Market Disequilibrium and the Impact of News Sentiment

Abstract

The market price of any stock or financial instrument listed on an exchange would be assumed to be observed price, though we believe there is always a true price (or fair market value) of any. The market value and market price are identical only under conditions of market efficiency, equilibrium, and rational expectations. According to the theory of efficient market, if the observed price of a stock is beyond the market value, the price is expected to drop and vice versa. This is consistent with one of the properties of stock returns - mean reversion. However, the market doesn?t always reverse. Although the long-term mean of returns tends towards zero, if we look at the small proportion of the entire market movements, some show extreme up/downwards movements. We believe sentiment takes place if the market is not reversing. In this thesis, I have used a systemic approach with two parts of the basic regression models to explore market behaviour: first, I have applied parametric models to check whether investor sentiment exists at both market- and firm-level. Second, we focused on whether we can find a non-parametric method to visualise the relationship between return/residual and sentiment. There are mainly five case studies in this thesis. The first case study confirmed the relationship between the proxy of sentiment (extracted from the textual corpus) and DJIA market returns using linear regression models. The second and the third case studies used the same method to observe the relationship between sentiment and index returns in Danish and Chinese markets. The fourth study visualised the relationship between the proxy of sentiment and the return residual (error term from the linear models) using a LOWESS model. In the fifth case study, we test all the procedures using the firm-level stock returns (23 top companies in Fortune 500) into the same model instead of the market-level returns, revealing the impact of investor sentiment on the firm-level stock returns. Results have shown that although impact of sentiment is always observed at market-level (because the coverage of market-level sentiment is better as any news of the companies included in the index can be regarded to the market news), it is not always obvious at the firm-level (with difficulties in availability of data and limitation of method).