Background Previous research has identified the role of social capital in explaining variations in health in the countries of the former Soviet Union. This study explores whether the benefits of social capital vary among these countries and why.
Methods The impact of micro social capital (trust, membership and social isolation) on individual health was estimated in each of eight former Soviet republics using instrumental variables to overcome methodological hazards such as endogeneity and reverse causality. Interactions with institutional variables (voice and accountability, effectiveness of the legal system, informal economy) and social protection variables (employment protection, old age and disability benefits, sickness and health benefits) were examined.
Results Most social capital indicators, in most countries, are associated with better health but the magnitude and significance of the impact differ between countries. Some of this variation can be explained by interacting social capital indicators with measures of institutional quality, with membership of organisations bringing greater benefit for health in countries where civil liberties are stronger, whereas social isolation has more adverse consequences where there is a large informal economy. A lesser amount is explained by the interaction of social capital indicators with selected measures of social protection.
Conclusion When considering interventions to improve social capital as a means of improving population health, it seems advisable to take into account the influence of macrocontextual variables, in order not to overstate or understate the likely impact of the intervention.
- Instrumental variables
- politics, social capital
Statistics from Altmetric.com
It is now clear that, despite its superficial trappings of superpower status, the Soviet Union was a deeply dysfunctional society and, from the 1960s onwards, failed to achieve either economic growth or improvements in population health.1 It was also a society where social interaction was characterised by distrust and isolation. The communist regime maintained a pervasive control over the lives of its citizens and the centrally planned economy discouraged individual initiative. Political opposition was banned, as were most forms of criticism. Social interaction was channelled through formal organisations, such as the party, trade unions and other associations controlled by the regime, and spontaneous interaction was implicitly and explicitly discouraged. Personal relationships were typically limited to a narrow circle of relatives and friends and it was difficult to trust outsiders who might pass information to the authorities, leading to long periods of incarceration or worse.2 Survival frequently required the use of imaginative circumvention of official restrictions, bringing the rule of law into disrepute.
The legacy of this situation was a low level of what has been termed ‘social capital’.3 4 The term ‘social capital’ was introduced in 1977 by Bourdieu5 and subsequently operationalised and tested empirically by Coleman.6 It comprises a combination of trust, social support, norms, and information channels and has been linked to progress in areas ranging from economic growth to the functioning of institutions.7 8
The importance of the elements comprising social capital became apparent in the transition to a market economy. Many of the formal support systems, often linked to places of employment, were no longer available. Economic restructuring preceded the creation of new social institutions and many people were unsure where to turn to for help when faced with personal crises such as job losses.3 9 The scale of the problem can be seen in table 1, which shows levels of trust, participation in local voluntary organisations, and confidence in key institutions (the army, press, labour unions, police and parliament) using data from the third round of the World Values Survey (WVS), undertaken mainly between 1999 and 2000 (see notes in table 1).
The mean degree of participation in local voluntary organisations is only 14% in this region, far below the 54% seen in high-income countries. Although the average level of trust is at 25%, comparable to other low- and middle-income countries, it is also below the 36% observed in high-income countries. Confidence in institutions (except the army and press) is also lower than elsewhere in the world, reflecting a yet incomplete transition.
