Page 367 - Special Topic Session (STS) - Volume 3
P. 367
STS550 Baoline Chen et al.
Combination nowcasts of advance estimates of
private consumption of services in the U.S
National Accounts
*†
Baoline Chen , Kyle Hood*
University of the Philippines
Abstract
This paper evaluates combination forecast methods for nowcasting advance
quarterly estimates of private consumption (PCE) of services in the U.S.
national accounts using data from 2009 to 2018. In a previous study, we
showed that both the bridge-equation and bridging-with-factors frameworks
could improve the accuracy of advance estimates of detailed PCE services
components by reducing revisions when quarterly data become available.
However, degrees of reduction in revisions vary over time and across PCE
services categories. Studies have shown that despite unstable performances
of individual nowcasting or forecasting models, combination forecasts often
improve upon individual nowcasts or forecasts. In this study, we evaluate
alternative methods to combine nowcasts from general-bridge-equation (GB)
and bridging-with-factors (BF) frameworks. We consider weights for
combination based on simple averaging (mean and median), information-
criterion averaging, and Bates-Granger averaging with leave-one-out cross-
validation (LOOCV) errors. We evaluate the performances of combined
forecasts by comparing their root mean squared revisions (RMSR).
Keywords
Nowcasting; forecast combination; model averaging; national accounts
1. Introduction
In the United States, Gross Domestic Product (GDP) is released three times
each quarter. The first “advance” estimate is made approximately one month
after the quarter’s end and is based on the most limited source data, while two
more estimates, the “second” and “third,” are based on more-detailed or less-
preliminary source data. Revisions in GDP result from using these new or revised
source data and can be rather large for some of its components. Quarterly GDP
estimates are built up from detailed components of the major subaggregates
* Bureau of Economic Analysis, Washington, DC. The views expressed herein are those of the
authors and do not reflect the views of the Bureau of Economic Analysis or the Department of
Commerce.
† Corresponding author. Email: baoline.chen@bea.gov.
356 | P a g e