Determinants of trust in Turkey Open Access

It is a hard fact that generalized trust¹ as an important component of social capital is crucial in both ‘facilitating coordination and cooperation for mutual benefit’ (Putnam 1995: 67) and constructing social solidarity in a society; and its absence is then argued to be a key cause of social, ethnic and political tension as it works like a glue that holds society together (Luhmann 1979; Ward et al. 2014). In addition, trusting societies rank better on corruption and governance measures (Uslaner 2008a, 2008b). Moreover, a better trust² level is also associated with a better economic environment in countries, which triggers economic development and growth. Fukuyama (1996: 151), for example, argues that ‘it is very difficult to conceive of modern economic life in the absence of a minimum level of informal trust’. Thus, it is essential to analyze the determinants of trust in a society, which will help create efficient and effective social and economic policies. In addition to studying the general determinants of a trust, it is also crucial to find out ‘which groups are distrusted or which groups are more likely to distrust others’ (Ward et al.2014: 2) so that proper social and economic policies can be formulated to have a better functioning society.

Various factors affect the level of trust in a society. For example, the level of diversity – whether it is ethnic or religious – is regarded as an important factor affecting political and economic development in a country, especially through its effect on interpersonal trust. Apart from being a cause of social conflict, it has been argued in the literature that low levels of generalized trust stemming from ethnic diversity in a country not only have a direct negative impact on economic factors such as economic growth (Easterly and Levine 1997; Knack and Keefer 1997; Whiteley 2000; Zak and Knack 2001; Knack and Zak 2003; Beugelsdijk et al. 2004; Alesina and La Ferrara 2005) but they also have an indirect negative impact on economic growth through its influence on the state of democracy and quality of governance in a country.

There are two important streams of literature on the determinants and effects of trust; one relates trust to economic development and the other focuses more on the relationship between trust and social development. Economics literature focuses on the centrality of trust in economic development, especially on the link between economic growth and trust, arguing that increased trust has a direct positive effect on the GDP per capita growth and this direct effect is mainly achieved due to lower transaction costs (Knack 1999) which then helps increase overall investment levels. Furthermore, Knack and Keefer (1997) argue that there is a positive impact of trust on the innovation level, which is a well-known input of production. The other stream of the literature focuses upon the socio-political determinants of trust, arguing that ethnic diversity, good governance and religiosity (Zak and Knack 2001; Delhey and Newton 2005; Hooghe et al. 2009) are strongly related to trust and, it has been argued that low-trust societies have lower institutional quality and higher political corruption (Uslaner 2008a). Moreover, La Porta et al. (1997) found that a higher level of trust is positively linked with judicial efficiency and negatively with government corruption. Besides, some studies argue that social trust has institutional roots (Rothstein 2000; Beugelsdijk 2006) while some others claim that ‘trust in people and trust in government have different roots’ (Uslaner 2002: 7). On this issue, however, Putnam (1993) argued greater civic engagement and social trust lead to better governance in Italy.

In terms of the relationship between diversity and trust, it is argued that people trust their group members (peers of the same ethnic roots, hereafter co-ethnics, or peers of the same faith, hereafter co-religions) more than other groups’ members due to kinship altruism. Thus, this means that the structure of a society in terms of its ethnic and religious divisions has an impact on its level of trust. In other words, the more heterogeneous a society, the lower the level of trust there. However, recent literature finds a non-linear relationship between trust and ethnic diversity. Dincer (2011), for example, argues that there is a trade-off between ethnic diversity and trust up to a certain threshold, but that above this level there exists a positive relationship between ethnic fractionalization and trust, which means that there should be a higher level of interpersonal trust in a highly diverse society than a polarized one. Therefore, he suggests that policies promoting less diversity are an ineffective way to improve trust.

