Preference Robust Generalized Shortfall Risk Measure Based on the Cumulative Prospect Theory When the Value Function and Weighting Functions Are Ambiguous

TL;DR

This paper introduces a robust generalized shortfall risk measure based on cumulative prospect theory, accounting for ambiguity in value and weighting functions, with a computational approach and numerical experiments demonstrating its effectiveness.

Contribution

It proposes a novel preference robust approach to CPT-based shortfall risk, handling ambiguity in value and weighting functions through optimization and elicited preference data.

Findings

The robust model can be computed via linear programming.

Numerical experiments show improved risk assessment under ambiguity.

The approach effectively incorporates preference information to reduce ambiguity.

Abstract

The utility-based shortfall risk (SR) measure introduced by Folmer and Schied [15] has been recently extended by Mao and Cai [29] to cumulative prospect theory (CPT) based SR in order to better capture a decision maker's utility/risk preference. In this paper, we consider a situation where information on the value function and/or the weighting functions in the CPT based SR is incomplete. Instead of using partially available information to construct an approximate value function and weighting functions, we propose a robust approach to define a generalized shortfall risk which is based on a tuple of the worst case value/weighting functions from ambiguity sets of plausible value/weighting functions identified via available information. The ambiguity set may be constructed with elicited preference information (e.g. pairwise comparison questionnaires) and subjective judgement, and the ambiguity reduces as more and more preference information is obtained. Under some moderate conditions, we demonstrate how the subproblem of the robust shortfall risk measure can be calculated by solving a linear programming problem. To examine the performance of the proposed robust model and computational scheme, we carry out some numerical experiments and report preliminary test results. This work may be viewed as an extension of recent research of preference robust optimization models to the CPT based SR.