News Release: National Income and Product Accounts Comprehensive Revision
Sign up for e-mail subscriptions.
EMBARGOED UNTIL RELEASE AT 8:30 A.M. EST, WEDNESDAY, DECEMBER 10, 2003 Brent Moulton: 606-9606 BEA 03-48 Carol Moylan: 606-9715 COMPREHENSIVE REVISION OF THE NATIONAL INCOME AND PRODUCT ACCOUNTS 1929 THROUGH SECOND QUARTER 2003 Today, BEA is releasing revised estimates of gross domestic product (GDP) and other national income and product accounts (NIPAs) series from 1929 through the second quarter of 2003. Comprehensive revisions, which are carried out about every 4 to 5 years, are an important part of BEA's regular process for improving and modernizing its accounts to keep pace with the ever-changing U.S. economy. Most of the tables in this release present revised estimates beginning with 1992; estimates beginning with 1929 are available on the BEA Web site (<www.bea.gov>) and will be published in the January 2004 issue of BEA's monthly journal, the Survey of Current Business. The picture of the economy shown in the revised estimates is very similar in broad outline to the picture shown in the previously published estimates. The similarity in outline and some of the differences in detail can be seen in the following: o For 1992-2002, the average growth rate of real GDP is 3.2 percent, the same as in the previously published estimates. Consumer spending and residential investment increased a little more than in the previously published estimates, but exports and government spending increased a little less. For the subperiod 1994-2000, the rate of growth is also the same as in the previously published estimates (3.8 percent). o Similarly, for 1992-2002, the average rate of change in prices paid by U.S. residents is the same as in the previously published estimates (1.8 percent). ===== BOX This release presents revised estimates through the second quarter of 2003. Revised estimates for the third quarter of 2003 will be released on December 23, 2003. Until that time, the 8.2-percent estimate for the rate of growth in real GDP for the third quarter of 2003 (released on November 25) remains BEA's best estimate. ===== o From the fourth quarter of 2000 to the third quarter of 2001, real GDP decreased 0.5 percent; in the previously published estimates, it decreased 0.6 percent. In the revised estimates, real GDP decreased slightly in the third quarter of 2000; in the previously published estimates, it had increased throughout the year. o The pace of the current expansion has been revised down slightly; from the third quarter of 2001 to the second quarter of 2003, the average growth rate of real GDP is revised from 2.7 percent to 2.6 percent. o Corporate profits is revised up substantially for 2002; profits of domestic industries as a percentage of gross domestic income is revised up from 6.3 percent to 7.1 percent. Improvements incorporated in this comprehensive revision Comprehensive revisions incorporate three major types of improvements: o Changes in definitions and classifications that update the accounts to more accurately portray the evolving U.S. economy, o Presentational changes that make the NIPA tables more informative and easier to use, and o Statistical changes that introduce new and improved methodologies and that bring in newly available and revised source data. The improvements incorporated in the estimates being released today have been previewed in a series of articles in the Survey and are available on BEA's Web site at: /national/cr2003content.htm Changes in definitions and classifications. The changes in definitions and classifications introduced in this comprehensive revision include the following: o The implicit services provided by property and casualty insurance are recognized, and the treatment of insured losses is improved; as a result, large swings in measured insurance services associated with catastrophic losses such as those of September 11th and of Hurricane Andrew are eliminated. o A portion of the implicit services of commercial banks is allocated to borrowers, thereby recognizing that both borrowers and depositors receive these services. o The definition of national income is broadened to include all net incomes (net of consumption of fixed capital) earned in production. For example, under the new definition, national income will include taxes on production and imports (formerly, indirect business taxes). Presentational changes. Noteworthy changes in presentation include the following: o The reference year for chain-type quantity and price indexes and for chained-dollar estimates has been updated from 1996 to 2000. o Many new tables show percent changes in GDP components, contributions to percent change, and relative shares of GDP and of gross domestic income. o A new organization and numbering system is introduced for the NIPA tables that groups together the different measures for a given set of components and that allows the addition of new tables more easily. o Industry estimates are based on the North American Industrial Classification System (NAICS), which deals well with new and emerging industries. Statistical changes: Improved methodologies. Several important methodology changes are also incorporated in this revision and include the following: o The estimates of used motor vehicles reflect new methods and source data for estimating net transactions and total dealers' margins. o The estimates of personal consumption expenditures for hotels and motels reflect new source data on U.S. residents' lodging expenditures by sector and nonresidents' spending