Learning to Signal in Markets

We formulate a dynamic learning-and-adjustment model of a market in which sellers choose signals that potentitally reveal their types. If the dynamic process selects a unique limiting outcome, then that outcome must be an undefeated equilibrium; though to be undefeated does not suffice to be the sole limiting outcome. If a Riley outcome exists that provides "high" type sellers with a higher utility than any other equilibrim outcome, then that outcome is the unique limiting outcome of our model. In the absence of a Riley outcome,. or if high type workers obtain higher utility in a pooling equlibrium than in the Riley outcome, a unique limit outcome will only emerge under very stringent conditions. If these conditions fail, the market will cycle between various equlibria and, possibly, nonequilibrrium outcomes.