As good as married: a model of long-term cohabitation, learning and marriage

This paper develops a two-sided search-matching model with imperfectly observed types and learning. Since agents do not observe one another’s type accurately, they first engage in a probationary partnership to learn one another’s true type. Using the metaphor ofpremarital cohabitation and marriage, we demonstrate that long-term cohabiting individuals eventually learn each other’s true type. We also demonstrate that singles ofeither sex are partitioned into classes and are matched in the same class in equilibrium. We show that sequential learning reduces signalling errors so that the Bayes estimator of the true type converges almost surely to true type. As noisy information is filtered over time, the mismatch risk disappears and the aggregate matching pattern based on true types is restored.