ABSTRACT
We investigate the correlation between institutions of economic freedom and generalized trust. The data we use are derived from the Eurobarometer surveys and measure how trustworthy nationals of a particular country are in the eyes of nationals of other countries. By using these data we avoid the presence of double causality between institutions of economic freedom and generalized trust that has been recognized as a major challenge in the literature. Indeed, external perceptions of trustworthiness are exogenous to the internal institutional building process. We find that during the 1980s and 1990s a 10% increase in the Economic Freedom of the World Index was associated with a 2.5% increase in generalized trust in the countries in our sample.
1. Introduction
It has been largely argued that generalized trust, defined as trust in people one does not know or knows very little about, positively affects various economic and social variables including economic growth, trade, happiness, democratic stability, etc. Of course, the findings of this literature have spurred a natural interest in explaining how generalized trust is formed.
A hypothesis that has received increasing attention in recent years stems from the literature of Institutional Economics and contends that generalized trust is largely shaped by contemporary institutions (the laws and informal conventions in societies that both permit and bound economic behavior and social interaction). In particular, institutions designed to promote economic freedom (defined by Gwartney et al. (2013) as those institutions that support personal choice, voluntary exchange, freedom to compete and security of privately owned property) have been argued to provide a suitable environment for the development of generalized trust (see, for example, Rose-Ackerman 2001; Acemoglu 2005; Berggren and Jordahl 2006). The argument is twofold. First, given that institutions of economic freedom are founded on securing private property rights, their establishment increases the cost of using non-cooperative strategies even in anonymous interactions. Second, institutions of economic freedom tend to reduce the active participation of the government in the economy and, therefore, tend to reduce bureaucracy and the opportunities for corruption associated with it. As a result, individuals learn to rely on predictable and enforceable rules of the game, expect others to do the same and eventually develop a trusting habit.1
Various studies have provided empirical evidence on the effect of institutions of economic freedom on generalized trust (see, for example, Rothstein and Stolle 2002; Uslaner 2003; Berggren and Jordahl 2006). As some of these authors have recognized, however, one of the major challenges of this empirical exercise is the presence of double causality. While it may be possible that institutions of economic freedom are conducive of generalized trust as explained above, it is also possible that the relationship goes in the other direction. As argued by La Porta et al. (1997) and Bjørnskov (2004), for example, an efficient legal system is easier to establish and maintain in an environment characterized by generalized trust. The presence of this endogeneity makes it challenging to provide empirical evidence consistent with the hypothesis mentioned above.2
Part of the empirical challenge of double causality relies on the nature of the data used to capture generalized trust levels. Most of the empirical studies that we have reviewed use data from the World Values Survey which capture the level of generalized trust placed by survey respondents on their fellow nationals. Therefore, when this information is regressed or correlated against data measuring the extent to which institutions of economic freedom are established in the same country, the presence of double causality is difficult to avoid.3
We contribute to this discussion by proposing the use of a particular dataset in which data on generalized trust levels are exogenous to the domestic institutional building process. The dataset proposed is derived from the Eurobarometer surveys of public opinion which have been implemented by the European Commission since 1973. During the first decades of their implementation, the Eurobarometer surveys asked the following question: ‘How much trust do you have in people from various countries? For each, please tell me whether you have a lot of trust, fair amount of trust, not very much trust or no trust at all.' Although this question was dropped from the Eurobarometer surveys after 1997 and, therefore, we only have data for the 1970s, 1980s and 1990s, we believe that the data collected provides a valuable opportunity to study the correlation of institutions of economic freedom with generalized trust avoiding the empirical challenge of double causality. Indeed, while institutions of economic freedom in a particular country may determine how trustworthy nationals of that country are in the eyes of nationals of other countries, the latter is not likely to determine the former.
One could validly argue that external perceptions of trustworthiness may not perfectly proxy the level of internal generalized trust. Could it be possible, for example, that nationals of a particular country trust each other to a large extent but they are not perceived as highly trustworthy by nationals of other countries? There are reasons to believe that this will seldom be the case. As Putnam (1995) argues, when a person develops trust in people with whom he or she interacts repeatedly, that person tends to also develop trust in all other people as a matter of habit and unreflected internalization. Moreover, a trusting habit is a two-way street. A person that develops trust in people in general becomes less likely to use non-cooperative strategies as preemptive or retaliatory measures. In other words, trusting individuals let their guard down and this attitude makes them, in turn, trustworthy. It follows then that, if foreigners perceive nationals of a particular country as trustworthy, it is very likely that that country enjoys high levels of domestic generalized trust as well.4
It is important to clarify two methodological points before moving forward. First, as mentioned above, the hypothesis we explore in this article stems from the literature of Institutional Economics which has made important contributions at the crossroads of economics and sociology. In this context, the premise of our enquiry is to reliably establish the correlation between institutions of economic freedom and generalized trust once the potential endogeneity mentioned above is removed from the empirical analysis. While the theoretical framework that we present below establishes a specific hypothesis as to why institutions of economic freedom may be conducive of generalized trust, we do not aim at providing a comprehensive sociological interpretation of the relationship between these two variables. We do not, therefore, aim at making any normative arguments regarding the virtues or the quality of institutions of economic freedom as compared to other types of institutions.
Second, it should be evident that the data from the Eurobarometer surveys described above capture the level of trust placed by foreigners on a social representation of ethnic others. That is, the respondents answered the surveys based on their individual perceptions of nationals of other countries as shaped, of course, by limited and imperfect information. While individual responses may not closely proxy generalized trust in the country in question because of these idiosyncrasies, our empirical exercise relies on weighted averages of thousands of responses over multiple countries which, thanks to the law of large numbers, provide a much closer approximation to the variable we are trying to instrumentalize.
