1
The Effect of Trust on Trade:
When does it no longer matter?
Shu Yu Supervisor: Prof. Jakob de Haan
Prof. Sjoerd Beugelsdijk
Abstract
This paper shows that the effect of trust on trade is conditional on the dissimilarity of judicial quality between two trading countries. To avoid using a trust measure that instead captures the well-functioning of institutions, we focus on countries’ current judicial quality, the effectiveness of countries’ judicial system, and use a historical proxy for generalized trust, which is largely cultural-rooted and historically determined. We found that when two countries are similar in judicial quality, the effect of trust on trade disappears. In contrast, if one country has a much more effective judicial system than the other, trust significantly influences trade volumes. Our findings suggest that the uncertainty in judicial quality in the setting of international transactions will have a negative impact on trade. Well-functioning institutions can foster trade via familiarity and certainty.
Key Words: Trust, international trade, judicial quality
2
I.
Introduction
Does trust influence economic exchange? According to Guiso, Sapienza and Zingales (2009), higher bilateral trust does lead to more trade between European countries. When traders from one country show a high level of trust in people from another country, they are more willing to believe that traders from that country will take actions that are beneficial rather than detrimental to them(Child, 2001). If true, trust has a similar positive effect on trade volumes as a common border and a common language. It influences trade by eliminating transaction costs, particularly contract enforcement costs. Since contracts are usually incomplete, traders have an incentive to expropriate the rights of the other trader. To prevent opportunistic behavior, such as hold-up, traders have to divert resources to monitoring and contract enforcement activities. Costs induced by these activities will shy away potential traders (Anderson, 1999). When traders trust their counterparts, they will be more certain about a fair distribution at the end of the transaction.
Although bilateral trust can reduce this type of transaction costs involved in trade, it is not the only way to mitigate it. The formal channel to reduce it is through the judicial system, where courts interpret and apply the law to resolute disputes. When judicial systems impose strong contract enforcement, will agents still base their decision on bilateral trust?
3
Instead of verifying the effect of trust on trade, this study shows that the effect is conditional on the dissimilarity of judicial quality between two trading countries. By using a series of Eurobarameter surveys, we obtain a proxy for generalized bilateral trust. Since the proxy for trust is largely cultural-rooted and historically determined, the international trade statistics after the survey period are used to test the hypothesis for a group of European countries. Empirical results generated by gravity model show that when two countries do not differ in judicial quality, the effect of trust on trade disappears. However, when two countries have different levels of judicial quality, the effect of trust on trade turns significant.
By using a proxy for bilateral trust, which is mainly determined by historical factors (Guiso et al., 2009), we avoid the risks of only capturing the well-functioning of institutions (Beugelsdijk, 2006). Since bilateral trust is a kind of perception rooted in culture, it is more influenced by cultural determinants than current functioning of institutions. Some may argue that the cultural determinants that influence bilateral trust, such as legal origin, can influence legal system as well. However, it should have a strong impact on the type of legal system but a minimal impact on the effectiveness of the judicial system. The results of this study have policy implications for trade development. Countries that receive a low level of trust can eliminate the unfavorable effect of trust on trade through improving the effectiveness of laws applied in courts. When countries share the same level of judicial quality, bilateral trust will no longer influence trade.
The paper is structured as follows. Section II reviews recent literature linking trust, judicial quality and international trade. In Section III, the data will be described and the empirical strategy will be explained in detail. Section IV tests the hypothesis with a gravity model. Robustness checks will be reported in Section V. Finally, Section VI will conclude and discuss some policy implications.
II.
Literature review
4
economic mechanisms, such as investment, financial market development, international trade, and eventually economic growth.
Trust, which is based on deeply rooted cultural traditions (as suggested by Fukuyama, 1995; Putnam et al, 1993; and Guiso et al, 2009), is defined as the willingness to permit others’ actions to influence one’s welfare (Sobel, 2002). More precisely, an agent is willing to take a risk on the actions of others based on the belief that potential trustees will “do what is right” (Hoffman, 2002).
This belief influences trade through its impact on mitigating transaction costs. During all stages of a trade transaction, trust can eliminate transaction costs in different forms (den Butter and Mosch, 2003). In the contact stage of a potential transaction, agents are looking for information about the reliability of trading partners and the price and quality of the goods. The distribution of high-quality information among mutual-trusting members of a network is certainly beneficial (Casson, 1997). Empirical studies, such as Rauch (2001) and Rauch and Trinade (2002), found co-ethnic business networks to reduce the information costs needed for trading more differentiated goods.
In the contract stage of trade, trust can reduce negotiation costs by limiting the number of outcomes that should be considered and listed. Contracting is a time-consuming task. It is impossible to go through all possible outcomes (den Butter and Mosch, 2003). By trusting the counterpart, it is unnecessary for trading partners to specify how to proceed under each circumstance, which makes negotiations easier and faster. In the last stage of trade (control), trading partners have to decide whether and to what extent monitoring and contract enforcement are needed. With high mutual trust, trading partners are less likely to divert resources to monitoring and contract enforcement activities-through briberies, private security services and other tactic methods (Knack and Keefer, 1997). The positive effect of bilateral trust on international trade is empirically verified by Guiso, Sapienza and Zingales (2009). By using a series of Eurobarometer surveys, they show that a pair of European countries with high mutual trust tends to trade more intensively, even after controlling for other characteristics of the two countries.
5
tariffs do. By classifying traded goods into differentiated and non-differentiated goods, Ranjan and Lee (2004) verify the effect of contract enforcement on the volume of trade in both types of goods while the impact is larger for differentiated goods. Recent literature examines whether institutional quality determines a country’s comparative advantage. In particular, Levchenko et al. (2007) demonstrate that countries with better institutions specialize in goods that are institutionally dependent. Nunn (2007) reports that countries with good contract enforcement specialize in the production of goods for which relationship-specific investments are most important. According to Nunn (2007), contract enforcement can explain more of the trade pattern than physical capital and skilled labor in combination.
