Social Capital and Health - Findings from the Scottish Health Survey and Scottish Social Attitudes Survey

2 Analytical Approach

2.1 The analysis conducted for this paper used logistic regression to explore whether or not various aspects of social capital are independently associated with (a) how people view their health in general, and (b) their level of mental wellbeing, as measured by the Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS).

2.2 Logistic regression is a statistical technique that allows you to examine the relationship between a dependent variable (in this case, either general self-assessed health or WEMWBS score), and various independent variables (like sex, age, income, etc.). The analysis identifies which of these independent variables are significantly and independently related to the dependent variable, after controlling for the inter-relationships between variables.

2.3 As the Scottish Government's interest is particularly in the contribution that social capital makes to health, the analysis was conducted in two stages. First, regression models were created, looking at the relationship between the chosen health outcomes and various demographic and socio-economic variables commonly associated with those outcomes (for a full list of variables included, see regression models in Annex A). Second, a number of questions that aimed to measure different aspects of social capital were added to each model, to see how much social capital adds to our ability to explain differences in general health and wellbeing.

2.4 By including both demographic variables and measures of social capital in a regression model, it is possible to assess whether higher or lower levels of social capital appear to be significantly related to better or worse health outcomes even after controlling for demographic differences. This is important, since there are significant demographic differences in the distribution of both social capital and general health and wellbeing. For example, recent UK analysis found that people with higher incomes tend to score more highly than those with lower incomes on various measures of social participation^[3] (Ferragina et al, 2011), while previous analysis of SSA data has shown variations in levels of 'community-connectedness' by gender, age, income and area deprivation (Anderson & Dobbie, 2008). At the same time, analysis of the Scottish Health Survey has shown that general perceptions of health vary with age (with participants less likely to rate their health as 'good' or 'very good' as they get older), while mental wellbeing as measured by WEMWBS varies significantly with socio-economic classification and equivalised household income (Bromley et al, 2009). Without controlling for these demographic factors, it is possible that any association we find between social capital and health is simply a reflection of the fact that both health and social capital vary along particular demographic or socio-economic lines.

2.5 While the findings reported in this paper are primarily based on multivariable regression analysis (the full results of which are presented in Annex A), for ease of interpretation by general readers, significant differences identified by this analysis are illustrated in the main text using bivariate statistics (percentages based on cross-tabulations). Any differences between bivariate and multivariable findings are noted in the text or in footnotes.

Limits of the analysis

2.6 Logistic regression is a useful technique for exploring the relationship between multiple 'independent' variables and a given outcome. However, it can only tell us if a statistically significant and independent relationship exists between each variable and the outcome. It cannot tell us whether or not these 'independent' variables cause this outcome. In the context of this paper, this means that we cannot conclusively say that scoring more highly on any particular measure of social capital is the reason why someone has a higher level of health or wellbeing. It might be that having limited social networks reduces people's resilience and contributes to poor health, for example, but equally it could be that being in poor health limits people's ability to see their friends and relatives.

2.7 We cannot say conclusively which of these interpretations is correct. All this paper can say is whether or not a relationship between the two appears to exist. Further work is required to establish the direction of this relationship. This might involve, for example, a before and after study of participants in an initiative that explicitly aims to improve health via improving levels of social capital. Alternatively, future linkage of SHeS data to data collected about respondents by the NHS should enable analysts to assess whether those who exhibit relatively higher or lower levels of social capital now have better or worse health outcomes in the future.

2.8 Any analysis is also inevitably limited to the data available. For this analysis, the Scottish Government was interested primarily in questions that relate to aspects of social networks and social support, reciprocity and trust, and feelings of efficacy in relation to civic participation. The questions included are shown in Table 1, below. It is of course possible that had we included a different set of social capital variables we would have found different results. In particular, future analysis might usefully look at the relationship between health and volunteering, given the findings cited in Putnam suggesting a positive relationship between the two. However, as neither of the data sources used for this paper included a question measuring volunteering, this was not possible within the scope of this analysis.

2.9 As indicated in Table 1, SHeS 2009 and SSA 2009 each included a slightly different set of questions relating to three aspects of social capital included in the framework for measuring social capital in the UK, developed by ONS (Harper 2002). In particular, questions in SHeS on social networks tap into wider social support (how many people respondents could turn to in a crisis, and frequency of contact with relatives, friends or neighbours), while SSA questions on this topic are framed with reference to support in respondents’ local areas.

Table 1 - Summary of questions included in analysis

2.10 Similarly, this paper focuses on the relationship between social capital and (a) self-assessed general health and (b) mental wellbeing. The former was measured in both SHeS 2009 and SSA 2009 by a question which asks simply:

How is your health in general? Would you say it is …

… very good,
good,
fair,
bad,
or very bad?

2.11 Mental wellbeing was measured (in SHeS 2009) by a series of statements developed by a team of researchers at Warwick and Edinburgh Universities, known as the Warwick-Edinburgh Wellbeing Scale (WEMWBS - see Stewart-Brown, S and Janmohamed, K, 2008 for full details of the items included in this scale). Each person's combined responses to these statements provides a score, which can then be analysed. This paper focuses on patterns in those with a below average wellbeing score (defined as a score at least one standard deviation below the mean).

2.12 These questions were chosen on the basis that they are broad, general measures of health and wellbeing, and that analysis of the relationship between social capital and health should start at this level. It is, of course, possible that there may be different relationships between people's social capital assets and specific physical and mental health outcomes, such as cardiovascular disease or depression. These are possible topics for future analyses.