The next section presents a theoretical framework that develops a working hypothesis as to why institutions of economic freedom may be conducive to generalized trust. The implications of the hypothesis are tested empirically in Section 3 using the data mentioned above.
Our main result indicates that, for the years and countries in our sample, other things equal, a 10% increase in the Economic Freedom of the World Summary Index (our main measure of the extent to which institutions of economic freedom are established in a particular country) is associated with a 2.5% increase in generalized trust as proxied by the data derived from the Eurobarometer surveys.
2. A theoretical framework
Consider two individuals who meet for the first time and establish a relationship in which some unobservable factor determines the profits for one of the parties. This could be the case, for example, of a manager hiring an employee, an investor hiring a financial advisor or an insurance company selling insurance policies to a client. Notice that, after a contract is written, the manager cannot perfectly observe the effort that the employee exerts, the investor cannot perfectly observe the diligence of the financial advisor in selecting an investment portfolio and the insurance company cannot perfectly observe how risky the client's behavior is. Economists characterize these type of interactions as moral hazard relationships. The fact that there is an unobservable factor in the relationship generates incentives for opportunistic non-cooperative behavior and, therefore, makes these arrangements costly. It is easy to see how, in this context, generalized trust can be instrumental in allowing potentially mutually profitable interactions possible.
As we have seen in our introduction, one of the cornerstone characteristics of institutions of economic freedom is the security of property rights. In practice, this characteristic is typically reflected in the ability to write contracts that are relatively easy to enforce (e.g. the laws and regulations tend to be easy to navigate and the judicial system is relatively efficient restoring property in the case of tort). It follows that, in a social context in which institutions of economic freedom are established, interactions like the ones described above are easier to develop because non-cooperative behavior is more costly and both parties are consciously aware of that reality.
But while securing property rights is instrumental in allowing the interactions described above, it is not sufficient for the development of generalized trust – at least not in a one-shot game. The key determinant is the evolution of prior beliefs over time. If relationships characterized by the presence of an unobservable factor happen under the continuous support of secure property rights, non-cooperative behavior will be continuously costly and individuals will develop expectations that counterparts will most likely not behave in such manner. Over time, these expectations will be internalized and individuals will form a trusting habit not consciously related to the establishment of institutions of economic freedom that first allowed such interactions, that is, individuals will develop generalized trust.5
There are two important stylized facts suggested by our theoretical framework. First, generalized trust develops over time as secure property rights are continuously built. Second, although our theoretical framework uses only one characteristic of institutions of economic freedom to reach its results (i.e. the security of property rights) this does not imply that other characteristics of institutions of economic freedom are not capable of producing the same results. Indeed, lower regulatory constraints or freedom to trade can lead to individuals developing increasing incentives to interact with others and, over time, also develop generalized trust. By the same token, we cannot discard that yet other characteristics of institutions of economic freedom (like smaller government size) fail to present the same correlation. These stylized facts guide our empirical analysis in the next section.
3. Empirical analysis
Given that one of the main results was that generalized trust develops over time as secure property rights are continuously built, we need to assess the relationship between institutions of economic freedom and generalized trust in a time-series context. Since we are fortunate to also have data for several countries, our regressions use a panel model.
where Trust is a measure of how trustworthy nationals of country i are in the eyes of nationals of countries other than i in period t; EFW is a measure of the extent by which institutions of economic freedom have been established in country i and period t and CV is a vector of control variables in country i and period t.
3.1. Data
3.1.1. Trust
Our proxy for generalized trust is derived from the Eurobarometer surveys. The Eurobarometer surveys asked the following question to nationals of several European countries during the 1970s, 1980s and 1990s:
How much trust do you have in people from various countries? For each, please tell me whether you have a lot of trust, fair amount of trust, not very much trust or no trust at all.
As this question measures how trustworthy nationals of country X are in the eyes of nationals of other countries, the linear model in (1) avoids double causality. Perceptions of how trustworthy nationals of country X are in the eyes of nationals of other countries are not likely to determine the extent by which institutions of economic freedom are established in country X.
The question of interest was asked in the Eurobarometer surveys of 1970, 1976, 1980, 1982, 1986, 1990, 1993, 1994, 1996 and 1997. However, the countries that the surveys asked about varied significantly over the years. As a result, we have complete data only for a group of 14 countries for 1980, 1986, 1990 and 1996. These 14 countries are: Belgium, Denmark, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, the UK, Japan and the US.
For each one of the four years of data available, the results for each of the 14 countries are tabulated in five columns to correspond to the four options regarding trust (‘a lot of trust’, ‘fair amount of trust’, ‘not very much trust’ and ‘no trust at all’) plus a column for ‘don't know/no answer’. The rows in the table correspond to the countries in which the survey was implemented.6 We calculate the totals for each column corresponding to an option regarding trust and find the percentage that each sum represents of the total number of respondents.7 Next, we assign a weight to each of the percentages found and add the weighted values. The weights we used were 2 and 1 for the columns ‘lot of trust’ and ‘fair amount of trust’, respectively, and –1 and –2 for the columns ‘not very much trust’ and ‘no trust at all’, respectively. Thus, strongest opinions regarding trust perceptions received twice the weight as did mild answers. The weighted sum is an index of generalized trust for each country and year that ranges from –1 to 1, where 1 represents the maximum level of trust and –1 represents the lowest level of trust.
