Information Criteria for Impulse Response Function Matching Estimation

We propose a new Information Criterion for Impulse Response Function Matching estimators of the parameters of a structural model based on classical Minimum Distance estimation. The advantages of our procedure are that: (i) it improves the efficiency of the estimates of the model's deep parameters; (ii) it allows the researcher to select the impulse responses that are more informative about the deep parameters. Our criterion applies to impulse responses estimated by VARs, local projections, as well as simulation methods. An empirical application to the estimation of representative Dynamic Stochastic General Equilibrium models show that our method can substantially improve inference.