This paper aims to develop a model of trading in the stock market that can shed light on the sources of several widely reported empirical features of stock markets, including occasional predictability of excess returns using public information, 'excess volatility', and predictability of trading volume and volatility. Traders, drawing on both public and private information, make use of recursive modelling techniques to select and update their statistical model used to predict future excess returns of the various stocks traded in the market. This leads to continued disparities in traders' beliefs about future excess returns even in the absence of unobserved variations in the net supply of shares. It is shown that aggregation across traders as well as stocks then implies that an econometrician having access to public information only may find the presence of features such as 'excess volatility' and predictability of trading volume and volatility in market excess return/price data. This is even though the scope for individual traders to successfully predict excess returns on individual stocks is very limited. Simulation analysis is used to assess the model's effectiveness in explaining U.S. stock market data quantitatively.