As noted in the list of 14 countries, the Eurobarometer surveys not only asked respondents for their perceptions of trustworthiness of nationals of European countries but also of other countries with which Europeans are fairly familiar. Japan and the US are the only two of these other countries for which we have complete data for 1980, 1986, 1990 and 1996.
3.1.2. Institutions of economic freedom (EFW)
Our measure of the extent to which institutions of economic freedom are established in a particular country is given by the Economic Freedom of the World Index (Gwartney et al. 2013). This index is measured on a 0 to 10 scale (higher numbers indicate a stronger presence of institutions of economic freedom) and includes five broad areas:
Size of government (EFW 1): Measures the extent to which countries rely on the government and the political process to allocate resources and goods and services.
Legal system and property rights (EFW 2): Measures the protection of persons and their rightfully acquired property as determined by the rule of law, the security of property rights, an independent and unbiased judiciary, and impartial and effective enforcement of the law.
Sound money (EFW 3): Measures the stability of the country's currency and the ability of individuals to access alternative currencies to trade.
Freedom to trade internationally (EFW 4): Measures a wide variety of restraints that affect international exchange: tariffs, quotas, hidden administrative restraints and controls on exchange rates and capital. The lower the measures of these restraints the higher the index EFW 4.
Regulation (EFW 5): Measures regulatory restraints that limit the freedom of exchange in credit, labor and product markets. The lower the measures of these restraints the higher the index EFW 5.
While the use of the summary index that aggregates the results of all five areas (EFW) is preferable for capturing the overall establishment of institutions of economic freedom, we also use areas EFW 1 through EFW 5 individually to gauge what characteristics or components of such institutions are most relevant in determining Trust. We are particularly interested in the results of EFW 2 as the security of property rights is the characteristic upon which our theoretical framework reaches its results.
The Economic Freedom of the World Index is available since 1985 and every five years until the year 2000. From that year onwards the index is available yearly. Thus, to closely match the data from the Eurobarometer surveys, we use the years 1980, 1985, 1990 and 1995.
The Economic Freedom of the World Index has been extensively used in the literature of Institutional Economics. De Haan and Sturm (2000) and Paldam (2003), for example, consider the Economic Freedom of the World Index an appropriate indicator of such type of institutional building process.8
Trust against EFW for all countries and years in the sample.
3.1.3. Control variables
The control variables used in our empirical model are variables that have been traditionally argued to potentially determine generalized trust as well as external perceptions of trustworthiness.
3.1.3.1. GDP per capita
Foreigners may trust nationals of a particular country because nationals of that country are relatively wealthy. Income is also a standard control variable in empirical studies seeking to explain generalized trust. The results are mixed. Several authors have found a positive relationship between GDP per capita and generalized trust (e.g. Knack and Keefer 1997; Alesina and La Ferrara 2002) but others have failed to find such result (e.g. Bjørnskov 2005; Uslaner 2002).
3.1.3.2. Gini coefficient
It has been argued that social cohesiveness drives generalized trust. It may also be possible that foreigners are more willing to trust nationals of a particular country if that country presents relatively low levels of income inequality.
3.1.3.3. Religious fractionalization
Social cohesiveness could also be derived from lower levels of religious fractionalization. Knack and Keefer (1997), Alesina and La Ferrara (2002) and Uslaner (2002, 2003), for example, have all included measures of social cohesiveness such as the Gini coefficient and religious fractionalization in their empirical studies.
3.1.3.4. Young
The share of the population under 35 years of age. We include this variable in our regressions as perceptions of trustworthiness may be determined by the average age of the nationals of the country in question. The average age may capture life cycle patterns that make individuals more or less trustworthy in the eyes of foreigners. For example, young societies are more active in the labor market and are typically more economically and culturally integrated with the rest of the world.
3.1.3.5. Education
The percentage of the population above 35 years of age that completed secondary school. It could also be possible that foreigners trust nationals of a particular country because nationals of that country are well educated.
3.1.3.6. Roman Catholicism and Protestantism
3.2. Timing
To maximize the number of observations while keeping a balanced panel, our analysis restricts itself to level-to-level data for the dependent, independent and control variables for 1980, 1985, 1990 and 1995. Notice, however, that for the periods of 1985 and 1995, data on Trust correspond to 1986 and 1996, respectively, as this variable is not available for 1985 and 1995.
3.3. Results
Given that we use a panel model, our starting point was to investigate if a fixed effects or a random-effects model was appropriate for our analysis. The typical rationale for these models is that attributes specific to each country in the sample, not captured by explanatory and control variables, may affect the dependent variable.
As the bottom rows of Table 1 report, the F-test and Breusch-Pagan LM test indicate that we cannot reject fixed or random effects in either one of our regressions. The Hausman test indicates, in turn, that we can reject the null hypothesis favoring fixed effects only in three of the six regressions. As these results were not conclusive, we conducted a modified Wald test for group heteroskedasticity and a Breusch-Pagan LM test for contemporaneous correlation. As also reported in Table 1, these last tests indicate a strong presence of both group heteroskedasticity and contemporaneous correlation in all six regressions. Thus, to account for the characteristics of our data, we utilized the Panel Corrected Standard Errors methodology based on Beck and Katz (1995, 1996) where the parameters are estimated by OLS or Prais-Winsten regressions assuming that the errors are, by default, heteroskedastic and contemporaneously correlated across panels.
