Cultural Biases in Economic Exchange

Cultural Biases in Economic Exchange Luigi Guiso University of Sassari, Ente L. Einaudi & CEPR Paola Sapienza Northwestern University & CEPR Luigi Zingales University of Chicago, NBER, & CEPR First Draft: December 2003 Very Preliminary and Incomplete Abstract How much do cultural biases affect economic exchange? We try to answer this question by using the relative trust European citizens have for their fellow countrymen and for citizens of other countries. We find that lower relative levels of trust toward citizens of a country lead to less trade with that country, less portfolio investment, and less direct investment in that country. This effect persists after controlling for all the objective characteristics of that country, including the quality of their legal system. Furthermore, it is more pronounced when the transactions are more trust-intensive. We conclude that cultural perceptions are important (and generally omitted) determinants of economic exchange. JEL: Keywords: Culture, Trade, Home Bias in Portfolio, Foreign Direct Investments, Trust. Luigi Guiso acknowledges financial support from MURST, and the EEC. Paola Sapienza acknowledges financial support from the Center for International Economics and Development at Northwestern University. Luigi Zingales acknowledges financial support from the Center for Research on Security Prices at the University of Chicago. We also thank Roc Armenter for research assistantship. 1

Economic models predict a great deal of exchange among economi agents (trade in goods, risk sharing, equity investment, international portfolio diversification). In fact, they predict much more exchange than we observe in practice. People do not share risk anywhere close to what standard models predict (Cochrane, 1996). They also trade much less than expected. For example, McCallum (1995) finds that trade between Canadian provinces and U.S. states is twenty times smaller than trade among Canadian provinces themselves. Economic agents also participate in the equity market much less than classical portfolio theory predicts. And when they do, they are grossly under-diversified (Guiso et al., 2000). For example, at the end of the 1980s Americans were holding 94% of their equity wealth in the United States and Japanese 98% in Japan, when models predict they should not hold much more than a third (French and Poterba, 1991). Finally, economic models predict savings should massively flow from rich and developed countries, where capital is abundant, to poor ones, where capital is scarce, but they do not (Lucas, 1995). There are numerous attempts to explain these paradoxes individually and a few also together (e.g., Obstfeld and Rogoff, 2002). All these attempts exploit some standard friction (trading costs, informational asymmetries, etc.) to explain why agents do not behave according to what the more simplistic models would predict. While these more sophisticated models go some way toward explaining these puzzles, they can hardly eliminate them. In this paper, we propose an alternative type of friction, based on cultural stereotypes. Our starting point is that in most forms of exchange contracts are incomplete and thus exchange requires some trust between the parties involved. In forming their expectations about their counterparts, economic agents use not only objective information, but also cultural stereotypes. This phenomenon is particularly pronounced across countries, where we focus our attention to. In a a recent survey carried out by the 3i/Cranfield European Enterprise Center, European managers of five different nationalities were asked to score themselves and managers in other countries against twelve quality indicators (competency, efficiency, tenacity, reliability, hard work, entrepreneurship, education, trustworthiness, punctuality, humour, compassion, and ability to get on with others). 1 The average results are summarized in the table below. The highest 1 In total 1,016 managers (managing companies under 500 employees) responded from five major EC countries: Britain (433 responses), France (127), Germany (135), Italy (185) and Spain (136). See http://www.cranfield.ac.uk/docs/spss/spss.html. 2

ranking is 1 and the lowest ranking is 5. Britain France Germany Italy Spain British view 1 3 2 4 5 German view 3 2 1 4 5 French view 5 2 1 3 4 Italian view 4 3 1 2 5 Spanish view 4 3 1 5 2 The survey results suggest that European managers share some common views about each others (e.g. some countries are systematically ranked as overall better managers (Germans) or worse managers (Spanish), also, everybody but the Italians think that Italians are the least punctual managers) but also managers have some very idiosyncratic views. For example, only French managers ranked British managers as the worst managers. Also, the main diagonal of the matrix shows that managers tend to like managers from their own country more than managers from other countries. These idiosyncratic views may be affected by different level of information, but also by cultural stereotypes. In this paper, we attempt to quantify the importance of these idiosyncratic cultural opinions for economic exchange. In order to do this, we first study whether the pattern showed in the table above generalizes to a broader sample of individuals across countries. Second, we measure the impact of these opinions on economic transactions. Finally, we try to distinguish whether these opinions reflect mostly information or cultural stereotypes. To study the different opinions that individuals across countries have about each others, we use a cross national survey (Eurobarometer) that has data on the degree of trust that European citizens have towards citizens of other countries (both in Europe and outside). In general, it is difficult to distinguish between the objective component of trust (I trust that my counterparty will not cheat me because I know that his legal system will punish him for that) and the subjective one (I trust my counterparty because I consider him a trustworthy person). The structure of our data helps us in dealing with this issue, because we can control for both a country of origin fixed effect (which should capture possible systematic forces affecting the average level of trust) and for country of destination fixed effects (which should represent the objective characteristics of the citizens of that country). What is left is how citizens of a country 3

