Rejoinder - An approach for identifying and predicting economic recessions in real-time using time-frequency functional models

摘要

Ghysels and Owyang mention that the business cycle is often modeled as the movement between two economic regimes and go on to say that forecasting the transition from one regime to another may be of particular importance. Additionally, they point out that previous studies have revealed a number of issues with the setup and performance of model-based forecasting of recessions and note that one important issue is how the turning points are defined. We agree with this assessment and feel it is important to emphasize that the intent of our approach is not necessarily to forecast turning-points. Although in principle, our model could effectively achieve this goal, the data being considered are not well suited for this task. Specifically, because the NASDAQ began electronic trading in February 1971, our analysis begins with the second quarter (Q_2) of 1971. Consequently, to have enough data for model development, we do not begin conducting nowcasting and forecasting until the third quarter of 1989. The span from Q_2-1971 through Q_1-1989 contains three recessionary periods and hence six turning-points. Similarly, the span from Q_2-1989 through Q_2-2010 contains three recessionary periods and hence a total of six turning-points. Thus, our training and validation data have six turning-point observations each; half are associated with going from expansion to recession and the other half going from recession to expansion. As a result, if these quarters behave differently from the other recessionary and expansion periods, these will be more difficult to identify without any additional data.