Version classiqueVersion mobile
OpenEdition Books

L’ouverture de l’Europe vers l’est

 | 
Gabriel Poulalion

Chapitre IX. Institutions and Trade in Transition économies1

Oxana Koukhartchouk et Mathilde Maurel

Texte intégral

INTRODUCTION

  • 1 This research was supported by the RECEP (European TACIS Programme).

1The break-up of the Former Soviet Union, a highly integrated economy of 15 regions, into 15 independent states, has reduced mutual trade flows between these components. Nevertheless, everyone agrees that 10 years after this break-up took place, intra-FSU trade still remains at a very high level. Our purpose is to explore the reasons why it is so. On one hand, the experience of the Austro-Hungarian empire which broke-up in 1919 shows that regionalism leaves its mark during a long time over newly created independent states (De Menil, Maurel, 1994 and Maurel 1998). On the other hand, the heritage of 70 years of strong state intervention still stamped in post-communist institutions may as well explain a part of the « excess » intra-FSU or intra-CIS trade (Duchene 1994).

  • 2 See Jeffrey Frankel, Ernesto Stein and Shang-Jin Wei, (1997).
  • 3 The implied norm will be described in a subsequent section.

2Considering the CIS as a group of countries sharing common histories, cultures, and protectionist behaviours, we propose to consider this region as a customs union and to treat it as such, with the specific institution problem mentioned above. Our objective is thus twofold. We want to estimate the degree of trade creation within CIS and that of trade diversion ; Our second focus will be the impact of institutions in a broad sense on trade integration. For that purpose, we use the gravity equation, which has been extensively used in the literature on regionalism versus globalisation, for instance to deal with the impact of EU, ASEAN, NAFTA, on trade integration and trade blocs2 ; similarly, we will use the gravity framework to answer the question of the impact of being member of the CIS on bilateral trade flows, and the long run potential for trade increase if CIS were to be member of the EU. Furthermore, we augment the gravity equation with institutional variables, like the extent to which property rights are protected in partner countries, the share of the black market in the whole economy, or more directly the importance of tariff, non-tariff barriers, and corruption in the custom service. Gravity estimâtes are subsequently used for computing the extra trade implied in the short run by converging towards the EU institutional standards3.

  • 4 Institutional variables include variables allowing to evaluate the degree of economic freedom, assu (...)

3The structure of this paper is as follows : the first section describes the background and related literature. In the second part, we present the specification of the gravity equation. This latter is augmented by the variable « institution »4, whose expected impact on trade integration is documented ; the third section presents the results and simulation exercises. Conclusion draws the main policy implications.

I. BACKGROUND AND RELATED LITERATURE

  • 5 See Frankel and Romer (1999), who documented the impact of trade on growth in the very long run.
  • 6 See Acemoglu, Johnson and Robinson (2001) and Hall and Jones (1999).
  • 7 See Engermann and Sokoloff (1997).

4An important literature is concerned with the interrelationships between good institutions, trade and growth. There are actually two strands : one strand of the literature put the stress on the impact of trade openness on convergence and growth5 ; the other sheds light on the importance of good institutions in driving the pattern of growth. Institutions that protect property rights for instance are shown to matter for growth in the long run. Some of them are determined by historical investment in institution-building by colonial powers6, or by natural resources endowments7. In a recent World Bank paper, David Dollar and Aart Kraay (2002) attempt to isolate the partial effects of trade and institutions in the long run. The paper rests on estimating two growth equations. The first is based upon cross section analysis, and fails in disentangling the respective impact of trade and institutions because of multicollinearity problems : good institutions are likely to be associated with trade openness. The second exploits the dynamics of trade, institutions and growth over three decades, from the 70ies up to now. One striking fact is that the three variables are far from inertial, giving some a priori relevance to the dynamic analysis. The latter shows that a substantial effect can be attributed to changes in trade shares for predicting changes in growth rates, while the effect of institutions on growth cannot be shown at least on the equation in first-differences (that is in the short run).

5For taking into account the fact that while institutions influence growth, the reverse causality may exist as well, David Dollar and Aart Kraay (2002) instrument their explanatory variables by trade and institutions. For instance Table 6 reports estimates of the regression of twice lagged GDP per capita in logarithm and institutional variables like the Freedom House indicator, the ICRG Rule of Law index, etc., on the GDP per capita in logarithm. Like Rodrik (2000) has pointed out, increases in trade are associated significantly with improvement in institutional quality, the latter making countries more attractive for their trading partners.

  • 8 Our empirical analysis covers the period from 1994 to 2001.

6In this paper we focus on the causality running from institutions to trade intensity, and disregard the impact on growth as well as the very long run8. The direct impact of institutions on trade has been analysed in James E. Anderson and Leslie Young (2000), who show using a model of trade under two regimes, the rule of law (where institutions are present) and anarchy (institutions are absent), that the former is favourable to trade enforcement while the latter undermines the incentives to make an exchange. In James E. Anderson and Leslie Young (2002), the non-linear empirical relationships between the degree of contract enforcement and that of trade integration is rationalized in a model of imperfect contract enforcement. Finally Harry G. Broadman and Francesca Renati (2001) emphasise that economic development and maturation of democratic processes, but also the degree of trade openness, may explain the increase in corruption. While the latter is recognized as been able to seriously jeopardize the reforms, a well-established system of market institutions favours economic growth, supports the reforms towards the market economy, and reduces the incentives for corruption. In our analysis, we consider that corruption ranges in the set of explanatory variables and stress the causality running from corruption to trade.

7Our analysis aims at assessing both the impact of institutions on the intensity of bilateral trade flows, and the medium run perspective through dummy variables set equal to one when a given country is member of a trading bloc, the CIS (Commonwealth of independent States) or EU (European Union) for instance. The rationale behind such a strategy is that the transition from the Plan to the Market in CIS and CEECs (Central Eastern European Countries) entails big changes over a short span of time, in institutions and blocs geography from one hand, and trade integration from the other hand, allowing to measure the impact of the latter on the former ; According to the literature, this increase in trade openness can be expected to influence rapidly the rate of growth. The medium run is roughly understood as a period when some CIS will be members of the EU, which includes many things : once CIS will have converged in terms of GDP per head, once their patterns of specialisation will have narrowed that of the EU countries, allowing them to develop the basis for a much more intensive intra-industry trade, or once the inheritance of the former CMEA will have disappeared. Those assumptions are assessed using dummy variables estimates in a gravity equation.