idiosyncratically perceive the citizen of another country: for example, how much Germans trust Italians relative to the average trust Germans exhibit in general and relative to the average trustworthiness of Italian. We use this measure of relative trust to test the impact of trust on various form of exchange across countries. We find that higher degree of relative trust can explain cross country trade beyond what extended gravity models can account for. At sample means, increasing the trust of both exporter and the importer by one standard deviation raises the share of exports over GDP by 31 percent; if only the trust on one of the two partners were increased, the share of exports should increase by about 15 percent. More interestingly, relative trust also affects the type of trade that takes place. More trust toward a country increase the amount of durable goods (goods where the quality of the product is an important unobservable) imported from that country. By contrast, more trust toward a country increase the amount of commodities (goods where the quality of the product is less important than the financial reliability of the buyer) exported toward that country. We also show that the degree of relative trust affects the pattern of international diversification. Portfolio investments are tilted toward countries whose citizens are considered relatively more trustworthy. While this helps explain the home bias portfolio puzzle (there is a universal tendency to trust your fellow citizens more), the difference between trust toward fellow citizens and foreigners is not sufficient to eliminate the puzzle. We find the same results when we analyze the pattern of foreign direct investments. A country is more willing to make foreign direct investment in a country whose citizen it trusts more. Hence, the correlation between trust and economic exchange seems to be both economically important and pervasive. To interpret these results, however, we need to answer two questions. First, is it trust that impacts economic exchange or the other way around? Second, does the degree of relative trust reflects cultural stereotypes or differences in information across countries? In order to ascertain that the direction of causality is from trust to trade, rather than from trade to trust we use two approaches. First, we document that the impact of relative trust is more important when theory predicts so. Thus, we show that the trust of the country of origin is more important for trade of durable good, while the trust of the country of destination is more important for other goods. Second, we instrument the trust Italians exhibit toward Germans 4

with the trust all the other countries exhibit toward Germans (and similarly for all the other pairs). Since the IV estimates are virtually identical to OLS ones, we conclude reverse causation is not an issue. Another important question is whether the degree of relative trust reflects cultural stereotypes or differences in information across countries. In other words, The second question whether the data on average trust reflects information or stereotypes is more complex. Why do the British tend to trust the French as much as they trust the Italians and the Spanish, but much less than they trust the Belgians and the Dutch? And why the French reciprocate, mistrusting the British as much as they mistrust the Greeks? Looking at data within Italy Guiso, Sapienza, and Zingales (2004) show that economic agents form their expectations using not only objective information, but also cultural stereotypes. To answer this question, we use some other survey data that ask to a subset of the countries in our original data how much do they know citizens of other countries. We correlate the answer to this question and the extent they trust the same citizens. The correlation is negative, suggesting that information is not playing an important role in our results. That we show that trust does not reflect information, it does not necessarily imply that information has no role. Portes and Rey (2002), for example, relate portfolio investments and foreign trade patterns to differences in information. They measure availability of information as telephone traffic between two countries and as number of local branches of foreign banks. As for trust, these measures are potentially endogenous. We instrument our measures of trust to rule out the possibility of reverse causation and show that our results are robust. Our paper is also related to Morse and Shive (2003) and Cohen (2003). Morse and Shive (2003) relate portfolio choices to the degree of patriotism of a country. Cohen (2003) shows that employees bias toward investing in their own company is not due to information, but to some form of loyalty toward their company. Both these papers, thus, illustrate one specific dimension in which cultural biases can affect economic choices. Our paper uses a broader definition of cultural bias and tries to show the pervasiveness of its effects. The rest of the paper proceeds as follows. Section 1 introduces our data and show that 40% of the variation in trust is due not to objective characteristics, but to idiosyncratic opinions. Section 2 presents the effect of relative trust on trade, Section 3 on portfolio investments, Section 4 on foreign direct investments. Section 5 addresses the question whether the differences in 5

relative trust are explained by information or cultural stereotypes. Finally, Section 6 concludes. I Data A Measuring trust We obtain our measures of trust from a set of surveys conducted by Eurobarometer and sponsored by the European Commission. The surveys were designed to measure public awareness of, and attitudes toward, the Common Market and other European Community institutions, in complementary fashion (see the Data Appendix for details). They have been conducted on samples each time of about 1,000 individuals per country from a set of the European countries. The number of countries sampled varies over time: they were 5 in 1970 (France, Belgium, The Netherlands, Germany and Italy), when the first survey was conducted, and have grown to 18 in 1995, the last survey to which we have access (besides the 5 countries above, included are Luxembourg, Denmark, Britain, Northern Ireland, Greece, Spain, Portugal, East Germany, Norway, Sweden, Finland, and Austria). One distinct and unique feature of these surveys is that respondents have been asked to report how much they trust their fellow citizens and how much they trust the citizens of each of the countries belonging to the European Union. More specifically, they have been asked the following question: I would like to ask you a question about how much trust you have in people from various countries. For each, please tell me whether you have a lot of trust, some trust, not very much trust or no trust at all. In some of the surveys this same question was also asked with reference to citizens of a number of non EU countries, which include the United States, Russia, Switzerland, China, Japan, Turkey, and some Eastern and Central European countries (Bulgaria, Slovakia, Romania, Hungary, Poland, Slovenia and Czech Republic). In addition, each survey collects basic information on the demographic characteristics of the interviewed (year and country of birth, sex, marital status, level of education, income, family size, occupation, city size where respondent lives etc.) which makes it possible to filter trust data in order to control for differences in sample composition across countries. For our purposes, we have first re-coded the answers to the trust question setting them =1 ( no trust at all), = 2 (not very much trust), =3 (some trust), =4 (a lot of trust). We have then 6

