Social capital and health: Does egalitarianism matter? A literature review

The use of social factors to explain community health status is not a new phenomenon. Since Durkheim's classic work on suicide, the importance of social integration and social capital has been recognised for population well-being [1]. Nonetheless, the notion of 'social capital' has attracted wide-ranging attention in the social sciences and public health literature over the last decade [e.g., [2‐10]]. In the public health arena, the concept attracted attention following the work of James Coleman and Robert Putnam [11, 12]. A growing body of empirical research has been conducted on the links between individual (micro) and area (macro/meso) level social capital and population health [e.g. [13‐16]].

Social capital is a multifaceted phenomenon [2]. Among other views, social capital can be considered as a by-product of social relationships resulting from reciprocal exchanges between members involved in social associations or networks and can be recognized as a public good that generates positive externalities facilitating cooperation for the achievement of common goals [17]. It is thought that social capital may generate material/market and non-material/non-market returns to the individual. Material return may include higher wage, better employment prospects or reduced transaction costs, while non-market returns may include improvements in the quality of the individual's relationships and improvements in health or even happiness [3]. It is also hypothesised that there are both direct and indirect returns on the production and accumulation of health and social capital. Direct returns stem from the fact that both health and social capital enhance individual welfare, while indirect returns come about as a result of the observation that health capital increases the amount of productive time, and social capital improves the efficiency of the production technology used for producing health capital [18].

Whilst 'egalitarianism' is a multifaceted notion in social and political thought, common forms of egalitarianism include economic egalitarianism, moral egalitarianism, legal egalitarianism, democratic egalitarianism, political egalitarianism, gender egalitarianism and opportunity egalitarianism [19]. In modern democratic societies, the term "egalitarian" is often used to refer to a situation that favors (for any of a wide range of reasons), a greater degree of equality of income and wealth across a population. Economic egalitarianism, (popular with liberals throughout much of the 20th Century), has given way to a concern not that everyone be strictly equal in material possession, but rather that everyone be equal in having enough material goods to effectively fulfill his or her native human capacities [19].

In the present paper the notion of egalitarianism is considered from material aspects and operationalized on the basis of a country's overall level of income inequality and total public social expenditure as a percentage of GDP. In a recent Luxembourg Income Study (LIS) report, Mahler & Jesuit [20] note that the Gini-index of private sector income inequality in the reputed egalitarian countries such as Sweden (44.1) and the Netherlands (45.8) are almost at the same level as in the USA (44.7) or the UK (47.5). However, the authors point out that the egalitarian distribution of disposable income (post-government income) through government transfers and taxes in Sweden and the Netherlands make the difference with the USA or the UK. Therefore, as a marker of income inequality we consider Gini-coefficients based on disposable income rather than private income. Despite this clarification, there may still remain some difficulties in classifying countries by their degree of egalitarianism. However, for the sake of simplicity we consider a country with a comparatively lower Gini-coefficient based on disposable income and a higher share of public social expenditure to be more egalitarian than a country with the reverse case. Based on country-level Gini-coefficients (based on disposable income) and share of public social expenditure in percentage of GDP, we classified countries as 'egalitarian', 'moderately egalitarian' and 'not egalitarian' as provided in Table 1[21‐23]. Still, there may remain some cautions. Due to the fact that the UK is a country with a relatively higher Gini-coefficient and higher public social expenditure compared with (in our classification) moderately egalitarian countries, its position may be seen as tentative in this study. Although it may be seen as somewhat arbitrary to categorize countries as egalitarian or not, for the sake of parsimony we define Nordic countries (namely, Finland, Norway and Sweden), Germany and the Netherlands as relatively more egalitarian compared to other countries.

Why would a more egalitarian distribution of income and wealth matter to social capital and its impact on health? As a complementary question, one could inquire whether egalitarianism makes any difference in the formation of social capital in a society. In trying to respond to this query researchers have been puzzled by an apparently paradoxical coexistence of a greater stock of social capital and an egalitarian distribution of income and wealth through extensive welfare-state arrangements such as seen in the Scandinavian countries [24]. For instance, Alan Wolfe argued that in societies where citizens are protected "from cradle to grave" by the state, civil society and norms of reciprocity or social capital are "crowded out" [[25], p.142, also cited in [24]]. However, in reality this is not the case. Empirical research shows that in contrast to the trends described in the USA [12] there is little evidence of a decline in social capital in Sweden. Rather, in terms of trust, associational membership and informal interaction, social capital seems to have been maintained at comparatively high levels in Scandinavia [26]. For instance, the World Values Surveys in 1981, 1991–92 and 1995–96 show that 60% of the population in Sweden (second highest, 61% in Norway) agree with the statement that 'most other people can be trusted'; the percentage for the USA is 49 [27]. Recently, Kumlin and Rothstein tried to resolve this paradox empirically using Swedish data. Their findings indicate that welfare-state institutions have a capacity for both making and breaking social capital in Sweden [24]. They show that contacts with universal welfare-state institutions tend to increase social trust; by contrast, means-tested programmes tend to diminish trust.