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| EFW EFW1 | 0.049** (0.021) | −0.004 | ||||
| EFW2 | (0.018) | 0.040* | ||||
| EFW3 | (0.021) | 0.027** | ||||
| EFW4 | (0.013) | 0.039** | ||||
| EFW5 | (0.019) | 0.039*** | ||||
| GDPpc | 1.15e-05*** | 1.4e-05*** | 1.24e-05*** | 1.29e-05*** | 1.24e-05*** | (0.015) 9.56e-06*** |
| (3.5e-06) | (3.87e-06) | (3.62e-06) | (3.80e-06) | (3.36e-06) | (3.61e-06) | |
| Gini | −0.017*** | −0.019*** | −0.018*** | −0.017*** | −0.016*** | −0.020*** |
| (0.004) | (0.004) | (0.004) | (0.004) | (0.005) | (0.004) | |
| Young | 0.012*** | 0.014** | 0.014*** | 0.017*** | 0.014*** | 0.006 |
| (0.005) | (0.006) | (0.005) | (0.005) | (0.005) | (0.005) | |
| Education | 0.001 | 0.002** | 0.001* | 0.001 | 0.002** | 0.002* |
| (0.001) | (0.001) | (0.001) | (0.001) | (0.001) | (0.001) | |
| Religious Frac | −0.314*** | −0.245*** | −0.282*** | −0.293*** | −0.255*** | −0.342*** |
| (0.089) | (0.070) | (0.103) | (0.084) | (0.074) | (0.010) | |
| Roman Catholicism | 0.001 | −0.004 | −0.018 | −0.010 | −0.013 | 0.026 |
| (0.036) | (0.043) | (0.047) | (0.033) | (0.034) | (0.039) | |
| Protestantism | 0.115*** | 0.122*** | 0.093* | 0.111*** | 0.095** | 0.142*** |
| (0.040) | (0.047) | (0.052) | (0.036) | (0.040) | (0.044) | |
| Constant | −0.431 | −0.151 | −0.452* | −0.560* | −0.566* | 0.073 |
| (0.263) | (0.223) | (0.274) | (0.325) | (0.291) | (0.251) | |
| Wald test | 58752.74*** | 3644.89*** | 4539.42*** | 9149.39*** | 122051.14*** | 4707.52*** |
| R-squared | 0.685 | 0.683 | 0.679 | 0.682 | 0.686 | 0.688 |
| FE test | 6.97*** | 7.08*** | 6.47*** | 6.88*** | 6.34*** | 7.81*** |
| RE test | 5.66*** | 2.43* | 1.71* | 2.28* | 5.37** | 3.36** |
| Hausman | 5.12 | 5.78 | 30.7*** | 12.96** | −1.57 | 13.74** |
| Heteroskedasticity | 38.35*** | 108.38*** | 67.61*** | 43.16*** | 57.52*** | 119.52*** |
| Cont correlation | 122.205** | 134.299*** | 126.938*** | 125.632*** | 122.794** | 120.559** |
| N | 56 | 56 | 56 | 56 | 56 | 56 |
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
| EFW EFW1 | 0.049** (0.021) | −0.004 | ||||
| EFW2 | (0.018) | 0.040* | ||||
| EFW3 | (0.021) | 0.027** | ||||
| EFW4 | (0.013) | 0.039** | ||||
| EFW5 | (0.019) | 0.039*** | ||||
| GDPpc | 1.15e-05*** | 1.4e-05*** | 1.24e-05*** | 1.29e-05*** | 1.24e-05*** | (0.015) 9.56e-06*** |
| (3.5e-06) | (3.87e-06) | (3.62e-06) | (3.80e-06) | (3.36e-06) | (3.61e-06) | |
| Gini | −0.017*** | −0.019*** | −0.018*** | −0.017*** | −0.016*** | −0.020*** |
| (0.004) | (0.004) | (0.004) | (0.004) | (0.005) | (0.004) | |
| Young | 0.012*** | 0.014** | 0.014*** | 0.017*** | 0.014*** | 0.006 |
| (0.005) | (0.006) | (0.005) | (0.005) | (0.005) | (0.005) | |
| Education | 0.001 | 0.002** | 0.001* | 0.001 | 0.002** | 0.002* |
| (0.001) | (0.001) | (0.001) | (0.001) | (0.001) | (0.001) | |
| Religious Frac | −0.314*** | −0.245*** | −0.282*** | −0.293*** | −0.255*** | −0.342*** |
| (0.089) | (0.070) | (0.103) | (0.084) | (0.074) | (0.010) | |
| Roman Catholicism | 0.001 | −0.004 | −0.018 | −0.010 | −0.013 | 0.026 |
| (0.036) | (0.043) | (0.047) | (0.033) | (0.034) | (0.039) | |
| Protestantism | 0.115*** | 0.122*** | 0.093* | 0.111*** | 0.095** | 0.142*** |
| (0.040) | (0.047) | (0.052) | (0.036) | (0.040) | (0.044) | |
| Constant | −0.431 | −0.151 | −0.452* | −0.560* | −0.566* | 0.073 |
| (0.263) | (0.223) | (0.274) | (0.325) | (0.291) | (0.251) | |
| Wald test | 58752.74*** | 3644.89*** | 4539.42*** | 9149.39*** | 122051.14*** | 4707.52*** |
| R-squared | 0.685 | 0.683 | 0.679 | 0.682 | 0.686 | 0.688 |
| FE test | 6.97*** | 7.08*** | 6.47*** | 6.88*** | 6.34*** | 7.81*** |
| RE test | 5.66*** | 2.43* | 1.71* | 2.28* | 5.37** | 3.36** |
| Hausman | 5.12 | 5.78 | 30.7*** | 12.96** | −1.57 | 13.74** |
| Heteroskedasticity | 38.35*** | 108.38*** | 67.61*** | 43.16*** | 57.52*** | 119.52*** |
| Cont correlation | 122.205** | 134.299*** | 126.938*** | 125.632*** | 122.794** | 120.559** |
| N | 56 | 56 | 56 | 56 | 56 | 56 |
Standard errors in parenthesis.
