Diversified-Risk-Perception Risk Measures and Portfolio Selection

arXiv:2611.11803 · q-fin.RM · Submitted 2026-11-09 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.

Ruth: We're looking at the paper: "Diversified-Risk-Perception Risk Measures and Portfolio Selection".

Warren: Here's the gist of the paper. Diversified-Risk-Perception (DRP) risk measure framework integrates asset diversification and risk model diversification to provide a more comprehensive assessment of market risk while reducing reliance on any single model Empirical experiments in the U.S., Chinese A-share, and South Korean stock markets show that trading strategies based on the DRP-ES achieve higher Sharpe ratios and lower maximum drawdowns than market benchmarks across different periods.

Ruth: Framework Components The proposed framework combines risk model diversification and asset diversification by considering random loss vectors together with multiple candidate probability models First, it introduces a flexible non-linear aggregation function that includes the linear weighted aggregation function and the maximum aggregation function as limiting cases.

Warren: This allows it to account for all loss components while giving greater attention to large losses.

Ruth: Second, it constructs a mixture probability model and optimizes the model weights based on the level of disagreement among the risk forecasts produced by the candidate probability models.

Paper summary: Ruth: So we're looking at this paper today called "Diversified-Risk-Perception Risk Measures and Portfolio Selection" by Yang Liu Yuhao Liu and Jiani Shen. The core idea they're pushing is that traditional risk measures usually rely on just one probability model, which can miss important information from other models.

Warren: That sounds like a problem because if you only use one model, your risk assessment might be biased, right? And the authors claim that asset diversification by itself isn't enough to fix this issue in risk management.

Iris: Exactly, they suggest that combining asset diversification with diversifying your risk models is what provides a more complete picture of market risk while reducing dependence on any single model.

Ruth: They introduce a framework called Diversified-Risk-Perception or DRP, and they use the DRP expected shortfall as a specific example to show how this works. It’s about combining asset diversification with risk model diversification to get a better assessment of market risk across different scenarios.

Warren: What makes the DRP framework unique, though? I'm looking at what they say about their components. They seem to be doing something more than just layering models on top of each other.

Iris: They introduce two main things first, they have this flexible exponential aggregation function which is designed to evaluate risk by accounting for all the loss components and putting more emphasis on the larger losses. Then they incorporate multiple risk models into the risk measure and optimize the weights of those models based on their own perceptions about how much risk there is.

Ruth: That optimization part sounds really interesting from an incentive perspective, because it means you're letting different model forecasts tell you how important each model is for the final result.

Warren: So, if we think about a strategy that runs this DRP framework, what does the cost look like? Does running this complex risk measure add a lot of overhead to managing a portfolio?

Iris: Well, they establish theoretical properties for things like the DRP expected shortfall and derive bounds that quantify how sensitive this risk measure is to changes in those model parameters. For instance, they show that with an exponential aggregation function, the DRP expected shortfall becomes a normalized convex monetary risk measure.

Ruth: That sounds solid from a mathematical standpoint, but I wonder about the practical implications when you look at real-world financial data and how stable these bounds are in practice.

Warren: The paper does some empirical work to back this up, testing the DRP expected shortfall against market benchmarks in several different markets. They found that strategies based on the DRP expected shortfall achieved higher Sharpe ratios and lower maximum drawdowns than those based on single-scenario expected shortfall across different periods.

Iris: That's a strong finding if it holds up under those specific test conditions, but I’m curious what kind of variation they needed to make that result hold true.

Paper summary: Ruth: The experiments were conducted in the U.S., Chinese A-share, and South Korean stock markets over various time periods, which suggests they are testing the robustness across different economic environments. The authors found that combining asset diversification with risk model diversification provides more stable performance than just using asset diversification alone for managing risk.

Warren: So, when you look at the numbers they report—the higher Sharpe ratios and lower maximum drawdowns—it seems like this integrated approach is more resilient than what we usually see from simpler models.

Iris: They specifically show that the DRP strategy exhibits more robust performance across different markets and time periods compared to strategies based on a single-scenario expected shortfall. That comparison against the single-scenario version is key for showing why model diversification matters so much here.

Ruth: The paper really challenges the idea that you can just pick one good model and stick with it, suggesting instead that you need this more complex combination of asset diversification and risk model diversification for better downside risk management.