on lodging. o The estimates of real nonresidential structures and of photocopy equipment reflect the use of new quality-adjusted price indexes. o The estimates of corporate profits for recent periods for which Internal Revenue Service tabulations of corporate tax returns are unavailable reflect a new adjustment for the exercise of stock options. Statistical changes: New and revised data. The revised estimates also reflect the incorporation of newly available and revised source data. The most important source data that affect the current-dollar and "real" estimates prior to 2000 are the following: BEA's benchmark 1997 input-output (I-O) accounts; data on inventories, on receipts and expenses of business establishments and of governments, on sales by detailed commodity and by merchandise line, and on final industry and product shipments from the 1997 Economic Census, and on trade margins from both the 1997 Economic Census and the 1997 annual surveys of merchant wholesale and retail trade; and data on housing units from the 2000 decennial Census of Housing. In addition, estimates that are based on BEA's international transactions accounts (ITAs) -- primarily net exports of goods and services and rest-of-the-world income receipts and payments -- reflect improvements to the ITAs that have been introduced since 1999. Estimates of private structures reflect Census Bureau revisions to the surveys of value of construction put-in place data that go back to 1978. Other data that were incorporated include: Data on expenditures and receipts of state and local governments for fiscal years 1997-99 from the Census Bureau; final data on employer pension and profit-sharing plans for 1995-98 from the Department of Labor; revised data on mortgage debt outstanding beginning with 1982, and on consumer debt outstanding and the effective rate of interest on consumer debt outstanding beginning with 1980, from the Federal Reserve Board; and new Census Bureau "exact-match" files based on tabulations of Internal Revenue Service (IRS) individual tax returns for 1996 and 1999. The revised estimates for 2000 forward also reflect the incorporation of newly available and revised source data that became available since the last annual NIPA revision in July 2002. The most important of these data are the following: Census Bureau data from annual surveys of state and local governments (for fiscal years 2000 (final) and 2001 (preliminary)), of manufactures, of merchant wholesale trade, and of retail trade (for 2000 (revised) and 2001 (preliminary)); Census Bureau data from the services annual survey and from the surveys of the value of construction put- in-place for 2001 and 2002; Census Bureau data from the American Housing Survey for 2001; federal government budget data (for fiscal years 2002 and 2003); ITA data for 2000-02 (revised); Bureau of Labor Statistics (BLS) tabulations of wages and salaries of employees covered by state unemployment insurance for 2001 and 2002 (revised); newly available IRS tabulations of corporate tax returns for 2000 (final) and 2001 (preliminary); U.S. Department of Agriculture (USDA) farm statistics for 2002; and newly available IRS tabulations of sole proprietorship and partnership tax returns for 2001. In the aggregate, changes in definitions and classifications lowered current-dollar GDP, and statistical changes (improved data and methodologies) raised GDP, as shown in the following chart for 1992-2002. News release tables. This release includes most of the tables that will regularly be shown in future GDP news releases; in addition, special tables have been included to highlight the effects of the comprehensive revision. The special tables are: o Tables 1A, 2A, and 4A, which show revised and previously published estimates for percent changes in real GDP, contributions to percent change in real GDP, and percent changes in price indexes, respectively; o Table 1B, which shows revisions to current-dollar GDP and selected income components; and o Tables 7A-7C, which show annual levels, percent changes, and revisions to percent changes for current-dollar GDP, for real GDP, and for price indexes for GDP, respectively. Most of the tables show annual estimates beginning with 1992; quarterly estimates (if shown) begin with the first quarter of 1998. Three of the regular tables -- tables 3, 11, and 12 -- are split into A and B segments in this release in order to accommodate this longer-than-usual time span. The regular news release tables on real GDP percent change from quarter one year ago and on gross value added of nonfinancial corporate business, as well as appendix table A are not yet available; they will be posted on BEA's Web site in a few days as part of the regular NIPA tables. The revisions For this comprehensive revision, most current-dollar series are revised back to 1993, and many are revised back to 1929, the earliest year for which the NIPA estimates are available. Real GDP growth. For 1929-2002, the average annual rate of growth of real GDP, is 3.4 percent, the same as in the previously published estimates. Over the shorter period, 1992-2002, the growth rate is 3.2 percent in both the revised and previously published estimates. For the 1994-2000 subperiod, the growth rate is also the same in both the revised and previously published estimates (3.8 percent). Since 1992, personal consumption expenditures (PCE) and residential fixed investment grew somewhat faster in the revised estimates, and exports of goods and services and government consumption expenditures and