How Much Do Institutions Matter for Trade? Evidence from Transition Countries

How Much Do Institutions Matter for Trade? Evidence from Transition Countries Astghik Mavisakalyan 1 ABSTRACT What is the effect of institutions on trade? A line of recent literature has observed an apparent positive correlation between measures of institutional quality and volume of trade. But this could be due to endogeneity of institutions, rather than causality. This study endogenizes institutions in estimating gravity equation for bilateral trade exploiting the differences in years under the governing system of communism as an instrumental variable for institutions. The longer a country had been under communism, the stronger might have been its institutions supporting the centrally-planned features of economy and thus, the weaker would have been their adaptability to market conditions. The reduced form equation estimates provide evidence for highly economically and statistically significant effect of the years under communism on the quality of current institutions. We find that institutions indeed have a beneficial effect on bilateral trade. However, the extent of their influence might have been misestimated while not having taken care of endogeneity. Overall, the results obtained via this approach shed a new light over the extent of explanatory power of exporter s and importer s institutions over trade that might be helpful in explaining the missing trade costs phenomenon having been 1 Economics Discipline, SEPS, University of Sydney, email: a.mavisakalyan@econ.usyd.edu.au. This paper is a version of my MA thesis submitted to the Graduate School of Humanities and Social Sciences, University of Tsukuba, Japan. I would like to thank Neantro Saavedra-Rivano, Daiji Kawaguchi, Tadashi Yamada for their excellent supervision and seminar participants at the University of Tsukuba for valuable comments. All remaining errors are of course mine. Financial support for graduate studies under the Joint Japan/World Bank Graduate Scholarship Program is gratefully acknowledged. 1

increasingly gaining attention in recent literature on international trade. Moreover, the results suggest that the number of years under communism indirectly affected trade, pointing out at the largely objective reasons behind the differences in the amount of trade and economic success of countries of Central and Eastern Europe and Former Soviet Union on the one hand, and on the other hand, at the lasting comparative disadvantage of transition countries in general, in inheriting the institutions from the communist regime. JEL Classification: F13, F15, P33 Keywords: Institutions; Trade; Gravity equation; Transition countries 1. INTRODUCTION It has been popularly argued that with the increased integration of the world economy over the recent decades the importance of borders and distance should have been declining. However, interestingly, expectations about the integration of the world economy and the reduced impact of geography have not received support by evidence on the change in importance of distance and borders in trade between the countries (see, for example, Brun, Carrere, Guillaumont and De Melo, 2002, Carrere and Schiff, 2004). Moreover, even countries like the US and Canada - neighboring economies, well integrated in terms of trade policy, have been shown to have much more intense trade between regions within a country than across the common borders in a number of studies on border effect in trade (see, for example, McCallum, 1995, Helliwell, 1998). Formerly having been almost fully isolated from the world economy, the transition countries appeared to be particularly striking examples of the processes of trade liberalization. However, after the collapse of closed centrally planned economic systems together with associated with them formal barriers to trade, the difference between actual and potential trade of some of the formerly communist countries remains large (see, for example, Babetskii, Babetskaia-Kukharchuk and Raiser, 2003, Elborgh-Woytek, 2003). Obstfeld and Rogoff (2000) in examining a number of puzzles in international macroeconomics, argue that the home bias in consumption may be largely attributed to 2

the unobserved costs of international trade. The missing trade costs phenomenon having been increasingly gaining attention in recent literature on international trade suggests that barriers to trade are much larger than the directly measurable transport costs and governmentally imposed trade constraints (Anderson, 2000). With the reduction in tariffs and transportation costs, a number of studies have focused on informal barriers to international trade as an explanation for missing trade costs (see, for example, Roberts and Tybout, 1997, Rauch and Trindade, 2002 for different approaches to informal barriers to trade). Among those barriers, differences in institutions have received considerable attention in recent years. The underlying argument for this approach is that insecurity and risk associated with inadequate institutions that do not support a favorable environment for economic activities protecting against diversion (problems in contract enforcement, corrupt customs, etc.) act as an additional tax on trade transactions negatively affecting the amount of trade (Anderson and Marcouiller, 2002). Anderson (2001) provides an overview on the effects of institutions on trade. This study considers institutions as mechanisms by which authority in a country is exercised focusing on formal regulations and their enforceability realized by bureaucracies in general. While the impact of institutions on transaction costs has received a considerable attention in the literature on economic growth and development (see, for example, Hall and Jones, 1999, Acemoglu, Johnson and Robinson, 2001), the impact of institutions on international trade flows has received relatively little attention so far. Anderson and Marcouiller (2002) have been one of the first to study the impact of institutions on international trade finding that institutions located in the importer s country constrained trade as much as tariffs did testing the model of import demand in which insecurity acted as a hidden tax on trade. This new view on the role of institutions has received some support from cross-country correlations between measures of institutions and trade found in several studies followed (Berkowitz, Moenius and Pistor, 2002; Berkowitz and Moenius, 2003; De Groot, Linders, Rietveld and Subramanian, 2004; De Groot, Linders and Rietveld, 2003). With respect to transition countries, it has been argued that the quality of institutions in those countries and the risks associated with trade explained the considerable trade 3