The literature mentioned above generally studies the issue in Western societies, so the determinants of trust in many other countries are still a question awaiting an answer. Whether the lack of interpersonal trust is due to ethnic and religious diversity, or individual and/or societal characteristics, international survey results reveal that Turkey has a very low level of trust. The level of interpersonal trust in Turkey, for instance, was far below the OECD average (59% of people expressing a high level of trust in others on average considering all OECD countries) in 2008; Turkey ranked 29th out of 30 member states. The results of a 2008 poll by World Values Survey also strengthens the position of Turkey being an extreme example of a distrustful society as it finds that only 4.9% of respondents agree with the statement that ‘most people can be trusted’, putting Turkey at the bottom of the ranking again (57th country out of 60). As the level of trust is linked to the various social and economic factors of a country, this high level of distrust in Turkey is not only a bottleneck for its socio-economic development but also for a healthy dialogue and cooperation between different groups in society – be they defined in terms of ethnic, religious or political preferences.

However, even if Turkey is infamous for its low level of trust in society, there is still no quantitative analysis of trust in Turkey linked to identity preferences, ethnicity and the voting behavior of its citizens. It is crucial to study its determinants so that we can understand how the level of trust can be increased and, thus, social dialogue can be raised. This paper aims to fill this gap and trigger further research on the issue by analyzing the sources of interpersonal trust, group trust and political trust in a distrustful and diverse society like Turkey. Another aim of this study is to address the role of national,³ ethnic, religious identity preferences and voting behavior in addition to other individual characteristics in shaping trust. Through incorporating various individual level variables into the empirical model, this study in a way also provides an answer to the question of who has higher or lower trust within the country. With this respect, this paper does not only provide evidence for the level of trust in Turkey but will also contribute to the within-country analysis of trust literature (see, for example, Ward et al. 2014 ⁴) following the footsteps of Fukuyama's (1996) between-country analysis of trust.⁵

This paper comprises four sections: Section 2 provides a brief overview of the data and methodology. Section 3 provides the empirical findings and Section 4 contains the conclusion.

Micro-level public opinion data collected by KONDA Barometer⁶ in September 2012 in 12 regions of Turkey is used for the analysis in this paper. Since the data are cross-sectional, this analysis only explains the variations in trust between individuals at a given point in time. The list of variables in the empirical models is provided in Table 1. Furthermore, average trust levels by ethnicity, voting preferences, and identity choices are presented in Tables 2–4.

As can be observed from these descriptive statistics, generalized trust is at very low levels in Turkey, although it is higher in people who belong to the same ethnic or religious group. On average only 6.3% of respondents agree with the statement that ‘most people can be trusted’, but the average level of trust is far above this level when it comes to trust in co-ethnics and co-religions.

Determinants of trust are analyzed under three different categories. First, a generalized trust measure is examined. The survey question to measure the generalized trust in the data is similar to the most widely used measure of trust in many other survey studies: ‘Generally speaking, would you say that most people can be trusted or that you can't be too careful in dealing with people.’ This question started to be the widely used measure after Rosenberg (1956), and it provides a binary variable for the measure of trust, which takes the value of ‘1’ if respondents agree that ‘most people can be trusted’ and the value of ‘0’ if they state that ‘you can't be too careful’. Based on the averages of the data, the West Marmara region and Mediterranean region have the highest share of trusting people in Turkey; the lowest share of trusting people is in the Central Anatolia region. Moreover, the share of trusting people in the Southeastern region is 6.07%, which is below the same share in Istanbul.

Second, group trust levels are measured under four sub-categories: trust in co-ethnics, non-co-ethnics, co-religions and non-co-religions. The questions related to these categories are similar in their structure. For example, for co-ethnics the question is ‘how much do you trust in people in the same ethnic group with you?’ Respondents choose one answer on a 5-point scale ranging from ‘1 = complete distrust’ and ‘5 = complete trust’. It seems that there is a higher level of group trust compared to interpersonal trust. For example, 35.1% of respondents stated that they trust people in the same ethnic group, while 46.3% says that they trust people in the same religious group as them, which shows the importance of religion in the society.

