Paper: arXiv 2310.02014

Authors: Marcin Pitera, Miklós Rásonyi

Abstract

In this short paper we introduce a new class of performance measures based on certainty equivalents defined via scaled utility functions. We analyse their properties, show that the corresponding portfolio optimization problem is well-posed under generic conditions, and analyse the link between portfolio dynamics, benchmark process, and utility function choice in the long-run setting.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 1.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper is highly theoretical, featuring advanced mathematics like risk measures, utility theory, and limit theorems, with no empirical backtests or data mentioned.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Utility-based Acceptability Indices<br>Performance Measures"] --> B["Methodology<br>Scale Utility via Certainty Equivalent<br>Analyse Properties"]
  B --> C["Data/Inputs<br>Multi-Asset Portfolio Dynamics<br>Benchmark Process"]
  C --> D["Computational Process<br>Long-Run Optimization<br>Generic Conditions Check"]
  D --> E["Key Findings<br>Well-Posed Optimization<br>Link Dynamics & Utility Choice"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper introduces a new class of performance measures based on certainty equivalents defined via scaled utility functions.

Source sentence
In this short paper we introduce a new class of performance measures based on certainty equivalents defined via scaled utility functions.

Assumes 4

  • The utility function U is concave, strictly increasing, bounded from above and twice continuously differentiable on ℝ. model
    Source sentence
    To this end, we consider a concave, strictly increasing, and bounded from above utility function U ∈C2(R); for brevity, we use notation U(∞) := limx→∞U(x).
  • U is scale‑aversion regular, i.e., µγ1(X) ≤ µγ2(X) for 0 < γ1 ≤ γ2. model
    Source sentence
    UTILITY-BASED ACCEPTABILITY INDICES 3 We use parameter γ > 0 to quantify the risk aversion using the convention that the bigger the value of γ, the more risk-averse we are. To this end, we say that the utility function U is scale aversion regular, if for any X ∈L∞and 0 < γ1 ≤γ2 < ∞ we get (2.2) µγ1(X) ≤µγ2(X), that is (uγ(·))γ>0 is a non-decreasing function of γ.
  • For some examples the log‑returns rt are assumed to be i.i.d. statistical
    Source sentence
    For simplicity, let us assume that φ is such that (rt)t∈T is an i.i.d. sequence.
  • Exponential utility is used in several sections as a specific utility choice. model
    Source sentence
    Let us assume that we are given the exponential utility U defined in (2.6).

Data

No dataset named in the text.

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • The utility‑based acceptability index α(X) is defined as the supremum of risk‑aversion parameters γ such that the corresponding certainty‑equivalent risk measure µγ(X) is non‑positive. theoretical
    Source sentence
    Then, the acceptability index α: L0 →[0, +∞] based on utility U is UTILITY-BASED ACCEPTABILITY INDICES 7 given by (3.1) α(X) := sup{γ > 0 : µγ(X) ≤0}, X ∈L0, where (µγ)γ>0 is a family defined in (2.1); in (3.1) we use convention sup ∅:= 0.
  • For exponential utility and normally distributed one‑step log‑returns rt ~ N(m,σ²) with m>λ, the index evaluates to α = 2(m−λ)/σ². theoretical
    Source sentence
    (4.11) For instance, if rt ∼N(m, σ), where m > λ and σ > 0, then α(ln VT (φ) −λT ) = 2(m −λ)/σ2, as µγ(r1 −λ) = −(m −λ) + γ 2 σ2; note that this is the value of the standard risk-aversion coefficient in the mean-variance optimisation.
  • With a non‑exponential (modified exponential) utility, the long‑run acceptability index can become infinite (J∞(φ,λ)=∞). theoretical
    Source sentence
    The choice of a non-exponential utility in a (near) i.i.d. setting may lead to a situation when we always get J∞(·, λ) = ∞, i.e. the acceptability index could be somewhat too lenient on a long time horizon. In this setting we always get J∞(φ, λ) = ∞.
  • The long‑run performance depends on the memory of the return process: super‑diffusive (σ=∞) yields J∞=0, diffusive (σ∈(0,∞)) yields a finite positive J∞, and zero variance (σ=0) yields a degenerate case. theoretical
    Source sentence
    In the second case, we are in the super-diffusive regime and get J∞(φ, λ) = 0. Finally, in the third case, we are in the intermediate (diffusive) regime where the variance of ST scales linearly in T thus we obtain a finite, positive J∞(φ, λ).

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 10 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.