Разнообразие, дискриминация и дипломатия: политическая экономика международной торговли в многоэтнических обществах

Размещено на http://www.allbest.ru/

ПРАВИТЕЛЬСТВО РОССИЙСКОЙ ФЕДЕРАЦИИ

ФЕДЕРАЛЬНОЕ ГОСУДАРСТВЕННОЕ АВТОНОМНОЕ ОБРАЗОВАТЕЛЬНОЕ

выпускная квалификационная РАБОТА

Разнообразие, дискриминация и дипломатия: политическая экономика международной торговли в многоэтнических обществах

М. С. Чупилкин

Москва, 2020 г.

Аннотация

Большинство исследований этнического разнообразия сосредоточены на экономике и политике внутри страны. В данной статье рассматривается роль эт- ничности в международной экономике путем изучения влияния этнического разнообразия на внешнюю торговлю. Влияние этничности зависит от политических институтов. Когда у правящей группы есть стимулы причинять вред другим группам посредством дискриминации или конкуренции за власть, этническое разнообразие оказывает негативное влияние как на импорт, так и на экспорт. Эффект снижается в иерархически ранжированных этнических системах, в которых у правящей группы нет угрозы потерять власть. Данная работа также подчеркивает пробел в исследованиях в области связи торговли и демократии, поскольку этническая политическая конкуренция включена в индексы демократии как демократическая и, следовательно, смещает результаты вниз. Хотя эффект демократии не оказывает значимого эффекта в данных, после вычитания периодов этнической конкуренции из периодов демократии эффект демократии становится более точным и значимым.

Research on the effects of ethnic diversity has been mostly concentrated on domestic politics. This paper examines the role of ethnicity in international economics by studying the effect of ethnic diversity on foreign trade. The effect of ethnicity is contingent on political institutions. When the governing group has incentives to harm other groups through discrimination or competition for power, there is a negative effect on both import and export. The effect is lower in hierarchical ranked ethnic systems in which the power of the governing group is secure. This finding also highlights a gap in research on the effect of democracy on trade as ethnic political competition is included as democratic in the indexes of democracy and therefore the results are biased downward. While democracy per se had no significant effect in the data, subtracting periods of ethnic politics from periods of democracy made the effect of democracy more precise and significant.

Introduction

International trade does not happen naturally, but demands public action and therefore involves political competition. High levels of trade are contingent on lowering formal and informal barriers, building infrastructure, adjusting laws and matching domestic producers (or buyers) with foreign buyers (or producers). The presence of several ethnic groups can both increase potential trade by providing ethnolinguistic ties with other countries and decrease it by making it more difficult to come up with foreign trade policy that satisfies all actors. Similarly to other domains with active presence of the state and redistributive conflict, foreign trade policy opens space for sectarian politics which can take place both through simple maximisation of group-specific and not national benefit and through active harm inflicted on trade of other groups.

This study shows that excluded groups in conflict with the government (through active discrimination by the governing group or separatism by the excluded group) have a negative effect on international trade. The main mechanism is not the simple absence of representation, but the active harm inflicted by the government on excluded groups or by separatist groups on the government. Case studies show that the negative effect is caused by a wide array of policies from manipulating the exchange rate and banning specific import or export to creating excessive customs rules.

However, providing equal access to executive power to all ethnic groups is not a recipe for trade promotion. Sectarian politics is often the result of insecurity of the governing group or excessive competition between different groups involved in the power-sharing regime. Long-term national development is likelier to be provided by the dominant group which feels secure in power. Due to the need of centralised and consistent long-term policy, trade is higher in ranked ethnic systems and lower in power-sharing regimes.

This study is also of interest to the scholars of democracy and international trade. Ethnic political competition is coded as democratic in the indexes of democracy such as Polity IV while having a negative effect on trade and consequently the effect of democracy in ethnically diverse countries is biased downward by construction.