gross investment grew somewhat slower. The revisions to year-to-year real GDP growth are generally less than 0.1 percentage point between 1959 and 1971, but more sizable thereafter. The largest upward revision before 1959 is 0.3 percentage point for 1940; downward revisions of 0.2 percentage point are recorded for 3 years (1937, 1947, and 1952). The largest upward revision after 1959 is 0.4 percentage point for 1999, and upward revisions of 0.3 percentage point are recorded for 1985, 1991, and 1992; the largest downward revision after 1959 is 0.3 percentage point recorded for 1976. For the 3 most recent years, growth is revised as follows: o down from 3.8 percent to 3.7 percent for 2000, o up from 0.3 percent to 0.5 percent for 2001, and o down from 2.4 percent to 2.2 percent for 2002. Business cycles. From the fourth quarter of 2000 to the third quarter of 2001, real GDP decreased 0.5 percent; in the previously published estimates, it decreased 0.6 percent. In the revised estimates, real GDP also decreased slightly in the third quarter of 2000. The pace of the current expansion has been revised down slightly; from the third quarter of 2001 to the second quarter of 2003, the growth rate of real GDP is revised from 2.7 percent to 2.6 percent. Earlier business cycles show very little revision. Price changes. For 1929-2002, the average annual increases in both the price index for gross domestic purchases and the price index for GDP are 3.0 percent, the same as in the previously published estimates. For 1992-2002, the average annual increase in the price index for gross domestic purchases is 1.8 percent, the same as in the previously published estimates; the price index for GDP is 1.9 percent, also the same as in the previously published estimates. Real disposable personal income (DPI) growth. For 1929-2002, the average annual increase in real DPI is 3.3 percent, the same as in the previously published estimates. For 1992-2002, the average annual increase in real DPI is 3.2 percent; 0.1 percentage point more than in the previously published estimates. Personal saving. The revisions are relatively small through 1986. Beginning with 1987, the revisions are larger and all downward. The downward revisions reflect upward revisions to PCE and downward revisions to personal income (discussed below). The downward revisions to personal saving result in corresponding downward revisions to the personal saving rate -- personal saving as a percentage of DPI -- that range from 0.2 percentage point for 1999 to 1.4 percentage points for 2002. Revised current-dollar estimates The revisions to current-dollar GDP and to selected measures of income are shown in table 1B. This table shows the "revisions in level," that is, the revised estimates less the previously published estimates, and the revisions as a percent of the previously published estimates for selected years. GDP. Since 1959, GDP is revised down for 31 years and revised up for 13 years. Over the entire period, GDP is generally reduced by the reallocation of a portion of the implicit services of commercial banks from depositors to borrowers; relatively large reductions begin with 1987. (Services imputed to business borrowers are treated as intermediate purchases and, therefore, are not part of GDP.) In contrast, GDP is raised by the recognition of the implicit services provided by the investment income of property and casualty insurance companies. Generally, the reduction in GDP resulting from the new treatment of banking services outweighs the increase in GDP resulting from the new treatment of insurance services. Excluding all the changes in definitions and classifications, current-dollar GDP is revised down by small amounts for 1959-80 (except for 1979 when there is a small upward revision) and revised up beginning with 1981. The revisions tend to increase over time; the largest is an upward revision of about 1 percent of GDP for 2002. PCE. Revisions to PCE are relatively small through 1988; PCE for services is generally the main source of the revisions. From 1989 forward, PCE is revised up. PCE for durable goods is the main source of the upward revisions through 1999; for 1995-99, there are partly offsetting downward revisions to PCE for nondurable goods. For 2000-02, upward revisions to durable goods and to services are partly offset by downward revisions to nondurable goods. For the entire period, the revisions reflect changes in definitions and classifications; until 1982, the revisions attributable to new source data and methodologies are small. o Services. For 1978 forward, downward revisions to services resulting from the new treatment of banking services more than offset upward revisions resulting from the new treatment of insurance services. Banking and insurance are not the only notable sources of revision to PCE for services. Beginning with 1960, PCE for religious and welfare activities is revised down, reflecting the improved allocations to PCE commodity categories made possible by NAICS. Beginning with the 1980s, PCE for housing is revised up, reflecting, in part, improved data on lodging at hotels and motels. Also beginning with the 1980s, PCE for transportation services is revised up, reflecting, in part, the new treatment of motor vehicle insurance services. PCE for medical services is revised up (and the revisions increase over time), reflecting the recognition of the implicit services provided by health insurance companies, the improved allocations made possible by NAICS, as well as the incorporation of newly available