Social trust and democratization in the post-communist societies

Social trust and democratization in the post-communist societies

Recent assessments of the transformation processes in the ex-communist societies tend to agree on the increasing diversity in the quality and extent of their democratization (q). One possible category of suspects for accounting this variation relates to the level of social capital in these countries (q). In his 1993 book, Putnam deplores the fact that proposals for strengthening market economies and democratic institutions [of developing and transitional countries] center almost exclusively on deficiencies in financial and human capital.' (Putnam 1993 b, p38) He considers that the deficiencies in social capital in these countries are at least as alarming, and asks, as a solution, for more efforts to encourage social capital formation by 'patiently reconstructing those shards of indigenous civic associations that have survived decades of totalitarian rule' (Putnam 1993 b, p39).

In studies that followed Putnam's book, generalized trust, which in most conceptualizations of social capital is considered as one of its basic components (q), has been shown to have different distributions among the public of the post-communist societies than among the one from consolidated democracies. More specifically, the proportions of people who state that 'most people can be trusted', are found to be systematically lower in surveys conducted in the East than in those conducted in Western Europe (q).

This chapter will focus on one category of consequences that trust may have on transition within the former communist countries. The main assertion is that generalized trust is a resource for democratization processes.

The fact that a positive correlation between trust and volunteer membership exists is a well established result, both theoretically and, in the case of the Western nations, also empirically. It is highly debated what is the direction of the causal arrow between the two terms, but a general agreement exists on the fact that generalized trust has something to do, even when the effect of other factors is controlled, with membership to associations (q). Yet, proving the existence of a positive correlation in the case of the East European countries is not as straightforward as expected. The fact that at the macro level a non-positive relation can be found is a result that holds with different data and measures, and that seems to contradict the existing pattern found among the Western nations.

My aim is this chapter is to demonstrate that a more detailed methodological treatment leads to a different image then one produced by macro-level analysis, showing, in fact, important similarity across both West and East-European countries. Thus, when analysis is performed at an individual level and the effect of contextual factors at a national level is controlled, in any of the thirteen ex-communist countries under study is found that volunteer members in associations tend to be more trustful than ordinary citizens. Still, the relations are generally weaker in transitional countries than in the western ones. I will bring evidence in favor of two possible explanations: the validity problems of the standard measure of generalized trust, and specific attributes of volunteer associations which are linked to the ethnic context, could each decrease the intensity of the observed relation between trust and membership.

Finally, I will discuss the relationships among activity in associations, civil society and democracy, within the former communist countries. I will argue that despite of the fact that not all volunteer organizations are democratic, the overall effect of volunteerism on democracy is positive in each of these countries.

Trust and civic involvement

There are several possible lines of argumentation in favor of a positive causation from social trust to level of democratization. An important part of the reasoning is based on the assertion that trust is a resource for collective action. Citizen activism is significantly affected by the capacity of citizens to identify common goals and to pursue them collectively. First, it is expected that a more trustful citizen become embedded in more dense and extended social networks. Social trust, in particular, may have the potential to increase the connectives through networks that favor mobilization for participatory acts. Second, a high level of social trust seems to ease empathy towards other interest, by identification with their own. Finally, formal models and experiments showed that more trustful citizens tend to be better in overcoming collective action dilemmas.

A throughout empirical check of the causal relation from social trust to citizen activism is especially problematic because of the difficulty in measuring the later term. One possible strategy is to focus on volunteer membership to associations, considered as an important component, but also as a proxy of a larger range of citizen participatory acts. It certainly misses a broad category of relevant activities, consisting in ad hoc or in regular but uninstitutionalized collective behavior, which are by no means negligible for democratic processes (Tarrow, 1994). However, these activities are difficult to record by standardized interviews, and even more to make their measures comparable across different societies. The analyses in this chapter will rely mainly on estimations of volunteer membership in associations, as they are recorded in the 1990-1993 and 1995-1998 World Values Surveys and the 1999 European Values Surveys.

