Affiliations: [a] Abt Associates Inc., 55 Wheeler Street, Cambridge,
Massachusetts 02138-1168, USA. Tel.: +1 617 492 7100; Fax: +1 617 492 5219 | [b] Boston College, Newton, Massachusetts 02167,
USA
Note: [1] The comments of Christopher F. Baum, Sengul Dagdeviren, and Ayla Ogus of the
Department of Economics at Boston College and Joseph W. Duncan of Dun &
Bradstreet Corporation are also acknowledged. Data were provided courtesy of
Maureen Haver of Haver Analytics, Inc. and Allen H. Young of the Bureau of
Economic Analysis. Any errors remain the responsibility of the authors.
Abstract: This paper reviews the performance of the National Income and Product Accounts
(NIPA) over the period 1986 through 1991. Forecasts published jointly by the
American Statistical Association and the National Bureau of Economic Research
are decomposed into two components, pure prediction error and data measurement
error. The results show that more than two-thirds of observed forecast error
could be attributed to data measurement error over all forecast horizons
examined, including those as brief as one quarter ahead. Outdated estimation
methods for large portions of GNP, namely PCE and S&L, are identified as
potential driving forces behind the problems caused by measurement error.
Keywords: Prediction error, consensus forecasts, business cycles
