State space methods in asset pricing

AbstractThe Arbitrage Pricing Theory currently popular in the finance literature is based on the assumption that risky asset returns are generated by an approximate factor model in which the factors are static state variables. Since the factors are often associated with macroeconomic phenomena, in this paper they are modeled as dynamic processes using the system-theoretic time series approach recently developed by Masanao Aoki. The forecasts from the model are evaluated using a nonparametric test due to Henriksson and Merton. The evidence suggests that the number of states required to explain asset pricing is large relative to the number of assets, and that the states vary across assets, in contradiction of previous evidence from static models