It is widely acknowledged that social capital can produce not only social goods but also social bads (Portes 1998). Social capital has the potential to empower groups of people for any kind of collective action, which sometimes have desirable consequences for their members and bad consequences for other groups. The former communist countries tend to be characterized by low level of social capital on some of its dimensions, and by high level on some of the other. Ethnic heterogeneity, more prevalent traditional social relations, the numerous cases of ineffective state institutions, have enhanced the role of personal networks (Rose .), but in the same time, have lowered interpersonal and institutional trust (q). Especially the multi-ethnic character of a society could significantly affect the equation between trust, volunteerism and democratization. The simple and optimistic view is that trust positively affects membership to associations, which in turn serves as a resource for democratic transition. However, not all associations are necessarily good for democracy, and, in particular, a high membership to groups that stress ethnic cleavages could enhance the level of conflict and undermine those collective actions which transgress ethnic lines (q). Therefore, in the case of the transitional societies, the effect of trust on democracy, mediated by volunteerism, depends on which of the following statements holds true:

(1) social trust tends to be a resource only for those associations that have a net positive effect on democratization;

(2) social trust is a resource for any associations, including the ones that can be harmful for democracy.

Consequently, the ideal empirical testing should distinguish between associations that are 'good' or 'bad' in terms of their democratic effects. However, not only that information on activities performed by associations are scarce but, as I will discuss later, assessing the net democratic effect of a certain type of organization is not straightforward. The available data will allow me to do an indirect test regarding the effect of ethnicity on the relationship between social trust and democratization.

Civil society, volunteerism and democracy.

While almost absent from intellectual and political life for nearly a century and a half, the concept of civil society gained a wider popularity during the '70s by its use to the context of Eastern Europe. The fall of communism has brought a new situation where the term embraced additional meanings and, mostly as a result of developments in the social capital field, raised new expectations concerning its explanatory power (Foley and Edwards, 1998). Within and with regard of the ex-communist countries, the language of civil society has become a visible but polisemic part of the discourse of politicians, academics, business leaders, foundation executives, and citizens. For many actors that oversee democratic transition within the region, civil society refers to 'dynamic webs of interrelated nongovernmental institutions' (Keane, 1998, p. 6). Thus, from this perspective, the level of development of civil society can be assessed by looking at the number of nongovernmental organizations, the number of people involved in their activities, and by various aspects regarding their activity. To the others, civil society and nongovernmental sector do not necessarily overlap. Timothy Garton Ash, for instance, considers as an open question whether in Eastern Europe nongovernmental organizations are 'like seeds to the beautiful turf of civil society or like that thing called astroturf--artificial grass that covers the ground so the real grass cannot grow.' (2000, p. 400). Mark Warren considers volunteer behavior as central to a definition of civil society, in this way excluding a significant part of nongovernmental associations that are active in the East European countries. He makes the distinction between 'pure' associations, and associations that are 'of' civil society, but not 'in' civil society. The former consists in organizations, within which voluntary associative relations are dominant, whereas the later are organizations that mediate between 'pure' associations, states and markets. Political parties and unions are two main examples for the category of mediating forms of associations (Warren 2001 p. 58). In this perspective, the proportion of people volunteering in associations, other than parties and unions, provides evidence on how developed civil society is in a certain country.

On any of these conceptualizations, the term civil society designates very different realities from a nation to another. It is well documented the large variance that exist among the western nations, for instance between the Scandinavian ones and the U.S. (q) East-European countries present certain specificity but, also a large variance among themselves, in both quantity and quality. Clear differentiation existed during the communist period, and there are compelling arguments that a previously more developed civil society accounts at a large extent for a well functioning civil society in the present. One the one hand, countries such as Poland, and Hungary had, during their communist regimes, a significant amount of population involved in organizations, formal and informal, that retained a high degree of autonomy from the state. Poland stand at one extreme with up to ten million people in Solidarity movement, and also the only private higher education institute in the communist East Europe, a catholic university. Similarly, in Hungary there were some officially accepted and even supported nonprofit organizations, such as trade unions, Red Cross, Adult Education Society, Patriotic Front, Chamber of Commerce, etc., and a large variety of voluntary associations (e.g. fishermen's, hunters' associations, sport clubs, pet fans' societies, voluntary fire brigades, amateur theatre groups, youth clubs, intellectuals' organizations, folk dance houses, etc.) considered to be harmless and therefore tolerated by public authorities (Kuti, 2002).

