Abstract
To date, most research finds education to have a positive effect on trust. Education increases people’s social intelligence, making them better able to distinguish between trustworthy and opportunistic types. Alternatively, education allows people to attain privileged social status, making them more resistant to deceit and exploitation by opportunistic types. In this article we show that this is not always the case. The relationship between education and trust is mediated by state efficacy; where the state is relatively efficacious, trustworthy types largely survive, while the opposite is true with relatively weak states. In weak states, highly educated people should be the least trustful. We empirically demonstrate this theoretical insight with survey data from three continents, Europe and Africa at the extremes and Latin America in the middle. We provide some indirect evidence in favor of social intelligence as the key mechanism linking education and trust.
Introduction
There is a wide variation in levels of trust across countries. There are societies where people routinely trust strangers, economic transactions are conducted without cumbersome paperwork and free riding is an infrequent event. At the other extreme, there are countries where trust is limited to closed networks of friends, acquaintances, or members of an ethnic group. Within countries, there is also variation between individuals in their predisposition to trust strangers: the country-level differences between high-trusting societies and those where people do not dare to trust beyond the boundaries of their own closed networks is mimicked in within-country variation in levels of trust between individuals of different social strata. How can we explain this variation? Explanations at the country-level include economic inequality (Rothstein and Uslaner, 2005) and government effectiveness (see Robbins, 2011). At the individual level, education is usually considered a good predictor of trust: more educated people, through mechanisms probably related to the development of ‘social intelligence’ – the capacity to read the social environment and to distinguish trustworthy from untrustworthy individuals – are usually more trusting than the low educated. As it is claimed in a recent study, ‘it is an almost universal finding that there is a strong and positive relation between education level and trust’ (Hooghe et al., 2012: 609).
In this article we claim that in complex societies, state efficacy is an umbrella under which the individual propensity to trust, driven by variables such as education, can be understood. Education, we argue, will be positively related to trust only in countries blessed with highly efficacious states; by contrast, in relatively inefficient states, both the lowly and the highly educated will show minimum levels of trust. The key to these differential patterns of trust across countries lies, we will argue, in the capacity of states to protect trustworthy types from being exploited by opportunistic types. We define trustworthy types as individuals with pro-social preferences – altruists who care about others and strong reciprocators, people willing to reciprocate cooperation and to sanction opportunistic individuals even at a cost to themselves. In small, simple societies, trustworthy individuals can be protected from exploitation by opportunistic agents through decentralized sanctions. In more complex, large societies, by contrast, the state is necessary for the trustworthy types not to be crowded out. In failed states, pro-social preferences will die out, and opportunistic types will predominate in the population. This variation in the capacity of state institutions to protect trustworthy individuals affects the relationship between individual-level variables such as education and trust. Contrary to what is usually claimed in literature related to this subject, the relationship between education and trust is not universally positive; in failed states, where pro-social preferences are lacking, highly educated individuals will probably trust less than the lowly educated, as they are more capable of discerning the distribution of types in society. By contrast, in efficacious states, where pro-social individuals largely survive, the highly educated will be considerably more trusting than the lowly educated. There is, in sum, an intervening effect of state efficacy on the relationship between education and trust.
We test these ideas by comparing the patterns of the relationship between trust and education in three continents: Europe, Latin America and Africa. They can be placed on a continuum of state efficacy, going from relatively high state efficacy (Europe) to weak (Africa), with Latin America somewhere in between. Our analysis shows that, according to our expectations, education has a strong positive relation to trust in Europe while showing a considerably weaker positive relationship in Latin America and a substantially negative relationship in Africa.
Literature review and theory
Our idea is that the relationship between trust and education depends on the degree of state efficacy. In efficacious states, more educated people will be more trusting than lowly educated people, while, at the other extreme, in weak states, either there will be no relationship between the two variables (as no one will trust, regardless of how educated they are) or it will be a negative one. By state efficacy we mean the infrastructural capacity of the state to enforce its laws, norms and regulations. This is slightly different to the concept of ‘quality of government’, which stresses the idea of impartiality in the exercise of public power, although the two could indeed be related to each other, as impartiality in the recruitment of the public service can positively affect state efficacy (Rothstein, 2011: 31). In the rest of this section we will develop this hypothesis further. It rests basically on three building blocks: a definition of trust, an argument on the role of the state in promoting trust in complex societies and an assumption on the connection between education and trust. We will deal with these three building blocks in turn.
