Paper: arXiv 2507.23138

Authors: Alejandro Rodriguez Dominguez

Abstract

A recent line of research has argued that causal factor models are necessary for portfolio optimization, claiming that structurally misspecified models inevitably produce inverted signals and nonviable frontiers. This paper challenges that view. We show, through theoretical analysis, simulation counterexamples, and empirical validation, that predictive models can remain operationally valid even when structurally incorrect. Our contributions are fourfold. First, we distinguish between directional agreement, ranking, and calibration, proving that sign alignment alone does not ensure efficiency when signals are mis-scaled. Second, we establish that structurally misspecified signals can still yield convex and viable efficient frontiers provided they maintain directional alignment with true returns. Third, we derive and empirically confirm a quantitative scaling law that shows how Sharpe ratios contract smoothly with declining alignment, thereby clarifying the role of calibration within the efficient set. Fourth, we validate these results on real financial data, demonstrating that predictive signals, despite structural imperfections, can support coherent frontiers. These findings refine the debate on causality in portfolio modeling. While causal inference remains valuable for interpretability and risk attribution, it is not a prerequisite for optimization efficiency. Ultimately, what matters is the directional fidelity and calibration of predictive signals in relation to their intended use in robust portfolio construction.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10 (capped from 7.0: transaction costs not modelled; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 7.0)

Why this score: The paper presents advanced mathematical derivations, including theoretical analysis of misspecified models and a quantitative scaling law linking Sharpe ratios to signal alignment. It also validates these findings on real financial data (S&P 500) and synthetic data, demonstrating a strong empirical component.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Is causality necessary<br>for efficient portfolios?"]
  
  B["Methodology<br>Theoretical Analysis & Counterexamples"]
  C["Data/Input<br>Simulated & Real Financial Data"]
  
  D["Computational Process<br>Testing Structural Misspecification<br>vs Predictive Validity"]
  
  E["Outcome 1<br>Distinguish Alignment, Ranking, Calibration"]
  F["Outcome 2<br>Misspecified signals can still<br>yield convex efficient frontiers"]
  G["Outcome 3<br>Quantitative scaling law:<br>Sharpe ratios contract smoothly"]
  H["Outcome 4<br>Predictive signals support<br>coherent frontiers despite imperfections"]

  A --> B
  A --> C
  B --> D
  C --> D
  D --> E
  D --> F
  D --> G
  D --> H

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes that causal identification is not required for portfolio efficiency and introduces a three‑way geometric distinction—directional alignment, ranking preservation, and calibration—governing efficient frontiers.

Source sentence
We introduce and formalize a three-way distinction between directional alignment, ranking preservation, and calibration in portfolio optimization, showing that directional correctness is necessary but not sufficient for portfolio efficiency.

Assumes 2

  • Portfolio efficiency is analyzed within a static mean‑variance and closely related quadratic optimization framework. model
    Source sentence
    causal identification is not necessary for portfolio efficiency within static mean–variance and closely related quadratic portfolio optimization frameworks.
  • Directional alignment of predictive signals is necessary for portfolio efficiency. model
    Source sentence
    directional correctness is necessary but not sufficient for portfolio efficiency when signal magnitudes are miscalibrated.

Data

Backtest sample: global bond universe · fixed-income · daily

  • Global bond universe · Bloomberg market-prices commercial
    1,061 trading days · 1,350 instruments across multiple currencies, countries, sectors, maturities, seniority cl
    Source sentence
    Appendix E further reports a large‑scale empirical validation using a global bond universe comprising 1,350 instruments across multiple currencies, countries, sectors, maturities along the term structure, seniority classes, and credit ratings, observed over 1,061 trading days (from Bloomberg).
  • Synthetic simulation data synthetic unknown
    asset universes with N ∈{300, 500, 800, 1000}
    Source sentence
    Simulations are fully replicable, implemented in Python with fixed pseudo‑random seeds, and involve no ex post parameter tuning. We consider asset universes with N ∈{300, 500, 800, 1000}.

Costs

Not modeled A trading or execution result is reported without costs.

Finds 4

Reported performance (mean‑variance portfolio using shrinkage covariance and long‑only constraints): Sharpe 1.48

  • Causal factor models are not required for portfolio efficiency; directionally aligned but misspecified signals still yield smooth, convex efficient frontiers. theoretical
    Source sentence
    We show that structural misspecification, including omitted variables and nonlinear confounding, does not inherently invalidate portfolio optimization. Predictive models without causal identifiability can nevertheless yield directionally informative signals that sustain smooth and convex efficient frontiers.
  • Mis‑calibration alone attenuates Sharpe ratios while preserving alignment and ranking, leading to smooth degradation rather than collapse of performance. performance
    Source sentence
    miscalibration alone attenuates Sharpe ratios even when alignment and ranking are preserved. Robustness is characterized as smooth degradation rather than collapse, with explicit attenuation behavior and continuity of performance under increasing misspecification.
  • Empirical results on a global bond universe confirm that portfolio efficiency degrades smoothly with misspecification and is highly sensitive to covariance estimation but robust to constraints. performance
    Source sentence
    Performance is highly sensitive to covariance estimation, with shrinkage materially improving both average Sharpe ratios and significance rates. Portfolio constraints act as a stabilizing regularizer rather than destroying efficiency, consistent with Corollary 2.
  • Simulation experiments demonstrate that directional alignment preserves frontier convexity even under structural cancellation, while mis‑calibration reduces achievable risk‑adjusted returns. theoretical
    Source sentence
    The preservation of sign across weights indicates directional alignment, while attenuation of magnitudes reflects miscalibration. This validates that misspecification need not invalidate diversification, but primarily reduces achievable risk‑adjusted returns.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 25 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (62)

