Anomaly Evaluation

What is a financial market anomaly? How can investors determine whether an apparent anomaly is real (economically material)? In his March 2011 book chapter entitled “Perspectives on Capital Market Anomalies”, Mozaffar Khan provides a framework for interpreting academic research on anomalies and evaluating the exploitability of specific anomalies. His context is market efficiency: “Respect for the efficient markets theory, and an acknowledgment that it sometimes fails (i.e., that mispriced stocks can be identified), can coexist.” Key points are:

• Anomalies are deviations from efficient markets theory, manifested as predictable non-zero risk-adjusted returns.
• The typical approach for evaluating anomaly alpha is to: (1) rank stocks into (for example) deciles based on some historical firm or stock characteristic; (2) calculate the future return for a portfolio that is long the extreme outperforming decile and short the extreme underperforming decile; and, (3) debit this raw future by the expected return based on the risk of the portfolio (as specified by a risk adjustment model such as the Fama-French three-factor model).
• While many academic studies do not, practitioners must debit information, search and trading costs from the alpha to assess anomaly exploitability.
• The statistical reliability of an anomaly measures the level of assurance that its alpha differs from zero, with at least 95% assurance a typical criterion. Generally, if the variability of alpha over time is small compared to its average value, the associated anomaly is reliable. Uncertainties that confound determination of alpha reliability include:
  • The risk adjustment model may be defective, thereby either exposing false alphas or obscuring real ones.
  • The statistical tests used to assess alpha reliability may be defective (as evidenced by conflicts with alternative tests), perhaps because they have some built-in bias or they are unsuitable for the anomaly return distribution.
  • A discovered alpha may be a lucky result of intensive in-sample data snooping, either directly by the immediate researcher or indirectly by follow-up researchers using the same data. Luck is unlikely to persist out of sample. Tying the result conceptually to an otherwise testable economic rationale helps mitigate data snooping bias.
• Major competing explanatory frameworks for anomalies (often difficult to distinguish empirically) are:
  • Investors lacking complete information about asset valuation parameters and/or being otherwise uncertain about them may rationally misprice associated assets.
  • Inherent investor behavioral and cognitive biases (such as emotional sentiment, overconfidence, self-attribution, conservatism and representativeness) may generate mispricings via underreactions and overreactions to information. Obstacles to market efficiency such as transaction costs, short sale constraints, absence of close substitutes and non-scalability may allow such mispricings to persist.

In summary, the typical process for evaluating the exploitability of financial market anomalies invokes a range of uncertainties that undermine investor confidence in associated implementation strategies.

The author does not address in this book chapter:

• Potential asset return distribution wildness (susceptibility to Black Swans) and consequent statistical intractability.
• The extreme difficulty of addressing trading frictions, which vary considerably over time, across assets and among investors.
• Potential market adaptation to publicly known anomalies over time.

Compare and contrast this overview with that presented in “Investing Demons”.

You May Also Enjoy...

• Beat the Market with Hot-Anomaly Switching?
• The Earnings Yield Anomaly Revisited
• Evolution of the Accruals Anomaly (to Extinction?)
• A Tradable Accruals Anomaly
• Robert McHugh Objects to Our Evaluation of His Commentaries