Dynamic Regulation with Stochastic Costs: Signal Dampening, Experimentation and the Ratchet Effect

Regulators and the firms they regulate interact repeatedly. Over the course of these interactions, the regulator collects data that contains information about the firm’s invariant private characteristics. This paper studies the case in which the regulator uses information gleaned from past cost observations when designing the current period’s contract (that is, contracts are short-term). When costs are stochastic, the regulator’s learning process is slowed compared to a deterministic setting. In the absence of additional messages, a separating contract no longer gives her complete information about the firm’s type as soon as the first period cost is observed; thus, the second period game is one of asymmetric information. Second period beliefs depend on the first period contract. In particular, the first period contract determines how much the regulator updates her prior beliefs for any given cost realization. By increasing the distance between first period cost targets (“experimenting”), the regulator increases expected second period welfare, and by decreasing this distance (“signal dampening”), she preserves the good firm’s expected second period rent, thus reducing the first period transfer. Given reasonable assumptions on the distribution of noise, the overall effect of the first period contract is to reduce the flow of information; thus, the good agent’s efforts are ratcheted up over time.