| Popis: |
Estimators of Dynamic Stochastic General Equilibrium (DSGE) Model parameters, as well as impulse response functions, can be wildly inaccurate when data used in the estimation process are de-trended; even if the data are de-trended in the same manner, the model is de-trended. However, little is known about inferences of DSGE parameters and impulse response functions when raw data are used. This may be attributable to difficulties in applying the law of large numbers and the central limit theorem on sample means or functions of sample means when data are not derived from stationary processes. The good news for DSGE models is that the equilibrium conditions, represented by the first-order conditions of agent problems used to build the impulse response functions, are usually written as a non-linear combination of stationary variables at the true value of the parameters. In this study, we exploited that property to suggest the conditions under which the generalized method of moments (GMM), the indirect inference (II) estimators, and the minimum chi-square estimators are consistent and asymptotically Gaussian distributions. We also suggested procedures and conditions under which the GMM bootstrap, the indirect inference bootstrap, and the minimum chi-square bootstrap for DSGE model parameters are valid. For empirical application, we used U.S. data to assess the impulse response functions -- due respectively to money supply shock, government spending shock, and productivity shock -- in a DSGE framework in which the Federal Reserve Bank set the policy rate that controlled the raw value of the nominal gross domestic product. This empirical analysis would have been very difficult without our theoretical results. |