Unlike domestic trade, international trade contracts must be enforced across borders. Since international trade involves multiple governance systems, the effectiveness of both domestic and foreign institutions should matter. Existing literature suggests that sharing similar institutions promotes trade. According to Anderson (1999), if traders in both countries experience similar level of institutional quality, they will be more familiar with other country’s formal procedures and more able to cope with foreign governance. When agents from effective judicial systems trade with each other, they know what can be expected from the court. Agents from ineffective judicial systems know to secure their benefits via other informal channels (such as briberies) in another ineffective judicial system. When trading with countries with different level of judicial quality, the lack of knowledge in coping with another system will create problems for agents. De Groot et al. (2004) show that having a similar institutional framework promotes bilateral trade by 13% on average.
6
III.
Empirical Strategy
To test whether the effect of culture-rooted bilateral trust on trade is conditioned by judicial quality, our empirical strategy consists of two steps. First, we deduct the time dimension of a series of social surveys conducted discontinuously between 1970 and 1996. This approach results in a generated proxy for trust that captures the time-invariant part of bilateral trust. Then, the trust proxy will be used to test the marginal effect of bilateral trust on international trade after 1995. Since the proxy for trust is generated before the trade period, we make sure the causality runs from trust to trade and the trust proxy does not capture institutions’ functioning situation. The use of this time-invariant trust variable in the second stage is validated by the fact that trust is largely determined by historical factors, which means its time-invariant part will not alter after the survey period.
3.1 Time-invariant trust
The measure of bilateral trust used in this study is obtained from a subsample of surveys conducted by Eurobarometer in various years and countries. The first survey was conducted in 1970 in only five countries while the last survey in 1996 included 17 countries. A detailed description of the selected subsample is reported in Appendix-E. In view of data availability and the degree of homogeneity in trading rules and income level, we restrict the analysis to 16 countries1 belonging to the European Economic Area (EEA): 15 European Union members plus Norway.
In each country, about 1,000 representative individuals of age sixteen2
were asked the following: “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.” According to Guiso, Sapienza, and Zingales (2004), this type of question measures generalized trust, which shows the trust people have towards a random member of an identifiable group. In a separate survey in Guiso et al. (2009), evidence showed that the Eurobarometer survey question indeed reports the subjective probability that a random person is trustworthy. It differs from personalized trust that people develop through rounds of repeated interactions (Greif, 1993). Since this study examines the general pattern of trade with random agents, it is more suitable to use generalized trust.
1
France, Belgium, The Netherlands, Germany, Italy, Denmark, Ireland, Great Britain, Northern Ireland, Greece, Spain, Portugal, Norway, Sweden, Finland, and Austria; Moreover, weighted average of the data gathered from Great Britain and Northern Ireland are further combined to represent UK (see Guiso et al., 2009).
7
Answers to the above trust question are recoded to 1 (no trust at all), 2 (not very much trust), 3 (some trust) and 4 (a lot of trust). Then, they are recoded by pairs of countries and year. Since countries differ in their years of participation and the survey methods varied over time, there are potential measurement errors that can lead to biased results. Thus, we follow Betrand, Duflo, and Mullainathan (2004) and collapse the dataset by averaging over time the residuals of regressing Trustijt4on calendar-year dummies t. Then, the generated averaged residuals, Trustij, will be used to show the trust a representative individual from country i has towards a random individual from country j. Cases are dropped when i is equal to j. The value of Trustij is reported in Table-1.
[Insert Table-1 here]
As suggested by the data in Table-1, there are systematic differences in how much a country projects trust and how much a country is trusted. People from Scandinavian countries and the Netherlands tend to project more trust towards people from other countries. It may suggest that people excessively apply the level of trustworthiness of their own countrymen to people from other countries, which is consistent with the experimental founding in Sapienza et al.(2007). Meanwhile, people from the south part of Europe receive a lower level of trust from other European countries. In all, Sweden receives most time-invariant trust from other countries in the sample while Portugal receives the least trust from other countries.
Guiso, Sapienza and Zingales (2009) report that trust is not only affected by the characteristics of the country being trusted, but also by the cultural aspects of the match between trusting country and trusted country. In combination with country-fixed effects, factors, such as history of conflicts and religious, genetic, and somatic similarities, can account for more than 80 percent of the variation in the generalized trust proxy. Since both these factors and country-fixed effects are time-invariant, bilateral trust shall not vary extensively over time.
In order to test whether the trust variable used here is indeed time-invariant, we follow Guiso, Sapienza and Zingales (2009) and improve upon their model specification. The results support their conclusion that bilateral trust is largely determined by fixed historical factors. Detailed empirical results
4
8
and model specifications are reported in Appendix-B. Thus, it is valid to use this generated proxy to explain trade flows after the survey period.
3.2 Gravity Model
The data on commodity trade is obtained from Feenstra et al. (2005)3
, which is constructed from United Nations trade data and is available from 1960 to 2000. The advantage of this dataset is that it provides bilateral trade statistics among the 17 countries included in this study. Primacy was given to the commodity trade flows reported by the importing country, assuming that these are more accurate than those from exporting countries (Feenstra et al, 2005). Since the focus of this study is on the aggregate trade flow between countries, we chose the total trade volume4
between countries in thousands of US dollars directly reported by Feenstra et al. (2005).
Since we are interested in the effect of judicial system on eliminating transaction costs, we judge judicial quality based on how effective the protection of property rights and contract enforcement are in one country given the existing laws. As our primary measure of judicial quality, the “rule of law” indicator from Kaufmann, Kraay and Matruzzi (2009) captures perceptions of individuals in the effectiveness and predictability of the judiciary and the enforcement of contracts. One advantage of this measure is that it reflects the effectiveness of judicial system (de facto judicial quality), which differs from de jure judicial quality. Another advantage is that both “rule of law” and bilateral trust measure perception of agents. Kaufmann et al. (2009) provide data for 1996, 1998, 2000 and sequential years afterwards. To avoid the complication induced by the launch of Euro, the sample of trade flows is restricted to 1996 -2000. In order to test the sensitivity of the results to the use of alternative measures of judicial quality and the time period of study, an alternative measure of judicial quality, “legal structure and security of property rights”, developed by Gwartney and Lawson (2009), is used in Section V as a robustness check.
