Learning and the Price Dynamics of a Double-Auction Financial Market with Portfolio Traders

Abstract

In this paper we study the dynamics of price adjustments in a market where portfolio traders with bounded rationality and limited resources interact through a continuous, electronic open book. The market trading activity depends on the heterogeneity of agents' beliefs. We allow agents to hold arbitrary priors about the univariate marginal distribution of returns, while we assume that agents have a constant common view of the assets' association structure. We make agents update prior marginal distributions using past realized market prices. We show that asset price dynamics is strongly affected by the structure of the learning process. Under learning the price series show long run positive trends and become non-stationary. In particular, we underline the role of the assumed assets association structure in shaping asset price dynamics.