This paper introduces a new way to describe how firms choose actions in a noisy market. The authors borrow mathematical tools from physics — path integrals and a Schrödinger-type equation — to study an economy that satisfies Walrasian equilibrium and Pareto efficiency while firms interact non-cooperatively in Markovian feedback Nash equilibrium. The goal is to find optimal firm strategies when the system evolves under random shocks and firms look ahead to future consequences of their actions.

Rather than building a value function and solving a Hamilton-Jacobi-Bellman (HJB) equation, the paper casts the firm's problem as a Lagrangian stochastic control problem. A Lagrangian here means an objective summed (integrated) over time. The stochastic model is approximated over short time steps and Ito’s lemma (a rule for calculus with random processes) is applied. After a mathematical transform called a Wick rotation, the authors obtain a Schrödinger-type equation. Optimal policies come out of a continuously differentiable Ito process constructed with integrating factors, which is a standard tool to solve differential equations.

This route matters because standard HJB methods can be hard to use when the underlying dynamics are strongly nonlinear or when decisions today affect future constraints (so-called forward-looking dynamics). The path-integral formulation naturally links individual decision rules to the evolution of the population distribution of firms through a forward Fokker–Planck equation (a way to track how probabilities evolve). That connection makes it possible to compare noncooperative Nash outcomes with mean-field game limits, which describe interactions among many small agents.

There are important caveats. The solutions produced by this path-integral approach need not be unique, consistent with Feynman–Kac type representations mentioned in the paper. The method avoids some restrictive assumptions required by HJB approaches, but it does not eliminate all difficulties with equilibrium selection or with the computational tasks that arise in high-dimensional models. The paper also contrasts its results with those from the Pontryagin maximum principle and with promised-utility and saddle-point formulations, noting that alternative formulations each have their own feasibility and computational issues.