The gravity model used here is developed from Anderson and van Wincoop (2003), which is consumption-based and assumes that trade is driven by love of variety instead of resource
3 Available at www.nber.org/data. For 1984-2000, Feenstra et al. (2005) obtained data from UN Comtrade.
4 It is not corrected for intra-firm trade. However, we think firms’ location decision might be determined by the same
9
endowments. Thus, the trade pattern tested in this study is driven by the demand side. The regression model is: ijt t j t i ij jt it ijt ij ijt ij
jit Trust Q Trust Q GDP GDP Dist k Year Year
LnTrade =β1 +β2 +β3 * +α1 +α2 +α3 + * +λ * +ε
(3) Where LnTrade jit represents the total trade volume (in logarithm form) from country j, the exporting country, to country i, the importing country, in year t; Trustijis the averaged residuals of regressing the mean of answers to the trust question from country i to country j in survey year t on calendar-year dummies t, Qijt is the difference between judicial quality of country i and country j in year t, GDPi(j)t denotes the logarithm of real GDP per capita in country i (j) in Year t, and Distij measures the physical distance between country i and country j. By using the interactive terms between the importing-country fixed effect, ki, the exporting-country fixed effect, λj, and time trend, Yeart, we take those time variant frictions in trade volume into account. The data for real GDP per capita is obtained from Penn World Table 6.3, while data for geographic distance come from Jon Haveman’s website5
. Although there are alternative measures of geographic distance, the results of this study do not change when using a difference distance measure6.
The marginal effect of bilateral trust on trade, has the following form:
ijt ij jit Q Trust LnTrade * 3 1 β β + = ∂ ∂ (4)
The corresponding estimated standard error is:
ijt ijt Q Q var( ) 2*cov( , )* ) var( ˆ 2 3 1 3 1 β β β β σ = + + (5)
By estimating the marginal effect of bilateral trust on trade and its corresponding standard error, we can test whether the dissimilarity in judicial quality creates uncertainty so that bilateral trust has a significant effect on trade.
5 Measured in the log of distance in kilometers between the major cities (mainly, capital cities) of the respective
countries. See http://www.macalester.edu/research/economics/page/haveman/Trade.Resources/ tradeconcordances.html
6 We also tried other distance measures obtained from http://www.cepii.fr/anglaisgraph/bdd/distances.htm.
10
IV.
Empirical Results
Before investigating the effect of trust on trade, we start our estimation with a standard gravity model. All empirical results are reported in Table-3, and all model specifications include an interaction between fixed effects for the exporting country and year and for the importing country and year. In Column (1), trade flows from country j to country i are regressed upon the levels of their Gross Domestic Product and the geographic distance between them. The results confirm other studies’ finding that GDP positively and significantly affects trade while distance has a negative impact on trade. Moreover, it suggests that the GDP level of the exporting country has a larger impact on trade than the GDP level of the importing country. While a 1% increase in the exporter’s GDP will raises trade by 1.329%, the equivalent increase in importer’s GDP raises trade by 0.880% on average.
[Insert Table-3 here]
In Column (2), we add the proxy for trust (Trustij) from the importer towards the exporter to the gravity model. The proxy has no significant impact on trade flows. As a robustness check, an alternative proxy for bilateral trust (AvTrustij) is used in Column (3), which is simply the mean of Trustijt among all the survey years country i participated. The coefficient for AvTrustij does not significantly differ from zero, which indicates the insignificant impact is not driven by the proxy we constructed. Thus, we use Trustij in the following model specifications.
11
systems would like to import more from countries with ineffective judicial systems but specialized in producing other type of goods.
As Acemoglu, Johnson, and Robinson (2005) found, international trade can have an impact on the development of legal system as well. To test whether the results in Column (4) are driven by a simultaneity problem, we include the lagged value of the exporter’s judicial quality (Qjt-1) in Column (5). The coefficient for Qjt-1 has the same negative sign and significant level as the one for Qjt. Since the results in Column (5) agree with those in Column (4), the simultaneity problem is not severe in this sample.
To test whether the difference in two trading countries’ judicial quality has an impact on trade volumes, we use the difference between Qit and Qjt (which is denoted as Qijt) in Column (6). When Qijt is close to zero, it means that the courts in two trading countries are perceived to be similar in terms of effective contract enforcement and property protection. In contrast, if its value differs from zero, one country is superior to the other in terms of effectively protecting property rights and enforcing contracts. The coefficient for Qijtis negative and significant at the level of 10 percent. It implies that the difference in judicial systems’ effectiveness has a direct negative impact on international trade. When the lagged value of Qijt (Qijt-1) is included in Column (7), the results show that the difference in two trading countries’ judicial quality in the previous year does not influence trade volumes. Since using Qijt provides us with more observations, Qijt is used in the following model specifications.
Figure-1 Marginal effect of trust from the demand side on commodity trade flows
12
Note: the figure shows the marginal effect of Trustij on commodity trade flows for various values of Qijt. Furthermore, (in green) the 95% confidence interval for the marginal effect is plotted. The dots refer to the observations indicate various values of Qijt for all observations included the sample and their corresponding marginal effects. Kaufmann et al. (2009) is the data source for Qijt.
When adding Trustij and Qijt to the standard gravity model in Column (8), they both turn out to be insignificant. By further including the interactive term between Trustij and Qijt, we obtain the baseline model of the study in Column (9). Although Trustij is not significant, both Qijt and the interactive term are significant at least at the level of 5%. The marginal effect of trust on trade is plotted in Figure-1 with the 95% confidence interval.
According to Figure-1, the marginal effect of bilateral trust on trade is not constantly positive. It has a downward slope, which suggests that the marginal effect of trust on trade will drop when the importer’s judicial quality becomes increasingly superior to the exporter’s judicial quality.
Moreover, the results support the conjecture that when there is uncertainty concerning judicial quality, bilateral trust will significantly influence trade volume. When approaching the right end of the figure where the importer has a more effective judicial system, both the lower bound and the upper bound of the confidence interval are below zero, indicating higher trust from the importing country leads to lower trade flows from the exporting country. In contrast, at the left end of the figure, both the lower bound and the upper bound of the confidence interval are above zero, suggesting a significant and positive impact of trust on trade. Meanwhile, when Qijt is close to zero, the confidence interval includes zero. It means the effect of trust on trade becomes insignificant when two countries have similar judicial quality. In addition, when the importer starts to have a better judicial quality than the exporter, the effect of trust on trade turns significant and negative.