gap compared to trade levels in industrialized countries (see, for example, Babetskii, Babetskaia-Kukharchuk and Raiser, 2003; Elborgh-Woytek, 2003). In addition, the EBRD Transition Report 2003 identified the improvements in governance and the quality of institutions as key to increasing the integration of countries of FSU into the world economy. Although to a certain extent it s obvious that institutions matter for trade, we face a problem of reliable estimates of the effect of institutions on trade. It is quiet likely that open economies choose or are enforced to have better institutions. One reason behind that suggested by Ades and Di Tella (1999) is that competition from foreign firms that results from increased imports reduces the rents enjoyed by domestic firms and this in turn, reduces the rewards from corrupt behavior of government officials. On the other hand, exports present exogenously created rents for domestic firms which increase rewards from corruption. Another reason might be that international integration itself also places significant demands on a country s institutions. The market discipline imposed by being an open economy imposes good governance (see, for example, Wei, 2000; Islam and Montenegro, 2002). In addition, there may be many omitted determinants of trade that in the meantime will be correlated with institutions. Finally, the fact that the institutions variable is measured with considerable error and corresponds poorly to the cluster of institutions that matter in practice may create attenuation (Acemoglu, Johnson and Robinson, 2001). Thus, incorporating the institutional features in the analysis of trade determinants encounters the challenge to estimate their effect on trade flows. The cross-country correlation observed in most of the previous studies suggests a possible causal relationship between institutional quality and trade, but tells little about the direction of causality. Consequently whether institutions may help to explain the missing trade costs phenomenon cannot be determined while not taking care of this problem. Moreover, consistently estimating the causal effect of institutions may be important in determining the full causal effect of institutions on economic success in a wider sense. Each of the institutions and trade have been identified as key determinants of income in the voluminous literature on the subject (see Rodrik, Subramanian and Trebbi, 2002, for an overview). However, whereas the impact of institutions on income has received a 4

careful treatment in previous studies (see, for example, Acemoglu, Johnson and Robinson, 2001, Hall and Jones, 1999), still there might be a concern over underconsideration of their impact due to neglecting their intermediated influence on income through trade. Thus, consistent estimates of the causal effect of institutions on trade might be helpful also in deriving conclusions on a fuller effect of institutions on income. Estimation of the causal effect of institutions on trade encounters the challenge of finding an appropriate instrument for institutions. A line of recent literature on economic growth and development has investigated the exogenous sources of variation in institutions across countries with the purpose of isolating the causal effect of institutions on income or growth. Mauro (1995) instrumented for corruption using ethnolinguistic fragmentation. Hall and Jones (1999) used distance from the equator as an instrument for social infrastructure as according to them, latitude is correlated with western influence which leads to good institutions. Acemoglu, Johnson and Robinson (2001) exploited the differences in European mortality rates to estimate the effect of institutions on economic performance. As argued by them, in places where Europeans faced high mortality rates, they could not settle and were more likely to set up extractive institutions which persisted for a long time. A number of studies have been specifically devoted to investigating the determinants of institutions themselves. Examples include the comprehensive study by La Porta, Lopez-de-Sinales, Shleifer and Vishny (1999) where they tested the effect of ethnic heterogeneity, religious affiliation of population, legal origins of countries on the quality of institutions. Engerman and Sokoloff (1997) examined the implications of factor endowments and inequality for the differences in institutional development. The findings of these studies have been important for understanding what factors interacted in shaping the pattern of institutional adaptation and to what extent policy measures could be instrumental for influencing them. Transition economies are an interesting context to investigate the patterns of institutional change because the entire set of institutions appeared in a necessity to be remodeled in 1990s. The move from a system based on state planning and an allocation of resources by government dictate to a system of decentralized market allocation in transition countries necessitated substantial change in laws and regulations. The outcome 5