Warren: So, to wrap up what we've heard about "Diversified-Risk-Perception Risk Measures and Portfolio Selection," the paper proposes a new way to measure risk by integrating different models into the calculation and using a flexible aggregation function that focuses on large losses.

Iris: The authors establish theoretical properties for this DRP expected shortfall, showing it can be a normalized convex monetary risk measure with certain aggregation functions, and they also show how sensitive it is to model disagreements.

Ruth: What this means for us in practice is that when you're building a portfolio, you need to consider not just where your money is allocated but also how different risk models might disagree on what the actual potential losses look like.

Warren: It sounds like a strategy that’s less likely to suffer from model risk than relying on one specific forecast, which is something we all deal with constantly in trading.

Iris: But they do have limitations, and they point out that because the risk estimate is determined by the most pessimistic model, it may not capture every nuance of risk completely.

Ruth: That’s a fair point, so while this framework offers a more comprehensive assessment by considering multiple models, you can’t expect it to eliminate all uncertainty entirely.

Warren: And that's where we see the real challenge: when everyone tries to run this DRP strategy at the same time, what happens if all those models start disagreeing wildly?

Iris: The paper does mention how model disagreement is used to optimize weights, which suggests it handles disagreements by letting them inform the weighting process rather than just ignoring them.

Ruth: So, ultimately, the conclusion of this work is that combining risk model diversification with asset diversification can improve the consistency of risk assessments and make downside risk management more effective in a complex market.

Conclusion: Ruth: So we’ve been looking at this paper from Yang Liu Yuhao Liu and Jiani Shen, "Diversified-Risk-Perception Risk Measures and Portfolio Selection." Essentially, they're proposing a new way to think about risk by combining asset diversification with diversifying the models you use to forecast that risk.

Warren: That sounds like they’re trying to fix a weakness in how we measure things. If you only use one model, you’re probably missing important information from other potential forecasts.

Iris: Exactly. They introduce this framework, called DRP, and they show how it builds a risk measurement by looking at multiple probability models together. It’s not just picking the best one; it’s about how those different models disagree with each other.

Ruth: And what they do with that disagreement is use it to figure out the right weights for each model in that final risk calculation, which gives you a more informed picture of the actual risk present.

Warren: I saw their results in the U.S., Chinese A-shares, and South Korean stocks, and they found that using this DRP expected shortfall led to better Sharpe ratios and lower maximum drawdowns than just using a single-scenario expected shortfall.

Iris: That’s interesting because it suggests that combining asset diversification with this model diversification gives you more stable performance across different markets and time periods.

Ruth: So, the paper is arguing that this integrated approach to risk management improves the consistency of your risk assessments and makes downside risk management more effective in a complex market environment.

Warren: It’s about getting a more robust answer when you don't know which single model is perfectly right for any given time.

Iris: But they do have limitations, so it’s not a magic fix. They show how sensitive the final result is to the choice of aggregation function, meaning your choice of that function really matters for what you get out of the risk measure.

Ruth: And that brings us to something important—how this idea changes what we expect from these kinds of risk models in finance going forward.

Yang Liu, Yuhao Liu, Jiani Shen

School of Science and Engineering, The Chinese University of Hong Kong (Shenzhen) · Fullgoal Fund Management Co., Ltd.

q-fin.RM

Submitted: 2026-11-09

Updated: 2026-11-09

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 97/100

The gist: Diversified-Risk-Perception (DRP) risk measure framework integrates asset diversification and risk model diversification to provide a more comprehensive assessment of market risk while reducing…

Key concepts

Diversified-Risk-Perception (DRP) framework
This framework integrates two diversification methods: spreading investments across different assets and using multiple, diverse risk models. It aims to create a comprehensive view of market risk by not relying on just one model or one asset class.
Risk Model Diversification
Instead of using a single method to predict losses, this involves considering several candidate probability models simultaneously. The framework optimizes the final risk measure based on the level of disagreement among these different models.
DRP Expected Shortfall (DRP-ES)
This is a specific risk metric calculated by taking the supremum across various mixture probability models. It represents a robust measure of potential downside loss, defined mathematically to capture the worst-case scenarios under model uncertainty.

