 99-01 9.12 7.92 8.76 7.61 13.73 12.18 4.85 3.42
border effect is computed as the antilog of the coefficient of each variable.
15
Table 5: Estimations with regional integration Period 93-01 Dependent variable: ln exports (5) ln(GNPit ) 0.86a ln(GNPjt )
0.01
0.01
0.74a
0.74a
0.009 a
0.009 a
-0.81
ln(Distanceij )
0.019 a
Homejt
agreements 93-01 (6) 0.86a
-0.81 0.02
2.45 0.07
Soviet Uniont
2.37a
Former Yugoslaviat
3.01a
0.07
0.13
2.68a
Czechoslovakiat
0.07
ln(Index of Reform Progressi )
1.16
a
0.05
0.05
ln(Index of Reform Progressj )
0.63
a
0.63a
0.05
0.05
ln(Remotenessi )
0.10a
0.10a
0.03
0.03
ln(Remotenessj )
0.05
0.06c
0.03
0.03
EUij
1.36
a
0.03
CEFTAij
0.43
0.43a 0.05
Baltic Free Trade Areaij
0.47
a
0.55a
0.08
0.09
Association Agreementsij
0.45a
0.45a
0.03
0.03
-0.03
0.04
0.21
0.22
Yes 21409 0.64
Yes 21409 0.64
Custom Union in FSUij Year fixed effects # of observations R-squared a, b
1.37a 0.03
a
0.04
Notes:
1.16a
and
c
define 1%, 5% and 10% significance level respectively. Standard Errors (heteroskedasticity robust) are presented in italic.
16
Table 6: Border effect with regional integration agreementsa Period Average Border Effect Border Effect Soviet Union Border Effect Former Yugoslavia Border Effect Czechoslovakia a The
Without IRP 93-95 96-98 99-01 10.49 9.12 7.54 9.97 8.50 6.89 13.06 14.73 14.73 27.66 17.29 9.87
93-95 17.12 16.61 20.90 20.90
With IRP 96-98 99-01 10.91 9.02 9.87 8.25 21.11 18.91 15.18 9.21
border effect is computed as the antilog of the coefficient on each variable.
17
5
Conclusion
The disintegration of Soviet Union, former Yugoslavia and Czechoslovakia resulted in the creation of new boundaries between the constituent Republics. In this paper we show that the political disintegration itself is not the only cause of trade disintegration and that a more plausible explanation is the transition process. The transition of the newly independent states towards the market economy created the conditions for redeployment of their trade and the trade disintegration of the area. Even if Successor States were still together in the same country they would surely trade relatively less between themselves than in the eighties. However the impact of transitional reforms have been different for the three considered areas. They affected the most trade between successor states of former Yugoslavia whereas they had almost no impact on trade integration between the Czech Republic and Slovakia.
18
References Alesina A. and E. Spolaore (1997), “On the number and size of nation”, Quarterly Journal of Economics, 112(4), 1027-1056. Alesina A., E. Spolaore and R. Wacziarg (2000), “Economic integration and political disintegration”, American Economic Review, 90(5), 1276-1296. Anderson J.E. and E. Van Wincoop (2004), “Trade costs”, Mimeo. Anderson J.E. and E. Van Wincoop (2003), “Gravity with gravitas: a solution to the border effect puzzle”, American Economic Review, 93 (1), 170-192. Andreff M. and W. Andreff (1995), “Economic disintegration in Eastern Europe: towards a new integration?” in B. Dallago and G. Pegoretti, eds., Integration and disintegration in European Economies, Dartmouth Publishing, 113-141. Baier S. and Bergstrand J.H. (2001), “The Growth of World Trade: Tariffs, Transport Costs, and Income Similarity”, Journal of International Economics, 53(1), 1-27. Bergstrand J.H. (1985), “The gravity equation in international trade: some microeconomic foundations and empirical evidence”, Review of Economics and Statistics, 67 (3), 474-481. Bergstrand J.H. (1989), “The generalised gravity equation, monopolistic competition and the factor-propositions theory of international trade”, Review of Economics and Statistics, 71(1), 143-153. Bolton P. and G. Roland (1997), “The break-up of nations: a political economy analysis”, Quarterly Journal of Economics, 112(4), 1057-1090. Disdier A.-C. and J.-L. Mucchielli (2002), “Biais domestique et concurrence des processus d’int´egration dans les ´echanges de l’Europe du Sud-Est”, Economie Internationale, 89-90, 209-225. Djankov S. and C. Freund (2002), “Trade flows in the Soviet Union - 1987 to 1996”, Journal of Comparative Economics, 30(1), 76-90. Eaton, J. and S. Kortum (2002), “Technology, Geography and Trade”, Econometrica, 70(5), 1741-1779. EBRD (2002), Transition Report, London. 19