of concerted efforts across the region to redraw the set of laws and regulations governing economic exchange has differed widely across transition countries, with institutional performance showing significant divergence after the first decade of transition (Raiser, Di Tommaso, Weeks, 2000). This study shows that the length of communist regime in transition countries appears to be a reasonably exogenous determinant of such variation in institutional performance that largely constrained the institutional change in those countries. The argument is based on the following premises: - One of the main characteristic traits of centrally-planned economies inherent to those systems, that might have left a lasting mark on economic structures, is the fact that under central planning inherent to the governing system of communism there was no need for the legal and institutional framework underpinning a market economy, the state s intent to create institutions to maintain its power and extract resources without much regard for protecting the economic interests or the liberties of population has been prevalent. - As longer a country had been under the governing system of communist, as stronger might have been both its formal and informal institutions supporting the centrallyplanned features of the economy because of more investment made throughout the time in building them. Consequently, the weaker would be their adaptability to market conditions. - Because setting up institutions that place restrictions on government power and enforce property rights is costly, institutions usually exhibit a feature of long-term persistence to change (Acemoglu, Johnson and Robinson, 2001). Thus the weaker the adaptability of institutions to market conditions supporting restrictions on government power and enforcement of property rights, the stronger would be the persistence to adapt institutions to market conditions and thus more inadequate institutions would a country have at the observed period of time. This approach is similar in its underlying logic to the one by Schleifer, 1997, arguing that transition of government from a communist state to an institution supporting market economy is crucial in the success of transition process. The approach is also supported by economists who, in their attempt to understand what accounts for differences in the 6

outcomes for economic reforms, concluded that initial conditions are endogenous to economic liberalization. Examples include studies by Raiser, Tommaso and Weeks (2000), De Melo, Denizer, Gelb and Tenev (2001), Falcetti, Raiser and Sanfey (2002) who have tested the consolidated impact of initial conditions including, the length of communism, on different aspects of economic performance of transition countries. However, we are not aware of studies that have explicitly pointed out and have empirically traced the link between the time under the communist regime, the quality of institutions after almost ten years of transition and trade in a similar pattern. We find a strong negative first-stage relationship between the number of years under the governing system of communism and the quality of current institutions, which should be an interesting finding in its own way. We consistently estimate a gravity equation controlling for the institutional quality along with a range of traditional gravity variables, using the 2SLS estimation technique on a cross-country data set for 1998. Even after endogenizing the institutional quality, we find that it is positively correlated with trade although the extent of correlation would have been misestimated while not having taken care of endogeneity. Thus, while it is likely that trade openness affects institutions, we find evidence that institutions, in their turn affect trade. Moreover, our instrument suggests that the existing gap in trade and subsequently, economic success in general between the transition countries of CEE and FSU is a largely objective phenomenon, explained by the relative length of the communist regime in laters. While not understating the frequently claimed argument on the relative success in reforms of the CEE countries as compared to those in FSU, this finding rather comes to point out at a possibility of largely objective reasons behind that success. Overall, this paper contributes to the literature on the effect of institutions on trade in two ways. First, it consistently estimates the extent to which better institutions cause trade taking into account the endogeneity of institutions, that may help to shed light on their explanatory power over the missing trade costs. Second, it applies a new instrument for institutions that helps to reveal the largely decisive impact on current economic performance of transition countries of the number of years under communist regime. 7

In addition, the study indirectly contributes to deriving conclusions over the fuller impact of institutions on cross-country income and growth levels. The causal effect of institutions on trade is a potential mechanism through which institutions may affect overall income and growth rates. In that sense the paper can be seen as an attempt to advance the understanding of how institutions affect real outcomes. The next chapter presents the empirical strategy and the results. Chapter III concludes. 2. EMPIRICAL EVIDENCE 2.1. Empirical Methodology 2.1.1. The Gravity Approach To estimate the effect of institutions on trade, we rely on the standard gravity model of bilateral trade which is evidently one of the most widely used empirical approaches in the literature to estimate the determinants of bilateral trade flows. Inspired by Newton s Law of Gravitation, the gravity model at the core predicts that the volume of trade between two countries should increase with their size and decrease with transport costs by analogy to the formula for gravitational attraction between two masses. In its earliest form, the gravity model has been introduced by Tinbergen (1962) and Pöyhönen (1963). While the core gravity model has been used empirically since the time of those studies, the theoretical justification behind the core gravity model has evolved gradually. A number of studies have shown that the gravity model is consistent with different classes of models of international trade (Deardorff, 1998, provides a survey of literature on theoretical foundations of gravity model). According to Helliwell (1998) the gravity model has gone far from being a theoretical orphan to being a favored child of main theories of international trade. As for the applicability of the gravity model for transition countries, Paas (2002) points out that despite continuing discussions and uncertainty about the foundations of the gravity model it is possible to say that those theoretical 8