II. GRAVITY EQUATION SPECIFICATION

8Our empirical strategy is based upon the well-known gravity equation. One can measure the impact of institutional variables through qualitative variables and running cross section procedures, but this is likely to provide with biased estimates, by not taking into account non-observable variables, like cultural, linguistic, and historical ties. We rely in this paper upon a rich panel characterised by a very large fluctuation of institutions across countries and over the nineties. Therefore, the focus can be put on changes from protectionism and a limited amount of economic freedom to an increasing level of markedly determined behaviours, which is expected to translate into more trade integration. In other words, the panel structure of the data-set eliminates any bias occurring from omitted (time-invariant) variables and which could « pollute » the variables of interest.

9According to the gravity equation, bilateral trade depends upon a variety of « structural » factors, among which are both nations’ market size, measured by their national products. The gravity equation provides a natural benchmark to which a number of other explanatory variables can in turn be added. In this paper two sets of additional variables are taken into account. One set is made of institution variables9, whose effect on trade we seek to measure. The second set is made of dummy variables, set equal to one when countries are members of a given bloc. In practice the following basic specification is used :

10where LnTradeij denotes the bilateral trade, that is the natural logarithm of exports of country i from country j in thousands dollars. Bilateral trade flows, in thousands constant dollars, come from the Chelem-CEPII dataset. Missing values are filled in using IMF-DOTS and Goskomstat Rossijskoj Federatzii. LnTradeij depends upon :

11The market size of country i (respectively j) is measured as the log of real GDP in PPP. Data come from Chelem-CEPII and IMF World Economic Outlook (for Slovenia).

12Transportation costs are proxied by the log of distance (LnDIST) between the capitals, considered as representing the economic centres. For taking into account possible bias for countries like the United States, Canada and Russia, which are big countries and as such are likely to have multiple economic cores, we considered distances from several economic centres. The data for distances were collected from CEPII database.

  • 10 See, for example, Bruny, Carrère, Guillaumont, Melo (2002), Djankov, Freund (2000), Bougheas Demetr (...)
  • 11 The World Bank provides a database, The World Development Indicators, which ends up in 1999. Our da (...)

13Distance entails not only transportation costs, but also infrastructure, which has a big impact on transportation costs10. Unfortunately we cannot include infrastructure because it would imply a large reduction of our sample due to missing values for selected countries. For instance, the widely used Database of World Infrastructure Stocks by David Canning11 provide data from 1950 up to 1995 which for transition countries consider the former territories and former political blocs, like the Soviet Union. By putting the stress on those countries whose geography and political borders have changed dramatically we cannot include a variable for which data have not been updated. Finally even if infrastructure have a great impact on trade, we can consider that over a short period of time they do not vary too much and that they can be considered as fixed effects.

  • 12 According to the following formula : VOL = σ [(eij –eavg)/ eavg], where σ is the standard deviation
  • 13 See IMF [1984] estimation of 42 gravity equations, where exchange rate volatility has a positive in (...)

14VOL, the bilateral exchange rate volatility, calculated as the standard deviation of the ratio of the monthly exchange rate over its yearly average12. Data come from the Pacific Exchange Rate Service (for the EU, China, Kazakhstan), from the Central Bank of Russia for Kazakhstan over the period 1994-1997, and from the IMF for other countries in the sample. Exchange rate volatility is not necessarily correlated negatively with trade flows : it can be seen either as an additional transaction cost or as a way of absorbing asymmetric shocks, and the sign of the correlation is a matter of empirics13.

  • 14 For more details see Annual Report of Economic Freedom of the World.

15To measure the degree of State intervention in the economy and the role of institutions we use the Index of Economic Freedom. This index is composed of « the factors that most influence the institutional setting of economic growth »14. It was computed for measuring economic freedom around the world and for emphasising the empirical strong correlation between freedom and growth. Here we are interested in the causality running from institutions to trade.

  • 15 Trade policy, Size of government, Monetary policy and Price stability, Capital flows and Foreign In (...)

16INST_i (Institutional score in country i) is an index varying from 1, which corresponds to the lowest level of State Intervention, to 5, which denoted the lowest level of freedom. The ten variables identify the consistency of institutional arrangements and policies with economic freedom in ten major areas15. The variables are therefore as follows :

17INST1 = Trade Policy, an indicator measuring tariff and non-tariff barriers, obviously reduces trade ;

18INST2 = Fiscal Burden of Government and

19INST3 = Government intervention in the economy : here the link with trade is not so evident, but we believe that market incentives for trading are much lower when the State plays a predominant role in the economy ;

20INST4 = Monetary Policy measured through the inflation rate: like volatility, inflation is expected to be negatively correlated with trade ;

21INST5 = Capital Flows and Foreign Investment : measuring restrictions on foreign business. FDI might play an important role in transition countries insertion into the world economy : this is the case in Hungary, where foreign companies invest in the country to serve the domestic market or to benefit from low input costs (wage costs) and re-export towards Home country. In both case, legislation on FDI is likely to have an impact on bilateral trade flows intensity ;

22INST6 = Banking and Finance : all restrictions on credit and finance obviously tend to hinder foreign trade.

23INST7 = Wages and Prices : more freedom from government influence and higher labour market flexibility favour absolute advantage based on low wage costs. What regards regulation of prices, it is clear that imposing for instance low prices on energy must be accompanied by strict measures of trade controls.

