Ancestral Distance as a Barrier to International Trade Irene Fensore Stefan Legge Lukas Schmid∗ July 28, 2016 Preliminary Version‡ Abstract We investigate the impact of ancestral distance on international trade flows. Using a new data set covering the universe of global trade, we document that ancestral distance is an important barrier to bilateral trade in addition to transportation costs. We use ge- netic differences between populations as a proxy for ancestral distance. Our results show that country pairs with a high genetic distance are less likely to trade with each other (extensive margin) and if they do trade, we find that genetic distance negatively affects the volume and number of goods traded (intensive margin). These findings are robust to including a vast array of micro-geographic controls as well as linguistic and religious distance variables. We provide evidence suggesting that the inverse relationship between bilateral trade flows and ancestral distance arises from both increased trade costs. JEL Classification: F14, F15, O33 Keywords: Ancestral Distance, Trade Barriers, Trade Flows ∗ University of St.Gallen, Department of Economics, SIAW-HSG, Bodanstrasse 8, CH-9000 St.Gallen, Switzer- land. E-mail: [email protected]; [email protected]; [email protected] ‡ Please do not cite without permission. The latest version is available upon request. For helpful comments and discussions, we are grateful to James Anderson, Reto F¨ollmi, Beata Javorcik, Dennis Novy, Enrico Spolaore, Romain Wacziarg, Josef Zweimu¨ller as well as seminar participants at the University of St.Gallen and the University of Zurich. 1 Introduction Studies on international trade broadly agree that bilateral trade costs are of sizable magnitude. Including transportation, border-related and local distribution costs, Anderson and van Win- coop (2004) estimate a 170% total trade barrier. However, there is still a limited understanding of the factors that account for this large magnitude. Despite vast technological improvements as well as large reductions and often complete abolishment of tariffs, even in border regions trade across countries is far smaller than trade within countries. This suggests that other factors beyond geography and tariffs constitute a significant barrier to trade. This paper argues that ancestral distance, measured by genetic differences between popu- lations, reflects a significant part of this barrier. In particular, we find that bilateral ancestral distance affects international trade in a globalized world. We can link this finding to previous research pointing out that countries with a more recent common ethnic ancestry trust each other more and tend to have similar preferences (Spolaore and Wacziarg, 2015). More broadly, Spolaore and Wacziarg (2016a) argue that there is a close association between ancestry, lan- guage and culture. Furthermore, genetic distance is highly correlated with answers to the World Values Survey (Desmet et al., 2011). Hence, ancestral distance can reflect both differ- ences in preferences as well as an obstacle to communication, social interaction, and learning across different societies. Interpreting genetic ancestral distance as a barrier to communication and collaboration, Spolaore and Wacziarg (2009) find that long-term divergence in populations negatively affects the diffusion of institutional and technological innovations. Using genetic distance as a proxy for common linguistic and cultural roots, it has further been shown that European countries with similar populations are more likely to trade with each other (Guiso, Sapienza and Zingales, 2009). It remains unclear, however, whether this finding extend beyond the set of relatively homogeneous European countries.1 Moreover, previous research was limited to the period prior to 1997 and the intensive margin of trade even though the extensive margin might be more relevant (Hillberry and Hummels, 2008). Hence, it is an open question whether genetic distance is in fact a significant barrier to international trade. 1Guiso, Sapienza and Zingales (2009, p. 1128) emphasize that their “results are obtained within the bound- aries of the old European Union, which comprises fairly culturally homogeneous nations” and that the impact of genetic distance “might be much larger on world trade”. 1 Using a novel data set, we document the sizable impact of ancestral distance on bilateral trade flows for a wide range of 172 countries. In a gravity equation following Tinbergen (1962) as well as Anderson and van Wincoop (2003), we find that a larger ancestral distance not only reduces the probability that a trade relationship exists (extensive margin), but also constitutes an important obstacle to the amount and number of goods traded (intensive margin). These findings are robust to the inclusion of a vast array of measures for micro-geographical distance as proxy for transportation costs, as well as measures of linguistic and religious distance. A one standard deviation increase in ancestral distance decreases the probability that countries estab- lish trade relations by about twenty percent of a standard deviation. Moreover, the volume of trade is up to 17 percent lower if genetic distance is one standard deviation larger. While an- cestral distance is highly (though not perfectly) correlated with geographic distance, we show that ancestral distance constitutes a barrier to international trade on top of geographic dis- tance. Our results reveal that after netting out the impact of geography, we still observe that country pairs with unexpectedly large ancestral distances are less likely to trade and, if they do trade, ship smaller amounts.2 In addition, we show that ancestral distance has a significant and positive impact on estimated trade costs taken from Simonovska and Waugh (2014). This indicates that neglecting genetic differences yields