Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 Measures of Government Productivity: Concepts, Methods, and Sources Previous chapters have discussed the concepts of productivity. We turn now to the way in which these concepts have been defined and measured in practice. This chapter describes the data sources and methods used by government agencies, particularly the Bureau of Labor Statistics (BES), in measuring productivity. It also includes brief comments on the adequacy and reliability of the official measures; major evaluation of these measures is included in Chapters 5 and 6. Productivity measures by private researchers that are almost as well known and widely used as the government measures are also covered in later chapters. Following a brief overview of government measures, the rest of this chapter is divided into sections on each of the major classifications of the economy for which official productivity measures are prepared: economy- wide, industry divisions, selected detailed industries, and the federal government. The last section, "Omissions from Existing Measures," deals with sectors and external effects not covered in any official Pleasures of productivity. Although this survey points up some gaps and weaknesses in the current official programs, we do not mean to imply that the current measures are not useful or that the agency personnel responsible for the series are not well aware of the weaknesses and gaps. Indeed, in many cases it was agency personnel who called our attention to the problems. For example, much of our analysis of weaknesses in the measurement of labor input came from a report of the Bureau of Labor Statistics Task Force on Hours Worked (1976~. 50
Government Measures of Productivity OVERVIEW OF GOVERNMENT MEASURES 51 BES iS the major producer of government productivity measures.) It publishes or maintains measures for a wide array of sectors and industries from aggregate measures covering almost the whole economy to measures for detailed industry classifications. However, BES publishes only measures of labor productivity. Measures of output per unit of labor and physical capital combined, although widely estimated by private investigators, are not now part of the official program to measure productivity.2 Our review considers the various sector and industry measures separately since issues about data and concepts diner somewhat for the aggregate economy, industry divisions, and detailed industries. Two basic considerations apply to the official measures at all levels of aggregation, however, and it is useful to set them out in advance. One is the use of many secondary data sources in measuring productivity. The other is the difficulty of measuring real output for some sectors and industries. DATA SOURCES Unlike the government's series of measures of unemployment and prices, official measures of productivity are made up of data components that come from separate and independent surveys that are conducted for purposes other than measuring productivity. Even in measuring productiv- ity for detailed industry categories (e.g., cereal breakfast foods), BES must usually combine data from two independent surveys of the establishments in the industry. Data for labor input and the value of output come from one survey (taken by the Census Bureau); data on prices, which are needed to convert the value data into measures of real output change, come from another survey (usually taken by BES). For measuring productivity at high levels of aggregation (e.g., the private business sector), many different data sources must be combined. Not only do the data on value of output, price change, and labor input come from different sources, but also for each of the three components a number of different sources must be used (especially for value of output). Moreover, the data sources are not all governmental; trade associations and large companies are important sources of data for some industries. Neither are all the sources statistical surveys; some data are by-products of activities of regulatory agencies. This situation makes it difficult to develop objective measures of the margin of error in productivity estimates, especially the more aggregate ones. The well-established methodologies for estimating sampling error that are applied to other official statistics, like the unemployment rate,
52 REPORT OF THE PANEL cannot be applied to the official productivity measures. Holland and King (in this volume) attempt to measure overall sampling error of output-per- hour measures from the sampling error in individual data series that are used in constructing these measures. Recommendation 3. The Panel recommends that the Bureau of Labor Statistics and the Bureau of Economic Analysis explore methods for estimating the implications of error reduction in component measures for the reduction of overall error in productivity measures beyond that corrected by routine revisions. The Panel recognizes that BES must continue for the present to use many sources of data because of the very high cost of carrying out a large- scale special survey to collect all the data components needed for productivity measures.3 The full expense of such a survey would involve not only direct outlays by the federal government but also the cost to firms to respond. (Ways to improve indirect measures of output per hour and other government statistics through cooperation among the many statisti- cal collection agencies are explored in Chapter 8.) PROBLEMS OF MEASURING REAL OUTPUT The difficulty of measuring changes in real output varies greatly across sectors and industries. At one extreme is an industry like copper whose output is a simple standardized product that does not change either over time or from place to place. At the other extreme are firms and agencies in the not-for-profit and government sectors, for which output measurement is extremely di~cult.4 Along a spectrum between these extremes are products and services for which good measures of output are available but only with increasing degrees of difficulty: apparel, consumer durables, new construction, producer durables, prescription drugs, the services of doctors and dentists, and the services of financial intermediaries are some examples of the more difficult ones. The current official program only publishes measures for specific industries when the concepts and data permit relatively good measures. Because there is also a need for productivity measures that cover broad sectors of the economy, BES has a program of measures covering the private business sector and major industry divisions. However, the data for constructing these measures vary in quality. In making aggregate measures, the better data are combined with weaker data, and the resulting measures necessarily suffer from some of the drawbacks of the weaker
Government Measures of Productivity - c' .o _ 8 An IL y J 1 LO o Cry 5 o I to LL ~ 4 J o 53 9 7 6 Productivity at constant rate of growth (3.3 percent per year) '' ' ~Actual productivity / 3LI 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1948 1 1952 1956 1960 1964 1968 1972 1976 YEAR FIGURE 4-1 Productivity in the private business sector. (U.S. Department of Labor.) components. Our presentation will try to allow the reader to identify these weak spots. ECONOMY-WIDE MEASURES The most inclusive productivity measure regularly published quarterly and annually by BES iS real output of the private business sector per unweighted hour of all workers. Figure 4-1 shows the trend since World War II in this measure. This measure undoubtedly has an important influence on public policy formation. Its movements over long periods have given rise to the current concern over the productivity slowdown, while its movements from quarter to quarter and year to year are closely watched as an important cyclical indicator.5 The productivity measure in Figure 4-1 is composed of a numerator that measures real output produced and a denominator that measures the labor input used in producing that output. Table 4-1 shows the value of labor
54 REPORT OF THE PANEL TABLE 4-1 Components of the Bureau of Labor Statistics' Productivity Measure for the Private Business Sector, 1''73-1977 Labor Inputb Output per Annual Change in Outputa (billions (billions of Hour (billions Output per Hour Yearof 1972 dollars) hours) of 1972 dollars) (id) 1973974.5 134.0 7.27 2.0 197495 1.3 134.6 7.07 .-2.7 1975929.1 129.1 7.20 2.4 1976993.6 132.6 7.49 4.0 19771,053.4 137.1 7.68 2.5 aFrom last row of Table 4-2. bUnpublished data from the Bureau of Labor Statistics, U.S. Department of Labor. input, output, and the ratio of the two for the private business sector for 1973-1977. BLS generates its own measure of the denominator, relying largely on data it collects itself. For the numerator, BES uses, with minor modifications, measures produced by the Bureau of Economic Analysis (BEA) in the U.S. Department of Commerce as part of the program of national income and product accounts. BEA uses data from a large number of sources to estimate income and product. REAL OUTPUT The numerator of the private business sector series is derived from measures of the gross national product (GNP) and its components. The GNP measures the market value of the output of all capital and labor resources that generate observable income transactions, plus the imputed product of some resources for which there are no observable market transactions, such as the services of owner-occupied homes and food produced and consumed on farms. The GNP includes the output of domestic workers hired by households (although not that of people who do the same work for themselves). The GNP also includes the output of employees of local, state, and federal government agencies and of not-for-profit organizations. However, BES subtracts the measures of output for these three sectors- household employment of domestic workers, employees hired directly by government, and not-for-profit organizations-from the GNP because at present the output measures for these sectors are based only on the amounts of labor inputs that are used. In the absence of any observable transactions involving output, BEA uses changes in inputs to measure changes in real output. When BES measures changes in labor productivity, it excludes these sectors from the output measure because, by definition,