24INST8 = Property rights : we expect violation of creditor rights, poor enforcement of international law, unfavourable climate for business activities to affect both domestic and foreign activity ;

25INST9 = Regulation and Corruption within the bureaucracy, licensing requirements to operate a business are likely to discourage potential trade partners ;

  • 16 The bias however works in both directions : imports are over-invoiced, but export are under-invoice (...)

26INST10 = Black Market. To the extent that trade statistics if trade registered officially, an high level of unofficial trade not captured through official agencies may well affect negatively trade flows. Moreover, capital flights, import duty fraud, income tax evasion, or money laundering, more or less linked to the phenomena of illegal and black activities, can lead to abnormal pricing in international trade16.

  • 17 X (respectively Y) can be EU, CEECs, CIS, and World, which is the rest of the world.

27Finally, bloc dummies are used to identify the various regimes involved. EU captures the EU trade bias, CEEC the bias in trade within CEECs, CIS the CIS bias ; X-Y is set equal to one when country i is member of the bloc X and j member of the bloc Y17. It reflects the extent to which a trade diversion effect may possibly occur.

28We dropped the usual « population » variable. Usually, population appears with a negative sign which is interpreted as the influence of the development level (the wealth of nations) on trade, or is included as a part of an « economic distance » argument ; in the first case, if the development level is measured by the GDP per capita, then the population variable will be negatively related to trade ; in the second case, countries with very different GDP per capita are considered « economically distant » and will trade less. Without entering the details, we consider that both arguments do not reflect well established theoretical results, and that dropping the population variable is more convenient to reproduce Krugman-Helpman (1985) theoretical framework.

29The sample consists of 14 European countries (Austria, Belgium and Luxembourg, Denmark, France, Germany, Greece, Italy, Netherlands, Portugal, Spain, Sweden, United-Kingdom, Finland, Ireland) plus the CEECs (Bulgaria, Hungary, Poland, The Czech Republic, Romania, Slovakia, and Slovenia), the three Baltic States (Estonia, Latvia, Lithuania), Russia, Kazakhstan, Ukraine, Belarus, which constitute the CIS, and the Rest of the World (Brazil, Canada, China, South Korea, Egypt, The United States, The States of the Golf, India, Israel, Japan, Norway, Switzerland, Thailand, Turkey). The time span covers the period from 1994 to 2001. The large size of the panel (13 712 observations, despite some limitations on unobservable bilateral flows) provides a basis for consistent estimation and the inclusion of a large number of explanatory variables. Fixed effects can be included in the equation, which as a result does not necessarily include the traditional distance variable used as a proxy for transaction cost, and excludes as well all time-invariant variables (geography in its broad sense, including distance, adjacency, linguistic and cultural ties, and dummy variables as well). Getting consistent estimates of the time invariant variables requires the use of the Hausman-Taylor procedure, which will be presented below. Annex 2 summarises OLS, GLS, Within and Hausman-Taylor results, which are commented subsequently.

III. TRADE AND INSTITUTIONS : OLS AND GLS RESULTS

30As can be seen in Table 1, which contains OLS and GLS results, GDP coefficients are between.89 and.87, both significant at 1 % level and positive : as expected, trade depends upon supply and demand forces. Distance between economic centres of both countries, which is a proxy for transaction costs, affects negatively trade : more distant countries trade less. For what regards bilateral volatility, the coefficient is negative and significant, which means that higher transaction costs have a negative impact on trade. Bilateral volatility has thus the same effect as a larger distance between two partners.

Table 1 - OLS and GLS Results

Table 1 - OLS and GLS Results

Standard Error in paratheses.

*, **, *** define 1 %, 5 % and 10 % significance level respectively

Source : See Annex 2

31Overall, institutions influence trade in the expected way : more freedom attracts foreign partners and stimulates bilateral trade flows. Higher tariff and non-tariffs barriers constitute an impediment to mutual trade ; a legislation which favors activity, by limiting the number of restrictions on foreign ownerships of business and land, on the repatriation of earnings, by treating equally foreign and domestic companies, goes in hand with larger trade volumes. The existence of a well-functioning financial and banking system, able to finance efficiently trade business, affects positively and significantly the latter. The higher wage and price flexibility, the higher bilateral trade. The stronger the degree of protection of property rights, the higher the incentive to engage in bilateral trade activities. Regulation is not strongly correlated with the endogenous trade variable. Finally the extent of the Black Market influences trade in the expected way : more informal activity, not registered, influences negatively the level of official, registered trade.

Table 1 (followed) - Institutional variables estimates ; Dependent variable : LnTRADEij

Table 1 (followed) - Institutional variables estimates ; Dependent variable : LnTRADEij

Source : See Annex 2

32Four variables coefficients have the wrong signs or are not significant. Two reasons can be mentioned. First, it is well known that the size of the government is significantly bigger in richer countries, which are also more likely to be more largely involved in international trade. As a consequence, Government Intervention, and Regulation scores, could be larger (or not very different) in EU countries, lower in transition countries, and as a result they may affect positively bilateral trade flows. This assumption is partially confirmed in Table 2 :

Table 2 - Institution scores averaged over countries blocs

Table 2 - Institution scores averaged over countries blocs

33The second reason lies in the mutlicollinearity between some variables and other institutions, which capture already the extent of economic freedom in a given country. This is the case for Monetary Policy, which according to Annex 3 is highly correlated with Property Rights and Black Market.

34The last set of estimates regards dummy coefficients estimates.

Table 1 (followed) - dummy estimates

Table 1 (followed) - dummy estimates

SOURCE : SEE ANNEX 2

Table 3 - Effective Trade in percent of Potential trade

Table 3 - Effective Trade in percent of Potential trade

Source : From table 1, Authors Computation

  • 18 Potential trade : the trade level above what it would be if it was determined solely by the gravity (...)

35An EU estimate of 0.22 implies that trade within EU is 125 % above its potential level18. This extra trade is usually called trade creation effect, which in the case of EU does not occur at the expense of the trade with the Rest of the World. From the dummy World-EU indeed, we can infer that the counterpart of the trade creation effect, labelled trade diverting effect, is not significant, or that the potential of trade is not significantly different from the effective trade (since this dummy is not significant, its not reported in the table).

