Estimating fertility trends from retrospective birth histories: Sensitivity to imputation of missing dates

Abstract In the birth history section of the Core Questionnaire adopted by the countries participating in the World Fertility Survey a question on the date of birth of each child born to the respondent is asked. When the woman cannot provide a date, as is often the case in many developing countries, she is asked how many 'years ago' the birth occurred. If the default is used, the month and year of the birth is imputed by a computer program: However, there can be two plausible interpretations of the 'years ago' response: as completed years, the demographer's usual definition of age, or as rounded years. In this paper, data from the WFS Survey in Bangladesh are used to determine the sensitivity of recent fertility estimates to the interpretation of 'years ago'. It is found that if the woman meant rounded years, but completed years were assumed for imputation, the resulting evidence of a recent decline could be either exaggerated or false. The results have implications for other surveys in which the 'years ago' response is not an explicit option.