The determinants of intra-regional trade in the Western Balkans *1

Abstract

The purpose of this paper is to identify factors that have an influence on bilateral trade among the Western Balkan countries for the period from 1995 to 2012. Unlike a majority of works that have dealt with the subject of trade in the Balkans or the SEEs, not only geographical, economic or political factors are taken into account, but also factors constituting cultural, communicational and historical types of the so-called "distance" between countries. In order to assess their influence on trade values, an augmented version of the gravity model is employed. It was estimated threefold: as pooled data by OLS, as a random effects model and as a fixed effects model with an additional estimation of time-invariant variables following the method of Cheng and Wall (2005). The results of the research are surprising, as the strongest influence on trade values were exhibited by variables representing ease of a direct communication and similarity of religious structures. In addition, war and one-year-post-war effect showed a strong and statistically important influence. Thus, the main conclusion is that non-economic factors in the region of the Western Balkans play the most important role in determining trade values between countries.

Headnote

Abstract

The purpose of this paper is to identify factors that have an influence on bilateral trade among the Western Balkan countries for the period from 1995 to 2012. Unlike a majority of works that have dealt with the subject of trade in the Balkans or the SEEs, not only geographical, economic or political factors are taken into account, but also factors constituting cultural, communicational and historical types of the so-called "distance" between countries. In order to assess their influence on trade values, an augmented version of the gravity model is employed. It was estimated threefold: as pooled data by OLS, as a random effects model and as a fixed effects model with an additional estimation of time-invariant variables following the method of Cheng and Wall (2005). The results of the research are surprising, as the strongest influence on trade values were exhibited by variables representing ease of a direct communication and similarity of religious structures. In addition, war and one-year-post-war effect showed a strong and statistically important influence. Thus, the main conclusion is that non-economic factors in the region of the Western Balkans play the most important role in determining trade values between countries.

Key words: trade determinants, Western Balkans, gravity model, panel data, distance JEL classification: F14, F15, C23

Sazetak

Cilj ovog rada je identificirati cimbenike koji imaju utjecaj na bilateralnu trgovinu izmeðu zemalja Zapadnog Balkana u razdoblju od 1995. - 2012. godine. Za razliku od vecine radova koji se bave problematikom meðusobne trgovine zemalja Balkana ili zemalja Jugoistocne Europe, osim zemljopisnih, ekonomskih ili politickih cimbenika u radu su uzeti u obzir i cimbenici koji se odnose na kulturoloske, komunikacijske i povijesne tipove varijabli, a koji cine takozvanu "udaljenost" izmeðu zemalja. Kako bi se procijenio njihov utjecaj na trgovinu, koristen je prosireni gravitacijski model koji je ocijenjen trostruko: kao model grupiranih OLS podataka, kao model slucajnih efekata i kao model fiksnih efekata s dodatnom procjenom vremenski nepromjenljivih varijabli u skladu s metodom u Cheng i Wall (2005). Rezultati istrazivanja su iznenaðujuci, s obzirom da su najjaci utjecaj na vrijednosti trgovine imale varijable koje predstavljaju jednostavnost izravne komunikacije i religijske slicnosti. Takoðer, varijabla rata i varijabla jednogodisnjeg post-ratnog efekta pokazali su snazan i statisticki znacajan utjecaj. Dakle, glavni je zakljucak da u regiji Zapadnog Balkana ne-ekonomski faktori igraju najvazniju ulogu u odreðivanju vrijednosti trgovine izmeðu zemalja.

Kljucne rijeci: odrednice trgovine, Zapadni Balkan, gravitacijski model, panel podaci

JEL klasifikacija: F14, F15, C23

(ProQuest: ... denotes formulae omitted.)

1. Introduction

The problem of finding factors that determine trade between countries has been constantly present in economic literature since Adam Smith's theory of the absolute advantage. However, just very recently authors tended to investigate trade determinants with a wider scope of potential answers, than just geographical, economic or political factors, which constituted typical explanations so far. Specifically those "other" issues may play a prominent role in the analyzed region of the Western Balkans (further referred to as WB).

Firstly, the region is situated in the joint of three big religions: Catholicism, Orthodoxy and Islam. This brings upon a question, if faith matters in trade relations or, as it was put by Helble (2007), "Is God good for trade?" (for related analysis see Guiso et al., 2004, Lewer and van der Berg 2007, Melitz and Toubal 2012). Secondly, spoken languages in the WB region show a high level of similarity (except Albanian)4. According to recent studies on language similarity and trade, the relation should be very strong, especially when we look at languages as means of communication (different measures of the language similarity are presented in Boisso and Ferrantino 1997, Fearon 2003, Guiso et al. 2004, Laitin 2000, Melitzt 2008, Melitzt and Toubal 2012). Thirdly, in the 1990s the Balkan countries suffered wars and conflicts which could have a distorting influence on the natural trade pattern among the economies. This issue seems to be rather neglected in the empirical research concerning trade (for exceptions see Guiso et al. 2004, Martin et al. 2005, Melitz and Toubal 2012, Sarkees and Wayman, 2010). The relatively small region of the Western Balkans, which is inhabited by ca. 22 million people, is so diversified internally, that it condenses many problems that may distort natural trade patterns not only there, but worldwide.

As for the object of the research, the Western Balkans is a political, geopolitical and economic term used in terminology of the European Union since the early 1990s to describe countries of the former Yugoslavia5 minus Slovenia plus Albania6. Our research thus deals with the analysis of bilateral trade among five/six countries: Albania, Bosnia and Herzegovina, Croatia, Macedonia and Serbia and Montenegro (as one country until 2007 or as two entities afterwards). Although the Western Balkans does not represent a significant trading partner for the European Union, the largest trading partner of all countries of the Western Balkans is the EU, constituting nearly two-thirds of the total trade of the region (European Commission, 2014). EU granted a free trade regime to all countries of WB, while the reciprocal elimination of barriers by WB countries on imports from the EU has been gradual and in the framework of the Stabilization and Association Processes. The European Union concluded Stabilization and Association Agreements with Macedonia and Croatia in 2001, Albania in 2005 and Bosnia and Herzegovina, Montenegro and Serbia in 2008. This relatively late involvement in the EU-integration of the Western Balkan countries might have caused stronger intra-regional trade relationships. As Association Agreements required that the exports to the EU would contain minimum levels of input originating from the EU-members or from the countries that signed the Agreements, the WB countries could have been left behind not only as a potential exporter of final goods to the EU, but also as a supplier of inputs for companies from the Central and Eastern Europe, which had signed the Association Agreements much earlier (see Stojcic (2012) and Tkalec and Vizek (2014) for the case of Croatia).

In December 2006, WB countries signed CEFTA-2006 (Central European Free Trade Agreement)7, which replaced a network of 32 bilateral free trade agreements between them. All tariff barriers were thus removed while removal of non-tariff barriers is carried out in the framework of CEFTA-20068. By acquiring a full membership in the EU in July 2013, Croatia has emerged from new CEFTA. Trade characteristics of the WB countries show that they can be regarded as small open economies with high share of trade in gross domestic product. Over last three years of our analysis, between 2010 and 2012, Albania had an average trade to GDP ratio of 85.9, Bosnia and Herzegovina 97.8, Croatia 82.2, Macedonia 121.9, Montenegro 103.7 and Serbia 92.59.