36Contrary to what happens in the EU, trade within CEECs is higher than the level, which would imply demand, supply forces, and geography, but at the expense of trade of CEECs with EU and with other countries.

37Finally, trade within the CIS is between 9 and 14 times what it would be if it were « normal ». With the rest of the world, it is lower than expected. The latter figures are simply unbelievable, and suggest that the procedure is too simplistic for delivering results reliable enough. CEEC-CIS trade was non significant and not present here.

38Towards more reliable estimates of the bloc effect : Panel Data procedures.

  • 19 See Graziani (1981).

39For explaining the huge impact of being member of the CIS on trade, one can resort to at least three explanations. The first explanation emphasizes the aggregation bias. Nitsch (2002) in a paper whose primary focus is the effect of currency unions on trade, argues that the latter must be disentangled : unilateral currency unions do not behave the same way as multilateral currency unions. Melitz (2001) emphasises that the political dimension of a currency union adds something, which must be taken into account ; Otherwise, putting everything together may produce biased estimates of the pure currency union effect. The same line of reasoning can be found in Klein (2002), who distinguishes countries, which dollarized their economy, from other countries in the sample. For what regards the issue of CIS trade, by summing up bilateral trade flows between Russia and other CIS from one hand, and between CIS excluding Russia form the other hand, one hides two facts : first, effective trade within CIS is highly depressed, hence 40 times higher implies an absolute value which is not necessarily very high; second, bilateral trade flows may be well above their normal level for what regards trade with the Russian core, an inheritance of the dependency structures in former CMEA19, while peripheral or radial trade between Russia and Belarus would not significantly differ from its normal level.

40Persson (2001) stresses that if there is a non-linear relationships between a bloc dummy variable and other variables of the gravity equation, then the dummy coefficient may capture the impact of this non linear effect, and may be over-estimated. In order to avoid any such bias, he proposes to introduce a control group sharing the same characteristics than the countries, which belong to a currency union. This control group is determined by a propensity score, which is the probability of being selected in the bloc group, as a function of the explanatory variables. Comparing country-pairs within and outside the control group allows estimating the effect of the bloc on trade flows.

  • 20 This explanation is latent already in the aggregation argument, which recommends splitting the vari (...)

41The last explanation rests on the possibility that unobservable variables are omitted20. For addressing the bias induced by the correlation between omitted and bloc dummy variables, one is faced with the following three options : either estimate a system of simultaneous equations, where explanatory variables correlated with the residual are endogenously determined in a separate equation (i), or instrument the explanatory variables responsible for the problem of endogeneity (ii), or introduce all possible (but time-invariant) omitted variables through dummy variables set equal to one when the pair (i,j) is considered (iii). The first option is illustrated in Tenreyro

  • 21 She estimates simultaneously a selection model for the formation of a currency union and a gravity (...)

42(2001)21, or Flandreau and Maurel (2001) ; The second option can be found in Flandreau and Maurel (2001) and Maurel (2002) ; the last and third option, which is equivalent to introduce fixed effects, is illustrated in Pakko and Wall (2001), Rose and Van Wincoop (2001), and Maurel (2002).

43In what follows we are going to exploit both the cross-section and dynamic dimensions of our panel data set. But the dummy variables we are interested in are time-invariant, which means that they do vanish when we apply the within operator. The random effect model, more efficient but still faced with the problem of endogeneity, is rejected by the Hausman test. Fortunately, Hausman and Taylor (1981) proposed a procedure, which basically consists in regressing all time-invariant variables on time-variant exogenous variables, then in using the new variables in a model with fixed effects. In other words, they proposed a procedure, which combines the advantages of taking into account the fixed effects and keeping in the equation the time-invariant variables whose impact on trade we want to estimate. This procedure is used in Julie Lochard (2002) for assessing the effect of a currency union on bilateral trade flows.

IV. THE HAUSMAN TAYLOR METHOD

  • 22 Notice that within estimates allow assessing the impact of institutions on growth, since institutio (...)

44We are thus concerned with the potential correlation of the fixed effects with the explanatory variables. The traditional technique to overcome this problem is to eliminate the individual effects in the sample by transforming the data into deviations from their individual means. Unfortunately, the OLS coefficient estimates, known as within or fixed-effects estimators, have two important defects : all time-invariant variables are eliminated by the transformation, so that we cannot estimate the bloc dummy we are interest in22 ; Under certain circumstances, the within-group estimator is not fully efficient since it ignores variation across individuals in the sample. Hausman and Taylor (1981) basically show that if the fixed effects are not correlated with a subset of explanatory variables, then the time-invariant variables coefficients can be consistently and efficiently estimated.

45Assume that our the gravity model can be written in the following way :

46Yit = Xitβ + Ziγ + εit i=1,...N; t=1...T; (1)

47Where : εit =ut +ui +eit

  • 23 Reported in Annex 2.

48The Hausman test23 rejects the null hypothesis of the absence of correlation between the residuals and certain variables among the X and Z : X2, correlated with the specific effect and varying over time, and Z2, correlated with the specific effect and not varying over time. In other words :

49E(ui/X2,Z2 )≠0 but E(ui/X1,Z1 )=0

This presentation is taken from Julie Lochard (2002) ; the number of variables is indicated in brackets ;

  • 24 P denotes the operator, which transforms the variables into their individual means; Q denotes the o (...)

50From what precedes, it follows that X2 and Z2 must be instrumented. The list of instruments includes the deviations from the group means of X1 and X2, and the means of X1 and Z1: [QX1, QX2, PθX1, θZ1]24.

51One identification condition is that k1 must be higher than g2 ; in other words the number of time invariant variables to be regressed on time-variant and exogeneous variables must be lower than the number of the latter, which are used as instruments.

52The algebra for transforming (and instrumenting) the model is the following :

53Ω-1/2Yit = Ω-1/2Xit β + Ω-1/2Ziγ+Ω-1/2 ε it (2)

54Where:

55Ω-1/2 =θP+Q=Itn-(1-θ)P;

56T is the number of periods for each country-pairs, a2 and are the variance of the time varying error component and of the specific effect.

