About

10-K filings are a rich source of information for investors and are expected to be objective and factual. As a midterm project in Data Science for Finance, I explored whether 10-K filings implicitly contain value-relevant information in the sentiment of the text.

I did this by examining the relationships between sentiments found in 10-K filings and returns following the 10-K's release using two valence-based dictionaries and three topic-based schemes (with financial stability, corporate governance, and e-commerce themes). I found higher absolute correlations for positive sentiments in general, with the largest magnitudes for positive e-commerce topics, positive LM sentiment, negative LM sentiment, and positive financial stability topics.

The Details

Please see the report for a detailed explanation of the data, methodology, and topic selection, along with the final analysis.