Improved estimates of gross product by industry, 1959-94.

Yuskavage, Robert E.

In this article, the Bureau of Economic Analysis (BEA) presents new estimates of gross product, or gross product originating (GPO), by industry for 1994 and revised estimates for 1959-93.(1) The estimates reflect the results of the recent comprehensive revision of the national income and product accounts (NIPA'S), and they incorporate newly available source data and methodological changes for GPO by industry.(2)

The following major improvements are incorporated into this GPO revision and were also incorporated into the comprehensive NIPA revision. Improved chain-type measures of real GPO that eliminate the overstatement of real growth for periods after the base year and the understatement of real growth for periods before the base year, a new treatment of government investment that provides a more complete picture of investment through the consistent treatment of fixed assets whether purchased by the public or the private sector, and a quality-adjusted BEA price index for selected semiconductor products.(3) Other major improvements include the following: An improved industry allocation of commodity taxes, a newly available 1987 employment matrix for the estimation of profits and capital consumption on an establishment basis, and newly available information on the composition of inputs from the 1987 benchmark input-output table and the 1992 Economic Censuses. In addition, new and redesigned tables are introduced to up-date the presentation of the GPO estimates to reflect the improved measures of real output and to provide detail on the cost components Of GPO.

The release of the new and improved GPO estimates is the latest step in a continuous GPO improvement program that BEA initiated in 1988.(4) Earlier improvements included improved estimation techniques for most services-producing industries, the incorporation of services prices and import prices into the estimation of real inputs, and the resumption of an annual publication schedule. Future improvement efforts will focus on integrating the GPO series with the benchmark input-output accounts and with other BEA industry estimates.(5)

The first part of this article discusses the relative performance of various industries in terms of growth rates and industry shares. The second part discusses the revisions to the GPO estimates, and the third part describes the methodology used to prepare the GPO estimates. Detailed tables following the text present the current-dollar GPO estimates for 1959-94 and the real estimates for 1977-94.

Industry Growth Rates and Shares

Comparisons of gross product growth rates and shares of gross domestic product (GDP) across industries provide indications of the relative performance of particular industries or industry groups. For example, a comparison of the growth rate of real gross product for an industry with the growth rate of real GDP indicates whether that industry is adding to (or is reducing) the economy's growth and whether that industry is becoming a larger (or a smaller) part of the total economy. A comparison of the share of current-dollar GDP accounted for by the gross product of an industry over time indicates whether that industry's claim on the economy's resources is increasing (or decreasing).

Real growth rates

Real GDP increased at an average annual rate of 2,6 percent for 1977-94 (chart 1 and table 1). The gross product of all industry groups increased over this period, the increases ranged from percent for wholesale trade to 0.9 percent for mining. Manufacturing increased 2.3 percent, 0.3 percentage point less than the increase in GDP; durable goods increased 2.5 percent, 0.1 percentage point less than GDP.

[TABULAR DATA OMITTED]

Growth rates for 1977-94 for more detailed industry groups are shown in table 15.(6) Among industries with comparable definitions over the entire period, nine recorded average annual increases in real gross product of 5 percent or more. The two fastest growing industries were security and commodity brokers, which increased 10.5 percent, and agricultural services, forestry, and fishing, which increased 7.2 percent. Other fast growing industries included coal mining, industrial machinery and equipment, rubber and miscellaneous plastics products, transportation by air, transportation services, motion pictures, and social services.(7) Increases of more than 5 percent were also recorded in two industry combinations: Communications (which consists of the telephone and telegraph and the radio and television industries) and "business, miscellaneous professional, and other services."

The growth rates of nine industries decreased for 1977-94. The three largest decreases were in manufacturing. Tobacco products, down 5.1 percent; leather and leather products, down 2.4 percent; and primary metal industries, down 1.1 percent. The decreases in the remaining six industries - two in manufacturing, two in transportation and public utilities, one in mining, and one in services - were all less than 1 percent.

Average annual growth rates for the subperiods 1977-87 and 1987-94 are also shown in table 15.(8) For 1977-87, real GDP increased at an average annual rate of 2.8 percent. The growth rates of all industry groups increased during this period; the largest increase was in wholesale trade (4.8 percent). Manufacturing increased 2.7 percent, 0.1 percentage point less than the increase in GDP.

For 1987-94, real GDP increased 2.3 percent. The growth rates of all industry groups increased; the largest increase was in wholesale trade (4.9 percent). Manufacturing increased 1.6 percent, 0.7 percentage point less than the increase in GDP. Nondurable goods increased 0.9 percent, largely reflecting a decline in petroleum and coal products; durable goods increased 2.2 percent. In contrast. for 1992-94, manufacturing grew at an annual rate of 4.8 percent, compared with a 2.8-percent increase in GDP, durable goods increased 7.1 percent.