13
V.
Robustness Checks
5.1 Robustness Check I: International Trade between 1970 and 1996
In this section, we test the sensitivity of the results in Section IV to the use of alternative measures of judicial quality and the time period of study. The alternative measure of judicial quality, “legal structure and security of property rights”, is developed by Gwartney and Lawson (2009). It is constructed by using multiple other studies, including Global Competitiveness Reports, International Country Risk Guide, and World Bank’s Doing Business surveys. It covers several dimensions, such as judicial independence, impartial court and legal enforcement of contracts.
Data is available for 1970, 1975, 1980, 1985, 1990 and 1995. The measure is matched with trust survey data available in the closest years and trade flows in these corresponding years. For instance, while the judicial quality measure is available in 1970, we match it with the trust survey data obtained in 1970. When the survey data is not available in the same year the judicial quality is reported, we use the survey data from the following year. In all, the judicial quality measure in 1975, 1985 and 1995 is matched with the survey (and trade) data in 1976, 1986 and 1996.
Following Guiso et al. (2009), the following regression model is used:
ijt t j t i ij jt it ijt ijt ijt ijt
jit Trust Q Trust Q GDP GDP Dist k Year Year
LnTrade =β1 +β2 +β3 * +α1 +α2 +α3 + * +λ * +ε (6)
where Trustijt is the mean of answers to the trust question from country i to country j in survey year t and represents the trust a representative individual from country i has towards a random individual from country j. Here, year t gets value from years when the survey data is taken. Other variables are in the same form as those used in equation (3).
We report all the empirical results in Table-4. First, a standard gravity model driven by the demand side is reported in Column (1). In line with our expectations, geographic distance has a negative and significant effect on trade volume while both the exporter’s and the importer’s GDP level have a positive and significant impact on trade. Similarly, the exporter’s GDP level shows a stronger impact on trade than the importer’s GDP level.
14
with finding in Section IV, the sample between 1970 and 1996 are dominated by importing countries have more effective judicial systems than exporting countries.
To test solely the effect of judicial quality on trade pattern, we include the exporter’s judicial quality (Qjt) to the gravity model in Column (3). The coefficient has a negative sign and is significant at the 5 percent level. As argued in the previous section, the negative sign is probably caused by the assumption that trade flows are driven by the love of variety. Since the difference in judicial quality leads to difference in comparative advantage, importing countries with better judicial quality prefer to trade with countries with poorer judicial quality.
When the difference between the judicial quality of exporter and importer (Qijt) is added to the gravity model in Column (4), the coefficient for Qijt suggests that the difference in judicial quality between two trading countries does not have a direct impact on trade volume. Based on the model in Column (4), Trustijt is further added to the model in Column (5). The coefficient for judicial quality difference is still not different from zero while the coefficient for Trustijt is significant at the level of 5% but negative.
Column (6) reports the final model with Trustijt, Qijt and their interactive term. The coefficient for Trustij is still negative and significant at the level of 1%. Meanwhile, the coefficients for both Qijt and the interactive term are significant at the level of 5%. Now, the coefficient for Qijt is positive while the coefficient for the interactive has a negative sign. The marginal effect of trust on trade is plotted in Figure-2 with the 95% confidence interval.
Figure-2 generally resembles Figure-1. The marginal effect of bilateral trust on trade has a downward slope. As the importer’s judicial quality improves, the effect of trust on trade drops and turns negative when it exceeds the export’s judicial quality. Additionally, when approaching the right end of the figure where the importer has a much better judicial system, both the lower bound and the upper bound of the confidence interval are below zero. It indicates under this situation, higher trust from the importing country leads to lower trade flows from the exporting country.
15
the better judicial quality. It is also worth noticing that when Qijt gets close to zero, the marginal effect turns insignificant. Though the results do not fully agree with the proposed hypotheses, they are largely in line with it. It could be caused by the limitation of the available data. So far, only European countries are used in this sample. The difference in judicial quality between a pair of European countries is not large enough for the effect to turn significant at the left end of the figure.
Figure-2 Marginal effect of trust on trade (Robustness Check I)
-3 -2 -1 0 1 -4 -2 0 2 4
Note: the figure shows the marginal effect of Trustijt on commodity trade flows for various values of Qijt. Furthermore, (in green) the 95% confidence interval for the marginal effect is plotted. The dots refer to the observations indicate various values of Qijt for all observations included the sample and their corresponding marginal effects. Gwartney and Lawson (2009) is the data source for Qijt.
5.2 Robustness Check II: International Trade between 2001 and 2009
If we assume that the launch of Euro and European Union do not cause a major change in people’s general perception of trustworthiness of people from other European countries, we could use our generated proxy for trust (Trustij) to explain the trade pattern between 2001 and 2009. The data on
commodity trade is obtained from UN comtrade, which is the original source for Feenstra et al.(2005) for period 1984 to 2000. Moreover, we tried the two proxies for judicial quality: Kaufmann et al. (2009) and Gwartney and Lawson (2009), which have a correlation of 91.02%
In Table-5, we redo those model specifications in Table-3. In Column (1), generated proxy for trust (Trustij) is added to the standard gravity model. Similar to the results in Table-3, Trustij alone does not
16
obtained from Kaufmann et al. (2009). Judicial quality of the exporter and its lagged value do not differ from zero, while the difference in judicial quality between countries has a significant but negative effect on international trade. It suggests that although judicial quality of the exporter alone does not influence trade volumes, countries tend to trade less with countries that have very different level of judicial quality. Since the coefficient for the lagged value of the difference in judicial quality does not significantly differ from zero, the use of Qijt will not lead to simultaneous problem. In Column(7), Trustij , Qijt and their interactive term are added to the standard gravity model. The results do not differ from the results obtained in Table-3: Trustij is not significant while both Qijt and the interactive term are significant at least at the level of 5%. Since the interactive term has a negative coefficient, it indicates that as the difference in two countries’ judicial quality increases, the effect of trust on trade decreases.