We define trust as the truster’s expectations about the likelihood of the trustee reciprocating cooperation for intrinsic reasons, regardless of the existence of exogenous incentives for cooperation. Pro-social preferences are crucial components of these intrinsic reasons for reciprocal cooperation. They include strong reciprocity – the predisposition to cooperate, to reciprocate others’ cooperation and to punish violations of cooperative norms even at a net cost to the punisher (Fehr and Gintis, 2007) – and altruism: a concern for the welfare of others as an end in itself (Sloan Wilson, 2015: 4). There is strong empirical evidence of the existence of these pro-social preferences. Most of it comes from experiments. In the standard trust game, most trustees tend to reciprocate cooperative behavior, probably out of moral concerns (Camerer, 2003: 100), and much experimental evidence shows that other-regarding motives are common in all human populations (Bowles, 2016: 4). Individuals with these pro-social preferences are likely to reciprocate cooperation and, therefore, can be considered trustworthy.
This definition has some similarities with other notions of trust. It is, as noted in Hardin (2002), a three-part relationship (‘A trust B to do X’) and an expectation. This differentiates it from other definitions that consider trust a decision made under uncertainty (Coleman, 1990). If trust is defined as a decision, it is not possible to differentiate it from the act of cooperation itself. By defining it as a belief, trust can have an independent effect on cooperation: the truster will cooperate, at least partly, as a consequence of their belief about the trustee’s trustworthiness. Another difference is that, in our account, to qualify as trust, these beliefs should be referred to the trustee’s intrinsic motivations to cooperate; otherwise, trust would be a somewhat redundant concept.
There are several reasons why the truster could offer cooperation to the trustee. One of them is that both are immersed in an ongoing relationship. Provided that both strongly value future payoffs and that there is uncertainty about when this relationship is going to end, cooperation may arise (Axelrod, 1984). It may also arise, even if interactions are just ephemeral, if both the truster and the trustee are members of a network where the information about a breach of trust is cheaply communicated to other members of the network. In this case, if the trustee values the benefits of cooperation with those other members, he will have an incentive to keep a reputation of trustworthiness (Greif, 2006). Finally, cooperation may be the result of a third party, typically the state, providing sanctions against would-be opportunistic players. Notice that in all these cases it is not necessary to include the term ‘trust’ to explain why cooperation arises. Long-term relationships, non-myopic preferences, dense networks and a third-party enforcer of private agreements are all that is needed to explain why the outcome is cooperation. As Guseva and Rona-Tas (2001) conclude in their comparison of US and Russian credit markets, the presence of institutions makes trust redundant.
By contrast, if trust is limited to beliefs about the trustee’s intrinsic motivation to reciprocate cooperation, then it may have a role in explaining cooperation. The truster may have an idea about whether the person they are interacting with is an altruistic or an opportunistic individual. If exogenous incentives of the sort described above are in place, these considerations may be irrelevant in order to decide whether to cooperate or not. But if these incentives are not in place, for example, if the truster is dealing with a stranger in a one-shot interaction, then those beliefs may be crucial to determine whether the outcome is mutual cooperation or not; in a one-shot prisoner’s dilemma, trust will be crucial for cooperation (Gambetta, 1988: 216). Even when there are exogenous incentives to cooperation, the truster could put them aside and simply decide to cooperate based on their beliefs about what the trustee would do in case those incentives were not available.
This definition affects the role of the state regarding trust. This is the second building block of our theory. Some studies claim that, by acting as a third-party enforcer of private agreements, an efficacious state directly promotes trust. The argument is quite straightforward: the trustee will reciprocate cooperation because otherwise the state will punish them (see Cook et al., 2005; Levi, 1998). Knowing this, the truster will offer cooperation. However, according to our definition of trust, in this case the truster is not really expressing trust towards the trustee. They may trust that the state will efficaciously sanction opportunistic behavior, and that the trustee will know this and act accordingly. In order to trust the trustee in the proper way, the truster has to believe that they will reciprocate cooperation even if they think that the state is not going to punish them if they cheat. The presence of an efficacious state does not affect this belief.
