Investor Pessimism And The Market: Exploring Google Search QueriesVW Staff
Investor Pessimism And The German Stock Market: Exploring Google Search Queries
University of Tuebingen – Department of Statistics and Econometrics
University of Tuebingen – Faculty of Economics and Social Sciences
January 28, 2016
We analyze the effects of retail investor sentiment on the German stock market by introducing four distinct investor pessimism indices (IPIs) based on selected aggregate Google search queries of households. We assess the impact of weekly changes in sentiment captured by the IPIs on both contemporaneous and future DAX returns, volatility and trading volume. The indices are found to have individually varying but overall remarkably high explanatory power. If retail investor pessimism increases, contemporaneous market returns tend to decrease, accompanied by increases in volatility and trading volume. Moreover, future returns tend to increase while future volatility and trading volume decrease. However, the observed effects are only of transitory nature and vanish after at most two weeks. The outcome can be interpreted as correction effects. Overall, the results are well in line with modern investor sentiment theory.
Investor Pessimism And The German Stock Market: Exploring Google Search Queries – Introduction
“Welcome to the Age of Big Data” the New York Times cheered in 2012 (Lohr, 2012). Big Data, an abstract term describing vast and continuously increasing information flows has become prevalent in modern societies. Without doubt, now, in the second decade of the 2000s we are at a turning point where we steadily learn how to process and interpret this enormous volume of collected and collective information. The probably most successful internet search engine Google has a market share of about 97% in Germany (Haucap and Heimeshoff, 2014). Not only is Google capable of instantly providing all the information one requires but also, in turn, it gains tremendous insight from the collective search behavior. Aggregating billions of individual search queries might be the key to revealing and understanding linkages and dynamics that were technically inexplorable before. The first study that successfully employed Google search volume data is Ginsberg, Mohebbi, Patel, Brammer, Smolinski, and Brilliant (2009) who predict waves of influenza epidemics.
In Economics, the availability of Google’s search data paves the way towards an understanding of individuals’ behavior. Notable early contributions are Choi and Varian (2012) who find that it is possible to predict near term economic indicators with search queries as well as Da, Engelberg, and Gao (2011) who introduced a novel measure of investor attention based on Google search volume. By being able to directly monitor internet search behavior of households new possibilities arose with respect to assessing the effects of both investor attention and investor sentiment on financial markets compared to earlier attempts based on indirect market based proxies or observable media coverage.
This article contributes to the promising field of investor sentiment analysis and is inspired by the recent article of Da, Engelberg, and Gao (2015) who use Google search queries to construct a novel daily Financial and Economic Attitudes Revealed by Search (FEARS) index of investor sentiment which the authors then use to analyze S&P 500 returns, volatility and fund flows. We extend their methodology by proposing and comparing different variable selection methods such as the LASSO (cp. Tibshirani, 1996) and (sparse) principal component analysis with a slightly modified version of the linear regression based variable selection of Da et al. (2015). The four variants of a composite index of investor pessimism are then related to the German stock market. We shed light on the question whether there are sentiment-induced dynamics of stock market returns, volatility and liquidity.
Our results indicate that the four distinct indices have individually varying, but overall remarkably high explanatory power with respect to both present and future returns, volatility and trading volume. Two of the indices are found to be strongly related not only to present but also to one week ahead market returns. Two other indices appear to have a strong effect not only on contemporaneous but also on two week ahead trading volume. Moreover, all four indices appear to be strongly related to both contemporaneous volatility and one week ahead market volatility. Finally, the findings reveal the existence of correction effects and reversal patterns which is well in line with the literature dealing with the impact of investor sentiment.
The article is organized as follows. In Section 2 we provide a brief overview of the related literature on theory and empirical assessment of investor sentiment and its impacts. Subsequently, the data used in the analysis and the methodology for the construction of four distinct investor pessimism indices (IPIs) consisting of Google search queries are introduced in Section 3 . Section 4 relates the IPIs to key indicators of the German financial market, namely DAX index returns, volatility, and DAX trading volume. Section 5 summarizes the findings and concludes.
See full PDF below.