|
Activity Number:
|
320
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Business and Economics Statistics Section
|
| Abstract - #309962 |
|
Title:
|
Real-Time Estimation of Revised GDP Based on an Estimated Model of Initial and Revised GDP
|
|
Author(s):
|
Baoline Chen*+ and Peter Zadrozny
|
|
Companies:
|
Bureau of Economic Analysis and Bureau of Labor Statistics
|
|
Address:
|
1441 L Street NW, Washington, DC, 20230,
|
|
Keywords:
|
Kalman filtering ; Missing data ; revised data ; reduced form modeling
|
|
Abstract:
|
This paper develops and illustrates with U.S. GDP a two-step method based on an estimated VAR model for estimating in any period the final GDP for that period. Step one estimates the model for included variables by applying either OLS or maximum likelihood using three initial releases and three annual revisions of GDP from 1977 to 2003, indexed either by periods to which they pertain or by periods in which they are released. Step two estimates final GDP in every period using only sample information which is available through that period. This can be done by applying the missing-data Kalman filter to the estimated model using data in sparse real-time form. In the application, filtered estimates of final GDP are compared to eventually released final values in terms of RMSE measures of closeness.
|
