Negative words in 10-K’s and filing period stock returns
A a replication study using sentiment analysis
Abstract
Previous research has developed a word list specifically designed to classify the negative tone of financial texts. Research has shown evidence that classifying 10-K documents by frequency of negative words can predict returns after the filing date. Using this negative word list, I analyze if the return predictability patterns still emerge. In order to control for implementation differences, I collect two samples of 10-K documents with daily stock data: from 1994 to 2008 as the control sample and a new sample from 2008 to 2018. Negative tone analysis with the frequency based method is not sufficient enough, to consistently predict stock returns. Results from previous research using this method can only be replicated in their original time frame from 1994 to 2008.
Bachelor's thesis paper
Original study
When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10‐Ks
Data profiling
- Profiling of un-filtered dataset 2008-2018
- Profiling of un-filtered dataset 1994-2008
- Sample Filtering Summary
- Sample profiling 2008-2018
- Sample profiling 1994-2008
- Regression results 1994-2008
- Regression results 2008-2018
