NOTE.This special note was prepared by Robert P. Parker.
THE TWO ARTICLES that precede this note present two independently derived, but conceptually equivalent, measures of current-dollar gross product by industry for 1992 that are prepared by the Bureau of Economic Analysis (BEA).
This note explains BEA's use of the NIPA data for the GPO estimates, discusses the differences between the GPO estimates and the I-O estimates for 1992, and describes the steps BEA is taking to address these differences.
BEA views the GDP estimate that is derived in the benchmark I-O accounts as the most accurate estimate available. It is based on the most reliable source dataprimarily detailed and comprehensive information from the most recent quinquennial economic censusesand it is calculated within the framework of the I-O accounts, which track the detailed input and output flows in the economy./1/
In order to prepare timely annual estimates of GPO by industry, BEA uses the industry distributions of the NIPA components of gross domestic income (GDI). The GDI estimates are available annually, while the I-O value-added estimates are available at roughly 5-year intervals. In addition, because of a lack of comprehensive source data on intermediate inputs, the I-O estimates of industry value added reflect a widespread use of indirect estimating methodologies; the missing source data are primarily on business purchases of services and purchases of goods by nonmanufacturing industries./2/ As a result, while the I-O-based estimate of GDP is viewed as a more accurate measure of GDP than the GDI-based estimate of GDP, the I-O-based estimates of the distribution of GDP by industry are not necessarily more accurate than GDI-based estimates of GDP by industry.
The industry distributions of GDI are available on a more timely basis, but they also reflect the use of less-than-adequate source data. In particular, IRS tabulations of corporate income tax returnswhich are the source data for the estimates of corporate profits, depreciation, and net interestare available only on an enterprise, or company, basis, so they must be converted by BEA to an establishment, or plant, basis. However, the methodologies used for this conversion are less than adequate, and they are not applicable to net interest, for which no conversion is made./3/
Another source data problem that affects both the I-O value-added and the GPO estimates is the lack of consistency in industry classification at the establishment level. The I-O estimates largely reflect the industry classifications assigned by the Bureau of the Census in the economic censuses. The GPO estimates reflect a mix of classifications: The wage and salary component of GDI is based on industry classifications assigned by the Bureau of Labor Statistics for Employment and Wages, and the nonfarm proprietors' income component is based on industry classifications assigned by the Internal Revenue Service for the Statistics of Income program. In addition, GDI in theory should equal GDP, but in practice, these measures differ because they are estimated using largely independent and less-than-perfect source data. The difference between these two NIPA measures is called the "statistical discrepancy." For the GPO series, the statistical discrepancy is presented as a component of private industries because BEA assumes that it does not affect the estimates for government./4/ For the I-O accounts, this difference does not exist, because the components of both final expenditures and value added result from the internally consistent I-O framework and because these accounts do not include independent estimates of the detailed components of value added.
The accompanying table presents an approximation of the differences in the industry distribution of GDP for 1992 on the basis of the presently published GPO and the new I-O estimates. In order to make a valid comparison, it was necessary to adjust both series. The GPO estimates were adjusted to reflect the new