What are the most pressing systematic weaknesses in financial research, and how should the investment community address them? In the May 2015 version of his article entitled “The Future of Empirical Finance”, Marcos Lopez de Prado identifies three major problems in empirical finance and proposes ways to mitigate them. Based on his experience and common sense arguments and references to some research, he concludes that:
- Advances in data collection/processing have made multiple hypothesis testing easy, and the associated data snooping bias ubiquitous. Researchers generally ignore this bias, even though corrective adjustments are known (see, for example, “Navigating the Data Snooping Icebergs” and “Taming the Factor Zoo?”). The Econometric Society should denounce uncorrected multiple hypothesis testing, and leading financial journals should reject papers that obviously present it.
- Financial markets adapt to (published) anomalies, thereby dampening or extinguishing them in future data (see “Effects of Market Adaptation”). Adaptive behavior in complex social systems is not amenable to the scientific method. Empirical research in finance should focus not on discovery of anomalies but rather on quantifying parameters of well-reasoned and strongly formalized mathematical theories of anomalies.
- Academics, unlike practitioners, do not perform live testing. True discoveries (and true refutations) in empirical finance are therefore more likely to come from research-driven investment firms than universities. Academic journals should engage and collaborate with industry.
In summary, the investment research community should demand methodological rigor to avoid “becoming a pathological science, a collection of ‘cold fusion’ claims.”
Cautions regarding conclusions include:
- Material snooping bias inherited from prior research may not be obvious, or avoidable.
- Practitioners have competitive incentives to keep discoveries private.
See “Chapter 3: Avoiding or Mitigating Snooping Bias” and “Chapter 5: Checking for Market Adaptation”.