From Column (8) to Column (14), proxy for judicial quality provided by Gwartney and Lawson (2009) is used. Although judicial quality of the exporter does not significantly influence the international trade, the coefficient for its lagged value is negative and significant at the level of 5%. When concerning the difference in judicial quality, both its current and lagged value have a negative and significant coefficient. It confirms previous findings that countries do not trade much with countries with different level of judicial quality. In case that the use of Qijt provided by Gwartney and Lawson (2009) leads to simultaneous problem, we provide results for baseline model specification using Qijt andQijt-1 in Column (13) and (14). The results in Column (13) and Column (14) do not differ much from the results in Column (7) and the results of the baseline model in Table-3. Coefficients for Qijt and the interactive term between Qijt and Trustij are negative and significant at the level of 1%. It further supports the previous finding that the effect of trust on trade drops as the difference in two countries’ judicial quality increases.
17
approaching the right end of the figure where the importer has a much better judicial system, both the lower bound and the upper bound of the confidence interval are below zero. It indicates higher trust from the importing country leads to lower trade flows from the exporting country.
In general, the results of this section are highly in line with our hypothesis. They suggest that the dissimilarity in judicial quality creates uncertainty that makes the informal channel, trust, have an impact on trade. Although the results in this section strongly support our hypotheses, they should be interpreted with caution. It is true that the trust proxy we used here is time-invariant and is largely determined by historical determinants. We expect that it should persist over time and will not have an extreme change after 2000. However, due to lack of data, we cannot rule out the possibility that the launch of Euro and European Union leads to a dramatic change in people’s perception of trustworthiness towards people from other European countries.
Figure-3.1 Marginal effect of trust on trade (Robustness Check II)
-2 -1 0 1 2 -2 -1 0 1 2
Note: the figure shows the marginal effect of Trustij on commodity trade flows for various values of Qijt. Furthermore, (in green) the 95% confidence interval for the marginal effect is plotted. The dots refer to the observations indicate various values of Qijt for all observations included the sample and their corresponding marginal effects. Kaufmann et al. (2009) is the data source for Qijt.
18
Figure-3.2 Marginal effect of trust on trade (Robustness Check II)
-2 -1 0 1 2 -4 -2 0 2 4
Note: the figure shows the marginal effect of Trustijt on commodity trade flows for various values of Qijt. Furthermore, (in green) the 95% confidence interval for the marginal effect is plotted. The dots refer to the observations indicate various values of Qijt for all observations included the sample and their corresponding marginal effects. Gwartney and Lawson (2009) is the data source for Qijt.
VI.
Conclusion
How does judicial quality influence the effect of bilateral trust on trade? This study does not only find that both judicial quality and bilateral trust matter for trade, but also discovers that the effect of bilateral trust on trade volume depends on the dissimilarity of two countries’ judicial quality. When two countries have similar judicial quality, traders in each country can predict transaction costs with certainty. As Anderson (1999) suggested, it is easier to deal with the familiar. Traders will know about the formal procedures and how to deal with the ineffectiveness embodied in the other judicial system. However, if two countries greatly differ in judicial quality, it creates uncertainty in which judicial system traders will deal with when disputes occur. This kind of uncertainty makes traders rely on additional information or belief, such as trust, to estimate future payoffs.
The empirical results further show the effect of trust on trade is not constantly positive. When the importer’s judicial quality is much better than the exporter’s, a higher level of generalized trust from the importing country would cause a drop in trade volume. However, if the exporter has a more effective judicial system, higher trust from the importing side has a positive effect on trade volume. The main contribution of this study is that it provides some insights for trade promotion. Since
19
bilateral trust is largely cultural-rooted and time-invariant, its positive impact on trade volume puts countries that do not receive much trust in a disadvantageous position. However, according to the results obtained in this study, countries that are not big recipients of trust can eliminate the unfavorable effect of trust on trade through improvement in judicial quality without altering the type of legal system (which should be much harder to accomplish, if possible). When countries share the same level of judicial quality, bilateral trust will no longer influence trade.
20
VII.
Appendix
1 Table-1 Matrix for time invariant trust
Country of destination
21
2 Determinants of Trust
To prove that trust is largely cultural-rooted and time invariant, we follow Guiso, Sapienza and Zingales (2009) and explain bilateral trust with match-specific variables. These match-specific variables are historical and do not vary over time. Additionally, both country-of-origin fixed effects and country-of destination fixed effects are included. Country-of-origin fixed effects are used to control for the fact that people apply the level of trustworthiness of their own countrymen towards people from other countries (Glaeser et al., 2000; Sapienze et al., 2007). The regression model for bilateral trust takes the following form:
ij ij j i ij k X Trust = + λ + β + ε (7)
Where ki is the country-of-origin (where trust originated) fixed effect, λj is the country-of
destination (where trust is received) fixed effect, and Xij is the match-specific variables that are rooted in culture and can explain bilateral trust. The values for Xij depend on country pairs.
According to Guiso et al. (2009), countries differ in their trust toward the same population for several possible reasons. One is the difference in the information sets countries possess: better information leads to a more accurate estimate. An alternative is that some sort of cultural-rooted bias can be passed over several generations and form the perception of trustworthiness today. To capture the two possible reasons, we select both proxies for information and proxies for cultural similarity. A description of all the match-specific variables can be found in Appendix-F.
22
of people belonging to each religious denomination in one country from Alesina et al. (2003). To take into account the level of fragmentation, I develop a new measure for religious similarity, which has the following form:
) * 1 ( 2 ) ( 2 jk k ik jk k ik ρ ρ ρ ρ
∑
∑
− − (8)where ρi(j)krepresents the fraction of individuals in country i (j) who have religion k.
Two measures are used to capture ethnic distance. One is the genetic distance between indigenous populations developed by Cavalli-Sforza, Menozzi, and Pizazza (1996). It is a measure of differences in the genetic composition between two populations by summing the differences in frequencies of these polymorphisms7
. Secondly, somatic distance developed by Guiso et al. (2009) is used to account for the fact that people trust people who look like them more than those who do not (DeBruine, 2002). Furthermore, we adopt the Djankov et al. (2007)’s classification and construct a dummy variable equal to one if the countries share the same legal origin. The last measure for historical interactions captures the conflicts between countries. Presumably, countries with a long history of wars and conflict will mistrust each other. We compute the number of years a country pair has been in a war between 1816 and 1970 by using COW (Correlates of War) dataset. Although including wars before 1816 will provide us with more incidences, it is hard to believe that those wars are relevant. European countries began to have their current shape after 1800. Another advantage of this dataset is that it provides detailed information on where the battles took place and which states were involved. By using this information, we took the border change into consideration.
