Semantic Analysis of MDA Section in 10K Reports

The study applies natural language processing tools to analyze discretionary disclosures in item 7 in 10K filings. Sentiment scores, longer filings and use of low-information-content words in management discussion and analysis (MD&A) section in annual reports correlate with worse performance, whereas higher numerical content suggest higher profitability and lower probability of financial distress. Over a 20-year period starting in 2003, messages to shareholders have gradually become more optimistic; this change parallels drop in numerical content and increase in use of parts of speech with low information content. Taken together, these findings provide evidence of impression management by publicly traded firms.