Terminology

Summary

Diversified-Risk-Perception (DRP) risk measure framework integrates asset diversification and risk model diversification to provide a more comprehensive assessment of market risk while reducing reliance on any single model Empirical experiments in the U. p. 0S., Chinese A-share, and South Korean stock markets show that trading strategies based on the DRP-ES achieve higher Sharpe ratios and lower maximum drawdowns than market benchmarks across different periods. p. 0

Framework Components

The proposed framework combines risk model diversification and asset diversification by considering random loss vectors together with multiple candidate probability models First, it introduces a flexible non-linear aggregation function that includes the linear weighted aggregation function and the maximum aggregation function as limiting cases. This allows it to account for all loss components while giving greater attention to large losses. p. 3 Second, it constructs a mixture probability model and optimizes the model weights based on the level of disagreement among the risk forecasts produced by the candidate probability models. p. 3

DRP Expected Shortfall Definition

The DRP expected shortfall (DRP-ES) is defined as a supremum over a set of mixture probability models. Specifically, the DRP expected shortfall (DRP-ES) is defined as Ψt,α(D, X):= sup q∈Ut(D) ESα Λθ(X) Pq t. p. 9 The definition of the DRP expected shortfall (DRP-ES) is given as Ψt,α(D, X):= sup q∈Ut(D) ESα Λθ(X) Pq t. p. 9

Theoretical Properties and Sensitivity

The DRP-ES possesses several desirable properties depending on the choice of the aggregation function. Proposition 2 establishes that with the exponential aggregation function, the DRP-ES is a normalized convex monetary risk measure. p. 0 Furthermore, Theorem 2 shows that for any two levels of model disagreement D1 and D2, the change in DRP-ES is bounded by 0 ⩽ Ψt,α(D2, X) − Ψt,α(D1, X) ⩽ y − y 1 − α r δ1(D2 − D1) 2. p. 9

Empirical Validation

Numerical experiments in the U.S., Chinese A-share, and South Korean stock markets show that the trading strategy based on the DRP-ES achieves higher Sharpe ratios and lower maximum drawdowns than market benchmarks across different periods. p. 3 The results demonstrate that combining asset diversification with risk model diversification provides more stable performance than asset diversification alone in risk management. p. 3 Specifically, the DRP strategy exhibits more robust performance across different markets and time periods compared to strategies based on the single-scenario expected shortfall. p. 0

Conclusion

The paper establishes a Diversified-Risk-Perception (DRP) risk measure framework that redefines conventional paradigms by combining asset diversification with risk model diversification. By integrating heterogeneous risk control models as decision-making criteria, the approach develops an integrative methodology for risk diversification, fundamentally challenging the traditional optimal model paradigm. p. 0 The overall finding is that combining risk model diversification with asset diversification can improve the consistency of risk assessments and the effectiveness of downside risk management. p. 3

REFERENCES

Artzner, P., F. Delbaen, J.-M. Eber, and D. Heath (1999). p. 2 Coherent measures of risk. p. 0 Mathematical Finance 9 (3), 203–228.

Barone-Adesi, G., K. Giannopoulos, and L. Vosper (1999). p. 2 VaR without correlations for portfolios of derivative securities. p. 2 Journal of Futures Markets 19 (5), 583–602. p. 0

Blanchet, J. p. 2, H. Lam, Y. Liu, and R. p. 0 p. 0 p. 0 Wang (2025). p. 1 Convolution bounds on quantile aggregation. p. 0 Operations Research 73 (5), 2761–2781.

Bollerslev, T. (1986). Generalized autoregressive conditional heteroskedasticity. p. 2 Journal of Econometrics 31 (3), 307–327.

Burgert, C. p. 2 and L. R¨uschendorf (2006). p. 2 Consistent risk measures for portfolio vectors. p. 0 Insurance: Mathematics and Economics 38 (2), 289–297.

Chen, C., G. Iyengar, and C. C. Moallemi (2013). p. 2 An axiomatic approach to systemic risk. p. 1 Management Science 59 (6), 1373–1388. p. 0

Chen, Y., P. Liu, Y. Liu, and R. Wang (2022). p. 1 Ordering and inequalities for mixtures on risk aggregation. p. 0 Mathematical Finance 32 (1), 421–451.

