Abstract:
|
This paper evaluates general bridge equation and bridging with factor frameworks as well as a set of forecast combination methods for nowcasting advance estimate of quarterly private consumption (PCE) of services in the U.S. national accounts using real time data from 2009 to 2018. We show that both individual frameworks could improve accuracy of advance estimate of detailed PCE services components by reducing revisions when quarterly source data become available. However, degrees of reduction in revision vary over time and across PCE services categories. To explore further improvements in the accuracy of advance estimate, we also evaluate alternative methods to combine nowcasts from different specifications of the general bridge equation and bridging with factors models. We consider weights for combination based on simple averaging (mean, median, trimmed means), information-criterion-based averaging, Bates-Granger averaging with leave-one-out cross-validation errors, and log-likelihood averaging. We evaluate the performances of all methods by comparing their root mean squared revisions in advance estimate of each detailed PCE services component.
|