Investor expectations, earnings management, and asset prices

This paper examines the implications of investor expectations for the joint determination of earnings manipulation and asset prices. Three alternative models of investor expectations are studied: constant-gain learning, regime-shifting beliefs, and accounting-information-system (AIS) beliefs. I use the simulated method of moments (SMM) to estimate the most plausible model that matches the actual data. AIS beliefs and regime-shifting beliefs are shown to best explain the empirical moments of 63% and 32% of S&P 500 firms, respectively. Regression analysis suggests that the three models offer different predictions on the existence and magnitude of several empirical regularities including a positive earnings response coefficient, the discretionary accruals anomaly, and return momentum.