Choice under Markovian Constraints

In this paper I provide a descriptive model of choice over time by a population of constrained maximizing agents. Agents’ choice sets are markovian in the sense that they depend on previous choices. The unperturbed dynamics turns out to be trapped into local maxima whatever the length of memory. In the presence of perturbations efficiency is got with a memory of at least two periods. This provides a useful insight for what drives to efficient evolution in this setting: perturbations create variety and a two period long memory allows comparisons and selection