On the other hand, the former soviet countries, Bulgaria, Romania and Albania, were at another extreme, with very limited autonomous activity of groups of people (Nelson 1996, Sampson ?). In Romania, for instance, a number of nongovernmental organizations were allowed to exist, some of them even with international affiliation. They included traditional craftsmen, philatelists, sport clubs, or associations for people with disabilities. A few of them, such as the Writers Union, were influent enough to negotiate privileges for their members, but their influence on policies were either insignificant or uncivic (Verdery ?). A special case is that of the so called Cenaclul Flacara, which consisted in a series of gatherings taking place in all major cities, and organized by a poet close to the circles of power. The audience, consisting in young people, sometimes in number of tens of thousands, was encouraged to sing, dance, create literature. Cenaclul Flacara can be regarded as an associative movement, providing support for social and cultural exchanges, but in the same as an instrument of mobilization by the communist regime in favor of nationalistic values. (Stoiciu, 2001).

The type and amount of civil society in the past is relevant because some of the old organizations have survived over time and count as part of civil society at the present; in addition, some of the skills, orientations and knowledge acquired during the past membership has been employed in the present organizations, and transmitted through family socialization. No less important is the fact that attitudes of population and of the officials towards the civil society have been influenced by its degree of development and by its role played under the communist regime. In Hungary, for instance, 'governments did not trust [associations] at all, the most dictatorial ones even tried to completely eradicate them, but they were held in high esteem by citizens.' (Kuti, 2002). Again, Romania presents a very different case, with civil society type of organizations facing mainly negative attitudes throughout the transition period[1].

One consequence of the extremely diverse picture of civil societies and of their dominant definitions across ex-communist nations, is that validity of the comparative measures for both civil society and civic engagement, and, consequently, their relationships with trust can be influenced by factors that are country specific.

There are three mechanisms through which volunteer associations might produce democratic effects in a transitional society:

1. Social learning effects. Associations may contribute to forming and enhancing those attitudes, knowledge and skills on which democratic practices are based. While in western nations family, school, place of work and friends provide important instances for transmitting democratic capacities, citizens of the East-European nations cannot rely at the same extent on such mechanisms, which are more resistant to change over time and, in their countries, still bear undemocratic traces. Less influenced by local tradition, and frequently bringing an import of organizational practices from their western counterparts, volunteer associations in transitional societies have an important potential to develop cognitive and deliberative skills, civic virtues and a sense of efficacy.

2. Public sphere effects. Associations in general, and volunteer ones in particular, provides 'the social infrastructure of public spheres that develop agendas, test ideas, embody deliberations, and provide voice' (Warren, 2001, p.61).

3. Institutional effects. These are the more established effects of associations on democratic governance. Associations speak on behalf of groups of citizens, contributing in this way to the aggregation of individual judgement and to representation of collective decisions. In addition, associations have the capacity to affect state and market by organizing demonstrations, strikes and civil disobedience. Civic associations in the Eastern-European countries have become known mostly because of the role played during the fall of communist regimes, and, later, of their constant input to shaping the new institutions.

How strong are these effects, and what is the relative importance of each for the transitional countries remain unanswered here. Although I fully acknowledge that not just volunteer but also non-volunteer associations have the potential to influence democratic processes, and that their influence can be sometimes negative, I will rely on the assumption that in each of the new democratic polity benefits from the activity of volunteer associations exceed loses.

Empirical testing: macro-level relationships between trust and civic engagement.

The relation between generalized trust and volunteerism, measured at a country level, has been consistently found to be positive in Western nations. In the 1999 European Values Survey, the correlation is 0.39 for all countries, and has the same value when the relation is limited to the West-European ones. In the case of the former communist countries, the same relation displays a completely different pattern (Figure 1), which is summarized by a correlation of -0.21.

Figure 1. Relationship between the proportion of people who 'trust other people' and the proportion of volunteers.  1999 European Values Survey.

What could explain this difference between the new and the established European democracies? It could be that generalized trust has different effects on civic involvement across the two categories of countries. Nevertheless, it is possible that specificity is only apparent, having in fact processes that are similar but reflected by distorted instruments. Measurement errors, model under-specification, statistical assumptions that are not respected, could each potentially explain why results are different at an aggregate level between the two categories of countries.