and revised source data. o Durable goods. Notable upward revisions to PCE for durable goods begin with 1987. For 1987- 90, the revisions mainly reflect improved estimates of net transactions in used light trucks. (The improvement involves the use of average auction prices to value the change in unit stocks.) For 1991 forward, a similar methodology has been applied to the estimation of net transactions in used autos. Beginning with 1994, upward revisions to furniture and equipment primarily reflect the incorporation of results from the 1997 I-O table. o Nondurable goods. Upward revisions for 1988-93 largely reflect revisions to gasoline and oil. Beginning with 1995, downward revisions are more than accounted for by food and by clothing and shoes; the revisions mainly reflect the incorporation of results from the 1997 I-O table. Private fixed investment. The revisions begin with 1978 and are relatively small through 1996, as downward revisions to nonresidential fixed investment are generally offset by upward revisions to residential fixed investment. For 1997-2000, the downward revisions to nonresidential outweigh the upward revisions to residential. For 2001 and 2002, the revisions are almost offsetting. The downward revisions to nonresidential fixed investment primarily reflect revisions to equipment and software; revisions to nonresidential structures are generally smaller. o Nonresidential structures. Upward revisions occur for 1978-92 and downward revisions occur for 1993 forward (except for 1995). The revisions (except for 1997 and 1998) are relatively small and primarily reflect revisions to Census Bureau estimates of the value of construction put-in- place and the incorporation of data from the 1997 I-O accounts. o Equipment and software. Downward revisions for 1978 forward start out very small but become large. For 1978-86, the revisions mainly reflect improvements to the measurement of computer software that were introduced in the 1997 I-O accounts. For 1987-2002, the revisions reflect improvements to the measurement of used trucks and of computer software. o Residential fixed investment. Upward revisions for 1983 forward start out very small but become large and reflect an improved methodology for estimating residential improvements and revised Census Bureau data on the value of construction put-in place. Change in private inventories. The revisions begin with 1991 and are small through 1996, primarily reflecting the inclusion of farm materials and supplies in the definition of private farm inventories. Beginning with 1997, a mixed pattern of revisions continues, but the magnitudes are generally larger; nonfarm inventories dominate these revisions, though for 2002, an upward revision to nonfarm inventories is largely offset by a downward revision to farm inventories. The revisions to nonfarm inventories primarily reflect newly available and revised Census Bureau data on inventory book values, the incorporation of new commodity price weights from the 1997 I-O accounts, and revised producer prices. Exports and imports of goods and services. Revisions to net exports of goods and services are small through 1987 -- at first upward and then downward. Thereafter, revisions are downward for all years; they tend to grow larger over time, but the revision for 2002 is small. o Exports. Through 1988, revisions are relatively small and generally upward; thereafter, larger downward revisions occur. For most years, the revisions are mainly accounted for by services (2002 is an exception, in that goods account for most of the revision). The revisions to exports of services mainly reflect the new treatments of insurance and of banking services, as well as a reclassification of military grants from federal government consumption expenditures to exports of services. o Imports. Upward revisions begin with 1986 and tend to be small through 1997; imports is revised down slightly for 1998. Larger upward revisions occur for 1999-2001, and a downward revision occurs for 2002. The new treatment of insurance services contributes to the upward revisions throughout the period; for 2002, the incorporation of revised estimates from the ITAs more than offsets the contribution from this definitional change. Government consumption expenditures and gross investment. Revisions prior to 1959 are very small. Beginning with 1959, the revisions for all but 5 years are downward. Through 1996, the revisions generally remain small (except for 1986, when the new treatment of banking services more than accounts for a large upward revision); for 1997-2002, the revisions are much larger. Downward revisions for 1959-84 are primarily accounted for by the federal government, although state and local government also contributes. Through 1975, the revisions to federal spending primarily reflect the reclassification of military grants. Beginning with 1972, new source data and methodologies become important and contribute to the downward revisions; these revisions are mainly to consumption of general government fixed capital, reflecting improved price estimates for military equipment and, beginning with the 1990s, revised estimates of software investment. As noted, the new treatment of banking services also contributes to the revisions, most notably beginning with 1985, sometimes raising spending, other times reducing spending. Beginning with 1997, state and local spending accounts for most of the downward revisions; the revisions for 1997 and 1998 primarily reflect the incorporation of newly available medical insurance data