Reliability Of The Quarterly National Income And Product Accounts Of The United States, 1947–1971

The reliability of the quarterly national income and product (NIP) accounts of the United States is examined from several standpoints. First, possible sources of error in the quarterly NIP accounts are explored, the most important being the lack of appropriate data, seasonal adjustment errors, sampling errors and biases, and the nature of the U.S. statistical system. Next, four ways of assessing the reliability of the accounts are considered. The most weight is given to measures of revisions in early estimates of the quarterly NIP aggregates. Results of previous studies of revisions are reviewed, and a summary of a major study of revisions for the period 1947–71 is given. The other ways of assessing reliability which are examined are the effect of errors on economic policy making, analysis of the statistical discrepancy, and expert judgment on sources and methodology. The degree of accuracy is judged to be generally sufficient for the policy decisions for which the NIP estimates are used. The early estimates of a quarter's change in GNP almost always distinguish whether the ultimate estimate will be large or small and will usually distinguish whether the ultimate estimate will be larger or smaller than the preceding quarter. While the accuracy of the estimates has generally been sufficient, the accuracy for 1965 was judged insufficient by policymakers. There is some evidence that errors have been reduced over time.