Contributions to real GDP growth. - Differences in growth rates alone do not indicate the extent to which industries contribute to the growth of real GDP; their contribution also depends on the industry's size in the first year of the period being measured. Table 2 shows contributions by industry groups to real GDP growth for each year in 1978-94.(9) For example, real GPO for mining increased 6.6 percent in 1994, 3.1 percentage points more than the increase in GDP. However, because mining is a relatively small industry group, it contributed only 0.1 percentage point to the growth in GDP.

[TABULAR DATA OMITTED]

Shares of current-dollar GDP

For 1959-94, the share of current-dollar GDP accounted for by private services-producing industries increased from 48.8 percent to 62.0 percent, while the share accounted for by private goods-producing industries declined from 38.8 percent to 24.2 percent (table 3 and chart 2).(10) The increase in the share for the private services-producing industries is the result of both above-average real growth and above-average growth in prices. The increase was more than accounted for by "services" and by finance, insurance, and real estate (FIRE), whose shares rose 9.9 percentage points and 4.8 percentage points, respectively. Among the industries in these groups, the increase was broad-based, but is particularly noticeable in banking, business services, and health services (table 11).

[TABULAR DATA OMITTED]

The decline in the share of goods-producing industries was concentrated in manufacturing. Manufacturing's share fell 10.4 percentage points, from 27.7 percent to 17.3 percent. Within manufacturing, the share of durable goods declined 6.4 percentage points, and the share of nondurable goods declined 4.0 percentage points.

The share of government increased from 12.8 percent to 13.4 percent. The increase was more than accounted for by State and local government; the share of the Federal Government declined.

Composition of current-dollar GPO. - Current-dollar GPO is measured as the incomes and profits earned in production in each industry; it is equal to gross domestic income, whose components can be divided into categories that approximate returns to labor and returns to capital. Differences over time and among industry groups in returns to labor and capital can thus be observed using these approximations. The return to labor from production can be approximated using compensation of employees. The return to capital from production can be approximated using the remaining components of GPO except indirect business tax and nontax liability, which is excluded because it can be viewed as a part of the pre-tax return to capital that accrues to government rather than to business.(11)

For the economy as a whole, compensation of employees as a share of GDP increased slightly, from nearly 56 percent in in 1959 to nearly a 58 percent in 1944, while the share of "other" GPO declined (table 12). Labor and capital shares varied among industry groups. The labor share of GPO in manufacturing consistently exceeded the average for the total economy, but it declined slightly over the period, largely because the share in nondurable goods declined. The labor share of GPO in both FIRE and "services" increased substantially, while the capital share, mainly proprietors' income, in those industries fell.

Revisions to the GPO Estimates

The first section in this part of the article discusses the impact of the revisions to the GPO estimates. The second section discusses the major sources of the revisions, and the third section discusses the changes in the presentation of the GPO estimates.

Impact of the revisions

Current-dollar estimates. - The pattern of the revisions to current-dollar GPO largely reflects the pattern of the comprehensive NIPA revisions to GDP and of the revisions by industry to the components of gross domestic income. Table 4 presents revisions to current-dollar GDP by industry group for selected years.

[TABULAR DATA OMITTED]

For 1959-82, the GPO of most private industries was revised very little. The exceptions are the revisions to nondurable goods manufacturing (mainly petroleum and coal products) and to wholesale trade, which resulted from the improved treatment of commodity taxes. For 1983-93 the largest upward revision was to FIRE (mainly due to nonfarm housing services) beginning in 1984. Other large upward revisions were to mining (mainly to oil and gas extraction) for each year 1984-92 and to wholesale trade, beginning in 1991. The largest downward revisions were to agriculture, forestry, and fishing (mainly to farms) beginning in 1988; to retail trade beginning in 1986 and to nondurable goods manufacturing (mainly to petroleum and coal products) for most years.

For government, revisions were relatively large for all years, reflecting the new treatment of government investment and, beginning in 1984, the new definitions for the Federal Government retirement programs. General government GPO was revised up for each year during 1959-93.

Real growth rates. - For 1977-93, revisions to the rates of growth in real GPO did not alter the picture of growth by industry that was shown by the previously published estimates (table 5). After revision, wholesale trade remains the fastest growing industry group. Previously, construction was the slowest growing industry group; now, construction and mining are tied for the slowest.

