Seminario dipartimentale a.a 2019/2020 nell'ambito del programma STaRs - Supporting Talented Researchers

Simulation-based Matching Inference with Applications to DSGE Models 

INTERVIENE: Prof. Lynda Khalaf, Carleton University, Canada

ABSTRACT: Simulation-based matching methods are commonly used to estimate structural parameters through auxiliary statistics that summarize key features of the data and model implications. This paper develops a general procedure for inference with auxiliary statistics neither assuming identification of parameters of interest nor a one-to-one binding function. Specifically, the conditions underlying asymptotic validity of our simulators in conjunction with appropriate bootstraps are characterized beyond the strict and exact calibration of Dridi, Guay and Renault (2007). Such settings include impulse-response (IR) matching for DSGE models, which we analyse in a laboratory environment and a variant of the DSGE model of Del Negro and Schorfheide (2008). In addition to usual Wald-type statistics that combine structural or reduced form IRs (Christiano, Eichenbaum and Evans (2005), Inoue and Killian (2013), Guerron-Quintana, Inoue and Killian (2017)), we analyze local projections (Jordà (2005), Jordà and Kozicki (2011)) through a factor-analytic measure of distance adapted from Bai (2013) which eschews the need to define a weighting matrix. We study exact and miscalibrated cases. Overall, our simulations illustrate the superiority of our proposed factor-analytic local projection approach to IR matching and associated asymptotic framework by documenting useful testable directions robust-to-calibration.

Per informazioni: giovanni.urga@unibg.it