Improvements in forecasting insurance stock excess returns: Comparing the investor sentiment endurance index with the CAPM and Fama-French models

Abstract

ABSTRACT This study applies the sentiment endurance (SE) index developed by He (2012) to forecast excess returns of insurance stocks. With the exception of the 12-month rolling forecasts of the Fama-French three-factor model (FF), forecasts of the SE model persistently outperform that of the CAPM and FF models in terms of lower absolute percent forecasting error (APFE) and significantly lower standard deviation of APFE. The accuracy of 6-month rolling forecasts of SE model is significantly higher than that of the FF model. Further, this study finds that the inclusion of SMB and HML in the SE model significantly deteriorates the accuracy and stability of forecasts. To a lesser degree, the addition of the market risk factor to the SE model hurts more than it improves the quality of forecasts. The results clearly suggest that compared to global variables, the SE index, as a local variable, more accurately reflects insurance investor sentiment and response to news and therefore can better forecast excess returns of insurance stocks.