Learning and assortative matching with private histories

The paper models an economy in which output is produced by teams of size two, match partners’ qualities are complements, and agents’ histories are private information. The initial situation is a random assignment caused by a technological change that alters agents’ productive skills. Agents must relearn their types by observing the output that they and their match partners produce. The economy adjusts toward its long-run positive assignment at a geometric rate when separating output-contingent contracts are available. When contracts fail and output is split equally, adverse selection slows adjustment to an arithmetic rate and, for some parameter values, stops it altogether. Along the transition path, the middle output state shrinks and mass moves to the extremes. Thus the speed of convergence to assortative matching depends on whether firms can commit to contracts that screen workers by their private histories.