eThesis - Kauppakorkeakoulun sähköiset gradut

OBJECTIVE OF THE STUDY:

In this thesis I analyze do annual statement releases of Finnish companies' include information about the value of a company's equity that the computational text analysis is able to extract. Furthermore, I investigate is this information immediately incorporated into the stock market prices or do markets under- or overreact to the information content of annual statement releases. The secondary focus is to analyze the readability of the annual statement releases' impact post-announcement period's market reaction.

DATA AND METHODOLOGY:

The study consists of 725 annual statement release from OMX NASDAQ Helsinki, which take place from 2007 to 2014. I use the dictionary approach and Loughran and McDonald -dictionaries to calculate positive and negative words. To determine the sentiment of the releases, I use proportional and "term-frequency - inverse document frequency" - word weights. To measure readability, I use the amount of words as a proxy. Finally, I calculate cumulative abnormal returns based on the market model and use the event study methodology to investigate sentiment's and readability's impact on cumulative abnormal returns.

FINDINGS OF THE STUDY:

At market-level, I find that the yearly annual statement releases' average sentiment correlates moderately with the changes in the market index. At firm-level, I find that increase of positive words are linked to the event day's positive shift in the abnormal returns and to the positive drift in the next five days. Furthermore, I find that negative words and "term-frequency - inverse document frequency" word weights are poor predictors of cumulative abnormal returns. Finally, I find no support for the hypothesis that the releases' length proxies readability and impact on post-event returns.