57Results

  • 25 Because they are not significant or have the wrong sign, we dropped the following variables: Fiscal (...)

58The first step consists is determining to what category belong each variable. We apply the Hausman-Taylor procedure according to the following ranking of the variables25 :

  • 26 See David Dollar and Aart Kraay (2002) for instance.

59We therefore consider than certain institutional variables are (more likely to be) correlated with the residuals and omitted fixed effects. It may be because increases in trade openness lead to a demand for and realization of better institutions, that is more open Trade Policy, higher enforcement of Property Rights, and lower share of Black Market. Indeed the probability for Trade Policy to be endogenously determined by the degree of trade openness, that is by the country’s dependency to foreign markets and the country’s sensitivity to changes in its own relative competitiveness, is higher than for other institutions, more exclusively concerned with internal policy. For Property Rights there are also good reasons to be concerned about the problem of reverse causation : Property Rights include the commercial code defining contracts, the sanctioning of foreign arbitration of contract disputes, and more generally the protection of property rights, whatever foreign or domestic. Similarly, the share of the Black Market in the whole economy may well be influenced by the frequency and intensity of foreign trade flows. Market size is often considered in the literature as being determined by external trade, which is one of the GDP components indeed ; the same line of reasoning can be found in the growth literature26.

  • 27 Indeed g2 = 0.

60Given that the identification condition is trivial27, we could easily run sensitivity analysis to the number of institution variables included in each category. For instance we included VOL, bilateral exchange rate volatility, in the category of variables correlated with the specific effects. The final decision rested upon a test comparing within and Hausman-Taylor estimates and requiring that the latter should not differ too much from the former.

61Table 4 below reports within and Hausman-Taylor results, which do not differ significantly from each other. Market size elasticity is lower than in Table 1 for country j and about 7-10 % percent higher for country i, while the distance coefficient diminished, from -1.09 (1.06) to 1.05 (0.94). This latter figure is more similar to the figure usually reported in the gravity literature. The VOL coefficient is remarkably stable across the different procedures.

Table 4 - Within and Hausman-Taylor Results

Table 4 - Within and Hausman-Taylor Results

Source : See Annex 2

62Institutions matter again. Our preferred estimates are that of HTIV II (last column), where we dropped institutions whose coefficients had either the wrong sign or were not significant, for the reasons mentioned above. Compared to OLS or GLS estimates, new elasticities are generally lower, but still significant. They will be used for computing the impact of a change in institutions, towards the EU average for instance, on bilateral trade.

Table 4 (followed) - Institutional variables estimates ; Dependent variable : Ln TRADEij

Table 4 (followed) - Institutional variables estimates ; Dependent variable : Ln TRADEij

Source : See Annex 2

63For what regards dummy coefficients as compared to former estimates, everything is different. The only exception is the EU and EU-CEEC dummy coefficient : trade within EU is as previously higher than what it would be without the existence of the European privileged trade area, and the order of magnitude did not changed significantly, the EU-CEEC trade is under its « normal » level and is stable too.

Table 4 (followed) - dummy estimates

Table 4 (followed) - dummy estimates

Source : See Annex 2

  • 28 The ratio of potential trade on effective trade is set equal to the exponential of the dummy bloc v (...)

64In Table 5 we computed effective trade in percent of potential trade28. Several striking facts are worth mentioning : trade within CIS is about six times what it should be, while CIS trade with other countries is very much depressed : with the rest of the world, trade flows are only at 16-31 % of their normal level. But with CEECs, that is former CMEA countries, flows do not differ from their expected level. This dual result points to a well-known phenomena characterising the former FSU (Former Soviet Union) trade, called by Pelzman (1977) the trade destruction effect, which exhibits some persistence today : trade within FSU is higher than what it should be but at the expense of very high tariff and non tariff barriers against non former CMEA countries. Those results are much more reliable than those found in the earlier section.

Table 5 - Effective Trade in percent of Potential trade

Table 5 - Effective Trade in percent of Potential trade

Source : Table 4 and Authors Computation

65According to table 5, there is a big potential for CEECs trade increases within CEECs, with EU countries, and with other countries in the world. This can be explained by the fact that we took GDP in Purchasing Power Parity, which are significantly larger than official estimates based upon markedly determined exchange rates, and which proxy the market size long run potential of those countries. As a result the trade potential is higher, and the ratio of effective to potential trade lower than unity.

  • 29 Indeed, 1/exp(-1.18)=3,25.

66According to table 5, trade within CIS should decrease by six times, while trade between CIS and EU should increase by about three times29.

67What would be trade if transition institutions were converging towards EU ones ?

68Table 2 reports institutional variables scores averaged across three sub-samples of countries. Those scores, as expected, are generally higher for transition countries, where the institutional quality is also lower. It must be recognised however that at the eve of EU enlargement, a very efficient process of convergence of CEECs institutions towards EU ones has taken place. Except Black Market, Property Rights, and Monetary Policy, where CEECs institutional scores are higher by respectively 54 % 48 %, and 67 % as compared to EU scores, the institutional differences are not huge, contrary to what happens we one compares CIS with EU institutions (see table 6).

Table 6 - How far are institutional scores in transition countries higher than in EU countries ?

Table 6 - How far are institutional scores in transition countries higher than in EU countries ?

Source : See Table 2 ; Authors Computations

69The room for institutional convergence towards better institutional quality implies a potential for trade increase, which can be assessed by comparing the effective trade with what the trade would be if institutions in CEECs (k or k’ = CEECs) or CIS (k or k’ = CIS) were at the EU level (k or k’ = EU). Assume for instance that the score for Trade Policy is set equal to 2 instead of 3,5, in CIS countries. In as much as Trade Policy has a negative impact on bilateral trade flows, this decrease from 3,5 to 2 can be expected to induce an increase in trade.

