Equilibrium master equations for time-inconsistent problems with distribution dependent rewards

TL;DR

This paper introduces a novel equilibrium master equation approach for solving time-inconsistent problems with distribution-dependent rewards, applicable to behavioral finance and economics, and demonstrates its effectiveness through portfolio choice applications.

Contribution

It develops a new non-local PDE framework on Wasserstein space, refining derivatives and Itô's formula, to find equilibrium solutions in time-inconsistent stochastic control problems, distinct from mean field games.

Findings

Reformulates extended HJB equation under weaker assumptions

Provides a new procedure for equilibrium solutions in mean-ES portfolio problems

Reexamines dynamic portfolio choice with rank-dependent utility

Abstract

We provide a unified approach to find equilibrium solutions for time-inconsistent problems with distribution dependent rewards, which are important to the study of behavioral finance and economics. Our approach is based on {\it equilibrium master equation}, a non-local partial differential equation on Wasserstein space. We refine the classical notion of derivatives with respect to distribution and establish It$\hato$'s formula in the sense of such refined derivatives. Our approach is inspired by theories of Mckean-Vlasov stochastic control and mean field games, but is significantly different from both in that: we prohibit marginal distribution of state to be an input of closed loop control; we solve the best reaction to individual selves in an intra-person game instead of the best reaction to large populations as in mean field games. As applications, we reexamine the dynamic portfolio choice problem with rank dependent utility based on the proposed novel approach. We also recover the celebrated extended HJB equation when the reward of the problem has a nonlinear function of expectation while reformulating and weakening the assumptions needed. Most importantly, we provide a procedure to find an equilibrium solution of a dynamic mean-ES portfolio choice problem, which is completely new to the literature.