aggregated responses by country and year computing the mean value of the responses to each survey. The result is a rectangular matrix of trust from European countries to European and non European countries which varies over time and in size. Obviously, for the EU countries the matrix is symmetric in each given sample year. Table 1 shows the trust matrix for the last year in the sample, the 1995. It shows the percentage fraction of citizens that report that they trust a lot their fellow citizens and the citizens of the other countries. The data refer to the European countries. Three features are noteworthy. First, there is considerable variation in the amount of individuals who in each country trust other individuals in other countries. The average response ranges from 4 percent (the fraction of Dutch that trust a lot Italians), to 69 percent (the fraction of Swedish that trust a lot the Norwegians). Second, individuals tend to trust more their fellow citizens, as the larger values on the main diagonal show. But there are exceptions to this pattern. For instance, Italians tend to trust Germans almost as much as they trust other Italians and Swedish trust their neighbors from Norway even more then themselves! This is hardly consistent with differences in trust being driven by differences in information concerning the other countries citizens (which should decay with distance), but can be explained by opinions on trustworthiness being highly affected by cultural stereotypes. Similarly, it is hard to explain only with information the fact that the British tend to trust the French as much as they trust the Italians and the Spanish (only 8 percent of the British trust fully the French) but much less than they trust the Belgians and the Dutch; and even more difficult would be to reconcile with information the fact that the French reciprocate, trusting the British as much as they trust (little) the Greeks. Finally, it is clear that there are systematic differences in how much a given country trusts and how much is trusted by others (see the last row and last column of Table 1). For instance, the Portuguese are those who trust the least and the Swedish those who trust the most (46 percent report they trust others a lot on average; furthermore, the Italians are the least trusted (11 percent trust them fully on average) and citizens of Norway the most (29 percent trust them fully on average). Obviously, these country of origin and country of destination effects may easily reflect systematic features of the country that trusts or is trusted. If all (or almost all) the variation in the data were explained by the attitude citizens of a country have to trust (being trusted), there would be little hope for relative trust to be able to affect the patterns of 7

bilateral trade. However, country of origin effects and country of destination effects leave al lot of variation unexplained. This effect is visible in Table 2 that shows the results of a regression of the average trust of a country versus others when full sets of country of origin dummies, country of destination dummies and years dummies are inserted. Characteristics of the country expressing and receiving trust can (controlling for time variation) at most explain between 44 and 64% of the variability in trust depending on how the aggregate trust of a country s citizens is computed. There remains a considerable portion of the trust to citizens of a country that cannot be explained by characteristics of either one of the two countries. It is this residual variation that we use to test whether trust affects the patterns of economic exchange. B Trade data The first variable we use to test whether trust matters for economic exchange is data on trade of goods and services assembled by the World Bank. The data contains trade, production and tariff data for 67 developing and developed countries at the industry level over the period 1976-1999. The sector disaggregation in the database follows the International Standard Industrial Classification (ISIC) and is provided at the 3 digit level (28 industries) for 67 countries. 2 As we will discuss, data at the industry level can help us discriminate between alternative channels through which trust can affect the patterns of trade since different goods entail different degrees of durability and exchanges of durables are more trust intensive. The sample statistics for the data are reported in Panel A of Table 3. C Portfolio data The second type of international exchange we focus on is data on international financial transactions. We use a dataset provided by Morningstar that has information of the geographical breakdown of equity investment (data are expressed in percentage) from mutual funds located in various European Countries. All types of funds that report their position to Morningstar have been included (including balanced and flexible funds, for example). In this dataset bonds investments are not included. Funds located in Luxembourg and Ireland affiliated with companies located in some European country have been excluded. The problem with this dataset is 2 See Alessandro Nicita and Marcelo Olarreaga (2001) for a description of the data. 8

that includes institutional investors. Institutions are likely to based their investment decision on cultural biases less than individual investors. This feature of the data may bias the results against us finding a significant effect. The sample statistics for the data are reported in Panel B of Table 3. D Direct investment data Statistics on FDI transactions and position are based on the OECD database developed by the Directorate for Financial, Fiscal and Enterprise Affairs. These statistics are compiled according to the concept used for balance and payments (flows) and international investment positions (stocks) statistics. The dataset covers 28 OECD economies from 1980 to 2000. Summary statistics are reported in Table 3, Panel C. E Controls We supplement our data on international exchange with additional variables that we use as controls in our regressions and that theory suggests are relevant in explaining the patterns of international trade. Besides complementing the data on trade with GDP data which gravity models suggest should affect bilateral trade, we augment our dataset with a measure of geographical distance, the percentage of people that speak the same language in each pair of countries, and an indicator variable equal to one if the pair of countries share a border and a measure of national stock market correlation. Summary statistics on these controls are reported in Table 3. The data Appendix describes in detail the source of these variables. F Other Eurobarometer data In order to test whether relative trust reflects information or cultural biases we use data from another Eurobarometer survey. Eurobarometer 38.0 was done in 1992 and focused on the current status and continuing development of the European Community (EC). In this survey respondents were asked to state how much they knew citizens of other countries. Specifically, the question asked was For each of the twelve countries of the European Community, please tell me if you know it very well, fairly well, not very well or not at all?. 3 We have coded the 3 This question was asked only in France, West Germany, Great Britain, Northern Ireland, Spain, and Italy. 9