After explaining the mechanism through which ethnic diversity impacts foreign trade and elaborating on how different institutions can make diversity both harmful and beneficial, I test my hypothesis empirically. I integrate the Ethnic Power Relations data from Vogt et al. (2015) in the gravity model of bilateral trade. Ethnic Power Relations data codes the level of access to power for each ethnic group in the country-year. As a result, it is possible to see how the effect of an additional ethnic group on foreign trade is conditional on whether it is not-represented, is in the state of animosity with the government, separatist or participates in the power-sharing regime. Ethnic Power Relations data also allows to create a variable for ranked ethnic systems which was shown by Horowitz (1985) as a good indicator of whether the governing ethnic group feels secure in power. международная экономика демократия торговля

The main result is that the number of excluded groups in animosity with the government (discriminated and separatist groups) is negatively correlated with both import and export on a statistically significant level. In Africa and Asia separatist groups which have a developed capacity to resist the government and govern themselves have an additional negative effect. These findings are partly replicated if the share of excluded population is used as an independent variable. Furthermore, trade is positively correlated with ranked ethnic systems and negatively with power-sharing regimes. It is important to note that the effect on trade is additional to the effect on economic development. Therefore, ethnic politics decreases trade more than it decreases economic growth. The main cause for this is that foreign trade is less natural than regular economic activity, demands more public action and is therefore more susceptible to sectarian politics

Second, there is a significant heterogeneity of the effect across continents. The greatest effect of ethnic politics is in Asia where trade is susceptible to most of the measures from discrimination and separatism to the negative consequences of power sharing regimes. Even though there are also significant effects in Africa and Europe they are more mild and less variables reach statistical significance. This points out that high levels of ethnic politics may be the feature of states in the middle of their development, when the ''trade pie” is already significantly big to provoke competition and there are political institutions for ethnic mobilisation, however the country is not yet as centralised as to have a unified foreign policy which is not susceptible to factional politics.

Third, there are important empirical implications for the study of the effect of democracy on trade. As a robustness check, I replicate my results using Polity IV data, specifically the component of the Polity index called factional competitiveness of political participation. Factional dummy is negatively correlated with trade and significantly so in the ethnically diverse environment. However, factional politics is counted as democratic by Polity IV, therefore the index of democracy by construction has a characteristic which is negatively associated with trade and studies of the effect of democracy on trade are biased downward. After subtracting periods of factional political competition from the periods of consolidated democracies, the effect of democracy dummy on trade becomes significantly positive and more precise.

The main contribution of this paper is the provision of evidence of negative correlation between ethnically based sectarian politics and international trade and the proposition of several possible mechanisms. As shown in the literature review, the effect of ethnic diversity has been mainly studied on a domestic level and there is a limited number of research studying connection between ethnic diversity and foreign policy. This study is the motivation for further research in interconnection between ethnic politics and international political economy.

1. Literature review

In economics and political science there is a long tradition of studying the effect of ethnic diversity. Ethnic fractionalisation has been proven to be one of the reasons of low economic growth in Africa in Easterly and Levine (1997), to undermine the provision of public goods in Alesina et al. (1999) and to impede participation in social associations and activities in Alesina and Ferrara (2000). The channel proposed or implied by most of the researchers is the negative effect of ethnic diversity on cooperation inside the country. Ethnic groups are being involved in the power struggle and engage in redistribution based on ethnic favouritism and not the national benefit.

The effect of ethnicity is contingent on political institutions. Easterly (1999) has shown that political institutions have power to mitigate the negative effect of diversity on economic development. Essential features of ethnic political institutions are exclusion and inclusion. Some groups are involved in politics, while others are denied access or discriminated against. Birnir and Waguespack (2011) have already suggested that inclusion may be an important factor in domestic politics by showing that cabinet diversity partly offsets the negative effect of ethnic fractionalisation on economic growth.