[TABULAR DATA OMITTED]

The average annual growth rate for manufacturing was revised down 0.1 percentage point for 1977-93, the same as the revision for GDP.(12) For durable goods, the real growth rate was revised down 0.5 percentage point. For 1977-87, the growth rate for manufacturing was revised up 0.3 percentage point, and for 1987-93, it was revised down 0.9 percentage point. This downward revision is largely attributable to revisions to the preliminary estimates for 1993 which were based on incomplete source data, and to the use of the chain-type measure of real growth for manufacturing, beginning with 1988.

Chain-type measures of real growth are more appropriate than fixed-weighted measures because they avoid the substitution bias that results from changes in relative prices as one moves farther from a fixed base year. This substitution bias is especially apparent in manufacturing, because the continuous decline in computer prices has had a major impact on relative prices in manufacturing. In the previously published estimates, the use of the benchmark-years-weighted measure for manufacturing had a significant effect on the estimate of the growth in real manufacturing GPO for 1977-87.(13) In the revised estimates, as previously noted, the use of the chain-type measure in place of the fixed-weighted measure for 1987-93 was a major factor in the downward revision of the growth rate in manufacturing for that period.

Shares of current-dollar GDP. - The industry shares of current-dollar GDP were not much affected by the revisions. However, government's share was raised each year during 1959-94; the amount of the revision ranged from a high of 2.3 percentage points in 1959 to a low of 1.4 percentage points in 1993. For each year during 1959-75, manufacturing's share was revised down by about 1 percentage point; during 1976-93, it was revised down by smaller amounts, but by no less than 0.4 percentage point, The drop in the manufacturing share was primary due to downward revisions to the share of petroleum and coal products.

Major definitional and statistical changes

The revisions to the GPO estimates arise from incorporating the definitional and statistical changes introduced in January 1996 in the comprehensive NIPA revision and from statistical changes introduced in this GPO revision.

NIPA revisions. - The comprehensive NIPA revision released in January included definitional, statistical, and other changes that affect the GPO estimates. The most important of these changes was the introduction of the annual chain-type quantity index as the featured measure of real growth. In the GPO revision, chain-type measures were introduced for each of the 51 GPO industries for which the "double-deflation" method is used to compute real GPO.(14) For the 15 industries for which real GPO is computed using other methods, real GPO was calculated by linking two fixed-weighted quantity indexes; 1987 weights were used for 1977-87, and 1992 weights were used beginning with 1988.

The revised GPO series also incorporates the new NIPA treatment of government investment and the definitional change to Federal Government compensation of employees. Recognition of government expenditures for structures and equipment as fixed investment results in the inclusion of the services of government fixed assets - measured as depreciation, or consumption of fixed capital - in GDP and general government GPO for all years.(15) In the previously published estimates, general government GPO was defined to consist only of compensation of employees.

The definitional change to Federal Government compensation of employees affects GPO for both general government and government enterprises. For civilian retirement programs, contributions beginning with 1969 now include payments to the Civil Service Retirement Fund for interest and unfunded liability. For military retirement programs, contributions beginning with 1984 will now be the actual contributions to the fund. Previously, the value of these contributions had been "imputed" to equal the value of the benefits paid.

Several other major statistical changes to the NIPA's also affected the GPO estimates. The incorporation of the newly available data from the 1991 Residential Finance Survey on rental payments and on the value of tenant- and owner-occupied units led to large upward revisions to both gross output and GPO of nonfarm housing services, and the improved adjustments for misreporting on tax returns substantially increased the GPO of several services industries. Other NIPA changes that affected the gross output estimates used in the double-deflation method of estimating real GPO included the following: Revised estimates of petroleum and natural gas exploration, which affected the oil and gas extraction industry; revised estimates of both residential and nonresidential construction, which affected the construction industry; and revised estimates of personal consumption expenditures, which affected several financial and services industries.

GPO statistical changes. - A number of statistical changes specific to the GPO estimates were also introduced.

Commodity taxes. - The most important of the statistical changes incorporated into the GPO estimates was the improved treatment of commodity taxes beginning with 1959. Commodity taxes - which are taxes that vary with the consumption, production, or sale of products - are included in industry gross output. These taxes are part of the "indirect business tax and nontax liability" component of GPO.

The most important change in the treatment of commodity taxes, which affected both current-dollar and real industry GPO measures, was the improved industry assignment of specific Federal Government excise taxes. The new assignments more closely reflect the regulations and practices that determine which industry is liable for tax collection during the periods the taxes were imposed. Of these new assignments, the one with the largest impact is the shift of the Federal excise tax on gasoline and gasohol from petroleum and coal products in manufacturing to wholesale trade, beginning with 1959.(16) This change in industry assignment recognizes that the tax is due when fuel is withdrawn from the terminals, which are classified in wholesale trade, whether the terminals are owned independently or by an integrated petroleum company.