The next issue to resolve is whether greater (lower) intensity of material egalitarianism can modify the health impact of social capital in a society. It is documented that an egalitarian distribution of wealth and income appears to imply a more cohesive, harmonious society and the level of income inequality may affect both a society's cohesiveness and its members' health [28]. By examining the relationships between income inequality, social capital and health, Wilkinson has argued that the social environment is more cohesive in more egalitarian societies; the members of these societies tend to be less violent, less prone to commit homicide, less hostile to one another, and trust other members more [29, 30]. Using data from 39 US states, Kawachi and colleagues demonstrated that income inequality is positively associated with disinvestment in social capital, which is in turn linked with increased mortality [17]. Kawachi et al. have also shown a strong correlation between trust and income inequality in the United States, while income distribution in states was linked to variations in mortality and, indeed, death from specific causes such as heart disease, cancer and homicide [17].

Within a defined economic and political system (such as a country), the effect of different forms of social capital on health may also be conditioned by the country's intensity of economic egalitarianism. More specifically, it is arguable whether findings from countries such as the USA are generalizable to other countries which are relatively more egalitarian. Besides, unlike the conclusions of most research conducted in the USA, it has been observed that in Canada and Sweden there are no relationships between income inequality and mortality, neither at the level of provinces nor in metropolitan areas [31]. In the Swedish context, Gerdtham and Johannesson examined whether mortality is related to individual income, mean community income, and community income inequality [32]. They concluded that municipality level income inequality had no effect on mortality. At the county level, they found some indications of an association between income inequality and mortality; however, the counties which were more unequal had lower mortality.

Different public social institutions within welfare-states and more equitable distribution of income and wealth in the countries may modify the health impact of social capital or the effect of different forms of social capital (e.g. individual level and area level). Area level social capital (either derived from individual responses or contextual in itself) may not be relevant or important to public health if the societal level of equity or public safety net provision ensures resources for the individual. Given societal variations in economic egalitarianism, it is interesting to consider whether the effect of social capital on health varies among geographical areas according to the degree of equity and safety net provision within countries as well.

The aim of this paper is three-fold. Firstly, we critically review the origins and different forms and dimensions of social capital as it has been operationalized in the empirical literature. Secondly, we systematically review the empirical studies that have examined the health impact of individual and area level social capital for different countries by surveying both single level and multilevel studies. Thirdly, we explore some analytical and interpretational issues that may be pertinent when assessing the health impact of area level social capital. In section 4 we interpret our review results according to both single and multilevel studies and also by country, as well as by fixed and random effect results from multilevel studies. The paper ends with a discussion and conclusions along with some suggestions on directions for future research.

(This section is mainly based on Portes [33] and Woolcock [34]).

In the literature there is no consensus on the intellectual origins or who first implicitly or explicitly introduced the notion of 'social capital'. Different authors with diverse background trace the idea from different intellectual origins. In particular, American sociologist A. Portes argues that the idea is imported from the 19^th-century foundations of sociology and he recognizes that the concept is implicit in the works of Emile Durkheim and Karl Marx [33, 35]. He further claims that Pierre Bourdieu [36] was the first to systematically and explicitly analyze the notion of social capital in the present sense. Bourdieu defines social capital as "the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalized relationships of mutual acquaintance or recognition" [[37], p.248].

After reviewing the intellectual history of social capital, Woolcock notes that though employed for different purposes, the word 'social capital' was used by two major proponents of economic sciences, Marshall and Hicks, 'to distinguish between temporary and permanent stocks of physical capital' [cited in [34], p 159]. He also acknowledges the works of David Hume and Edmund Burke as intellectual origins of the notion [[34] & also cited in [35]]. Refereeing to Swedberg [38], Woolcock further claims "the Durkheimian, Weberian and Marxist traditions within classical sociology were all heavily influenced by the economic debates and issues of that period, and much of what we now refer to as 'social capital' lay at heart of these concerns" [[34] p.160]. As an explicit pioneer of the notion, Woolcock gives credit to Lyda Hanifan (in discussing rural school community centers, see [39, 40] – also cited in [34], endnote 13]. Hanifan uses the term to illustrate "those tangible substances [that] count for most in the daily lives of people" ([40] p. 130). However, both Putnam [41] and Woolcock [34] refer to Jane Jacobs [42] for the first use of the term 'social capital' in its contemporary sense.

