An auxiliary particle filter for nonlinear dynamic equilibrium models

Abstract

We develop a particle filter algorithm to approximate the likelihood function of nonlinear dynamic stochastic general equilibrium models. The new algorithm reduces the Monte Carlo variance of likelihood approximation and accelerates the convergence of posterior sampler. It requires much fewer particles to achieve comparable results as currently available particle filters. We illustrate our algorithm in Bayesian estimation of a new Keynesian macroeconomic model.