Mispricing Factors

Most factor models — Fama-French style — build factors from a single characteristic at a time: book-to-market for value, past return for momentum. Mispricing factors take a different approach: pool many anomaly signals (accruals, net stock issuance, profitability, and others) into one or two composite scores meant to capture "how mispriced is this stock" directly, then build long-short factors from those composite scores instead of from any single characteristic.

The logic is that if dozens of unrelated anomalies are all, at root, symptoms of the same underlying mispricing — stocks that are systematically overvalued tend to score badly on several signals at once, not just one — then combining signals into a composite should isolate that common mispricing component with less noise than any single anomaly measure, and should better absorb (explain away) the individual anomalies as separate factors.

Worked example. The Stambaugh-Yuan model builds two composites, MGMT (management-quality-related anomalies like accruals and asset growth) and PERF (performance-related anomalies like profitability and momentum), each averaging rankings across roughly six underlying signals. In tests, a long-short portfolio sorted on MGMT alone captures much of the return spread that used to require six separate anomaly factors to explain, suggesting those six anomalies share a common driver rather than being six independent effects.