In a recent review work, Durlauf & Fafchamps [43] assert that the concept is first introduced into modern social science research by economist Glen Loury [44] who defined social capital as "naturally occurring social relationships among persons which promote or assist the acquisition of skills and traits valued in the marketplace an asset which may be as significant as financial bequests in accounting for the maintenance of inequality of our society" [[44], p.100].

Whilst it is not unambiguous who first implicitly or explicitly introduced the notion of social capital in its present sense, it is however clear that after Loury and Bourdieu, the concept is further developed, modified and disseminated in the diverse disciplines by the works of Coleman [11], Putnam et al. [45] and Portes [33]. The definitions given by these proponents of the notion are cited as follows:

Coleman: "consist of some aspect of social structure and they facilitate certain actions of individuals who are within the structure" [[11], p.302].

Putnam et al: "refers to features of social organization, such as trust, norms and networks that can improve the efficiency of society by facilitating coordinated actions" [[45], p.167].

Portes: "refers to the capacity of individuals to command scare resources by virtue of their membership in networks or broader social structure" [[33], p.12].

Although overlapping to some extent, four main theoretical ingredients can be identified in the definition of social capital: social trust/reciprocity, collective efficacy, participation in voluntary organizations and social integration for mutual benefit [9]. After surveying different empirical studies and definitions of social capital, Durlauf & Fafchamps distinguish three main basic ideas:"(a) social capital generates positive externalities for members of a group; (b) these externalities are achieved thanks to shared trust, norms and values and their effects on expectations and behavior; and (c) shared trust, norms and values arise from informal forms of organizations based on social network and association. The study of social capital is that of network-based process that generate beneficial outcomes through norms and trust". [[43], p.5].

A total of 42 papers were reviewed. These papers come mostly from OECD countries, representing a total of 30 single-level studies (17 individual level and 13 ecological studies) and 12 multilevel studies. These comprised of 13 papers (6 multilevel studies) using US data, 4 studies (1 multilevel) from Canada, 4 from eastern Europe (no multilevel study), 4 from western Europe (2 multilevel studies), 7 from Scandinavia (1 multilevel), 3 from Australia (no multilevel study), and 7 cross-country studies (1 multilevel). A concise summary of the studies on the associations between social capital and health status (mortality or self-rated health/health) by different countries is presented in Table 1. Tables 2a and 2b summarize the sources and characteristics of the studies, i.e. (a) study design (unit of analysis- whether ecological or individual based analysis, data etc), (b) conceptualization and operationalization of the social capital variable (how social capital is defined and measured and kind of measurement), and (c) outcome variable(s) (which response variable(s) was considered for the studies) along with (d) their main findings. In the last column of Table 2a, countries are classified according to their degree of egalitarianism.

In this review we have tried to provide an overview of the definitions and forms of social capital and the probable theoretical associations between social capital and health. Recent empirical studies on associations between individual and area level social capital with various health outcomes have also been reviewed systematically. To summarize, it is evident from our review that most of the studies consider some combination of cognitive and structural forms of social capital and operationalize social capital using individual level trust and reciprocity along with informal participation as indicators. Only one study considered the vertical dimension of contextual social capital using Irish data. Although different studies have conceptualized social capital and combined its several components rather differently, irrespective of a country's overall degree of economic egalitarianism, most single level studies have reported a positive association between social capital (both at the individual level and area level) and better health. However, the findings seemed different for studies employing multilevel analysis where the health impacts of 'aggregated'/contextual social capital have been separated from individual compositional effects. Multilevel studies based on US data have generally demonstrated area level fixed and random effects of social capital on population health with higher intra-cluster correlation. By contrast, Swedish and Canadian studies supports an association in the fixed effects, but area level random parameters are generally very low (i.e. low ICC).

Due to the small number of multilevel studies and very few studies reporting random part results, it is difficult to draw any definite conclusions on the basis of our observations. Nonetheless, the proportion of area level variance to total variance (ICC) was quantitatively similar to those presented in the literature emphasizing the significance of area effects. For instance, Fisher et al. examined the variation in self-reported physical activity among older adults in Portland, USA and concluded that social capital (social cohesion) is associated with increased levels of physical activity among older adults with an estimated ICC of 4% [117]. Using British household cross-sectional data, McCulloch tested variations in neighborhood social capital and social disorganization. For social capital he found ICCs of 9.0% for men and 11.3% for women, while for social disorganization the estimated ICCs were even higher, 21.2% for men and 25.5% for women [118]. The Scottish Heart Health Study found that ICC (district level variance of total variance) was between 0.5% (serum cholesterol) and 5.7% (alcohol consumption and smoking) for coronary heart disease risk factors [119].

