Fabio Canova, Mary Finn and Adrian R. Pagan*
Most real business cycle models have been assessed by studying the correspondence of their predictions to a set of stylised facts. This paper argues that such tests are not extensive enough and proposes to evaluate the models using standard econometric procedures. Specifically it is argued that these models should be studied by eliciting the restricted VAR representation underlying them and comparing it with the VAR estimated in an unrestricted way from the underlying data. Allowance is made for cases where the driving forces are integrated and when they are stationary. When forces such as technology shocks are integrated these models produce a specific set of predictions about the cointegrating vectors as well as a set of restrictions upon the dynamics. The approach is illustrated using a real business cycle model estimated by Burnside, Eichenbaum and Rebelo. This model has been subjected to some formal testing based upon stylised facts, and it therefore seems an appropriate one upon which to utilize the formal econometric procedures.
In the last decade, real business cycles (RBC) models have gone from the preliminary explorations of Long and Plosser (1983) and Kydland and Prescott (1982) to well developed and tested models such as Burnside; Eichenbaum and Rebelo (1993) and McGrattan (1991). Early models could be regarded as ‘idealised’, in the sense adopted in the philosophical literature summarised in
* Financial support from the U. K. Economic and Social Research Council under grant R000233447 is gratefully acknowledged by both authors. We are indebted to Neil Ericsson for helpful comments.