Trading the Similarity of Positive Language in Company Filings

Summary

This document describes a monthly equity strategy based on how similar the positive language in companies’ latest 10-K or 10-Q filings is to prior language. It uses a vendor’s cosine-similarity measure, ranks covered stocks into deciles, buys the lowest-similarity group, and shorts the highest. The stated universe is roughly the largest 1,000 US stocks with available price histories, favoring more liquid names. The document contrasts this approach with earlier research on overall filing similarity, noting differences in language type, holding period, and universe.

The cited study reports that low positive-similarity stocks outperform high-similarity stocks by about 5% annually, and gives a Sharpe ratio of 0.84 for the long-short implementation. It also says the effect is not explained by common pricing models or changes in filing sentiment, while the proposed management-incentive explanation remains uncertain. These are reported research findings, not a guarantee of future results; the page gives limited detail on costs, data construction, and implementation sensitivity.

Key ideas

• The signal measures cosine similarity in positive language in each company’s most recent filing.
• Stocks are ranked monthly, with a long position in the lowest-similarity decile and a short position in the highest.
• The described universe focuses on large US stocks covered by the data provider.
• The source study reports annual outperformance of about 5% for low-similarity versus high-similarity stocks.
• The proposed explanation is uncertain, and the reported anomaly may still be sensitive to implementation choices.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.