The next two sections of this chapter will investigate measurement aspects regarding the key concepts of this analysis. One compelling reason in favor of performing this inquiry is the extremely low reliability of measures across several comparative studies that were conducted in transitional countries. Thus, generalized trust and volunteerism, which in national studies have been found to be stable over time, display unsystematic fluctuations when measured in cross-national surveys (Table 1. and Table 2).

Table 1. Correlations between aggregate measures of generalized trust in three cross-national surveys: 1990 World Values Surveys, 1995 World Values Surveys and 1999 European Values Surveys.

Note: cell entries represent Pearson correlation coefficients between proportions of respondents who declare that 'most people can be trusted'. In parentheses are given the number of cases (countries).

Table 2. Correlations between proportion of members in volunteer associations* in three cross-national surveys: 1990 World Values Surveys, 1995 World Values Surveys and 1999 European Values Surveys.

Note: cell entries represent Pearson correlation coefficients between proportions of respondents who declare that are members in volunteer associations; there are slight differences in the way that question was asked in the three comparative studies. In parentheses are given the number of cases (countries).

Measuring volunteer activism.

Identifying people who did volunteer work in associations is a difficult task, particularly when standardized interviews are used as a tool[1]. Several results of an overview of the estimates of volunteer membership among the Romanian public, could also apply at some extent to other post-communist countries:

1. Slight variations in the way that questions are asked, elicit large variations in the proportions of respondents who acknowledge their volunteer behavior. Table 3. shows estimates between 3.7% and 23.3%. Because they are close in time, within few years or even months, differences are not attributable to changes in real membership.

2. Assessments which are based on other data than national representative surveys (e.g., surveys on associations, qualitative interviews, official statistics) indicate that the proportion of Romanians who were volunteers in associations did not exceed 3% at any time since 1990 (Badescu, 2001b). That indicates that most measures based on national surveys overestimate the real proportion of members. One of the important implications is that any correlation between membership measured by a citizen survey and other variable, at an individual level, will be underestimated.

Table 3. Proportions of volunteers in associations, as estimated by several surveys, representative for the adult population of Romania.

Notes:

estimates the proportion of those who performed volunteer activities during the last 12 months

estimates the proportion of the active members

Measuring social trust.

Trust is even more difficult to measure in a comparative study than membership to associations. First, cross-national surveys rely on the assumption of equivalence among translations, which in the questions used to assess the level of trust may not be accurate enough. A study on nine surveys conducted in Romania that contained various items of generalized trust, including three different translations of the same dichotomous question used in the World Values Surveys and European Values Surveys, showed significant variations of results induced by small changes in wording (Badescu 2001a). We can assume from here that translations of the same item in different languages can produce even more different estimates of the proportions of trusters in those societies. This type of problem, however, is not so important when relationships between the variable on trust and other variables are assessed at an individual level, as long as one can assume that estimates produced by different translations can be linked by linear transformations .

A second problem associated with the assessment of the level of social trust is that of validity. It has been long assumed that the questions of generalized trust, such as the standard one--'Generally speaking, do you believe most people can be trusted or can't you be too careful in dealing with people?'--, which is asked in the WVS and EVS, are operationalizations of that type of trust which is praised for a broad range of benefits, especially the ones that involve solving collective dilemmas. Only recently, it has been argued that trust is multi-dimensional, and that its dimensions are distinct from each other empirically as well as analytically. Moreover, the type of trust that 'helps us get things done' in general, and has the potential to increase membership in associations in particular, was found to be largely similar with the one which is measured by the standard question (Uslaner 2002). However, this result is based exclusively on Western data, and its extension to the case of the ex-communist countries needs to rely on empirical ground. This is a general principle that applies to any cross-national analysis, but in this case, there is an additional reason for performing a validity assessment. Eric Uslaner shows a clear distinction between, on the one hand, trusting 'most people', which means trust in strangers, and, on the other hand, to 'put our faith in people we know or we place our confidence in people we think we know well (folks like ourselves)', which he labels as particularized trust. Trust in strangers and trust in people that the subject knows (relatives, friends), or share some attributes with him/her (people of the same ethnicity, confession) are shown, using survey data on American public, to have distinctive empirical manifestations (Uslaner 2002, ch.3). However, the public from the post-communist societies presents several characteristics that may influence their representation of 'the other people' about which they are asked to evaluate in terms of trust. In comparison with the Western world, the former communist nations tend to have much lower internal mobility and, in most cases, an almost non-existent immigration. In addition, they have a larger proportion of people living in rural localities, which tend to be small, characterized by extended kinship relations, and low interaction beyond their borders. 'Most people' could have a different meaning for someone whose contacts are almost exclusively with relatives and people who has known for his entire life, than for a person who has moved several times, had travel extensively, and has lived in places that undergone rapid changes in their social structure as a result of migration and immigration. The former category of people is better represented in the East, whereas the later has a larger proportion in the West. From these reasons, I hypothesize that respondents from the former communist nations are influenced in their assignments of 'other people' as trustworthy or untrustworthy, not only by their level of trust in strangers, as it was shown to be the case of the American public, but also by their level of particularistic trust.

