Predictive Disequilibria and the Short Run Dynamics of Asset Prices

Abstract

The received theory of efficient capital markets is based on an equilibrium to a predictive game. As such, it does not cope with the observed volatility of many asset markets, or with the evident existence of contingent or actional rules such as chartism. Choosing a simple martingale (random walk) for the fundamental equilibrium model, we show that the coexistence of a commonly used chartist rule will imply an evolutionary mixture of rules, all of which represent attempts at unbiased forecasts, but which are unavoidably myopic in their concentration on the structure as it exists at any one point in time. Rule mixtures do not as such create patterns; these are instead generated as disequilibrium traverses by the continual revision of rules. Once a pattern becomes evident, ad-hoc rules such as adaptive or extrapolative expectations will have local success, leading to overshooting behaviour that ex-post can validate the original chartist recommendation and may also induce speculative bubbles. Empirically, such local patterns contradict the fractal character of the random walk.