considerations are also valid when exploring the changes in international trade patterns during transition. The gravity model has proven to be quite flexible in a number of empirical applications to test for the impact on bilateral trade flows of border effects (see, for example, McCallum, 1995, Anderson and van Wincoop, 2003), infrastructure and internet (see, for example, Bougheas, Demetriades and Morgenroth, 1999, Freud and Weinhold, 2004), regional trade blocks (see, for example, Frankel, Stein and Wei, 1995, Soloaga and Winters, 2001), common currency (see, for example, Rose, 2000, Frankel and Rose, 2002), specialized international organizations (see, for example, Rose, 2002, Subramanian and Wei, 2003) among other subjects. In recent years, gravity models have been successfully applied also in empirical studies of changes in international trade pattern and reintegration of economies in transition (see, for example Firdmuc and Firdmuc, 2003, Babetskii, Babetskaia-Kukharchuk and Raiser, 2003). Numerous gravity model specifications are used in the literature with a listing of variables correlated with bilateral trade. We adopt a version of the conventional gravity model with a specification of the following form: ln( Export α X + u (1) ij + 1) = 0 + α1 lngdpi + α 2 lngdpj + α 3 ln Distij + α 4 In addition to core gravity variables of log GDPs and bilateral distance, other control variables, X ij, reflecting the factors affecting the bilateral trade costs between two countries are included (including institutions of each of the countries). Our purpose is to extend the gravity model beyond the core to explain regional trade patterns and to reduce worries about inconsistency in estimated coefficients due to a variety of possibilities of endogeneity involved in the model. Thus we augment the basic gravity equation with a number of extra conditioning variables but limit the number of variables affecting trade only to those being naturally exogenous to be able to achieve consistent estimation for the coefficients for institutions after controlling for endogeneity of institutions and GDPs in the model. As a number of potentially endogenous variables affecting trade and being correlated with institutions (tariffs and non-tariff barriers, exchange rates, etc.) are excluded from the specification, our estimates may be considered as of the full impact of ij ij 9

institutions on trade including the effect of those other variables (excluding the GDPs that are an explicit part of the gravity model specification). In the core gravity model, the bilateral distance between the capital cities of the two countries is used to measure transportation costs. However, there are several other geographic factors that can affect transportation costs and thus the volume of trade. In particular, the cost of moving goods between two adjacent locations is lower than the cost of moving goods through a third country. Similarly, the cost of shipping goods across water is lower than over land. Moreover, not all international trade terminates in the capital cities where the bilateral distance is measured. As a result, those countries with larger surface areas should have higher transportation costs ceteris paribus than those with smaller surface areas. Thus we consider three additional variables to measure geographic factors affecting trade. Gravity models allow for testing the impact of not only real geographical distances, but also of virtual distances on trade (Laaser and Schrader, 2002). It has been common to include measures of linguistic and historical ties to capture path dependence in trade flows. In particular, if two countries share the same language then trade is more likely to occur. A common language may directly lower translation costs and, thereby, transaction costs. Moreover, if countries have been in horizontal or vertical colonial relationships then the volume of trade among them is hypothesized to increase (Eichengreen and Irwin, 1998). We consider three measures of linguistic and historical ties in the model. We include importer s and exporter s characteristics as separate regressors which allows meaningfully differentiate importer and exporter features and their effects (e.g., importer s log GDP and exporter s log GDP are included as separate regressors) on trade. As most of the applications, we specify the model in double-log form, expressing the dependent variable and most of the independent variables in logs that allows interpret the coefficients as elasticities. However, considering that there might be country pairs with zero trade values and taking logs would omit those observations that might convey information about why low levels of trade were observed, we express the dependent variable as 1998. ln( Export ij +1). The estimation is carried out using the cross-sectional data for 10

Anderson and van Wincoop (2003) argue that trade between countries do not just depend on the characteristics between those countries but also on the barriers between those countries and the rest of the world ( the multilateral resistance ) and suggest including country-specific dummies to control for that effect. However, considering the time-invariant nature of our proposed instruments as discussed earlier, inclusion of dummies is not feasible because of the perfect multicollinearity problem in the first stage regression. Taking into account that the focus of this study is on institutions and controlling for their endogeneity in the gravity equation estimation, we consider giving up with including country-specific dummies as a shortcoming of reasonably low overall consequence. 2.1.2. Data To estimate the gravity model we employ data on bilateral exports, incomes, distance as well as on institutions, on a number of geographical, cultural and historical characteristics as additional variables affecting trade costs for 147 countries (see Table 1 for the list of countries in the sample) for 1998. We are missing observations for some of the regressors for some countries so the usable sample is limited to those countries 2. Most of our data for traditional gravity controls are from Rose (2004) as well as from the World Development Indicators 2003 and CIA World Factbook 2003. Nonstandard sources will be indicated when the relevant variable comes up. The detailed description of variables and sources is provided in Table 2. Mostly, those include controls traditionally applied in estimating gravity equations. Table 3 provides descriptive statistics for data used in the gravity model estimation. Log of real exports is the dependent variable in the gravity model estimation. The sample includes 18450 bilateral export relationships. There are large differences in exports in our sample and the standard deviation of log of exports is 5.425. Integrating the institutional environment in the analysis of bilateral trade encounters the challenge of finding appropriate indicators to measure it. Recent studies incorporating 2 Observations for some countries are excluded for other considerations as described later in the text 11

