Attribution Methods in Asset Pricing: Do They Account for Risk?

ArXiv ID: 2407.08953 “View on arXiv”

Authors: Unknown

Abstract

Over the past few decades, machine learning models have been extremely successful. As a result of axiomatic attribution methods, feature contributions have been explained more clearly and rigorously. There are, however, few studies that have examined domain knowledge in conjunction with the axioms. In this study, we examine asset pricing in finance, a field closely related to risk management. Consequently, when applying machine learning models, we must ensure that the attribution methods reflect the underlying risks accurately. In this work, we present and study several axioms derived from asset pricing domain knowledge. It is shown that while Shapley value and Integrated Gradients preserve most axioms, neither can satisfy all axioms. Using extensive analytical and empirical examples, we demonstrate how attribution methods can reflect risks and when they should not be used.

Keywords: XAI in Finance, Shapley Values, Integrated Gradients, Asset Pricing Axioms, Risk Attribution, Equities

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Holy Grail
  • Why: The paper presents heavy mathematical formalization with axioms, proofs, and stochastic calculus concepts (like Greeks and Black-Scholes), while also using empirical examples and demonstrating practical implications for risk management.
  flowchart TD
    A["Research Goal: Validate if attribution methods<br>account for risk in asset pricing"] --> B["Domain: Finance & Risk Management"]
    B --> C["Methodology: Derive Asset Pricing Axioms"]
    C --> D["Process: Test Shapley Values<br>vs Integrated Gradients"]
    D --> E["Data: Empirical & Analytical Examples"]
    E --> F{"Key Finding"}
    F -->|Shapley Values| G["Preserves most axioms<br>but fails specific risk axioms"]
    F -->|Integrated Gradients| H["Preserves most axioms<br>but fails specific risk axioms"]
    G & H --> I["Outcome: Attribution methods require<br>domain-specific adjustments for risk"]