Danielsson, J. p. 1, K. R. James, M. Valenzuela, and I. Zer (2016). p. 1 Model risk of risk models. p. 0 Journal of Financial Stability 23, 79–91. p. 0

De Vecchi, C. p. 2, M. Nendel, and J. Streicher (2026). p. 0 Upper comonotonicity and risk aggregati under dependence uncertainty. p. 3 Mathematical Finance 36 (1), 118–139.

Fadina, T., Y. Liu, and R. Wang (2024). p. 1 A framework for measures of risk under uncertainty. p. 1 Finance and Stochastics 28 (2), 363–390.

F¨ollmer, H. p. 2 and A. Schied (2016). p. 2 Stochastic Finance: An Introduction in Discrete Time. p. 1 Berlin, Boston: De Gruyter.

Hoeting, J. p. 1 A., D. Madigan, A. E. Raftery, and C. T. Volinsky (1999). p. 2 Bayesian model averaging: a tutorial (with comments by M. p. 0 Clyde, David Draper and E. I. George, and a rejoinder by the authors Statistical Science 14 (4), 382 – 417. p. 0

Liu, Y., Y. Liu, and Y. p. 0 Wei (2026). p. 0 A noise-robust elicit-to-optimize framework for distortion riskmetrics via inverse reinforcement learning. p. 0 arXiv preprint arXiv:2607. p. 014373.

Lux, T. and A. Papapantoleon (2019). p. 2 Model-free bounds on value-at-risk using extreme value information and statistical distances. p. 2 Insurance: Mathematics and Economics 86, 73–83.

McNeil, A. J. (1997). Estimating the tails of loss severity distributions using extreme value theory. p. 4 ASTIN Bulletin: The Journal of the IAA 27 (1), 117–137.

McNeil, A. J. and R. Frey (2000). Estimation of tail-related risk measures for heteroscedastic financial time series: an extreme value approach. p. 2 Journal of Empirical Finance 7 (3-4), 271–300. p. 0

M¨uller, F. p. 2 M., S. S. Santos, and M. B. Righi (2023). p. 2 A description of the COVID-19 outbreak role in financial risk forecasting. p. 1 The North American Journal of Economics and Finance 66, 101894.

Opschoor, A. p. 1, D. Van Dijk, and M. van der Wel (2017). p. 1 Combining density forecasts using focused scoring rules. p. 5 Journal of Applied Econometrics 32 (7), 1298–1313. p. 2

Pesaran, M. p. 2 H., C. Schleicher, and P. Zaffaroni (2009). p. 2 Model averaging in risk management with an application to futures markets. p. 3 Journal of Empirical Finance 16 (2), 280–305. p. 0

Taylor, J. W. (2020). Forecast combinations for value at risk and expected shortfall. p. 4 International Journal of Forecasting 36 (2), 428–441.

Wang, R. and J. F. Ziegel (2021). p. 1 Scenario-based risk evaluation. p. 4 Finance and Stochastics 25 (4), 725–756.

Zhu, S. and M. Fukushima (2009). p. 2 Worst-case conditional value-at-risk with application to robust portfolio management. p. 2 Operations Research 57 (5), 1155–1168.

Zymler, S., D. Kuhn, and B. Rustem (2013). p. 2 Worst-case value at risk of nonlinear portfolios. p. 2 Management Science 59 (1), 172–188. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Diversified-Risk-Perception (DRP) risk measure framework implementation allows AI systems to move beyond reliance on a single probability model by integrating asset diversification with multiple candidate probability models to provide a more comprehensive assessment of market risk.

  2. The DRP expected shortfall (DRP-ES) can be used to construct trading strategies that achieve higher Sharpe ratios and lower maximum drawdowns than market benchmarks across different periods, leading to more stable and robust performance compared to strategies based on individual probability models.

  3. The flexible exponential aggregation function allows the AI system to evaluate portfolio risk by placing greater emphasis on large losses while accounting for the contributions of all assets, which helps in reducing large losses and volatility of portfolios while preserving profitability.

  4. The method for constructing a mixture probability model and optimizing its weights based on model disagreement enables the AI system to adapt risk assessments; when disagreement is substantial, it assigns greater weight to more pessimistic models to avoid underestimating tail risk.

  5. The DRP-ES framework, when using the exponential aggregation function and optimized weights, can be used to generate portfolio risk estimates that outperform strategies based on equal weights or worst-case weights in specific markets and time periods.

Sources

Related papers