Box: The Statistical Discrepancy

Gross domestic product (GDP) measures output as the sum of final expenditures—consumer spending, private investment, net exports, and government consumption and investment. Gross domestic income (GDI) measures output as the sum of the costs incurred and the incomes earned in the production of GDP. In theory, GDP should equal GDI; in practice, they differ because their components are estimated using largely independent and less-than-perfect source data. In the national income and product accounts (NIPA's), the difference between GDP and GDI is called the "statistical discrepancy"; it is recorded in the NIPA's as an "income" component that reconciles GDI with GDP (see NIPA table 1.9).

Recently, there has been considerable public debate about the growth rate of the U.S. economy because since the early 1990's, growth measured by real GDI has increased faster than growth measured by real GDP./1/ Some analysts maintain that the higher rate of growth of real GDI is more consistent with declines in the unemployment rate and with anecdotal information about increases in productivity in services-producing industries. This debate has important implications for market participants and policymakers.

BEA views GDP as a more reliable measure of output than GDI, because it considers the source data underlying the estimates of GDP to be more accurate. For example, most of the annual source data used for estimating GDP are based on complete enumerations, such as Federal Government budget data, or are regularly adjusted to complete enumerations, such as the quinquennial economic censuses and census of governments. In addition, all the expenditure components of GDP are revised every 5 years to reflect BEA's benchmark input-output accounts, which are prepared within an internally consistent framework that tracks the input and output flows in the economy. For GDI, only the annual tabulations of employment tax returns and Federal Government budget data are complete enumerations, and only farm proprietors' income and State and local government budget data are regularly adjusted to complete enumerations. For most of the remaining components of GDI, the annual source data are tabulations of samples of income tax returns.

To improve the accuracy of the components of GDI that are based on tabulations of income tax return information and of employment tax returns, BEA adjusts the tax return data for misreporting, largely using information from audit studies conducted by the Internal Revenue Service. For 1994, the major adjustments were to wages and salaries ($74.0 billion), to nonfarm proprietors' income ($199.1 billion), and to corporate profits ($78.1 billion). Because the Census Bureau uses tax return information in the preparation of the economic censuses, BEA also adjusts the sales and receipts data from the censuses used in estimating GDP. These adjustments are smaller than those for GDI primarily because the adjustments to GDI for misreporting of wages and salaries on employment tax returns and misreporting of deductions on income tax returns do not affect the sales or receipts data from the censuses. Consequently, errors in the misreporting adjustments have a greater impact on the estimates of GDI than on the estimates of GDP.

BEA also views GDP to be more accurate than GDI because more of the critical annual source data are available on a timely basis. For example, for this year's annual revision, preliminary 1996 results and final 1995 results were available for more of the GDP source data; among the source data used for GDI, final 1995 tabulations of corporate income tax returns were not available.

The relative accuracy of GDP and GDI is also affected by the extent to which each measure has components for which there are no direct source data. The estimates of GDP for 1996 are missing direct source data for several components of consumer spending for services, of exports and imports of services, of home improvements, and of State and local government spending. Estimates of GDI for 1996 are missing direct source data for most of other labor income, of nonfarm proprietors' income, of corporate profits of small businesses, of interest paid and received, and of depreciation. Past trends in these series indicate that it is more difficult to project the missing data for the components of GDI than of GDP. Conversely, there also are new rapidly growing services, such as Internet and cellular phone services, for which gaps in source data are likely to cause larger errors in GDP than in GDI because GDI would likely include the associated wages and salaries though not the business incomes. However, it is BEA's view that overall, the missing source data for GDI is more of a problem than the missing source data for GDP.

In addition to the adjustments for tax return misreporting, adjustments are made to conform the accounting concepts underlying the source data to the accounting concepts underlying the NIPA's. The major NIPA-accounting adjustment to the estimates of GDP is the inventory valuation adjustment, which converts inventories valued at historical cost to replacement cost; however, errors in this adjustment do not significantly affect the difference between GDP and GDI, because a similar adjustment is also made to business incomes in GDI. For the estimates of GDI, there are many NIPA-accounting adjustments. BEA is concerned most about the accuracy of the adjustment that restates business incomes and depreciation to conform to the NIPA definition of investment, because the adjustment does not appear to correctly account for purchases of software. In the NIPA's, these purchases are treated as intermediate inputs; for tax return reporting, a large amount is treated as investment. Errors in the adjustment for these purchases would likely result in an overstatement of GDI that is large and growing.

BEA will continue to work to reduce the size of the statistical discrepancy, but it is highly unlikely that it can be eliminated completely; source data from sample surveys reflect sampling errors, and source data from complete enumerations reflect nonresponse errors. Currently, BEA is developing estimates of business purchases of software that are treated as investment in tax returns in order to improve the present adjustments for this type of difference; BEA also is evaluating a recent Census Bureau report on a potentially large understatement of the reported value of exports. As part of BEA's Strategic Plan for improving the national economic accounts, BEA is also working with other agencies to close several of the data gaps. These efforts include developing annual surveys of transportation, finance, insurance, and real estate; extending the annual wholesale trade survey to cover all inventories; improving the coverage of nonresidential reconstruction in the value-put-in-place survey; expanding the monthly establishment series to cover hours and earnings of all workers; developing complete and consistent surveys of fringe benefits; and speeding up the availability of the surveys on State and local governments.


1. For a discussion of this issue, see "Economic Report of the President" (Washington DC: U.S. Government Printing Office, 1997): 72–74.