arxiv: 2605.02666 · v1 · submitted 2026-05-04 · 💱 q-fin.MF

Recognition: unknown

Pareto frontier of portfolio investment under volatility uncertainty and short-sale constraints market

Jing He, Shuzhen Yang

Pith reviewed 2026-05-08 01:34 UTC · model grok-4.3

classification 💱 q-fin.MF

keywords portfolio optimizationsublinear expectationvolatility uncertaintyPareto frontiermean-variance modelshort-sale constraintsefficient frontier

0 comments

The pith

A single risk factor w produces an analytical polynomial expression for the Pareto frontier in the SLE-MUV portfolio model under volatility uncertainty.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The authors model volatility uncertainty in stock portfolios using sublinear expectations, which naturally produce both an upper and lower bound on variance for any given allocation. They introduce a single risk factor w that combines these maximum and minimum risks into one optimization problem while respecting short-sale constraints. They prove that the set of optimal portfolios traces a continuous convex curve whose weights are given by an explicit polynomial formula in w. In tests on simulated data and real US and Chinese stock markets, this SLE-MUV approach delivers better risk-adjusted returns than the standard mean-variance model.

Core claim

The Pareto frontier of the SLE-MUV model is a continuous convex curve, and its optimal solution can be expressed as a polynomial analytical expression with respect to the risk factor w. The model is tested empirically on simulated data, US stock market data, and A-share market data, where it significantly improves the risk-adjusted return of the investment portfolio compared to the traditional Mean-Variance model.

What carries the argument

The SLE-MUV model, which uses a risk factor w to couple the maximum and minimum variances arising from volatility uncertainty under sublinear expectations.

Load-bearing premise

That sublinear expectations accurately represent volatility uncertainty and that a single risk factor w suffices to couple maximum and minimum risks without creating inconsistencies in the optimal portfolios.

What would settle it

Finding a value of w where the polynomial-derived weights do not solve the underlying optimization problem, or showing that out-of-sample portfolios from the model have risk-adjusted performance no better than or worse than mean-variance portfolios.

Figures

Figures reproduced from arXiv: 2605.02666 by Jing He, Shuzhen Yang.

**Figure 1.** Figure 1: Geometric illustration of a smooth convex Pareto frontier between the upper and lower view at source ↗

**Figure 2.** Figure 2: Pareto frontier of portfolios constructed with synthetic data on the first sample day view at source ↗

**Figure 3.** Figure 3: Cumulative Wealth as of End Sample,Sharpe Ratio: SLE-MUV Model vs. Traditional view at source ↗

**Figure 4.** Figure 4: Protfolio Cumulative Wealth for SLE-MUV model and traditional MV Model under view at source ↗

**Figure 5.** Figure 5: Cumulative Wealth as of December 22, 2025 and Sharpe Ratio : SLE-MUV Model vs. view at source ↗

**Figure 6.** Figure 6: Protfolio Cumulative Wealth for SLE-MUV model and traditional MV Model under view at source ↗

**Figure 7.** Figure 7: Cumulative Wealth as of December 22, 2025, Sharpe Ratio: SLE-MUV Model vs. view at source ↗

**Figure 8.** Figure 8: Portfolio Cumulative Wealth for SLE-MUV model and traditional MV Model under view at source ↗

read the original abstract

In this paper, we investigate a portfolio investment problem under volatility uncertainty and short-sale constraints market via sublinear expectation which is used to model volatility uncertainty. We assume the stocks admit volatility uncertainty. Thus the related portfolio has upper variance (maximum risk) and lower variance (minimum risk). By introducing a risk factor $w$ to conduct coupled modeling of the maximum and minimum risks, a simplified Sublinear Expectation Mean-Uncertainty Variance (SLE-MUV) model is constructed. Theoretically, we show that the Pareto frontier of the SLE-MUV model is a continuous convex curve, and its optimal solution can be expressed as a polynomial analytical expression with respect to the risk factor $w$. Empirically, we systematically test the practical performance of the SLE-MUV model and conduct comparative analysis with the traditional Mean-Variance (MV) model as the benchmark based on three sets of samples -- simulated generated data, data of the US stock market and the A-share market. The empirical results show that the SLE-MUV model can significantly improving the risk-adjusted return of the investment portfolio.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper's core claim of a single polynomial expression for optimal weights under short-sale constraints is the part that needs checking first, since inequality constraints normally produce case splits.

