We study how people search for partners and how beliefs about their own attractiveness shape whom they approach, using the detailed search log of a marriage matching platform. Requests to more popular partners are less likely to be accepted, but members at every popularity level send most of their requests to the most popular partners. This could reflect a strong preference for popular partners over remaining single, or optimistic beliefs about their chances of being accepted. The strength of the data is that search behavior over time is directly observed rather than inferred from final matches, so we can distinguish between these explanations. Members gradually target less popular partners, especially after unexpected rejections. A predictable rejection has little effect, and an unexpected acceptance raises their aim. This pattern is consistent with belief updating, because unexpected outcomes carry more information than predictable ones. We develop and estimate an equilibrium search model in which members are uncertain about their own attractiveness and learn from the responses they receive. The model recovers members’ beliefs, so we can separate their effect on search from that of preferences. We then use the model to quantify how optimistic beliefs at entry affect the timing and quality of the final match as well as the decision to quit the platform. We also evaluate policies that provide feedback on acceptance probabilities or limit monthly requests, allowing beliefs and acceptance rates to change in equilibrium.
Joint with Chihiro Inoue and Suguru Otani.