On the other hand, in a recent study using a four-year follow up study of the entire Swedish population aged 40–64, Sundquist et al. examined whether neighborhood deprivation (another area attribute, an index measured at small area market statistics level by the use of Care Need Index) predicts incidence rates of coronary heart disease (CHD) [120]. They concluded that high levels of neighborhood deprivation independently influence CHD for both men and women, although the ICCs were reported at 0.9 % for men and 2.1% for women. Other Swedish studies also reported similarly low ICCs. For instance, Lindström and colleagues observed that neighborhood variance in daily tobacco smoking was mainly affected by individual factors other than individual social capital (assessed by social participation) and the intra-neighborhood correlation was 1.9% (ICC) [121]. In another study, Lindström et al. concluded that leisure time physical activity is mainly influenced by individual factors, and area level social capital (aggregated) did not explain the neighborhood difference in physical inactivity. They reported the ICC (intra-neighborhood correlation) to be 2.1% [122].

Some caveats may be identified from our review. Firstly, as we have observed, different studies conceptualized social capital differently within the same country and/or across the countries. How robust are these observations considering the fact that different researchers conceptualize social capital differently? In other words, one could raise the question if we are indeed comparing the same phenomenon across the countries? In particular, social participation and trust have been suggested in the literature as representing different aspects of social capital and both these aspects of social capital are thought to mutually enhance each other [123]. However, in reality this is not always the case. According to Fukuyama's notion of 'the miniaturization of community', the more ideologically narrowly defined social networks and organizations destroy generalized trust between people and this phenomenon is observed in the USA [124]. In the Swedish context, Lindström also makes it evident that poor self-reported health is significantly more prevalent in miniaturized (high social participation/low trust) communities compared to high social capital (high social participation/high trust) communities [92]. However, because of limited numbers of studies across countries, the study found difficulties in further disaggregating studies with similar social capital indicators or health outcomes.

The second issue is a corollary of the first; because of small numbers of studies it is difficult to draw definitive conclusions on the basis of our observations. Moreover, social capital may not always generate better health outcomes. Because of the potential importance of social capital externalities (positive and/or negative), the path from individual to aggregate social capital may be difficult to identify [3]. The benefits that social capital creates for one group may disadvantage another, so that the net spill-over effect on society need not be positive [43]. It was also not possible to scrutinize intra-group effect of social capital through our review.

Finally, most of the studies included in our review had a number of methodological limitations. We may exemplify some of the statistical issues that could be regarded as important here. Firstly, for cross-sectional studies (which is the basis of most of the studies), one cannot exclude the possibility of reverse causality. It is recognized that in many contexts, social capital is endogenous and studies have not employed instrumental variables to allow for consistent estimation of parameters. This problem may be expected to be more problematic for individual level social capital studies. In addition, due to imprecision in the indicators of area level social capital executed in different studies, there is the possibility of measurement errors in the explanatory variable and any such errors would lead to inconsistent estimation of models [125]. A general limitation is failure to include other relevant area level predictors in the models. This could mis-specify the fixed part of a multilevel-level statistical model. However, researchers should be cautious about including other contextual predictors and keep in mind the multicolinearity problem. Keeping these methodological limitations in mind, the conclusions observed in some of the reviewed studies need to be interpreted with caution.

It is evident from the fixed effect results across studies (both single level and multilevel) and countries that the positive health impact of social capital and a country's degree of egalitarianism seemed rather unimportant factors in modifying the effects of social capital on health. Secondly, although a significant fixed effect association was observed between area level social capital and better health in multilevel studies, low variability in health across areas reported in some studies suggests that the differences in health were predominantly affected by individual factors rather than by area characteristics, especially in egalitarian countries.

We have raised the question whether egalitarianism matters in mediating the health impact of social capital and have tried to find some answers through reviewing existing literature. However, many questions still remain unanswered about the notion of social capital and interrelations between social capital and health. Before making further definitive conclusions that social capital is an important determinant of population health (or health differences) or that the egalitarian distribution of income and wealth matters for conditioning the health impact or differences of social capital, careful exploration is warranted through employing appropriate statistical methods (e.g. multilevel analysis) and data. It is expected that the observations from this review will have implications for the content and direction of future research concerning social capital and social determinants of health, on policy formation as well as on the implementation of health service delivery.