On Scoring Asymmetric Periodic Probability Models of Turning-point Forecasts.

To forecast the turnaround of an economy we do not usually take seasonal effects into account. Recently, the author showed that business cycle turning points as well as durations do not appear to be uniformly distributed throughout the year (see Ghysels, 1991a). In this paper we suggest improving the forecasting performance of turning-point predictions by adopting periodic hazard models. Following Diebold and Rudebusch (1989), we construct several formal probability models and score their prediction performance. The results indicate that for sequential forecasting rules significant gains can be made by exploiting periodicities in turning-point probabilities. [ABSTRACT FROM AUTHOR]