Problems in using birth-history analysis to estimate trends in fertility.

Summary The problem investigated is the adequacy of birth-history analysis as a method for estimating fertility change. The analysis demonstrates that inaccurate reporting of the dates of birth of live-born children can, under reasonable assumptions, significantly distort cohort fertility schedules in such a way that estimates of change in fertility will be biased in the direction of exaggerating declines in fertility. This kind of bias is shown to exist in fertility estimates obtained from survey data in El Salvador and Bangladesh. An important implication is that birth history questionnaires should begin with the most recent, rather than the earliest, event in a respondent's experience.