Statistical significance: *< 0.1, **< 0.05, ***< 0.01.
FE test is the F test for fixed effects.
RE test is the Breusch and Pagan Lagrangian multiplier test for random effects Heteroskedasticity is the Wald test for group heteroskedasticity.
Cont Correlation is the Breusch-Pagan LM test for contemporaneous correlation.
Overall, the results are consistent with our hypothesis and the results of previous literature: institutions of economic freedom are significant and positively correlated to generalized trust as proxied by external perceptions of trustworthiness.
In our first regression, when using the EFW summary index as the explanatory variable, our results indicate that, other things equal, during the period of study, a 10% increase in EFW was associated to a 2.5% increase in Trust.10
We obtain similar results for the different areas of the Economic Freedom of the World Index with the exception of EFW 1 whose coefficient is not significantly different from zero. Indeed, a 10% increase in EFW 2 (legal system and property rights), EFW 3 (sound money), EFW 4 (freedom to trade internationally) and EFW 5 (regulation) was associated, other things equal, to a 2, 1.35, 1.95 and 1.95% increase in Trust, respectively.
These results indicate that the security of property rights was the most important characteristic of institutions of economic freedom in terms of its association with Trust (it has the largest significant coefficient). This finding is consistent not only with the insights of our theoretical framework but also with the results of previous literature. Knack and Keefer (1997), Zak and Knack (2001) and Berggren and Jordahl (2006) have all found that legal structure and security of property rights are the institutional elements most closely related to generalized trust.
The fact that EFW 1 is not significantly correlated to Trust may be reflecting two contradicting mechanisms. On the one hand, a large participation of the government allocating resources and goods and services may hinder generalized trust due to excessive taxation and bureaucracy. On the other hand, large governments may favor generalized trust by funding efficient judicial and educational systems.
One could speculate that the positive sign and significance of EFW 3 suggests that solid macroeconomic environments promote economic interaction and voluntary contracts which, in turn, promote generalized trust.
One could also speculate that the positive sign and significance of EFW 4 suggests that the absence of barriers to trade (including tariffs, quotas and administrative restraints) favor economic interaction with the rest of the world which generates, in turn, an increase in domestic interaction as more productive opportunities become available.11
Finally, the positive sign and significance of EFW 5 is also a very important result as it suggests that excessive market regulation tends to increase bureaucracy and corruption, both of which weaken the security of property rights and, ultimately, weaken generalized trust.
We are also able to verify that, for the most part, our control variables are significantly correlated to Trust and present the expected sign.
GDP per capita is positive and significantly correlated to Trust in all six regressions, suggesting that higher incomes are positively associated with generalized trust.
The Gini coefficient and our measure of religious fractionalization are both separately negative and significantly correlated to Trust suggesting that less cohesive societies are associated with less generalized trust.
The variable Young, the share of the population under 35 years of age, is positive and significantly correlated to Trust in five of the six regressions suggesting that young societies are better suited for the development of generalized trust.
The coefficient of the variable Education is positive in all six regressions and significantly different from zero in four of the six regressions. This result suggests that, for the most part, more educated societies are associated with generalized trust. One could speculate that education produces individuals more capable to navigate the rules of the game in society and, therefore, reduces the need for non-cooperative behavior. Education also makes societies more cohesive and tends to be strongly correlated with the level of income which, as we have seen, both are positively correlated to Trust.
The dummy variable Protestantism is positive and always significantly correlated to Trust while the dummy variable Roman Catholicism is never significantly correlated to Trust. One could speculate that the strong work ethic that characterizes the Protestant culture is highly effective in driving generalized trust. In our sample, Denmark, Germany and the US received a 1 for Protestantism. All other countries, except Japan, received a 1 for Roman Catholicism. These results are consistent with La Porta et al. (1997).
Notice that, when considered as a whole, our results capture three major trustful representations: legal certainty (EFW 2), market economy (EFW 3, EFW 4 and EFW 5) and social cohesiveness (Education, Gini coefficient and religious fractionalization). Indeed, our results suggest that generalized trust is associated to individual freedom firmly harnessed by an ideal combination of collective legal rules that protect property rights, opportunities for social mobility and cultural cohesiveness. Interestingly, this interpretation of our results is consistent with Adam Smith's ‘duties of the sovereign’ related to (i) the protection of society from the invasion of other societies, (ii) the protection of every member of society from the injustice or oppression of other members of society and (iii) the provision of public goods for which large externalities will make their private provision too difficult (e.g. education) (Smith 1937). In fact, the result found in terms of EFW 1 (the size of the government) could be interpreted as a special case of Adam Smith's third duty.12
To test the robustness of our results for the inclusion of non-European countries in the sample, we removed the observations for Japan and the US and ran the same regressions again. Our results, available upon request, are qualitatively very similar. The two important differences that we noted were:
The coefficients of EFW and four of its areas increased in value. Other things equal, a 10% increase in EFW, EFW 2, EFW 3, EFW 4 and EFW 5 were associated to a 3.5, 2.3, 1.6, 2.5 and 2.3% increase in Trust, respectively (EFW 1 continued to be non-significantly correlated to Trust).
The variable religious fractionalization was not significantly correlated to Trust in either one of the regressions.
