Extrapolative Investor Beliefs: A Study of Variations in the Stock Market and across Stocks

This study examines potential determinants of extrapolation bias in the stock market and across stocks. We obtain estimates of the degree of extrapolative weighting from the X-CAPM by Barberis et al. (2015), which is an extrapolative consumption-based asset pricing model with behavioral foundations. To estimate the degree of extrapolative weighting, we calibrate a proposition by the X-CAPM related to autocorrelation of the price-dividend ratio to the stock market data using the generalized method of moments. We find support for a salience-based explanation of variation in the degree of extrapolation bias in the stock market and across stocks. Time variation in the degree of extrapolative weighting is mostly related to financial indicators and sentiment, and less to macroeconomic explanations. Furthermore, we find that the ability of price-scaled variables to predict the equity premium is conditional on the degree of extrapolation bias. Cross-sectional dispersion in the degree of extrapolative weighting can be explained by the salience of stock’s return distribution. A stock’s sector is also indicative of the degree of extrapolation bias. Despite the cross-sectional variation, we do not find evidence that the degree of extrapolative weighting is a priced factor.