Robust optimal decisions with stochastic nonlinear economic systems

Abstract

Optimal decisions for nonlinear systems are usually computed on the basis that certainty-equivalent decisions are sufficiently accurate. This premise is tested in the present paper using an approach that evaluates, using Monte Carlo simulations, any expectations bias that nonlinearities may introduce. Optimal decisions are determined by incorporating this bias with a robust decision formulation. The expected value of the policy objective function is optimized simultaneously with the sensitivity of the problem to given sources of uncertainty in the model. Numerical results, based on the National Institute of Economic and Social Research model of the UK economy, are used to highlight two specific points. The first is that robust decision performs better than simple certainty-equivalent deterministic decision and that increased robustness makes the policy more risk-averse. The second point is that the bias due to the nonlinearity of the model is significant in terms of the policy and the endogenous variables.