Social capital, income inequality, and self-rated health in 45 countries

Abstract

There has been growing interest in the relationship between the social environment and health. Among the concepts that have emerged over the past decade to examine this relationship are socio-economic inequality and social capital. Using data from the World Values Survey and the World Bank, we tested the hypothesis that self-rated health is affected by social capital and income inequality cross-nationally. The merit of our approach was that we used multilevel methods in a larger and more diverse sample of countries than used previously. Our results indicated that, for a large number of diverse countries, commonly used measures of social capital and income inequality had strong compositional effects on self-rated health, but inconsistent contextual effects, depending on the countries included. Cross-level interactions suggested that contextual measures can moderate the effect of compositional measures on self-rated health. Sensitivity tests indicated that effects varied in different subsets of countries. Future research should examine country-specific characteristics, such as differences in cultural values or norms, which may influence the relationships between social capital, income inequality, and health.

Introduction

There has been growing interest in the relationship between the social environment and health (Institute of Medicine, 2001; Institute of Medicine & National Research Council, 2002; National Research Council, 2001). Among the concepts that have emerged over the past decade to examine this relationship are socio-economic inequality (Lynch & Kaplan, 1999; Marmot, Bobak, & Smith, 1995; Wilkinson, 1996) and social capital (Kawachi & Berkman, 2000; Kawachi, Kennedy, Lochner, & Prothrow-Stith, 1997). Following Putnam's work on the importance of civic society and community ties (Putnam, 2000; Putnam, Leonardi, & Nanetti, 1993), social capital seems to have especially captured the interest of many social epidemiologists (Islam, Merlo, Kawachi, Lindström, & Gerdtham, 2006; Kawachi & Berkman, 2000; Kawachi et al., 1997; Lochner, Kawachi, & Kennedy, 1999; Macinko & Starfield, 2001; Poortinga, 2006; Szreter & Woolcock, 2004), although not all are convinced of its utility in implementing effective health policy (Lynch, Due, Muntaner, & Smith, 2000).

Sociologists Pierre Bourdieu and James Coleman were among the first theorists to define social capital. Bourdieu called it “the sum of the resources, actual or virtual, that accrue to an individual or a group by virtue of possessing a durable network of more or less institutionalized relationships of mutual acquaintance and recognition” (Bourdieu & Wacquant, 1992, p. 119). Coleman (1988) defined social capital by its function, stating that “social capital inheres in the structure of relations between actors and among actors” (p. S98). Like Bourdieu, Coleman identified it as a means of providing actors with resources that could be used to achieve their interests. Expanding on Coleman and Putnam, Kawachi and Berkman (2000) defined social capital as “those features of social structures—such as levels of interpersonal trust and norms of reciprocity and mutual aid—which act as resources for individuals and facilitate collective action” (Kawachi & Berkman, 2000, p. 175). A recent literature review has documented two components of social capital, a cognitive component that includes “norms, values, attitudes and beliefs” such as “people's perceptions of the level of interpersonal trust, sharing and reciprocity” and a structural component that “refers to externally observable aspects of social organization, such as the density of social networks, or patterns of civic engagement” (Islam et al., 2006, p. 5). Along these lines, most discussion of social capital has considered it to be a positive asset for a society to have. Portes (1998, p. 15) cautioned, however, that social capital can lead to negative outcomes such as “exclusion of outsiders, excess claims on group members, restrictions on individual freedoms, and downward leveling norms”. Negative outcomes such as these are often ignored in research (Islam et al., 2006).

A great deal of the research in social capital and health has been linked to the literature on socio-economic inequality and health. Most studies of the latter have focused on the distribution of income, starting with Wilkinson's (1996) seminal work in which he demonstrated that higher income inequality is associated with lower life expectancy in wealthier countries. Wilkinson's findings generated an explosion of research into income inequality and health, some supporting (Kawachi, 2000; Marmot, 2002; Marmot et al., 1995; Subramanian & Kawachi, 2004; Wilkinson & Pickett, 2006), some refuting (Lynch et al. (2001), Lynch et al. (2004); Mackenbach, 2002; Osler et al., 2002; Ross et al., 2000; Shibuya, Hashimoto, & Yano, 2002) his findings. Results varied in different studies depending on the time period, the sample of countries used, and the level of measurement. Studies using more recent data measuring income inequality at the state or country level have generally supported Wilkinson's findings (Ram, 2006; Wilkinson & Pickett, 2006).

The literature has focused on two theories describing the pathways leading to health inequalities. Using a psychosocial approach, social epidemiologists such as Wilkinson (1996), and Kawachi and Berkman (2000) describe how social comparisons can increase stress levels and how this can be modified by access to social capital. Other social epidemiologists take a more materialist approach. For example, Lynch and Kaplan (2000) have maintained that the material conditions at different SES levels are the fundamental mechanism by which inequality is related to the health gradient. More recently it has been recognized that both approaches can contribute to the understanding of how income inequality and social capital may be related to health (Schnittker & McLeod, 2005; Szreter & Woolcock, 2004). There is, moreover, growing consensus that income inequality itself is not the cause of poor health outcomes, but it can be an indicator of a broader social context of structural inequality that affects health through discrimination, dominance hierarchies, violence, and under-investment in human or social infrastructure (Deaton, 2003; Eckersley, 2006; Islam et al., 2006; Lynch & Kaplan, 1999; Subramanian & Kawachi, 2004; Wilkinson, 2005).

