Browsing by Subject "Google"

Now showing 1 - 1 of 1

Investor beliefs and their impact on financial markets
(2021) Hartmann, Carolin; Burghof, Hans-Peter
The idea of this thesis is to use new data sources to approximate investor beliefs. It investigates whether the approximation improves the measurement of return and volatility in existing model frameworks. The findings are that differences in implied volatility, Google Search volume and Twitter Volume can be proxy variables for investor beliefs. They have an impact on financial market indicators and on the prediction of future market movements. Comparison of the trading behaviour of individual and institutional investors to predict market movements The first approach is to create a new sentiment index which compares the difference between retail investor behaviour at the Stuttgart Stock Exchange (SSE) and professional investors at the Frankfurt Stock Exchange (FSE). The measure is a comparison between the implied volatility measures for the DAX at the FSE (VDAX and VDAX-NEW) and a newly created implied volatility index (VSSE) for the SSE. The sentiment index is significant in predicting the daily returns on a size-based long-short portfolio over a four-year period. The analysis shows the persistent inconsistence between prices of structured products for retail investors on the SSE and option prices of professional investors on the FSE. The results provide empirical evidence that there are significant persistent behavioural differences between the two investor types which is reflected in persistent mispricing. Measurability of investor beliefs and their impact on financial markets The second approach is to measure individual investor beliefs with Google search volume (GSV) and Twitter volume (TV) to analyse their impact on financial markets. The basis is a daily panel of 29 Dow Jones Industrial average index (DJIA) stocks over a time period of 3.5 years in a panel data set-up. The impact on trading activity measured by turnover, is positive for GSV and TV on the same day and the next day which indicates their predictive power. The impact on realized volatility (RV), indicating the share of noise traders on the market, is only positive and significant for TV. It is significant on the same day and the next day. The impact of GSV is not significant. The results support the idea that GSV and TV capture the beliefs of individual investors. Although they suggest that the impact of TV on financial markets is more important than the impact of GSV. Predictive power of Google and Twitter The third approach is to use GSV and TV as a proxy for investor attention and investor sentiment, to assess their predictive power on the RV of the DJIA. The basis is a time-series set-up with a vector autoregression (VAR) model over a period of 2.5 years. The findings show that GSV and TV granger cause RV, controlling for macroeconomic and financial factors. Again, the effect of TV on RV is more important than the effect of GSV. In-sample, the linear prediction model with GSV and TV outperforms a standard AR (1) process. Out-of-sample the AR (1) process outperforms the standard model with GSV and TV. Clustering for high and low volatility groups, the analysis shows that the effect of GSV and TV on RV changes. Especially in times of high and low RV, GSV and TV seem to contain new information, as they improve the model fit compared to a standard AR (1) process. However, the results are not persistent in- and out-of-sample. This underlines that the results of GSV and TV are not generally persistent but depend on the selected criteria. Overall, the results of this thesis show that investor beliefs have an impact on financial markets. The measures, such as a sentiment index based on implied volatility, GSV and TV are proxy variables for investor beliefs. Future research should further improve the comprehension of investor beliefs to improve causality and economic significance in the long term.