The real measures of ad valorem commodity taxes assigned to manufacturing industries were improved by holding constant both commodity prices and tax rates from the base period. In the previously published estimates, the procedure held constant only tax rates and yielded the correct result only for quantity-based commodity taxes.

Semiconductor prices. - For the comprehensive NIPA revision, BEA prepared quality-adjusted annual price indexes for memory and for micro-processor metal-oxide semiconductor integrated circuits (chips) for 1974-94. These indexes, which were incorporated into the NIPA estimates of exports and imports of semiconductors, have also been incorporated into the GPO estimates of real industry gross output and real intermediate inputs, beginning with 1977.

For gross output, the new price indexes were weighted together with appropriate producer price indexes from the Bureau of Labor Statistics to develop a composite deflator to cover all products of the semiconductor manufacturing industry. For intermediate inputs, the same composite index was used for the purchased by other industries of domestically produced semiconductors; the NIPA import price index was used for imported inputs. The incorporation of these new price indexes resulted in upward revisions to the growth rates for both semiconductor output and inputs, especially after 1922; the most noticeable effect was to raise real gross output and real GPO in the electronic and other electric equipment industry for 1992-93.

Employment matrix - For the current-dollar GPO estimates, a newly available Census Bureau employment matrix for 1987 that converts the NIPA industry estimates of corporate profits and capital consumption allowances from a company basis to an establishment basis was introduced. The new matrix is based on data from the 1987 Economic Census and covers all private nonfarm industries except private households. A matrix based on 1982 Economic Census data had been used in the previously published estimates. The new matrix was used to revise the profits and depreciation estimates beginning with 1983.

Changes affecting real GPO. - Revisions to the estimates of real GPO reflect the previously discussed changes to current-dollar GPO and to real gross output and intermediate inputs, as well as the shift from benchmark-years-weighted and fixed-weighted measures to chain-type measures. In addition, the revisions reflect other changes to gross output and intermediate inputs and changes in the methods used for estimating the composition of inputs.

Revisions to gross output primarily resulted from benchmarking to the final - rather than to the preliminary - levels of the 1987 input-output (I-O) table and from incorporating the preliminary gross output levels from the 1992 I-O table, which are largely based on data from the 1992 Economic Censuses. Benchmarking to the 1987 I-O table also affected the estimates of gross output for 1983-86. In addition, beginning with 1988, new and revised annual survey data from the Census Bureau were used to interporlate between 1987 and 1992 and to extrapolate from 1992.

New and improved estimates of the composition of inputs were introduced for all "double-deflated" industries. The revised estimates incorporate the input composition from the final 1987 benchmark I-O table, which affects input compositions beginning with 1983, and data from the 1992 Economic Censuses on the cost of materials, fuels, and energy for selected industries. In the previously published estimates, input compositions for 1987 were from a preliminary I-O table. The revised estimates for 1983-86 are primarily interporlations based on the 1982 and 1987 input compositions; for the revised estimates for 1988-94, the input composition for all double-deflated industries is generally assumed to be the same as that for 1987.(17)

Presentational changes

The following changes have been made to the presentation of the GPO estimates: China-type measures of real GPO are shown in chained dollars and as quantity indexes; the industry group contributions to the annual percent changes in real GDP are shown; a new table is added presenting information on the composition of current-dollar GPO; the definition of "private industries" is changed to include the statistical discrepancy; the definition of the category "not allocated by industry" is changed to reflect the introduction of chained-dollar estimates; and the tables showing industry shares of real GPD and fixed-weighted measures of real GDP are dropped.(18) Each of these changes is discussed below.

In the previously published estimates, real GPO for 1977-87 was shown in two tables: One table showed benchmark-years-weighted indexes for GDP and for all manufacturing industries and fixed (1987) weighted indexes for all other industries; and the other table showed constant-dollar GPO for all industries calculated using fixed (1987) weights. (Beginning with 1988, both the indexes and the constant-dollar estimates were shown using fixed (1987) weights.) In the revised presentation, the table of indexes (table 13) shows chain-type indexes for selected years from 1977-94, and the table of real dollars (table 14) shows chained dollars for all years 1987-94.

A new table (table 12) shows the composition of current-dollar GPO, which is measured as the sum of industry distributions of the components of gross domestic income. (Previously this detail was available only by request.)

