Content area
Full Text
1. Introduction
The traditional asset-pricing models are premised on the assumption that market prices reflect the available information (Fama, 1976). However, in the real world, investors have access to limited information, because the attention is a scarce cognitive activity (Kahneman, 1973). Investors’ attention is mostly related to the determination of stock prices and liquidity (Merton, 1987). Previous research in this strand of literature observes that the stock prices are guided by new information and hence follow a random path as the new information occurs randomly (Fama, 1965).
In the search for new information, researchers have increasingly focused on the impact of investor sentiment (Baker and Wurgler, 2006). Some researchers have relied on available data from news articles (Tetlock, 2007), Twitter (Bollen et al., 2011) and Wikipedia (Moat et al., 2013). However, it is difficult to measure the degree of investors’ attention. In their seminal paper Da et al. (2011) propose a novel and direct measure of investor attention using an aggregate search frequency in Google which we term as Google search volume index (GSVI). The recent literature in this domain provides support to the “price pressure hypothesis” or “attention theory” as it reveals that the Google search intensity provided by the Google Trends is positively related to stock returns and the trading volume (Bank et al., 2011; Joseph et al., 2011; Vlastakis and Markellos, 2012). In a recent study, Challet and Ahmed (2013) seek to test the claims that GSVI contains enough data to predict future financial index returns, and find that strategy based on financial keywords does not outperform the strategy based on completely unrelated keywords.
Three papers most related to this paper Da et al. (2011), Joseph et al. (2011) and Bijl et al. (2016) observe that a high GSV predicts high future returns for the first one to two weeks with a subsequent reversal. The first two papers use data for the period from 2004 to 2008 and the third one uses data covering 2008–2013. We use a more recent data for the period from 2012 to 2017 to investigate whether search query data on company names can be used to predict weekly stock returns for individual firms. This study complements the prior studies by investigating the...