Lastly, trust in two political institutions – namely, the government and parliament – are measured using again a 5-point scale question. The percentage of people who stated that they trust parliament and the government (Justice and Development Party [hereafter, AKP]) are almost the same: 43.4% and 43.3%, respectively. Concerning the fact that the AKP has the majority of seats in parliament, it is not surprising that these two trust levels coincide. However, it is important to note that these levels are far higher than the level of interpersonal trust and above the level of trust toward co-ethnics. Table 3 shows that although 69% of AKP voters in the 2011 elections trust in the AKP government, the same group's interpersonal trust level is only 5.38%, trust in their co-ethnics 34% and their trust in their co-religionists 50.6%. One can argue that this is an important barrier for an environment of healthy social dialogue in the country as the majority of the people trust the political parties they vote for⁷ more than their fellow citizens.

Determinants of these trust categories are analyzed below using variables of individual characteristics, ethnicity, voting preference and identity choices of respondents. Scaling information and the average values of these variables are listed in Table 1.

The survey questionnaire has four ethnic background categories respondents can choose from: Turk,⁸ Kurd, Arab and Others. On average, 84.3% of respondents stated that they are Turks, 12.2% stated that they are Kurds, 0.7% are Arabs and 2.8% belong to other ethnic groups. Interestingly, individuals from all ethnic backgrounds have higher trust in people from the same religious group; this level of trust is even greater than for their co-ethnics. This shows the importance of religion for the construction of dialogue between different ethnic groups of the same faith. Another striking result is the comparatively lower trust levels of Turks than members of other ethnic groups at all trust levels. Knowing that a majority of population in Turkey define themselves as ‘Turk’, the lower level of trust in this group is worrisome as this high level of distrust may increase discrimination towards minorities and, thus, prevents the formation of social cohesion.

Apart from the role of ethnic background in determining trust, the association of the voting preferences of respondents with three different trust categories is also analyzed. For the voting preference variable, respondents’ choice in the 12 June 2011 election is used. As can be seen in Table 1, 47.4% of respondents voted for the AKP, 20% voted for the CHP, 9.6% voted for the MHP and 2.5% voted for the BDP.⁹ Moreover, 11.8% of respondents did not vote or submitted a blank ballot even though they were eligible to vote.

When we check the average trust levels of respondents with respect to their voting preferences, it seems that BDP voters trust the AKP government the least. Furthermore, there is a large gap between AKP voters and other parties’ voters in terms of trust in government. In terms of the general level of trust, results show that it is lowest among AKP voters (5.38%) and highest among BDP voters (9.68%). This is quite striking as one can argue that the generalized trust level should be lower for BDP voters, the majority of whom are Kurds and have been affected by the country's ‘Kurdish Issue’ for a long time. However, it seems that AKP voters, who are in the majority, are less trustful, and the reasons for this should be analyzed to check if in-groups are constructed by political decisions or ethnic preferences in Turkey. Another interesting result is that trust in co-religions is higher than trust in co-ethnics among all voters. Although the gap between trust in co-religions and co-ethnics is very small among CHP voters (only 3.67%), the same gap is the largest among AKP voters (16.59%).

Furthermore, another distinctive feature of this paper is its analysis of the role of identity choice of people in determining levels of trust. The questionnaire asks respondents ‘Which of your identities precedes the others? Which one describes you the best?’ There are three possible answers to this question: ‘Citizenship of Turkey’, ‘Ethnicity’ and ‘Religion/Sect’. Analyzing the relationship of the answers to this question with trust levels is important as it can signal which should be used as an instrument for an overarching definition of identity.

It was especially after the ‘Kurdish opening’ or ‘Peace Process’¹⁰ of 2009 that the use of the term ‘Turk’ to encompass all citizens of Turkey began to be discussed in the media and academia; it has often been argued that it needed to be replaced, without any other umbrella identity definition being suggested. Until then, national identity, defined on the grounds of citizenship of Turkey, was offered as an overarching identity, the argument being that it would defuse tensions, especially among Kurds. Thus, the results of this part of the analysis will at least enable us to discuss if this hypothesis works on the issue of trust and dialogue among different groups in society.