the concept of institutional quality have been relying on subjective measures on the perceived quality of institutions that allows capture the institutional environment closely and directly (Grogan and Moers, 2001). We also take this approach in choosing the variables for institutional quality in this study. While a serious problem in doing that is the potential existence of serious measurement error in dealing with such type of highly subjective data, our approach of treating institutions as endogenous in this study should be relevant to control for this problem. Following Babetskii, Babetskaia-Kukharchuk and Raiser (2003) we construct our measure of institutional quality as the average of the World Bank s new governance indicators for quality of regulations, rule of law and extent of corruption. This measure is appropriate for our purpose since the focus in this study is on general institutional environment that matters for trade. However, as an additional test, we provide estimations for the separate impact of each of those institutional variables on trade. The quality of regulations summarizes various indicators of the ability of the government to formulate and implement sound policies. It includes in particular measures of perceptions of the burdens imposed by excessive regulation in areas as foreign trade and business development. The rule of law measures the extent to which agents have confidence in and abide by the rules of society, including perceptions of incidence of crime, the effectiveness and predictability of the judiciary, and the enforceability of contracts. The control of corruption measures perception of corruption, conventionally defined as the exercise of the public power for private gains. The indicators are based on several hundred individual variables measuring perceptions of governance, drawn from 25 separate data sources (for detailed description of governance indicators see Kaufman, Kraay and Mastruzzi, 2004). All scores lie on average between -2.5 to 2.5, with higher scores corresponding to better outcomes. Our composite index of institutional quality varies between -2.110 (Democratic Republic of Congo) to 2.137 (Singapore) with the mean score of 0.185. The institutional quality for transition countries varies between -1.417 (Tajikistan) to 0.873 (Hungary), excluding Germany with 1.763. Our first instrument for institutions is the number of years under governing system of communism which comes from De Melo, Denizer, Gelb and Tenev (2001) (Table 4). 12

The countries with non-zero values for the years are the transition countries in CEE and FSU. As there is a difference of only few years in the collapse of communist regimes in those countries, we assume that that difference has not been of a qualitative impact on the potential of adaptability of institutions. Countries in East and South East Asia with socialist legal origin (China, Vietnam, Laos and Cambodia) are excluded from the analysis because of the qualitatively different mode of their institutional development under the governing system of communism which might not be comparable to that of our target group of countries and in addition, because of at least de jure persistence of communist regimes in those countries up to now (excluding Cambodia). For the case of Germany, to account for the influence of the institutional heritage of Eastern Germany on the unified country, we calculate a measure of the relative length of the communist regime. According to the last national census (1986) held before the unification of two parts, GDR had a population of 16.6 million and FRG of 58.1 million - 74.7 million in total. Thus considering the ratio of the population of GDR to the total population of Germany (=0.22), we calculate a measure of relative length of communist regime in Germany (0.22*41=9). Overall, the number of years under communist regime for transition countries, excluding Germany, ranges from 41 (Poland) to 74 (Ukraine and Russia) years. The second instrument for institutions that we employ primarily with the purpose of making it possible to test the exogeneity of instruments, is the legal origin of the Company Law or Commercial Code of each country that comes from La Porta, Lopezde-Silanes, Schleifer and Vishny (1999). National legal traditions are divided into English Common Law, French civil law, German civil law, Scandinavian law and Socialist/Communist law. Those traditions were developed in England, France, Germany, Scandinavia and the Soviet Union but then spread through the world through conquest, colonization, imitation and voluntary adoption. The legal systems are viewed as indicators of the relative power of the State vis-à-vis property owners. Thus having an explanatory power over the current institutional qualities of countries and in the meantime, being largely exogenous in their nature, they can serve as an instrument for institutions. As our purpose in this case is to simply find an exogenous source of variation of institutions rather than to find out the separate effect of each legal tradition on 13

