An Analysis of Speculative Language in SEC 10-K Filings

This study applies sentiment analysis techniques to model the usage of speculation within a collection of financial documents. The model is trained on the MPQA corpus to extract features that correlate with speculative sentences and applied to a collection of SEC 10-K documents from a five year period. The documents with the highest amount of speculation contained a different concentration of terms compared to the entire collection, and the sentences mostly consisted of explaining potential risks concerning projects, taxes, and pensions.