Government Measures of Productivity TABLE 4-2 Components of the Difference Between the Gross National Product and the Numerator of the Bureau of Labor Statistics' Productivity Measure for the Private Business Sector, 1973-1977 (billions of 1972 dollars) 55 Component1973a1974al97sa1976a 1g77b ._ . Gross national product 1,235.0 1,217.8 1,202.1 1,274.71,337.5 Less: rest of world 7.6 6.8 4.9 6.77.4 Gross domestic product 1,227.4 1,211.0 1,197.2 1,268.01,330.1 Less: government 138.9 141.9 144.6 145.8147.5 Gross domestic private product 1,088.5 1,069.1 1,052.6 1,122.21,182.6 Less: households and institutions 38.1 38.0 38.9 40.241.4 Gross domestic business product 1,050.4 1,031.1 1,013.7 1,082.01,141.2 Less: Gross product of owner-occupied housings 71.0 74.5 79.0 83.286.1 Less: "residual" 4.9 5.3 5.6 5.21.7 Bureau of Labor Sta tistics numerator (private business sector) 974.5 951.3 929.1 993.61,053.4 aBureau of Economic Analysis (1977) Survey of Current Business, July, Table 1.8. bBureau of Economic Analysis (1978) Survey of CurrentBusiness, February, Table 3. CUnpublished estimates by the Bureau of Economic Analysis, U.S. Department of Commerce. productivity change is zero.6 BES also excludes from output the imputed rental value of the services of owner-occupied homes. This is done because there is no way of obtaining good measures of the labor input (e.g., maintenance activities) associated with this output. Finally, BLISS deducts the output of resources owned by U.S. citizens but located in foreign countries and adds the output of resources owned by foreigners but located in the United States. This is done because the data for measuring labor input include only labor at work in the United States. In BEA terminology the numerator of the BES measure is equal to gross domestic business product in constant dollars minus the imputed value (in constant dollars) of the services of owner-occupied homes. Table 4-2 shows the components of the difference between GNP and the numerator of the BES productivity measure. BEA estimates gross domestic business product in constant dollars in two ways. One involves measuring the flow of real goods and services to the final-demand categories, and the other involves measuring the real
S6 REPORT OF THE PANEL gross product originating in the industries that make up the private business sector.7 If all the data used in the two approaches were without error, the two measures would be equal. In practice, of course, the data have errors that lead to a difference between the two measures. The difference, called the residual, has been between 0.1 and 0.5 percent, with the final-demand measure always exceeding the product-originating measure. The rest of this section describes the BEA method based on the sales to final-demand approach. The following section, which is devoted to measures for industry divisions, describes the data and procedures used by BEA to measure real gross product originating by industry.8 The BEA procedure for estimating domestic business product in constant dollars involves two steps. First, BEA estimates output in current dollars from data on business sales to the major categories of final demand: personal consumption expenditures (PCE, 72.2 percent), producers' dura- ble equipment (PDE, 7.5 percent), structures (7.7 percent), government purchases of goods and services from private business (11.0 percent), change in inventories (0.8 percent), and net exports (0.9 percent).9 Second, it converts the current-dollar measure to one in constant dollars by a detailed deflation procedure. i. . r . ~ . BEA eStlmaleS OI current-aollar components come in a sequence ot preliminary and revised estimates for any given year (called a reference year) with each subsequent revision being based on more comprehensive and more direct sources of data. Table 4-3 shows some of the data sources used by BEA for making the annual current-dollar estimates. The table is arranged by final-demand category and revision sequence.l° The data sources listed in Table 4-3 are a subset of all the sources used; the other data used are more specialized and are usually collected either by a trade association or by a government regulatory agency. The data sources diner in terms of sample size and the randomness of sampling procedures as well as in the degree to which they directly measure the magnitudes of interest. Some detail on the estimation of current-dollar PCE in 1974 will illustrate the characteristics of some of the data sources. In 1974, 42.8 percent of current-dollar PCE was based on data from the current-dollar sales of goods (durable and nondurable) reported in the monthly and annual Census Bureau Surveys of Retail Trade (see Once of Federal Statistical Policy and Standards 1977, Chapter 5~. For the remainder of PCE goods-new and used cars and trucks, and gasoline and oil (13 percent of PCE - BEA uses data from business and trade association sources. For new cars, four data sources are used. One trade source provides data every 6 months on auto production and list prices broken down by detailed categories of nameplate, model, and options. These are used to establish the average list price. Another trade source provides data
Government Measures of Productivity 57 on overall auto sales (i.e., sales to consumers, businesses, and govern- ments) in quantity units. BES provides data on auto price discounts from the consumer price index Icy. Finally, a third industry source provides data used to estimate what part of the unit sales went to consumers. For consumer expenditures on gasoline and oil, BEA uses data provided by the Ethyl Corporation on the quantity of gasoline sold in a sample of gasoline stations and the data from the BES consumer price index for gasoline prices. For the services component of PCE (44 percent of PCE), a number of data sources are used. For hotel, personal, recreational, medical, and legal servicesii (6 percent of PCE), the data come from the Census Bureau's Monthly Selected Services Receipts Survey (see Once of Federal Statisti- cal Policy and Standards 1977, Chapter 5~. For the rents paid for rented dwellings (4 percent of PCE), BEA combines information on the number of tenant-occupied housing units from the Annual Housing Survey (of the Census Bureau and the U.S. Department of Housing and Urban Development) with information on rental prices from both the latest decennial census and the consumer price index. To estimate expenditures on auto repair services (2 percent of PCE), BEA combines trade source data on the number of car registrations with cat information on prices of auto repairs. For household utilities gas, electricity, telephone, etc.-(6 percent of PCE), BEA uses either company or industry sources. For example, the American Telephone and Telegraph Company supplies data on residential local and long distance telephone calls. The discussion above relates to estimation of output in current dollars; to measure output in constant dollars, BEA deflates the current-dollar data. For the large part of PCE that involves direct and explicit money transactions between consumers and private business firms (about 80 percent of PCE), BEA uses price change information from the cat for deflation purposes. The deflation is carried out at a fairly detailed level of product and service disaggregations.~3 The remainder of private business PCE iS made up almost entirely of the services of life insurance companies and banks and other financial intermediaries. For this sector, because of conceptual and data gaps, BEA measures real output change by measuring input change. For life insurance companies, the current value of operating expenses is deflated;~4 for financial institutions, changes in unweighted employee-hours between the base year and the current year are used to extrapolate the base-year value of operating expenses. For deflating PDE, which involves direct money transactions between business firms, BEA relies mainly on the price change information in the producer price index UPPED.
58 o 50 C,3 V7 ._ as ._ V3 o o C,3 Cal ._ Cal ·5 a: Cal - C~ ._ ._ - o Cal Cal Cal 3 o V) a' C) _I Cal C) _ .= 3 C ._ '~ V) V) C) Ct ._ =' _ 3 c: ._ ._ _ O A_ C: ~ - _ . ._ ,. _ O _. '- 3 C.) .= ~ Cal ~ Ct O O 3 3 V, r' C: Cal Cal Cal ._ Cat o Cal in: C) =~ ~ U ~ C) ~= ._ ~ U) ~ Ct C) . ~ _ ~ q Cal ~ C) :> 3 So: Cal ~ ~ ·C~ ;> tt I_ 3 -~ O O C~. ~ ~ - o ._ _ Cal 3 ~ .'' As c,5 ¢ Cat ¢ Cal .= C) C) at: C) Cal ~ ._ C) ~ Cal Cal U) C) _ ~ ~ V) ~ :^ _ _ _ ~ _ O O V: C) ._ V, - 5 - '-_ ~ O cr lo, _ .= ~ _ ~ O 3 ~ a: 3 Cal - Cal o - - C) Cal 3 ,~ 3 Ed o :^ - U) - Ct 3 ~ ~ C: ¢ U - o ._ 3 - V) o o 3 V) C) C: - _) o _1 C) G.) Cat ~ Can Cal O ~ ;^ C) ~ 3 (A If ~ - o , O) ¢ Cal sit O ^ ~ . ~ - Lea ~- V) ~> - ~. _ ¢ V'^ ~- C) ~ ~-Cal Ax ~ C: ~ O - - - Is: _ - ~ - O - o _ =3 _ ~ (t E - ' V: - C~: ° Cs V, ' ~ ' ~ := ~, c~ ~ ~ C~ o ~ o O v) C) 3 - C~ o :^ 3 - Ct 3 ' ¢ ~; C~ - ._ :t C) ~ C) o C:} V3 ~ C~ C) ~ :> ~ _ C) V) - C~ ~ C) - _ ~ O ~ - o 3 - V) C) 3 - I ~Q O X . . - O _ ~, ~ C~: ~ C) .o O ~- ) ~= . _ ~ ~.- .) ~r O ~ O _ 3 V) - ~: - ~: ¢ 0- o ~ , C~ o - X C) z C~ ~ , , .C ~ o Q~, ~ C~ _ ~ o O ._ e~ _ ;> t, cn D - - c'3 D · ~ ~ C ~ O - O ~ ~ C ~ ~ C~ ~ , O.Q C ~ ·_. - Ce O .o ~, C,0 ·t .~ O V) ,_, ~ O ~ C :D 1- c,, O 0 t4 ~ bc - ~ '~ _ ~ ,~> 3 ~_ ~, ~ t_ _ · t'; -0 X - _ ~ O ~ ~ = ~c' 3 _ :e ~ ~ ._ ~ , _ ~ _,_= o :^ o oo _ - C~ ~ , ~ ~ _ ~ _ ~_ _ ~ ~ o _ - _ tC ^ 3 Am - ~ ; ~ ~ ~ ~_ ~ ~ ~ 00 ~ ~ ·Q 1 1 ~ ~ ~ e~ ~ ~ _ ~ _ _ W ~ _ ~ _ _ ~ ~ - ~ ·' .' .' _ I I c,) O C ~_ ,D ~ ~) ~ ~ :t ~ ~ ~ ~ ~ ~ CQ
Government Measures of Productivity 59 The remaining major components of current-dollar business product shown in Table 4-3 are structures and government purchases from private firms, for both of which there is very little coverage of prices by government statistical agencies. For deflation, BEA uses price indexes from a wide variety of trade associations, regulatory agencies, and other governmental and private sources. When the constant-dollar measures of change in inventories and net exports are added to the estimates described above, the result is BEA'S official gross domestic business product in constant dollars (see Office of Federal Statistical Policy and Standards 1977, Chapters 5 and 8~; BES subtracts the constant-dollar value imputed to the services of owner- occupied dwelling units to arrive at the numerator of its measure of productivity for the private business sector. LABOR INPUT The concept of labor input used by BLS iS the unweighted sum of the hours of all workers engaged in the production of the output measured in the numerator of the productivity ratio. The data used to measure labor input come from different sources and sometimes involve imputations or assumptions. Unlike the output data, which come from many sources, the hours data are based primarily upon sample data from two sources: the current employment statistics (CES) program and the Current Population Survey (cPs). The CES covers all wage and salary workers in the nonagricultural business sector, who account for about 80 percent of total hours of the private business sector. For workers not covered by the CES, primarily those in agriculture and the self-employed, BES relies largely on the cPs. Table 4-4 summarizes the sources and imputations of hours of labor input for major components of the private business sector.~5 Like the output data, the input data come from programs not specifically designed to measure productivity. The CES iS designed to provide timely information on employment, hours, and earnings for the nonagricultural economy and also for detailed industries and for metropol- itan areas and state. The cPs, conducted by the Census Bureau for BES, iS designed primarily to provide timely information about the labor force, employment, and unemployment. The CES data are collected in a mail survey from the payroll records of about 165,000 sample establishments. State employment agencies, cooper- ating with BES, choose the state samples and survey the establishments that participate in the program, which is voluntary. The data cover 1 week, the survey week containing the twelfth day of the month, and become the basis of monthly estimates of employment, average hourly