institutions (which has been done in several studies devoted to that purpose), we use a variable ranging from 1 to 5 to identify the differences in legal traditions that have an explanatory power over differences in institutions. 2.2. Empirical Results 2.2.1 Ordinary Least Squares Estimation The estimated form of the gravity model is specified in the following extended way: ln( Export + α Inst i ij + α Border 5 11 ) = α ij + α Inst 0 + α1 ln i + α 2 ln j + α 3 ln ij + α 4 ln( i j ) + α Landl 12 6 j i + u GDP + α Landl ij 7 j GDP + α Lang 8 ij Dist + α ComCol 9 ij Area Area + α Colony 10 ij + + (2) The coefficients of interest throughout this paper are α 11 and α 12, the effect of exporter s and importer s institutions on trade. The first column of Table 5 reports the OLS regression result of the gravity equation. It shows that there is a strong correlation between the measures of institutional quality and log of exports. However, due to the number of reasons specified in the previous chapter leading to endogeneity of institutions, the effects reported in column 1 may not be interpreted as causal. Besides that problem, the reported coefficients might be biased because of the endogeneity of GDPs in the model. This could be due to a number of reasons. First, there is a simultaneous determination of exports and exporter s and importer s GDPs in the model stemming from the income identity: the higher the exports from country i to country j, the higher the GDP of country i and the lower the GDP of country j. Second, there may be many omitted determinants of trade that will be correlated with GDPs. Finally, the fact that the variables for GDPs are often measured with error may create an additional source of endogeneity. In our context, especially important may be the problem of unreported activities taken place in underground economy which might be of a large impact on market capacities of each of the trading partners that may be neglected due to the measurement error. 14

We next adopt the 2SLS estimation technique to estimate the model. 2.2.2 The Quality of Instruments We use the share of working age population in total population as an instrument for GDP. The results from the reduced form regression are reported in Table 7. We reject the null hypothesis of no correlation between the log of share of working age population and log of GDP with a high significance level (t statistics of 109.45). Thus, our instruments are correlated with the endogenous variables they are used for. In addition, the coefficients that we obtain for the log of the share of working age population in total population are positive and sensible in size. To evaluate the causal effect of institutions on exports as our first strategy, we exploit the years a country has been under the governing system of communism as a source of exogenous variation in institutions. As expected, we obtain a negative effect of the years in communist regime on institutions in estimating the reduced form equation. According to our results, one additional year under the governing system of communism would result in a decrease of the institutional variable of 0.023 points. Considering that the institutional variables in our sample fall in the interval of -2.110 to 2.137, this result is of sensible economic significance. In particular, 72 years under the governing system of communism (which is the average number of years for the FSU republics), would result in a decrease in the institutional variable on average, of 1.656 points. In addition, 28 additional years under the governing system of communism (which is the difference between average number of years for the FSU and CEE) would lead on average, to a decrease in the institutional variable of 0.644 points. We reject the null hypothesis of no correlation between the years in communist regime and the institutions at a high significance level (t statistics of 104.41). As our second strategy, we consider countries legal systems which have been tested for potential explanatory power over the quality of institutions. Based on the results from the reduced form regression we reject the null hypothesis of no correlation between the legal origin and the institutions at a high significance level (t statistics of 5.36). 15

In addition we jointly test the coefficients for legal origin and years under communist mode for correlation with institutions and we again, reject the test at high significance level (F statistics of 3946.73). Although our instruments appear to be correlated with the endogenous variables they are used for, the validity of our approach would be threatened if they were correlated with the error term. We conduct an over-identification test to ensure that our instruments are truly exogenous (Table 8). According to the results that we get we cannot reject the null hypothesis that all the instrumental variables are uncorrelated with the error term. Thus we admit that our results do not generate any evidence for the endogeneity of instruments. In order to completely rule out the concerns over possibility of endogeneity of the years under communist regime in the model coming from frequently expressed approaches that the artificial trade patterns developed throughout the years in communist regime might needed time to adjust due to the need for changes in production structures and thus might have directly affected the current trade patterns, we include the number of years under communist regime as an additional regressor in the gravity equation treating institutions and GDPs as endogenous in estimation. We obtain insignificant coefficients for the years under communist regime thus proving once again, that our suggested instrument is truly exogenous 3. Our results support the argument by Gros and Suhrcke (2000) that artificial trade patterns inherent to centrally-planned economies have been changed almost immediately and thus the length of the period under communist regime have not had directly affected trade through the same long-term adaptability channels, as in the case of institutions. 2.2.3 Two Stage Least Squares Estimation The second column of Table 5 reports 2SLS estimates of gravity equation. The 2SLS estimates of the impact of institutions on income are positive and significant. The 2SLS estimate of the impact of exporter s institutions on log of exports is equal to 0.520 and that of importer s institutions is 1.396. As we have semi-log relationships, 3 The estimation results are available upon request 16

