Recover Dynamic Utility from Observable Process: Application to the Economic Equilibrium
Nicole El Karoui,Mohamed Mrad
DOI: https://doi.org/10.1137/18M1235843
2021-02-23
SIAM Journal on Financial Mathematics
Abstract:SIAM Journal on Financial Mathematics, Volume 12, Issue 1 , Page 189-225, January 2021. Decision making under uncertainty is generally considered as the selection of an optimal sequence of actions in an uncertain environment. Its calibration raises the "inverse" problem to recover the criterion from the data. A classical example in economy is the theory of "revealed preference" introduced by Samuelson [Economica, 5 (1938), pp. 61--71]. The observable at a given date , , is an increasing function of a real parameter (the wealth in economy). The process is called the characteristic process. The objective is to recover a dynamic stochastic utility , "revealed" in the sense where its performance is without bias, more formally when " is a martingale." The increasing of in and the concavity of the utility lead to privilege the so-called adjoint process in the linearization of the problem; the one to one correspondence between and for a given characteristic process is used intensively. We focus on the triplets, bringing great attention to their initial conditions. We establish a necessary and sufficient condition for the existence of at least one solution to the "revealed" utility problem. An operational version is as follows: " is a supermartingale for any and is a martingale for ." Moreover, there is an equivalent intrinsic framework, where in addition the processes " , , " are supermartingales. Itô's semimartingale framework is used to illustrate this characterization. The operational version ensures that the revealed utility is a solution of a nonlinear stochastic PDE. Less obvious is its interpretation as a stochastic value function of some optimization problem. The financial markets framework appears as a special case, under stronger assumptions. Then, we revisit the dynamic equilibrium problem as in He and Leland [Rev. Financial Stud., 6 (1993), pp. 593--617], by considering it as a revealed utility problem. We solve the problem in a random environment, by characterizing all the equilibria, in showing that the only possible conjugate dynamic utilities are the mixture of stochastic dual power utilities.
social sciences, mathematical methods,mathematics, interdisciplinary applications,business, finance