Policy Evaluation and DSGE Models

One of the most forceful condemnations of the previous generation of structural econometric models was the Lucas [1976] critique. It was based on the lack of adequate microeconomic foundations of these models and argued that the estimated parameters were not truly structural, because they consisted of combinations of both structural and policy parameters.

In principle, DSGE models should not suffer from the Lucas critique. After all, their parameters are structural and are clearly distinguished from policy parameters. In addition, it is much easier to conduct full-fledged welfare comparisons with DSGE models.

However, in practice, empirical DSGE models fall short of this ideal. The dynamics implied by theoretically consistent DSGE models suggest much faster adjustment toward steady state than the dynamics found in the actual data. As a result, reconciling the models with the data has resulted in significant ad hoc additions of sources of persistence in empirical DSGE models. This approach is not unlike the methodology applied by Klein [1950] in his original Keynesian econometric model, which was initially derived from microeconomic foundations, but in which a data-determined lag structure was allowed.

For example, in estimated DSGE models, habit persistence has been used to explain the slow adjustment of consumption, convex costs of changing investment have been added to account for the dynamics of investment, and indexation on past inflation has been used to account for the dynamics of the Phillips curve. These additional assumptions are neither suggested by the theory, nor do they enjoy independent empirical support. They are introduced simply because they improve the statistical properties of the estimated models. Hence, in practice, estimated DSGE models also contain exogenous nonstructural parameters that are not totally immune to the Lucas critique.

23.4