for government employees, and the revisions for 1999-2002 primarily reflect the incorporation of Government Finances data. Personal income. For 1929-81, downward revisions to personal income mainly reflect revisions to personal interest income; through 1977, the revisions are small -- proprietors' income contributes to the downward revisions, and rental income of persons moderates them. For 1982-86, the revisions to personal income are upward, and again personal interest income is the main contributor. For 1987-97, downward revisions to personal income are accounted for by downward revisions to personal interest income, to supplements to wages and salaries, and to proprietors' income; they are moderated by upward revisions to rental income of persons. For the 5 most recent years, the revisions to personal income can be summarized as follows: o 1998: A small downward revision is accounted for by personal interest income and by wages and salaries; it is moderated by supplements. o 1999: An upward revision is mainly accounted for by supplements and by personal dividend income; it is moderated by personal interest income and by wages and salaries. o 2000: An upward revision is mainly accounted for by supplements, by personal current transfer receipts, and by proprietors' income; it is moderated by personal interest income and by wages and salaries. o 2001: An upward revision is mainly accounted for by supplements, by proprietors' income, by rental income, and by personal current transfer receipts; it is moderated by personal interest income, by dividend income, and by wages and salaries. o 2002: A downward revision is mainly accounted for by personal interest income, by personal dividend income, and by wages and salaries; it is moderated by supplements, by proprietors' income, and by rental income. The revisions to the components of personal income are discussed below. o Wage and salary disbursements. Through 1988, revisions are upward and small. Beginning with 1989, the revisions are downward and become more notable beginning with 1994, reflecting downward revisions to private wages and salaries that are partly offset by upward revisions to government wages and salaries. A large downward revision for 2002 primarily reflects the more complete incorporation of tabulations of wage and salary data from the BLS for private and state and local government employees who are covered by state unemployment insurance. o Supplements to wages and salaries. Through 1986, revisions are small and downward; they remain downward but become larger through 1995. Beginning with 1996, the revisions are upward, becoming large beginning with 1997. The revisions reflect the pattern of revisions to employer contributions for pension and insurance funds. Upward revisions to employer contributions for pensions begin with 1989, reflecting methodology improvements and more complete source data, including more complete pension data from the Department of Labor for 1988-98. Downward revisions to employer contributions for health insurance for 1987-96 reflect an improved methodology that incorporates the results of the annual medical expenditure panel survey; beginning with 1998, upward revisions reflect newly available source data. o Proprietors' income. Through 1997, most revisions are downward. For 1998, a small upward revision occurs, and for 1999, a slight downward revision occurs, followed by large upward revisions for 2000-02. Through 1983, the revisions generally reflect downward revisions to both farm and nonfarm proprietors' income. Beginning with 1984, the revisions primarily reflect the pattern of revisions to nonfarm proprietors' income. The revisions to farm proprietors' income reflect the reclassification of farm housing services from proprietors' income to rental income of persons and, for recent years, newly available data from the USDA. Prior to 2001, the revisions to nonfarm proprietors' income primarily reflect a number of statistical improvements. The large revisions for 2001 and 2002 also reflect the incorporation of newly available IRS tabulations of sole proprietorship and partnership tax returns for 2001. o Rental income of persons. Through 1977, the revisions are upward and small, after which the pattern of revisions is mixed although the magnitudes remain relatively small through 1991. Beginning with 1992, the revisions are mostly upward with particularly large upward revisions for 2001 and 2002. The reclassification of farm housing services contributes to the upward revisions throughout the period. In addition, the revisions for 2001 and 2002, reflect the incorporation of newly available source data from the Census Bureau's American Housing Survey. o Personal interest income. The revisions are relatively small and mostly downward through 1977. Beginning with 1978, the revisions become larger and remain downward (except for 1982-86). For most years, the revisions reflect generally downward revisions attributed to the new treatment of banking services that are partly offset by upward revisions attributed to the new treatment of insurance services. The revisions also reflect the incorporation of revised and newly available source data for estimating net interest and miscellaneous payments (see below) and data on consumer debt outstanding from the Federal Reserve Board. o Personal dividend income. The revisions are relatively small through 2000 (except for 1999, which shows a notable upward revision); the revisions primarily reflect newly incorporated data from the ITAs on dividends from