answers setting them =1 ( not at all), = 2 (not very well), =3 (fairly well), =4 (very well). In the same survey, for the same subset of countries the following question was asked: Which countries of the European Community are in your opinion the most pleasant (maximus 3 answers possible)?. We coded 1 if country j was mentioned by citizen of country i and we use the percentage of times in which country j was mentioned by all the citizens of country i, as a measure of how much citizens of country i think citizens of country j are pleasant people. II The Effect of Trust on Trade Table 4 estimates the effect of relative trust on the amount of trade between two countries. The first three columns use the log of export over GDP as a dependent variable. Column I presents the standard trade regression, where we insert fixed effects for both importing and the exporting country. After we control for these, the distance between two countries negatively affects the level of export, while the presence of a common border and of a common language positively affect it. In column II we insert two measures of trust: the trust of the exporting country toward the importing country and the trust of the importing country toward the exporting one. In principle both could matter: the country receiving the good has to trust that it is shipped; the country that sells the good has to trust that it receives the payment from the buyer. Since we control for country fixed effects, these measures of trust are relative measures of trust. We find that both the trust of the exporting country and that of the importing country matter and the effect is of similar magnitude. At sample means, increasing the trust of both exporter and the importer by one standard deviation raises the share of of exports over GDP by 31 percent; if only the trust on one of the two partners were increased, the share of exports should increase by about 15 percent. As column III shows, these results are unchanged if we use instrumental variables. We instrument the trust of country i toward country j with the trust that all the other countries have toward country j. This instrument should be correlated with the characteristics that make country j more trustworthy for country i, but uncorrelated with the error (such as more trade between country i and j generates more trust between these two countries). That the instrumental variable estimates are virtually identical to OLS ones suggests that reverse causation is not really an issue. 10

Columns IV-VI replicate the estimates but using as a left hand side variable (the log of) total bilateral trade (exports plus imports) at the industry level scaled by industry output of the exporter. The results are very similar to those in the first three columns. Economically, raising the trust of both exporter and importers by one standard deviation increases total trade (as a share of output) by about 29 percent. In Table 5 we test whether the impact of trust on trade varies according to what theory would suggest. We split industries according to the level of technological sophistication of their products. Computers and machinery in general are durable goods, where the buyer s trust toward the seller would play a very significant role in the purchase decision. By contrast, in industries such oil, leather goods, and other commodities, the trust of the buyer is less important. In these industries, the most important moral hazard is that the buyer will not pay upon delivery. Hence the trust of the seller toward the buyer will be playing a more important role. We find that these predictions are borne out by the data. Relative trust toward the exporting country has a coefficient twice as large and statistically significant for the sample of sophisticated goods. By contrast, for non sophisticated goods the trust of the exporting country toward the importing one matters 30% more. In both instances the estimates are unchanged when we use instrumental variables. III The Effect of Trust on Portfolio Investment Table 6 investigates the effects of trust on portfolio allocations. Table 6 uses the pecentage of investments done by European mutual funds in other countries (Morningstar source). As a benchmark we use an enlarged portfolio model. A traditional portfolio model would only include the inverse of the covariance of stock market returns and the weight of the stock market of country i in the world portfolio. Since we include country fixed effects, this latter variable is absorbed by those. Our dependent variable is the percentage of investment of European mutual funds located in country i into country j. In columns I and II we only include information whenever country i is different from country j. Column I shows that the trust of the investing country toward the people of the destination country has a positive and statistically significant effect on the amount of in- 11

vestment. Inserting this variable increases the R-squared of the regression from 0.75 to 0.78. Raising the trust of the investing country toward all the partners by one standard deviation of trust (while keeping constant that towards the fellow citizens) would increase the share of investment abroad by four percentage points and reduce that in domestic assets by the same amount. This effect is unchanged when we instrument the trust of country i toward country j with the trust all the other country (except i) have toward country j (column II). In column III, IV, and V, we include also the observation of the main diagonal of portfolio investment (the percentage of portfolio investment that funds located in country i invest in country i. The results confirm our previous finding. The magnitude of the effect is larger. However, the coefficient changes sign and become insignificant, if we insert a dummy for the home country (column V). That trust loses statistically significance can be easily explained with the fact that the variable home is measured without error, while the variable trust is clearly noisy. Nevertheless, these results suggest that differences in trust alone cannot explain the home bias. IV The Effect of Trust on Foreign Direct Investment Table 7 reports the effect of trust on foreign Direct Investments. Column I reports the basic specification with country fixed effects, border, language, and distance. To this specification, in column II we add the mean trust of the people of the investing nation toward the people of the country where the investment takes place. The impact of trust is positive and statistically significant. This effect remains unchanged when we use instrumental variables. Economically, increasing the trust of the people in the country of origin versus the country of destination by one standard deviation, increases the stock FDI by 12.6 percent at sample means. V Information or cultural bias? So far, we have shown that the level of relative trust has a positive and significative impact on trade, portfolio investment, and foreign direct investments. To interpret these results we need to answer the question whether the relative differences in trust reflect information. At first sight, by looking at table 1 we may be tempted to think that this is the case. The main diagonal of the trust matrix shows that citizens of most European countries tend to trust their 12

