Is more memory in evolutionary selection (de)stabilizing?

Abstract

We investigate the effects of memory on the stability of evolutionary selection dynamics based on a multi-nomial logit model in an asset pricing model with heterogeneous beliefs. Whether memory is stabilizing or destabilizing depends in general on three key factors: (1) whether or not the weights on past observations are normalized; (2) the ecology of forecasting rules, in particular the average strength of trend extrapolation and the spread in biased forecasts, and (3) whether or not costs for information gathering of economic fundamentals have to be incurred.