In game-theoretic terms, an efficacious state leads to a pooling equilibrium, in which all types cooperate. Maybe they do it for different reasons: the altruistic type does it because they always cooperate; while the opportunistic type does it because they fear being sanctioned. But the truster only sees the outcome, and, as all types of players cooperate, they cannot update their beliefs about the trustee’s intrinsic preferences towards cooperation. There is, however, a way through which the state can, albeit indirectly, promote trust: through its role as protector of trustworthy types. In the aforementioned pooling equilibrium both trustworthy and untrustworthy types cooperate. Were the state incapable of enforcing the agreement between the two players, the result would be different: the opportunistic type would have cheated the truster while the trustworthy type would have complied. But then, the payoffs of the opportunistic type would have been higher and, in the long run, opportunistic types would have outperformed the trustworthy types to their extinction.
An efficacious state changes things radically: by enforcing private agreements, trustworthy types can survive. Rule of law efficaciously enforced, therefore, protects trustworthy types from opportunism; this is crucial for pro-social preferences to survive (Bowles, 2016: 206). 1 Besides, if trustworthy types do survive, trust can survive too. By contrast, in the absence of an efficacious state or other exogenous incentives for cooperation, the truster will adjust their expectations about other people’s trustworthiness downwards, as society will increasingly be made of opportunistic types. Additionally, although in many cases in countries with efficacious states cooperation can be reached without trust, trust will still have a role to play, especially in one-shot interactions with unknown people, where exogenous incentives to cooperative behavior do not exist. Notice that this does not mean that in countries with efficacious states all people will have pro-social preferences; it is not a place of unlimited altruism or inhabited by a breed of strong reciprocators. However, it will be a place where trustworthy types coexist alongside some opportunistic types. By contrast, in countries where the rule of law is deficiently enforced, pro-social types will be crowded out.
We have laid out, therefore, the two first building blocks of our theory: a definition of trust as expectations about other people’s pro-social preferences, and the state as a protector of trustworthy types. The third is the connection between education and trust. The literature relating to this subject habitually, but not always (see Gleave et al., 2012), shows that where there is a high level of education, social trust is higher (Rothstein, 2005: 94). Attempts to explain this relationship cite one or other of the following mechanisms: (a) high levels of education that broaden people’s worldview and their aspirations; (b) the development of social intelligence, born of education, which confers the ability to differentiate between those who should and should not be trusted; and (c) the ability of education to broaden our horizons and educate us about what is different from us (Yamagishi et al., 1999). Greater education increases trust because it may teach individuals the experiences and history of unfamiliar groups, and provides people with a greater understanding of trust-warranting information; that is, they may learn how to assess trustworthy signs and signals. So, basically – through several mechanisms – formal education provides individuals with the social intelligence they need to better ascertain other people’s trustworthiness.
These three building blocks allow us to formulate a theory about the relationship between education and trust as one that is mediated by the effectiveness of the enforcement of the rule of law. We claim that a positive relationship between education and trust is only possible in those countries where the rule of law efficaciously protects trustworthy types from being exploited by opportunistic individuals. In countries without an adequate rule of law, education will either have no effect on trust, or even a negative one. This is a corollary of the three building blocks outlined before. First, trust is an expectation about the likelihood of the trustee having pro-social preferences. Second, the survival of these preferences is a direct consequence of the state protecting them against opportunistic individuals. Third, people with high levels of education will be particularly well placed to distinguish trustworthiness. Regardless of the state’s capacity to protect trustworthy types, people with low levels of education will be relatively low trusters. This is probably because they lack the social intelligence that formal education provides to distinguish between trustworthy and opportunistic types so, as they are faced by greater uncertainty when dealing with people, they will probably refrain from trusting others and rely instead on closed networks of friends and kin. If the state does protect trustworthy types, highly educated people will probably be much more trustful. But in countries where the state is weak, pro-social preferences are not likely to survive and, therefore, highly educated people will negatively adjust their expectations about other people’s trustworthiness accordingly.