own fellow citizens more than citizens of other countries. It is plausible to assume that people know their own fellow citizens more than citizens of other countries. So, we may be tempted to conclude that our measure of trust is a proxy for information. First of all, it is not clear why more information should lead always higher trust. In principle, more information could lead to higher or lower levels of trust. A possible explanation why higher information may correlate with higher trust is that on average when individuals share the same background with some people, they also are able to understand their culture better. Cornell and Welch (1996) show that under the plausible assumption that it is easier to screen people with the same background, individuals tend to discriminate in favor of people more similar to them. While this mechanism may explain why the main diagonal of our trust matrix exhibits high levels of relative trust, this story does not seem consistent with the pattern of relative bilateral trust off the main diagonal. For example, it is not plausible that the reason why British people should trust more Belgians than French is attributable to the greater amount of knowledge that British have of Belgians rather than of French. In order to understand how much of the relative trust reflects information, we use another Eurobarometer survey where citizens of some countries (the question was asked only in France, West Germany, Great Britain, Northern Ireland, Spain, and Italy) are asked how well they know citizens of the twelve European Countries. When we correlate the average country response with our measure of trust, we find that the correlation is negative (-.08) and insignificant. This result is not explained by the fact that some countries are more trusting than others. In table Table 8 we show the result of regressing the average level of trust of citizens of country i to citizens of country j on the average level of knowledge that citizens of country i have of citizens of country j. In the regression we include both country of origin and destination fixed effects. The findings confirm our previous results. The partial correlation coefficient is statistically insignificant. Interestingly, when we introduce the percentage of citizens of country i that have mentioned citizens of country j as the most pleasant citizens in European Union, we find that the coefficient of perceived pleasantness is positive and significant. A one-standard deviation increase in the perceived pleasantness increases trust by 18 percent. When we control for perceived pleasantness the coefficient of information becomes negative and significant, suggesting that the relationship between relative trust and information is negative, and not positive. 13

These findings suggest that our trust variable is not a proxy for information and that sympathy plays a big role in explaining relative trust among countries. VI Conclusions In this paper we document that the difference in relative trust between people of two countries affects the level of economic exchange between these countries, even after controlling for the objective characteristics of these two countries. This effect appears to be causal, since it varies according to what theory predicts and remains unchanged when we estimate it using IV. There are two possible explanations for this effect. The first, more consistent with economic orthodoxy, is that people from different countries differ in the information they have about foreigners. These differences translate into different priors, which affect the level of trade and investment. The second possible explanation is that cultural stereotyping affect people s attitudes. We show evidence consistent with the latter explanation. These results are consistent with the existing evidence that people trust toward others is affected by religious beliefs (Guiso et al, 2003) and their place of birth (Guiso et al, 2004). 14

References Cochrane, John. A cross-sectional test of an investment- based asset pricing model. Journal of Political Economy, 1996, 104, pp. 572.621. Cohen, Laurent 2003 Loyalty Based Portfolio Choice working paper University of Michigan. Cornell, Brad, and Ivo Welch 1996 Culture, Information and Screening Discrimination. The Journal of Political Economy 104-3, 542-571. French, Kenneth and Poterba, James. A cross-sectional test of an investment- based asset pricing model. Journal of Political Economy, 1996, 104, pp. 572.621. Guiso, Luigi; Sapienza, Paola, Zingales, Luigi. 2003, People s opium? Economic Attitudes Journal of Monetary Economics 50:225-282. Religion and Guiso, Luigi; Sapienza, Paola, Zingales, Luigi. 2004, forthcoming, The Role of Social Capital in Financial Development The American Economic Review, forthcoming McCallum, John. National Borders Matter: Canada-U.S. Regional Trade Patterns American Economic Review. 1995, 85, pp. 615-23. Morse, Adair and Shive, Sophie Patriotism and the Home Equity Bias, 2003, working paper University of Michigan. Nicita, Alessandro and Olarreaga, Marcelo. Trade and Production, 1976-1999. World- Bank. Obsfeld, Maurice and Rogoff, Kenneth. The Six Major Puzzles in International Macroeconomics: Is There a Common Cause?. in Bernanke, Ben S.; Rogoff, Kenneth, eds. NBER macroeconomics annual 2000. Volume 15. Cambridge and London: MIT Press, 2001, pp. 339 90. OECD. Policies and International integration: influences on trade and foreign direct investment. Annex 2: Foreign Direct Investments and other miscellaneous data, OECD, Paris. 15