70The algebra for calculating the increase in trade following an institutional improvement is the following :

71

72Tables 7a and 7b below report the potential for trade increase implied by the changes in institutional variables :

Table 7a - Impact of institutional improvement on trade

Table 7a - Impact of institutional improvement on trade
  • a Trade increases calculus is based on HTIV II gravity’s estimates of Table 4 ;
  • b « X — Y » means that institutions in X are set equal to institutions in Y ;

Note aa
Note bb

  • 30 Guriev, Makarov and Maurel (2002) show at the firms level that the violation of property rights and (...)

73The move from CIS institutions towards EU institutions implies a 122 % increase in trade. Much of this percentage (50.9 %) is due to the decline in the Black Market, followed by the improvement in the protection of Property Rights30 (23,5 %). According to table 7b, Russian and Ukrainian Institutions being slightly better than Institutions in the neighbouring countries, the potentials for trade increases due to institutional improvements turn out to be lower : 103 % in Russia and Ukraine against 144 % in both Belarus and Kazakhstan.

Table 7b - Impact of institutional improvement on trade by CIS country

Table 7b - Impact of institutional improvement on trade by CIS country
  • c Trade increases calculus is based on HTIV gravity’s estimates of Table 4 ;
  • d « X — Y » means that institutions in X are set equal to institutions in Y ;

Note ac
Note bd

CONCLUSION

74The objective of this paper was to estimate two sources of potential for trade increases between CIS and EU countries : the first induced by the progressive decline in the trade diversion effect, and the second due to the institutional improvement in CIS. For that purpose, we used the gravity framework allowing to answer the question of the impact of being member of the CIS on bilateral trade flows. The result is that trade within CIS can be expected to decrease by six times, while at the same time trade between Russia and EU countries can be expected to increase by three times.

75Then, we augmented the gravity equation by adding institutional variables, like the extent to which property rights are protected in partner countries, the share of the black market in the whole economy, or more directly the importance of tariff, non-tariff barriers, and corruption in the custom service. Gravity estimates were subsequently used for computing the extra trade implied in the short run by converging towards the EU institutional standards: this extra trade reaches 122 % of the actual level of trade. Much of this percentage (51 %) is due to the decline in the Black Market, followed by the improvement in the protection of Property Rights (23 %).

76Those estimates are obtained by applying the procedure introduced by Hausman-Taylor (1981), which allows simultaneously estimating time invariant variables coefficients and taking into account fixed effects. Those results rely upon a very rich panel dataset, which contains 42 countries, over 8 years, for a total of about 13 712 trade flows.

Bibliographie

REFERENCES

Acemoglu, Daron, Johnson, Simon, Robinson, James A., 2000, « The Colonial Origins of Comparative Development : An Empirical Investigation », NBER Working Paper 7771, Cambridge, MA.

Alesina, Alberto, Barro, Robert, « Currency Unions », Quarterly Journal of Economics, 117, pp. 409-36.

Anderson J.E., Young L., 2000, « Trade implies law : The power of the weak », NBER Working Paper 8847, Cambridge, MA.

Anderson J.E., Young L., 2002, « Imperfect contract enforcement », NBER Working Paper 8847, Cambridge, MA.

Bougheas S., Panicos O.D., Morgenroth E. L.W., 1999, « Infrastructure, transport costs and trade », Journal of International Economics, 47, pp. 169-189

Brada, J.C, Mendez J.A., 1988, « Exchange Rate Risk, Exchange Rate Regime and the Volume of International Trade », Kyklos, n° 41, pp. 263-280.

Broadman H.G., Recanatini F., 2001, « Seeds of corruption - Do market institutions matter ? », MOCT-MOST, 11, pp. 359-392.

Bruny, J.-F., Céline Carrère, C., Guillaumont, P., Jaime de Melo, 2002, « Has Distance Died ? Evidence from a panel Gravity Model », CERDI, Working Paper, June 2002.

De Boyrie, Maria E., Pak, Simon J., Zdanovicz, John S., 2002, « Detecting abnormal pricing in international trade: the Russia-USA case », paper presented at the Third International Conference on Public Economics, July, 4th-6th.

De Ménil, Georges, Maurel, Mathilde, 1994, « Breaking up a custom Union : The case of the Austro-Hungarian Empire in 1919 », Weltwirtschaftliches Archiv, Band 130, Heft 3 : 553-575.

Dollar, David, Kraay, Aart, 2002, « Institutions, Trade and Growth », The World Bank, mimeo.

Djankov S., Freund C., 2000, « Disintegration and Trade Flows : Evidence from the Former Soviet Union », Policy Research Workings Paper, 2378.

Duchêne, Gérard, 1994, « Intégration ou Désintégration économique dans l’ex-URSS », Revue

Economique, 45(3), pp. 575-88.

Engermann, Sokoloff, 1997,

Flandreau, Marc, Maurel, Mathilde, 2001, « Monetary Union, Trade Integration, and Business Cycles in 19th Century Europe : Just Do It », CEPR Discussion Paper 3087.

Frankel, Jeffrey, Stein, Ernesto, Wei, Shang-Jin, 1997 Regional Trading Blocs in the World Economic System, Institute for International Economics, Washington, DC.

Frankel, Jeffrey A., Romer, David, 1999, « Does Trade Cause Growth ? », The American Economic Review, (June), pp. 379-399.

Graziani, Giovanni, 1981, « Dependency Structures in COMECON », Review of Radical Political Economics, 13(1), pp. 67-75.

Guriev, Sergei, Makarov, Igor, Maurel, Mathilde, 2002, « Debt Overhang and Barter in Russia », forthcoming in the Journal of Comparative Economics.

Hausman, Jerry, Taylor, William, 1981, « Panel data and unobservable individual effects », Econometrica, 49 (6), pp. 1377-1398.

Helpman, Elhanan, Krugman, Paul R., 1989, Trade policy and market structure, Cambridge, Mass. and London : MIT Press, 1989, p. 191.

IMF [1984], « Exchange Rate Volatility and World Trade », International Monetary Fund

Occasional Papers, 28/7/84.

