Quantum Monte Carlo for Economics: Stress Testing and Macroeconomic Deep Learning

Monte Carlo (MC) simulations are used throughout economics, such as to solve economic models and extend the sample size of small datasets. However, MC suffers from a slow convergence rate, which often results in a computational bottleneck and limits the usefulness of MC. Yet new developments in quantum computation may soon remove this bottleneck with the help of the quantum Monte Carlo (QMC) algorithm, which speeds up the time taken by classical MC.

This paper is the first to apply the QMC algorithm to problems in economics and among the first to apply quantum computation more generally to this field. We compare MC and QMC for typical economics problems and introduce economists to quantum computation. Our paper focuses on two such problems: (a) stress testing banks subject to credit shocks and fire sales and (b) solving a neoclassical dynamic stochastic general equilibrium model with deep learning. In stress testing, we use QMC to estimate the expected system-wide capital loss of the banking system after a few periods of financial stress. In the deep learning solution of the neoclassical macroeconomic model, QMC estimates how a large economy behaves in the presence of random productivity shocks.

Our paper is written such that an economist without any knowledge of quantum computation can gradually progress to fully implementing the QMC algorithm for their problem of interest. In addition, the paper offers several improvements to the algorithm itself and supplies code for performing QMC and its benchmarking.