Estimating the co-variates of age at marriage and first birth

Abstract The purpose of this paper is to explore how best to estimate the co-variates of age at marriage or age at first birth. Some investigators have approached this problem by estimating separate schedules for different classifications of the population. However, such attempts cannot have a truly multivariate flavour since cell sizes become small as ambitious cross-classification is attempted. We examine two competing models, one an extension of that proposed by Coale and McNeil and the other, proportional hazards, arising from the analysis of failure time data. Their performance is contrasted on WFS data for Colombia. The conclusion we reach is that both methods perform well; which is chosen depends upon the purpose of the analysis and the type of data available. Specifically, the proportional hazards method cannot be applied to data on ever-married women only and cannot be used to project the experience of incomplete cohorts. However, Coale and McNeil's extended model is harder to estimate since software is not so readily available.