4. Conclusion
Institutions of economic freedom are positively correlated to generalized trust. At the theoretical level, the working hypothesis that we used was that, by securing property rights, institutions of economic freedom encourage interactions in which unobservable factors determine the gains for one of the parties. Over time, under the continuous support of secure property rights, individuals develop trusting habits not consciously related to the establishment of institutions of economic freedom that first allowed such interactions, that is, individuals develop generalized trust.
At the empirical level, our contribution resides on the use of data that avoids the presence of double causality between institutions of economic freedom and generalized trust, which has been recognized as a major challenge in previous empirical literature. The data used are derived from the Eurobarometer surveys and measure how trustworthy nationals of a particular country are in the eyes of nationals of other countries. We argue that this variable adequately proxies generalized trust.
We find that, for the sample and period of study, a 10% increase in the Economic Freedom of the World Index was associated with a 2.5% increase in generalized trust. We also find that four of the five areas of the Economic Freedom of the World Index were positive and significantly correlated to generalized trust. Consistent with our theoretical framework, the area most strongly correlated with generalized trust was the one measuring the security of property rights. We also find that, for the most part, our control variables are significantly correlated to trust and present the expected sign.
When considered as a whole, our results capture three major trustful representations: legal certainty, market economy and social cohesiveness. Indeed, our results suggest that generalized trust is associated to individual freedom firmly harnessed by an ideal combination of collective legal rules that protect property rights, opportunities for social mobility and cultural cohesiveness. This interpretation is consistent with Adam Smith's ‘duties of the sovereign’ related to (i) the protection of society from the invasion of other societies, (ii) the protection of every member of society from the injustice or oppression of other members of society and (iii) the provision of public goods for which large externalities will make their private provision too difficult (e.g. education).
Disclosure statement
No potential conflict of interest was reported by the author.
Notes on contributor
Antonio Saravia holds a Ph.D. in Economics from Arizona State University, an M.S. in Economics from Arizona State University, an M.A. in Economics and Social Policies from Georgetown University, and a B.S. in Economics from Bolivian Catholic University. Prior to joining Mercer University, Dr. Saravia held faculty appointments at Georgia State University, American University of Sharjah and Arizona State University. Dr. Saravia's research interests focus on the areas of institutional economics, political economy and the determinants of economic freedom. His work has appeared in highly reputed academic journals such as Constitutional Political Economy and Review of Development Economics among others.
Footnotes
Yamagishi and Yamagishi (1994) define generalized trust as trust that reflects ‘a belief in the benevolence of human nature in general'. Different authors have used different terms to refer to generalized trust. Knack and Keefer (1997), for example, refer to generalized trust as anonymous trust whereas Uslaner (2002) uses the term moralistic trust. Some of the studies that have analyzed the positive effects of generalized trust on economic and social variables include Putnam (1993), Knack and Keefer (1997), Dearmon and Grier (2009), Den Butter and Mosch (2003), Bjørnskov (2003, 2006) and Uslaner (2003).
The World Values Survey has been administered in six waves between 1981 and 2014. Studies that have used this database such as Rothstein and Stolle (2002), Uslaner (2003) and Berggren and Jordahl (2006), assess the level of generalized trust in a country by using responses to the question ‘Generally speaking, would you say that most people can be trusted, or that you can't be too careful in dealing with people?'
Various experimental studies have produced evidence of a positive correlation between trust and trustworthiness (see, for example, Knack and Keefer 1997; Lazzarini et al. 2005; Glaeser et al. 2000). In addition to relying on this argument for the use of external perceptions of trustworthiness as our proxy of generalized trust, our empirical strategy also controls for non-institutional country-level variables that could possibly influence external perceptions of trustworthiness (i.e. income levels, religion, inequality levels, educational levels and age).
The evolutionary component of our theoretical framework is consistent with the social learning perspective theory of trust (see, for example, Rotter 1971; Offe 1999; Hardin 2002) in which people make their ‘skeptical judgment largely by generalization from past encounters with other people' (Hardin 2002).
We made sure to remove the row for the country for which the question was asked. For example, if the question asked trust perceptions about Italians, we removed the answers given by Italians themselves.
The average number of respondents in each country was 1000. Depending on the year, the surveys were implemented in 5–17 countries.
An alternative index that measures the extent to which institutions of economic freedom are established is the Index of Economic Freedom published by The Heritage Foundation. We were, however, unable to use this index as it provides data starting only in 1995.
We do not add dummy variables for Buddhism, Orthodoxy and Anglicanism as adding these variables will be equivalent to adding dummy variables exclusively for Japan, Greece and the UK, respectively (that is not the case for Catholicism and Protestantism as these religions are the dominant religions in more than one country). Adding Buddhism, Orthodoxy and Anglicanism, therefore, will be equivalent to using a fixed effects model. As we will see below, however, the econometric model that fits the characteristics of our data most efficiently is Panel Corrected Standard Errors. We did check, however, what the results would be without using Panel Corrected Standard Errors. We found that, in general, Protestantism, Buddhism and Anglicanism are positive and significantly correlated with Trust while Orthodoxy and Catholicism are not significantly correlated with the same variable.
Remember that the EFW ranges in a scale from 0 to 10 and Trust ranges in a scale from –1 to 1.
This result could also be capturing, of course, that foreigners have more opportunities to interact with nationals of the country in question which gives the former the opportunity to develop trust in the latter.
Our results are consistent with those of Delhey and Newton (2005) who use the World Values Survey and find that ‘high trust countries are characterized by ethnic homogeneity, Protestantism, good government, GDP per capita and income inequality'. These authors also find that ‘ … Simmel (Simmel and Wolff 1950) rather than Tönnies (1963) was correct to argue that modern money economies with greater individual freedom and personal interdependence have ways of maintaining social trust.'