The limitation of the previous studies is the overwhelming focus on domestic politics. Foreign trade has been studied from the point of a single ethnicity dispersed around the world from Maghribi traders in Greif (1993) to Chinese bamboo networks in Rauch and Trindade (2002), but the effect of ethnic diversity inside each state on its foreign trade policy is understudied. To my knowledge, the only two effects found in the literature are the positive correlation between ethnic diversity and trade of high- and middle-income countries in Ahmad and Amin (2019) and lower levels of trade liberalisation in countries with excluded capital dominant ethnic groups in Monroe (2017).

In the same time, there is an elaborate literature on the effect of domestic politics on foreign policy. The logic of two-level games by Putnam (1988) has introduced the notion that every international agreement should be ratified by domestic actors. Mansfield et al. (2000) and Mansfield et al. (2002) have shown that the legislative ratification process in democracies makes democracies trade more with other democracies than with autocracies and that democratic leaders are interested in international agreements to overcome the commitment problem at home. Dutt and Mitra (2002) have demonstrated that leftwing governments adopt more protectionist policies in capital rich economies and free- trade policies in labour-rich economies and Baldwin and Magee (2000) have found that congressional voting on trade liberalisation is affected by previous campaign contributions.

Since Max Weber (1978) ethnicity is thought to be not an entity in itself, but a characteristic that facilitates a formation of a political group. Political science literature has a range of findings that show that ethnicity is manipulated and used by elites (see Fearon and Laitin (2000), Humphreys and Weinstein (2008), Wilkinson (2004)) as a tool of political mobilisation. If ethnicity is a strong force of domestic political mobilisation and domestic politics has an effect on international relations ethnic diversity should have an effect on foreign policy.

2. Mechanism

2.1 Cases

In the domain of international trade ethnic politics also plays out through maximisation of sectarian benefit and ethnic competition. Foreign trade does not reach its full potential naturally, but demands consistent pro-trade policies and matching buyers and suppliers in the home country with potential exporters and importers abroad. Both of these conditions are mostly realised as a public good. The central government negotiates tariffs and often promotes domestic firms through trade agreements and informal economic cooperation venues such as forums. Another possible requirements in the domain of development are trade infrastructure such as ports and homogeneous legal environment.

Historical cases point out that ethnic politics often leads to inconsistent policy and breakdown of foreign trade. In the seminal paper on political institutions, ethnic diversity and economic development Easterly and Levine (1997) showed how ethnic politics essentially killed the export of Cocoa from Ghana. Ashanti ethnic group which controls most of Cocoa production ruled Ghana from 1969 to 1971. The Ashanti government promoted export by supporting farmers and devaluing the currency. However, the new government controlled by Akan ethnic group which got power in 1971 was not interested in Ashanti farmers and distributed rents to its ethnic supporters involved in reselling imported products by overturning the economic policy and inflating the official exchange rate. As a result the Cocoa export have fallen from 19% of GDP in 1955 to 3% of GDP in 1983.

Trade is also constrained specifically to harm the development of the excluded ethnic group. Studying ethnic politics in Nigeria in the early 1980s, Ilorah (2009) writes that: “the government had banned all importation of stockfish for no apparent economic or health reason other than to punish the Igbo, who had just been defeated in a civil war and for whom trade in stockfish was an important mainstay at the time”. Other studies of post-Civil War policies in Nigeria such as Uzoigwe and Nwachuku (2004) confirm that the ban on the importation of stockfish and second-hand clothes was introduced by the Yoruba ethnic group specifically to harm Igbo's economic progress and were “tantamount to an effort at economic strangulation” of Igbos.

The case of Nigeria is especially revealing as ethnic politics played out not only in the formal trade policy but also in the illicit sphere. One of the main illicit channels through which ethnic groups may impede international trade is simple corruption and ethnic favouritism. Ilorah (2009) writes that in the 1980s members of Nigerian ruling elite (most of them co-ethnics) deliberately delayed products at the Lagos port to push prices higher and reap bigger benefits. Friends and relatives from the same ethnic groups as politicians were also often prioritised in granting import and export licenses.