imprecise estimates of true trade barriers. As pointed out by Giuliano, Spilimbergo and Tonon (2014), a careful analysis of the impact of genetic distance has to take into consideration micro-geographic variables. This comprises, for example, the presence of mountains, seas, or rivers which influence the ease with which two countries can enter into contact and thus their genetic relatedness. Following this research, we collect measures of common official languages, contiguities, access to the same sea, latitude, longitude, terrain ruggedness, shares of fertile soil, desert, and tropical climate, as well as the average distance to the nearest ice-free coast. We find that the inverse relationship between geneticdistanceandinternationaltraderemainsrobusttotheinclusionofsuchcontrolvariables. Moreover, we control for a set of political variables including corruption, civil and political liberty, as well as free trade agreements. None of these variables alter the finding that bilateral trade flows are negatively associated with ancestral distance. 2Unexpectedly large ancestral distance refers to the difference between true genetic distance and projected genetic distance based on geographic distance. 2 Our work is related to several strands of literature. First, we contribute to prior research on the determinants of trade costs and their impact on trade flows. This literature has documented that transportation costs —the costs of shipping goods from one country to another— only constitute a fraction of total trade costs.3 Several papers, including Melitz (2008), Ku and Zussman (2010), Lohmann (2011), Egger and Lassmann (2014, 2015), as well as Melitz and Toubal (2014) find that countries sharing a common language trade more with each other. Their findings are in line with a meta-analysis based on 81 academic articles by Egger and Lassmann (2012). In a broader context, Felbermayr and Toubal (2010) construct a proxy for cultural proximity based on score data from the Eurovision Song Contest. The authors find that their measure of proximity is positively correlated with bilateral trade volumes. Similarly, we document that countries with a shorter ancestral distance are more likely to trade with each other. This adds to prior research by Sapienza, Zingales and Guiso (2006) on the impact of cultural differences on trade flows. In particular, we contribute to the work by Guiso, Sapienza and Zingales (2009) who show that countries with lower bilateral trust trade less and also show lower levels of portfolio as well as direct investment. Second, our findings add to the literature on the diffusion of technology. As suggested by Keller (2010), Alvarez, Buera and Lucas (2013) as well as Allen and Arkolakis (2014), international trade plays a key role in the diffusion of technology. As a result, barriers to trade andinformationalfrictions(Allen,2014)maypreventpoorcountriesfromcatchingup. Previous research has documented that the diffusion of development is closely related to genetic distance (Spolaore and Wacziarg, 2009, 2013a,b). In particular, the authors argue that countries with larger genetic distance to the United States (i.e., the technological frontier) will be slower in the adoption of new technologies. Hence such countries will have a lower per capita income. In our paper, we replicate their finding and suggest that trade is an important channel through which genetic distance affects the diffusion of technology and hence the spread of economic growth. Having data on the universe of global trade flows, we provide evidence for a statistically and economically significant negative impact of genetic distance on trade flows. Finally, our work adds to the literature investigating the role of genetic distance for un- 3This is not to argue that transportation costs do not matter. For example, Limao and Venables (2001) find that infrastructure and transport costs are important constituents of bilateral trade costs. 3 derstanding several economic and political phenomena. Spolaore and Wacziarg (2016b), for example, show that international conflict is more likely to arise between populations that are closely related. The remainder of the paper is organized as follows. In Section 2, we discuss which mech- anisms could explain a relationship between ancestral distance and trade. Section 3 provides information on the construction of our data set as well as several descriptive statistics on the relationship between genetic and geographic distance. Section 4 describes the econometric ap- proach, shows the main empirical results as well as a series of robustness checks. In Section 5, we link genetic distance to estimated trade costs and discuss channels through which ancestral distance affects trade. Finally, Section 6 concludes. 2 Theoretical Considerations Why would we expect ancestral distance to affect bilateral trade flows? What mechanism would be reflected by an association between trade and ancestral distance? Are information andcontractenforcementcostshigherthelargerthebilateralancestraldistance? Doesancestral distance capture differences in tastes and consumer preferences? In this section, we discuss such questions on how ancestral distance could affect bilateral trade through various channels. In terms of theory, consider a non-linear gravity model with a multiplicative error term as in Anderson and van Wincoop (2003) where exports x from country i to country j are a function i,j of GDP (y), trade costs (τ ), and price indices (P , P ): i,j i j y y (cid:18) τ (cid:19)(1−σ) i j i,j x = (1) i,j y P P w i j We consider ancestral distance to capture several aspects of trade costs τ . This includes i,j information, contract enforcement or communication costs. Alternatively, we could model CES preferences such that demand for country-specific varieties depends on bilateral ancestral dis- tance. 