All countries of the Western Balkans are regarded as countries in the process of transition (EBRD, 2013) but there are significant differences in the economic welfare among them. In 2000, Albania was the poorest country of the region with GDP per capita measured by purchasing power parity (PPP) of 3,846 international US dollars while in the same year Croatia had 11,148 US dollars of GDP p.c. In 2012, the poorest country by GDP p.c. was Bosnia and Herzegovina with 8,047 international US dollars, whereas Croatia as the leader of the region had the highest GDP p.c. of 18,10110 US dollars.

Taking into account all the so far presented internal diversity of the Western Balkan region, the main hypothesis that was tested in our research stated that there were more than just geographical, economic and political factors that influenced trade among WB countries over the time period from 1995-2012. Auxiliary hypotheses were appointed to certain types of "distances" (communicational, cultural, political and historical) and their specific impact on trade values11.

The rest of the paper is structured as follows. In the second part the existing literature on trade in Western Balkans and South East Europe is reviewed, especially positions that explore the problems with use of the gravity model. In the third part, the methodology of the current research is presented. It is followed by the fourth part which encompasses data sources and analysis. Results and discussion are presented in the fifth part, whereas conclusions are drawn in the sixth and the closing one.

2. Literature review

In general, analysis of trade in the South East Europe or in the Western Balkans was in focus of many researchers and scholars especially in the context of regional cooperation and free trade arrangements (World Bank, 2008, EBRD 2003, Mehic and Babic-Hodovic, 2011, Nikolic et al., 2011, Botric, 2012). On the other hand, a majority of authors devote their attention to the research of bilateral trade of one (their) country with the rest of the world (Jovicic et al., 2001, Miljovski and Uzunov, 2002, Bilas, 2007, Mojsoska-Blazevski and Petreski, 2010, Druzic et al., 2011, Cejvanovic and Dzafic, 2011, Bjelic and Dragutinovic Mitrovic, 2012, Fetahu, 2014) most likely because integration processes in South East Europe, in contrast to Central Europe, lasted several years, leaving researchers with the problem of how to define a region or group of countries they want to explore. There is a lack of literature which would analyse trade among a group of countries constituting the WB region, especially one that would be based on the gravity model of trade (Begovic (2011) analysed the CEFTA-2006 countries). Although in the last ten years, this topic is becoming more important and more interesting for a number of scholars, not only from this region.

As for the purpose of studies, a significant number of authors were trying to assess the potential trade of a country or the region (see Miljovski and Uzunov, 2002, Christie, 2002, Kaminski and de la Rocha, 2003, EBRD, 2003, Bussière et al. 2005; Kernohan, 2006, Damijen et al., 2006, Kucharcuková et al., 2010, Pllaha, 2012). However, many of them aimed at finding out specific determinants of trade, most often focusing on political issues of trade liberalization or regional integration (Adam et al., 2003, Herderschee and Qiao, 2007, Bjelic et al., 2013, TosevskaTrpcevska and Tevdovski, 2014). Other possible trade determinants, like common history, religion or language were usually neglected.

Some of the first attempts to explain regional trade were made by Christie (2002). By using the classical gravity model, the author came to the conclusion that in the context of intraregional trade, South East Europe - 7 and South East Europe - 11 cannot be considered as a region, while division in three sub-regions better fits the classical model. This was probably due to the time period of analysis as there were still strong negative effects of the recent war, especially in trade between FR Yugoslavia (Serbia and Montenegro) and Croatia and between Bosnia and Herzegovina's trade with both - Croatia and FR Yugoslavia. Kaminski and de la Rocha (2003: 41) focusing on the future of regional trade liberalization among SEE-5 (Albania, Bosnia and Herzegovina, Croatia, Serbia and Macedonia) showed similar results. Using a very simple gravity model due to the poor quality of data, they found out that part of the intraregional trade was below potential (trade between FR Yugoslavia and Croatia) and part of it was above (between Bosnia and Herzegovina and both FR Yugoslavia and Croatia) which suggested a division of the region into sorts of dyads and enabled them to better predict intraregional trade by means of the gravity model. An important conclusion drawn from both researches (Chrisite, 2002, and Kaminski and de la Rocha, 2003) is that recent history of the region, i.e. dissolution of ex-Yugoslavia as well as very specific and similar cultural ties among these countries, played an important role in determining potential trade among the region.

The EBRD (2003) study employed a more complex gravity model to compare the actual level of trade with the predicted level between seven SEE countries in transition. The baseline gravity model which included only country's GDP in PPP, the distance between capitals and a measure of bilateral exchange rate volatility - showed that the actual trade is around 25% of the predicted trade in the SEE. When some other determinants of trade were included - the size of a country, the number of borders a country has to cross to get to a target market, the quality of transport infrastructure, the extent of trade liberalization and the quality of institutions - the actual trade was around 42% of its potential indicating still that there are some other region-specific factors that should be included in order to better explain intraregional trade.

Instead of using cross section data as Christie (2002) and Kaminski and de la Rocha (2003), Damijen et al. (2006) employed the panel-data approach in the gravity model to explore trade potential in SEE intraregional trade and with the EU for the period between 1994 and 2002. Using the model augmented with several additional variables - exchange rate volatility and a set of dummy variables for six different country groups, authors, in respect to Western Balkans or SEE-5, found out that the actual trade between the WB countries is close to its potential, thus indicating that future liberalization would not have a significant impact on trade flows. However, the actual trade between ex-Yugoslav countries was far above potential trade which was predicted by the gravity model, indicating again that some other factors should be included in order to explain trade among ex-Yugoslav countries.

Inspired by a World Bank study from 2005, which confirmed below-potential trade among Western Balkan countries and importance of regional trade integration and co-operation, Kernohan (2006) analyzed trade potential among SEE-9 (Albania, Croatia, BiH, Macedonia, Serbia, Montenegro, Romania, Bulgraia and Turkey) with the gravity approach. His results suggests that the SEE region could be regarded as two blocks - one block encompasses ex-Yugoslav countries whose trade is far above its trade potential while another block of non-ex-Yugoslav countries is not integrated enough with the rest of the region, especially Turkey. This is in line with Pjerotic (2008: 511) who concluded that the significance of intra-SEE trade varies considerably across countries.

Using the enhanced gravity model, Bussière et al. (2005) tried to assess the future integration of the Central and South Eastern European economies with the EURO area based on bilateral trade flows from 1980 to 2003. Besides standard determinants such as economic size and geographical distance, several additional variables were used: common language, common border, being a part of the same territory, participation in a free trade agreement and valuation effects of exchange rates. According to Bussière et al. (2005: 29-30) countries of the Western Balkans - Albania, Bosnia and Herzegovina and Macedonia, and to a lesser degree Croatia, still have significant space to raise their trade intensity with the euro area. Moreover, these countries show a low degree of trade integration with the world economy. Similarly to Busier et al. (2005), Herderschee and Qiao (2007) proved by a simple gravity model that CEFTA-2006 would in future enhance integration of the region with the EU and inside the region itself, suggesting a rise of intraregional trade in CEFTA-2006 by 38-49 per cent.