The United Nations (1989) study of Palestinian trade under Israeli occupation shows the similar dynamic. Israel has prevented Palestinian exporters from using trade infrastructure such as bridges and ports and issued only a limited number of export and import licenses to protect Israeli producers from competition. The Paris Protocol on Economic Relations signed as a part of the Oslo accords further limited the ability of Palestinians to develop independent trade policy: there were limits on products which might be imported to Palestine from countries other than Israel and all Palestinian trade had to pass through the Israeli customs system to which Palestinian traders had limited access (see Khalidi and Taghdisi-Rad (2009)). By combining direct policies constraining Palestinian trade and informal barriers such as very long and inefficient licensing process, Israel significantly diminished Palestinian foreign trade.

Summing up, exclusion of one or more ethnic groups results in its preferences being unaccounted for by the government. Historical cases reviewed above range from simple corruption to neglect reaching the level of conscious harm. There is a space for sectarian politics as trade is contingent on specific policies implemented by the government and therefore the governing group can use it for discrimination and competition for power.

2.2 Theory

For the sake of clarity, the public action in trade development susceptible to sectarian politics can be divided in internal and external. Internal is setting the policy itself -- establishing tariff levels, import and export quotas and implementing the bureaucratic controls. The cases reviewed above were mostly touching upon the internal setting of foreign trade policy. However, the theory of international trade also points out the external mechanism -- domestic producers and consumers have to be matched with foreign firms and customers. This matching procedures mostly concentrated on providing information about domestic firms to foreign counterparts are often conducted by the government and therefore create the space for ethnic favouritism and exclusionary politics.

Internally, the negative effect is a result of the governing group having incentives to prioritise sectarian over national benefit or specific motivation to harm the development of other groups. Such strategies have long been studied in the domain of domestic politics. As the example of prioritising the benefit of the governing group over the benefit of the country Burgess et al. (2015) find that in Kenya regions populated by the ethnic group of the president receive twice as much investment in road building and have four times more roads built than other regions. Importantly, this effect completely disappears in the periods of democracy.

As foreign trade involves a distributive conflict -- groups control different resources of production and experience different gains or losses from different patterns of trade liberalisation -- ethnic favouritism will be present in foreign trade policy. The main harm of ethnic favouritism is invoked by suboptimal economic policy driven not by economic but by ethnic motivation. The related effect is the inconsistency of foreign trade policy. If decisions on tariffs, trading infrastructure investments and legal barriers change with the governing ethnic group, foreign partners face difficulties in building long-term trading relationship with the country.

Policy inconsistency was shown by Aqeel and Nishat (2004) to decrease foreign direct investment. As Carballo et al. (2018) have demonstrated that starting a trading relationship is a type of investment in the country -- the firm has to bear sunk costs of establishing supply chains, passing the regulatory barrier and making contacts with new customers or suppliers -- the effect of policy inconsistency on trade is likely to be similarly negative.

Externally, the main problem is informational asymmetry. The preferences for import and export on the ground are not known by foreign businesses. The supply of export and demand for import is often promoted by local and central governments, however excluded groups have no institutions of representation and are not on the agenda of the central government. The information about their trade preferences does not reach trade partners, they engage in less trade than their counterparts and the trading potential of the country is not fulfilled.

In the seminal paper Rauch (1996) introduced the idea of foreign trade as a network in contrast to market. Rauch wrote that the process of searching for trading partners is costly and firms do not engage in it in an optimal way. The severe constraints on information result in trade being arranged through family and ethnic ties, government- sponsored trade missions and diversified trading companies such as Japan's sogo shosha. In the following paper Rauch and Trindade (2002) have quantitively shown that in SouthEast Asia Chinese ethnic networks boost trade exactly by providing information and solving information asymmetry.

Foreign partners cannot observe the preferences of ethnic groups on the ground. The preferences are usually expressed through media, analytics or public diplomacy. However, these means favour only formally represented groups. For example, the governor of the region or the head of ethnic business organisation is covered by newspapers and makes speeches on different forums. In contrast, when the group is discriminated -- its representatives are absent from the mainstream coverage and it has less power to engage in the matching process. As a result, a significant potential for trade is lost. As Kincaid (2002) writes, there is a need in not national but constituent diplomacy, because when the world is the global marketplace local governments compete for foreign economic ties with their counterparts at home and abroad.

