A0317
Title: Likelihood based estimation of nonlinear dynamic stochastic general equilibrium models
Authors: Elnura Baiaman - Hitotsubashi University (Japan)
Roberto Leon-Gonzalez - GRIPS (Japan) [presenting]
Abstract: A new likelihood-based approach is proposed using perturbation methods to estimate nonlinear DSGE models. A nonlinear approximation is implicitly used for the policy function that is invertible with respect to the shocks, implying that shocks can be recovered uniquely from some of the control variables in the approximation. Based on this approximation, the likelihood can then be obtained by using a standard change of variables theorem and a Lagrange inversion formula. This technique is implemented to estimate the DSGE model. In contrast with previous likelihood-based approaches, this method allows for unobserved non-stochastic state variables and requires neither additional shocks nor simulation to evaluate the likelihood. Using US data, the proposed approach is demonstrated to the well-known neoclassical growth model of a prior study. In addition to the baseline model, versions of the model are also considered in which the structural shocks have time-varying variances. It is found that a nonlinear heteroscedastic model has much better empirical performance. It is a much better fit for the observed data than the linearized model. In addition, the monetary policy shock is found to primarily drive the time changes in the uncertainty in the economy.