the rest of the world. For 2001 and 2002, large downward revisions primarily reflect the incorporation of newly available IRS tabulations of corporate tax return data for 2001 and data from company financial statements. o Personal current transfer receipts. Through 1990, the revisions are generally small and upward. For 1991-98, the revisions are downward, and for 1999-2002, they are upward. The revisions mostly reflect the pattern of revisions to "other current transfer receipts, from business (net)" (formerly "business transfer payments"). For 2002, the revision also reflects an upward revision to old-age, survivors, disability, and health insurance benefits, reflecting newly available data on Medicare benefits, and downward revisions to government unemployment insurance benefits and to state and local Medicaid benefits. o Contributions for government social insurance. The revisions for contributions for government social insurance (which is deducted in the calculation of personal income) are small throughout the period. o Personal current taxes. This component consists of tax components that were included in the former aggregate "personal tax and nontax payments"; however, the nontax components are reclassified as personal current transfer payment to government. Therefore, the pattern for the new series is similar to that of the previously published series, although the revised levels are somewhat lower. o Disposable personal income (DPI). The pattern of revisions to disposable personal income reflects the revisions to personal income and to personal current taxes. Through 1974, the revisions to DPI are very small; beginning with 1975, there is a mixed pattern: Upward revisions for 1975 and 1976, downward revisions for 1977-81, upward revisions for 1982-86, downward revisions for 1987-95 (except for 1989), and upward revisions for 1996-2002. o Personal outlays. This series now consists of PCE, personal interest payments, and personal current transfer payments. The revision to personal outlays, primarily reflects the revisions to PCE that were previously described. In addition, the revised levels of the series reflect the inclusion of personal current transfer payments to government, which was formerly classified as personal nontaxes. National income. The revisions are upward for all years and grow progressively larger over time. They mainly reflect the redefinition of national income to include nonfactor incomes, notably taxes on production and imports, which grew from $6.8 billion for 1929 to $760.1 billion for 2002. Corporate profits with inventory valuation and capital consumption adjustments. Revisions are generally upward and small through 1980, growing larger thereafter; the revision for 2002 is very large. Through 1999, the revisions largely reflect the pattern of revisions to the capital consumption adjustment. The revisions for 2000-02 reflect revised 2000 and newly available 2001 IRS tabulations of corporate tax returns. In addition, the revision for 2002 reflects a new stock-options adjustment that offsets a timing problem between the treatment of nonqualified stock options in reported wages and salaries and the treatment of exercised stock options in the corporate profits data. Profits from the rest of the world are revised beginning with 1982; the revisions are relatively small and generally upward through 2000, with larger upward revisions for 2001 and 2002. Net interest and miscellaneous payments. The revisions are very small until the mid-1970s, when they become more notable. The pattern of revisions is mixed through 1986; beginning with 1987, the revisions are downward and generally become larger. The overall revisions reflect mostly downward revisions attributable to the new treatment of banking services that are partly offset by upward revisions attributable to the new treatment of insurance services. The revisions also reflect the incorporation of revised and newly available source data from the Federal Reserve Board on mortgage debt outstanding, from BEA's ITAs, and, beginning with 2000, IRS tabulations of business tax returns. Consumption of fixed capital (CFC). The revisions to the NIPA measure of depreciation are downward for most years. They are generally small through 1980 and become larger thereafter. The revisions reflect revised investment estimates and three statistical improvements: The incorporation of new service lives for private aircraft, separate estimation of CFC for light trucks, and a revised depreciation schedule for autos. Statistical discrepancy. For 1959-77, revisions to GDP and to gross domestic income (GDI) are in the same direction and of similar magnitude; thus, there is little effect on the statistical discrepancy, which is the difference between GDP and GDI. (In theory, GDP should equal GDI; in practice, they differ because their components are estimated using largely independent and less-than-perfect source data.) For 1978-86, the revisions to GDP and to GDI continue to be in the same direction but the magnitudes differ more; thus, for some of these years, the statistical discrepancy is revised up, for others, it is revised down (the revisions to the discrepancy remain relatively small). Beginning with 1987, the statistical discrepancy is revised up (becoming more positive or less negative): For 1987-90, GDI is revised down more than GDP; for 1991-94, GDP is revised up, and GDI is revised down; for 1995-2000 (except for 1996), GDI is again revised down more than GDP; and for 2001 and 2002, GDP is revised up more than GDI. ======