read the letter

The paper introduces an SLE-MUV model that uses sublinear expectations to handle volatility uncertainty, adds a tunable risk factor w to couple maximum and minimum portfolio variances, and asserts both a continuous convex Pareto frontier and a closed-form polynomial solution for the weights even with non-negativity constraints on holdings. They back this with comparisons to standard mean-variance on simulated data and two real equity markets, claiming better risk-adjusted returns for the new model. That combination of uncertainty modeling and an analytical frontier is the genuinely new piece relative to the cited literature. The empirical section at least tries to show practical relevance rather than stopping at theory. The main soft spot is the polynomial claim itself. Short-sale constraints are inequalities, so any KKT-style derivation will have complementary slackness conditions and multiple active-set regimes; each regime gives its own algebraic expression for the weights. A single polynomial that works for all w would require either that the active set never changes or that the pieces miraculously combine into one polynomial, neither of which is obvious. The abstract gives no hint of case analysis or verification that the pieces join smoothly, so the central theoretical result rests on an unshown step. The dependence on w also creates a practical circularity: the reported optimum shifts with the choice of w, yet the paper offers no guidance on how to set or estimate it. This work is aimed at quantitative finance researchers who already use sublinear expectations or robust mean-variance variants. It is coherent enough on its own terms to deserve a serious referee who can verify the derivation and the empirical controls, even if the polynomial result turns out to need qualification. I would send it out rather than desk-reject.

Referee Report

3 major / 1 minor

Summary. The manuscript proposes the SLE-MUV model for portfolio optimization under volatility uncertainty using sublinear expectations, incorporating short-sale constraints. A risk factor w is introduced to couple the maximum and minimum variances, leading to the claim that the Pareto frontier is a continuous convex curve and that the optimal weights are given by a polynomial expression in w. The paper also presents empirical comparisons with the mean-variance model on simulated and real stock data, asserting improved risk-adjusted returns.

Significance. If the central theoretical claim is rigorously established, the work offers an analytical solution to a mean-uncertainty-variance optimization problem, which could be significant for robust portfolio management in uncertain volatility environments. The convexity of the frontier and closed-form solution would be valuable contributions to mathematical finance. The empirical tests, if they include proper statistical validation, could support practical applicability. However, the introduction of the free parameter w requires careful justification for the results to be fully convincing.

major comments (3)

[Abstract] Abstract: The claim that 'its optimal solution can be expressed as a polynomial analytical expression with respect to the risk factor w' is difficult to reconcile with the short-sale constraints, which impose non-negativity inequalities on the weights. Standard optimization theory (KKT conditions with complementary slackness) suggests that the solution would be piecewise, depending on which constraints are binding for different ranges of w, rather than a single polynomial for all w. This needs to be addressed with explicit case analysis or a proof that the expression remains polynomial across the domain.
[Theoretical section] Theoretical development: The abstract states that the Pareto frontier is a continuous convex curve and the solution is polynomial in w, yet provides no derivation steps, proof outline, or verification that the polynomial satisfies the original constrained problem. The role of w in coupling max and min variances must be shown to preserve convexity without introducing kinks at active-set boundaries.
[Empirical section] Empirical analysis: Superiority over the MV benchmark is asserted without reported metrics (e.g., Sharpe ratios, specific risk-adjusted returns), confidence intervals, or details on how w is selected or fitted in the tests on simulated, US, and A-share data. This prevents evaluation of whether the claimed improvements are statistically significant or robust.

minor comments (1)

[Model setup] The notation for upper and lower variances under sublinear expectation could be introduced more explicitly before the model formulation to aid readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments on our manuscript. We address each major comment below and indicate the revisions we will make to strengthen the presentation of the SLE-MUV model.

read point-by-point responses

Referee: [Abstract] Abstract: The claim that 'its optimal solution can be expressed as a polynomial analytical expression with respect to the risk factor w' is difficult to reconcile with the short-sale constraints, which impose non-negativity inequalities on the weights. Standard optimization theory (KKT conditions with complementary slackness) suggests that the solution would be piecewise, depending on which constraints are binding for different ranges of w, rather than a single polynomial for all w. This needs to be addressed with explicit case analysis or a proof that the expression remains polynomial across the domain.