these results can be interpreted as 1 point increase in exporter s institutional quality leading to 0.682 fold increase in the amount of trade ( e 0.520 1) and 1 point increase in importer s institutional quality leading to 3.039 fold increase in the amount of trade ( e 1.396 1). To get a sense of the magnitude of the results, let us consider the case of Ukraine, a country with a maximum number of years in communism. If Ukraine would have spent as much time in communism as Poland, with the shortest time in communism among transition countries, 41 years, (excluding the special case of Germany), its institutions would be better by 0.759 and if it would not have spent any time at all, they would be better by 1.702 points. In the first case the Ukraine s imports with a given trade partner would be 1.885 times higher and in the second case, they would be 9.762 times higher. Similarly, its exports with a given partner would be respectively 0.484 and 1.423 times higher. In the case of exporter s institutions, the coefficient is smaller than the OLS estimate which suggests that the problem of reverse causality, explicable with the proposition that more open economies have better institutions, that had created an upward bias on OLS estimates has been important. For the importer s institutions, now we have a larger coefficient than the one for OLS estimation suggesting the relative importance of measurement error in institutions having led to attenuation bias in OLS estimates. In addition, the relative importance of exporter s and importer s institutions for bilateral trade is reversed with importer s institutions playing a larger role for bilateral trade flows. Inadequate institutions in importer s country may create a risk of predation at the border, as shown by Anderson and Marcouiller (2002). In addition, importer s institution may matter for timely and accurate payments to exporters. Berkowitz and Moenius (2003) note that the risk faced by exporter of not getting paid can be offset by contractual methods such as letters of credits. However, reservation should be made to the end that mechanisms ensuring reliability of those contractual methods should be highly correlated with the overall institutional environment of the importer s country and thus the possibility of the risk faced by exporters related to payments may not be ruled out. The deliveries of goods comes before their effective payments as argued by Turrini and van 17

Ypersele (2002) and thus there may be a wide range for both objective and subjective reasons for delayed payments. On the other hand, exporter s institutions may matter for offsetting the risk of getting substandard products (Berkowitz, Moenius and Pistor, 2002). Our results thus imply that risks related to predation at importer s border and to problems in timely and accurate payments that may be associated with poor institutions, influence international transaction costs more that the risk that the exporter would not comply with product specification. For coefficients of GDPs again, the problems of biases are rectified. The OLS estimate of exporter s GDPs might have been underestimated due to relative importance of measurement error in GDP having led to attenuation bias. The OLS estimate of importer s GDPs might have been overestimated due to possibility of unobserved shocks having negatively affected both trade and importer s GDP (for example, the Asian Crisis during 1997-1998). The Hausman t-statistics reported in the third column of Table 5 points out at the significant difference of our OLS and 2SLS estimates thus proving at the validity of our endogeneity assumption. In addition we complete a test of endogeneity that yield results (reported in Table 6) confirming once again the validity of that assumption. The remaining gravity controls are highly significant, sensible in size and bear the signs in line with those in previous studies. In addition, we estimate the separate effect of each of the components of institutional environment on trade. As the results reported in Table 9 indicate, the coefficients for each of the reported institutional components are positive and statistically significant. The quality of regulations appears to be of the highest impact on trade, while the extent of corruption has the lowest impact. This result may be interpreted as the relative importance of formal institutional environment reflected in the quality of regulations over the informal institutions related to the practice of enforceability of formal rules that may be interpreted by the fact of availability of more reliable information about the formal institutions that trading partners are able to rely on while making their decisions. 18

2.2.4 Robustness Checks We check the robustness of our results presented in Table 10. We consider robustness along four dimensions. One feature of the gravity model regressions that is often considered as problematic for calculating standard errors, is that the same country s characteristics are represented repeatedly on the right hand side. Defining those repetitions as groups, error terms within those groups are likely to be correlated with each other, while error terms across groups should not correlate (Berkowitz and Moenius, 2003). In order to test whether the significance obtained in benchmark results would be preserved while accounting for that grouping effect, we test the benchmark specification with calculating standard errors with clustering by exporter and importer reported in columns 1 and 2 accordingly. As a result the standard errors are considerably higher than those in benchmark results however with the statistical significance of estimates of only exporter s institutions hurt with insignificant estimate obtained in case of clustering by exporter in column 1. The high statistical significance of the estimates of importer s institutions remains robust to controlling for this feature of gravity equation. Columns 3 and 4 examine the sensitivity of the results with respect to the choice of the instrument. In columns 3 and 4 we report the results of estimation with institutions instrumented separately with the years in communist regime and the legal origin accordingly. While the results remain robust in the first case, in the second case the estimate of institutions of exporter looses statistical significance while that of institutions of importer rises in size while preserving the statistical significance. In addition, in order to minimize the possibility that our variables for institutions are proxying for omitted variables in the regression, we include variables that could be correlated with institutions and with trade at the same time. In column 5 we report results of estimation of augmented baseline model which controls for the WTO membership of countries. The intuition is that countries that are especially open to trade might also be more likely to reform their institutions. To test the possibility that the omission of variables that proxy for openness to trade might affect the results, we add dummy 19

