Dependence matters! Investor sentiment and stock returns: A sliced inverse regression approach

We construct a new investor sentiment index by exploiting the information of the sentiment proxies with the sliced inverse regression approach. We show that the new index is a strong negative predictor of future aggregate stock returns in both in-sample and out-of-sample tests. Further evidence indicates that the new sentiment index generates large utility gains for a mean-variance investor who optimally allocates between equities and risk-free assets. In addition, we show that this sentiment index exhibits the strongest return predictability for portfolios sorted on size, value, momentum, and industry.