On-the-Job Search with Hidden Attributes

This paper studies optimal search strategy and equilibrium contract design when one attribute of an offer is difficult to evaluate ex ante, but an individual eventually learns its true value and can conduct `on-the-job' search. Focusing on a type of agents who cannot infer the value of the hidden component (``benefits") from the headline attribute (``wage"), but nonetheless take future learning into account, I show that agents who naively overestimate the option value of searching on the job adopt a lower reservation wage initially. Endogenising the firms' offers leads to two main results. First, a symmetric equilibrium with inefficiently low benefits is easier to sustain than an equilibrium with socially desirable high benefits, even though the firms' profits suffer. Second, there may also exist asymmetric equilibria, in which low wages are necessarily paired with high benefits, and vice versa. The logic behind this result is very different from the theory of compensating differentials, however