Not logged in.
Quick Search - Contribution
Contribution Details
Type | Dissertation |
Scope | Discipline-based scholarship |
Title | Measuring Media Sentiment – Essays on Its Impact on the Economy and the Financial Markets |
Organization Unit | |
Authors |
|
Supervisors |
|
Language |
|
Institution | ETH Zürich |
Faculty | Department of Management, Technology and Economics |
Number of Pages | 114 |
Date | 2011 |
Abstract Text | This thesis considers sentiment in the print, TV and financial markets news media in order to evaluate if and how sentiment can explain and predict consumer and investor behavior. In the empirical section, we introduce novel and unique datasets in order to show with various models the explanatory and predictive power of media sentiment. The first chapter introduces the subject matter, sets out the methodological framework, and introduces the datasets. The second chapter tests the models of Carroll et al (1994) and Sommer (2007) by adding news sentiment to their consumption behavior equation. It is shown that Autoregressive Moving Average (ARMA) models are suited for modeling private consumption with macroeconomic and sentiment variables. The introduced news sentiment index was created from over 100,000 newspaper articles, and it is shown how such an index can be created. The third chapter takes the nowcasting approach from Kholodilin et al (2010) and Schmidt and Vosen (2011) by showing that TV sentiment is a valid variable to nowcast private consumption. The TV sentiment index was created from over 10,000 TV news shows. In the last chapter, we examine the explanatory power of sentiment from Reuters news on stock returns with Vector Autoregression (VAR) models, accounting for the findings of Tetlock (2007), Brown and Cliff (2005) as well as Menkhoff and Rebitzky (2008). In total, we consider over 3.6 million Reuters news pieces. The main results are the following: first, news sentiment is a valid variable to add to private consumption behavior models, and news sentiment has similar explanatory power than personal income. Nevertheless, consumer sentiment is the more powerful variable compared to news sentiment. Second, TV sentiment is more suited to nowcast private consumption than consumer sentiment. And third, we show that both positive and negative Reuters sentiment matter for explaining stock returns. The sentiment effect is measurable and present over several months, and not only days, as previously assumed in the literature. |
Free access at | Official URL |
Digital Object Identifier | 10.3929/ethz-a-006838119 |
Other Identification Number | merlin-id:12473 |
PDF File | Download from ZORA |
Export |
BibTeX
EP3 XML (ZORA) |