Bayesian Analysis of Event History Models with Unobserved Heterogeneity via Markov Chain Monte Carlo

Abstract

This article describes an interesting application of Markov chain Monte Carlo (MCMC). MCMC is used to assess competing explanations of marital fertility decline. Data collected during the World Fertility Study in Iran are analyzed using methods developed to perform discrete time event history analyses in which unobserved heterogeneity is explicitly accounted for. The usual age-period-cohort identifiability problem is compounded by the presence of a fourth clock, duration since previous birth, and a fifth clocklike variable, mother's parity. The authors resolve this problem by modeling some of the clocks parametrically using codings suggested by alternating conditional expectation (ACE) and Bayes factors to decide which clocks are necessary. Compound Laplace-Metropolis estimates are used to compute Bayes factors for comparing alternative models. The new methods enable the authors to conclude that Iran's fertility decline was primarily a period effect and not a cohort effect, that it started before the Family Planning Program was initiated, that it was the same for women at all educational levels but varied depending on husband's education, and that it was greatest in the largest cities, particularly Tehran.