Quantitative finance continues to debate the reliability and limits of model-driven investment strategies. One central question is how much weight investors should place on backtesting.
In The Factor Mirage: How Quant Models Go Wrong, Marcos López de Prado, PhD, and Vincent Zoonekynd, PhD, outline why investors should move beyond accepting historical performance at face value and focus on understanding why a model works. That is a valuable contribution to strengthening the rigor of quantitative investing — and one that invites further reflection on how that reasoning is structured.
It may help to frame the issue not as a binary choice between correlation and causation, but as a layered problem in which different forms of reasoning play distinct roles.
In practice, the choice is rarely between simple correlation and fully specified causality. Most investment research operates somewhere in between. Sometimes we can describe and test a mechanism directly. Sometimes we cannot. The system may move too quickly, key variables may be only partially observable, or the time and resources required to build a richer model may not be available.
In those settings, association-based reasoning still has value. That is not a defect of finance; it is a general feature of decision-making under uncertainty.
Association Under Constraint
Human beings often rely on associations when there is no time to construct a full causal account. That is not necessarily irrational; it can be adaptive. A fast association can guide action before slower, more elaborate reasoning is possible.
The same is true in investment practice. When relevant drivers cannot be directly observed or causal structure is only partly understood, associational signals may still contain useful information.
Association is not explanation. The question is not whether association has value, but whether it is sufficient. For institutional investors, this distinction has practical implications for due diligence, including how managers justify the inclusion and exclusion of variables in systematic models. When stronger structural knowledge exists, ignoring it is not sophistication; it is a loss of information. Association has a place, but it should not become a stopping point.
The call for greater causal discipline in finance is not new. The more interesting question is how to incorporate that discipline without oversimplifying the nature of markets themselves.
Epidemiology as a Model of Structured Reasoning
An epidemiologist would not analyze an epidemic as a purely statistical pattern detached from what is known about transmission. If susceptible individuals can become infected and infected individuals can recover or be removed, that knowledge becomes part of the model’s structure.
Compartmental models such as SIR (susceptible, infected, recovered) and SEIR (susceptible, exposed, infected, recovered) formalize those transitions. Statistical methods remain essential for estimating parameters and testing fit. But the analysis does not begin from a blank slate; it begins from established causal structure.
Finance can draw a similar lesson. Where durable mechanisms are reasonably well understood, they should be represented explicitly. If leverage amplifies forced selling, refinancing conditions shape default risk, inventories influence pricing power, passive flows affect demand, or network structures transmit distress, these are more than recurring correlations. They are mechanisms that can be modeled, tested, and challenged.
Dynamic models can be especially useful here. A regression captures co-movement; a dynamic model represents stocks, flows, delays, and feedback. In finance, that may mean balance-sheet capacity, funding conditions, capital flows, or adoption dynamics. Such models help clarify how the state of the system evolves and how today’s conditions shape tomorrow’s outcomes.
Reflexivity and Adaptive Markets
Finance differs from epidemiology.
Markets are reflexive. Beliefs influence prices, and prices in turn reshape beliefs, incentives, and financing conditions. A narrative can attract capital; capital flows can move prices; rising prices can reinforce the original narrative. What appears to be a durable relationship may, for a time, reflect a self-reinforcing loop.
Causal reasoning remains essential, but the relevant structure may itself include feedback between beliefs, flows, and outcomes.
A Three-Layered Framework
Investment research can operate on three distinct but related layers:
Association: What appears to predict, even imperfectly?
Causal: What mechanism could plausibly generate that relationship?
Reflexive: How might the use of the signal itself alter behavior, crowd the trade, change flows, or reshape the environment being modeled?
Seen this way, the debate is not about choosing correlation over causation. It is about knowing when association is sufficient, when mechanisms must be modeled explicitly, and when reflexive feedback makes the system more adaptive than either approach assumes.
Few serious quantitative researchers would defend correlation without scrutiny. Robust practice already includes stress testing, economic intuition, and structural reasoning. The question is not whether causality matters, but whether we are explicit about which layer is doing the work — and how those layers interact.
Toward a More Disciplined Quantitative Practice
We should use causal knowledge when it is available and test causal hypotheses when we have them. When a phenomenon involves accumulation, delay, or feedback, dynamic models may be more appropriate than static statistical fits.
Association-based thinking retains an important role, especially under constraints of time and observability. But where established structure exists, ignoring it is not sophistication; it is a loss of information.
The opportunity for quantitative finance is not to replace one methodological slogan with another. It is to become more disciplined and more transparent about how different forms of reasoning contribute to robust investment research — when patterns are enough, when mechanisms are required, and when reflexivity demands that we treat markets as adaptive systems shaped in part by our own participation.
The future of investment research is therefore unlikely to be purely correlational or narrowly causal. It will be more plural, more dynamic, and more explicit about the difference between patterns that merely appear stable and mechanisms capable of sustaining them.
References
López de Prado, Marcos, and Vincent Zoonekynd. The Factor Mirage: How Quant Models Go Wrong. Enterprising Investor, CFA Institute, 30 October 2025.
Delli Gatti D, Gusella F, Ricchiuti G. Endogenous vs exogenous fluctuations: unveiling the impact of heterogeneous expectations. Macroeconomic Dynamics. 2025;29:e125. doi:10.1017/S1365100525100345
Gigerenzer, Gerd, and Daniel G. Goldstein. “Reasoning the Fast and Frugal Way: Models of Bounded Rationality.” Psychological Review 103, no. 4 (1996): 650–669.
Kermack, W. O., and A. G. McKendrick. “A Contribution to the Mathematical Theory of Epidemics.” Proceedings of the Royal Society of London. Series A 115, no. 772 (1927): 700–721.
Greenwood, Robin, Samuel G. Hanson, and Lawrence Jin. “Reflexivity in Credit Markets.” NBER Working Paper No. 25747, April 2019.