[TABULAR DATA OMITTED]

The definition of "private industries" was changed to include the statistical discrepancy, which is the difference between GDP and gross domestic income. Previously, the statistical discrepancy was treated as a separate "industry," and current-dollar GPO was equal to the sum of GPO for private industries, GPO for government, and the statistical discrepancy. The change reflects BEA'S judgement that the GPO of private industries is affected by most of the measurement problems that cause the current-dollar expenditure components used to measure GDP to be more accurate than the current-dollar estimates of gross domestic income.

The definition of the category "not allocated by industry" was changed to reflect the introduction of chained-dollar estimates. Previously, this category consisted of the following. The "residual," or the statistical discrepancy in constant dollars; the difference between constant-dollar GDP and the sum of the estimates of constant, dollar GPO; and, for 1977-86, the effect of using the benchmark-years-weighted formula for calculating real GDP and real manufacturing GPO. In the new presentation, "not allocated by industry" consists of the difference between chain-dollar GDP and the sum of chain-dollar GPO for the detailed industries - private and government - and the statistical discrepancy. The value of not allocated by industry" reflects the lack of additivity of the detailed chain-dollar GPO estimates because of the formula used to calculate chain-type measures of real output and the use of the expenditures measure of real GDP.

The table showing industry shares of real GDP has been replaced because of BEA'S judgement that, for most analytical uses, shares in current dollars (as shown in table 11) are a better indicator of an industry,s relative size in the economy in any one period. Industry shares in real dollars, whether measured in chained dollars or in constant dollars, are dependent on the choice of the base period and therefore are not good indicators of relative size.

[TABULAR DATA OMITTED]

Methodology

This part of the article discusses the methodologies - that is, the source data and estimating procedures - used to prepare the estimates of current-dollar and real GPO. Previously published tables that summarize the methodology have been updated to incorporate the major changes introduced in this revision of the GPO estimates.

Current-dollar estimates

As noted in the box "Gross Product Originating: Definition and Relationship to Gross Domestic Product" on page 133, the current-dollar GPO estimates are prepared as the sum of the distributions by industry of the components of gross domestic income. This section describes the methodology for distributing the current-dollar estimates of these components,

For most components of gross domestic income, the estimates are based on source data that provide industry distributions on either a company basis or an establishment basis. Only the estimates with distributions based on establishment data can be used directly to calculate industry GPO. For those components that are estimated from Internal Revenue Service (IRS) tabulations of business tax returns, which have company-based distributions, the industry distributions may need to be converted to an establishment basis. This conversion is particularly necessary for large multi-establishment companies that typically have establishments classified in different Standard Industrial Classification (SIC) industries. For the components of gross domestic income for which the source data provide no industry distribution, BEA has developed establishment-based industry distributions from related sources. Table 6 shows the major source data for each component of gross domestic income, the availability an,d type of industrial distribution in the source data, and the data or assumptions used, when necessary, to develop establishment-based industry distributions.(19)

The methodology used to convert corporate profits before tax and corporate capital consumption allowances is based primarily on special Census Bureau matrices of the employment of establishments of corporations. These matrices present employment of these establishments cross-classified by (1) the company-industry classification assigned by IRS in preparing the tabulations of corporate tax returns and (2) the establishment-industry classification assigned by the Census Bureau in the economic censuses. For integrated petroleum companies, the results of applying this matrix are supplemented by information from Department of Energy tabulations of the net income and depreciation of energy companies on an establishment basis. Adjustments to the matrix also are made, when necessary, to reflect publicly available information about large mergers, acquisitions, or changes in company diversification that have occurred since 1987, the year covered by the latest matrix.

Real (chained-dollar) estimates

The real, or chained-dollar, GPO estimates for each industry, industry group, and for all private industries are derived as the product of the chain-type quantity index (divided by 100) and the corresponding 1992 current-dollar value. As in the previously published real GPO estimates, which were calculated using both benchmark-years-weighted and fixed-weighted indexes, three methods are used to calculate quantity indexes: Double deflation, extrapolation, and direct deflation.(20) The method chosen depends on the availability and reliability of source data.

* In the double-deflation method, real GPO is calculated using the chain-type formula, as shown in the box "Computation of the Chain-Type Quantity Indexes for Double-Deflated Industries" on page 142. For this method, separate estimates of gross output and of intermediate inputs enter into the calculation of real GPO. For farms and for nonfarm housing services, complete and consistent current-dollar series are available for gross output and for intermediate inputs. For most other industries, suitable intermediate input series are not available; instead they are obtained by subtracting current-dollar GPO from current-dollar gross output.

* In the extrapolation method, real GPO is calculated by extrapolating the current-dollar value Of GPO for 1992 (the present base period) in both directions by the quantity indicator. In most industries for which this method is used, the quantity indicator is usually the number of persons engaged in production or the number of hours worked.