Authors: We acknowledge that short-sale constraints generally produce piecewise solutions via KKT conditions. In our derivation, the coupled modeling via the risk factor w yields an explicit polynomial form for the optimal weights that satisfies the non-negativity constraints over the relevant domain of w. To make this fully rigorous, we will add an explicit case analysis in the revised theoretical section, verifying that the polynomial expression remains valid without requiring separate cases for active constraints, and confirm it meets the KKT conditions for the constrained problem. revision: yes
Referee: [Theoretical section] Theoretical development: The abstract states that the Pareto frontier is a continuous convex curve and the solution is polynomial in w, yet provides no derivation steps, proof outline, or verification that the polynomial satisfies the original constrained problem. The role of w in coupling max and min variances must be shown to preserve convexity without introducing kinks at active-set boundaries.

Authors: We agree that additional derivation details are needed. In the revised manuscript we will insert a step-by-step proof outline: first showing how w couples the maximum and minimum variances under sublinear expectation, then proving that the resulting objective produces a continuous convex Pareto frontier, and finally verifying that the polynomial solution satisfies the original constrained optimization problem without introducing kinks at constraint boundaries. revision: yes
Referee: [Empirical section] Empirical analysis: Superiority over the MV benchmark is asserted without reported metrics (e.g., Sharpe ratios, specific risk-adjusted returns), confidence intervals, or details on how w is selected or fitted in the tests on simulated, US, and A-share data. This prevents evaluation of whether the claimed improvements are statistically significant or robust.

Authors: We recognize that quantitative metrics and selection details are essential for evaluating the empirical claims. In the revision we will report specific risk-adjusted performance measures (including Sharpe ratios), confidence intervals for the improvements, and a clear description of how the risk factor w is chosen or fitted for each dataset. These additions will allow readers to assess statistical significance and robustness of the SLE-MUV results relative to the MV benchmark. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation proceeds from model definition to closed-form solution

full rationale

The SLE-MUV model is explicitly constructed by introducing the tunable risk factor w to couple maximum and minimum variances under sublinear expectations. The subsequent theoretical result—that the Pareto frontier is a continuous convex curve and that optimal weights admit a polynomial expression in w—is obtained by analytically solving the resulting constrained quadratic program for varying w. This is a direct derivation from the stated objective and constraints rather than a re-labeling or self-referential fit. Short-sale constraints are acknowledged in the model but do not alter the non-circular character of the algebraic solution step. No self-citation chains, ansatz smuggling, or renaming of known results are required for the central claim. The empirical section compares the model against MV benchmarks on external data and is independent of the theoretical derivation.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on sublinear expectation as the modeling framework for volatility uncertainty and on the ad-hoc introduction of the scalar risk factor w; no independent evidence is given that w can be chosen without reference to the target data.

free parameters (1)

risk factor w
Single scalar introduced to couple maximum and minimum portfolio variances; its specific value determines the location on the frontier and the optimal weights.

axioms (1)

domain assumption Sublinear expectation properties suffice to represent volatility uncertainty for all stocks
Invoked to justify upper and lower variances for every portfolio.

pith-pipeline@v0.9.0 · 5485 in / 1391 out tokens · 41303 ms · 2026-05-08T01:34:50.818452+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

32 extracted references

[1]

Probability, Uncertainty & Quantitative Risk , volume=

Linear regression under model uncertainty , author=. Probability, Uncertainty & Quantitative Risk , volume=
[2]

Available at SSRN 5225465 , year=

Conditional Value-at-Risk Under Reward-Penalty Mechanism with Applications to Robust Portfolio Management , author=. Available at SSRN 5225465 , year=
[3]

A Worst-Case Risk Measure by

Pei, Ziting and Wang, Xishun and Yue, Xingye , year =. A Worst-Case Risk Measure by. Acta Mathematicae Applicatae Sinica, English Series , doi =
[4]

The Journal of Finance , volume=

Portfolio selection , author=. The Journal of Finance , volume=
[5]

Peng , journal=

S. Peng , journal=. Nonlinear expectations and stochastic calculus under uncertainty with robust
[6]

2015 , note =

Li He and Qinghuai Liu , title =. 2015 , note =

2015
[7]

Journal of Financial Econometrics , year=

Improving Value-at-Risk prediction under model uncertainty , author=. Journal of Financial Econometrics , year=
[8]

Pei, Z. T. and Yue, X. Y. and Zheng, X. T. , journal=. Numerical methods for two-dimensional
[9]

White , keywords =

D.J. White , keywords =. Epsilon-dominating solutions in mean-variance portfolio analysis , journal =. 1998 , issn =

1998
[10]

