Evaluating Portfolio Strategies with Asymmetric Risk–Reward Measures
Summary
The document asks how to compare portfolio allocation strategies beyond the performance measures used in a cited study of optimized portfolios and the naive equal-weight strategy. The answer points to research on alternative risk and reward specifications for long-run portfolio optimization. Its focus is on measures that treat gains and losses asymmetrically, rather than summarizing risk through symmetric measures alone.
The cited work reports that objectives based on partial moments often performed better than counterparts based on central moments such as variance, and says the results were later replicated on another dataset. This offers one empirical direction for evaluating allocation methods: test whether downside-sensitive risk measures lead to different portfolio choices or performance. The document does not provide detailed definitions, datasets, numerical results, or a broad comparison of evaluation frameworks. Its evidence is limited to the authors’ reported findings, and the answer itself notes the empirical orientation of the cited work.
Key ideas
- Portfolio strategies can be evaluated with alternative specifications of risk and reward.
- Asymmetric measures distinguish between gains and losses, unlike symmetric measures such as variance.
- The cited research reports stronger performance for partial-moment objectives than for central-moment counterparts.
- Replication on a different dataset is mentioned, but detailed evidence and evaluation criteria are not provided.
Tags
Full text
# Reference Request: Horse Race for Portfolio Allocation # Reference Request: Horse Race for Portfolio Allocation Probably the most popular horse race study for portfolio strategies is > Optimal versus Naive Diversification: How Inefficient Is the 1/N Portfolio Strategy?, with DeMiguel, L. Garlappi and R. Uppal. The Review of Financial Studies 22(5), 1915--1953 (2009) In their Paper DeMiguel et al. use several datasets as well as simulated data and compare different portfolio strategies with several performance measures, for example: Sharpe Ratio, CE, Mean, Turnover, Return-Loss... Obviously the drawback of such strategies can always be found in their 'relatively' arbitrary choice of the datasets, the evaluation methods etc. I would like to learn other approaches (purely theoretical ones are also welcome) in literature to evaluate the goodness of a portfolio strategy. I really think that is important to improve the evaluation process in order to justify all these fancy allocation methods that are proposed in literature. What papers did you come across that propose different performance measures than DeMiguel? ## Answer by Enrico Schumann (score 1) https://quant.stackexchange.com/a/43466 An Empirical Analysis of Alternative Portfolio Selection Criteria and Risk-Reward Optimisation for Long-Run Investors: An Empirical Analysis look into alternative risk and reward specifications for portfolio optimisation, from a purely-empirical point of view. (Disclosure: I am one of the authors.) We looked particularly into such risk/reward functions that allow for an asymmetric treatment of returns (i.e. losses and gains are treated differently). We found that alternative risk functions often performed better than their symmetric counterparts (e.g. objective functions based on partial moments worked better than central moments such as variance). We later replicated the results on a different dataset in Risk-Reward Ratio Optimisation (Revisited).
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.