Discrete Actions in Information-Constrained Tracking Problems

Optimal actions of an agent facing a Shannon capacity constraint on the translation of an uncertain signal into an action can easily turn out to be discretely distributed, even when the objective function and the initial distribution of uncertainty contain no discrete elements. We show this result analytically in a broad class of cases. It has implications for the interpretation of observed intervals between changes in prices or other economic choice variables in micro-data as indicators of costs of adjustment or of the degree of “stickiness” in responses to aggregate policy changes or business cycle fluctuations.