Fuzzy Optimization and Decision Making , volume=

Portfolio selection problems with Markowitz’s mean--variance framework: a review of literature , author=. Fuzzy Optimization and Decision Making , volume=. 2018 , publisher=

2018
[11]

Journal of business , volume=

The variation of certain speculative prices , author=. Journal of business , volume=. 1963 , publisher=

1963
[12]

2007 , organization=

Peng, Shige , booktitle=. 2007 , organization=

2007
[13]

Distributional uncertainty of the financial time series measured by

Peng, Shige and Yang, Shuzhen , journal=. Distributional uncertainty of the financial time series measured by. 2022 , publisher=

2022
[14]

Journal of Machine Learning Research , year =

Steven Diamond and Stephen Boyd , title =. Journal of Machine Learning Research , year =
[15]

Nature Methods , year =

Virtanen, Pauli and Gommers, et.al , title =. Nature Methods , year =
[16]

Worst-case values of target semi-variances with applications to robust portfolio selection , journal =

Jun Cai and Zhanyi Jiao and Tiantian Mao , keywords =. Worst-case values of target semi-variances with applications to robust portfolio selection , journal =. 2025 , issn =

2025
[17]

Worst-case risk measures of stop-loss and limited loss random variables under distribution uncertainty with applications to robust reinsurance , journal =

Jun Cai and Fangda Liu and Mingren Yin , keywords =. Worst-case risk measures of stop-loss and limited loss random variables under distribution uncertainty with applications to robust reinsurance , journal =. 2024 , issn =

2024
[18]

Distributionally robust optimal allocation of financial assets under the uncertainty and irrationality , journal =

Jianping Li and Jiaxin Yuan and Jun Hao , keywords =. Distributionally robust optimal allocation of financial assets under the uncertainty and irrationality , journal =. 2026 , issn =

2026
[19]

Jones and Helenice O

Dylan F. Jones and Helenice O. Florentino , title=. The Palgrave Handbook of Operations Research , chapter=. 2022 , month=

2022
[20]

Decision sciences , pages=

Multi-objective optimization , author=. Decision sciences , pages=. 2016 , publisher=

2016
[21]

Fifty years of portfolio optimization , journal =

Ahti Salo and Michalis Doumpos and Juuso Liesiö and Constantin Zopounidis , keywords =. Fifty years of portfolio optimization , journal =. 2024 , issn =

2024
[22]

The Review of Economic Studies , volume=

Mean-variance analysis in the theory of liquidity preference and portfolio selection , author=. The Review of Economic Studies , volume=. 1969 , publisher=

1969
[23]

The Review of Economic Studies , volume=

A note on uncertainty and indifference curves , author=. The Review of Economic Studies , volume=. 1969 , publisher=

1969
[24]

The Review of Economic Studies , volume=

The fundamental approximation theorem of portfolio analysis in terms of means, variances and higher moments , author=. The Review of Economic Studies , volume=. 1970 , publisher=

1970
[25]

Journal of Financial and Quantitative Analysis , volume=

General proof that diversification pays , author=. Journal of Financial and Quantitative Analysis , volume=. 1967 , publisher=

1967
[26]

International Journal of Systems Science , volume=

Mean-variance-skewness model for portfolio selection with transaction costs , author=. International Journal of Systems Science , volume=. 2003 , publisher=

2003
[27]

Computers & Operations Research , volume=

Neural network-based mean--variance--skewness model for portfolio selection , author=. Computers & Operations Research , volume=. 2008 , publisher=

2008
[28]

European Journal of Operational Research , volume=

Mean-variance-skewness model for portfolio selection with fuzzy returns , author=. European Journal of Operational Research , volume=. 2010 , publisher=

2010
[29]

Optimization , volume=

Uncertain random mean--variance--skewness models for the portfolio optimization problem , author=. Optimization , volume=. 2022 , publisher=

2022
[30]

Soft computing , volume=

Uncertain random variables: A mixture of uncertainty and randomness , author=. Soft computing , volume=. 2013 , publisher=

2013
[31]

European Journal of Operational Research , volume=

Mean-variance model for portfolio optimization problem in the simultaneous presence of random and uncertain returns , author=. European Journal of Operational Research , volume=. 2015 , publisher=

2015
[32]

Statistics, Optimization & Information Computing , volume=

Uncertain Portfolio Optimization Problems: Systematic Review , author=. Statistics, Optimization & Information Computing , volume=