“Experimentation, Full Revelation, and Inspiration and Collapse” by Giulio SECCIA
University of Southampton
Michael T. RAUH
University of Liverpool
In this paper we develop a simple two-period experimentation model where the noise component of the signals has bounded support and need not satisfy the monotone likelihood ratio property (MLRP). The motivations for these departures from standard assumptions are that we obtain new and very different results and that the standard assumptions are implausible in many realistic situations. The main result of the paper is that what the MLRP is violated, small changes in either the agent's characteristics or the volatility of noise can induce discontinuous changes in the optimal first period action. We use this result to explain qualitative differences in effort across workers in terms of small differences in ability or the volatility of their stochastic environments.