  1. The C omprehensive T ex A rchive N etwork ( CTAN ) (1993)
  2. Portfolio Selection: Efficient Diversification of Investments (1959)
  3. Portfolios from sorts (2005)
  4. Honey, I Shrunk the Sample Covariance Matrix (2004) — link
  5. Estimation with Quadratic Loss (1961)
  6. Predicting Risk/Return Performance Using Upper Partial Moment/Lower Partial Moment Metrics (2016) — link
  7. Portfolio Management under Stress: A Bayesian-Net Approach to Coherent Asset Allocation (2014) — link
  8. Epistemic Limits of Empirical Finance: Causal Reductionism and Self-Reference — → 2311.16570
and 54 more
  1. Adaptive Markets: Financial Evolution at the Speed of Thought (2017)
  2. The probability of backtest overfitting (2017) — link
  3. Evaluating trading strategies (2014) — link
  4. Adaptive Markets: Financial Evolution at the Speed of Thought (2017)
  5. Hierarchical Causality in Financial Economics (2014) — link
  6. Portfolio Selection (1952)
  7. Causal Factor Investing: Can Factor Investing Become Scientific? (2023) — link
  8. Correcting the Factor Mirage: A Research Protocol for Causal Factor Investing (2026) — link
  9. Causal Factor Analysis Is a Necessary Condition for Investment Efficiency (2025) — link
  10. When Factors Collide: Mapping Causal Spillovers across Global Asset Networks (2026) — link
  11. Causal Network Representations in Factor Investing (2025) — link
  12. Towards Automating Causal Discovery in Financial Markets and Beyond (2024) — link
  13. Causal discovery from nonstationary time series (2025) — link
  14. Causal Factor Analysis Is a Necessary Condition for Investment Efficiency (2025) — link
  15. Causal Factor Investing: Can Factor Investing Become Scientific? (2023) — link
  16. Causal Network Representations in Factor Investing (2025) — link
  17. Why Has Factor Investing Failed?: The Role of Specification Errors (2024)
  18. Causal Discovery from Nonstationary Time Series (2025) — link
  19. The Case for Causal Factor Investing (2024) — link
  20. PORTFOLIO SELECTION (1952) — link
  21. To Explain or to Predict? (2010) — link
  22. Statistical modeling: The two cultures (2001) — link
  23. Big data: New tricks for econometrics (2014) — link
  24. Machine learning: An applied econometric approach (2017) — link
  25. The credibility revolution in empirical economics: How better research design is taking the con out of econometrics (2010) — link
  26. On the sensitivity of mean-variance-efficient portfolios to changes in asset means (1991) — link
  27. The effect of errors in means, variances, and covariances on optimal portfolio choice (1993) — link
  28. The sampling error in estimates of mean-variance efficient portfolio weights (1999) — link
  29. The Markowitz optimization enigma: Is “optimized” optimal? (1989) — link
  30. Global portfolio optimization (1992)
  31. Risk reduction in large portfolios: Why imposing the wrong constraints helps (2003) — link
  32. Optimal versus naive diversification: How inefficient is the 1/N portfolio strategy? (2009) — link
  33. The surprising robustness of dynamic mean-variance portfolio optimization to model misspecification errors (2021) — link
  34. Portfolio selection with parameter and model uncertainty: A multi-prior approach (2007) — link
  35. Robustness (2008)
  36. Empirical asset pricing via machine learning (2020) — link
  37. Commonality in the determinants of expected stock returns (1996) — link
  38. … and the cross-section of expected returns (2016) — link
  39. Portfolio optimization based on neural networks sensitivities from assets dynamics respect common drivers (2023) — link
  40. A Portfolio’s Common Causal Conditional Risk-Neutral PDE — → 2401.00949
  41. Correlation and Beyond: Positive Definite Dependence Measures for Robust Inference, Flexible Scenarios, and Stress Testing for Financial Portfolios (2021) — link
  42. Long Story Short: Omitted Variable Bias in Causal Machine Learning (2021) — link
  43. Machine Learning Methods That Economists Should Know About (2019)
  44. Predictability in securities price formation: differences between developed and emerging markets (2020) — link
  45. When Factors Collide: Mapping Causal Spillovers across Global Asset Networks (2026) — link
  46. Causal Network Representations in Factor Investing (2025) — link
  47. Towards Automating Causal Discovery in Financial Markets and Beyond (2024) — link
  48. Causal discovery from nonstationary time series (2025) — link
  49. Deciphering the Causal Drivers of Bond Indexes: A Double Machine Learning Approach (2025) — link
  50. Linear and Nonlinear Causality in Financial Markets — → 2312.16185
  51. Causal Inference for Asset Pricing (2025) — link
  52. Causal Portfolio Optimization: Principles and Sensitivity-Based Solutions — → 2504.05743
  53. Causal Inference in Finance: An Expertise-Driven Model for Instrument Variables Identification and Interpretation (2024) — link
  54. A Comprehensive Review of Causal Inference in Banking, Finance, and Insurance (2025) — link

4 of 62 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.