According to Kucharcuková et al. (2010) who used the augmented gravity model with several additional variables, the below-potential international trade of SEE countries given their income levels and geographical location could be explained by the low quality of economic institutions and below-potential regional cooperation. Furthermore, authors suggest that the regional conflicts in the past could be an initial cause of bad institutions which is in line with Adam et al. (2003: 3) who blame the horrific conflicts for the small scope of the regional co-operation. Based on a gravity model with several variables - economic size, geographical distance, similarities in demand, regional cooperation, common border and common language in the case of BAFTA and CEFTA from 1996 to 2000, authors predicted a necessity of multilateral free trade agreement among SEE countries (Adam et al., 2003: 14). Nine years later, for nine SEE countries - Albania, Bosnia & Herzegovina, Bulgaria, Croatia, Macedonia, Romania, Serbia & Montenegro, Slovenia and Turkey, Pllaha (2012) confirmed again that most SEE-9 countries trade below their potentials. The gravity model based on the panel data approach was augmented with a dummy variable called a colonial link indicating belonging to the same territory in the case of ex-Yugoslav countries while the overall scope of the analysis was to determine the importance of bilateral trade agreement on trade flows. In the case of the colonial link, which approximated common history and cultural similarities, the model showed a positive influence of this variable on bilateral trade. The same effect was found for the free-trade-agreement dummy variable which was also statistically significant, thus indicating that the removal of trade barriers contributed to trade intensification among SEE-9 countries during the period 2002-2010 (Pllaha, 2012: 21).

Tosevska-Trpcevska and Tevdovski (2014) used an augmented gravity model primarily to asses the impact of customs and administrative procedures on trade among South East European countries in the period between 2008-2012 and found that sharing the same border, being a part of the ex-Yugoslavian market and sharing a common history were still important factors of trade, while customs and administrative procedures were important determinants of the volume of traded goods in the region but depended on how these variables were measured in the model. Similarly to the above authors, Bjelic et al. (2013) analyzed bilateral trade of the Western Balkans as a group and the rest of the world, focusing on trade with EU as the main trading partner of the WB countries. They found the border effect as a significant determinant of trade while signing of the Stabilization and Association Agreements was not. Regarding the variables they were focusing on - administrative and technical barriers, among the Western Balkans and between Western Balkans and EU, the results were mixed: technical barriers were an important impediment for the export of WB countries to the EU as well as administrative barriers, while the latter ones were not important factors of intraregional trade.

Related to our study, an analysis of the influence of cultural, historical, and political ties on patterns of trade in the SEE transitional countries based on a simple gravity model could be found in Gjipali et al. (2012). Authors showed that countries that share common border trade almost twice more than countries that were not, while countries with similar languages trade three times more, and countries that had been a member of ex-Yugoslavia trade even four times more. Similarly, an interesting but not well-grounded analysis of the impact of cultural ties between Balkans and Turkey was developed by Gumuscan and Kahveci (2012).

3. Methodology and method of analysis

The gravity model of trade has been present in economic literature since 1960's. The early concepts were presented by Tinbergen (1962), Pöyhönen (1963) and Linneman (1966), who employed in their analysis of trade assumptions similar to the Newtonian gravity concept. It stated that the gravitational force depends on the "masses" of entities and "distance" between them. The first attempts to justify the gravity equation with a theoretical model were made by Anderson (1979), who based on the Armington assumption that the products are differentiated by the country of origin. The approach of Anderson was followed by Bergstrand (1985) who derived the gravity equation from CES preferences and the Armingtondifferentiated goods. The New Trade Theory-justifications came with the works of Helpman (1981) and Helpman and Krugman (1985) who abandoned the neoclassical assumptions in favour of the product differentiation among firms and monopolistic competition. Monopolistic competition also formed the ground for the New Economic Geography, which introduced endogenous location of firms and demand and led to a conclusion that factor's movement and profitability caused reallocation of the production12.

In trade analysis, the common proxies of "mass" are country GDP and/or population, whereas "distance" used to be interpreted mostly as physical distance between capital cities or main economic regions (see Kandogan, 2009, Wydymus, 2012). One of the most popular versions of the gravity equation is presented in Equation (1).

... (1)

where:

EXPijt - export from country i to country j in year t

GDPit - country's i GDP in year t

GDPjt - country's j GDP in year t

POPit - country's i population in year t

POPjt - country's j population in year t

DISTij - great circle distance between capital cities of countries i and j

a0 - constant

a1 - a5 - coefficients

δijt - error term

Export, import or total trade can serve as dependent variables, although the latter one would not provide any information on whether determinants influence more export or more import between two countries. The model is usually expressed as a power function and further log-linearized, which allows the coefficients to be interpreted as elasticities.

Although the so-called standard gravity model was very directly referring to the Newtonian gravity concept, in the past forty years numerous variations of it were tested, in order to capture different trade determinants. This process showed clearly that "distance" in trade is something more than just a geographical one.

One of the earliest attempts to categorise and measure different forms of "distance" between countries showed up in non-English literature. Namely Zelias (1991) in his analysis of trade among ex-Soviet block countries used the following division of the "distance" types:

1. Physical (in kilometres)

2. Time (in the meaning of time needed to transport goods from one country to another)

3. Economic (eg. transport cost)

4. Social (eg. unemployment)

5. Political (binary variable indicating political preferences between countries, eg. "1" for the ex-Soviet block).

However, it should be noted, that the first three types of "distance" are de facto different measures of the same category - a physical distance. No matter if measured in miles or hours needed to overcome the distance or cost of this process, they still characterise a "physical" distance.

Over a decade after Zelias (1991), in their often cited works Anderson and van Wincoop (2001, 2004) started a broad discussion on the role of what they called the "cost" of trade. The starting point of their analysis is the so-called "border effect" described by McCallum (1995). Anderson and van Wincoop spread the research for a wider set of countries and presented their own division of "costs" or "distance":

1. Policy barriers

2. Transport costs

3. Wholesale and retail distribution costs.

Anderson and van Wincoop (2001, 2004) point out another type of "distance" related to a common language in both trading countries. Moreover, they characterise "distance" in total as price dispersion between trading countries, which should proxy all barriers to trade. On the other hand, they emphasise that the inclusion of a "multilateral trade resistance" variable in the model is needed in order to encompass relative barriers to trade.

Another milestone in the theory of a "distance" in trade are papers by Melitz (2008) and Melitz and Toubal (2012). Unlike most of the other studies, which stress the physical, economic or political barriers to trade, these authors draw our attention to cultural and communicational obstacles. Although many other previous studies raised the language issue, it was usually simplified as binary variable taking value of "1" when countries had the same official language. In the above mentioned works, Melitz and Melitz and Toubal present more complex measures of communicational possibilities among people in trading countries13.

In our analysis we divide the "cost" or "distance" between countries into six categories:

1. Physical, which refers to geographical position of one country in relation to another.

2. Political, which refers to effects of political decisions, e.g. WTO membership, membership in a FTA.

3. Economic, proxying diversity in development in trading countries.

4. Cultural, referring to cultural diversity of two countries.

5. Communicational, reflecting problems in effective, direct communication between citizens.

6. Historical, which shows how a common history determines trade relations.

The first three of the above types of "distance" are commonly used, whereas the latter three, especially cultural and communicational "distance", tend to gain more attention of the researchers mainly in more recent studies. By incorporating all six types in the gravity model we obtain Equation (2), where the first two vectors encompass variables which characterize "masses" of exporting and importing countries, whereas the following six vectors represent variables of six types of "distance": physical, political, economic, cultural, communicational and historical.