Descriptive statistics show us that the general level of trust is higher among people who state that national identity (the identity choice of people who state that citizenship of Turkey describes them the best) precedes any other of their identities (6.62%), though this figure is not far above the average level of interpersonal trust in the sample. Moreover, although 81.84% of CHP voters prefer national identity to describe themselves, only 15% for BDP voters and 59.49% of AKP voters state the same. Thus, it seems that, on average at least, the hypothesis of presenting ‘citizenship of Turkey’ as a new overarching identity fails for BDP voters and for 67.11% of Kurds. Instead, the majority of BDP voters and Kurds define themselves in terms of their religion/sect or ethnicity.

Although these descriptive statistics present us a vague picture of the issue, an empirical analysis should be employed to observe if the association of all these different variables on the level of trust stay the same when we also control for other individual-specific variables that are related to trust in the literature. As part of methodology, the ordinary least squares (OLS) estimation technique is used if our dependent variable is defined in several point scales, thus they are assumed to be continuous variables. If the dependent variable is binary, then a non-linear probability model (logit model in our case) is employed. Furthermore, regional fixed effects are included in all of the regression models in order to get rid of the unobservable regional factors affecting trust levels. The results of the empirical analysis are presented in the following section.

This section aims to answer important questions such as ‘What are the determinants of trust?’, ‘Are in-groups defined in terms of identity preferences, ethnicity or political behaviors?’, ‘What should be Turkey's overarching identity which will help improve trust in this diverse society?’ and so on. In order to answer these questions, regression analysis is used and its results are described below.

In regression models, a set of control variables such as education level, age, type of residency, income level, gender, risk aversion level, individual economic expectations and religiosity are included in the regressions not only to minimize the omitted variable bias but also to test the individual theories about interpersonal trust (Uslaner 2000), which argue that personality characteristics such as personal optimism and a sense of control over one's own life (Delhey and Newton 2003) are the main determinants of trust. As Dincer (2011) argues, ‘trust also depends on how optimistic people are about their future and how optimistic they are controlling their own fate’. Thus, perceptions about the future of one's personal economic situation are also added to regression models as a control variable.

Education is another control variable as it is argued in the literature that it has a bearing on trust and a successful professional life makes people more trusting (Knack and Keefer 1997; Alesina and La Ferrara 2002; Dincer 2011). Furthermore, in analyzing the case of Britain, Li et al. (2005) find that education is relatively more important in determining the level of trust compared to other socio-cultural factors and, they found no evidence for the effect of marital status on trust. Moreover, age is used as the control variable as it has been found that interpersonal trust has a concave relationship with age (Alesina and La Ferrara 2002). Li et al. (2005), Hall (1999) and Putnam (2000) find that the generalized level of trust of the young is less than that of the elderly. In terms of gender, women are argued to be more risk averse than men. However, Newton (2001) could not find any significant impact of gender on trust levels.

The dummy variable for whether the individual is living in an urban, rural or metropolitan area is also added to the regression to control for the effect of residency type on the level of trust. Putnam (2000) argues that the level of trust increases among Americans as they live in small towns and rural areas. However, empirical studies are not in consensus about the effect of local communities on social trust levels. For example, Knack and Keefer (1997) could not find any significant relationship between trust level and urbanization, although Delhey and Newton (2005) argue – in their analysis of 60 nations of the world – that people who live in rural areas trust less compared to those in urban societies. Additionally, Dincer (2011) used the share of the people in urban areas in his regression and found that there is an inverse U-shape relationship between urban shares and trust in the US data. Moreover, Uslaner (2006) argues that the level of trust towards different ethnic groups will be higher if there is no lack of contact between different group members. Income level is another control variable in the regression analysis. We can assume that the higher the income level of an individual, the happier she/he will be and thus those people will have higher levels of trust.

Regression analysis is completed in three steps. First, determinants of group trust are analyzed. Then, generalized trust indicators are examined in subsection (b). Lastly, the variables affecting levels of trust in political institutions are investigated.

