Assortative matching through signals

The matching of likes is a frequently observed phenomenon. However, for such assortative matching to arise in a search model, often implausibly strong conditions are required. This paper shows that, once signals are introduced, a search model can generate even perfect assortative matching under weak conditions: supermodularity of the match production function is a necessary and sufficient condition. It simultaneously drives sorting and functions as a single-crossing property ensuring that agents choose truthful signals. The information thereby transmitted allows agents to avoid all unnecessary costs of random search, which creates in effect an almost frictionless environment. Hence the unique separating equilibrium in the model achieves nearly unconstrained efficiency despite frictions. -- Assortative matching ; sorting ; search ; signals