... (2)

where:

EXPijt - export from country i to country j in year t

xitβ1 - vector of variables representing "mass" of the exporting country (i)

mjtβ2 - vector of variables representing "mass" of the importing country (j)

dfijtβ3 - vector of variables representing physical "distance" between countries i and j

dpijtβ4 - vector of variables representing political "distance" between countries i and j

deijtβ5 - vector of variables representing economical "distance" between countries i and j

dkuijtβ6 - vector of variables representing cultural "distance" between countries i and j

dkoijtβ7 - vector of variables representing communicational "distance" between i and j

dhijtβ8 - vector of variables representing historical "distance" between countries i and j

a0 - constant

β1 - β8 - vectors of parameters

δijt - error term

In the case of two out of the eighteen explanatory variables used in the model we found it necessary to use new indexes that would be a proper approximation of phenomena they describe - similarity of religious structures of both of the analysed countries and ability of their societies to communicate directly. The first of them (variable RELIG), one of the two "cultural" determinants of trade in the Western Balkans, was an index constructed as a sum of differences of shares of three main religions represented in the region, which altogether accounted for at least 95% of population in each of the analysed countries. The reason for including this variable of structural religious similarity was the fact that there is religious diversity in those countries (see Table 1). Next to quite homogenous countries like catholic Croatia (91%) or orthodox Serbia and Montenegro (71%) and Macedonia (67%), there are states of more complex religious structures, like Albania or Bosnia.

Inclusion of differences in religious structures in the gravity models of trade could be seen in literature after the year 2000. In case of some of the papers this problem is addressed by including the sum of products of percentage of people in both countries who share the same religion (see Helble 2007, Guiso et al., 2004, Melitz and Toubal 2012). However, this measure brings low results for internally diversified countries, which is a significant shortcoming. On the other hand, Kandogan (2009) suggests the use of the Euclidean distance to measure religious distance between countries.

In this research, the applied measure (Equation 3) shows in what percentage the shares of certain religions in certain countries coincide14. It takes values between <0;1>, which is a desirable feature in a regression analysis where other variables (like LANG) share the same scope.

... (3)

where:

RELIGij - index of similarity of religious structures of countries i and j

Rik - share of people belonging to religion group k in country i

Rjk - share of people belonging to religion group k in country j

In this research, the use of n = 3 religions was quite obvious, as the sum of their shares together with a share of non-religious people accounted for at least 95% of countries' total populations. However, in other studies it remains as open question as to how to choose the number of religions investigated.

Unlike the issue of religion, the issue of languages that citizens of both trading countries speak was long used in the gravity models. In the majority of instances it was simplified to a binary variable taking the value of "1" when countries shared the same official language. How poor such measure would perform, shows an example of the Western Balkans, where its value for a country-pair Croatia and Serbia would be "0", though people in these countries can communicate easily in their mother tongues.

Similarly to the "religion" variable, more complex indicators showed up in literature after the year 2000. Firstly, authors began to include the heritage of the linguists researching actually spoken languages, not the official ones (see Guiso et al. 2004, Boisso and Ferrantino 1997, Melitz and Toubal 2012). Secondly, it was finally noticed that some languages are more similar to one another, than others. This can result in situations where people that speak different languages, but belonging to the same branch of the language tree could, to some extent, communicate. Thus, a variable representing similarity of the main languages in both trading countries was introduced by some authors. It was operationalised as the "distance" coming from the language tree (see Fearon, 2003, Laitin, 2000, Melitz and Toubal, 2012). The missing part was the connection of the actually spoken languages with their similarity. The solution came in the work of Klimczak (2014a), who constructed a matrix of percentage products of shares of people speaking each of the "n" languages in two chosen countries (see an example of the matrix for a country pair Albania-Bosnia in Appendix 1). In the second step, this matrix was multiplied by ratios reflecting the similarity of certain languages15 (see Appendix 2 for details), giving back the spoken languages similarity index of two countries (Equation 4).

... (4)

where:

LANGij - index of spoken languages similarity of country i and j

Sik - share of people speaking language k in country i

Sjl - share of people speaking language l in country j

simkl - similarity ratio of languages k, l

Values of RELIG and LANG indexes are presented in Appendix 3.

In order to control for unobserved country-specific factos, like eg. multilateral resistance or internal distance, dummy variables for the exporter country were added to the model.

As for the data and estimation techniques, in a majority of the studies before the 1990's, cross-sectional data was used and only more recent studies tend to estimate panel data (Baltagi et al. 2003). However, there is no one commonly accepted estimation technique. Thus in this paper the panel data model was estimated threefold:

- as pooled data model with OLS;

- as fixed effects model, with additional estimation of the time-invariant variables;

- as random effects model.

In order to decide upon the most appropriate model specification, three tests were run. At first, a test was run on differing group intercepts, in order to check if groups have a common intercept. Then the Breusch-Pagan test was run to verify a hypothesis, that variance of the unit-specific error = 0. At the end, the Hausman test was carried out to see if the GLS estimates used in the random effects model are consistent.

Various studies show that the model with fixed effects is the most appropriate for the estimation of gravity equations (Cheng and Wall, 2005, Baltagi et al., 2003, Bussière et al., 2005). An important obstacle in this case is the inability of the fixed effects models to estimate time-invariant variables. Cheng and Wall (2005, further referred to as "ChW") propose as a solution to this problem a two-stage estimation. In the first stage a fixed effects model is estimated. Then, in the second step all time-invariant variables are being regressed on country-pair fixed effects which come from the first stage estimation. This method was used i.a. by Bussière et al. (2005) who investigated processes of trade integration of the Middle and East European countries16.

4. Empirical data and analysis

Apart from the typical variables of the gravity model, GDP and population of both countries17 and physical distance between them, in the presented analysis a number of additional variables were added. Based on the trade theory, results of other research projects or regional specifics of the Western Balkans, all of them could have been presumed to have a substantial influence on trade values. A majority of the variables were specified in a way that has been already present in the literature. However, taking into account the extent to which the Western Balkan region is diversified, less known measures were also applied to proxy cultural, communicational and historical "distance" (see Klimczak 2014a).

Table 2. includes a description of certain variables representing "distance" which are tested in the model presented further. They are classified into groups according to the division of types of "distance".

Two measures were implemented in order to capture the importance of the physical (or geographical) "distance", both of them in the form most commonly used in the literature. DIST was supposed to provide information on great circle distance between capital cities of the analysed countries. On the other hand, BORDER was a binary variable taking the value of "1" in instances when countries share a common land border. As the region is geographically limited, both variables were expected not to have any greater influence on trade values.

Political "distance" variable FTA reflected a situation where countries had a Free Trade Agreement signed or both underwent a higher stage of the process of regional integration. In the case of the Western Balkans before 2013, this meant either bilateral trade agreements or joining CEFTA-2006. The issue of bilateral trade liberalisation comes into meaning, as two authors came up with two totally different conclusions for the WBs. Ilic (2012) proved that Serbia exports more to CEFTA-2006 countries than to other countries. On the other hand, Damijan et al. (2006) found little effect of the trade liberalisation in the region from 1996 - 2000, as well as a minor expected influence on trade of the planned signing of CEFTA-2006.

The economic "distance" was represented by a variable PERCAP_DIFF, a difference in GDP per capita, which proxied relative factor endowments (capital/ labour) of the countries. According to the New Trade Theory, countries that are similar in economic development (and factor endowments) tend to trade more with one another. Thus, a high value of the variable's coefficient would indicate H-O type of trade between countries.