60 REPORT OF THE PANEL TABLE 4-4 Source Data for Hours in the Private Business Sector, 1977 Percentage of Components' Average Weekly Hours to Total Total and Components Hours (AWH) Employment Hours (Jo) Private business sectora 100.0 Manufactunngb 29.3 Production workers C CES CES 20.8 Nonproduction workersC assumed constantd CES 8.0 Self-employment CPS CPS 0.5 Nonmanufactunuge 63.8 NonsupervisoryC'f CES CES 46.0 Supervisory workersC same as nonsupervisory CES 7.9 Self-employed CPS CPS 9.2 Unpaid family workers CPS CPS 0.7 Government enterprises CPS BEA 2.2 Farm-all employees CPS CPS 4.7 CES = Consumer Expenditure Survey, CPS = Current Population Survey, BEA = Bureau of Economic Analysis. aNot-for-profit institutions are excluded from measures of the private business sector, so the Bureau of Labor Statistics subtracts their hours from those of the entire sector. Esti- mates of not-for-profit institutions" hours are calculated by dividing compensation in not-for-profit institutions by the compensation per hour figure for the entire industry. BEA is the source of the compensation data, collected annually. bBureau of Labor Statistics assumes no unpaid family workers in manufacturing. CEstimates of employment are benchmarked to unemployment insurance, Social Se- curity, regulatory and other government agencies, and private sources. dBureau of Labor Statistics has held AWH for nonproduction workers in nondurable manufacturing constant at 39.1 and for durable at 39.7 since 1962. Prior to 1962, it estimated these hours from its Area Wage Surveys but has found them changing very gradually and so has not changed estimates in recent years. eNonmanufacturing includes mining, construction, transportation, wholesale and retail trade, finance, insurance and real estate, and public utilities. Construction workers in the construction industry. SOURCE: Derived from unpublished data provided by the Bureau of Labor Statistics, U.S. Department of Labor. earnings, and average weekly hours. (When a survey week contains a holiday, BES makes adjustments so that data for that week correspond to a regular 5-day week.) The cPs sample consists of about 56,000 households. Since it is a probability sample, reliable estimates can be constructed from the sample data for the entire labor force. The sample households are visited by a Census Bureau enumerator or are interviewed by telephone: a respondent is asked to report hours at work and where worked for each member of the household who is 16 years of age and older. Like the CES data, the monthly cPs data are based on 1 week, the survey week containing the twelfth day
Government Measures of Productivity 61 of the month. An important difference between the two surveys is that the cPs depends on the responses and memories of individuals while the CES iS based on payroll records. Periodically, there is a benchmark of adjustment of CES estimates of employment, but not hours, to more comprehensive sources. The benchmark adjustment is made to unemployment insurance data, which cover nearly all workers in the CES target population. For the remaining workers, BES relies on data from other government sources and from private agencies. The employment benchmark in recent years has been 2 or 3 years behind the current estimate. Nonetheless, revisions in the employment estimates at the time of the adjustment are small, primarily because a correction factor is used to minimize the bias that would result from the entry of new firms between benchmark adjustments. There is no benchmark adjustment for the cPs. The CES collects data on both employment and hours for production workers (in manufacturing and mining) and nonsupervisory workers (elsewhere). The hours recorded are hours paid for, rather than hours at the workplace: this includes hours spent at the workplace plus hours of paid leave. (For further discussion of measurement of hours, see Chapter 6.) For nonproduction and supervisory workers, only employment is reported. For nonproduction workers, hours are assumed to be constant at their 1962 level; for supervisory workers, hours are assumed to equal those of nonsupervisory workers in the same industry. In 1977, about 16 percent of the labor input in the productivity measure for the private business sector was based on assumptions about hours worked rather than direct measurement. The cPs covers the entire labor force and is therefore an alternative to the CES as a source of hours and employment data for nonagricultural employees. Formerly, BES published two measures of labor productivity for the private business sector, one based on cPs data and other based primarily on CES data. It now publishes only the latter, however, because the cEs-based productivity series is smoother (fluctuates less) than the one based on cPs data and so gives a clearer indication of trends in productivity. The alternative measure of productivity based on cPs data is maintained by BES and used as a check on the published series. REVISIONS Both the numerator and denominator of BES productivity measures undergo a number of revisions. The revised estimates of output and labor input can provide information on the probable reliability of the earlier estimates of productivity change. At present, BES does not provide an
62 REPORT OF THE PANEL estimate of the probable size of the subsequent revision when it releases its preliminary estimates, even though the revisions are often quite large (see De Sha in this volume). Recommendation 2. The Panel recommends that the Bureau of Labor Statistics study the size, direction, and other characteristics of past revisions in estimates of output per hour and, on the basis of its findings, consider publishing in its press releases a range of probable revision (based on historical experience) for the preliminary estimates. INDUSTRY DIVISIONS In addition to measuring productivity for the whole private business sector, BES also produces measures of output per hour for the industry divisions that comprise it. These measures are an important supplement to the aggregate measures just described; however, only the measure for the manufacturing division is published.~7 In addition to providing indicators of the performance of the individual subsectors, these measures also provide clues to the sources of overall productivity change. For example, knowledge of differences in the level and rate of growth of labor productivity across industry divisions allows one to measure the contribu- tion of shifts in industry composition to the aggregate growth rate. The sources of productivity growth may also differ across industry divisions- technology may be important in one industry and managerial or organizational innovations in another so that knowledge of individual industry growth rates can help to identify underlying causes of productivi- ty change. Figure 4-2 shows changes in the measure for 1948-1976 for the manufacturing division and compares it with changes in the measure for the private business sector. Our discussion of productivity measures for industry divisions is limited to the annual published measure for manufacturing. Since the procedures for estimating labor input by industry division are essentially identical to those used for the private business sector, our description of sources and methods is also limited to those used for measuring real output. DATA SOURCES AND ESTIMATING METHODOLOGY For the numerator of the measures of productivity by industry division, BES uses the BEA measures of real gross product originating (GPO) by industry. The BEA concept of industry output is that of value added or product originating, rather than of total output. The value-added approach measures only the additional contribution to total industry output made by
Government Measures of Productivity loo J 6 oh o 6 75 50 Manufacturing C 63 } ~Private business sector 950 ~ 955 ~ 960 1 965 ~ 970 ~ 975 YEAR FIGURE 4-2 Productivity for the private business sector and manufacturing (1967 = 100~. (Bureau of Labor Statistics productivity press releases.) the factors of production located in the given industry, as distinguished from the contribution made by factors in other industries. This contribu- tion in current dollars is measured by the difference between the value of the industry's total output and the value of the goods, materials, and services purchased from other firms (and used up in the production of its current output). As was the case with measuring GNP by output sold to the final-demand categories, this same quantity (in current dollars) can also be measured from data on the incomes of the factors in the given industry and on the non-factor costs of production (e.g., excise taxes paid. To measure changes in real value added or gross product originating, BEA uses the method of "double deflation" for about half the industry divisions and other procedures for the rest. The term "double" indicates that both production and purchases must be deflated in order to measure changes in the real output attributable to the factors employed in a given industry. This procedure as applied to manufacturing is described below.3 9 To help readers understand the procedure, Table 4-5 shows an example of the digit levels of the Standard Industrial Classification Psych. The manufacturing division is made up of 450 detailed 4-digit industries, e.g., meat packing plants (sac 2011), cereal breakfast foods (sly 2043), women's full-length and knee-length hosiery (sac 2251~. For each of these detailed industries, the Annual Survey of Manufactures provides
64 REPORT OF THE PANEL TABLE 4-5 An Example of Standard Industrial Classification (sac) Codes for Industries and Products Level SIC Code Industry or Product Description 1-digit 2-digit 3-digit 4-digit S-digita 7-digit 2 Manufacturing 27 Printing, publishing, and allied industries 273 273 1 27311 2731111 7-digit 2731113 4-digit 5-digit 7-digit 2732 27321 2732111 7-digit 2732113 Books Books: publishing Textbooks Textbooks: elementary, hardbound Textbooks: high school, hardbound Includes establishments engaged in durable or nondurable manufacturing Includes establishments engaged in printing or performing ser vices for printing trade Includes establishments printing and publishing books Includes establishments primarily engaged in publishing only of books and pamphlets This is a product class for the industry SIC 2731 This is a specific product in the product class SIC 27311; SIC 27 31 1 12 is for elementary paperbound textbooks This is another specific product in SIC 27311; SIC 2731114 is for high school paperbound text books; SIC 2731115 and 27311 16 refer to college text books, hardbound and paper bound Books: printing Includes establishments primarily engaged in printing and bind ing books This is a product class for a speci fied method of manufacture This product code includes ele mentary and high school books together Books: printing only, lithograph Books: printing only, lithograph, elementary, high school, hard- bound, paperbound Books: printing only, lithograph, all tech- nical books This includes books for medicine, . . engineering, and so on aThis breakdown is not exhaustive. There are other 5-digit product classes and 7-digit products within SIC 2 7 3 1 and 2 7 3 2. SOURCE: Bureau of the Census (1972), Numerical List of Manufactured Products. Washington, D.C.: U.S. Department of Commerce.
