Style-based investor attention, event sentiment and style comovement

In this project, we propose a style-based investor attention by calculating the average search volume index with the stocks sharing the same style characteristic. We test the cross-sectional return predictability of the style-based investor attention after controlling the effect of past style returns. We also test the impact of the style-based investor attention and the past style returns on the comovement for the stocks within a style portfolio. At the same time, we consider the stock characteristics that is used to construct the risk factors or that induce return anomalies in the literature as possible style characteristics to explore the possible style preferences of investors. Through these tests, we investigate the possible reasons for the style preferences shifts of investors other than their chasing the past style returns. We further examine whether news events and their event sentiments induce the style preferences shifts. The event sentiment is different from the news sentiment which is simply constructed by calculating the number of positive and negative words in an article. The event sentiment score is produced by categorizing news into different event types and analyzing the words and sentences around the specific event text. We test whether the occurrence of a particular event of a stock and the magnitude of event sentiment can induce the style preferences shifts toward the style of that stock. We then explore the possible causes of the style preferences shifts and their impact on the future returns of the style portfolios.