4 2.1 Ancestral Distance and Trade Costs In the trade literature, there is an ongoing discussion of which factors account for the large magnitude of estimated trade costs. Head and Mayer (2013) coined the term ‘dark costs’ and argue that 72–96% of the rise in trade costs associated with distance is attributable to the dark sources of resistance.4 In a related study, Allen (2014) estimates that search costs account for 90% of the distance effect, leaving only 10% for transport costs. Thereareseveralreasonswhyancestraldistancebetweentwocountriesincreasestradecosts. In the presence of imperfect information, familiarity declines with geographic distance (Gross- man, 1988). Hence, information costs are higher among distant countries.5 Given that our measure of genetic distance positively correlates with geographic distance and that it provides a summary statistic for intergenerationally transmitted traits (Spolaore and Wacziarg, 2015), higher ancestral distances could result in higher informational asymmetries between countries. A second explanation for smaller trade volumes between ancestrally distant countries could be the lack of established trading networks. Information Costs — Interpreting the detrimental effect of ancestral distance on trade flows as reflecting transaction and communication costs would be in line with prior research on the impact of common languages (Melitz, 2008). However, why should information be subject to iceberg melting with distance? A related set of evidence on the impact of distance on informa- tion focuses on knowledge flows, based primarily on citation patterns and their spatial decay. Peri (2005) analyzes citations using a gravity equation, Li (2014) estimates a more standard gravity specification and finds distance elasticity on citation of -0.12 (when she excludes self- citation). Griffith, Lee and Van Reenen (2011) examine the speed at which new patents are cited at home compared with when they are cited in other countries. The key finding is that the bias towards citing domestic patents first at home is declining over time. Finally, Comin, Dmitriev and Rossi-Hansberg (2012) examine adoption of 20 major technologies in 161 coun- 4Feyrer(2009)usestheclosingoftheSuezCanalbetween1967and1975andconcludesthatdarktradecosts account for 50%–85% of the effect of distance on trade flow. 5Huang (2007) investigates an implication of Grossman’s hypothesis that distance proxies for familiarity. Huangestimatesthatcountriesthataremoreaversetouncertaintyhavelargerdistanceeffectsontheirexports. 5 tries over the last 140 years. They estimate that the frequency of interactions decays by 73% every 1000 km for the median technology. Does technology reduce trade costs? Horta¸csu, Mart´ınez-Jerez and Douglas (2009) find that eBay transactions have a distance elasticity of -0.07. This is much smaller than the near unit elasticities estimated for commodity flows within the United States. More surprising is the finding that transactions are 75% more likely to occur within the same state (after controlling for distance and state fixed effects). The in-state effect should have been negative given that sales taxes are only applicable when the buyer and seller reside in the same state. Horta¸csu, Mart´ınez-Jerez and Douglas (2009) suggest different levels of trust may underlie these effects, perhaps because ‘increased possibility of direct enforcement of the trade agreement, either by returning the good in person or by compelling the seller to deliver on his or her promise.’ An alternative interpretation would consider ancestral distance as a proxy for historical relations betweencountriesasinRauchandTrindade(2002)orfollowBellocandBowles(2013)andargue that long-term cultural and institutional differences between countries are linked to bilateral trade flows. Fixed versus Variable Costs — Does ancestral distance reflect a fixed or variable trade cost? In other words, do we consider ancestral distance to be a temporary or permanent barrier to trade? Spolaore and Wacziarg (2016a) argue that there is a close association between ancestry, language and culture. Hence, ancestral distance can be understood as an obstacle to communication, social interaction, and learning across different societies. In our empirical analysis, we explore the effect of ancestral distance on both the extensive and intensive margin of trade. Any evidence that trade at the extensive margin is reduced would be interpreted as evidence of a fixed cost. In contrast, the effect with respect to the total trade volume would reflect a variable trade cost. 2.2 Ancestral Distance and Preferences As we discussed above, ancestral distance can also affect bilateral trade flows through individual preferences if countries produce country-specific varieties. Following Linder’s hypothesis, two 6 countries may trade more with each other if they have more similar demand structures. If ancestral distance increases differences in preferences, we should observe a negative correlation of genetic distance with trade flows.6 As suggested by the ‘home bias’ phenomenon in Trefler (1995), the same observation could be made in the presence of very localized tastes, which are historically determined and change only slowly. There is number of papers discussing the impact of geography on preferences. Research has addressed this association in the context of oil versus butter (Head and Mayer, 2013), music (Ferreira and Waldfogel, 2013), websites (Blum and Goldfarb, 2006), cereals (Bronnenberg, Dub´e and Gentzkow, 2012; Atkin, 2013). 