Table-2 includes all the regression models derived from eq(7). Both of-origin and country-of-destination fixed effects are included. In Column (1), we use all the match-specific variables. To improve the model, we take a general-to-specific model selection procedure. Starting with the model specification in column (1), the least significant variable from the regression is dropped one at a time. The procedure ends when only significant variables remain. The resulted model is reported in column (2).
[Insert Table-2 here]
7
23
In column (2), only three variables remain significant. Countries with a long history of conflicts and wars will have a tendency to mistrust each other. Moreover, it supports the argument of Guiso et al. (2009) and DeBruine (2002) that people trust people who look like them more than those who do not. When two populations lack common somatic traits, they tend not to trust each other. The effect of somatic distance on bilateral trust is constantly significant at the level of 1 percent. Although the dummy variable indicating whether two countries have the same legal origin is only significant at the level of 10 percent, it has the expected positive sign. It suggests that the similarity in formal institutions facilitates the formation of trust between two populations.
24
Table-2 Determinants of Trust
Trustij (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) War -0.012** -0.011** -0.012* -2.188 -2.124 -1.928 Somatic distance -0.016*** -0.017*** -0.019*** -6.035 -7.705 -9.505
Same legal origin 0.041 0.047* 0.172***
1.254 1.756 6.301
Same official language -0.044 0.097
-0.600 1.503 Geographic distance 0.006 -0.104*** 0.192 -4.963 Common border 0.029 0.119*** 0.708 4.577 Religious distance -0.022 -0.197*** -0.398 -4.546 Genetic distance (fst*1000) 0.000 0.000 0.573 0.128
Linguistic common roots 0.075 0.524***
0.627 5.063
Observations 205 207 207 207 207 207 207 207 206 206 207
R-squared 0.846 0.847 0.755 0.841 0.798 0.754 0.775 0.768 0.768 0.751 0.790
F test 0 0 0 0 0 0 0 0 0 0 0
25
3 Effect of Trust on Trade
Table-3 Effect of Trust on Trade
LnTradejit (1) (2) (3) (4) (5) (6) (7) (8) (9) Distij -1.088*** -1.071*** -1.072*** -1.144*** -1.053*** -1.117*** -1.031*** -1.125*** -1.131*** -30.850 -27.030 -27.530 -23.810 -19.980 -23.500 -19.830 -21.200 -21.260 GDPit 0.880*** 0.969*** 0.957*** 0.625** 0.526 1.183*** 0.981** 1.320*** 1.300*** 3.724 3.981 3.936 2.139 1.371 3.532 2.107 3.833 3.832 GDPjt 1.329*** 1.230*** 1.219*** 1.717*** 1.828*** 1.048*** 1.182** 0.918*** 0.943*** 5.603 5.035 4.950 5.665 4.444 3.127 2.530 2.657 2.772 Trustij 0.084 -0.065 -0.161 0.640 -0.362 -0.882 AvTrustij 0.082 0.623 Qjt -0.567** -2.426 Qjt-1 -1.059** -2.478 Qijt -0.291* -0.324* -0.457*** -1.696 -1.885 -2.747 Qijt-1 -0.102 -0.311 Qijt* Trustij -0.825** -2.422
Exporting-country fixed effects*years Yes Yes Yes Yes Yes Yes Yes Yes Yes Importing-country fixed effects*years Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 1050 1035 1035 630 420 630 420 621 621 R-squared 0.998 0.998 0.998 0.998 0.999 0.998 0.998 0.998 0.998 F test 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Note: 1)Qijt=Qit-Qjt, data obtained from Kaufmann et al. (2009);
26
4 Robustness Checks
Table-4 Robustness Check for International Trade between 1970 and 1996
LnTradejit (1) (2) (3) (4) (5) (6) Distij -1.302*** -1.025*** -1.307*** -1.283*** -1.022*** -1.033*** -32.42 -19.60 -31.44 -31.91 -19.45 -19.33 GDPit 0.673*** 0.561* 0.589*** 0.676*** 0.520 0.554* 4.404 1.874 3.847 4.005 1.647 1.764 GDPjt 1.560*** 1.740*** 1.698*** 1.555*** 1.775*** 1.750*** 10.16 5.450 10.58 9.169 5.349 5.317 Trustijt -0.579*** -0.568*** -0.583*** -3.116 -3.048 -3.111 Qjt -0.0665** -2.162 Qijt 0.00187 0.0287 0.495* 0.0663 0.755 1.740
Qijt* Trustijt -0.180*
-1.654
Exporting-country fixed effects*years Yes Yes Yes Yes Yes Yes
Importing-country fixed effects*years Yes Yes Yes Yes Yes Yes
Observations 1,259 570 1,231 1,203 570 570
R-squared 0.996 0.998 0.996 0.996 0.998 0.998
F test 0.000 0.000 0.000 0.000 0.000 0.000
Note: 1)Qijt=Qit-Qjt, data obtained from Gwartney and Lawson (2009)
27
Table-5 Robustness Check for International Trade between 2001 and 2009
I: Kaufmann et al. (2009) II: Gwartney and Lawson (2009)
LnTradejit (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) Distij -1.128*** -1.092*** -1.173*** -1.087*** -1.165*** -1.045*** -1.053*** -1.122*** -1.114*** -1.130*** -1.130*** -1.128*** -1.136*** -1.136*** -33.62 -35.78 -26.44 -35.82 -26.11 -30.53 -30.96 -37.07 -37.1 -37.91 -37.95 -33.71 -34.26 -34.30 GDPit 0.451* 0.221 0.855** 0.519* 0.834** 0.673** 0.607** 0.390* 0.453* 0.539** 0.569** 0.719*** 0.561** 0.597** 1.879 0.79 2.443 1.835 2.199 2.326 2.118 1.656 1.94 2.187 2.329 2.829 2.243 2.41 GDPjt 1.747*** 1.993*** 1.404*** 1.650*** 1.390*** 1.469*** 1.541*** 1.765*** 1.657*** 1.659*** 1.629*** 1.480*** 1.643*** 1.608*** 7.25 7.014 3.942 5.812 3.698 5.059 5.357 7.162 6.798 6.705 6.642 5.812 6.559 6.48 Trustij -0.0436 0.222** 0.128 -0.0361 -0.105 -0.105 -0.386 1.971 1.102 -0.322 -0.948 -0.94 Qjt -0.275 0.0469 -1.396 1.159 Qjt-1 -0.189 0.0934** -0.765 2.372 Qijt -0.239* -0.281* -0.352** -0.0623** -0.0766*** -0.107*** -1.676 -1.958 -2.534 -2.256 -2.772 -3.937 Qijt-1 0.103 -0.0664** -0.118*** 0.552 -2.384 -4.28 Qijt* Trustij -0.618*** -0.262*** -3.353 -3.935 Qijt-1* Trustij -0.266*** -4.09