I will test this assertion by using the data of a survey representative for the Romanian adult population, the October 1999 Romanian Barometer Survey, one of the few in the post-communist countries that ask about trusting specific groups. This survey provides indicators about trust in other people, people of other ethnicity, people of other religion, other Romanians, family members and neighbors. I will use structural equation models in order to test if the six observed measures can be considered as manifestations of two latent concepts, the notion of trust in strangers and that of particularized trust, respectively. More specifically, the questions asking trust in family and trust in neighbors are expected to be significantly stronger correlated to one of the latent variables than with the other. Because both questions refer to people about whom the subject has at least some degree of familiarity, the first factor would be considered as a measure of particularized trust. Similarly, I assert that the questions asking trust in people of a different ethnicity and trust in people of a different religion are stronger correlated with the second latent variable. Following Uslaner's argument in his analysis on the U.S. data, this later factor will be considered to be a measure of trust in strangers (Uslaner 2002 ch.3). Trust in people of the same ethnicity is also expected to correlate more to the variable of particularized trust and less with trust in strangers. Because in the Romanian society ethnicity has been a salient issue during the last ten years, and also the public is well aware of its role played in some neighboring countries as a source of violence, war and territorial disputes, it would not be a surprise to find that subjects tend to consider the other's ethnicity as a relevant attribute in judging its level of trustworthiness. In other words, the tendency to put faith in own family or neighbors, that is people on whom the subject has previous knowledge, may work similarly in the case of people with whom the subject had no previous contacts but share a common attribute, considered as important.

Table 4. summarizes the main results of data analysis. It is according to expectations that five of the six measures of trust are significantly influenced (p < .05) by only one or another of the two latent variables. The levels of trust in people of other ethnicity and in people of other religion are strongly correlated with the measure of trust in strangers (0.74 and 0.66 respectively) and not significantly correlated with particularized trust. Trust in own family, in neighbors and in people of the same ethnicity are positively correlated with the 'particularized trust' factor (0.34, 0.56 and 0.72, respectively) and not significantly correlated with trust in strangers. The only measure of trust that is correlated with both factors is trust in other people. It is different to the result on the U.S. data that particularized trust has a stronger effect on this variable than trust in strangers (0.30, compared to 0.19). Another dissimilarity is that the two factors have a stronger correlation in the Romanian data than in the U.S. one  (0.61, compared to 0.39).

The overall result shows more similarity than dissimilarity, which is remarkable giving the difference in the wording of questions and statistical technique. If the same result holds in two largely different societies, then the possibility of its generalization to other contexts is significant. However, the difference, should make us cautious when comparing countries based on the standard trust questions. If trust in strangers has a stronger influence than particularized trust on the quantity and quality of social involvement of citizens in democratic societies, then we should expect that the standard variable on trust is more useful in explaining democratization where it is a better measure of trust in strangers.

Table 4. Structural model with two latent variables for trust, Trust in Strangers, and Particularized Trust, in October 1999 Romanian Barometer Survey.

Trust Measure Particularized Trust Trust Strangers

Most People Can Be Trusted  .30 .19

Trust People of Other Ethnicity  - .74

Trust People of Other Religion  - .66

Trust People of Similar Ethnicity  .72 -

Trust Your Family .34 -

Trust Your Neighbors  .56 -

Notes:

Entries are standardized coefficients (Lambda-X); maximum modification index for the coefficients constrained to zero is 1.41; Comparative Fit Index (CFI) = 0.99; Adjusted Goodness of Fit Index (AGFI) = 0.99; Correlation between the two latent factors = 0.61.