2.3 Disentangle the Channels It is important to note that both the trade cost explanation (ancestral distance increases trade costs) and the Linder hypothesis have the same prediction: smaller trade flows between ances- trally distant countries. In order to disentangle whether ancestral distance affects trade through its impact on trade costs (i.e., trust; information or contract enforcement costs) or preferences, we explore differences across commodities. If the main channel is in fact trade costs, we should observe smaller effects: (i) in the presence of migration networks, and (ii) between non-corrupt countries Additionally, we should see an effect even when restricting the set of commodities to homogeneous goods that do not differ horizontally or vertically (e.g., raw materials). The first point refers to the fact that existing migration networks —the presence of people from coun- try A in country B — reduces information and enforcement costs. Similarly, if both countries have low levels of corruption, contract enforcement should also be less of a problem. Finally, in the case of homogeneous commodities like coal or oil, differences in taste do not matter. The coal from China is essentially the same as from Germany. Hence, ancestral distance does not reflect different varieties but simply trading costs. 6Empirically, Hallak (2010) finds support for the Linder hypothesis after controlling for the effect of inter- sectoral determinants of trade. 7 3 Data In this section, we describe our data sources and how we combine them into a single data set. Moreover, we provide descriptive statistics on all variables employed in the analysis. Our empirical work is based on a novel data set which contains information on international bilateral trade flows, country characteristics, and numerous measures of genetic, linguistic, religious, and geographic distances. We explain the source and definitions of each part separately. 3.1 Trade Data Our data on international trade flows is taken from UN COMTRADE, a database that contains all bilateral trade flows for the year 2000. For each recorded trade flow, the data includes both the value and weight, which is available at the 6-digit commodity code level. Notably, every reporting country (‘reporter’) has a large set of partner countries (‘partners’). For the set of countries that do not report imports and exports (i.e., a large set of poorer countries), we follow the method by Feenstra et al. (2005) as well as Helpman, Melitz and Rubinstein (2008) who impute exports and imports of non-reporting countries from the reports of (richer) countries trade flows. For example, Albania might not provide information on their exports to the United States. In this case, we use the import data from the United States. By using this method, our data set contains virtually all of the world’s countries and their trade flows.7 A significant shortcoming of the UN COMTRADE data is that it only includes positive trade flows. In other words, the missing (or zero) trade flows are not recorded. To overcome this issue, we save the full list of (reporter and partner) countries. Using this list we create a template that contains all possible country pairs.8 For every pair, our data set has a separate entrywitheach6-, 4-, 3-, or2-digitcommoditycode. Asaresultourtemplatedatafilecoversall possible trade flows. This allows us to investigate not only the intensive but also the extensive margin of trade. 7The only trade flows we miss are those between two countries, both of which do not submit information to the UN COMTRADE data base. These trade flows, however, comprise a negligible fraction of world trade. 8In terms of countries, we only remove those nations whose population is smaller than ten thousand. These countries account for only a tiny fraction of international trade. Moreover, crucial information such as GDP is usually not available. 8 3.2 Country Information We merge the trade flow data with country-level information. In particular, we add data on GDP and population size for each country. As primary source for this information, we use the Penn World Table 8.1, for which we take into account the most recent update by Feenstra, Inklaar and Timmer (2015). If there is no information for a particular country, we use the World Development Indicators as secondary or, if necessary, UNdata as third data source. Note that we use the secondary (or tertiary) data sources to predict the GDP or population value that is missing in the PWT. This makes the GDP (per capita) values comparable even if they stem from different sources. The literature on political regimes and trade has found empirical support for the hypothesis that democracies are more likely to set up free trade areas and trade more with each other (Mansfield, Milner and Rosendorff, 2000). We follow this insight and use data from the Polity IV Project to test whether regime types affect our estimates. In particular, we use a dummy variable that takes the value one if both countries’ democracy score (which ranges from 0 to 10 with higher values indicating more democratic) is above eight. To account for trade policy, free trade areas (FTA) as well as political unions, we extend the list of variables by dummy variables for each country’s membership in the EU, NAFTA, EFTA, AFTA, and Mercosur. Furthermore, we add data by Baier, Bergstrand and Feng (2014) as well as Bergstrand, Larch and Yotov (2015) who provide a database on Economic Integration Agreements (EIA). For each bilateral pair, this indicator ranges from 0 to 6 with higher values reflecting deeper integration. Finally, we use data on political rights and civil liberties from Freedom House as well as information about corruption from Transparency International. 3.3 Geographic Variables ´ We add a large set of geographic information to our data. The Centre d’Etudes Prospectives et d’Informations Internationales (CEPII) provides a database that comprises both information for each country as well as bilateral variables. The former includes each country’s continental location, currency as well as a dummy for being landlocked. The bilateral variables provide information on geodesic distance between largest cities, contingency, common official languages, 9