Table 1: The trust matrix The matrix shows the percentage share of citizens of a given European country who report they trust a lot their fellow citizens and citizens of the other European countries. The last column is the average percentage share of those that in a given country report they trust a lot and gives a summary measure of how much citizens of a given country trust citizens of their own or other countries; the last row shows the average share of the citizens of different countries that report they trust a lot their own citizens or the citizens of a given country. It gives a summary measure of how trustworthy are the citizens of the country in each column. Fra Bel NL Ger Ita Lux Den Ire UK Gre Spa Por Nor Fin Swe Aus Mean Fra 33 23 18 16 7 23 23 18 10 9 12 11 19 16 20 11 17 Bel 23 40 24 19 8 39 23 15 18 9 11 10 19 18 20 18 20 NL 12 29 37 15 4 34 35 15 21 8 8 9 34 30 36 14 21 Ger 21 17 22 57 8 24 23 13 15 11 14 11 24 20 26 26 21 Ita 12 9 14 19 19 10 13 8 11 7 11 5 15 16 18 11 12 Lux 21 16 22 18 11 53 17 11 12 9 12 12 19 19 19 22 18 Den 18 30 40 30 11 32 46 26 35 14 13 13 54 34 47 34 30 Ire 15 15 20 18 10 15 18 43 18 9 10 10 14 13 13 14 16 UK 8 17 29 15 8 17 27 15 39 11 8 12 22 18 20 15 18 Gre 26 19 21 18 12 18 21 17 16 51 21 17 9 10 13 8 19 Spa 14 17 20 20 15 17 17 14 10 13 49 14 19 14 20 13 18 Por 21 10 11 11 7 11 10 7 12 6 13 44 6 6 6 5 12 Nor 21 31 37 27 12 32 57 27 38 14 13 13 61 - - - 29 Fin 23 29 33 27 10 27 42 25 34 15 12 13 55 72 47 41 32 Swe 34 42 48 41 24 45 63 45 53 31 29 33 69 59 64 58 46 Aus 17 25 23 36 12 30 21 15 15 15 14 14 27 24 29 65 24 Mean 20 23 26 24 11 27 29 20 22 15 16 15 29 25 27 24 16

Table 2: Bilateral trust and country of origin and destination characteristics The table shows how much of the trust of the average trust of a country s citizens versus the other countries citizens is explained by observed and unobserved characteristics of the country receiving and giving trust. Mean trust is the average trust across individuals of a given country; median trust uses the median to aggregate across individuals; share of individuals trusting a lot is the fraction of interviewed individuals in a given country that report they trust a lot the citizens of another country. Fraction of individuals Mean trust Median trust trusting a lot Year 1970-0.0813-0.0836-0.0048 (0.0438) (0.0688) (0.0171) Year 1976 Excluded Excluded Excluded Year 1980.0635** 0.0221.0284*** (0.0266) (0.0418) (0.0104) Year 1986.1345***.1949***.0395*** (0.026) (0.0408) (0.0101) Year 1990 0.2080***.3012***.0429*** (0.0254) (0.0399) (0.0099) Year 1993.1561***.1908***.0308*** (0.0265) (0.0417) (0.0103) Year 1994.1322***.1932***.0293*** (0.0265) (0.0417) (0.0103) Year 1996 0.0264 0.0515 0.0047 (0.0254) (0.04) (0.0099) Dummies for country F(17, 1694) = 31.84 F(17, 1694) = 9.49 F(17, 1694) = 25.98 of origin: F-test p-value= 0.000 p-value = 0.0000 p-value = 0.0000 Dummies for country F(28, 1694) = 88.41 F(8, 1694) = 39.51 F(28, 1694) = 33.67 of destination: F-test p-value=0.0000 p-value=0.0000 p-value=0.0000 Adjusted R-square 0.6358 0.4302 0.4369 N. of observations 1747 1747 1747 17

Table 3: Summary Statistics The description of the variables is in the Appendix. Panel A contains summary statistics for the trade dataset. Panel B shows summary statistics for the portfolios datasets. Panel C shows summary statistics for the foreign direct investment data. Panel A: Trade data Mean Median Std. Dev. Min Max Observations Export over output to individual partner country by ISIC 0.020208 0.004418 0.067478 0 2.23075 6218 Export plus import over output to partner country by ISIC 0.064641 0.015071 0.239442 0 8.449433 6218 Distance 1330.98 1243 730.3393 176 3367 6218 Border 0.107108 0 0.309276 0 1 6218 Language 260.4374 0 1563.881 0 10000 6218 Average trust from country to each partner 2.754606 2.737242 0.309339 2.008086 3.569197 6218 Average trust from all the other countries to each partner (instrument) 2.731168 2.697652 0.155096 2.512563 2.963402 6218 Panel B: Morningstar Data Mean Median Std. Dev. Min Max Observations Percentage invested in partner country 0.090476 0.035335 0.15527 0.00024 0.71733 86 Inverse Covariance of stock market returns -1.6E-05-5.20E-07 0.000308-0.00255 0.00062 86 Border 0.255814 0 0.438877 0 1 86 Language 144.0581 0 775.3672 0 6800 86 Distance 1037.419 998.5 723.9855 1 2955 86 Average trust from country to each partner 3.001507 2.952854 0.330547 2.353828 3.691296 86 Average trust from all the other countries to each partner (instrument) 2.943187 3.024507 0.216038 2.487168 3.197232 86 Panel C: Foreign Direct Investments (OECD) Mean Median Std. Dev. Min Max Observations Outward stock of FDI (log) 20.78821 21.18551 2.367365 9.608683 25.29763 1130 Distance 3662.118 1307.406 5449.992 173.011 19838.47 1130 Border 0.060177 0 0.23792 0 1 1130 Language 20.69027 0 403.9575 0 10000 1130 Average trust from country to each partner 2.762178 2.792833 0.32253 2.075855 3.462598 1130 Average trust from all the other countries to each partner (instrument) 2.746398 2.743249 0.012807 2.725259 2.782643 1130 18

