Factor investing has emerged as a methodology designed to clarify the reasons behind the varying performance of stocks. It emphasizes the notion that markets reward exposure to certain risks, including value, momentum, quality, and size. Despite its theoretical appeal, practical outcomes often disappoint investors and fund managers, necessitating a reevaluation of the underlying models used in this strategy.
Research indicates that the issue may not reside in the data itself but rather in the construction of the models.
A recent study suggests that many factor models are flawed because they confuse correlation with causation, leading to what is termed a factor mirage. This misunderstanding carries significant implications for investment strategies, ultimately affecting capital allocation and market efficiency.
Table of Contents:
The foundation of factor investing
The theoretical foundation of factor investing rests on the premise that certain risks can lead to consistent outperformance of specific assets. Trillions of dollars have been invested based on this principle, yet the results reveal a contrasting narrative. For example, the Bloomberg–Goldman Sachs US Equity Multi-Factor Index has maintained a low Sharpe ratio of only 0.17 since 2007, indicating that factor investing has not adequately rewarded investors, especially when considering costs.
Understanding model limitations
Traditionally, critics attribute these shortcomings to issues like backtesting overfitting or p-hacking, a process where researchers sift through data until they identify patterns that seem valuable. While this explanation holds some merit, it overlooks a more profound flaw: the systematic misspecification of models. The methodology often adopted in constructing factor models leans heavily on standard econometric practices, such as linear regressions and significance testing, which can inadvertently misrepresent relationships among variables.
By incorporating a collider—a variable influenced by both the factor and the returns—analysts may skew their estimates, leading to misleading conclusions. This misrepresentation can result in investors making poor decisions, such as investing in securities that should have been sold.
The danger of the factor mirage
A particularly concerning phenomenon is the factor zoo, which refers to numerous published anomalies that do not withstand real-world scrutiny. However, the factor mirage extends this concept further. It arises not merely from data manipulation but from inherent flaws within the models themselves, despite adherence to conventional econometric principles.
Models that include colliders might appear to provide better fit statistics, such as higher R² values and lower p-values. Unfortunately, this can lead to a false sense of security regarding their validity. For instance, if a researcher includes return on equity—an outcome influenced by profitability—this can create a misleading correlation with the quality factor being analyzed.
Real-world implications
This methodological error has far-reaching consequences. In practical settings, models that seem effective during backtesting may falter in actual trading scenarios, resulting in unexpected capital losses. For individual fund managers, these miscalculations can gradually erode portfolio returns, while for the broader market, they can distort capital allocation, leading to inefficiencies.
Research from ADIA Lab emphasizes that without accurately specified causal factor models, achieving an efficient portfolio is unattainable. Even if means and covariances are estimated perfectly, flawed underlying factors will yield suboptimal investment decisions. This realization indicates that investing is not merely about prediction but necessitates a deeper understanding of causal relationships.
A call for causal reasoning
To restore trust in factor investing, a shift towards causal reasoning is essential. The field of finance can benefit from lessons learned in other domains, such as medicine, which transitioned from correlation-based approaches to evidence-based practices decades ago. Embracing causal inference will enable finance professionals to identify the true sources of risk and return, resulting in more reliable capital allocation.
Research indicates that the issue may not reside in the data itself but rather in the construction of the models. A recent study suggests that many factor models are flawed because they confuse correlation with causation, leading to what is termed a factor mirage. This misunderstanding carries significant implications for investment strategies, ultimately affecting capital allocation and market efficiency.0
Research indicates that the issue may not reside in the data itself but rather in the construction of the models. A recent study suggests that many factor models are flawed because they confuse correlation with causation, leading to what is termed a factor mirage. This misunderstanding carries significant implications for investment strategies, ultimately affecting capital allocation and market efficiency.1