Exporting-country fixed effects*years Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Importing-country fixed effects*years Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1,449 1,260 840 1,260 840 1,242 1,242 1,470 1,470 1,470 1,470 1,449 1,449 1449
R-squared 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998 0.998
F test 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Note: 1)Qijt=Qit-Qjt, data obtained from Kaufmann et al. (2009) in I and from Gwartney and Lawson (2009) in II;
28
5 List of Surveys
No. Country Sampled N. of years present in Survey Years Present
1 Austria 1 1996 2 Belgium 8 1970, 1976, 1980, 1986, 1990, 1993, 1994, 1996 3 Denmark 7 1976, 1980, 1986, 1990, 1993, 1994, 1996 4 Finland 2 1993, 1996 5 France 8 1970, 1976, 1980, 1986, 1990, 1993, 1994, 1996 6 Greece 6 1980, 1986, 1990, 1993, 1994, 1996 7 Ireland 7 1976, 1980, 1986, 1990, 1993, 1994, 1996 8 Italy 8 1970, 1976, 1980, 1986, 1990, 1993, 1994, 1996 9 Netherlands 8 1970, 1976, 1980, 1986, 1990, 1993, 1994, 1996 10 Norway 1 1993 11 Portugal 5 1986, 1990, 1993, 1994, 1996 12 Spain 5 1986, 1990, 1993, 1994, 1996 13 Sweden 1 1996 14 Great Britain 7 1976, 1980, 1986, 1990, 1993, 1994, 1996 15 Northern Ireland 7 1976, 1980, 1986, 1990, 1993, 1994, 1996 16 (West)Germany 8 1970, 1976, 1980, 1986, 1990, 1993, 1994, 1996 Source: http://www.gesis.org/dienstleistungen/daten/umfragedaten/eurobarometer-data-service/ 6 Data Description
Data for Baseline Model
Variable Description Obs Mean S.D. Min Max Source:
Qi(j)t
Judicial quality (de facto)in country i(j)
in year t 630 1.5591 0.3474 0.721 2.0007 Kaufmann et al. (2009)
GDPi (j)t
The natural logarithm of GDP per capita
of country i(j) in year t 1050 10.1649 0.1893 9.721 10.6401 PWT 6.3
lnTradejit
Log of import (value in thousands of USDs) from Country j to Country i in
year t 1050 14.5656 1.5966 10.015 17.7543 Feenstra et al.(2005)
Trustij
The averaged residuals of regressing the mean of answers to the trust question from country i to country j in survey year
t on calendar-year dummies t 1035 0.0074 0.3012 -0.672 0.8496 Eurobarometer
Qijt
Judicial quality (de facto) distance in
29
Data for Robustness Check I I I I
Variable Description Obs Mean S.D. Min Max Source:
lnTradejit
Log of import (value in thousands of USDs) from Country j to Country i in year t
1259 13.3042 1.9670 6.4846 17.6193 Feenstra et al.(2005)
Qi(j)t Judicial quality (de jure)of
country i(j) in year t
1232 7.2534 1.3033 1.4000 9.3000 Gwartney and Lawson (2009)
GDPi (j)t The logarithm of GDP per capita
of country i(j) in year t
1260 9.8326 0.2759 8.9399 10.5218 PWT 6.3
Trustijt
The mean of answers to the trust question from country i to country j in survey year t
571 2.7301 0.2938 2.0081 3.6504 Eurobarometer
Qijt Judicial quality (de jure) distance
in year t: Qit-Qjt
1204 -0.0066 1.3859 -6.9000 6.9000 Gwartney and Lawson (2009)
Data for Robustness Check II II II II
Variable Description Obs Mean S.D. Min Max Source:
lnTradejit
Log of import (value in thousands of USDs) from Country j to Country i in year t
2828 14.8394 1.6083 10.0151 18.5517 UN comtrade
GDPi (j)t
The natural logarithm of GDP per capita of country i(j) in year t
2520 10.2501 0.1983 9.7209 10.7871 PWT 6.3
Qi(j)t [I]
Judicial quality (de facto)of country i(j) in year t
2100 1.5223 0.4068 0.3372 2.0431 Kaufmann et al. (2009)
Qi(j)t [II]
Judicial quality (de jure) of country i(j) in year t
1680 8.1067 1.0655 5.6000 9.6000 Gwartney and Lawson (2009)
Qijt [I]
Judicial quality (de facto) distance in year t: Qit-Qjt
2100 0.0000 0.5933 -1.6587 1.6587 Kaufmann et al. (2009)
Qijt [II]
Judicial quality (de jure) distance
in year t: Qit-Qjt 1680 0.0000 1.5321 -3.9000 3.9000 Gwartney and Lawson (2009)
Data for Appendix-B: Determinants for Trust
Variable Obs Mean Std. Dev. Min Max Source:
Trustij 207 0.00738 0.301656 -0.67224 0.849558 Eurobarometer
Same official language 207 0.028986 0.168173 0 1 Jon Haveman’s website
Geographic distance 207 7.086713 0.643405 5.153484 8.120583 Jon Haveman’s website
Common border 207 0.140097 0.347929 0 1 Jon Haveman’s website
War 207 1.024115 2.139889 0 10.58356 COW version 3.0
Religious distance 206 0.459115 0.277538 0.011283 0.851569 Alesina et al. (2003)
Somatic distance 207 9.47343 5.117837 0 20 Guiso et al. (2009)
Genetic distance (fst*1000) 206 55.49029 43.19188 4 204 Cavalli-Sforza et al. (1996)
Same legal origin 207 0.270531 0.445311 0 1 Djankov et al. (2007)
30
VIII.