Government Measures of Productivity 65 data on the current-dollar value of total shipments (both primary and secondary products).20 Also available on an annual basis is the current- dollar value of the shipments ("wherever made") of each of the 5-digit product classes that are primary to a 4-digit industry.2i Each 5-digit product class is made up of individual 7-digit products. BEA attempts to obtain price change data for each 7-digit product category, primarily from the PP'. BEA then deflates the current-dollar value of the wherever-made shipments of 5-digit primary product classes, using estimates of the 7-digit product mix of each 5-digit product class based on information from the latest available quinquennial census. The current- and constant-dollar values of all the 5-digit product classes primary to a 4-digit industry are summed and used to make an implicit deflator to apply to the current- dollar value of the industry's total shipments.22 After deflation, the measures of constant-dollar shipments for each of the 450 industries are summed to the 2-digit industry level.23 The same type of deflation procedure is applied to data on the value of the current purchases of materials. However, the level of product detail for which annual data are available is significantly more aggregated for purchases than for shipments. Whereas data on shipments are available at the 5-digit product class level, the current data on purchases are at the industry level in BEA'S 85-order input-output (I-O) table. This level is more detailed than the sac 2-digit industry level. Each of the industries in the 85- order I-O table is composed of more detailed industry categories, and BEA begins the deflation process with those categories. It uses price change data mostly from PA and estimates of the more detailed industry mix based on a more detailed I-O table, which in turn is based on the most recently available quinquennial census. After both shipments and purchases in constant dollars have been summed to the 2-digit industry level, the difference between them (shipments minus purchases) is adjusted to account for changes in real stocks of inventories of both work-in-process and finished goods. This step yields an estimate of real output that is net of the contribution of purchases of materials and supplies but includes the contribution of such purchased services as legal services, transportation, etc., for which industry purchase data are not now available. The last step in the double- deflation procedure is to use the estimates of the differences between shipments and materials purchases, adjusted for inventory change, in both current and constant dollars to compute an implicit deflator, which is then applied to current-dollar estimates of gross product originating derived from data on the income payments originating in the 2-digit industry. This step assumes that the price of unobserved purchased services changes like those of purchased materials. The sum of these deflated 2-digit industry
66 REPORT OF THE PANEL values over all manufacturing becomes the estimate of real gross product originating for manufacturing, which is the numerator of the BES annual productivity measure for that industry division. Currently, the preferred double-deflation procedure is used for manufac- turing, construction, agriculture, railroads, and utilities. For the remaining major industry groups, BEA uses a variety of other methods to measure changes in real product originating. The data on current-dollar value of shipments and purchases and on price changes come from a variety of sources. RELIABILITY OF MEASURED CHANGES The issues of reliability and validity of output and input measures are dealt with in detail in Chapters 5, 6, and 7. One point especially important in connection with the industry division program relates to the problem of estimating changes in the real volume of purchased materials and services used per unit of output. For many of the major industry sectors mining, trade, insurance, finance, real estate, transportation (except railroads), communications, and services~urrent data gaps preclude measurement of changes. Even for industries for which the preferred double-deflation measure is used, existing data sources seldom contain direct data on the current product mix of these purchased inputs. Deflators are constructed from information in past LO tables, which may not reflect the current mix of purchased inputs. DETAILED INDUSTRY PROGRAM BES also publishes labor productivity indexes (i.e., output per labor-hour of all workers) for the 65 2-, 3-, and 4-digit sac industries, shown in Table 4- 6. The included industries are within the industry divisions of manufactur- ing, mining, transportation, communications, public utilities, trade, and services. The 44 measures in the manufacturing division cover 26 percent of all the employment manufacturing, while the 8 measures in the mining division cover 52 percent of all the employment in mining. The remaining 13 measures cover 27 percent of the employment in the other sectors. Although BES publishes productivity measures only for industries for which it has reliable output and input indexes, it maintains 551 unpublished productivity measures covering all 2-, 3-, and 4-digit sac manufacturing industries. The focus of the recent expansion in coverage of the detailed industry program has been in the trade and service sectors. Measures developed in the last few years include such industries as
Government Measures of Productivity 67 gasoline service stations, retail food stores, eating and drinking places, and hotels, motels, and tourist courts. The detailed industry program is the most disaggregated level at which productivity measures are published. These measures provide information on the dispersion in productivity growth among the industries that are included in the more aggregated measures so that leading and lagging industries can be identified. The detailed industry measures also give a company the opportunity to compare its productivity change with that of its industry. The comparison must be made carefully, however, for even companies in the same industry may produce different products or buy or sell products in different markets. (Chapter 8 discusses productivity measurement at the company level.) METHODS AND DATA SOURCES Productivity measures for detailed industries are based on total output rather than on value added as in the industry division program. Because of data limitations, reliable measurement of changes in the real quantity of intermediate purchases at detailed levels of industry disaggregation is difficult, although not impossible (see Chapter 6 and Myers and Nakamura in this volume). For the detailed industry program, BES prepares its own measures of real output, primarily from Census Bureau and PP~ data, and uses measures of labor input based primarily on data from the same Census Bureau survey that provides the data on the value of shipments and inventories in current dollars. In measuring real industry output, BES prefers to combine individual products with labor-requirement weights rather than price weights as in the industry division program. This prevents changes over time in the industry's product or service mix from causing changes in the labor-productivity measure (see Chapter 3~. The concept of labor input is the same one used in the two programs described above the unweighted sum of the hours of all workers in the industry engaged in producing the output measured in the numerator. The output series approximate the preferred BES form as closely as the availability of data permits (see Table 4-6~. Unit labor-hour requirements are used as weights whenever possible. However, data for such weights are unavailable at the product level for many industries, and substitute weights are used. Generally, these are unit values, which are used when they are believed to be proportional to unit labor-hours. For some industries, unit values and employee-hours are used as weights at different levels of aggregation. Unit values are used to combine the industry's individual
68 REPORT OF THE PANEL TABLE 4-6 Methods of Constructing Output Indexes for Published Bureau of Labor Statistics (BES) Measures of Productivity for Detailed Industries Industry Mining Iron mining, crude ore Iron mining, usable ore Copper mining, crude ore Copper mining, recoverable metal Coal mining Bituminous coal and lignite mining, Nonmetallic minerals Crushed and broken stone Manufacturing Canning and preserving Grain mill products Flour and other grain mill products Cereal breakfast foods Rice milling Blended and prepared flour Wet corn milling Prepared feeds for animals and fowls Bakery products Sugar Candy and confectionery products Malt beverages Bottled and canned soft drinks Tobacco products (total) Cigarettes, chewing and smoking tobacco Cigars Hosiery Sawmills and planing mills (general) Paper, paperboard, and pulp mills Corrugated and solid fiber boxes Synthetic fibers Pharmaceutical preparations Paints and allied products Petroleum refining Tires and inner tubes Footwear Glass containers Hydraulic cement Structural clay Clay construction products Clay refractories Concrete products Ready-mixed concrete Steel Annual Output Index Benchmarked to Census Levelsa QU QV QU QU QUb QU QVH QVH DWH QV/QH/DW/DWHC QH/DWH DW QV DW DW DwHc DwHc QVC DWH QVH QV QUC QUC QU QVHC DWH QVH/DWHd QV QH DWH QH QVH QVH QH QV QV DWH DwHc DWH DWC DW QH Yes (QU) Yes (QV) Yes (QU) Yes (QU) Yes (QU) Yes (QU) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes No Yes Yes Yes Yes No Yes No Yes Yes Yes Yes No No
Government Measures of Productivity TABLE 4-6 (Continued) 69 Industry Annual Output Index Benchmarked to Census Levelsa Manufacturing (continued) Gray iron foundries Steel foundries Primary copper, lead, and zinc Primary aluminum Copper rolling and drawing Aluminum rolling and drawing Metal cans Major household appliances Radio and television receiving sets Motor vehicles and equipment Other Railroads, revenue traffic Railroads, car-miles Intercity trucking Intercity trucking (general freight) Air transportation Petroleum pipelines Telephone communications Gas and electric utilities Retail food stores Franchised new car dealers Gasoline service stations Eating and drinking places Hotels and motels DWH DwHc QV QU QH QVH DW DwHc QVH DWH/QV/QH c QVHe QU QVHf QVHf QV QU DWg QVh DCL QU/DCL DC DCL DCL Yes Yes No Yes Yes Yes No Yes Yes No No No No No No No No No Yes (DCL) Yes (QU/DCL) Yes (DC) Yes (DCL) Yes (DCL) QU = Unweighted quantity index. QV = Unit-value weighted quantity index. QH = Unit-employee-hour weighted quantity index. QVH = Unit-value weighted at product level to obtain product class indexes. Product class indexes combined using employee-hour weights. DW = Deflated value using BLS industrial price data at the industry level. DWH = Deflated value using BLS industrial price data at the product class level; product class indexes combined using employee-hour weights. DC = Deflated value using BLS consumer price data at the industry level. DCL = Deflated value for subcomponents of the industry using BLS consumer price data. Subcomponents combined using employee-hour, labor cost, or gross margin weights. aUnless noted in parentheses, the initial stages in benchmark construction utilize data on price change for individual products from the Bureau of Labor Statistics and from the Bureau of the Census to develop constant-dollar primary product class quantity indexes. The primary product class quantity indexes are combined into an industry level index with employee-hour weights. The industry level index of primary products is then ad justed to account for changes in inventories and for changes in the output of secondary products. (Codes in parentheses indicate the exceptions to this procedure and give the type of benchmark index that has been developed using data from the Bureau of the Census.) bThe indexes for bituminous coal and lignite and the indexes for anthracite coal (un published) are combined with production-worker-hour weights.
