Learning and Strategic Delay in a Dynamic Coordination Game

Abstract

Heterogeneously informed agents decide their optimal action timings while observing past activities over time. We construct such a dynamic global coordination game to investigate the impact of learning and delay options on coordination behaviors and outcomes. A unique monotone equilibrium is characterized, which is analytically convenient for all ranges of learning efficiencies, and we demonstrate that learning improves coordination success, while the delay options alone have no impact, relative to the one-shot game. Dynamics of agents' behaviors and welfare implications are then presented. In addition, we show that full learning about the state achieves in the limit, and find the condition on which observing actions reveals more accurate information about the state than directly observing it.