The econometric approach to business-cycle analysis reconsidered

The research performed and reported in this article represents an econometric analysis of the model generated by John Blatt (1978). A major point made by Blatt was that the standard statistical procedures used by economists would have led to egregious errors in inference; namely, estimating a stable linear relationship in a world generated by an unstable nonlinear model. The main point of this paper is that the use of specification error tests pioneered by one of the authors leads to immediate and overwhelming rejection of the incorrect model. That empirical economists must learn to test their models before attempting to make inferences is the major policy conclusion.