Two variables representing cultural and communicational "distance", RELIG and LANG respectively, were described in the previous section.

A series of historical "distance" variables were included in the study. Unlike frequently used colonial links, not applicable to this particular research, more important seemed to be the effects of the wars and military conflicts which the region suffered in the 1990's. In the literature, there appear three ways the effects of historical conflicts are included in trade modelling:

- as mutual reluctance being an outcome of the even far away conflicts (Guiso et al. 2004; Melitz and Toubal 2012, Sarkees and Wayman, 2010);

- as a real barrier to trade connected with warfare in the current (analysed) year (Martin et al. 2005);

- as a real barrier to trade connected with warfare in the current (analysed) year or a number of years after the end of the war (Klimczak, 2014a).

In this research, the latter one was applied for two reasons. Firstly, it does not equate conflicts which happened hundreds of years ago with conflicts from the recent past. Secondly, it captures the effects of the post-war periods which, intuitively taken, still have a negative effect on business relations between the previously conflicted countries.

Apart from the distance-type variables, stock of the foreign direct investment in the exporting country (FDI) was added to the model (regardless of the investment's country of origin). It can proxy technological advancement of a transition economy. However, the theory does not give a clear hint, whether to expect rise or decrease of trade values as a consequence of foreign direct investments. They can substitute trade with certain goods previously exported from the investing country, but on the other hand they can boost exports of these goods to the third countries (Fontagné and Pajot, 1997). The impact much depends on the type of the FDI - horizontal investments may restrain trade, whereas vertical ones may cause its increase, but mostly with the investing country (Camarero and Tamarit, 2003). As for the SEE region, Mehic and Babic-Hodovic (2011) show positive and statistically important impact of the FDI on the SEEs exports. However, the latter study focuses on total exports, whereas our research is about intra-regional trade.

The dataset used in the analysis deserves an explanation. There was a lack of data for some of the trade flows, namely from Albania to Bosnia (years 1995 - 2000), from Bosnia to Albania (1995 - 1996) and from Croatia to Yugoslavia (1995). As these constituted just nine observations, they were simply not estimated. No zeroflaws occurred in the dataset, as we worked on the aggregated values, which on the other hand could have led to an aggregation bias. Further research is welcomed in order to see results for disaggregated flows. As for the other data limitations, the sources of religion and language data were usually limited to just one year and did not cover all countries of interest, so there was a necessity to transfer the data on religion and language from Serbia to Montenegro and to Serbia & Montenegro (or Federal Republic of Yugoslavia). Finally, the used dataset did not encompass Kosovo due to a complete lack of comparable data19.

5. Results and discussion

The results of the model which included additional variables proved to be well-fitted to data, with the adjusted R-squared values in pooled and fixed effects specifications 0.82 and 0.89 respectively20. Furthermore, only 0.07 pp difference between the two specifications (pooled and fixed) suggests a significant part of the bilateral effects (meaning "distance") to be covered by the proposed explanatory variables.

In order to evaluate which model specification (pooled, fixed or random effects) was best suited to the empirical data, a series of three tests was conducted. At first, a test was run on differing group intercepts. According to the null hypothesis, groups have a common intercept. The test statistics provided: F(35, 326) = 7.76 with p-value = P(F(35, 326) > 7.76) = 2.90067e-026. This brought a conclusion that the null hypothesis was rejected, which meant the fixed effects model was more appropriate then the pooled one. In the next step, the Breusch-Pagan test was run to verify a hypothesis, that variance of the unit-specific error = 0. Test results showed asymptotic test statistic: Chi-square = 147.566 with p-value = 5.90087e-034, proving that the random effects model suited data better than the pooled one.

At the end, the Hausman test was carried out to see if the fixed effects model was more appropriate than the random effects model. The null hypothesis suggested that the GLS estimates used in the random effects model are consistent. Asymptotic test statistic: Chi-square = 63.295 with p-value = 1.34284e-008 showed that the null hypothesis should be rejected and thus the fixed effects model would be more appropriate than the random effects one.

Regarding the interpretation of the variables which were suspected of having an influence in trade values, we would concentrate on the ones augmenting the standard model, because the influence of GDP, population and distance between main cities on trade between countries is widely known. The first additional variable, representing the physical "distance", was BORDER. This binary variable took value of "1" in the instances when the countries are adjacent. In the fixed effects model, as time-invariant, the variable had to be estimated using the ChW method, which brought a statistically important value of the coefficient of 5.17. For the other two models, it also showed statistical importance at p < 0.01 and a positive influence on export (1.09 in pooled and 1.13 in random effects model).

Being a representative of the economic "distance", the absolute value of the difference in GDP per capita of trading partners describes not only the difference in their welfare and in the stage of development, but also in factor endowments (labour/capital). As for the relatively poor countries, a big difference in factor endowments should stimulate inter-industry trade, which was empirically confirmed for Serbia's trade by Bjelic and Dragutinovic Mitrovic (2012). The regression results of this study were contradicting this thesis. Just in the pooled model was the variable statistically important. Its coefficient values were negative and close to zero (-0.24; -0.06 and -0.09 for pooled, fixed and random models respectively). Comparable value of this variable's co-efficient were estimated for the CEFTA-2006 countries in the model of Begovic (2011).

The variable constituting the cultural "distance", similarity of religious structures (RELIG), in the fixed effects model, was estimated with the ChW method and the value of its coefficient was 9.19. In the pooled model this variable appeared to have a positive (0.66), though statistically not important influence on the value of export. However, in the random effects model, this influence was stronger (1.94) and statistically important by p < 0.1. These results can be interpreted in a way that similarity in religious structures is one of the key determinants of trade values and directions in the region of the Western Balkans21. As previously mentioned, the region lies on the joint of the three big religions: Catholicism, Orthodoxy and Islam, so the high value of the coefficient standing by the RELIG variable indicates problems which have to be overcome by exporters who want to compete successfully on a market that is different in terms of a dominant religion.

Another variable which was important for the purpose of this research was the ability of people in different countries to communicate directly (LANG). The variable is the sole representative of the communicational "distance". In the research, it appeared to be statistically important, and its coefficients had the highest values of all the examined variables. In the pooled model it accounted to 5.13, whereas in the random effects model to 4.87). Like in the case of the RELIG variable, estimation by the ChW method produced a very high value of the coefficient - 10.64. It means that communicational barriers belong to the most important obstacles in trade in the region. This finding reconfirms recent results of Begovic (2011), Fearon (2003), Guiso et al. (2004), Laitin (2000), Melitz (2008) or Melitz and Toubal (2012).

The political "distance" was represented in the research by the binary variable FTA, taking the value of "1" in instances when the countries had signed a Free Trade Agreement. The variable proxied trade liberalisation, which is one of the most common potential trade determinants in the literature. In this research, the coefficients took positive values regardless of the type of model, 0.93, 0.22 and 0.41 for pooled, fixed and random effects respectively. In the pooled and random effects models it was statistically important, whereas in the fixed effects model it was not, which slightly mitigated earlier prognoses of Herderschee and Qiao (2007) of a hidden potential which was supposed to be revealed while signing of the CEFTA-2006 by the countries of the region. Those results show that the trade liberalisation process, which was expressed in signing of the bilateral Free Trade Agreements and CEFTA-2006, had positive influence on trade values. However, when taking into account all bilateral effects for country-pairs (the fixed effects model), it wasn't statistically important.