• Determinants of group trust

Understanding the group trust is especially important in a diverse and/or polarized society as the only difference between in-group and out-group members is group identity and, therefore, it leads us to figure out if the group identity is an obstacle for the social solidarity. Moreover, as it is argued in the literature (see Giddens 1990; Govier 1998), the distrust toward a stranger, foreigner or an out-group member is not necessarily established by prior individual experiences but is rather more about the social or cultural norms that are ‘often based on a constructed characterization or stylized notion of them’ (Ward et al. 2014: 2).

As it is easier to make generalized ‘constructed’ characterizations of groups rather than individuals and, as this way of out-grouping or labeling the ones who belong to the out-group is frequently done by various politicians, it may have an effect on the public discourse and, especially, on the majority-groups in a society which may further increase segregation of the minority from the social fabric. Thus, it is key to analyze both determinants of group trust and the ones who are more distrustful in order to decrease the trust deficit society.

Tables 5–7 present the regression results of determinants, respectively, by ethnicity, identity preferences and voting behaviors. It is found that the level of education is a very important variable to determine group trust. It seems that a higher education level has an increasing effect on group trust and this impact is even larger with respect to trust among different group members. This finding is in line with the literature, as mentioned above. Moreover, the higher the income level, the higher the level of trust toward non-co-ethnics and non-co-religions. These findings therefore show that a successful professional life makes people more trusting toward people from different ethnic and religious groups.

Furthermore, it has been found that religiosity has a statistically significant impact on increasing the level of trust in co-religions, when group trust levels are controlled by other variables. It can therefore be argued that religion cannot be used as an element for the overarching identity in Turkey as it does not make people more trusting toward people from different ethnic and religious groups. Furthermore, as is also argued by Delhey and Newton (2005), trust is not low in urban areas. On the contrary, there is a positive impact of urban residency in trust levels towards co-ethnics and non-co-ethnics, although highly urbanized areas (metropolitan regions) do not have the same impact. Another individual variable used is the level of risk aversion. It was found that people who are more hesitant to take risks have less trust in non-co-ethnics and non-co-religions.

The impact of identity preferences of respondents on group trust levels is also analyzed. Table 5 shows that respondents who state that ‘national identity’ describes them best are less trustful towards co-ethnics, non-co-ethnics and non-co-religions with respect to respondents who choose ethnic identity. Thus, it can be argued that national identity (in other words ‘citizenship of Turkey’) is not a proper overarching identity since it does not have a positive impact on group trust levels and, thus, is not able to improve dialogue between different group members in Turkey. When we checked the effect of religious identity, it is found that respondents who identify themselves by their religion are more distrustful of all groups of society except their co-religions.

Table 6 presents the regression results of the analysis of group trust levels by ethnicity. As can be seen from the table, Kurds trust all other groups more than Turks when group trust levels are controlled for other individual variables. Arabs also have higher levels of trust than Turks. The Turks’ low levels of trust compared to other ethnic groups towards their fellow citizens are an important problem to be analyzed in future research.

The impact of voting preferences of respondents on group trust levels is presented in Table 7. Interestingly, it is found that AKP voters trust less people from different ethnic and religious groups compared to CHP and BDP voters. This shows that politics have an impact in decreasing trust levels and, thus, one can argue that the environment of dialogue in Turkish politics would have a positive effect on the group trust levels.

• (b) Determinants of generalized trust

Table 8 presents the regression results that are discussed in this subsection. The first finding to note is that the voting preference, ethnicity and identity choice of respondents do not have any statistically significant impact on generalized trust levels – except the finding that defining oneself by national identity has a negative impact on trust in people who are seen for the first time. This would also suggest that ‘citizenship of Turkey’ is again found to be an improper overarching identity, as it does not also have an improving effect on the generalized trust levels in Turkey.

It seems that other individual variables are important in determining the generalized trust level in Turkey. First, it was found that females are less trustful. Second, lower levels of education have a negative impact on the probability of having interpersonal trust. Third, people who live in metropolitan regions are less trustful of people they did not know before. And lastly, the more risk-averse people are, the less trustful they are in general.

