Can Google Search Volumes Predict Stock Market Movements? : In Search of a Better Measure of Investor Attention for the US Stock Market
Tynys, Ville (2019-06-14)
Can Google Search Volumes Predict Stock Market Movements? : In Search of a Better Measure of Investor Attention for the US Stock Market
Tynys, Ville
(14.06.2019)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
suljettu
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2019061921353
https://urn.fi/URN:NBN:fi-fe2019061921353
Tiivistelmä
Google’s search volume data has many attributes that could make it a powerful tool for studying social behavior and stock market anomalies. Multiple studies have found evidence that search volumes for the stock ticker symbols can be used as a direct measure of investor attention. However, the tickers have also disadvantages. First, the tickers could collect a lot of noise as the three to five letter combinations of the stock tickers can be easily be connected to other unrelated topics. Second, it is unlikely that a non-professional investor is familiar with the ticker of the company, especially at the beginning of the information gathering.
Therefore, this study sets out to search for a better proxy for investor attention that could be more intuitively reasoned with. The possibilities of semantic ambiguities should also be minimized. The proposed alternative measure is the combination of the commonly used names of a company and the word “stock” (CN+S). Implicitly the study also tests the attention-driven buying hypothesis presented by Barber and Odean (2007). The study focuses on the companies of Standard & Poor’s 100 index from March 2013 to December 2017. The financial data was gathered via DataStream and the search volume data was downloaded via Google Trends webpage by using a web-scraping program. The effectiveness of the tickers and the CN+S was tested by their ability to predict stock market trading volumes and abnormal returns. The statistical methods used were fixed effects and random effects regressions of which the preferred method was chosen by using the Hausman test (Hausman 1978).
The results indicate that both search terms can be used to predict the stock market trading volumes, but the tickers are four times more effective than the CN+S. However, neither of the search terms predict the trading volumes very well. In addition, neither of the search terms could be used to predict abnormal returns. Therefore, this study finds no support for the attention-driven net buying hypothesis in the Standards & Poor’s 100 index. The most likely explanation is that this study was focused only on companies that are mainly large and well-established.
Therefore, this study sets out to search for a better proxy for investor attention that could be more intuitively reasoned with. The possibilities of semantic ambiguities should also be minimized. The proposed alternative measure is the combination of the commonly used names of a company and the word “stock” (CN+S). Implicitly the study also tests the attention-driven buying hypothesis presented by Barber and Odean (2007). The study focuses on the companies of Standard & Poor’s 100 index from March 2013 to December 2017. The financial data was gathered via DataStream and the search volume data was downloaded via Google Trends webpage by using a web-scraping program. The effectiveness of the tickers and the CN+S was tested by their ability to predict stock market trading volumes and abnormal returns. The statistical methods used were fixed effects and random effects regressions of which the preferred method was chosen by using the Hausman test (Hausman 1978).
The results indicate that both search terms can be used to predict the stock market trading volumes, but the tickers are four times more effective than the CN+S. However, neither of the search terms predict the trading volumes very well. In addition, neither of the search terms could be used to predict abnormal returns. Therefore, this study finds no support for the attention-driven net buying hypothesis in the Standards & Poor’s 100 index. The most likely explanation is that this study was focused only on companies that are mainly large and well-established.