Data
The dependent variable: Trust
For our analyses we rely on three different databases: Afrobarometer, Latinobarometer and the European Social Survey. 2 These three databases cover most of the countries in the three continents under study. For trust we use the same question in all models, ‘Most people can be trusted or you can’t be too careful?’, the usual survey question on social or generalized trust. We are aware that the use of this trust measurement has been subject to several criticisms. First, concerning how to interpret what people may want to say when responding to this question. There are those who believe that the answer has more to do with the propensity towards cooperation of the respondent. According to others, it only reflects a kind of fixed and static trust, and it is incapable of evaluating its intensity and distribution. (Knack, 2000; Wuthnow, 2002; Letki and Evans, 2005; Thöni, et al., 2012). However, this indicator has been supported with evidence derived from experimental data, and it has proved to be a relatively stable measure in comparisons between countries (Cox et al., 2009; Ermisch et al., 2009; Robbins, 2011). As far as this variable primarily captures trust in strangers (‘most people’), it is also consistent with our idea of trust as expectations about the trustee’s type. Trust in strangers can be based on the truster’s experiences with other trustees, but not on past experiences with that particular person. It is more likely, then, a general evaluation about the likelihood of encountering trustworthy and untrustworthy types when dealing with strangers.
In Latinobarometer and Afrobarometer trust is a dummy variable, with value 0 for people who are totally distrustful, and value 1 for those who think that most people can be trusted, but in the European Social Survey it goes from 0 (‘You can’t be too careful’) to 10 (‘Most people can be trusted’). This undoubtedly poses problems of comparability between Europe on the one hand, and Africa and Latin America on the other, but it has to be taken into account that what our theoretical hypotheses demand from the empirical model is to reveal what is the direction of the relationship between education and trust in the three continents.
Figure 1 shows the distribution of trust across countries. In Latin America and Africa, the bars indicate the percentage of respondents who think that ‘most people can be trusted’. As in the European Social Survey there is no binary answer, the bars reflect the percentage of people that placed themselves between 7 to 10 in the trust question, who can be considered as high trusters. We only dichotomize the trust question for Europe for the purpose of comparability between continents in Figure 1. For our empirical models, we use the whole range of the variable.
Trust in the three continents.
The figure shows a sharp difference between Europe and the other continents. The average share of social trusters is considerably higher in Europe (34.5%) than in Latin America (16.6%) and Africa (19.3%). There is, nonetheless, considerably variation within each continent. Some southern and eastern European countries (such as Slovenia, Poland and Portugal) have a percentage of social trusters similar to Latin American and African countries, while countries such as Burundi, Niger and Burkina Faso have levels of social trust similar to some North European countries.
Independent variables
The three continents can be located on a continuum of state effectiveness: Africa is the weakest link in state infrastructure capabilities, with some countries being close to state failure. At the other extreme, Europe is endowed with fairly efficacious states, while Latin America lies somewhere in between.
For government effectiveness we use the Worldwide Governance Indicators from the World Bank (2014). The estimated values range from 2.5 (more Government Effectiveness) to -2.5 (less Government Effectiveness). This variable reflects perceptions of the quality of public services, the quality of the civil service and its independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies. This is the most widely used indicator of government efficacy.
Figure 2 shows the distribution of state efficacy across countries and the average state efficacy in each continent. Levels of state efficacy are relatively high in Europe (1.43), quite low in Africa (-0.45) and, although closer to Africa than to Europe, middling in Latin America (-0.24). However, these averages hide an important degree of heterogeneity within each continent. Some Latin American countries, such as Chile (1.08) and Uruguay (0.54), show levels of state efficacy similar to some European countries, while state efficacy in African countries such as Botswana (0.51) and Mauritius (1.04) is considerably higher than in some Latin American countries such as Bolivia (-0.66) and Venezuela (-1.22). To capture this heterogeneity, we have included the variable for state efficacy in the models for each continent.
Government effectiveness in the three continents (2015).
