In this paper, we investigate whether individuals provide consistent responses to self-assessed health (SAH) questions in the UK Household Longitudinal Study (UKHLS), and the potential implications for empirical research in case of inconsistent reporting behaviour. We capitalise on an opportunity in the UKHLS, asking respondents the same SAH question twice: with a self-completion and an open interview mode, within the same household interview over four waves. We estimate multivariate models to explore which individual characteristics are systematically relevant for the likelihood and frequency of inconsistent reporting. About 11%-24% of those reported a particular SAH category in the self-completion reported inconsistently in the open interview. The probability of inconsistency is systematically associated with individual’s demographics, education, income, employment status, cognitive and non-cognitive skills. The same characteristics also predict the frequency of inconsistent reporting across four UKHLS waves. Analysis of the implications of reporting inconsistencies shows no impact of SAH measurement on the association between income and health. A set of dimensions of people’s physiological and biological health, captured using biomarkers, is associated equally with both SAH measures, suggesting that the interview mode does not play a role in the relationship between SAH and more objective health measures.