* In the direct-deflation method, real GPO is derived by deflating current-dollar GPO, usually by a chain-type measure of gross output prices or by earnings.

Table 7 identifies which of the three methods for calculating real GPO is used for each industry. For industries for which the double-deflation method is not used, the table also shows the key. source data used in the preparation of real GPO.

[TABULAR OMMITTED]

Real GPO calculated using the double-deflation method requires detailed information on gross output and intermediate inputs. Table 8 provides a summary description of the principal source data used to prepare the gross output estimates.

The current-dollar intermediate input estimates are derived in four steps: (1) The input compositions for 1977, 1982, and 1987 are derived from BEA'S benchmark I-O tables; (2) the input compositions for 1978-81 and for 1983-86 are estimated by interpolating the detailed compositions from 1977, 1982, and 1987; (3) the imported and domestically produced shares of each detailed input for 1977-87 are estimated" and (4) the input compositions for 1988-94 are estimated, primarily based on the 1987 composition and on information from the 1992 Economic Censuses.

Real intermediate inputs are prepared by deflating each of the detailed current-dollar inputs; imports and domestic production are deflated separately. Prices for domestically produced intermediate inputs are largely based on the prices used to prepare the estimates of real gross output, as shown in table 8. For services prices, additional detail is shown in table 9. The import prices are developed from a variety of sources, primarily from Bureau of Labor Statistics (BLS) import price series, and are the same as those used for the NIPA estimates of imports. (For years before 1981, however, many of the detailed BLS import prices are not available, and the prices used reflect rates of change of more aggregate BLS import prices or of corresponding domestic prices based on the producer price indexes.)

[TABULAR DATA OMITTED]

Tables 8 through 15 follow.

[TABULAR DATA OMITTED]

Gross Product Originating: Definition and Relationship to Gross Domestic

Product

Gross product, or gross product originating (GPO), by industry is the contribution of each private industry and government to gross domestic product (GDP). An industry's GPO, often referred to as its "value added," is equal to its gross output (sales or receipts and other operating income, commodity taxes, and inventory change) minus its intermediate inputs (consumption of goods and services purchased from other industries or imported).

In concept, GDP measured as the sum of GPO in all industries is the same as GDP measured in two other ways: As the sum of expenditures (consumer spending, investment, net exports, and government consumption expenditures and gross investment) and as the sum of the costs incurred (such as compensation of employees, net interest, and indirect business taxes) and the profits earned in production. In practice, BEA Uses only the latter two ways to estimate GDP, but because of less than perfectly consistent source data, the resulting totals are not the same.

The current-dollar estimate of GDP is defined as the sum of the expenditure components, and gross domestic income is defined as the sum of costs incurred and profits earned; the difference between GDP and gross domestic income is the statistical discrepancy. Because the current-dollar GPO estimates are measured as the sum of distributions by industry of the components of gross domestic income, the sum of the current-dollar GPO estimates also differs from current-dollar GDP by the statistical discrepancy.

Real GDP is also measured as the sum of the expenditure components, using the formula for calculating chain-type measures. However, estimates of real gross domestic income are not prepared, because price indexes cannot be associated with income measures as they can be with the goods and services that make up the expenditure measures. Real GPO estimates for most industries are derived using the formula for calculating chain-type measures with separate estimates of gross output and intermediate inputs. (See the box "Computation of the Chain-type Quantity Indexes for Double-Deflated Industries, on page 142 for information on how output and inputs are combined in these calculations.)

The sum of the chain-dollar GPO estimates differs from chain-dollar GDP by the chain-dollar statistical discrepancy, which is shown as part of private industry GPO, and by the discrepancy entitled "not allocated by industry," which is the difference between chain-dollar GDP and the sum of chain-dollar GPO for the detailed industries and the statistical discrepancy. The value of the category "not allocated by industry" reflects the lack of additivity of detailed chain-dollar GPO estimates because of the formula used to calculate chain-type measures of real output and because of the use of the expenditures measure of real GDP.

The statistical discrepancy is included in the GPO Of private industries partly because Of BEA'S judgement that the current-dollar expenditure components used to measure GDP are more accurate than the current-dollar estimates of gross domestic income and that most of the measurement problems affect the GPO of private industries. For example, the adjustments to source data to account for tax-return misreporting, which are based on periodic IRS audit studies and rough estimates of the amounts of income not detected by these audits, are larger for the cost components of gross domestic income than for the expenditure components of GDP. In addition, it is BEA'S judgement that the real expenditure components used to measure GDP are more accurate than the real GPO estimates. The amount of detailed expenditures data that are available for weighting the price indexes used in calculating GDP is greater than that for gross outputs and intermediate inputs used in calculating GPO, and little information is collected annually on the composition of inputs or of nonmanufacturing outputs. For some industries, no source data are available to measure gross output, and the resulting GPO estimates are prepared using less reliable methodologies.

