Tailoring Portfolio Choice via Quantile-Targeted Policies

TL;DR

This paper explores how investors can optimize portfolio choices by targeting specific quantiles of outcomes, using a novel distributional reinforcement learning approach to align strategies with risk preferences.

Contribution

It introduces a distributional actor-critic algorithm for learning time-consistent, quantile-targeted policies that are adaptable to various risk preferences and market conditions.

Findings

Downside-focused agents effectively reduce risk in volatile states

High quantile targeting concentrates on high-dispersion assets with upside potential

Empirical portfolios align with investor risk objectives

Abstract

We study the dynamic investment decisions of investors who prioritise specific quantiles of outcomes over their expected values. Downside-focused agents targeting low quantiles reduce risk in states with high variance, while those with a preference for high quantiles concentrate in sleeves with high dispersion when there is potential for upside. These results provide a microfoundation for volatility management, demonstrating that reducing exposure in volatile states is an optimal response for risk-averse investors and rationalising inverse-variance heuristics. We propose a distributional actor-critic algorithm that learns time-consistent policies tailored to these specific risks, irrespective of the utilitys functional form. The quantile value can be mapped onto interpretable tilts, and the performance of empirically chosen portfolios aligns with investors objectives.