Table 4: Effect of Trust on Trade The dependent variable is the log of the export volume divided by output. Exports and output are industry-level exports bilateral exports (industry classification is three digit ISIC code). Control variables include the log of distance in miles (between largest cities); dummies for common land border between exporter and importer and for language. Language is an indicator variable measuring the fraction of same language speakers in the two countries. All regressions include fixed effects for exporting country and importing country. The standard errors reported in parentheses are corrected for the potential clustering of the residual at the exporter country level. The symbols ***, **,* mean that the coefficient is statistically different from zero respectively at the 1,5, and 10 percent level. EXPORTS OVER OUTPUT (EXPORTS+IMPORTS)/OUTPUT OLS OLS IV OLS OLS IV Mean trust of people in exporting 0.5531** 0.5534** 0.3941*** 0.3958*** country to people in importing country (0.1830) (0.1828) (0.1256) (0.1255) Mean trust of people in importing country 0.4695** 0.4713** 0.5401*** 0.5414*** to people in exporting country (0.1925) (0.1924) (0.1502) (0.1503) Distance -0.9863*** -0.8998*** -0.8996*** -0.7822*** -0.7241*** -0.7239*** (0.1972) (0.1603) (0.1602) (0.1835) (0.1543) (0.1543) Border 0.5357* 0.5490* 0.5490* 0.3580* 0.3565* 0.3565* (0.2888) (0.2873) (0.2873) (0.1825) (0.1765) (0.1765) Language 0.0002*** 0.0002*** 0.0002*** 0.0002*** 0.0002*** 0.0002*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) Exporting country fixed effects YES YES YES YES YES YES Importing country fixed effects YES YES YES YES YES YES Year fixed effects YES YES YES YES YES YES Observations 6454 6454 6454 7210 7210 7210 R-squared 0.619 0.623 0.623 0.686 0.690 0.690 19

Table 5: Effect of Trust on Trade: Industry Splits The dependent variable is the log of the export volume divided by output. Exports and output are industry-level exports bilateral exports (industry classification is three digit ISIC code). Control variables include the log of distance in miles (between largest cities); dummies for common land border between exporter and importer and for language. Language is an indicator variable measuring the fraction of same language speakers in the two countries. All regressions include fixed effects for exporting country and importing country. The standard errors reported in parentheses are corrected for the potential clustering of the residual at the exporter country level. The symbols ***, **,* mean that the coefficient is statistically different from zero respectively at the 1,5, and 10 percent level. Non-sophisticated goods Sophisticated goods OLS IV OLS IV Mean trust of people in exporting 0.2870 0.2871 0.6311*** 0.6314*** country to people in importing country (0.1796) (0.1794) (0.1876) (0.1875) Mean trust of people in importing country 0.6271*** 0.6290*** 0.4395* 0.4412* to people in exporting country (0.1600) (0.1600) (0.2137) (0.2138) Distance -0.6696*** -0.6695*** -0.9703*** -0.9701*** (0.1389) (0.1389) (0.1726) (0.1726) Border 0.4924 0.4925 0.5775* 0.5775* (0.2776) (0.2777) (0.2944) (0.2944) Language 0.0001*** 0.0001*** 0.0002*** 0.0002*** (0.0000) (0.0000) (0.0000) (0.0000) Exporting country fixed effects YES YES YES YES Importing country fixed effects YES YES YES YES Year fixed effects YES YES YES YES Observations 1429 1429 4785 4785 R-squared 0.745 0.745 0.594 0.594 20

Table 6: Effect of Trust on Portfolio Investment The dependent variable measures the percentage of net portfolio investment of a given country into another country. Specifically, the dependent variable is the stock of cross-border holdings of equities and long- and short-term debt securities valued at market prices prevailing at the end of 2001 (from Morningstar data) divided by the sum of all foreign equity holdings plus market capitalization- foreign liabilities. Control variables include the inverse of the covariance of stock market returns, calculated using monthly data for each country (DATASTREAM); the log of distance in miles (between largest cities); dummies for common land border between the two countries and a variable measuring language. Language is an indicator variable measuring the fraction of same language speakers in the two countries. All regressions include fixed effects for the country investing abroad and for the receiver country. The standard errors reported in parentheses are corrected for the potential clustering of the residual at the country level. The symbols ***, **,* mean that the coefficient is statistically different from zero respectively at the 1,5, and 10 percent level. OLS OLS OLS IV IV Mean trust of people in the country investing 0.1323** 0.1323** 0.4981*** 0.4981*** -0.0211 to people in destination country (0.0562) (0.0562) (0.0670) (0.0670) (0.0344) Inverse Covariance of stock market returns -38.2295* -38.2295* 40.3196 40.3196 23.5253* of country of origin and destination (20.7495) (20.7495) (27.0270) (27.0270) (11.3498) Home 0.4802*** (0.0534) Investing country fixed effects YES YES YES YES YES Destination country fixed effects YES YES YES YES YES Observations 97 97 106 106 106 R-squared 0.782 0.782 0.475 0.475 0.936 21