The last potential distance-type determinants of export were six binary variables describing a state of war between countries (WAR) or a number of the following years after its end (WAR+1 to WAR+5). As expected and in accordance with findings of Adam et al. (2003), in all three models their coefficients had negative signs. Their highest values and statistical importance were recorded by WAR+1 and WAR. One could think why after the years that followed the end of a war, represented by variables WAR+2 to WAR+5, appeared to have small and statistically not important impact on trade22. Those results could be interpreted in a way that warfare has an impact on trade relations between countries, but when possibilities of its revival are seen, entrepreneurs tend to overcome mutual animosities.

The last variable was the stock of the Foreign Direct Investments (FDI) in the exporters country, which proxied the country's "mass" as a "technology" potential. The results showed that regardless of the model, the variable took negative values between -0.12 (in the fixed model) and -0.04 (in the random effects model), and that in neither of cases was the coefficient was statistically important.

6. Conclusions

The presented results of our analysis proved the main working hypothesis which stated that there were more than just geographical, economic and political factors that influenced trade among the Western Balkan countries over the time period from 1995 - 2012. However, not all of the auxiliary hypotheses, which were appointed to certain types of "distance" (communicational, cultural, and historical) and their specific impact on trade values, were proven to be significant. The ability of the citizens to communicate directly and similarity of religious structures were the two auxiliary factors that proved to determine the trade values the most. Moreover, the time of war and one year after had a strong and statistically significant influence on trade, as well as the fact of signing of a free trade agreement between countries. On the other hand, the auxiliary hypothesis about the importance of foreign direct investments and further post-war periods did not prove valid as trade determinants. These conclusions about language, religion and historical factors having significant influence on trade among the WB countries contribute to contemporary science, and so do the methodological aspects of categorization of the notion of "distance" in trade and particular measurement methods of language and religious proximity. The new facts obtained in this research as the contribution to the economic literature are that the ability of the citizens to communicate directly in their mother tongues and similarity of religious structures of the societies enhance trade values between countries. On the other hand, just the opposite effect is brought by warfare, but in only one year after the peace is achieved both previously conflicted countries continue to trade as usual.

A number of data limitations occurred in the research process. Lack of data for early years of the analysis also occurred in the case of export values between some of the countries (mostly from Albania to Bosnia). However, it represented just a small part of the database. Another limitation was the necessity of transferring the data on religion and language from Serbia to Montenegro and to Serbia & Montenegro (Federal Republic of Yugoslavia). Finally, Kosovo was not taken into account at all, due to a complete lack of comparable data.

The study did not, by any means, close the subject of trade determinants in the region of the Western Balkans. Further research should especially encompass issues of internal distance and multilateral distance, which were not raised here. On the other hand, in order to capture the specifics of different types of goods, a similar analysis could be run with a division for intra/inter-industry trade, final/ intermediary goods or with a breakdown for different SITC categories. Each of these would cast light on determinants of trade in certain types of goods. In the present times, when trade is subject not only to differences in factor endowments, but also a strong fragmentation of production and growing differentiation of goods, this proposed direction of further research could bring interesting results.

The study points towards the importance of communication for trade issues, whether it be the straight forward possibility of understanding one another in terms of language, or an intercultural dialog focused on improving tolerance to the other party's beliefs. Policy implications of this research imply using all measures in hand to work on regional dialogue that would enhance multilateral understanding, tolerance, and diminishing of formal and informal barriers to trade.

Footnote

* Received: 24-10-2014; accepted: 10-06-2015

1 The research project was financed by means of the National Science Centre of Poland (Narodowe Centrum Nauki) granted upon decision number DEC-2013/11/N/HS4/03642. Authors would like to express their gratitude to Mr. Slawomir Smiech, PhD, from the Cracow University of Economics, for valuable comments to this research project.

4 Serbian, Croatian and Bosnian belong to South-Western Slavic and Macedonian to South-Eastern Slavic languages (www.ethnologue.com, accessed: 30.09.2013).

5 In this paper Kosovo under Resolution 1244/99 is not included in the analysis due to lack of compa - rable data sets.

6 With respect to the time horizon of our research, Croatia is classified as a country of the Western Balkans, although in the EU terminology, after becoming its full member, it is no longer considered as the Western Balkan country.

7 Croatia and Macedonia entered the CEFTA yet before its transformation to CEFTA-2006 (in 2003 and earlier in 2006 respectively).

8 For the effects of CEFTA-2006 see (Molendowski, 2010, Molendowski, 2011).

9 Data were derived from Statistics Database of World Trade Organization in April 2014, Available from:<http://stat.wto.org/CountryProfile/WSDBCountryPFReporter.aspx?Language=E>.

10 Data were derived from World Economic Outlook of International Monetary Fund in April 2014, Available from:<http://www.imf.org/external/pubs/ft/weo/2014/01/weodata/index.aspx>.

11 This paper presents results of a research, earlier stages of which were outlined in Klimczak (2014a), Klimczak (2014b) and Klimczak and Trivic, (2014).

12 More detailed reviews of the theoretical foundations of the gravity models of trade are presented in: Cafiso 2007, Cieslik 2009 and Kandogan 2009.

13 A subsequent discussion on the typology of "distance" in trade has been taking place in the literature related to business studies (see for instance: Berry et al. (2010), Ellis (2007), Ghemawat (2001), Hofstede (1980), House et al. (2004), Williamson (1985).

14 This structure similarity measure is a special case of the Bray-Curtis index, which is mainly used in natural sciences (Tajoli and De Benedictis 2006). It was theoretically derived by Sun and Ng (2000).

15 As for four stages of similarity in the language tree, the values of the ratios were 0.25, 0.5, 0.75 and 1 (see Klimczak 2014a, Melitz and Toubal 2012).

16 Alternativelly to the ChW approach, the Hausmann-Taylor approach is used to deal with time-invariant variables in the fixed effects models (see Bjelic and Dragutinovic Mitrovic, 2012 as a recent example).

17 Data for GDP and population came from UNCTAD STAT.org (accessed 18.11.2013). GDP was taken in constant prices (2005) and exchange rates. Data on physical "distance" between capital cities comes from www.cepii.org (accessed 18.11.2013).

18 In the case of the Kosovo war, the indexes WAR and WAR+X apply to Albania.

19 The model was tested for multicollinearity and for existence of outliers and it appeared robust. All results available upon request.

20 Regressions were computed with the use of the GRETL 1.9.6. cvs software.

21 As the pooled model proved to be the least relevant out of all three specified in this paper.

22 Exceptions were statistically important variable WAR+4 in the pooled model, and positive signs by the variable WAR+5 in the fixed and random effects models.

References

References

Adam, A., Kosma, T. S., McHugh, J. (2003) "Trade-Liberalization Strategies: What Could Southeastern Europe Learn from the CEFTA and BFTA?", IMF Working Paper, No 03/239, Available from: <http://www.imf.org/external/pubs/ ft/wp/2003/wp03239.pdf>.

Anderson, J. E. (1979) "A Theoretical Foundation for the Gravity Equation", American Economic Review 69(1), pp. 106-116.

Anderson, J. E., van Wincoop, E. (2001) "Gravity with Gravitas: A Solution to the Border Puzzle", NBER Working Paper [Internet] No. 8079, Available from: <http://www.nber.org/papers/w8079>.

-(2004) "Trade costs", Journal of Economic Literature, 42(3), pp. 691-751.