• (c) Determinants of trust in political institutions

Table 9 shows that, when other variables are held constant, females, on average, are found to trust more in government and parliament than males. Interestingly, the positive impact of education levels on trust is reversed when we analyze its impact on the trust in political institutions. In other words, people with a lower level of education have higher trust in political institutions, as do more religious people. Another important finding is that if people are pessimistic about their economic expectations, they have lower trust in political institutions. Moreover, it was found that identity preference is an important determinant of trust in political institutions and people who prefer to describe themselves by ethnic reference trust less political institutions than people who prefer to describe themselves by their national and religious identities. Moreover, the ethnic background of respondents does not seem to matter at all when our model is controlled by other individual characteristics.

Finally, it is found that AKP voters are the only group that trusts parliament; and CHP voters, MHP voters and BDP voters trust neither parliament nor the government. Karakoc (2013) also found that AKP voters trust national political institutions more. Thus, one can argue that there is polarization in society with respect to political behavior, which also affects the trust levels in other groups in Turkey. Moreover, both voters and respondents who did not vote or were not eligible to vote have lower trust levels than AKP voters. Therefore, it can be argued that getting rid of polarization stemming from voting preferences is crucial in achieving a healthy dialogue in this highly distrustful society.

Whether its very low level of trust is due to ethnic and/or religious diversity, or individual and/or societal characteristics, Turkey is described as a distrustful society according to the results of different international public opinion polls. This high level of distrust in Turkey is a bottleneck not only for its socio-economic development but also for a healthy dialogue and cooperation between different groups in society, whether these groups are defined in terms of ethnic, religious or political preferences.

Although the level of trust is very low in Turkey and its negative effects on socio-economic factors are well known in the literature, there is still no quantitative analysis of interpersonal, group and institutional trust in Turkey that is linked with the identity preferences, ethnicity and voting behavior of its citizens. This paper aims to fill this gap by analyzing the sources of trust in a distrustful, diverse society and to address the role of identity preferences, voting behavior and ethnic background in addition to some other individual characteristics in shaping trust. Moreover, the analysis in this paper investigates if the concept of ‘citizenship of Turkey’ can be used as an overarching identity in Turkey instead of the current practice of defining every member of society as a Turk.

Although the findings of the analysis are described in Section 3 in detail, some of the striking outcomes can be summarized as follows: Investment in education and improvements in labor market conditions are necessary to increase group trust levels in Turkey as professional success makes people more trusting. The low levels of trust especially in individuals with less than a high school diploma indicate the importance of the primary education system in shaping trust levels. Uslaner (2008b: 291) notes that ‘moralistic trust does reflect experiences, but those early in life rather than as an adult, and trusting values formed early in life persist through adulthood’. Thus, the fact that social trust is lower for less educated people in Turkey can then be linked to the spurious ideas taught in the education system because students have for a long time been taught ideas such as ‘Turkey is surrounded by enemies’ and ‘Turkey has no friends apart from the Turks themselves’ and so on in history classes. The similar rhetoric is also employed by various politicians throughout the Turkish political history and, the negative impact of the education system can be observed in various spheres. For example, conspiracy theories are the main menu in any discussion about economics and politics. It is not striking to hear a university student, when asked about his or her views on the 2013 Gezi protests, say: ‘The youngsters are being fooled. This is a global set up.’¹¹

Moreover, it is found that national identity (‘citizenship of Turkey’ as an identity preference) is not a proper overarching identity since it does not have a positive impact neither on group- nor generalized trust levels and, thus, it is not able to improve dialogue between different group members in Turkey. Furthermore, Kurds and Arabs trust all other groups more than Turks do when group trust levels are controlled for other individual variables. This finding is especially striking because the literature for Western societies argues that minority groups have lower trust levels. Alesina and Ferrara (2002), for example, argue that ‘belonging to a group that historically felt discriminated against’ is one of the strongest factors associated with the low trust in the US. Furthermore, Uslaner (2002) argues that people with a high generalized trust level are more inclined to be tolerant toward people who are different from them. This difference could be in terms of race, ethnicity, place of birth, class and so on. As the Kurdish issue is one of the big obstacles for peace and socio-economic development in Turkey, one can imagine how important it is to increase the general trust level of especially Turkish residents in order to ensure that ethnic differences will be more ‘tolerated’ by the majority population.