70 FOOTNOTES TO TABLE 4-6 (continued) REPORT OF THE PANEL CThe industries included in this measure are developed separately. They are combined into an overall measure with industry employee-hour weights. dSeparate indexes of paper production and pulp production are combined with employee-hour weights. eUnit revenue weighted at the freight commodity level; freight ton-miles and revenue passenger miles are combined into an overall measure with labor cost weights. funit value weighted at the freight commodity level; freight ton-miles by class and by type of service are combined with employment weights. gThe price index used for deflation is based on data furnished by the American Tele phone and Telegraph Company and by the Department of Commerce. hThe measures for electric production and for gas production are combined with em ployment weights. SOURCE: Bureau of Labor Statistics, U.S. Department of Labor. 7-digit products into 5-digit product classes, and labor-hours are used to combine the product classes to the industry level. When quantity data are unavailable, the measures used are based on the current value of production adjusted for price change (deflated value). Such measures are conceptually equivalent to measures that combine quantity data with unit-value weights. Measures for some manufacturing and most service and trade industries are based on the deflated-value method. The output series for many industries are developed in two stages. First, benchmark output series are developed from the detailed data in the quinquennial censuses. Second, less comprehensive data, available annual- ly, are used to measure output for the intercensal years. The annual output series are adjusted to the benchmark levels.24 Benchmark Output Series In developing benchmark indexes, BES uses the product detail reported in the economic censuses. It also tries to match the input data, which are reported on an establishment basis, with the output data, which are reported on a wherever-made basis, and to make them more consistent. The benchmark indexes for manufacturing industries are constructed in two steps. First, wherever-made quantity indexes for the 5-digit primary product classes are constructed by combining the 7-digit product detail with base-year unit-value weights (see notes 20 and 21~. If adequate unit- value weights are not available, the current value of each 7-digit product is deflated with a matching BLISS producer price index.25 The primary product quantity indexes are then divided into the current-dollar value of the primary products to obtain wherever-made implicit price deflators for the 5-digit product classes. Second, these deflators are applied to the current
Government Measures of Productivity 71 dollar value of shipments of 5-digit primary product classes made in the industry, as reported in the quinquennial census. The resulting constant- dollar series of product classes are indexed and combined, using 5-digit product class employee-hour weights based on the census data, to develop an industry-level primary product index. This index is then adjusted for differences in inventory change and in primary product coverage between census years, resulting in a final industry output index. The benchmark indexes are used to adjust the annual output series for most manufacturing industries. The benchmark indexes for most mining industries use unweighted tonnage data from the Census of Mineral Industries. For most trade and service industries the benchmark indexes are developed from sales data from the Census of Business. The current-dollar sales data are deflated for various industry components using specifically prepared price indexes based on BES consumer prices. The components are combined to the industry level using labor-requirement weights. Annual Output Measures For some manufacturing industries, physical quantity data by product group are obtained from trade associations and other sources, along with appropriate unit-value or unit labor requirement weights. The product categories are not as detailed as in the quinquennial censuses but are detailed enough for measuring annual changes in real output. For other manufacturing industries, data from the Annual Survey of Manufactures on shipments (wherever made) by 5-digit product class are deflated using PA price change data combined with 7-digit product mix weights from the most recently available quinquennial census. The resulting constant-dollar 5-digit product classes are combined with the current-dollar values into an industry deflator using base-year employee-hour weights. For a few measures, a special BES industry price index is used to deflate shipment data at the 4-digit industry level. The deflated-value measures are adjusted for inventory change, and a coverage adjustment is made to reflect inclusion of secondary products. A few 4-digit measures are constructed using both physical quantity and deflated value for different components of the output series. The components are combined to the industry level using employee-hour weights. Employee-hour weights are generally used to combine the components of those manufacturing industries for measures that are published for levels above the 4-digit level. Like the benchmark indexes, the annual output measures for most mining industries use unweighted tonnage data. For most transportation
72 REPORT OF THE PANEL and utility industries, quantity data at the finest level of detail are used. Unit-value or unit-revenue weights are used to combine the services at the industry or service-group level. Data at the service-group level are combined to the industry level using labor-requirement weights. Output measures for trade and services are developed using the same deflated-value technique discussed in describing the benchmarks. For most industries, retail sales data are deflated for industry components and combined to the industry level using employee-hours, employment, labor- cost, or gross-margin weights. Data on physical quantity for most manufacturing industries come from the Current Industrial Reports, the Census Bureau, or trade associations. Data on quantity for the mining, cement, and some primary metals industries come from the Bureau of Mines in the U.S. Department of the Interior and the Energy Information Administration in the U.S. Depart- ment of Energy. Value data come from the Annual Survey of Manufac- tures. For transportation industries, data on quantity come from regulato- ry agencies. Retail sales data come from the Census Bureau and the Internal Revenue Service (IRS). While for most industries data for productivity measurement are obtained from secondary sources, for a few of the more important industries some data are obtained directly from industry. The motor vehicle industry provides detailed data on the quantity and value of passenger cars and optional equipment by calendar year. The American Iron and Steel Institute provides data for employee-hour weights for all steel products by means of a special survey. Labor Input Measures For manufacturing industries, measures are developed for hours and employment of all employees, both production and nonproduction workers. For mining industries, measures are developed for hours of production workers, but only on employment for all employees. Data to develop these measures come from BES and the Census Bureau, which have different definitions of employee-hours. Census Bureau data exclude paid vacation time, holidays, and sick leave, while BES data include all hours paid for, including those not worked. When there is a choice between BES and Census Bureau data for development of the input series, one important criterion has been compatibility with the output series. Therefore, for most manufacturing industries, Census Bureau data are used for both input and output. But for some manufacturing industries, BES data appear to be more appropriate because of the more frequent collection of data on employment of production and nonproduction
Government Measures of Productivity 73 workers and the inclusion of central administrative and office employees in the industry coverage. For mining industries, BES data are generally used and adjusted to Census Bureau levels. Data on the hours of nonproduction workers are not collected by either the Census Bureau or BES and must therefore be estimated. Since 1968, these hours have been estimated using data collected in the BES biennial surveys of employee compensation. For services and trade, the input series are built up from a number of sources. Employment of nonsupervisory and supervisory workers and nonsupervisory worker hours, which constitute the bulk of the input measure, are based on the Current Employment Survey. Supervisory worker hours come from the Census of Population, and data from the IRS are used for the number of partners and proprietors. Special tabulations from the cPs are used for the hours of partners, proprietors, and unpaid family workers. LIMITATIONS Measures of output per hour are subject to certain limitations. First, although some adjustments are made for changes in the quality of the goods and services produced, existing measurement techniques cannot fully take these changes into account. For purposes of the BES measures of labor productivity, a change in product quality refers to a change in product characteristics that reflects an altered production process requir- ing different base-period labor input. Changes in characteristics that affect the value of the output to the user but do not result from an altered production process having different base-period labor requirements are not incorporated in the current BES measures. Productivity indexes that include both types of quality change will show changes in efficiency not only when unit labor requirements change, but also when the quality of the output changes and unit labor requirements remain constant. Second, except for trade and service industries, the employee-hours data relate to the production of both primary and secondary products, while output data usually cover only primary products. In addition, the data reported for primary products usually include the output of establishments both within and outside the industry. Thus, there can be some discrepancy between the coverage of physical-output and employee-hours measures. This is a serious problem only if there is considerable year-to-year variation in the proportion of primary products to total products of an industry or if there is a change in the proportion of primary products made in other industries. The comparability of the employee-hours and the physical-output data is indicated by the specialization and coverage ratios