References
Appendix
| 1980 | 1986 | 1990 | 1996 | |
|---|---|---|---|---|
| Belgium | 0.574 | 0.500 | 0.586 | 0.421 |
| Denmark | 0.622 | 0.593 | 0.605 | 0.564 |
| France | 0.264 | 0.256 | 0.458 | 0.233 |
| Germany | 0.323 | 0.350 | 0.466 | 0.301 |
| Greece | 0.000 | 0.072 | 0.119 | 0.005 |
| Ireland | 0.215 | 0.237 | 0.307 | 0.210 |
| Italy | −0.168 | 0.042 | 0.197 | 0.010 |
| Luxembourg | 0.653 | 0.597 | 0.654 | 0.573 |
| Netherlands | 0.592 | 0.557 | 0.602 | 0.503 |
| Portugal | −0.058 | 0.111 | 0.181 | 0.208 |
| Spain | −0.035 | 0.107 | 0.312 | 0.196 |
| UK | 0.376 | 0.298 | 0.284 | 0.184 |
| Japan | 0.110 | 0.236 | 0.181 | 0.152 |
| US | 0.416 | 0.265 | 0.455 | 0.254 |
| 1980 | 1986 | 1990 | 1996 | |
|---|---|---|---|---|
| Belgium | 0.574 | 0.500 | 0.586 | 0.421 |
| Denmark | 0.622 | 0.593 | 0.605 | 0.564 |
| France | 0.264 | 0.256 | 0.458 | 0.233 |
| Germany | 0.323 | 0.350 | 0.466 | 0.301 |
| Greece | 0.000 | 0.072 | 0.119 | 0.005 |
| Ireland | 0.215 | 0.237 | 0.307 | 0.210 |
| Italy | −0.168 | 0.042 | 0.197 | 0.010 |
| Luxembourg | 0.653 | 0.597 | 0.654 | 0.573 |
| Netherlands | 0.592 | 0.557 | 0.602 | 0.503 |
| Portugal | −0.058 | 0.111 | 0.181 | 0.208 |
| Spain | −0.035 | 0.107 | 0.312 | 0.196 |
| UK | 0.376 | 0.298 | 0.284 | 0.184 |
| Japan | 0.110 | 0.236 | 0.181 | 0.152 |
| US | 0.416 | 0.265 | 0.455 | 0.254 |
| Variable | Definition | # obs | Mean | St Dev | Min | Max | Sources |
|---|---|---|---|---|---|---|---|
| Trust | Index of external perception of trustworthiness | 56 | 0.309 | 0.212 | −0.168 | 0.6542497 | EB |
| EFW | Economic Freedom of the World summary index | 56 | 7.085 | 0.874 | 5.1 | 8.48 | G |
| EFW1 | Size of government | 56 | 4.621 | 1.074 | 3 | 7.4 | G |
| EFW2 | Legal system and property rights | 56 | 7.616 | 0.893 | 5.6 | 9.1 | G |
| EFW3 | Sound money | 56 | 8.393 | 1.545 | 5.5 | 9.8 | G |
| EFW4 | Freedom to trade internationally | 56 | 8.371 | 1.038 | 5.3 | 9.7 | G |
| EFW5 | Regulation | 56 | 6.414 | 1.316 | 3.9 | 8.7 | G |
| GDP per capita | GDP per capita at constant 2005 US$ | 56 | 25,262.72 | 9,112.25 | 10,219.50 | 57,406.41 | WDI |
| Gini coefficient | Gini index of income distribution | 56 | 31.966 | 4.443 | 22.5 | 41.3 | WDI |
| Religious Fractionalization | Index of religious fractionalization | 56 | 0.086 | 0.150 | 0.001 | 0.568 | A |
| Young | Percentage of a country's population younger than 35 years | 56 | 51.314 | 4.228 | 44.29 | 62.8 | USBC |
| Education | Percentage of a country's population older than 35 years that has | 56 | 20.264 | 10.476 | 2.12 | 47.1 | BL |
| completed secondary school | |||||||
| Roman Catholic | Dummy variable equal to if Roman Catholic is the dominant religion | 56 | 0.571 | 0.499 | 0 | 1 | WCD |
| Protestantism | Dummy variable equal to if Protestantism is the dominant religion | 56 | 0.214 | 0.414 | 0 | 1 | WCD |
| Variable | Definition | # obs | Mean | St Dev | Min | Max | Sources |
|---|---|---|---|---|---|---|---|
| Trust | Index of external perception of trustworthiness | 56 | 0.309 | 0.212 | −0.168 | 0.6542497 | EB |
| EFW | Economic Freedom of the World summary index | 56 | 7.085 | 0.874 | 5.1 | 8.48 | G |
| EFW1 | Size of government | 56 | 4.621 | 1.074 | 3 | 7.4 | G |
| EFW2 | Legal system and property rights | 56 | 7.616 | 0.893 | 5.6 | 9.1 | G |
| EFW3 | Sound money | 56 | 8.393 | 1.545 | 5.5 | 9.8 | G |
| EFW4 | Freedom to trade internationally | 56 | 8.371 | 1.038 | 5.3 | 9.7 | G |
| EFW5 | Regulation | 56 | 6.414 | 1.316 | 3.9 | 8.7 | G |
| GDP per capita | GDP per capita at constant 2005 US$ | 56 | 25,262.72 | 9,112.25 | 10,219.50 | 57,406.41 | WDI |
| Gini coefficient | Gini index of income distribution | 56 | 31.966 | 4.443 | 22.5 | 41.3 | WDI |
| Religious Fractionalization | Index of religious fractionalization | 56 | 0.086 | 0.150 | 0.001 | 0.568 | A |
| Young | Percentage of a country's population younger than 35 years | 56 | 51.314 | 4.228 | 44.29 | 62.8 | USBC |
| Education | Percentage of a country's population older than 35 years that has | 56 | 20.264 | 10.476 | 2.12 | 47.1 | BL |
| completed secondary school | |||||||
| Roman Catholic | Dummy variable equal to if Roman Catholic is the dominant religion | 56 | 0.571 | 0.499 | 0 | 1 | WCD |
| Protestantism | Dummy variable equal to if Protestantism is the dominant religion | 56 | 0.214 | 0.414 | 0 | 1 | WCD |
EB: Eurobarometer surveys, European Commission 2014 <http://ec.europa.eu/public opinion/index en.htm>; G: Gwartney et al. (2013); WDI: World Development Indicators, The World Bank 2014 <http://data.worldbank.org/data-catalog/world-development-indicators>; A: Alesina et al. (2003); USBC: US Bureau of the Census, International Database <http://www.census.gov/population/international/data/>; BL: Barro and Lee (2000); WCD: World Christian Database <http://www.worldchristiandatabase.org/wcd/>.
