Presented in English by Emmanuel Olamijuwon, discussed by Valérie Golaz and moderated by Valentine Becquet.
Accurate reporting of the timing of marriage is crucial for interventions that address the consequences of marital status for the health outcomes of women and families. Yet the reporting of the onset of marriage may be shaped by structural factors, including those that stigmatise unmarried women and single mothers, as well as by differences in recall, calendar familiarity, and data quality. Using data from 303,244 couples sampled in the Demographic and Health Surveys (DHS) conducted across 55 countries, I investigate the magnitude of inconsistent reporting of the year of first marriage by comparing the years independently reported by women and their husbands. I also examine how such inconsistency intersects with socio-demographically relevant characteristics. Beyond documenting these patterns, I develop and validate a machine-learning framework that harmonises discordant spousal reports into a single, more reliable estimate of the year of first marriage.