74 REPORT OF THE PANEL that the Census Bureau publishes. All the industries for which productivi- ty measures are published have high specialization and coverage ratios. Third, because total output rather than value added is used to measure output, changes in the degree of vertical integration in an industry can cause changes in the productivity measure. If establishments undertake additional operations (such as the manufacture of components that had previously been purchased from suppliers), measured labor-hours will increase but there will be no corresponding increase in final output. In such cases, an index of output per hour would be biased. To learn of possible changes in integration and specialization, BES examines relation- ships such as the change in the cost of materials as a percentage of shipments and conducts plant visits and interviews. Productivity measures are not published it if is believed that changes in integration and specialization bias them. Fourth, indexes using nonproduction worker hours are subject to a wider margin of error than are the indexes using production worker hours because the hours of nonproduction workers must be estimated. Possible errors in the estimates for the hours of nonproduction workers, however, have a relatively small eject on the trend in hours for all employees. PROGRAM TO MEASURE PRODUCTIVITY IN THE FEDERAL GOVERNMENT As noted above, activities of the federal government (except government enterprises) are excluded from the summary measures of productivity. Since 1973, however, BES has operated a measurement program for selected federal government activities. This program prepares indexes of labor productivity for about 65 percent of the 1977 federal civilian work force. This section describes the method of constructing the indexes and the possible use of the indexes in an economy-wide productivity measure. METHODOLOGY AND DATA SOURCES Ideally, a productivity index should relate the final output of a firm, industry, or sector to the inputs that were used in its production. For the private business sector, identifying final product is relatively simple: it is sales to final-demand categories (households and government) and sales of investment goods to firms. For the government sector, for the most part, goods and services are not sold.26 Indeed, for many government functions there are no measurable units of output flowing from resources to final users. For example, all employees of agencies that provide national defense, such as the four military services and the Central Intelligence
Government Measures of Productivity 75 Agency, are helping to produce a service that most of us value highly, but it is virtually impossible as yet to devise measures of change in the amounts of this service for productivity analysis. The same would apply to most of the final output of the U.S. Department of Justice, the U.S. Department of Health, Education, and Welfare, and many other parts of the federal government. The program to measure productivity in the federal government does not attempt to measure final output in this sense. Instead it focuses on work activities within selected agencies. It looks at final output from the point of view of the individual agency and tries to identify readily measurable output for these units, regardless of how far removed they are from the flow of final services. Using this approach, the BES since 1974 has regularly published productivity measures for 20 functional areas within the federal government and an overall measure for the agencies in its sample (see Table 4-7~. The productivity measure is the ratio of a composite output index to an unweighted employee-year index. To construct the composite index for one of the functional groupings in Table 4-7, individual output and labor inputs are aggregated at two levels-at the activity level within agency elements and at the agency element level.27 Organizational elements vary widely in size, number of different activities, and the degree to which the activities identified are removed from the flow of final services from the federal government to the public. Table 4-8 lists some of the agency activities and output indicators for some of the functional categories in Table 4-7. USES AND LIMITATIONS The program to measure productivity in the federal government occupies a unique position among official productivity measures. Although it was initiated for the purpose of helping improve the efficiency of government agencies, it has recently been suggested that these measures could also be used to improve the measurement of real product originating in the federal government sector for the national accounts. To fulfill the initial purpose of the program, BES provides agency managers with data on changes in labor utilization per unit of the agencies' output and with other related statistics. For this purpose, the conceptual issue of what is the final output of government is not important, but problems still exist with the use of agency activities as the output measure. Most federal activities are service oriented, and it is often difficult to specify output indicators. To avoid product mi. problems, output indicators should be specified in sufficient detail to represent a homoge- neous group of services. For example, in measuring the output of a motor
76 sit a' o o ._ 5~ no a, sit Cal Cal 5~ too, Cal o o o _ , 11 - a' o C`d ~ C~ o ~ V~ _, X a' ~ C;~` c: _ ~ ~, C~ _ C~ ~; V) ~ ~ _ ~ Ct Ce C~ _ C~ C~ ~ o ._ a' o ._ ._ ¢ ~ ._ Ct - . _ ~ - Ct V: V C) _ Ce , ~ ~ ~ C~ C~ V, . r. cr~ _ Ch - U~ - - C~ _ _ _ C~ _ O _ - 00 cr. - _ _ ~_ V, V3 0 ~=5 C: ._ ~ t>~) ~ C) ~ ~ C'3 _ O ~ ~ . ~ 0 ° - ~ ~ _ ~ ~ oo ~ oN ~ . . . ~oo ~ ~ ~- ~ ~ - - - ~ ~ u~ . . . . ~ ~ co ~- - oo - - - oo o ~ ~ . . . . o ~ k a~ - - - ~ ~ oo ~ ~ ~ ~ ~ - ~ - o - ~ o o ~ ~ o o ) u~ o ~ ~ o . .. . . . . . . . . . c~ cr~ ~ 0 ~ ~ ~ ~ ~ ~ ° ~ o- o c~ ~ o o ~ ~ o o - - - - - - - - - - - - oo ~ ~ cr ~ o~ ~ _ ~ u~ ~ ~ _ _ ~ o - ~ o o ~ ~ o - - - - - - - - - ~ O ~ ~ - O oo ~D - ~ . . . . . . . . . . oo cr oo ~ ~ ~ 0 u~ ~ o - ~ o o ~ ~ o ~ - - - - - - - - - . . ~ o - - o oo oo - ~ o ~ ~ . . . . . . . . crN '~ ~ => - oo oo o o ~ o o ~ ~ ~ oN - - - - - - o o ~oo - ~ 5- oo =) ~ ~, .. . . . . . . . . . . . O C~4 00 ~ ~ O C~ ~ ~ ~ ~ O =, - ~ -) O~ O - - - - - - - - ~ ~c~ . . . 1 0 cr - ,. - ~ kD ~ C~ _ ~ 00 - oo ~ ~ ~ ~ ~o ~ ~ ~ c~ o o ~ o - - -) o"~ ON - - - - - - - - oo ~oo ~ ~ ~ ~ - ~ ~ ~ u~ o ~1 _ ~ o oo ~ ~ ~ ~ ~ 0 o oN - - - - - - - - - - - ~ c~-)-O~ (r~ ON ~ ~ ~) - - ~ ~ .... . . . . . . . . . 1 -~ ~ ~ ~ ~ ~o ~ ~ _ _ - ~ o o c~ o o - ~ o o - - - - - - - - - - - ~-~ o4 ~ ~ o . .. . . . . 1 ooo ~ ~ O ~ ~o- o o - o -- - - - - ~0c~ 0 ~ oo r~ . . .. . . . . u ~oo 1 o~ 0 c~ ~ c~ o oN oo o o o o - -- - - - - o o o o . . . . o ° 1 1° 1 ° o o o o - - - - ~ o ~ :: ~ ~ o ~ . - . - ) c~ ~ ~ . ~ ~ - _> ~ ~ . _ - c~ c: - 1 ~ =1 u~ ~ o ~ ~ - ~ o ~ ~ o - - o o - - - - - ~ _ ~ cr ~ . . . . . cr ~ u~ _ ~ ~ o o o o - - - - o o o o o o o . . . . . . . o o o o o o o o o o o o o o - - - - - - - e~o . - . - L. ) o v, . c~ - ~ ~ o ~ ~ ~ o ~ ~ · - c) - ct c.) ct ~- ~- ~ ~ v: c~ - v) cD ~ c) t c) ~ oD :> c) v~ C >~ ~ ~ .O ~ - o ~ c~ c~ c) .- G~ v) c~ ~D cd - - .
77 ~ ~ o _ ~ CO oo . . . . . . . _ ~ ~ _ _ _ ~ ~ ~ C~ ~ ~ o o . . . . . . . o~ ~ ~ o o ~ ~ _ oo _ _ _ _ _ _ _ ~o ~ _ o ~ ~ . . . . . . . O ~ ~ ~ ~D - ~- _ ~ _ o oo _ ~ _ _ _ _ _ _ ~ ~ ~ C~ ~ o ~ . . . . . . . o~ oo oo ~ ~ oo ~ o C~ _ o oo o _ _ _ _ _ _ _ 00 ~D - <7\ -, v~ _~ . . . . . . .. o _ ~' ', ', ~ Cr~ _ ~ ~ o oo o o_ _ _ _ _ _ __ _ ~ _ ~ ~ oo oo . . . . . . . oo oo oo _ - - _ O oo cr _ _ _ _ _ _ ~ oo o ~ ~ U~ ~ . . . . . . . o ~ CO ~ ~ U) ~ _ _ _ o oo ~ _ _ _ _ _ _ ~ o oo ~ oo o oo . . . . . . . ~ ~ oo ~ ~ ~ _ o _ _ o oo ~ o _ _ _ _ _ ~ o ~ ~ o _ ~_ . . . . . . . . _ o ~ ~ ~ ~ o o o ~ Cr~ oo o o _ _ _ _ _ ~ ~ ~ C~ ~ ~ ~ . . . . . . . o _ _ oo ~ ') ~ 0 0 0 ~ ~ cr. 0 _ _ _ _ C~ ~ ~ o C~ o _ CO ~D oo ~ ~ 00 O O O cr ~ cr o~ _ _ _ o o o o o o o . . . . . . . o o o o o o o o o o o o o o _ _ _ _ _ _ _ o . ~ - ~ ~ o o ~ fi ~ ~ t - ~ hO ~ Ce C) ~ ~_ Ce _ _ ~3 ~ ~ a bo _ - , ~ oo _ ~ ~ ~ . . . . . . . . ( ~C~ ~ _ ~ 0 ~ ~ ~ 0 . . . . . . . . 0 C~ ~ 0 ~ oo ~ 0 ~ _ _ _ _ _ _ _ _ _ ~ ox ~ ~ ~ _ ~ . . . . . . . . _ ~ ~ ~ V~ ~ 0 ~ _ _ ____ __ ~ ~ ~ - 0 ~ ~ . . . . . . . . ~ ~ cr ~ ~ ~ 0 _ ~ ~ _ ~ ON ~ _ _ ____ __ . . . . . . . ~ o~ 0 ~ ~ ~ oo ~ _ ~ - C~ ~ O ____ __ ~ ~ a~ ~ ~D ~ c~ . . . . . . . . ~ - 0 ~ ~ ~ ~ _ ~ ~ ~ _ cr~ ~ 0 _ ____ __ oo _ ~ - ~ 0 0 ~ . . . . . . . . _ ~ ~ C~ 0 oo _ C~ - ~ - 0 - O _ _ _ ~ _ _ _ _ O ~ ~ ~ -~ ~D . . . . .. . _ o ~ 0 ~41 ~ ~ _ ~ _ ~ __ 0 _ ______ ~ _ - C~ O . . . .. . ~ oo oo 01 ~ e~ _ 0 _ _- 0 ______ _ 0 ~ C~ ~ ~ . . . . .. . \~) a~ ~ _ oo1 v~ ~ 0 0 0 - -- 0 _ ______ _ e' O oO ~1 o - 0 0 0 0 0_ 0 _ ______ 0 0 0 0 00 0 . . . . .. . o 0 0 0 0 1 o 0 0 0 0 0 0 0 0 _ ____ __ - o C,4 ~ .~: 0 ~ _ ^` .9 ~ O c, E.= ~a.O t4 ~ ,:' ~ ~ - - o CL ~ ~ ~ O ~ ~ ~ ~ ~ ~ ~C~ V~ V) ~ ~ ~ 5 Ct .0 Ct ~ £ £ ~ ~ o C~ ~, ~£, ~ ~ r · c~ 0 _ 0 - ~ \~) ° 11 C~ 11 ~ oo ~ C~ ~ ~ ~1 r ~ C~ Ct o~ Ct _ ~ _ ~ _ Ct _ 5) Ct ~ e~ ;^ ~ >, Ct---= .~ .U) .U~ .C~ ~ t.0 V) ba Ct ,_ Ct ~ D ct 5: ct ~ S C.) ' C) ~ C) C ~ O ~ O t ~ C5 ~ ~ X 13 ~ ~ bo S o U, O ~o C ~Ct 5 ~ O V, ~ Ct ~ C> C) _ Ct , C CL _ . ~ < ~r 0 r _ 0 ,-] ~ x "a b4, ~ u, ~ C ~ ON --- ~ ~ _ .`,, D ~ X =~,- Ce . ,_ ~ ~ O O-- ~o V:1 ~ O S Ct ~ ~ ~ .. S ~ O ~ ._ C) ~ ~ ~ _ ~ 0~ ~ ~
78 REPORT OF THE PANEL TABLE 4-8 Selected Activities and Output Indicators, by Selected Functional Groups, from the Bureau of Labor Statistics' Federal Government Productivity Program Finance and accounts Regulation-inspection and enforcement and rulemaking and li- censlng Commissary store operations; laundry and dry cleaning Dining facilities operations; recruit training Flight training Active duty personnel, pay ac- counts (Navy); civilian time cards (shipyard, Norfolk) Invoices and travel processing (shipyard, Norfolk); billing, bill and collect for GSA Accounting obligations and expenditures (USIA); pay- ment and reconciliation of of checks (Treasury) Background checks on persons assuming sensitive positions in Department of Defense; develop and publish rail safety standards (DOT) Inspections of manufacturing plants, retail stores, etc. (Consumer Product Safety Comm.); consumer com- plaints (CAB) New animal drug review and approval process (FDA); patent application exami- nation. Functional Grouping Military base services and military training Activity Output Indicator Deflated dollar value of sales; pieces processed Number of meals served; students enrolled Student-years trained Number of active duty ac- counts maintained; num- ber of time cards processed Number of invoices and travel claims processed; bills mailed Obligation and expendi- ture documents processed; checks paid and reconciled and tax deposit forms processed Number of cases closed Number of new standards and modifications com- pleted Number Inspections made Number of consumer com- plaints processed Number of new animal drug applications; ap- plication disposed DOT = Department of Transportation; GSA = General Services Administration; CAB = Civil Aeronautics Board; USIA = U.S. Information Agency; FDA = Federal Drug Ad- ministration. SOURCE: Productivity Programs in the Federal Government. Supplement to Vol. 1. The Measurement Data Base. Annual Report to the President and the Congress by the Joint Financial Management Improvement Program.