This could have adverse consequences for health. Social capital has been invoked as a determinant of health,6 8 10 11 acting through a variety of mechanisms.12 First, at the micro-level, social capital favours cooperation and interaction among individuals, which in turn enables them to assist and care for those falling ill and to provide economic support to those facing shocks, such as sudden unemployment, death of a breadwinner or adverse climatic conditions. Second, intense cooperation and social interactions facilitate the flow of health-related information. Third, positive interactions with other people can have positive consequences on psychological well-being. However, it cannot be assumed that benefits will flow, and much may depend on the characteristics of social capital in different settings, in particular the extent to which relationships bond members of existing groups or bridge divisions with others.13
The empirical evidence tends to support the idea that social capital may matter for health in transition countries, mostly based on ecological evidence14–16 but with some individual-level analyses,17 including a recent study in which data were pooled from eight former Soviet countries.18 Using instrumental variables to overcome certain methodological problems discussed below while simultaneously controlling for many individual, household community and country effects, it was found that those who trusted a majority of people were 7% more likely to report being in good health than those not trusting others. On the other hand, individuals who felt isolated were 11% less likely to be in good health than those who were not. However, there was no clear association between health and membership of what are referred to as ‘Putnamesque’ organisations, characterised as being based on horizontal, egalitarian relations and exemplified by sports clubs or religious and charitable organisations.7 These contrast with ‘Olsonian’ ones, such as political parties, trade unions and professional associations that tend to reconfigure redistribution systems in their favour at the expense of the rest of the society and which were dominant in Soviet society. The study findings were consistent with previous research from elsewhere.10
Earlier work did not, however, explore whether the effects of social capital on health varied between countries. The present study moves two steps further, taking advantage of the availability of consistent survey data from eight countries, by first documenting how the impact of social capital on health differs between countries and second, by attempting to explain any differential impacts. This draws on the work of Bobak et al, who assessed the impact of factors such as corruption on individual health in transition countries using multi-level analysis. They found that corruption is damaging to health even after controlling for individual socioeconomic factors.19
A better understanding of the contextual factors that promote or reduce the potential health benefits of social capital should help policymakers to explore ways in which improved social capital can improve health in a specific country. The present results suggest that diversity in areas such as civil liberties and political voice may influence the meaning of conventional measures of social capital and hence their impact on health.
Data and methods
Living conditions, lifestyles and health (LLH) surveys
The individual-level data are from the Living conditions, Lifestyles and Health (LLH) surveys, conducted in eight former Soviet countries - Armenia, Belarus, Georgia, Kazakhstan, Kyrgyzstan, Moldova, Russia and Ukraine. National samples number around 2000, except for Russia (4000) and Ukraine (2500). Samples were selected using multistage random sampling, with stratification by region and rural/urban settlement type. Within each primary sampling unit (about 50–200 per country), households were selected by random sampling from a household list (Armenia) or by standardised random route procedures (other countries). Out of each household one person was chosen, whose birthday was closest to the date of the interview. Interviews were carried out in all countries throughout autumn 2001. Response rates varied between 71% and 88% among countries. Further details have been reported elsewhere.20
Health production function
The determinants of individual health can be represented by the following equation:(1)where i stands for the individual, j for the community in which the individual lives and k refers to the country of residence; Hijk is a measure of health equal to 1 if individuals self-report to be in good health, and 0 otherwise, Cjk is a vector of explanatory variables at community level, Xijk is vector of explanatory variables at individual level, SCijk are the social capital indicators (at individual-level) and εijk is the disturbance term. For any individual i, community is defined as the set of individuals living in the same town or village, although in Armenia only the region was used as data did not provide information about the precise place of residence.
The set of covariates Xijk includes age, level of education, employment status and gender of the respondent. Also, household characteristics were controlled for related to the size and the number of working household members and the material and economic conditions of the household. Additionally, access to healthcare facilities was controlled for through the inclusion of two variables measuring, respectively, distance from the respondent's residence to the nearest hospital and to the nearest doctor and an indicator of material circumstances derived from the water quality to which the household has access. The variables are defined in the supplementary appendix. Finally, in order to capture all unobserved community heterogeneity, community fixed effects were included.
Three social capital indicators at individual-level are considered: trust, membership and social isolation, all represented by dummy variables. Trust takes the value of 1 if the respondent declares that the majority of people deserve to be trusted. Membership takes the value of 1 if the respondent reports membership of a voluntary local organisation (eg, religious, sporting, artistic, musical, neighbourhood, youth, women and charitable organisations). Social isolation equals 1 if the respondent feels alone and isolated from the community. These are all commonly used indicators in the empirical social capital literature.21
Assessing the impact of social capital on individual health at a micro-level must address at least three concerns. First, individual measures of social capital are self-reported and thus dependent on the respondent's attitude, state of mind during the interview and past experiences. Therefore, the available measure reported by the respondent is likely to differ from the true level of social capital, with an uncertain direction of bias. This is a non-random measurement error, which makes self-reported indicators of social capital endogenous and hence unsuitable for a standard Ordinary Least Squares regression.
