"The Evolution of Coordination under Inertia" Tom Norman Nuffield College, University of Oxford Abstract: This paper models the phenomenon of inertia driven by individual strategy switching costs in a stochastic evolutionary context. Kandori, Mailath, and Rob's (1993) model of a finite population of agents repeatedly playing a 2x2 symmetric coordination game is extended to allow for such inertia. Taking noise to the limit, a number of new short- to medium-run equilibria emerge, centred around the mixed-strategy equilibrium. Thus, unusually, an evolutionary model is seen to provide some justification for the controversial concept of mixed-strategy equilibrium. However, Kandori, Mailath, and Rob's long-run selection of the risk-dominant equilibrium continues to hold, both under fixed-rate mutations and under state-dependent mutations driven by stochastic switching costs. The key to this is the satisfaction of Blume's (1999) "skew-symmetry" of the noise process, which is shown to be crucial even under simultaneous strategy revisions. In fact, the presence of the new short-run equilibria can under certain conditions serve to reduce the expected waiting time before the risk-dominant equilibrium is reached - an instance of Ellison's (2000) idea that evolution is more rapid when it can proceed via a series of small "steps" between extremes. This suggests inertia to be a surprisingly efficient phenomenon, and also serves to moderate the force of the Ellison (1993) critique of excessively long transition times in models with vanishing noise.