variables indicating that the country is a member of the WTO (source: Rose, 2004). The estimated results of dummies appear to be statistically significant and affect the significance of the exporter s institutions while the estimate of importer s institutions remains robust to this change in specification of the baseline model. Next, in column 6 we present results with the baseline model augmented with the real exchange rates of exporting and importing countries (source: WDI, 2004). The main concern in doing this is that the real exchange rates data are available for fewer countries and using them sharply reduces the sample size. While the estimates for real exchange variables appear to be significant in the regression, they do not hurt the significance of the estimates of the institutions of exporter and importer, both of which remain to be of significant positive impact on trade. One limitation with augmenting the model with these potentially omitted variables is their possible endogeneity in the model which is not controlled for due to lack of appropriate instruments. Columns 7 and 8 examine the sensitivity of the results to the sample used in the estimation. As the sample changes, so do both the number of observations and the average values of the dependent variable both reported on the bottom of the table. Column 7 excludes purely the intra-developing country trade and column 8 excludes the intra-transition country trade (Germany is not excluded). This tests the idea that institutions effect is a purely developing country or transition country phenomenon which doesn t appear to be so as the estimates of the institutions for both exporters and importers preserve their significance. 3. CONCLUSION Countries with better institutions trade more than do countries where the quality of institutions is worse. The analysis suggests that good institutions protecting against public diversion help stimulate trade. The results point out at misestimation of the impact of exporter s and importer s institutions on trade while not having taken care of endogeneity of institutions in trade-institutions linkage. This helps to shed a new light over the explanatory power of institutions over the missing trade costs phenomenon. Moreover, 20

as the positive impact on incomes of both institutions and trade has been already established through careful treatment in previous studies, these results point out at a larger causal effect of institutions on economic development through their intermediated impact on trade. The results also suggest that trading partners are more sensitive to formal institutional environment reflected in their perceptions on burdens imposed by regulations in making their decisions to trade as compared to informal institutions related to the practice of enforceability of formal rules. This result may call for a policy priority for increased efforts towards bringing the business regulations in line with internationally accepted standards that appears to be of higher importance for trade as compared with fight against corruption and for enforceability of laws that still being important appear to be of a weaker impact on trade. In addition, the results suggest that the number of years under the governing system of communism indirectly affected trade through having affected institutions, pointing out at the largely objective reasons behind the differences in the amount of trade and economic success of countries of CEE and FSU on the one hand, and on the other hand, at the lasting comparative disadvantage of transition countries in general, in inheriting the institutions from the communist regime. There are many limitations existent in this study that we have not been able to overcome. Although the arguments posed in this study establish a basis for the claim that a successful economy requires appropriate institutions and thus we can conclude that institutions even more "matter" in a broad sense than we previously tended to know they did, it is still difficult to assess institutional quality in practice. Thus to the extent of difficulty of measuring the quality of institutions and virtually, impossibility to sort out the complex relationship between the different components of real-world institutions, this study has not been instrumental to yield sufficiently specific results. In addition, while looking at an integrated measure of institutions or at the independent effects of institutional components is useful for deriving conclusions at a conceptual level on the basic relationships of interest, looking at the effect of possible interactions of institutional components in affecting trade and income would be an interesting and fruitful direction for extension of the study. 21

Another limitation is that considering the specifics of institutions and the process of institutional change, this study based on cross-sectional results is under a serious threat of missing important information on the way the institutions-trade spiral functions. Unfortunately, obtaining relevant long time series data to study the dynamics of institutional change and its impact on economic performance still remains a challenge for the research in the field. In addition, although the established link between the number of years under the governing system of communism and trade and thus economic performance in general seems to shed light on largely objective reasons behind the poor performance of FSU states and formerly communist countries in general, still the results should be interpreted with a high degree of caution. As underlined in Rodrik, Subramanian and Trebbi (2002) a good instrument is simply a way of identifying an exogeneous source of variation in the independent variable of interest, and as such it is unable to lay out a full theory of causes and effects. Thus further research on establishing stronger theoretical background behind this finding remains pending. Another way for the study to be improved could be adjusting the number of years under the governing system of communism to meet the peculiarities of the system for each country, considering that the influence of the number of years on institutions of different countries has been different. An approach to adjustment of the nominal number of years under the communist regime based on some measures of the influence of the main values resulting from the regime on the society could be helpful to come up with real number of years under the communist regime. REFERENCES Acemoglu, D., Johnson, S. and Robinson, J. A. (2001). "The Colonial Origins of Comparative Development: An Empirical Investigation." The American Economic Review 91(5). Ades, A., and Di Tella, R. (1999). "Rents, Competition and Corruption." The American Economic Review 89(4). Anderson, J. E. (2000). "Why do nations trade (so little)?" Pacific Economic Review 5(2). 22