The other two variables at the country level are the Gini index, an indicator of income inequality, and per capita gross domestic product (GDP). Income inequality is one of the main predictors of social trust. In societies with high levels of economic inequality there is less concern for people of different background (Rothstein and Uslaner, 2005). This same idea also underlines Putnam’s (2015) analysis of inequality and social capital: increasing income inequality in the United States of America (USA), and its consequences in terms of growing levels of geographic and educational segregation, mean that rich and poor Americans are less likely to have first-hand knowledge of each other, and this has negative consequences for trust. Per capita GDP, in turn, is a fairly precise indicator of economic affluence. According to some studies, under certain circumstances income is related to trustworthiness (see Ermisch and Gambetta, 2016), and this, in turn, should probably lead to more trust. Besides, by factoring economic affluence, we exclude the presumption that what lies behind government effectiveness is economic development, which is certainly part of the story but perhaps not the whole story. It has been measured according to the World Bank database.
The models include three individual-level variables: education, gender and age. Education is our key independent variable, and it is measured as a continuous variable, with values ranging from non-completed primary education to doctoral degree. Age and gender are variables usually related to trust. Some studies find that older people tend to be more trusting than younger ones (Putnam, 2000), although this effect may be actually concentrated in the transition between childhood and early adulthood (Sutter and Kocher, 2007). There are several studies assessing the influence of gender on trust; most experimental evidence finding that, while women tend to be more trustworthy, men are more trusting (Buchan et al., 2008; Clark and Sefton, 2001).
In the online appendix we show the descriptive statistics of the variables used in our analysis (Table A1).
Results
Europe
We begin by analyzing the relationship between education and trust in Europe. According to our hypothesis, we expect a positive relationship between both variables in this continent. As opportunistic types will be kept in line, high trusters will not be generally exploited, and the beneficial effects of education on trust will naturally show up (or so we expect).
Spearman rank correlation allows us to have a first glimpse of the association between our key variables. It shows a fairly strong positive correlation between education and trust: as predicted, educated people tend to be more trusting than the less educated (a correlation of 0.1975). The correlation is even stronger with government effectiveness: trusters seem to be especially salient in countries with effective governments (a correlation of 0.2981). The Gini coefficient as expected, is negatively associated with trust (−0.1344).
We now go beyond descriptive analyses. Table 1 shows the results of multilevel lineal models for the following countries: Austria, Belgium, Switzerland, Czech Republic, Germany, Denmark, Estonia, Spain, Finland, France, United Kingdom, Hungary, Ireland, Israel, Lithuania, Netherlands, Norway, Poland, Portugal, Sweden and Slovenia. We have kept Israel in the sample, although, obviously, it is not a European country, in order not to diminish the number of cases at the country level; nonetheless the results are the same if we exclude it.
Education and trust in Europe.
p < 0.001; **p < 0.01; *p < 0.10.
Model 1, the null model, clearly indicates that there is important variation at both the individual and the country level. Model 2 shows that the positive relationship between education and trust that we have seen in the simple correlations largely holds when we control by the other independent variables. This positive association is graphically depicted in Figure 3, which shows the expected value of trust for different levels of education with the remainder variables at their means. The expected value of trust for a female of average age with a low level of education living in a country with average levels of state efficacy and income inequality is slightly above 4 in the trust scale. By contrast, the expected value of trust for a highly educated individual is above 6.
Education and trust in European countries.
Model 3 includes an interaction term between state efficacy and education. Although the main variation in state efficacy in our research design is captured by the differences between the three continents, there is enough between-country variation within each continent to allow us to estimate whether differences in state efficacy between European countries affect the relationship between education and trust. The interaction term is significant and points to the expected direction: while across the European continent the relationship between education and trust is positive, this is especially so in those countries with more efficacious states. Figure 4 depicts the conditional marginal effects of education on trust for different levels of state efficacy, with the other variables at their means. As can be seen, the effect of state efficacy within European countries is rather low: the effect of education on trust is always positive, and although it is more positive in relatively efficacious states, the differences are rather slim. In general terms, Figure 4 basically shows that European states are all efficacious enough for education to have a positive (and similar) effect on trust.
Effects of education on trust by state efficacy (Europe).
Latin America
We turn now to Latin America. In our tale of three continents, Latin America is the middle ground in the continuum of state efficacy, between the two extremes of Europe and Africa, although closer to the latter than to the former. We include in our analysis the following Latin American countries: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay and Venezuela.
Trust is measured as a dummy variable, with value 1 for those who think that ‘most people can be trusted’ and value 0 for those who think that ‘you can’t be too careful’ in dealing with others. We have estimated hierarchical logit models.
