Publication type
Understanding Society Working Paper Series
Series Number
2026-12
Series
Understanding Society Working Paper Series
Authors
Publication date
September 8, 2026
Abstract:
Integrating biological measures into population surveys has expanded opportunities to investigate how social and environmental factors shape health. The gut microbiome is a promising target, reflecting the combined influence of diet, lifestyle, medication and environmental exposures, and offering a route to understanding how social experience becomes biologically embedded. Collecting microbiome data, however, requires a stool sample, raising concerns that such a request may reduce survey participation or introduce selection bias. Using the Wave 16 pilot of Understanding Society, participants were randomly assigned to one of two groups: a group invited to an interview ('main survey'), health measurements and a blood sample; and a stool-sample group additionally invited to provide a stool sample during the interview, with advance letters reflecting this group assignment. We examined whether stool-sample group assignment affected participation in the main survey, and, among that group, which characteristics were associated with consent to provide a sample. Being informed in advance that they would be asked to provide a stool sample was associated with substantially lower participation. 38% of those told they would be asked to provide a sample took part in the main survey, compared with 56% of those not asked, halving the odds of participation (OR = 0.50, 95% CI 0.32–0.78, p = 0.002). The request also appeared to affect who took part: those asked and participating were older, and tentatively less likely to be employed, than non-participants. Among the 91 participants asked, 59% consented to provide a sample, with consent again more likely among older participants and, more tentatively, those of White ethnic background and homeowners. The request therefore acted primarily as a barrier to survey entry rather than to consent. Fewer people took part, and those who did were older, though a majority went on to agree to provide a sample. This points to a risk of selection bias in microbiome data that could distort observed associations between social factors and the gut microbiome, and underscores the need to understand how such requests shape who takes part as surveys increasingly incorporate biological measures.
Paper download#589175