Second, there is reverse causality, as membership and social isolation are likely to be influenced by individual health. A healthier respondent will find it easier to participate in activities of local voluntary organisations; a sick respondent, in contrast, might feel socially isolated because of his or her own constraints in meeting friends. To address these two issues, instrumental variable estimates were used, as in an earlier study using pooled data from all eight countries.18
Third, it is necessary to disentangle the role of social capital from other factors. For instance, economic and material conditions at individual-level and community-level determine both individual health and individual social capital. An affluent person will find it easier to obtain healthcare of good quality and to participate in charitable organisations or sporting associations. Omitting common covariates would lead to an underestimate or overestimate of the true impact of social capital on health. This has been addressed by controlling for many variables at individual-level and household-level, and by using fixed effects to capture all unobserved heterogeneity at the community-level.
The three social capital indicators are instrumented by ‘special’ community averages of the three social capital indicators: community averages are calculated for each individual as the mean of all other individuals living in the same community. These averages are ‘special’ in the sense that they vary at the individual-level. This is to avoid spurious correlation, especially in small villages.
Although the choice of instruments parallels that in an earlier study, where a detailed discussion is included, it is nonetheless necessary to present briefly the key elements of the justification. Individual self-reported perceptions of social capital are a distorted reflection of true social capital, a reflection shaped by individual idiosyncrasies. Technically, self-reported social capital is social capital plagued by measurement error, which is likely to be correlated with individual observable and unobservable characteristics. Instead, the averages of social capital over the (other) members of the community are likely to be uncorrelated with such idiosyncrasies and at the same time are likely to be correlated with true social capital, as they are the product of the same social context in which the individual lives. However, such instruments would not be acceptable if community social capital had a direct impact on individual health or an indirect impact through other community-level variables. To address this concern, all the community-level relevant determinants of individual health were captured by including community fixed effects into the model. Additional information on the validity of the instruments has been published previously.18
Observations containing missing data have been omitted from the analysis. The resulting sample comprises slightly more than 11 000 respondents with information on all relevant variables.
Country estimates: what drives country differences in the return to social capital?
The first step in the analysis was to estimate equation (1) separately for each country, while dealing with the endogeneity of social capital. The second step sought to explain the differences among countries. To that end, the data were pooled and each social capital indicator was interacted with variables capturing potentially relevant policy characteristics at the country-level. In particular, the influence of one set of proxies for the quality of governance and another set capturing the extent of formal social support policies were considered. Although a diverse set of governance indicators was used, the overarching hypothesis is that better governance will increase the ability of social capital to create greater health benefits. For instance, greater democracy should allow emerging social networks to work openly, free from repression. Indicators of formal support policies were selected because it was hypothesised that they may act as a substitute for informal social capital. Hence, formal social support would be expected to diminish the association between social capital and health because formal systems would substitute for informal ones.22–24
Table 2 shows the six variables selected: 1. the voice and accountability (VA) indicator is an index of political and civil liberties and of human rights in 2000.25 2. The logarithm of the number of days necessary to enforce a contract (DAYS) is a measure of the effectiveness of the legal system and, hence, of the transaction costs incurred by agents when contracting with others.26 3. The share of the unofficial economy (UE) is a proxy for the importance and frequency of informal transactions, outside the protection of the legal system. The last three variables capture different aspects of social protection: 4. the level of employment protection (EP), 5. old age and disability benefits (ADB) and 6. sickness and health benefits (SHB). The latter is an indicator that captures the generosity of social protection in the event of an employee's sickness. The variables UE, EP, ADB and SHB are available only for six out of eight countries. VA ranges between −2.5 and 2.5. UE, EP, ADB and SHB range between 0 and 1. Additional details are reported in the note to table 2.
