Stock-induced Google trends and the predictability of sectoral stock returns

In this paper, we consider Google trends (G-trends) as a measure of investors' attention in the predictability of stock returns across eleven major US sectors. The theoretical motivation for our paper is clear. In seeking information to guide investment decisions, investors' sentiments are shaped by news such as G-trends that could induce changes in the prices of stocks. Thus, we construct a predictive model that incorporates G-trends series as a predictor of stock returns and thereafter we account for evident asymmetry in G-trends to analyze the signficance of positive- and negative-worded news in the predictability of stock returns. We also compare single- and multi-factor predictive models augmented with distinctive statistical effects against the baseline time series model. We highlight three key findings: (1) G-trends record consistent negative correlations with stock returns across sectors. (2) The proposed predictive model with G-trends outperforms the baseline (random walk) model. (3) The inclusion of asymmetry and macroeconomic variables improves the outperformance of G-trends over the baseline model.