Government Measures of Productivity 79 pool garage, categories such as number of tune-ups, number of oil changes, etc., are preferable to the single category "number of cars serviced." In addition, as with other productivity measures, even the most detailed output can change in quality over time. The main approach used to minimize these difficulties is to collect data with as much output detail as possible. There are differences of opinion about whether productivity measures can be used to improve the current measure of real output for the federal government in the national accounts. BEA'S current method of measuring government output assumes that no productivity change takes place in the public sector; therefore, BES must exclude this sector in calculating its aggregate productivity measure. Some analysts note that almost all of the output measures are of intermediate activities that, although needed if the government sector is to produce anything, do not relate directly to the final services that taxpayers expect for their tax dollar national security, improved allocation of resources, etc.28 When any agency provides its intermediate services more efficiently, this only means that the potential for increasing overall government productivity has improved: we say "potential" because there is no direct link between increases in productivity at the intermediate level and at the level of final output. This may be shown by a hypothetical example. One agency produces its output faster than before-getting out a report for the use of another agency. But the second agency cannot implement the report any faster. Therefore a chain of events that might have reduced the amount of labor needed to produce the same level of national defense, or improved resource allocation, is choked oh. There is no way to confirm such a situation because there is no measure of final output. Other analysts stress that, for some programs, activities and output measures are closer to the flow of final services than for others, and that with further research and development the measures might be made worth using (see Searle and Waite 1980~. EXCLUSIONS FROM EXISTING MEASURES Several sectors of the economy are omitted from the official measures of output per hour. Some sectors included in GNP are excluded from productivity measures because BEA estimates their output by labor (and sometimes other) inputs. This results in essentially no measured productiv- ity change for these sectors (see the first part of Chapter 5~. General government, including federal, state, and local governments, is the largest of the excluded sectors. (As discussed above, the BES has a program to measure the productivity of the federal government that could, in
80 REPORT OF THE PANEL principle, be used to extend the coverage of the aggregate productivity measures.) Other sectors omitted for this reason include not-for-profit institutions, like private hospitals and universities, and domestic services purchased by households. Still other sectors are excluded both from GNP and from productivity measures. The largest of these is the household sector, which produces such output as child care, using nonpurchased inputs. The GNP measure also does not include, as a deduction from output, the effects of environmental pollution, congestion, or noise. This section describes measurement efforts by BEA and by private investigators for the omitted sectors; for further discussion of sectors and effects not included in GNP, see Moss (in this volume). STATE AND LOCAL GOVERNMENT There is evidence of widespread interest in and use of productivity measures by state and local governments throughout the country. One of the main activities of the National Center for Productivity and Quality of Working Life (see its report 1975) was to help make state and local government managers aware of organizational and other innovations, including the use of productivity measurement, that would lead to cost savings and enhance productivity. There is an important difference between state and local governments and the federal government in the potential for developing productivity measures that could be used to expand the economy-wide productivity measures, namely, that most activities of state and local governments deliver final services directly to taxpayers. This makes it possible to measure quantities of those final services. Activities such as garbage collection, recreation services, library services, public transportation, water supply, and elementary and secondary schooling deliver a service of value to final consumers. Some kind of quantity measure, such as tons of garbage collected, number of visitors to the zoo, or number of passenger- miles traveled, is conceptually straightforward and possible to develop. There remains the problem of quality change. Work currently under way at the Urban Institute seeks to develop objective measures of the quality dimensions of the services of state and local governments (see Hatry et al. 1977~. That work suggests that two quality characteristics for garbage collection, "spillage of garbage collections" and "damage to private property by collection crews," could be measured by means of household and business sample surveys. Changes in tons of garbage collected is a good measure of real output change if the responses to the questions about these characteristics do not change. If the responses to the
Government Measures of Productivity 81 quality questions do change, the problem remains of how to value the change in quality in terms of the appropriate amount of adjustment to the quantity measure. Of course, there is an analogous problem in the private sector whenever there are changes in the performance charactistics of goods and services sold. Private sales in current dollars are deflated by price change data that may not adequately allow for changes in the quality of the service sold. (For further discussion of measuring quality change in output, see Chapter 5.) NOT-FOR-PROFIT SECTOR Hospitals and colleges and universities are the two major industries in the not-for-profit sector. These organizations, although not-for-profit, have become increasingly concerned about cost and efficiency. There is an extensive literature by educators, medical and operations research special- ists, and economists on output, input, and productivity measurement and analysis for many aspects of hospital and university operations (see Froomkin et al. 1976, Institute of Medicine 1976, Stanford Center for Health Care Research 1976, Feldstein and Taylor 1977~. Much of the work (like that of the BES program for federal agencies, described above) is aimed at helping not-for-profit organizations improve their own efficiency rather than at developing measures that would permit comprehensive comparisons of productivity change between these sectors and the private business sector. One exception in higher education has been the use of weighted indexes of courses taken, using credit hours to aggregate the various kinds of course enrollments, such as full-time, part-time, degree credit, and nondegree credit (see O'Neill 1976~. The use of such output measures (and analogous measures, such as patient days, for hospitals) is similar to using tons of garbage collected to measure the output of garbage collectors; it is satisfactory as long as the quality of the basic unit of quantity does not change. This is particularly important for schools and hospitals because changes in the output quantity measure can themselves be associated with quality changes. If improve- ments in surgical procedures shorten the length of a hospital stay, they might result in lower output as measured by average patient load, and the productivity measure based on this output could decrease. If class size is significantly increased to reduce expenditures and the quality of instruc- tion falls, the output measured by credit hours would not change and the productivity ratio would increase. Indicators of changes in various quality dimensions (e.g., mortality rates adjusted by case mix, length of stay by type of ailment, class size, etc.) are needed to supplement pure quantity
82 REPORT OF THE PANEL measures. (For further discussion on measuring output in hospitals, see Scott in this volume.) HOUSEHOLDS In recent years there has been much work by economists analyzing the family as a producing unit. A behavioral model is used in which the goods purchased from business firms are considered intermediate inputs that are then combined within the family with other "inputs" the non-market time of family members to produce the final services of ultimate interest (good health, vacations abroad, dinner or theatre evenings out, etch. This approach is an outgrowth of earlier work by labor economists who, in analyzing trends and cyclical movements in the labor force participation of women and teenagers, were forced to expand their models of labor supply to include the interrelationships of the work and leisure choices of all family members. A related topic involving the use of consumers' time is the time used in obtaining market goods and services. For many goods and services, such as medical care, repair services, travel, banking, and shopping, the input of the time of the consumer is an important factor of production. Entrepre- neurs are aware of this in making decisions about resource allocation, location, and technical change. Despite the widespread interest in non-market uses of time, there has as yet been no systematic effort by government agencies to develop data series that measure the utilization and productivity of non-market time, either of time used in conjunction with purchased goods and services or of time used to purchase goods and services. There is one nongovernmental project of sufficient size and scope that it may produce information about the productivity of non-market time although it is not directly aimed at measuring productivity change. This is a project by researchers at six institutions called "Cooperative Research in Goals Accounting.''29 Its primary objective is to measure changes in various broad aspects of"well-being" (such as life expectancy, mental health, and safety) and to relate these changes to a number of categories of input: purchased consumer goods and services, expenditures and services of government, and the non-market time input of individuals. The project also seeks to explore exogenous factors that influence such broad goals as mental and physical health, including factors involving demographic and mobility trends. (These objectives go beyond the usual province of analysis of household productivity the saving of time and market goods in such activities as doing housework or in shopping.) Some of the work of the project, which combines survey data on the uses of non-market time by
Government Measures of Productivity 83 type of activity with national accounts data on purchased goods and services, may produce measures of secular changes in the time devoted to various activities. Moss (in this volume) discusses traditional measures of national income and broader measures of welfare. EXTERNAL EFFECTS AND OUTPUT The erects of production on the environment and on worker health and safety are now of great concern. Regulation of these erects is now pervasive and is widely believed to have had a significant erect on the change in measured productivity. To incorporate these effects fully into a productivity measure would require estimates of how much of every industry's activities has changed the quality of the environment at different times. This erect could be positive, as when regulations require firms to reduce their level of air or water pollution, or it could be negative or zero. The information could then be used to adjust the traditional measure of the industry's real output, which measures only its output of conventional goods.30 The change in the ratio of an index of adjusted output to an index of inputs-those devoted solely to producing conventional output and those being used specifically to influence the effects on the environment would yield a measure of productivity that fully reflected the ability of the industry to produce environmental erects in addition to conventional goods and services. It is not possible to perform this type of analysis with data now available. However, BEA iS planning to develop such information as part of its new Environmental and Nonmarket Economics Division. The data currently available primarily cover the resource costs that firms have incurred in meeting specific environmental standards imposed by regulatory agencies. Although these data alone do not allow for the computation of a comprehensive productivity ratio of the kind discussed above, they can be used to correct conventional productivity measures for the erects of requiring the firm to shift partially into the production of some favorable environmental output. These corrected ratios measure the trends in productivity that would have occurred if resources had all been used to produce conventional output and none had been used to reduce external environmental effects. Some researchers have used the available data to make such adjustments and thus to improve understanding of the trends in measured productivity in the period since 1966, which was one of rapidly increasing environmental and health and safety regulation. Chapter 7, which reviews the sources of measured productivity change, discusses the data and methods used in estimating these erects.
