Predicting Market Regimes and Stock Returns Using Investor Sentiment

Abstract

We study how investor sentiment affects market regimes and examine the predictive patterns of stock returns under different firm characteristics, investor sentiment, and the state of the regime. Two kinds of investor sentiment indices, Baker and Wurgler's (2006) sentiment index and CBOE's VIX index, and a new econometric method, the Qualitative Vector Autoregressive (QVAR) model proposed by Dueker (2005), have been used in this study. The QVAR model incorporates the information from qualitative binary variables, such as market regimes, into the VAR structure to facilitate exploring predictive patterns in a conditional way. Our empirical findings support those of Baker and Wurgler (2006) and Kumar and Lee (2006) that investor sentiment is important in the formation of stock returns, especially for sentiment-sensitive stocks. We find that (1) stock returns are in the presence of regime-shift levels, volatility, and the predictive power of the instrumental variables; (2) investor sentiment (or fear) has the ability to predict the subsequent market regimes as well as the cross-sectional stock returns over regimes, and (3) the appearance of size and book-to-market effects depends on the states of the regime and investor sentiment.