All these variables were rescaled so that one benchmark country takes the value of 0: for voice and accountability Moldova was defined as the reference (ie, the best scoring country) and for the remaining indicators the worst performer was used. This allows the coefficients of the non-interacted social capital indicators to be interpreted as the social capital impact in the reference country.
Finally, a community-level (rather than country-level) variable was interacted with the indicators of social capital: the population size of the community. In this way, the hypothesis that the impact of social capital would depend on the community population was tested: trusting others should be easier in small communities, with greater opportunities for repeated interactions and lower costs of monitoring and acquiring information.27 Hence, the impact of trust on health would be expected to be stronger in smaller communities.
Table 3 reports instrumental variable community fixed effects estimates for the pooled sample and country by country. The results for each country are broadly in line with those for the pooled sample, discussed at length in an earlier study: for most countries trust has a significant positive impact on health, social isolation a significant negative impact and membership no impact. Nevertheless, there are some differences in the magnitude of these effects. As for membership, Russia and Moldova are different in that they show a positive influence (significant only for Moldova).
Table 4 (columns 2–7), based on the full sample, reports the instrumental variable fixed effects estimates of the health equation, augmented by the interaction of each social capital indicator interacted with one of the measures of democracy, effectiveness of the legal system and social protection discussed in table 2. Column 8 reports the interaction with community size. As a reference, non-interacted estimates are reported in column 1.
Turning first to VA, in column 2, it appears that civil liberties and political participation do not affect the impact of trust or social isolation on self-reported health. By contrast, the impact of membership on individual health is influenced by VA in the sense that in countries with a high level of civil liberties and political participation (higher than about −0.5) there is a substantial positive impact of membership on health. Moldova is the country with the highest level of VA in 2000, among the eight countries of the sample. This explains one of the idiosyncrasies noticed in the country-specific estimations according to which the effect of membership on health was biggest in Moldova.
Column 3 reports the effect of a measure of the number of days necessary to enforce a contract (DAYS). The level of transaction costs does not influence the health effect of any of the social capital indicators. Column 4 shows that the larger the informal economy, the more dangerous it is (healthwise) to be socially isolated.
The next three columns report the results of the interactions with EP, ADB and SHB. The only significant interaction is with employment protection. When this is strong the impact of trust on health is diminished.
As predicted, in smaller communities trust has a larger impact although its size remains small in absolute terms (column 8). On the other hand, social isolation has a negative and significant impact in larger communities.
This study takes advantage of a unique dataset, with information collected consistently and simultaneously from eight countries with a shared cultural legacy that developed during 70 years of communist rule, yet with diverse social and political experiences in the subsequent decade. This provides a valuable opportunity to understand how national characteristics, many within the realm of government, influence the well-established contribution of social capital to health.
It is confirmed that, as in other parts of the world, social capital makes an important input to the production of health in this region. Indeed, as shown in an earlier study, it rivals some of the more traditional health determinants. Thus, the impact of trust on individual self-reported health turned out to be slightly higher than the impact of adequate access to water.18 However, both the level and health impact of social capital vary among these countries. It was attempted to interpret these differences by taking into account certain country-specific economic and institutional characteristics. Overall, the present results suggest that returns to social capital are larger in more democratic countries and in countries where contracts are more effectively enforced.
The finding that civil liberties have very little influence on how those aspects of social capital that relate primarily to the sphere of family and friends (such as trust) affect health is intuitive as it is here that the institutional framework is least relevant. This is also the case for the impact of social isolation on health. However, well-entrenched civil liberties, which encourage the emergence of voluntary associations, create a situation whereby the membership of such associations has an especially beneficial effect on health. This is consistent with other research showing how such associations support circulation of information and provide an extended network that can be relied upon in case of need.28–30 On the other hand, in undemocratic countries, which tend to repress and discourage voluntary associations, being a member of such an association can be associated with discrimination and other forms of persecution (such as restricted access to medical facilities or even physical and psychological violence). In such countries, associations are less likely to fulfil their role as a forum for information exchange and mutual support. In these circumstances, it is understandable that membership may negatively affect individual health.