Summing up, ethnic politics has negative effects on foreign trade either through suboptimal foreign trade policies or through the lack of information about the excluded groups. Both policy and trade promotion are provided by the government and therefore are determined by the interests of the ethnic group in power. As a result, when sectarian politics is salient, the country is likely to experience harmful trade policies, deficit of information on the trading preferences of excluded groups and consequently lower levels of foreign trade.

3. Empirical test

The empirical study is conducted in four steps. First, I test the general hypothesis on the effects of the number of excluded ethnic groups and the size of excluded population on trade. Second, I survey the regional heterogeneity of effects. Third, I study the effects of structural factors which cause ethnic competition such as unranked ethnic systems and power-sharing regimes. Fourth, I make a robustness check of my ideas by showing that the factional component of Polity IV index which captures ethnic political competition is negatively associated with trade. Robustness check also leads to an important and striking result that the factional component coded by Polity IV as democratic biases the correlation between democracy and trade downward.

3.1 The main relationship

HI: not represented ethnic groups decrease foreign trade

Animosity between the governing and the excluded group exacerbates the problem of divergence in state's and group's interests. When there are historic ethnic tensions between two groups or other types of ethnic conflict such as active discrimination of one group by another, the sectarian identities start to play out. Economic theory of identity often defines identity as a negative externality (i.e. Akerlof and Kranton (2000)). When one ethnic group flourishes, another experiences negative costs from it. As a result, the government will not be interested in promoting trade in the region of the ethnic group and the ethnic group will find it shameful or offensive to cooperate with the government.

Moreover, in the countries with high levels of ethnic tension, the governing group has strategic incentives to block the economic development of actively discriminated groups as loosing power is likely to mean for the governing group not just becoming powerless but becoming actively discriminated itself.

H2: animosity between the government and excluded group further decreases foreign trade

When the excluded and the governing group have high level of animosity, not only the government is interested in harming the excluded group, but the excluded group is interested at least not to cooperate with the government and at most to directly harm the governing group. However, different ethnic groups have different capacity to influence the state's trade policy. Capacity is the group's ability to exercise political action by imposing regulation, mobilising the collective action in the region and having financial resources independent from the central government. Groups with capacity are mostly separatist groups which self-excluded from the government.

Groups with high capacity (1) can have a sizeable impact on the central state's foreign trade policy and (2) are less susceptible to state's coercion than other groups. For example, separatist groups can easily diminish trade flows by blocking vital infrastructure. If the government imposes economic sanctions by, for example, blocking the transfers to the region -- groups with high capacity can live through these sanctions as they engage in their own formal or informal taxation.

H3: capacity of the not represented group in animosity with the government further decreases foreign trade

Figure 1: Each hypothesis narrows the subset.

To put the theory into data, representation, animosity and capacity are studied through Ethnic Power Relations data (Vogt et al. (2015)) which codes the access to executive power by different politically relevant ethnic groups. The ethno-political characteristics of the country-year of exporter and importer are added as independent variables to the gravity equation of international trade. The data allows to use both the number of excluded groups and the size of excluded population as explanatory variables.

Among excluded groups, Ethnic Power Relations data distinguishes three categories of access to power. Powerless groups which are simply excluded from power, discriminated groups which are subject to active discrimination and exclusion and self-excluded groups which are essentially separatist groups that self-declare to be independent or excluded from the national government.

Powerless groups are not represented, however have no conflict with the governing group and no significant capacity to defect. The exemplary group are Asians in the United Kingdom. Discriminated groups are not represented, in the state of animosity with the governing group and have low capacity. The example of discriminated population are African-Americans in the US before the civil rights era. Separatist groups are not represented, in conflict with the government and have capacity to implement regulation on their territory. The example of separatist population are Abkhazians in Georgia.

