Time-varying arbitrage and dynamic price discovery

We introduce time-varying measures of price discovery based on underlying profit maximizing behavior by combining the heterogeneous agent modelling literature with the market microstructure literature. We set up a heterogeneous agent model with arbitrageurs and trend chasers (chartists), and allow agents to switch between the strategies conditional on recent forecasting performance. Estimation of the model on Canadian-US cross-listed stocks on high-frequency data shows that there is significant heterogeneity and switching, causing ample variation in the information processing capacity of markets.