Title: Indirect inference estimation of misspecified DSGE asset pricing models using nonlinear vector autoregressions
Authors: Julie Schnaitmann - Universität Konstanz (Germany) [presenting]
Joachim Grammig - Eberhard Karls Universitaet Tuebingen (Germany)
Dalia Elshiaty - Eberhard Karls Universitaet Tuebingen (Germany)
Abstract: A two-step indirect inference strategy is proposed in order to analyze a class of DSGE asset pricing models. These models, which combine a one-sector stochastic growth model with external habit preferences and capital adjustment costs, have so far only been calibrated but not made subject to econometric analysis. The proposed strategy draws on and extends the idea of estimating misspecified DSGE models by sequential partial indirect inference (SPII). Drawing on the SPII philosophy, we use binding functions that facilitate the consistent estimation of some structural model parameters of interest, while treating others as nuisance parameters. We acknowledge that these parameters do not necessarily capture economic reality, but they are necessary to generate model-implied data. Moreover, we separate the estimation of the parameters dependent on the parts of the asset pricing model that they influence. Specifically, they are classified into technology parameters, which are estimated separately in the first step. The estimation of the investor preference parameters is performed in the second step, using the first-step estimates as input. The auxiliary parameters are delivered by the class of recently developed nonlinear vector autoregressive (NVAR) models, which allow for more flexible impulse-response functions than a standard VAR model.