I use two count variables from the Ethnic Power Relations data for both importer and exporter countries. Groups accounts for the number of politically relevant ethnic groups in the country-year which is different from the whole number of groups in the country as the main interest are the groups taking an active part in the political and economic life. Notrepresented is the number of powerless, discriminated and separatist groups in the country-year which are alternatively defined as excluded groups.

I also construct two count variables which allow to study separate effects of animosity and capacity. Notrepxanimosity is the number of discriminated and separatist groups in the country-year. Notrepxanimosityxcapacity is the number of separatist groups in the country year. Similar count variables from EPR (i.e. the number of excluded groups) are widely used both in political science and economics (for examples look Vogt (2018); Bluhm and Thomsson (2020)).

These variables do not account for the population of different groups. This allows to discuss the power of ethnic group agency without attachment to population. Population of the ethnic group is not the robust proxy of its power. Ethnic groups can have a significant impact even if small. Furthermore, as all the studied groups are ethnopolitically relevant it already means that they engage in political and economic life. Population also cannot serve as a proxy for the size of the market as it does not account for the purchasing and production powers of the group.

The effect of population is addressed in table 3, in which I use similar four measures of not represented population. Notreppop is the share of powerless, discriminated and separatist population in the country. Notrepxanimpop is the share of discriminated and separatist population. Notrepxanimxcappop is the share of separatist population. All measures can potentially vary from 0 to 1. This allows to check the effect of population and also serves as a robustness check of the initial study. All measures of the not represented population are also taken from the Ethnic Power Relations data.

This approach goes in contrast to using the widely used index of Ethno-Linguisitc Fractionalisation (ELF) (i.e. Alesina et al. (2003)). Even though ELF gives a good picture of diversity in the country, it faces difficulties in capturing ethnic politics. ELF does not show the number of groups that participate in the power contest and does not allow to distinguish between the size of governing and excluded populations. For example, ELF does not highlight highly consequential political arrangements in states such as Iraq during the rule of Saddam Husein, Syria during the rule of Assads and Rwanda before the genocide of 1994 when ethnic minorities were ruling over ethnic majorities. In contrast, both the number of excluded groups and the size of excluded population capture these dynamics.

My main dependent variable is bilateral trade flow. I use CEPII TRADHIST dataset from Fouquin and Hugot (2016), which aggregates export flows from different sources I focus on bilateral trade and not on overall trade of the country for two reasons. First, country's trade is a function of specific bilateral ties such as colonial past, common language and geographical distance to each trading partner. These ties cannot be captured in the study of total trade. Second, bilateral trade can be approximated with gravity equation, which has been used, checked and improved in the literature since 1954, therefore the empirical approach can be trusted to be robust and secure from omitted variables (i.e. Deardorff (1998); Bergstrand (1985)). However, I still address total export and import as an illustration and robustness check in table 3.

I include a range of controls from CEPII database on trade history. These controls include all the main variables proven to be significant in the gravity equation, including GDP, population and participation in GATT or WTO. Variables as distance, common language and colonial ties are not included in the models as all the models include country- pair fixed effects which allow to control for all time-invariant characteristics of the dyad. As ethnic diversity is often associated with political violence and low quality of political institutions, I also include for both countries the index of democracy polity2 from Polity IV index by Gurr and Marshall (2020) and the binary conflict dummy taken from the EPR data.

The panel is unbalanced and all models cover 161 countries and years from 1948 to 2014.

The unit of analysis in models 1-12 is a country-dyad-year. Regression models in tables 1 and 2 are based on the gravity equation which in the most basic form says that GDP_i ¦ GDP.j,

Exij = Jя~t where Ex. is export from country i to country j and Dist. is a

distance between two countries. All other factors such as colonial ties, common language and trade unions are added to the equation (examples of using all controls is in Glick and Rose (2002)). In the table 1 and models 1 to 6 the main explanatory variable is the number of not-represented groups. In the table 2 and models 7 to 12 the main explanatory variable is the share of not-represented population.

