Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This article evaluates the consistency of the NBER business cycle reference dates. It finds that the early reference dates are derived from detrended data, whereas the dates after 1927 are derived from data in levels. To evaluate the importance of this and other changes in technique, I derive a simple algorithm that matches the postwar NBER peaks and troughs closely. When this algorithm is applied to data for 1884 to 1940, the new prewar dates differ systematically from the NBER dates and challenge the conventional view that recessions have gotten shorter over time.