Education has a positive correlation with trust, but the association between both variables is much weaker than in Europe, which is basically what we expected (the Spearman rank correlation is 0.0302). As Latin American countries have relatively lower levels of state efficacy than European ones, we expected a less positive association between education and trust.
This result is largely confirmed in the multilevel analysis (Table 2). Once we include all the control variables, we see that the positive effects of education on trust stand, although the effect is rather weak. Figure 5 shows the expected probability of trusting for different levels of education with the remainder variables of the model at their means.
Education and trust in Latin America.
p < 0.001; **p < 0.01; *p < 0.10.
Education and trust in Latin American countries.
Figure 5 shows a difference of barely 5% in the expected probability of trusting between uneducated people and people with higher education for a female living in a country with average level of state efficacy as well as average levels of inequality. In Europe, as shown in Figure 3, the increase in the expected value of trust between people situated between those two extremes of the educational continuum (as always, with the other variables at their means) was around 50%. This largely confirms our theoretical expectations about the likely effects of state efficacy as an intermediary in the relationship between education and trust, and suggests that the role of the state in relation to trust has indeed to do with the protection of trustworthy types from being exploited by opportunistic individuals.
As in the European case, we have included an interaction between state efficacy and education. There is enough variation in levels of state efficacy within Latin America, from countries such as Chile, with levels of state efficacy above some European countries, to countries such as Paraguay, that seems more proximate to some of the weakest African states, for this interaction to make sense. Figure 6 shows a picture that clearly fits our theoretical expectations: in those Latin American countries with low government efficacy, the probability of trusting others is virtually the same, regardless of educational level. The figure actually shows that, at very low levels of state efficacy, education has a negative effect on trust: the less educated people are more trusting than the high educated. Things are much different in relatively high efficient states, such as Chile and Costa Rica. Here, education has a positive effect on trust.
Effects of education on trust by state efficacy (Latin America).
Africa
We now reproduce the results for the African continent. In this case, we rely on data from Afrobarometer for the following African countries: Benin, Botswana, Burkina Faso, Cape Verde, Ghana, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mozambique, Namibia, Nigeria, Senegal, South Africa, Tanzania, Uganda, Zambia, Zimbabwe, Mauritius, Sierra Leone, Niger, Togo, Burundi, Cameroon, Cote d’Ivoire, Guinea, Swaziland, Algeria, Egypt, Morocco, Sudan and Tunisia.
Africa is at the opposite extreme to Europe in our continuum of state efficacy. African states are notoriously inefficacious, and the list of African countries above includes some that could be easily classified as failed states, such as Liberia and Sudan. According to our hypothesis, we should expect a negative relationship between education and trust in African states, as lack of protection from opportunistic types will tend to undermine trustworthy individuals. If high education is truly associated with social intelligence, highly-educated people will then be the least trustful.
The descriptive statistics allow us to have a first confirmation of our hypothesis. There is a negative correlation of −0.1206 between education and trust in the African countries of our sample, quite a contrast to the results in both Europe and Latin America.
This result is largely confirmed in the multilevel models of Table 3. In Afrobarometer, trust is measured as a dummy variable, with value 1 for those who think that ‘most people can be trusted’ and value 0 for those who think that ‘you can’t be too careful’ in dealing with others. Therefore, as in the Latin American case, we have estimated hierarchical logit models.
Education and trust in Africa.
p < 0.001; **p < 0.01; *p < 0.10.
Model 2 shows that education negatively affects trust in African countries, confirming the idea that in places where the state is inefficacious the usually claimed positive relationship between education and trust does not show up. Far from being more trusting, as highly educated people are better placed to discern what the distribution of types in society is, they will probably trust less. Figure 7 shows the expected probability of trusting by different levels of education with the other variables of the model at their means. As can be seen, as people in Africa become well educated, their trust largely collapses. While individuals without formal education assign a (rather low) 0.25 probability to other people’s trustworthiness, the corresponding probability for individuals with higher education is barely 0.1.
Education and trust in African Countries.