84 REPORT OF THE PANEL Recommendation 5. The Panel recommends that research on the measurement of the output and productivity of the resources in excluded sectors be expanded. However, there should be no prema- ture selection or foreclosing of any of the alternative measures of output for such systems as health care and higher education. Although progress has been made in measurement in these areas, we do not yet know enough about the operation of such systems either to measure precisely their salient outcomes or to assume that we understand the processes that account for them. NOTES 1. The Economics, Statistics, and Cooperatives Service of the U.S. Depart- ment of Agriculture is the only other major government producer of productivity measures. It publishes both labor and multi-factor productivity measures for the farm sector. Since a task force of the American Agricultural Economics Association is currently assessing the data and methodology underlying USDA'S measures, our study did not include a review of those measures. 2. BES does develop measures of physical capital input (and other inputs, such as weighted labor input) on a periodic basis in order to analyze the factors underlying changes in labor productivity (see Norsworthy and Fulco 1977), but such measures are not published on an ongoing basis. 3. BES experimented with collecting all the productivity data from a single survey; the cost and complexity of the statistical design made the approach infeasible (see Mark 1961~. 4. BES does not publish any economy-wide productivity measures that include the not-for-profit and government sectors because there are no independent measures of output for those sectors (see Chapter 5~. 5. BES also publishes series on other, less-inclusive, aggregate measures: for the private nonfarm economy and for all non-financial corporations. Our discussion focuses on the measures for the private business sector. 6. This statement is not precisely true in all cases. The output measure used by BEA for employees of government involves a weighted sum in which senior and higher-grade employees count for more labor. Since the BES labor input measure is an unweighted sum, the resulting productivity ratio would show increases whenever the experience/skill mix increased. However, it is generally recognized that using inputs, even with this kind of adjustment, to measure real output is of limited value. There is interest in finding more direct measures of government output (see the section below on measuring productivity in the federal govern- ment). 7. The final-demand categories include sales of goods and services by business firms to households, other business firms on capital account (including the net changes in inventories of all firms), government, and the rest of the world (exports minus imports). The only business sales that are excluded are those that go to another firm on current account, so-called intermediate purchases. Gross product originating for an industry is defined as the value of its sales (adjusted for net
Government Measures of Productivity 85 change in inventories) minus all purchases from other firms on current account. If this quantity is summed over all industries, then all intermediate purchases and sales will drop out and what is left are the sales (net of inventory change) by firms to the final-demand categories. 8. There is a third way of measuring business product based on the accounting identity between total value of production and the sum of payments to the factors of production (including any indirect taxes and including a profit component defined so that the two sides will always balance). When these factor payments are summed over all finals, they will be equal, conceptually, to current-dollar business product measured as the sum of sales to the final-demand categories. In practice, they differ by an amount called the "statistical discrepancy." Thus, the factor- payments data could be used to estimate current-dollar business product, and real product can be derived by deflating current-dollar product by the ratio of current- to constant-dollar business product derived from the data on sales to final-demand categories. In effect, one first estimates real business product via the sales to final- demand approach and then uses this information to form a deflator to apply to the current-dollar product estimate derived from the income side. 9. The percentages in the text show the share of each category in final demand after subtracting estimates of the contribution to these categories of the non- business sectors and of the resources located in the rest of the world. Thus, total purchases by government in 1976 were $264.4 billion (in 1972 dollars) of which $145.8 billion was for direct hire of employees and $1 18.6 billion was for purchases of goods and services from private business. The percentage above for government reflects only the $1 18.6 billion, since we are excluding sectors with no independent measure of output (Survey of Current Business 1977, Tables 1.2 and 1.8~. 10. The most comprehensive data sources are quinquennial censuses, and although they are listed at the end of the sequence, they also play an important role in making the preliminary estimates for many subsequent years. For example, the data available from monthly and annual surveys to estimate current-dollar PCE and PDE do not contain enough product detail for carrying out the deflation process for conversion to real product. These more detailed current-dollar components are estimated using information contained in the large input-output (I-O) table maintained by BEA. The detailed cells of the I-O table, showing product shipments and purchases by detailed industry, are revised on the basis of information in the quinquennial censuses. Thus, when the data from the 1967 censuses became available in 1976, it meant that all the product estimates in the period 196~1976 were subject to revision. 11. For the hotel, personal, and recreational categories, the monthly Census Bureau survey is used for all the non-benchmark estimates-quarterly, and the first, second, and third July revisions. For medical and legal services the monthly survey is used only for the quarterly and first July revisions. The second and third revisions are based on the business receipts data in the Internal Revenue Service's Statistics of Income. 12. BEA uses a deflation procedure to obtain what it calls "output in constant dollars." This terminology is somewhat misleading because the procedure used by BEA iS a mixture of a constant-dollar quantity index (with base-year price weights) and a deflated-value index using price indexes with base-year quantity weights. BEA divides the current-dollar value of expenditures into individual product classes and
86 REPORT OF THE PANEL deflates each one by a price index. Each of these indexes is weighted by the relative quantities of individual products within the product class in the base year. If there was no change in product mix, the deflated value for each product class would precisely equal the value, in base-year prices, of the current-year mix of individual products, and the sum over all product classes would be equal to current output at base-year prices, i.e., "output in constant dollars." In general, however, the product mix of product classes changes, the amount of change being larger the less detailed the deflation categories. BEA iS aware of this problem and uses the most detailed categories. 13. Information on the actual detailed product breakdown is not in the quarterly and annual surveys that underlie the various July revisions. These detailed components are estimated using information contained in the large I-O table maintained by BEA (see note 10~. 14. The operating expense categories are paper, printing, telephone and telegraph, utilities, rent, computer leasing, advertising, medical service, postage, business travel and expense, and wages. Price change data for each of these categories are obtained from the producer price index (PP~), the cat, and other sources. The price change data are then used to deflate the current-dollar value in each category. 15. Some government services are sold directly to consumers, for example, those of the U.S. Postal Service. These activities, called "government enterprises" by BEA, are treated like private firms in the national accounts and are included by BES in the input and output of the private business sector. 16. The divisions of the private business sector are farming, mining, construc- tion, manufacturing, transportation, communications, utilities, wholesale and retail trade, finance, insurance and real estate, and services. 17. BES does not publish measures for the other divisions because the available measures of output either are based primarily on inputs, as in finance and life insurance, or are not considered reliable enough to justify publication on a regular basis. 18. Industry measures of labor productivity based on a value-added concept of industry output have some desirable properties compared with industry measures based on total output. Value-added measures do not change when the vertical integration structure of the industry changes, and they move in the right direction when firms in the industry economize on purchased inputs per unit of output. Neither of these properties is shared by measures based on total output. However, for purposes of analyzing the sources of an industry's output growth, its cost and price behavior, etc., a multi-factor productivity measure relating total output to all inputs (both intermediate purchases and labor and capital) is more relevant. 19. This description focuses on the internal logic of the double-deflation method and on how suitable the available data breakdowns (e.g., level of product detail available on a year-to-year basis) are for this methodology. The ability of the price change data to allow for changes in the performance characteristics of products and services within the deflation categories used is discussed in Chapter 5. 20. Primary products are the 5-digit product classes or 7-digit products listed under an industry in the Census Bureau's Numerical List of Manufactured Products (Bureau of the Census 1972; see Table 4-5~. Products of an industry not on this list are shown as secondary products.
Government Measures of Productivity 87 21. Wherever-made shipments include those from plants where the products are primary and from plants where they are secondary. Deflators calculated with these shipment data are called "wherever-made deflators" and are based on the product mix of all industries where the products are made. The use of these deflators for industries where the products are primary and where the mix may differ can result in some error. 22. For some 4-digit industries, shipments of secondary products are sig- nificant. For these industries, the 5-digit product class mix of their secondary products has to be estimated from the latest available quinquennial census. Also, if an industry has a low "coverage ratio" of its primary products (e.g., the industry's shipments of its primary products are less than 75 percent of the wherever-made shipments), the industry deflator, which is based on wherever-made weights, may have a large error if price behavior differed between primary producers and other producers (see Myers and Nakamura in this volume). 23. There are 20 2-digit industry categories in manufacturing, such as food and kindred products (sac 20), tobacco manufactures (sac 21), and textile mill products (sac 22). 24. In connection with output measures based on the quinquennial censuses, it should be noted that the Census Bureau publishes a volume showing real output indexes for all 4-digit manufacturing industries (U.S. Bureau of the Census, Census of Manufactures, Volume IV, Indexes of Production). See Myers and Nakamura (in this volume) for a discussion of the methods used by the Census Bureau. 25. Very often, the match between a Census Bureau 7-digit product line and the product specification used by the PP~ is approximate (Ruggles 1977~. 26. See note 15. 27. A total of about 1,600 individual output indicators are aggregated using their respective base-year unit labor requirements for weights. 28. See the dissenting comment by Denison in Office of Federal Statistical Policy and Standards (1977, p. 8~. 29. The project participants are Nestor Terleckyj, National Planning Associa- tion; Abraham Charnes, University of Texas; William Cooper, Carnegie-Mellon University; F. Thomas Juster, University of Michigan; Michael Levy, The Conference Board; and Milton Moss, Stanford Research Institute. 30. Such an adjustment would require some judgment about the value to consumers of the increment in environmental quality.