Relationships between social trust and volunteerism, at an individual level.

Performing individual level analyses, separate for each country, has clear advantages over studying only one country level model of the relation between generalized trust and volunteerism. On the one hand, it solves what statistical analysis describes as 'ecological correlation problems' (Huckfeld and Sprague ). On the other hand, it controls for any of the country level effects, including the ones induced by translation specificity. 

When strength of the two variables is compared across countries, there is another statistical aspect that needs to be taken into consideration: standard coefficients, such as Pearson correlation, tend to underestimate the intensity of relations between variables having non-normal distributions. This is exactly the case with both membership to associations and generalized trust. In almost every country under study, trusters and volunteers are less than half of the total, but in transitional countries their proportions are, on the average, closer to zero. Therefore, the relations between the two variables have to be assessed with statistics that are not influenced by their distributions. Table 5. summarizes the polichoric correlations and odds ratio-s for each of the East European country, as well as for their pooled sample and for the pooled sample of West European countries.

Table 5. Relationships between social trust and volunteer membership to associations.

This table shows a completely reversed image to the one from the country level analysis. Thus, in none of the thirteen East-European countries under study the relation between trust and membership is negative, and in eight of them is significantly positive (p < 0.05). This result suggests that the underlying social mechanism linking trust and volunteerism operates in similar ways in different contexts. Countries from Eastern Europe, having diverse results in their democratic consolidation and having quite different structure and amount of civil society, and countries from the Western Europe, all display remarkable similitude from this point of view. Still, the relation tends to be weaker in transitional countries than in the western one. Two possible explanations have been mentioned: the validity problems of the standard measure of generalized trust, and specific attributes of volunteer associations which are linked to the ethnic context, could each decrease the intensity of the observed relation between trust and membership. I will test both these assertions using the available data.

First, I will take again the case of Romania, one of the four countries having a positive but not statistically significant relation (p = 0.05). As I showed in the previous section, the standard measure of generalized trust has in its case a stronger correlation with a latent variable of particularistic trust than with one of trust in strangers. I expect than the aggregate measure of trust in strangers, which I argued that it provides a better operationalization than the standard question of trust, will have a positive correlation with volunteer membership, and stronger than both standard question and the latent variable of particularistic trust. Indeed, a structural equation model that includes these variables shows a statistically significant relation only from trust in strangers to membership: beta = 0.10 (t = 3.7). This result suggests that for Romania, and perhaps also for other post-communist countries, better measures of generalized trust would lead to stronger positive relations with volunteer membership in associations.

Second, I will test if the intensity of relation between trust and membership is affected by ethnic context. One possible reason for having the strength of the relation between the two factors lowered by ethnic diversity is that the proportion of associations whose functioning is based on ethnic division is expected to be higher when the proportion of ethnic minorities is higher. In the same time, I expect that this kind of associations do not benefit from a higher level of generalized trust, therefore the correlation between membership and trust would be non-positive in their case.

An analysis of the 1999 European Values Survey data shows a positive relation between the proportion of ethnic majority and the intensity of relation between trust and membership. If Moldova is added, using data from another survey[3], the correlation is 0.55 (p = 0.04). When data on ethnicity are more refined, the correlation coefficient exceeds 0.65 .

Figure 2. Relationship between the proportion of ethnic majority and odds ratio between trust and membership. 1999 European Values Survey and CID 2001 for Moldova.

Concluding remarks.

Main statements:

- generalized trust is positively correlated to membership

- the strength of relation increases when better measures are used

- it is unlikely that correlation is due exclusively to the causal relation from membership to trust (see this volume, Uslaner 2002)

- correlation between trust and membership depends on ethnic context. It makes plausible the fact that trust do not determine membership in associations stressing ethnic ties. So trust is a resource for associations, but rather selectively, possible more to the one which 'good' for democracy.

- volunteer associations are a resource for developing democracies

Conclusion: generalized trust is a resource for democratization in transitional countries.

References.