Klein, Michael, 2002, « Currency Unions and Trade : A Comment », mimeo, Tufts University.

Lochard, Julie, 2002, « Monetary Unions and Trade : Towards a More Reliable Method of Estimation », mimeo, ROSES and Université de Paris-1.

Maurel, Mathilde, 1998, Régionalisme et désintégration en Europe centrale et orientale : Une approche gravitationnelle, Editions du CNRS.

Maurel, Mathilde, 2002, « On the way of EMU enlargement towards CEECs : what is the appropriate exchange rate regime », CEPR Discussion Paper 3409.

Mélitz, Jacques, 2001, « Geography, Trade and Currency Union », CEPR Discussion Paper N° 2987.

Nitsch, Volcker, 2002, « Comparing Apples and Oranges : The Trade Effect of Multilateral

Currency Unions is Small », Mimeo, Bankgesellschaft Berlin.

Pakko, Michael, Wall, Howard, 2001, « Reconsidering the Trade-Creating Effects of a Currency Union », Federal Reserve Bank of St. Louis, 83, pp. 37-45.

Pelzman, J., 1977, « Trade Creation and Trade Diversion in the Council of Mutual Economic Assisstance : 1954-1970 », The American Economic Review, 67(4).

Persson, Torsten, 2001, « Currency Union and Trade: How large is the Treatment Effect ? », Economic Policy, 33, pp. 435-448.

Rose, Andrew, Van Wincoop, Eric, 2001, « National Money as a Barrier to Trade : The Real Case for Monetary Union », American Economic Review, 91, pp. 386-90.

Tenreyro, Sylviana, 2001, « On the Causes and Consequences of Currency Unions », mimeo, Harvard University.

Rodrik, Dani, 2000, « Trade Policy Reform as Institutional Reform », Manuscript, Harvard University.

Annexes

Annex 1 : Economic Freedom of the World

Annex 1 : Economic Freedom of the World

Annex 2: Econometric Results

Number of obs. = 13 712
Number of groups = 1 714
Obs. per group = 8
42 Countries
8 Years : 1994 – 2001, Balanced Panel

Note ******
Note ****
Note **
(Standard error in parentheses)

Annex 2 (followed) : Econometric Results

Annex 2 (followed) : Econometric Results

Note ******
Note ****
Note **
(Standard error in parentheses)

Annex 2 (followed) : Econometric Results

Annex 2 (followed) : Econometric Results

Note ******
Note ****
Note **
(Standard error in parentheses)

Annex 3 : Multicollinearity

Annex 3 : Multicollinearity

Notes

1 This research was supported by the RECEP (European TACIS Programme).

2 See Jeffrey Frankel, Ernesto Stein and Shang-Jin Wei, (1997).

3 The implied norm will be described in a subsequent section.

4 Institutional variables include variables allowing to evaluate the degree of economic freedom, assuming that more freedom guaranties more economic activities. The degree of government intervention (trade policy, the number of state banks, price controls), the extent to which corruption within bureaucracy or within the State can impede trade activity, the protection of property rights, in the absence of which contract execution is likely to be jeopardized, are examples of such institutions. For the exact set of variables used in this paper, see the econometric section.

5 See Frankel and Romer (1999), who documented the impact of trade on growth in the very long run.

6 See Acemoglu, Johnson and Robinson (2001) and Hall and Jones (1999).

7 See Engermann and Sokoloff (1997).

8 Our empirical analysis covers the period from 1994 to 2001.

9 Data can be found at the following address : http://database.townhall.com/heritage/index/indexoffreedom.cfm

10 See, for example, Bruny, Carrère, Guillaumont, Melo (2002), Djankov, Freund (2000), Bougheas Demetriades, Morgenroth (1999).

11 The World Bank provides a database, The World Development Indicators, which ends up in 1999. Our data cover the period from 1994 to 2001.

12 According to the following formula : VOL = σ [(eij –eavg)/ eavg], where σ is the standard deviation.

13 See IMF [1984] estimation of 42 gravity equations, where exchange rate volatility has a positive influence on trade in 26 cases (for a total of 42) ; Brada and Mendez [1988], who assert that the manipulation of exchange rate fluctuations by relaxing the external constraint favors trade ; For the rationale behind the ambiguous impact of exchange rate volatility on trade, see Flandreau and Maurel (2001) and Maurel (2002).

14 For more details see Annual Report of Economic Freedom of the World.

15 Trade policy, Size of government, Monetary policy and Price stability, Capital flows and Foreign Investment, Banking and Finance, Legal structure and Security of private ownership, Freedom to trade with foreigners and Freedom of exchange in capital markets, Size of the black market. Data come from the Heritage Foundation, Index of Economic Freedom. The methodology is described in any Annual Report of Economic Freedom of the World, see Annex 1 at the end of the Paper.

16 The bias however works in both directions : imports are over-invoiced, but export are under-invoiced, see de Boyrie, Pak, and Zdanovicz, (2002).

17 X (respectively Y) can be EU, CEECs, CIS, and World, which is the rest of the world.

18 Potential trade : the trade level above what it would be if it was determined solely by the gravity forces, above what it would be if EU countries were not differing from the remainder of the sample.

19 See Graziani (1981).

20 This explanation is latent already in the aggregation argument, which recommends splitting the variables into disaggregated ones, that is to focus on more specific bilateral dyads. Ideally indeed, this process of splitting should end up with the construction of a dummy variable for each pairwise (i,j).

21 She estimates simultaneously a selection model for the formation of a currency union and a gravity equation. This allows taking into account the problem of simultaneity, for properly assessing the impact of the currency union on the intensity of bilateral trade flows.

22 Notice that within estimates allow assessing the impact of institutions on growth, since institutions are varying over time.

23 Reported in Annex 2.

24 P denotes the operator, which transforms the variables into their individual means; Q denotes the operator, which transforms the variables into deviation from the individual means.

25 Because they are not significant or have the wrong sign, we dropped the following variables: Fiscal Burden of Government, Government Intervention, Monetary Policy, Regulation.