The effectiveness of the legal system has no impact on the effect of social capital on health as the social deals triggered and favoured by social capital are typically informal and do not require any legal protection.
The danger of being socially isolated in the presence of a large informal economy is explicable by the risk of being excluded from local networks, which can provide opportunities of economic transactions and occasional jobs, as well as opportunities to obtain goods otherwise not accessible.
The diminished impact of trust on health where employment protection is well developed, with formal insurance against the risk of being fired, is likely to be due to informal insurance mechanisms mattering less in such circumstances. However, the extension of this argument would suggest that higher old-age and sickness benefits should also reduce the importance of social capital, as formal and informal insurance act as (partial) substitutes. In particular, it would be expected that the negative impact on health of social isolation should be reduced. However, this was not the case and might indicate that informational and psychological benefits provided by aspects of social capital dominate those relating to economic protection and care provision. However, caution should remain about this interpretation as the indicators used display limited variability across countries.
The influences of community size on the relationship between trust and social isolation and health can be interpreted from a relative ‘deprivation’ perspective: the detrimental effect of social isolation on health is inversely related to the social isolation status of neighbourhoods. In large cities, there will be more opportunities for social exchange, so that the expected average level of social isolation is lower, and the perception of social isolation and its negative effect on health is more readily perceived.
In light of the present results, traditional health policies that have thus far focused on improving healthcare infrastructure should consider broader intervention strategies that incorporate social capital promotion, for instance, by supporting associations, voluntary groups and spontaneous networks. Such policies could be particularly beneficial in the former Soviet Union, where there is great scope for improving social capital, coupled with an even greater need to reduce what are some of the worst adult health outcomes in the world. In principle, policies to promote social capital may be pursued in two ways: by providing financial and/or in-kind support to allow social capital to develop more easily, or by generating ‘enthusiasm among communities and their leaders to develop social capital’.12 In practice, as Kawachi and Berkman noted earlier, there are hardly any examples in the literature analysing interventions that intentionally seek to improve social capital.31 A notable example is a recent, encouraging study by Pronyk et al, who demonstrate the positive effects of an intervention in rural South Africa, combining group-based microfinance with participatory gender and HIV training in order to promote changes in solidarity, reciprocity and social group membership as a means to reduce women's vulnerability to intimate partner violence and HIV.32
The testing of interventions to promote social capital in the transition countries context, where there appears to be particular scope for improving social capital, should be high on the agenda of applied social and health policy research.
What is already known on this subject
Social capital, assessed by levels of trust, membership of organisations, and social connections, has been linked to better health in studies from many parts of the world.
However, it is known to be low in the countries of the former Soviet Union, with recent research showing that this is a factor in the high levels of poor self-reported health in this region.
There are, however, differences in the association between the various measures of social capital and health in different countries in the region.
What this study adds
It has been shown how some of the observed variation in the relationship between components of social capital and health can be explained by measures of institutional quality.
Membership of organisations brings greater benefit for health in countries where civil liberties are stronger, whereas social isolation has more adverse consequences where there is a large informal economy.
Funding European Commission. Funding for the contributions of Béatrice d'Hombres (while she was still with the University of Padua) and Lorenzo Rocco from the WHO European Office for Investment for Health and Development, Venice, is gratefully acknowledged. The LLH surveys were funded by the European Commission's Copernicus Programme. Grant number ICA2-1999-10074.
Competing interests None.
Ethics approval The paper uses anonymised data: the original data were collected following approval by ethics committees in each of the participating countries, under the auspices of the teams collecting the data. See: http://www.llh.at/llh_partners_start.html.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.