Real business-cycle realizations, 1925–1995: A comment

Abstract

In their paper, Gregor Smith and Stanley Zin address an important question: how well do real-business-cycle models account for the historical behavior of the economy? This empirical exercise is in the spirit of traditional econometrics. It would hardly be surprising to see such comparisons in other contexts, but it is surprising in the case of real-business-cycle models. The standard empirical methodology, following Kydland and Prescott's (1982) original article on real-business-cycle models, is calibration. Typically, calibrated real-business-cycle models have been evaluated through subjective comparisons between selected sample moments for the actual economy and the corresponding moments from simulated data generated using the model. Smith and Zin's paper is especially attractive because it seeks to bridge the gap between the calibration methodology and traditional econometrics. The central empirical question about any model is, how well does the model reproduce characteristic features of the world? If a model does reproduce those features well--perchance, perfectly--one cannot logically deduce that the model is the true one that generated the data. What one can logically deduce is that the model belongs to a class--much narrower than the class of all possible models--that can reproduce the data to a given degree of accuracy. That is itself useful knowledge--especially if rival models do not fall within such a class. Calibrators have typically emphasized (e.g., Kydland and Prescott 1991 and 1996) that their models are highly stylized and are meant to capture certain dimensions of the data, while ignoring others. Kydland and Prescott, for example, argue that standard econometrics biases the investigator against the real-business-cycle model, because it makes the model responsible, as it were, for features of the data it was not designed to capture. The standards of traditional econometrics may undermine the usefulness of economic models for policy analysis, because policy analysis