Most of the models are based on the Poisson Pseudo Maximum Likelihood (PPML) estimation which is proven to be the the most robust to heteroskedasticity and which accounts for zero trade flows between countries (see Silva and Tenreyro (2006). However, in the models 7 and 11-12 I replicate my findings with the standard logged OLS model. As main explanatory variable varies on a country level, in all models standard errors are clustered three-way -- by exporter, importer and year. All models include country-pair (one sided) fixed effects. Most of the models include continentxyear fixed effects which control for both time and regional characteristics, however model 6 replicates the results without them and just uses year fixed effects instead.

After studying the correlation between ethnic exclusion and trade on a country-pair level, models 17-22 provide simple logged OLS regressions of country's total export and import on ethnic exclusion variables. Main explanatory variable is the number of excluded groups. Control variables are GDP, population, participation in WTO, polity2 and the presence of conflict. All models include country and continentxyear fixed effects, standard errors are clustered two-way by country and year. Even though trade is usually studied through gravity equation and not through total measures, these models show that the relationship between trade and ethnic exclusion is significant and robust even on the level of total trade.

Standard errors in parentheses and clustered three-way by exporter, importer and year Controls for population size and particiaption in GATT/WTO are included in all models * p < 0.05, ** p < 0.01, *** p < 0.001

**** All year fixed effects are absorbed by continentxyear fixed effects

Standard errors in parentheses and clustered three-way by exporter, importer and year * p <0.05, ** p <0.01, *** p <0.001

Standard errors clustered two-way by country and year Standard errors in parentheses * p <0.05, ** p <0.01, *** p <0.001

The study shows that only animosity between the governing and excluded groups has a significant negative effect both on import and export. It is possible to conclude that all the negative effect of ethnic diversity on trade comes from the governing group directly harming groups with which it has tensions or conflicted relationship. The absence of access to power per se or group capacity have no negative effects. This finding is supported by all models that use the number of groups as an explanatory variable. The effect is negative in all models based on the size of population, however it reaches statistical significance only in the OLS models. On a substantial level additional discriminated ethnic group decreases both export and import by around 2-4% depending on the model which is comparable to the effect of decreasing GDP by around 2.3 - 4.6%.

Therefore, the effect of animosity (H2) can be supported with the high degree of confidence when the number of groups is the object of study, however not if the size of population is studied. Using all measures it is not possible to reject null hypothesis that simple non-representation (H1) and group capacity (H2) have no effect on trade. Moreover, broad non-representation has a significant positive relationship with trade in all PPML models.

Interpreting the results, it is important to account for the presence of GDP as a control variable. Therefore, all significant effects of discrimination on foreign trade are effects additional to economic development. It is already established that ethnic diversity has an adverse effect on overall economic growth (i.e. Easterly and Levine (1997)). However, my models show that even controlling for this effect there is a further negative effect on trade. This indicates that the nature of trade is different from broad economic development -- it demands more coordination and more public action and is therefore more susceptible to be harmed by ethnic politics.

In all models the coefficient on the number of groups is positive for both import and export, however the coefficient becomes significant only in the OLS and not in the PPML models. This broadly supports ideas already present in the literature that ethnic groups can increase trade through contacts and efficient matching.

Models 1-6 and 13-18 which use the complete number of not represented groups as the main explanatory variable show that the absence of representation has no negative and even significantly positive effect on trade. Models 7-12 based on the size of population also show no significant negative effect of not represented population. This is likely caused by the fact that notrepresented includes a wide array of powerless groups which are not actively discriminated against and even in some cases have local representatives that lobby the interests of the group.

The effect of animosity between the governing and not-represented ethnic group is negative and significant for all models based on the number of groups. This effect holds both for dyadic models and the study of total trade on a country-level and is robust for the inclusion and exclusion of fixed effects. In the models based on the size of population the effect of animosity is similarly negative and significant in OLS models and negative albeit not significant in the PPML models..

The absence of statistical significance in the models based on the size of population possibly indicates that animosity matters as a group-...