Investor Sentiment – Implications from A Repeated Coordination Game Study

We study equilibrium selection in A. Gerber, T. Hens and B. Vogt’s experiment (in Rational Investor Sentiment in a Repeated Stochastic Game with Imperfect Monitoring, 2010), through investigation of subjects’ coordination behaviour and switching behaviour respectively. Quantal choice models are utilized to capture the idiosyncratic component of deviations from the best-response strategy. Our main conclusion is that subjects learn to coordinate through experience of repeated interactions, while their updating beliefs are biased by endogenous expectations systematically. Reference-dependent loss aversion is evidenced. When experiencing gains, i.e. with profit level exceeds endogenous expectation, subjects tend play passively and only follow last period’s realisation of the game. While under the pain of losses, i.e. with lower than expectation profit level, they become more aggressive as the discrepancy between expectation and profit level enlarges; and switch their action even if the game realisation did not switch in last period. Implications for investor sentiment in financial markets are proposed