UCLA

Marriage patterns can be well understood only if researchers employ measures of marriage rates that are appropriate for the question asked. In this paper, we provide evidence that the two classes of measures typically used in the literature, the number of new marriages per population and the share of individuals currently or ever married within an age range, generally lead to misleading inference when used to study the probability someone marries during his or her life or fertile life, how it evolves, and how it differs across populations. An alternative measure, the share of individuals ever married in a given cohort by a given age, is better suited for such studies. When researchers are interested in year-on-year changes in marriage probabilities of singles, age-specific marriage hazards are more reliable than population-based measures. We conclude by discussing implications of our findings for studies of the drivers and consequences of marriage formation.