Computation of the Chain-Type Quantity Indexes for Double-Deflated

Industries

For this comprehensive revision, BEA introduces annual chain-type quantity indexes as the measure of real gross output, intermediate inputs, and GPO for industries and industry groups. Each link in the chain-type quantity index is a Fisher quantity index for two adjacent years. Each annual Fisher quantity index, in turn, is the geometric mean of Laspeyeres and Paasche quantity indexes for the two adjacent years.

The formulas below summarize the computation of the Fisher chain-type quantity indexes of real gross output, intermediate inputs, and GPO for an industry or industry group. In the notation, L refers to the Laspeyeres quantity index, P refers to the Paasche quantity index; F refers to the Fisher quantity index, and C refers to the Fisher chain-type quantity index The subscripts indicate time periods; [L.sub.t-1,t] is the Laspeyeres quantity index for the two adjacent years, t - 1 and t. The superscript GO refers to gross output; II refers to intermediate inputs, and GPO refers to gross product originating. Lowercase p and q refer to detailed prices and quantities, respectively.

Laspeyeres quantity indexes for gross output, intermediate inputs, and GPO, respectively, are

[Mathematical Expressions Omitted]

Paasche quantity indexes for gross output, intermediate inputs, and GPO are

[Mathematical Expressions Omitted]

Fisher quantity indexes for gross output, intermediate inputs, and GPO are

[Mathematical Expressions Omitted]

Fisher chain-type quantity indexes for gross output, intermediate inputs, and GPO for years following the base year are

[Mathematical Expressions Omitted]

In the base year (1992 for this comprehensive revision),

[Mathematical Expressions Omitted]

The above formulas are applied to GPO industries, to industry groups such as durable manufacturing, and to aggregates such as private industries.

(1.) The previously published estimates of gross product by industry for 1947-90 appeared in the November 1993 issue of the Survey of Current Business, and the estimates for 1991-93, in the April 1995 issue. Revised GPO estimates for 1947-58 will be released after the release of the revised NIPA estimates for 1929-58 in the fall of 1996. (2.) The GPO and the gross domestic product (GDP) estimates in this article do not reflect the results of the recently released annual NIPA revision, which is presented in this issue of the Survey. The effect of not incorporating these results is small, The growth rate of real GDP for 1993 is revised up 0.1 percentage point to 2.3 percent, and the growth rate for 1994 is unrevised. (3.) For a description of the results of the comprehensive NIPA revision, "Improved Estimates of the National Income and Product Accounts for 1959-95 Results of the Comprehensive Revision," Survey 76 (January/February) 1996): 1-27. (4.) For information on the GPO improvement program, see "Gross Product by Industry, 1977-88: A Progress Report on Improving the Estimates, Survey 71 (January 1991): 23-37; and Robert P. Parker, Gross Product by Industry, 1977-90," Survey 73 (May 1993): 33-54. (5.) For additional information, see "Mid-Decade Strategic Review Of BEA's Economic Account: Maintaining and Improving Their Performance" Survey 75 (February 1995): 36-66; and "Mid-Decade Strategic Review of BEA'S Economic Accounts: An Update," Survey 75 (April 1995): 48-56. For a summary of BEA'S progress in implementing the plan, see BEA'S Mid-Decade Strategic Plan: A Progress Report," Survey 76 (June 1996): 52-55. (6.) As in the previously published series, estimates for 1959-86 are classified according to the 1972 Standard Industrial Classification (SIC), estimates for 1988-94 are classified according to the 1987 SIC, and estimates for 1987, are classified according to both the 1972 SIC and the 1987 SIC. The detailed data needed for an SIC conversion are not available. Consequently, directly comparable estimates for 7, of the 66 detailed industries cannot be computed for the entire period; combinations among these seven industries that are comparable over time are shown in the addenda to table 15. For all other industries, the definitions are comparable before and after 1987.