Table 7: Effect of Trust on Foreign Direct Investments The dependent variable is the log of outward investment (stocks). The percentage of net investment of a given country into another country. Control variables include the log of distance in miles (between largest cities); dummies for common land border between the two countries and a variable measuring language. Language is an indicator variable measuring the fraction of same language speakers in the two countries. All regressions include fixed effects for the country of origin and for the destination country. The standard errors reported in parentheses are corrected for the potential clustering of the residual at the country level. The symbols ***, **,* mean that the coefficient is statistically different from zero respectively at the 1,5, and 10 percent level. OLS OLS IV Mean trust of people in country of origin 0.3926* 0.3902* to people in destination country (0.2032) (0.2031) Border 0.3111 0.3071 0.3071 (0.3383) (0.3352) (0.3353) Language -0.0001-0.0001-0.0001 (0.0001) (0.0001) (0.0001) Distance -1.1948*** -1.1702*** -1.1703*** (0.2432) (0.2363) (0.2364) Country of origin fixed effects YES YES YES Country of destination fixed effects YES YES YES Years fixed effects YES YES YES Observations 1130 1130 1130 R-squared 0.719 0.721 0.721 22

Table 8: Cultural bias or information? The dependent variable is the average trust of citizens of country i toward citizens of country j. Trust is calculated by taking the average response to the following question: I would like to ask you a question about how much trust you have in people from various countries. For each, please tell me whether you have a lot of trust, some trust, not very much trust or no trust at all. The answers are coded in the following way:=1 ( no trust at all), = 2 (not very much trust), =3 (some trust), =4 (a lot of trust). Information is the average answer to the question: For each of the twelve countries of the European Community, please tell me if you know it very well, fairly well, not very well or not at all?. We have coded the answers setting them =1 ( not at all), = 2 (not very well), =3 (fairly well), =4 (very well). Perceived pleasantness is the percentage of times in which country j was mentioned by citizens of country i in the following question: Which countries of the European Community are in your opinion the most pleasant (maximus 3 answers possible)?. All regressions include fixed effects for the country of origin and for the destination country. The standard errors reported in parentheses are corrected for the potential clustering of the residual at the country level. The symbols ***, **,* mean that the coefficient is statistically different from zero respectively at the 1,5, and 10 percent level. Information 0.0745-0.2697* (0.1006) (0.1156) Perceived pleasantness 1.1586*** (0.1883) Country of origin fixed effects YES YES Country of destination fixed effects YES YES Observations 55 55 R-squared 0.747 0.860 23

Appendix A: The Data Data Appendix A.1. - The Eurobarometer surveys The Eurobarometer surveys are the products of a unique program of cross national and cross temporal social science research. The effort began in early 1970, when the Commission of the European Community sponsored simultaneous surveys of the publics of the European Community. These surveys were designed to measure public awareness of, and attitudes toward, the Common Market and other European Community institutions, in complementary fashion. They also probed the goals given top priority for one s own nation. These concerns have remained a central part of the European Community s research efforts which were carried forward in the summer of 1971 with another six-nation survey that gave special attention to agricultural problems. These themes were of central interest again in a survey of the publics of the European Community countries then nine in number carried out in September 1973. After 1973, the surveys took on a somewhat broader scope in content as well as in geographical coverage, with measures of subjective satisfaction and the perceived quality of life becoming standard features of the European Community public opinion surveys. In 1974, the Commission of the European Community launched the Eurobarometer series, designed to provide a regular monitoring of the social and political attitudes of the publics of the nine member-nations: France, Germany, the United Kingdom, Italy, the Netherlands, Belgium, Denmark, Ireland, Luxembourg. These Eurobarometer are carried out in the spring and fall of each year. In addition to obtaining regular readings of support for European integration and the perceived quality of life, each of the Eurobarometer has explored a variety of special topics. Also, attitudes toward the organization and role of the European Parliament have been explored in each Eurobarometer beginning with Barometer 7 in the spring of 1977. The Eurobarometer surveys have included Greece since Autumn 1980, Portugal and Spain since Autumn 1985, the former German Democratic Republic since 1990, Norway (irregularly) since the fall of 1990, Finland since the spring of 1993, and Sweden and Austria since the fall of 1994. Table A1 shows the number of observations from each country in our dataset, the number of years the country was samples and the years in which was sampled. Code Country sampled Number of observations N. of years present in survey Years present 1 France 11,464 8 1970,1976,1980,1986, 1990,1993,1994, 1995 2 Belgium 9,693 8 1970,1976,1980,1986, 1990,1993,1994, 1995 3 The Netherlands 10,123 8 1970,1976,1980,1986, 1990,1993,1994, 1995 4 Germany 11,332 8 1970,1976,1980,1986, 1990,1993,1994, 1995 5 Italy 11,016 8 1970,1976,1980,1986, 1990,1993,1994, 1995 6 Luxembourg 3,173 7 1976,1980,1986,1990,1993,1994, 1995 7 Denmark 7,020 7 1976,1980,1986,1990,1993,1994, 1995 8 Ireland 7,014 7 1976,1980,1986,1990,1993,1994, 1995 9 Great Britain 7,498 7 1976,1980,1986,1990,1993,1994, 1995 10 Northern Ireland 2,158 7 1976,1980,1986,1990,1993,1994, 1995 11 Greece 6,014 6 1980,1986,1990,1993,1994, 1995 12 Spain 5,031 5 1986,1990,1993,1994, 1995 13 Portugal 4,995 5 1986,1990,1993,1994, 1995 14 East Germany 3,210 3 1993,1994, 1995 15 Norway 994 1 1993 16 Finland 2,065 2 1993, 1995 17 Sweden 1,010 1 1995 18 Austria 1,995 1 1995 24