Evaluating the strength of identification in DSGE models. An a priori approach

The strength of identification in structural models is a reflection of the empirical relevance of the model features represented by the parameters. Weak identification arises when some parameters are nearly irrelevant or nearly redundant with respect to the aspects of reality the model is intended to explain. The strength of identification is therefore not only a crucial requirement for the reliable estimation of models, but also has important implications for model development. This paper proposes a new framework for evaluating the strength of identification in linearized dynamic stochastic general equilibrium (DSGE) models prior to their estimation. In a parametric setting, the empirical implications of a model are contained in the likelihood function, which, for DSGE models, is completely characterized by the underlying structural model. I show how to use standard asymptotic theory to evaluate the theoretical properties of likelihood-based estimators at any point in the parameter space associated with the model. Furthermore, in addition to assessing the informativeness of the likelihood as a whole, I show how to determine which particular features of the data, such as moments of a given variable or a set of variables, are most important for the identification of a given parameter. The methodology is illustrated using a medium-scale business cycle model.