Paper: arXiv 2603.01232
Authors: Ruodu Wang, Jingcheng Yu
Abstract
We study submodularity for law-invariant functionals, with particular attention to convex risk measures. Expected losses are modular, and certainty equivalents are submodular exactly when the loss function is convex. Law-invariant coherent risk measures are submodular exactly when they are coherent distortion risk measures, including Expected Shortfall (ES), and several deviation measures are also submodular. Beyond positive homogeneity, submodularity is restrictive for convex risk measures. We give a complete characterization for shortfall risk measures via the Arrow–Pratt measure of risk aversion, show that optimized certainty equivalents are always submodular, and prove that adjusted Expected Shortfall (AES) is submodular only when it reduces to ES. An empirical illustration for daily US equity returns finds no ES submodularity violations, many Value-at-Risk (VaR) violations, and relatively few AES violations.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 3.0/10
- Quadrant: Lab Rats — theoretically deep, empirically untested
Why this score: The paper presents a rigorous and deep theoretical analysis of submodularity across key classes of risk measures with detailed mathematical proofs. While it includes a small empirical illustration using US equity data, the core contribution is purely theoretical rather than empirical or computational.
Research Flowchart
flowchart TD
A["Research Goal: Characterize submodularity for law-invariant & convex risk measures"] --> B["Theoretical Framework: Evaluate Certainty Equivalents, Distortion, Shortfall, & AES Measures"]
A --> C["Empirical Input: Daily US Equity Returns Data"]
B --> D["Mathematical Characterization: Apply Arrow-Pratt measure & convexity constraints"]
C --> E["Empirical Testing: Check submodularity violations for VaR, ES, & AES"]
D --> F["Theoretical Outcomes: ES & OCE are always submodular; AES reduces to ES; VaR is non-submodular"]
E --> G["Empirical Outcomes: 0 ES violations, many VaR violations, few AES violations"]
Cited References (32)
- Measuring and Testing Dependence by Correlation of Distances (2007)
- Premium calculation: Why standard deviation should be replaced by absolute deviation1 (1990)
- Coherent measures of risk (1999)
- Submodularity in machine learning and artificial intelligence (2022)
- Adjusted E xpected S hortfall (2022)
- An old-new concept of convex risk measures: The optimized certainty equivalent (2007)
- Parametric Measures of Variability Induced by Risk Measures (2022) — link
- Choquet representability of submodular functions (2018)
and 24 more
- Submodular financial markets with frictions (2022) — link
- Risk measures: rationality and diversification (2011)
- Asset demand without the independence axiom (1989)
- Bayes risk, elicitability, and the Expected Shortfall (2021)
- Convex measures of risk and trading constraints (2002)
- Stochastic Finance. An Introduction in Discrete Time (2016)
- Submodular risk allocation (2019)
- Monotonic mean-deviation risk measures — → 2312.01034
- Quantitative Risk Management: Concepts, Techniques and Tools (2015)
- Introduction to the mathematics of ambiguity (2004)
- On concavity and supermodularity (2008)
- Risk aversion and insurance propensity (2025)
- Comparison Methods for Stochastic Models and Risks (2002)
- Risk aversion in the small and in the large (1964) — link
- Optimizing distortion riskmetrics with distributional uncertainty (2025) — link
- Mathematical Risk Analysis. Dependence, Risk Bounds, Optimal Allocations and Portfolios (2013)
- Integral representation without additivity (1986)
- Minimizing a Submodular Function on a Lattice (1978) — link
- Valuations on B anach lattices (2020)
- Characterization, robustness and aggregation of signed C hoquet integrals (2020)
- Distortion riskmetrics on general spaces (2020)
- Generalized deviations in risk analysis (2006)
- Diversification quotients based on V a R and E S (2023)
- The dual theory of choice under risk (1987)
1 of 32 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.