As with the European and Latin American cases, we have also tested whether variation in state efficacy within African countries has an impact on the shape of the relationship between education and trust. As we have seen in Figure 2, there are some states, such as Botswana, whose levels of state efficacy are only just lower than the least efficacious European states, whereas countries such as Sudan could probably be classified as ‘failed states’. Accordingly, there is enough variation to perform this analysis.
The interaction term indicates that there are certainly differences in the relationship between education and trust depending on levels of state efficacy. These differences can easily be seen in Figure 8, which shows the conditional marginal effects of education on trust by levels of state efficacy. For all levels of state efficacy, education has a negative effect on trust, although this is less so as state efficacy increases. In countries with high state efficacy there are slight differences between the highly and the lowly educated; basically, education has no effect on trust. The picture is quite different for countries with relatively low levels of state efficacy. In this case, trust clearly falls with levels of education. This suggests that the differences in levels of trust between the well-educated and the non-educated are considerably more attenuated (to the point of disappearing) in relatively efficacious countries.
Effects of education on trust by state efficacy (Africa).
This offers an interesting contrast to both Europe and Latin-America. In Europe, as we have seen, education has a positive effect regarding trust in all countries, reflecting the fact that European states are relatively efficacious. In the not-so-efficacious Latin American countries, education has a lower positive effect on trust, and there are indeed differences among countries: in relatively efficacious states education has a positive effect on trust, while in less efficacious states this relationship is mainly flat. In Africa we find this flat relationship (actually, a slightly negative one) for relatively efficacious states (such as Botswana or South Africa), while in weak states the relationship is openly negative. This pattern is what we expected in a continent that has weak states: the well-educated are better placed to grasp the distribution of types in society, and, as trustworthy types are exploited and crowded out in these states, levels of trust for the well-educated will collapse to zero, and there will be no differences between people’s trust by levels of education.
However, there are two results in the African case that are, to a certain extent, at odds with our theoretical argument. First, it is somewhat of a puzzle why the lesser educated are more trustful in countries with relatively strong states as compared to countries with weak or even failed states. The same pattern appeared in Latin America, as shown in Figure 6, but not in Europe. Uneducated people are usually low trusters, regardless of the degree of state efficacy. Perhaps below a certain degree of state efficacy, the difficulty of less-educated people to accurately grasp the distribution of types in society is accentuated, and for this reason they make an especially erroneous evaluation of strangers’ trustworthiness.
A second and related puzzling result is why in states with a level of efficacy similar to some in Latin America and even Europe the relationship between education and trust remains negative. At a level of state efficacy of 0.5 according to the World Bank indicator, the relationship between education and trust is positive in both Latin America and Europe, whereas in Africa it is slightly negative. It is true that, as the state becomes more efficient, the strong negative relationship found in low-efficacious states turns into a slightly negative one, but at certain levels of state efficacy the relationship should still be positive. Perhaps it is more difficult for African states to develop a reputation of efficacy than for Latin American and European states and this distorts people’s perception of how efficacious the state really is – even for people with a higher education and, supposedly, high cognitive skills.
Robustness checks
Now, we will conduct some robustness checks for the above results, estimating the models with different indicators of state efficacy. We will employ two alternative indicators of state efficacy: the ICRG (International Country Risk Guide) and the Freedom House indicators. The ICRG measures corruption (patronage, nepotism, job reservations, secret party funding, and suspiciously close ties between politics and business), law and order (an assessment of the strength and impartiality of the legal system and of observance of the law) and bureaucracy quality, scaled from 0 to 1. Higher values indicate higher quality of government. In turn, the Freedom House index captures to what extent the freely elected head of government and a representative national legislature determine the policies of the government; if the government is free from pervasive corruption; and if it is accountable to the electorate between elections and operates with openness and transparency. Countries are graded from 0 (worst) to 12 (best). Of the two, the ICRG indicator seems closer to the notion of state efficacy; the Freedom House indicator apparently is mostly about democratic quality, although its component of freedom from pervasive corruption can be related to state efficacy.
The results remain basically the same, albeit with some caveats. The most important result, referring to the different nature of the relationship between education and trust in the three continents, largely holds. Table A2 and Figure A1 in the online appendix show this relationship for the models using the ICRG and Freedom House indicators. The results are basically the same. As can be appreciated, education is positively and strongly associated with trust in Europe, it shows a considerably weaker positive association in Latin America, and a negative one in Africa.
