Estimation of DSGE Models by Non-Gaussian Vector Autoregressions

We propose a new impulse response matching procedure for estimating the parameters of dynamic stochastic general equilibrium models from observed macroeconomic time series. The estimator is based on an indirect inference framework in which the auxiliary model is a structural vector autoregressive model. Identification in the auxiliary model is achieved by exploiting non-Gaussianity through an estimator based on distance covariance, which belongs to the class of independent component analysis methods. We establish the asymptotic properties of the distance covariance estimator in general and within a vector autoregressive framework, and of the resulting indirect inference estimator. A Monte Carlo study evaluates the finite-sample performance of the proposed procedure. Finally, we illustrate the method with an application to a New Keynesian DSGE model.