Baltagi, B.H., Egger, P., Pfaffermayr, M. (2003) "A generalized design for bilateral trade flow models", Economics Letters, No 80 (3), pp. 391-397.

Begovic, S. (2011) "The Effect of Free Trade Agreements on Bilateral Trade Flows: The Case of CEFTA", Zagreb International Review of Economics & Business, Vol. 14 (2), pp. 51-69.

Bergstrand J. H. (1985) "The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence" Review of Economics and Statistics, 1985, 67(3), pp. 474-481.

Berry, H., Guillen, M.F., Zhou, N. (2010), "An institutional approach to crossnational distance", Journal of International Business Studies, No (2010) 41, 1460-1480.

Bilas, V. (2007) "Trgovinska povezanost Hrvatske i EU", Zbornik Ekonomskog fakulteta u Zagrebu, Vol. 5, pp. 67-78.

Bjelic, P., Dragutinovic Mitrovic, R. (2012) "The effects of competing trade regimes on bilateral trade flows: case of Serbia", Zbornik Radova Ekonomskog fakulteta u Rijeci - Proceedings of Rijeka Faculty of Economics, Vol. 30, No. 2, pp. 267-294.

Bjelic, P., Dragutinovic Mitrovic, R., Popovic Petrovic, I. (2013) "Administrative Barriers to Trade as Predominant Non-Tariff Barriers". In: The 3rd International Conference on International Trade and Investment, 4-6 September, Mauritius: University of Mauritius and WTO.

Boisso, D., Ferrantino, M. (1997) "Economic distance, cultural distance, and openness in international trade: Empirical puzzles", Journal of Economic Integration, No 12, pp. 456-484.

Botric, V. (2012) "Intra-industry Trade between the European Union and Western Balkans: A Close-up", Ekonomski Institut Zagreb Working Paper [Internet], No 1202, pp. 5-36. Available from: < http://hrcak.srce.hr/index.php?show=clanak&id_ clanak_jezik=125831>.

Bussière, M., Fidrmuc, J., Schnatz, B. (2005) "Trade Integration of Central and Eastern European Countries: a Lesson from a Gravity Model", European Central Bank Working Paper Series [Internet], No 545, Available from: <https:// www.ecb.europa.eu/pub/pdf/scpwps/ecbwp545.pdf>.

Cafiso, G. (2007) "The Geographic Space in International Trade: from Gravity to New Economic Geography", MPRA Paper 20269, University Library of Munich, Germany.

Camarero, M., Tamarit, C. (2003) "Estimating exports and imports demand for manufactured goods: the role of FDI", European Economy Group, Working Paper, No. 22.

Cheng, I.H., Wall, H.J. (2005) "Controlling for heterogeneity in gravity models of trade and integration", Federal Reserve Bank of St. Louis Review, 87, 1, pp. 49-63.

Christie, E. (2002) "Potential Trade in Southeast Europe: a Gravity Model Approach". WIIW Working Papers [Internet], No 21, Available from: <http://wiiw.ac.at/potential-trade-in-southeast-europe-a-gravity-modelapproach-p-520.html> [Accessed May 17, 2014].

Cejvanovic, F., Dzafic, Z. (2011) "Sporazum CEFTA-2006 i Ucinci vanjskotrgovinske razmjene Slucaj Bosne i Hercegovine". In: Trgovina kao pokretac razvoja Srednje i Jugoistocne Evrope, Zagreb: Ekonomski fakultet, pp. 124-139.

Cieslik, A. (2009) "Bilateral trade volumes, the gravity equation and factor proportions", The Journal of International Trade & Economic Development: An International and Comparative Review, Vol. 18, Issue 1, 2009, pp. 37-59.

Damijan, J., Sousa, J., Lamotte, O. (2006) "The Effect of Trade Liberalization in South-Eastern European Countries", The Vienna Institute for International Economic Studies Working Paper [Internet], No 070, Available from: <http:// wiiw.ac.at/the-effect-of-trade-liberalization-in-south-eastern-europeancountries-p-3233.html>.

Druzic, I., Anic, M., Sekur, T. (2011) "Gravity Model of Croatian Regional Foreign Trade". In: Proceedings of 8th International Conference Economic Integration, Competition and Cooperation, 6-9 April, Opatija, Rijeka: University of Rijeka, Faculty of Economics.

EBRD (2003) Transition report 2003: Integration and regional cooperation. London: European Bank for Reconstruction and Development.

- (2013) Transition report 2013: Stuck in Transition, London: European Bank for Reconstruction and Development.

Ellis, P.D. (2007) "Paths to foreign markets: does distance to market affect firm internationalisation?", International Business Review, Vol. 16, No 5, pp. 573-593.

European Commission (2014) Countries and Regions-Western Balkans, Available from: <http://ec.europa.eu/trade/policy/countries-and-regions/regions/westernbalkans/>.

Fearon, J.D. (2003) "Ethnic and Cultural Diversity by Country", Journal of Economic Growth, No 8, pp. 195-222.

Fetahu, E. (2014) "Trade Integration Between Albania and European Union - A Gravity Model Based Analysis", European Scientific Journal, Vol. 10, No 7, pp. 185-199.

Fontagné, L., Pajot, M. (1997) "How Foreign Direct Investment Affects International Trade and Competitiveness: an Empirical Assessment", CEPII document de travail, No. 97-17.

Ghemawat, P. (2001) "Distance still matters: The hard reality of global expansion", Harvard Business Review, No. 79(8), pp. 137-147.

Gjipali, A., Jorgji, E., Liko, E. (2012) "Intra-Regional Trade in Transitional Economy: Prospects from South-Eastern Europe", European Perspectives, Vol. 4, No 4, 2 (7), pp. 15-40.

Guiso, L., Sapienza, P., Zingales, L. (2004) "Cultural biases in economic exchange", NBER Working Paper [Internet], No 11005, Available from: <http://www.nber. org/papers/w11005.pdf>.

Gumuscan, I., Kahveci, M. (2012) "An Evaluation of Bilateral Trade between Turkey and Balkans: Based on Gravity Trade Model and Linder Hypothesis". In: The 2nd International Balkan Annual Conference IBAC 2012, Istanbul: Istanbul University.

Herderschee, J., Qiao, Z. (2007) "Impact of Intra-European Trade Agreements, 1990-2005: Policy Implications for the Western Balkans and Ukraine", IMF Working Paper [Internet], No 07/126, Available from: <https://www.imf.org/ external/pubs/ft/wp/2007/wp07126.pdf>.

Helble, M. (2007) "Is God good for trade", Kyklos, 60(3), pp. 385-413.

Helpman, E. (1981) "International trade in the presence of product differentiation, economies of scale and monopolistic competition: A Chamberlin - Heckscher - Ohlin approach", Journal of International Economics, 11, pp. 305-340.

Helpman, E., Krugman, P. (1985) Market Structure and Foreign Trade: Increasing Returns, Imperfect Competition, and the International Economy, Cambridge, MA: MIT Press.

Hofstede, G. (1980) Culture's consequences: International differences in workrelated values, Beverly Hills, CA: Sage.

House, R. J., et al. (2004) Culture, leadership, and organizations: The GLOBE study of 62 societies, Thousand Oaks, CA: Sage.

Ilic, S. (2012) Is there potential for trade growth? A gravity approach in the case of Serbia, Central European University.