We have also tested whether these differences across continents also show up within each continent, using these two indicators of state efficacy. We have found results that are basically consistent with those of the main models, albeit with some minor differences. The models estimated with ICRG basically show the same results as our main models. However, the results are not so consistent when we use the Freedom House indicator. The interaction between the indicator and education is only significant (and in the right direction) in the case of Latin America. In Europe and Africa, although the coefficients of the interaction go in the right direction, they are not significant.
Despite this last outcome, which may be the result of the Freedom House indicator not capturing the notion of state efficacy (or perhaps not as accurately as the ICRG indicator), in general terms the robustness check shows a similar story to the one we have developed in the previous sections.
Conclusion
Our results expose a paradox in trust: in those countries where trust is less needed (those endowed with efficacious institutions), trust is abundant, whereas in countries with weak and failed states, where trust is most needed to compensate for the lack of efficient institutions, it is in very short supply indeed. We have claimed that this is not because there is a direct relationship between state efficacy and trust, but because state efficacy is a precondition for the survival of trustworthy types and, where trustworthy individuals are crowded out, people will tend to distrust (with good reasons) their fellow citizens. This is especially so, we have claimed, for the highly educated, who will probably have more social intelligence, enabling them to distinguish between trustworthy and untrustworthy types. This runs counter to most of the literature on this subject, which claims that education has an independent, positive relationship with trust.
By contrast, our article shows that, as far as the mechanism that links education to trust is social intelligence, this positive relationship is only possible in countries endowed with efficacious states, where pro-social preferences are largely protected from exploitation by opportunistic types. This is not to say that the relationship between education and trust is simply explained by state efficacy; it is rather that whether the social intelligence provided by formal education leads to trust or distrust depending on the survival of trustworthy types in society, which, in turn, depends greatly on the efficacy of the state in sanctioning opportunistic behavior. In Europe and Latin America there is a positive association between education and trust, but in the latter continent this positive association is much weaker than in the former. In Africa, the relation is markedly negative, with highly educated people being the less trustful.
Our results echo other studies that analyze the different effects of education not on interpersonal trust, but on political and institutional trust, depending on levels of corruption: according to Hakhverdian and Mayne (2012), education is only positively associated with political trust in countries that are largely free of corruption, while in corrupt countries the relationship is mostly negative.
These results need to be viewed with some caution but we have uncovered an association between education and trust that is mediated by state efficacy. This points clearly in the direction of our theoretical hypotheses, although with some caveats in the case of relatively efficacious African states. This association offers some insight into the mechanisms that link education and trust. It is probably not due to the social status and privileged position that high levels of education allow. If this were the case, education would have a positive effect on trust in any context. If the shape of this relationship varies according to the degree of state efficacy and, as a consequence, with a variation in the distribution of types (trustworthy versus opportunistic) in society, it is most likely because the mechanism linking both variables is indeed social intelligence. To see whether this is really the case, more research has to be conducted, ideally using experimental data, on the mechanisms that lie behind the intimate relationship between education and trust.
Finally, these results point to the limitations of policies that, according to most of the literature related to this subject, are conducive to higher levels of trust in society. The universalistic policies associated with the welfare state, including the public funding of access to education, including higher education, are usually considered engines of social trust. The results reported in this article bring into question the idea that the welfare state can promote trust simply through its role in raising educational levels. An impartial and efficacious rule of law seems a precondition for trust to grow.
Supplemental Material
IPS779184_French_and_Spanish_abstracts – Supplemental material for Education and trust: A tale of three continents
Supplemental material, IPS779184_French_and_Spanish_abstracts for Education and trust: A tale of three continents by Cecilia Güemes and Francisco Herreros in International Political Science Review
Supplemental Material
Online_Appendix_(final_version) – Supplemental material for Education and trust: A tale of three continents
Supplemental material, Online_Appendix_(final_version) for Education and trust: A tale of three continents by Cecilia Güemes and Francisco Herreros in International Political Science Review
Footnotes
Funding
This work was supported by grant number CSO2015-65969-P, Spanish Ministry of Economy, Industry and Competitiveness.
Notes
Author biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