26 See David Dollar and Aart Kraay (2002) for instance.

27 Indeed g2 = 0.

28 The ratio of potential trade on effective trade is set equal to the exponential of the dummy bloc variable coefficient. For instance the potential trade of entering the EU is exp(0.35), that is 142 %, because being member of the EU implies an extra trade of 42 %.

29 Indeed, 1/exp(-1.18)=3,25.

30 Guriev, Makarov and Maurel (2002) show at the firms level that the violation of property rights and the low enforcement of creditors rights are responsible for the increase in the virtual (or barter) economy. The latter is not neutral on the level of economic efficiency and output equilibrium, which are lower than in a well-functioning market economy.

Notes de fin

* ** Significant at 1 % level

* * Significant at 5 % level

* Significant at 10 % level

* ** Significant at 1 % level

* * Significant at 5 % level

* Significant at 10 % level

* ** Significant at 1 % level

* * Significant at 5 % level

* Significant at 10 % level

a Trade increases calculus is based on HTIV II gravity’s estimates of Table 4 ;

b « X — Y » means that institutions in X are set equal to institutions in Y ;

c Trade increases calculus is based on HTIV gravity’s estimates of Table 4 ;

d « X — Y » means that institutions in X are set equal to institutions in Y ;

Table des illustrations

URL http://books.openedition.org/pufr/docannexe/image/1607/img-1.jpg
Fichier image/jpeg, 40k
Titre Table 1 - OLS and GLS Results
Légende Standard Error in paratheses.
URL http://books.openedition.org/pufr/docannexe/image/1607/img-2.jpg
Fichier image/jpeg, 96k
Titre Table 1 (followed) - Institutional variables estimates ; Dependent variable : LnTRADEij
Légende Source : See Annex 2
URL http://books.openedition.org/pufr/docannexe/image/1607/img-3.jpg
Fichier image/jpeg, 428k
Titre Table 2 - Institution scores averaged over countries blocs
URL http://books.openedition.org/pufr/docannexe/image/1607/img-4.jpg
Fichier image/jpeg, 204k
Titre Table 1 (followed) - dummy estimates
Légende SOURCE : SEE ANNEX 2
URL http://books.openedition.org/pufr/docannexe/image/1607/img-5.jpg
Fichier image/jpeg, 128k
Titre Table 3 - Effective Trade in percent of Potential trade
Légende Source : From table 1, Authors Computation
URL http://books.openedition.org/pufr/docannexe/image/1607/img-6.jpg
Fichier image/jpeg, 92k
Légende This presentation is taken from Julie Lochard (2002) ; the number of variables is indicated in brackets ;
URL http://books.openedition.org/pufr/docannexe/image/1607/img-7.jpg
Fichier image/jpeg, 80k
URL http://books.openedition.org/pufr/docannexe/image/1607/img-8.jpg
Fichier image/jpeg, 24k
URL http://books.openedition.org/pufr/docannexe/image/1607/img-9.jpg
Fichier image/jpeg, 156k
Titre Table 4 - Within and Hausman-Taylor Results
Légende Source : See Annex 2
URL http://books.openedition.org/pufr/docannexe/image/1607/img-10.jpg
Fichier image/jpeg, 112k
Titre Table 4 (followed) - Institutional variables estimates ; Dependent variable : Ln TRADEij
Légende Source : See Annex 2
URL http://books.openedition.org/pufr/docannexe/image/1607/img-11.jpg
Fichier image/jpeg, 504k
Titre Table 4 (followed) - dummy estimates
Légende Source : See Annex 2
URL http://books.openedition.org/pufr/docannexe/image/1607/img-12.jpg
Fichier image/jpeg, 92k
Titre Table 5 - Effective Trade in percent of Potential trade
Légende Source : Table 4 and Authors Computation
URL http://books.openedition.org/pufr/docannexe/image/1607/img-13.jpg
Fichier image/jpeg, 72k
Titre Table 6 - How far are institutional scores in transition countries higher than in EU countries ?
Légende Source : See Table 2 ; Authors Computations
URL http://books.openedition.org/pufr/docannexe/image/1607/img-14.jpg
Fichier image/jpeg, 132k
URL http://books.openedition.org/pufr/docannexe/image/1607/img-15.jpg
Fichier image/jpeg, 64k
Titre Table 7a - Impact of institutional improvement on trade
Légende Note aaNote bb
URL http://books.openedition.org/pufr/docannexe/image/1607/img-16.jpg
Fichier image/jpeg, 104k
Titre Table 7b - Impact of institutional improvement on trade by CIS country
Légende Note acNote bd
URL http://books.openedition.org/pufr/docannexe/image/1607/img-17.jpg
Fichier image/jpeg, 172k
Titre Annex 1 : Economic Freedom of the World
URL http://books.openedition.org/pufr/docannexe/image/1607/img-18.jpg
Fichier image/jpeg, 664k
Légende Note ******Note ****Note **(Standard error in parentheses)
URL http://books.openedition.org/pufr/docannexe/image/1607/img-19.jpg
Fichier image/jpeg, 384k
Titre Annex 2 (followed) : Econometric Results
Légende Note ******Note ****Note **(Standard error in parentheses)
URL http://books.openedition.org/pufr/docannexe/image/1607/img-20.jpg
Fichier image/jpeg, 588k
Titre Annex 2 (followed) : Econometric Results
Légende Note ******Note ****Note **(Standard error in parentheses)
URL http://books.openedition.org/pufr/docannexe/image/1607/img-21.jpg
Fichier image/jpeg, 212k
Titre Annex 3 : Multicollinearity
URL http://books.openedition.org/pufr/docannexe/image/1607/img-22.jpg
Fichier image/jpeg, 236k

Auteurs

CNRS-ROSES, Université de Paris I et State University HSE

CNRS-ROSES, Université de Paris I et CEPR

© Presses universitaires François-Rabelais, 2004

Conditions d’utilisation : http://www.openedition.org/6540