A Model of Market Sentiment

Several influential studies show that returns tend to  underreact to news in the short-run while overreact in the  long-run. The behavioral finance literature links the  emergence of this underreaction and overreaction to  investors' cognitive biases in the processing of new  information. These contributions are instead silent on the  possible effects entailed by the wealth dynamics deriving  from the interaction of heterogeneous investors. In this  paper we propose a stylized agent based model of a  financial market aimed at linking heterogeneous investors'  wealth dynamics with assets returns underreaction and  overreaction.  We assume that there are two groups of  investors trading  long-lived assets and interpret their  dividend payments as news. We assume that dividends follow  a Markov process that generates a positive  autocorrelation. Investors believe that the underlying  process is iid and have heterogeneous beliefs concerning  assets' dividend payment process. Using both analytical  and numerical techniques we show how the dividend positive  autocorrelation together with investors long-run survival  is able to generate both short-run underreaction and  long-run overreaction. A "market sentiment" emerges:  despite single investors hold constantly rebalanced  portfolios, their relative wealth dynamics is such that  the aggregate investor react to news as behavioral models  suggest. Our model suggests two testable hypotheses of  which we offer some empirical evidence: dividend  autocorrelation is positively related to underreaction and  negatively related to overreaction, instead divergence of  opinions is positively correlated to both underreaction  and overreaction.