It is also found that, on average, AKP voters trust less people from different ethnic and religious groups compared to CHP and BDP voters when other control variables are taken into account. This shows that politics in Turkey have a decreasing impact in trust levels and in constructing a healthy environment of dialogue in Turkey since political behavior has a negative and statistically significant impact on group trust levels. Putnam (1995: 71) differentiates tertiary associations from secondary ones (organizational affiliations that make people interact with each other such as parent–teacher associations, civic and fraternal organizations) where membership does not necessarily mean encountering any other member. Instead, as Putnam (1995: 70) states, ‘their ties, in short, are to common symbols, common leaders, and perhaps common ideals, but not to one another’, which puts the role of tertiary associations into suspicion when it comes to social trust. I argue that the finding of this paper (lower out-group trust levels for AKP voters which form the majority in the country) signifies the fragility of social bonds in the society as political parties may die but social trust should live long to ensure a healthy civic life of a society. Thus, governments and other political parties should cooperate in establishing social trust in Turkey and try to eliminate ‘partisanship’ being a strong barrier for a trusting society.

Furthermore, it is found that respondents’ voting preference, ethnicity and identity choice do not have any statistically significant impact on generalized trust – except for the finding that people who define themselves by national identity have, on average, less trust in people who they did not know before. Furthermore, when the determinants of trust in political institutions are analyzed, it is found that only AKP voters trust both parliament and the government, and this finding also leads one to suggest that getting rid of polarization stemming from voting preferences is crucial to achieve a healthy dialogue in this highly distrustful society as voting behavior is found to affect not only institutional trust but also social and group trust.

Moreover, as is discussed above, the majority population (Turks) and the majority voters (AKP voters) trust members of out-groups less. This is especially alarming in the sense that, as Uslaner (2004) argues, the ‘strong in-group and weak out-group trust’ phenomenon is linked to high levels of corruption. For example, Gambetta (1993) and Varese (2001) argue that mafia-type organizations exist in Italy and Russia, respectively, due to strong in-group trust but weak generalized trust in those societies. These examples do not necessarily lead one to conclude that low out-group trust is the reason for higher corruption levels, but they do show that there is at least a strong association between the two in those cases. Moreover, Uslaner (2008b: 293) argues that there is a vicious cycle between low trust and corruption, noting that ‘high inequality leads to low trust and, thus, to more corruption – and grand corruption, where high-level officials get rich as ordinary people stay poor, further erodes trust’. Therefore, it is very important to find out and implement necessary policies to improve out-group trust levels of especially majority population in Turkey to avoid negative externalities related with it.

To sum up, in a society with very low levels of social trust, ‘coordination and cooperation for mutual benefit’ (Putnam 1995: 67) will be at insufficient levels to solve dilemmas such as ethnic tensions. Therefore, it is of crucial significance for the government and policy-makers to improve the society's level of civic engagement through effective policies and, thus, encourage the emergence of higher social and group trust levels, which will boost socio-economic development in Turkey.

I would like to thank KONDA Research and Consultancy for their generosity in providing the micro-level data.

No potential conflict of interest was reported by the author.

Aysegul Kayaoglu is an Assistant Professor in Economics at Istanbul Technical University, Turkey. She earned her Ph.D. degree in Economics in 2014 from Université Catholique de Louvain, Belgium. Besides, she is a graduate of the European Doctoral Program for Quantitative Economics in 2014 during which she also studied at the London School of Economics and Political Science (LSE), UK, where she was a teaching fellow in the Departments of Economics and Management between 2011 and 2014. Her main research interests are applied development economics, population economics, political economy and informal institutions.

Aysegul Kayaogluhttp://orcid.org/0000-0003-1484-184X