References
Acemoglu, D., S. Johnson, and J. A. Robinson (2005), “The Rise of Europe: Atlantic Trade, Institutional Change and Economic Growth,” American Economic Review, 95, 546–579. Alan Heston, Robert Summers and Bettina Aten (2009), “Penn World Table Version 6.3”, Center
for International Comparisons of Production, Income and Prices at the University of Pennsylvania.
Alesina, A., A. Devleeschauwer, W.Easterly, S. Kurlat, and R. Wacziarg (2003), " Fractionalization," Journal of Economic Growth, Springer, 8(2), 155-94.
Anderson, J.E. (1999), “Why Do Nations Trade (So Little)?,” Boston College Working Papers in Economics, no. 428, Boston.
Anderson, James E., and Douglas Marcouiller (2002), “Insecurity and the Pattern of Trade: An Empirical Investigation,” The Review of Economics and Statistics, 84, 342–352.
Anderson, James E., and Eric van Wincoop (2003), “Gravity with Gravitas: A Solution to the Border Puzzle,” American Economic Review, 93, 170–192.
Anderson, James E. and Leslie Young (2006), “Trade and Contract Enforcement,” Economic Analysis & Policy, 5 (1), Article 30.
Arrow, K.J.(1973), “Information and economic behavior,” Stockholm: Federation of Swedish Industries.
Berkowitz, Daniel, Johannes Moenius, and Katharina Pistor (2006), “Trade, Law, and Product Complexity,” Review of Economics and Statistics, 88, 363–373.
Bertrand,M., Esther D., and S. Mullainathan (2004), “How Much Should We Trust Differences-in-Differences Estimates?,” Quarterly Journal of Economics,119, 249–275.
Beugelsdijk, S. (2006), “A Note on the Theory and Measurement of Trust in Explaining Differences in Economic Growth”, Cambridge Journal of Economics, 30, 371-387
Casson, M. (1997), “Information and organization,” Oxford: Clarendon Press.
31
Child, John (2001), “Trust-The Fundamental Bond in Global Collaboration,” Organizational Dynamics, 29 (4), 274-288.
Davis, D.R., D.E. Weinstein, S.C. Bradford, and K. Shimpo (1997), “Using international and Japanese regional data to determine when the factor abundance theory of trade works,” American Economic Review, 87, 421-446.
DeBruine, L.(2002), “Facial Resemblance Enhances Trust,” Proceedings of the Royal Society of London B, 269, 1307–1312.
De Groot, Henri L. F., Gert-Jan Linders, Piet Rietveld, and Uma Subramanian (2004), “The Institutional Determinants of Bilateral Trade Patterns,” Kyklos, 57, 103–123.
Den Butter, F.A.G., and R.H.J. Mosch (2003), “Trade, trust and transaction costs,” working paper, available on http://papers.ssrn.com/sol3/papers.cfm?abstract_id=459501.
Fearon, J. D., and D. D. Laitin(2003), “Ethnicity, Insurgency, and Civil War,” American Political Science Review, 97, 75–90.
Feenstra, R., R.E. Lipsey, Haiyan Deng, A. C. Ma and Hengyong Mo (2005), "World Trade Flows: 1962-2000," NBER Working Papers 11040, National Bureau of Economic Research, Inc.
Fukuyama, F. (1995), “Trust: the social virtues and the creation of prosperity,” New York: Simon and Schuster.
Glaeser, E., D. Laibson, J. A. Scheinkman, and C. L. Soutter (2000),“Measuring Trust,” Quarterly Journal of Economics, 115, 811–846.
Greif, A. (1993), “Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders’ Coalition,” American Economic Review, 83, 525–548.
Guiso, L., P. Sapienza, and L. Zingales (2004), “The Role of Social Capital in Financial Development,” American Economic Review, 94, 526-556.
Guiso, Luigi, P. Sapienza, and Luigi Zingales (2009), “Cultural Biases in Economic Exchange?,” Quarterly Journal of Economics, 124, 1095–1131.
32
Hoffman, A. (2002), “A Conceptualization of Trust in International Relations,” European Journal of International Relations, 8, 375.
Kaufmann, D., Kraay, Aart, and Mastruzzi, Massimo (2009), "Governance matters VIII : aggregate and individual governance indicators 1996-2008," Policy Research Working Paper Series 4978, The World Bank.
Knack, S. and P. Keefer (1997), “Does social capital have an economic payoff? A cross-country investigation,” Quarterly Journal of Economics, 112, 1251-1288.
Levchenko, Andrei (2007), “Institutional quality and international trade,” Review of Economic Studies, 74, 791–819.
Nunn, Nathan (2007), “Relationship-specificity, incomplete contracts, and the pattern of trade,” Quarterly Journal of Economics, 122 (2), 569–600.
Putnam, R.D. (1993), “Making democracy work: civic traditions in modern Italy,” Princeton, N.J.: Princeton university press.
Ranjan, Priya, and Jae Young Lee (2004), “Contract Enforcement and the Volume of International Trade in Different Types of Goods,” Mimeograph, U.C. Irvine.
Rauch, J.D. (2001), “Business and social networks in international trade,” Journal of Economic Perspectives, 14 (3),137-158.
Rauch, J.D., and V. Trindade (2002), “Ethnic Chinese networks in international trade,” The Review of Economics and Statistics, 84(1), 116-130.
Sapienza, P., A. Toldra, and L. Zingales (2007), “Understanding Trust,” NBER Working Paper No. 13387.
Sobel, J. (2002), “Can we trust social capital,” Journal of Economic Literature, 40 (1), 139-154. Trefler, D. (1995), “The case of missing trade and other mysteries,” American Economic Review, 85,
1029-1046.
