A considerable body of work claims that high levels of social capital are associated with better health (e.g. Kennedy et al 1998; Rose 2000; Islam et al 2006; Khawaja et al 2006). Thus, social capital has come to be widely considered to be one of the many social determinants of health. Social capital has a variety of definitions; so wide a variety indeed that its usefulness has been questioned:
the concept has been stretched, modified, and extrapolated to cover so many types of relationships at so many levels of individual, group, institutional, and state analyses that the term has lost all heuristic value there does not appear to be consensus on the nature of social capital, its appropriate level of analysis, or the appropriate means of measuring it
(Macinko and Starfield 2001:394-410).
This alone should mean that claims about the association of social capital and health should be treated judiciously. Additionally, the literature on social capital and health (SCH) reveals a set of conceptual and methodological problems that arise from the dominance in this literature of survey research that looks for associations between social capital and health indicators (e.g. Onyx and Bullen 2000; Rose 2000). The difficulty is that, as Bowling (2005) points out, most measures have not been validated. Furthermore, many studies are often based on secondary analysis of data sets gathered by research not designed to measure or explore social capital (e.g. Chavez et al 2004; Pevalin and Rose 2003). In these cases, researchers use answers to questions that may be regarded as proxy measures for social capital. For example, Pevalin and Rose (2003) select as indicators of social capital questions from the British Household Panel Survey on the following topics:
* social participation (meaning participation in organisations and associations);
* frequency of contact with three closest friends;
* perceptions of crime in the neighbourhood; and
* neighbourhood attachment.
They also include questions on social support. However, aspects of social capital emphasised elsewhere, such as trust and reciprocity, are absent. This illustrates how the way that social capital is conceptualised and operationalised may be influenced by the nature of the proxy indicators available.
One effect of such methods is to give social networks and social support (SNSS) a greater importance in the empirical SCH literature than they do in the theoretical literature on social capital. The three most commonly cited theorists offer rather less certainty about the central role of social networks. For Bourdieu (1986), it is not the pleasures and benefits of sociability that constitute social capital, but the resources that social contact and networking can bring to members of affluent and powerful elites. For Coleman (1988), social networks are valued for their normative effects in generating and policing rules of behaviour ('obligations, expectations and trustworthiness; information channels; norms and effective sanctions': S102). For Putnam et al (1993) the key features of social capital are participation in local activities and organisations, and generalised trust and reciprocity. Thus, there is no theoretical consensus about the importance of SNSS.
It would be misleading to imply that there is such a consensus in the empirical literature either. Some studies omit SNSS altogether (Kawachi et al 1997; Lochner et al 2003); some include social networks and social support as separate concepts (Pevalin and Rose 2003); some include social support as part of social networks (Coulthard et al 2002); some understand social networks and/or social support as the primary component of social capital (Snijders 1999); others explicitly exclude social support from social capital but include social networks (Cooper et al 1999); others do the reverse (Looman 2006). This variety illustrates the extent to which SNSS is included in the empirical literature on SCH and the various, and frequently inconsistent ways, in which it is employed in that literature. …