Falcetti E., M. Raiser and P. Sanfey (2002), “Defying the odds: initial conditions, reforms and growth in the first decade of transition”, Journal of Comparative Economics, 30(2), 229-250. Feenstra, R. (2004), Advanced International Trade: Theory and Evidence, Princeton, Princeton University Press. Fidrmuc J. and J. Fidrmuc (2003), “Disintegration and trade”, Review of International Economics, 11(5), 811-829. Harrigan, J. (2002), “Specialization and the volume of trade: do the data obey the laws?”, in Choi, K. and J. Harrigan, eds., Handbook of international trade, Basil Blackwell, Oxford, UK. Havrylyshyn O. and H. Al-Atrash (1998), ”Opening up and geographic diversification of trade in transition economies”, IMF Working Paper, 22. Helliwell J. F. (1998), How much do national borders matter?, The Brookings Institution Press. Kaminski, B. and M. de la Rocha (2003), “Stabilization and association process in the Balkans: integration options and their assessment”, World Bank Policy Research Working Paper, 3108. Lamotte (2004), “Disintegration and trade in Southeastern Europe”, Les cahiers de la MSE, S´erie Jaune, University of Paris 1. Maurel M. and G. Cheikbossian (1998), “The new geography of Eastern European Trade”, Kyklos, 51(1), 45-71. McCallum, J. (1995), “National borders matter: Canada-U.S. regional trade patterns”, American Economic Review, 85, 615-623. Melitz, J. (2004), “Distance, trade and political association”, Mimeo. Nitsch, V.(2000), “National Borders and International Trade: Evidence from the EU”, Canadian Journal of Economics, 33(4), 1091-1105. Pomfret, R. (2002), “National borders and disintegration of market area in central Asia after 1991”, in U. M¨ uller and H. Schultz, eds., National borders and economic disintegration in Modern East Central Europe, 261-79, Frankfurter Studien zur Grenzregion, vol. 8, Berlin Verlag arno Spritw Gmbh, Berlin Germany. 20
Sachs, J. D. (1996), “The transition at mid-decade”, American Economic Review, 86(2), 128-133. Wei, S.-J. (1996),“Intra-national versus international trade: how stubborn are nations in global integration?”, NBER Working Paper, 5531.
21
Appendix
Figure 1: Average border effect
Figure 2: Border effect of Soviet Union
22
Figure 3: Border effect of former Yugoslavia
Figure 4: Border effect of Czechoslovakia
23
Figure 5: Average border effect with geographic dummies
Figure 6: Border effect of Soviet Union with geographic dummies
24
Figure 7: Border effect of former Yugoslavia with geographic dummies
Figure 8: Border effect of Czechoslovakia with geographic dummies
25
Figure 9: Average border effect with regional trade agreements
Figure 10: Border effect of Soviet Union with regional trade agreements
26
Figure 11: Border effect of former Yugoslavia with regional trade agreements
Figure 12: Border effect of Czechoslovakia with regional trade agreements
27
Table 7: Index of reform progress (IRP) 1993 1994 Non transition countries 38,7 38,7 Albania 16,3 18 Armenia 13 13,6 Azerbaidjan 11 11 Belarus 13,7 13,7 Bosnia-Herzegovina 11 10 Bulgaria 17 19,3 Croatia 18,3 22,4 Czech Republic 26,7 28,7 Estonia 22,7 25,7 Georgia 12 12 Hungary 25,3 27,7 Kazakhstan 13 15 Kyrgyzstan 15 21,3 Latvia 18 23 Lituania 21 23 Macedonia 15,6 19,3 Moldova 14,7 17,7 Poland 26 27 Romania 16 19,3 Russia 19 21,1 Serbia-Montenegro 14,3 12,3 Slovakia 24,7 26,4 Slovenia 23,7 25,1 Tajikistan 12,7 12,7 Turkmenistan 9 9,7 Ukraine 11,7 12,7 Uzbekistan 12 17
1995 1996 1997 38,7 38,7 38,7 20,3 21,7 21,7 18,4 20,7 21,3 14,7 16 18,3 17,4 15,7 14 10 11,3 13,3 20,3 20,7 24,3 23,4 25,4 26 28,7 29,6 30 27 27,6 28,9 17 21 23,3 30 30,3 32,2 19,4 23 24 23 24 24 24 25,3 26,9 24,7 25,7 26,3 20,3 20,7 20,7 22 22,3 23 28,7 29,3 30,2 20,7 20,7 23,7 22,7 25 25,9 12,3 12,3 12,7 27,1 27,7 27,7 25,8 26,7 27,3 14,7 15 15,6 10,7 10,7 12,7 19 20 21,2 19,4 19,4 18,8
Sources: EBRD Transition Report 2002 & author’s calculation.
28
1998 38,7 21,7 22,6 20 13 17,3 24,3 26,3 30,3 29,6 23,6 32,8 24,3 23,6 26,7 26,6 22,4 23,6 30,8 24,3 22,3 12,4 28 28,3 17,7 12,4 20,9 18,4
1999 38,7 21,7 22,9 20 13 17,3 25,2 26,9 30,6 30,7 23,9 33,2 23,3 23,6 27 27,3 22,4 23,6 30,8 25,1 21,7 12,4 28,6 28,9 18 12,4 21,2 17,7
2000 38,7 23 22,9 20,7 13,7 18,3 27 27,5 30,9 31,3 25,2 33,6 23,9 23,6 27,6 28,4 23,7 23,9 31,5 25,7 22,3 12,7 28,9 29,3 18,6 11,7 22,3 17,4
2001 38,7 23,3 24,3 21,3 14,3 19 27 27,9 31,6 31,6 25,2 33,6 24,3 23,6 27,9 29,3 24,3 24,2 31,8 26,3 23 16 29,5 29,3 19,1 11 22,6 18,5