Jovicic, M., Zdravkovic, M., Dragutinovic Mitrovic, R. (2001) "Federal Republic Yugoslavia: Trade Potentials and Comparative Advantages", WIIW Working Papers [Internet], No 010, Available from: <http://wiiw.ac.at/federal-republicyugoslavia-trade-potentials-and-comparative-advantages-p-3343.html>.

Kaminski, B., de la Rocha, M. (2003) "Stabilization and Association Process in the Balkans: Integration Options and their Assessment". World Bank Policy Research Working Paper [Internet], No 3108, Available from: <http://elibrary.worldbank. org/doi/abs/10.1596/1813-9450-3108> [Accessed February 18, 2014].

Kandogan, Y. (2009) "A gravity model for components of imports", Review of Applied Economics, Vol. 5, No 1-2, pp. 99-115.

Kernohan, D. (2006) "Reverse Balkanisation? Trade Integration in South-East Europe", The Center for European Policy Studies Working Document [Internet], No. 249, Available from: <http://www.ceps.be/>.

Klimczak, L. (2014a) "Determinanty handlu wzajemnego panstw Balkanów Zachodnich", Zeszyty Naukowe Podkarpackiej Szkoly Wyzszej im. bl. ks. Wladyslawa Findysza w Jasle, No. 11, pp. 225-237.

Klimczak L. (2014b) "The gravity model as a tool for the international trade analysis - a case study of the Western Balkans". In: Proceedings of the 8th Professor Aleksander Zelias International Conference on Modelling and Forecasting of Socio-Economic Phenomena, May 6-9, Zakopane, Poland, pp. 67-76.

Klimczak, L., Trivic, J. (2014) "Determinants of Trade in the Western Balkans: A Gravity Approach". In: Conference Proceedings of the Third REDETE Conference Economic Development and Entrepreneurship in Transition Economies: Challenges in the Business Environment, Barriers and Challenges for Economic and Business Development, organized by the Faculty of Economics University of Banja Luka, April 10-12, 2014, Banja Luka, BiH, pp. 287-297.

Kucharcuková, O., Babecký, J., Raiser, M. (2010) "A Gravity Approach to Modelling International Trade in South-Eastern Europe and the Commonwealth of Independent States: The Role of Geography", CNB Working Paper [Internet], No. 4, Available from: < http://www.cnb.cz>.

Laitin, D. (2000) "What is a language community?", American Journal of Political Science, No 44, pp. 142-155.

Lewer, J., van den Berg, H. (2007) "Estimating the institutional and network effects of religious cultures on bilateral trade", Kyklos, No 60 (2), pp. 255-277.

Linneman, H. (1966) An econometric study of world trade flows, Amsterdam: North-Holland Publishing Co.

Martin, P., Mayer, T., Thoenig, M. (2005) "Make trade not war", Centre for Economic and Policy Research Discussion Paper Series, No 5218, Available from: <www.cepr.org/pubs/dps/DP5218.asp>.

McCallum, J. (1995) "National Borders Matter: Canada - U.S. Regional Trade Patterns". American Economic Review. Vol. 85, No 3, pp. 615-623.

Mehic, E., Babic-Hodovic, V. (2011) "Izravna strana ulaganja i trgovina u zemljama Jugoistocne Evrope". In: Trgovina kao pokretac razvoja Srednje i Jugoistocne Europe, 16th November Zagreb: Ekonomski fakultet, pp. 139-155.

Melitz, J. (2008) "Language and foreign trade", European Economic Review, No 52 (2008), pp. 667-699.

Melitz, J., Toubal, F. (2012) "Native language, spoken language, translation and trade", Centre for Economic Policy Research, Working Paper, No 8994, London.

Miljovski, J., Uzunov, V. (2002) "International and Regional Economic Integration in South East Europe: The Case of Macedonia", WIIW Working Papers, No. 019, Available from: <http://wiiw.ac.at/free-publications.html> [Accessed April 15, 2014].

Molendowski, E. (2010) "The Role of the "CEFTA-2006" Agreement in Preparation of the Balkan Countries Towards European Union Integration". In: Bukowski, S.I. ed. Global Economy Challenges of the XXI Century, Radom: Technical University of Radom.

- (2011) "Western Balkan countries - main issues concerning co-operation and integration with the European Union". In: Contemporary Issues In Business, Management and Education, Vilnius: Gediminas Technical University, Faculty of Business Management, pp. 235-236.

Mojsoska-Blazevski, N., Petreski, M. (2010) "Western Balkan's trade with the EU and CEFTA-2006: Evidence from Macedonian data", Munich Personal RePEc Archive, No. 41942, Available from: <http://mpra.ub.uni-muenchen.de/41942/1/ MPRA_paper_41942.pdf>.

Nikolic, G., Jovanovic, N., Todoric, T. (2011) CEFTA 2007-2010 - Iskustva, potencijal i perspektiva, Beograd: Centar za novu politiku.

Pjerotic, Lj. (2008) "Trade Liberalization in the South East Europe - Effects and Controversial Issues", Panoeconomicus, No 4, pp. 497-522.

Pllaha, A. (2012) "Free Trade Agreements and trade integration among South Eastern European countries; Gravity model estimations", Bank of Albania Working Paper. No 05 (36) Available from: <http://www.bankofalbania.org/>.

Pöyhönen, K. (1963) "Towards a General Theory of International Trade", Ekonomiska Samfundet Tidskrift, No 16 (1963), pp. 69-78.

Sarkees, M. R., Wayman F. (2010) Resort to War: 1816 - 2007. CQ Press.

Stojcic, N. (2012) "Two Decades of Croatian Transition: A Retrospective Analysis", South East European Journal of Economics and Business, Vol. 7, No. 2, pp. 63-76.

Sun, G.Z., Ng, Y.K. (2000) "The measurement of structural differences between economies: An axiomatic characterization", Economic Theory, 16, pp. 313-321.

Tajoli, L., De Benedictis, L. (2006) "Economic Integration and Similarity in Trade Structures", FEEM Note di Lavoro, April 2006.

Tinbergen, J. (1962) Shaping the World Economy Suggestions for an International Economic Policy, New York: The Twentieth Century Fund.

Tkalec M., Vizek, M. (2014) "Real estate boom and export performance bust in Croatia", Zbornik Radova Ekonomskog fakulteta u Rijeci - Proceedings of Rijeka Faculty of Economics, Vol. 32, No. 1, pp. 11-38.

Tosevska-Trpcevska K., Tevdovski, D. (2014) "Measuring the Effects of Customs and Administrative Procedures on Trade: Gravity Model for South-Eastern Europe", Croatian Economic Survey, Vol. 16, No. 1, pp. 109-127.

Williamson, O.E. (1985) The Economic Institutions of Capitalism, New York: Free Press.

World Bank (2008) Western Balkan Integration and the EU: An Agenda for Trade and Growth, Washington: World Bank.

Wydymus, S. (2012) "System wspólzaleznosci wskazników makroekonomicznych w grawitacyjno-potencjalowym modelu polskiego handle zagranicznego". In: Wydymus, S., Maciejewski, M. eds. Handel i wspólpraca miedzynarodowa w warunkach kryzysu gospodarczego, Kraków: Fundacja Uniwersytetu Ekonomicznego w Krakowie.

Zelias, A. ed. (1991) Ekonometria przestrzenna, Warszawa: PWE.

Appendix

(ProQuest: Appendix omitted.)