| Trust | EFW | EFW1 | EFW2 | EFW3 | EFW4 | EFW5 | GDPpc | Gini | Education | Young | Rel Frac | R Catholic | Protestant | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Trust | 1 | |||||||||||||
| EFW | 0.55 | 1 | ||||||||||||
| EFW1 | −0.03 | 0.54 | 1 | |||||||||||
| EFW2 | 0.58 | 0.79 | 0.21 | 1 | ||||||||||
| EFW3 | 0.50 | 0.83 | 0.22 | 0.66 | 1 | |||||||||
| EFW4 | 0.60 | 0.76 | 0.19 | 0.57 | 0.64 | 1 | ||||||||
| EFW5 | 0.39 | 0.76 | 0.42 | 0.51 | 0.42 | 0.43 | 1 | |||||||
| GDPpc | 0.65 | 0.64 | 0.06 | 0.63 | 0.63 | 0.50 | 0.48 | 1 | ||||||
| Education | 0.32 | 0.47 | 0.28 | 0.46 | 0.47 | 0.14 | 0.36 | 0.42 | 1 | |||||
| Gini | −0.44 | −0.19 | 0.10 | −0.20 | −0.27 | −0.44 | 0.11 | −0.22 | 0.05 | 1 | ||||
| Young | −0.14 | −0.23 | 0.12 | −0.26 | −0.51 | −0.29 | 0.15 | −0.49 | −0.02 | 0.23 | 1 | |||
| Rel Frac | −0.12 | 0.34 | 0.40 | 0.15 | 0.25 | −0.08 | 0.45 | 0.16 | 0.42 | 0.21 | −0.06 | 1 | ||
| R Catholic | 0.03 | −0.19 | −0.20 | −0.07 | −0.15 | 0.03 | −0.26 | −0.12 | −0.12 | −0.28 | 0.26 | −0.47 | 1 | |
| Protestant | 0.31 | 0.27 | 0.01 | 0.36 | 0.24 | 0.21 | 0.22 | 0.32 | 0.29 | 0.27 | −0.11 | 0.08 | −0.60 | 1 |
| Trust | EFW | EFW1 | EFW2 | EFW3 | EFW4 | EFW5 | GDPpc | Gini | Education | Young | Rel Frac | R Catholic | Protestant | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Trust | 1 | |||||||||||||
| EFW | 0.55 | 1 | ||||||||||||
| EFW1 | −0.03 | 0.54 | 1 | |||||||||||
| EFW2 | 0.58 | 0.79 | 0.21 | 1 | ||||||||||
| EFW3 | 0.50 | 0.83 | 0.22 | 0.66 | 1 | |||||||||
| EFW4 | 0.60 | 0.76 | 0.19 | 0.57 | 0.64 | 1 | ||||||||
| EFW5 | 0.39 | 0.76 | 0.42 | 0.51 | 0.42 | 0.43 | 1 | |||||||
| GDPpc | 0.65 | 0.64 | 0.06 | 0.63 | 0.63 | 0.50 | 0.48 | 1 | ||||||
| Education | 0.32 | 0.47 | 0.28 | 0.46 | 0.47 | 0.14 | 0.36 | 0.42 | 1 | |||||
| Gini | −0.44 | −0.19 | 0.10 | −0.20 | −0.27 | −0.44 | 0.11 | −0.22 | 0.05 | 1 | ||||
| Young | −0.14 | −0.23 | 0.12 | −0.26 | −0.51 | −0.29 | 0.15 | −0.49 | −0.02 | 0.23 | 1 | |||
| Rel Frac | −0.12 | 0.34 | 0.40 | 0.15 | 0.25 | −0.08 | 0.45 | 0.16 | 0.42 | 0.21 | −0.06 | 1 | ||
| R Catholic | 0.03 | −0.19 | −0.20 | −0.07 | −0.15 | 0.03 | −0.26 | −0.12 | −0.12 | −0.28 | 0.26 | −0.47 | 1 | |
| Protestant | 0.31 | 0.27 | 0.01 | 0.36 | 0.24 | 0.21 | 0.22 | 0.32 | 0.29 | 0.27 | −0.11 | 0.08 | −0.60 | 1 |