Social capital is typically considered to belong to the psychosocial pathway leading from the social context to health outcomes (Kawachi et al., 1997; Wilkinson, 1999). There is some question, however, as to whether social capital is best examined at the individual or ecological level (Islam et al., 2006). While Portes (1998) describes social capital as a property of individuals, much of the literature has described social capital as a collective characteristic (Bourdieu & Wacquant, 1992; Coleman, 1988; Putnam, 2000) that intercedes between the social context and health outcomes (Carlson, 1998; Islam et al., 2006; Kawachi & Berkman, 2000; Kawachi et al., 1997; Kennelly, O’Shea, & Garvey, 2003; Lochner et al., 1999; Macinko & Starfield, 2001).

Cross-national studies measuring social capital and socioeconomic position at the individual level have generally found fairly robust relationships with health (Islam et al., 2006). In two such studies comparing respondents in European countries, Carlson (1998), Carlson (2004) found that participation in civic activities had a positive effect on health, as did economic satisfaction. Carlson (1998, p. 1364) also found that self-rated health was consistently lower in the formerly Communist countries, demonstrating “that the so-called European health divide, documented for mortality, is also noticeable in self-perceived health”.

Results of ecological studies of the relationship between income inequality, social capital and health have been mixed, with supportive results typically found only in studies conducted within the United States (Islam et al., 2006). Until recently, international studies carried out since Wilkinson's (1996) publication had not generally supported associations between societal level income inequality, social capital and population health (Kennelly, O’Shea, & Garvey, 2003; Lynch et al. (2001), Lynch et al. (2004)). This trend changed when Ram (2006) conducted an ecological analysis comparing 108 countries using better data than had been used previously and reported a strong negative association between income inequality and life expectancy. Ram also found that social capital (measured by trust) was negatively associated with income inequality, although it was not related to health.

In ecological studies it is possible that the effects of social capital and income inequality on health may be confounded by other contextual characteristics, such as type of political system (Navarro & Shi, 2001) or differences in culture (Eckersley, 2006; Forbes & Wainwright, 2001). Furthermore, several authors have argued that it is important to distinguish between compositional and contextual effects of social capital measures (Lochner et al., 1999; Poortinga, 2006; Subramanian & Kawachi, 2004; Veenstra, 2005). Compositional effects are found when differences between groups are largely due to differences in the characteristics of the individuals belonging to the groups and contextual effects are found when group level characteristics are associated with differences in outcomes after controlling for the relevant individual-level confounders (Diez Roux, 2002). “The association between indicators of social capital and self-rated health could well be due to the fact that more socially isolated individuals are concentrated in communities lacking in social capital” (Poortinga, 2006, p. 293). Multilevel analysis can take both levels into consideration simultaneously and thus has the potential to distinguish between contextual and compositional effects (Diez Roux (1998), Diez Roux (2000); Duncan, Jones, & Moon, 1998). Further, it avoids the ecological and individualistic fallacies of using data at one level to draw inferences at another. Additional advantages are that multilevel methods statistically account for the non-independence of observations within groups and can examine random variation at both the group and individual levels.

Proponents of multilevel modeling have also suggested that interactions between group and individual measures of social capital may be important (Poortinga, 2006; Subramanian & Kawachi, 2004). In comparing 22 European countries using multilevel methods, Poortinga (2006) found that personal social support networks and trust were significantly related to health, but aggregate social capital measures were not significantly related to health directly. Instead, the aggregate measures exerted strong moderating effects in interactions with the individual level variables.

Olsen and Dahl (2007) conducted another hierarchical analysis of health in 21 European countries that controlled for various lifestyle and political and economic characteristics. Like Poortinga (2006) they found that individual measures of income satisfaction and social capital had stronger relationships with health than did societal measures. A history of Communism was more important at the societal level; like Carlson (1998), Carlson (2004), they found self-rated health to be consistently lower in Eastern Europe than in Western Europe. However, they did not take cross-level interactions into consideration.

To date, all of the cross-national studies using multilevel methods have been limited to Europe. Blakely and Woodward (2000) have stated that different countries may have different levels of exposure to any macro-level environmental risk factors such as socio-economic inequality or low amounts of social capital. To avoid a lack of sufficient variation in the ecological exposure studied, it is important to repeat the observations over time or across different societies. Therefore, it would be beneficial to use a larger sample of societies that includes non-European and developing countries to see if the same types of patterns hold across different social contexts. For this reason, we applied multilevel methods to a large and diverse sample of countries to test the hypothesis that health is affected by social capital and income inequality cross-nationally.