Absent but not silent: the impact of requiring a minimum number of contacts on egocentric social networks
Guy Harling, Elyse Jennings, Stephen Tollman, Lisa BerkmanBackground: Research shows that using different name generators or setting maximum numbers of social contacts can result in respondents naming different subsets of their social network. However, evidence on the effect of forcing a minimum number of respondents is limited, particularly in terms of whether respondents name “empty” or unimportant others, and who the ‘forced alters’ might be.
Methods: We took advantage of a change in the structure of a social network module between the first two waves of HAALSI, a longitudinal population-based study of rural South Africans aged over 40. In wave 1 respondents were asked to name up to six others important to them for any reason; in wave 2 they were asked to name exactly six (in both waves any unmentioned spouse was also added). We use this modification in the survey to evaluate how reported social networks changed when the response constraint was changed from a maximum of six names to a hard requirement for exactly six names.
We described changes in composition, (emotional, informational, physical and financial) support provision and egocentric network density. We then leveraged randomly assigned interview dates in wave 1, where interviewers increasingly acted to constrain the number of respondents alters could report, as a natural experiment to evaluate how social network composition changed with greater alter constraint. We did this by running regressions to see how month of wave 1 interview correlated with change in social network characteristics, interpreting changes across the 12-month interview process as reflecting the effect of increased constraint.
Results: The 3576 respondents (77% of baseline sample) who completed the social network module at both waves reported a mean 3.3 contacts in wave 1 and 6.1 in wave 2, with a similar approximate doubling from 66 to 130 monthly communication events. Mean levels of social support rose even more (between 119% for information and 170% for emotion) and conflict events (fighting, arguing, criticism) rose more than five-fold. Weighted egonet density fell from 0.55 to 0.46, but alter-alter density rose from 0.34 to 0.36.
Regression analysis showed that additional named individuals were similar to those already named in terms of kinship (~80% kin) and gender (80% same-gender) but were significantly younger and non-significantly more likely to live in the same household; regression also suggested that while additional named individuals led to more support of most kinds, this was not true for financial support.
Discussion: Forcing respondents to name more social contacts in this rural South African setting does not appear to have led to listing ‘empty’ alters, instead capturing unnamed individuals – notably younger household members – who provide substantial support and still formed part of a cohesive personal network. Further research evaluating the acceptability to respondents of forcing a fixed number of alters is needed to determine whether such approaches are a useful tool.