Growth rates for industries and industry groups are computed from the chain- type quantity indexes shown in table 13. (7.) For the industrial machinery and equipment industry, the growth rate for 1977-94 was computed using the 1977 value for the 1972 SIC "machinery, except electrical" industry, which is roughly comparable in definition to the 1987 SIC "industrial machinery and equipment" industry. (8.) The selection of 1987 as a breakpoint reflects that it was the base year for the previously published estimates and that it was also the year of a major updating of the SIC. (9.) For these calculations, current-dollar estimates for GDP, and for GPO for each industry group, are extrapolated for each year by the industry's chain-type quantity index to derive an estimate of change for the period in the prices of the initial year. Dollar-based contributions are then computed as a percentage of the total change in GDP for the period. These contributions are then used to estimate the percentage-point contributions of industry groups to the growth in real GDP. For years in which relative prices changed significantly, the detail may not add to the corresponding totals. (10.) For this comparison, private services-producing industries are defined to consist of the following industry groups: Transportation and public utilities; wholesale trade; retail trade; finance, insurance, and real estate; and "services." Private goods-producing industries are defined to consist of agriculture, forestry, and fishing; mining; construction; and manufacturing. (11.) "Other" GPO is the sum of proprietors' income, corporate profits, net interest, capital consumption allowances, business transfer payments, and the current surplus of government enterprises less subsidies. Proprietors, income is included in "other" GPO as a return to capital from production; however, an unknown portion of proprietors, income represents a return to labor. (12.) For 1977-87, the previously published rates of change for GDP and for manufacturing were calculated using benchmark-years-weighted quantity indexes. These quantity indexes differ from the chain-type indexes because the price weights are for adjacent benchmark years. For 1987-93, the fixed (1987) weighted quantity indexes that were used for nonmanufacturing industries for all periods were also used for GDP and for manufacturing. (13.) See the box "The Measurement of Change in Real GPO by Industry" in Parker, "Gross Product by Industry, 1977-90," 36-37. The use of the chain-type measure in place of the benchmark-years-weighted measure resulted in a minimal revision of the growth rate in manufacturing for the 1977-87 period. 14. For an explanation of this method, see the section Real (chained-dollar) estimates" in "Methodology." (15.) This change does not affect the GPO Of government enterprises, because the inclusion of the consumption of fixed capital in the gross product of government enterprises is offset by a corresponding reduction in the surplus of government enterprises. It should be noted that the new treatment of investment still does not provide an estimate of the full value of the services of general government fixed assets, the new component that enters GDP. These services, which are now recorded as current-account purchases, are measured using the convention that these services equal the estimate of general government consumption of fixed capital; that is, the net rate of return on general government fixed assets is assumed to be zero. (16.) The following Federal excise taxes were also affected by the reassignments: The diesel and special motor fuels taxes, which were shifted from retail trade to wholesale trade, beginning with 1988; the heavy-duty truck tax, which was shifted from motor vehicles and equipment manufacturing to wholesale trade, beginning with 1983; the imported petroleum products tax, which was shifted from petroleum and coal products to wholesale trade, beginning with 1990; and the imported chemical products tax which was shifted from chemical and allied products manufacturing to wholesale trade, beginning with 1990. (17.) In the previously published estimates, real inputs for three industries-construction, fabricated metal products, and industrial machinery and equipment-were estimated beginning with 1988 by assuming no change from the 1987 relationship between real inputs and real gross output. The special procedure was used for these industries because their input compositions were judged to have changed since 1987 in ways that would result in significant errors in industry real inputs; in the current revision, based on the new information from the final 1987, benchmark I-O table and 1992 Economic Censuses, this special procedure has been dropped. (18.) In addition to these changes, the industry "social services and membership organizations," shown in the previously published estimates, has been split into a social services, industry and a membership organizations, industry. (19.) For additional information about the methodology used to estimate the components of gross domestic income, see table 1 in "Updated Summary Methodologies" in this issue of the Survey. (20.) For information about the effects of the choice of methods, see Parker, "Gross Product by Industry, 1977-90," 43-46.

Acknowledgments

Robert E. Yuskavage, Chief of the GDP by Industry Branch of the Industry Economics Division (IED), supervised the preparation of the estimates. Ann M. Lawson and Robert P. Parker provided overall guidance. Felicia V. Candela, Sherlene K.S. Lum, Brian C. Moyer, Timothy F. Slaper, John Sporing, and Robert A. Sylvester prepared the estimates. A. Vanessa Clark, Sonia R. Jones, and Karen A. Newman provided support services. Other IED staff prepared tabulations from the preliminary 1992 benchmark input-output table.

Members of the staffs of the National Income and Wealth Division and the Government Division - particularly Sherman Hammack, M. Greg Key, Leonard J. Loebach, Brooks B. Robinson, and David B. Wasshausen - prepared special tabulations from the NIPA'S and contributed to the development of the estimates. Stephanie L. Howeu in the Office of the Director prepared the estimates of industry contributions to the growth in real GDP.