Google search volume as a proxy of investor attention : are previous findings robust?

Abstract

I assess the robustness to different periods and panel models of several findings in the literature that uses Google search volume as an investor attention proxy. With all S&P 500 stocks between 2004 and 2016, I confirm that weekly search volume is persistent and increases are associated with high share turnover as well as earnings announcements. When the CCEMG estimator of Pesaran (2006) is used, I find evidence against Da et al. (2011) but consistent with Barber and Odean (2008). Even for large stocks, surges in people’s attention predict positive abnormal returns one week ahead, which reverse within one year. I conclude the literature should adopt panel estimators more robust to the presence of both firm and time effects.