Click for next page ( 392


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 391
Data Adequacy for Productivity Analysis: A Case Study of the Primary Paper Industry JOHN G. MYERS Southern Illinois University at Carbondale and LEONARD NAKAMURA Citibank INTRODUCTION In this study we examine the adequacy of data sources for measuring and analyzing productivity change in the primary paper industries. We have chosen this small group of industries for several reasons. Together, they are highly specialized and their products are comparatively homogeneous, so that problems of quality change and new products, although present, are not as severe as in many fashion goods or durable goods industries. Also, the small number of industries in the group permits us to examine output and input measures at the lowest level of aggregation with current data sources. In this way, data limitations are revealed that would be missed at a higher level of aggregation. Finally, output estimates for this group of industries are prepared by four different government agencies. This means that we can examine several data sources and measurement methods. This report covers only the measurement of output and materials input. Productivity analysis also requires measurement of inputs of labor, capital, and purchased services. Their omission does not imply that there This is an abbreviated version of a larger study. The complete report is available through the National Technical Information Service, Springfield, Va. (PB-292 287 OGA). These are pulp mills (sac 2611); paper mills, except building paper (sac 2621); paperboard mills (sac 2631); and building paper and board mills (sac 2661). 391

OCR for page 391
392 PAPERS are no problems in these areas.2 Because of space limitations, we decided to review what appears to be the two most important elements of produc- tivity measurement for an industry. We include materials input rather than labor or capital input because of the pivotal role that materials play in productivity analysis at the industry level. Whether the analyst uses a value-added model or a total output model, reliable information is needed on materials and other inputs purchased from other establishments. For the value-added model, purchased inputs are subtracted from total output to obtain the value-added measure; for the total output model they are used as an input along with labor and capital. Thus, if the industry analyst can obtain good measures of output and intermediate inputs, he will have come a long way toward a useful industry productivity study. MEASURES OF OUTPUT There are four sources of primary data on output for the industries we are examining. They appear in the following sequence: weekly (followed by monthly and annual) figures, collected by the American Paper Institute (APl)3 on a product (wherever produced) basis; monthly and annual figures collected by the Census Bureau and published in Current Industrial Re- ports (CIR),4 also on a product basis; annual figures collected in the Annual Survey of Manufactures (ASM), on an industry basis; and quinquennial figures collected in the Census of Manufactures (CM), also on an in- dustry basis. The APT data are adjusted using the CIR data on bench- marks; they may therefore be considered to be preliminary estimates, valuable mainly for their timeliness. Because the coverage and specialization ratios are very high for the four paper industries taken together, the CIR annual product data move very similarly to the ASM data from year to year. However, the CIR data include a large amount of intermediate products (especially pulp), and there are trends in the ratio of intermediate to final products of the industries, in part due to physical integration of pulping with paper and paperboard mills. This means that product data will differ systematically and substan- tially from industry data over periods of more than a few years. When in- dividual industries are considered (rather than the four primary paper in 2For example, at present no annual data are collected on the services that establishments purchase from other establishments, such as legal and telephone services. 3AP! weekly figures are published in a Weekly Statistical Report, monthly and annual figures in the annual Statistics of Paper. 4Current Industrial Reports. Pulp, Paper. Paperboard and Building Paper Mills, Series M26-A.

OCR for page 391
Data Adequacyfor Productivity Analysis 393 dustries combined), product data can differ from industry data even more widely over time. For productivity analysis, therefore, where inputs are almost invariably collected on an industry basis, product data such as CIR will serve only as preliminary estimates until industry output data from CM and ASM become available. CM data differ from ASM data in three related ways: (1) the former are published in much more detail, making it possible to derive more precise summary measures of industry output; (2) because CM coverage is much more extensive, CM data suffer less from sampling variability; and (3) CM data are published with a much greater time lag because of the greater bulk of information to be processed and checked for accuracy and con- sistency before publication. Four secondary measures of output are prepared by federal government agencies for the industries we are studying. These are the CM indexes, prepared by the Census Bureau (referred to hereafter as Census), the (un- published) Bureau of Economic Analysis (BEA) output measures, the Bureau of Labor Statistics (BES) indexes, and the Federal Reserve Board (FRB) indexes. The CM indexes appear last and are based on the most com- prehensive data source; they are discussed first. Discussions of the BEA, BES, and FRB measures follow. Aside from timeliness and data sources, the four sets of indexes differ in method of computation, type of deflator, weighting system, and whether the index is on a product or industry basis. In the following discussion, special attention is given to these differences in concept and procedure, and to their consequences. Other problems examined are those of integra- tion and weighting bias, benchmark adjustments, and procedural incon- sistencies. CENSUS PRODUCTION INDEXES Beginning with the 1947 Census of Manufactures, the Census has prepared indexes to measure the change in production from the preceding census for each industry, a number of industry groupings, and total manufacturing. Although these indexes are usually published 5-7 years after the census year,5 they are still widely used as review benchmarks for other government indexes of production, including the FRB and Bus in- dexes. The belated appearance of the indexes is due primarily to the care with which they are constructed. Lengthy descriptions of the estimation pro sThe indexes for the 1972 Census of Manufacturers, showing changes from 1967, were published in December 1977.

OCR for page 391
394 PAPERS cedures are published with each set of indexes.6 We will restrict our discussion to a brief summary and concentrate on aspects particularly relevant to this study. Computational Methods Deflators are prepared for each 5-digit product class. The prices used, which correspond to the 7-digit products that make up the product class, represent one of the following: unit values drawn from CM itself or CiR, BUS wholesale prices, 7 or price data from other sources. A choice among these is made after an extensive analysis and review by a committee com- posed of representatives of several government agencies. The analysis in- cludes the preparation of distributions of unit values, for 7-digit products, drawn from individual establishment reports. If it is decided that no satisfactory prices are available for a S-digit product class, a deflator at a higher level of aggregation is used. The price relative of each 7-digit product is weighted by the relative im- portance of the product's shipments (wherever made) to obtain a 5-digit product class deflator. The S-digit deflators are used to deflate the value of the industry's shipments of each corresponding primary product class. Major miscellaneous receipts from secondary products are separately deflated. The deflated product-class shipments are then aggregated to the 4-digit (industry) level.8 A broad 4-digit wherever-made deflator is then derived and used to deflate the value of "other" miscellaneous receipts and a few additional small categories. An important characteristic of the Census deflation process is that the secondary products of an industry are deflated by prices that correspond directly to those products. This is preferable to the alternative procedure, which is to assume that the price behavior of secondary products follows that of primary products. In the four industries we are studying, no deflators in the category of price data from other sources were used for the 1967-1972 indexes. The distribution of deflators by type for these industries is shown in Table 1. It is noteworthy that the results of a lengthy and careful interagency review 6 See Bureau of the Census (1972, pp. 10-19). Our summary of estimation methods is based on this publication and conversations with Census Bureau employees. 7 The name of the wholesale price index has recently been changed to the producer price index; see Early (1978). We will use the term "producer prices" from now on. sin a few industries (66 in 1972), value-added weights are used to aggregate from the 5-digit to the 4-digit level. Paperboard mills (sac 2631) is the only one of the four primary paper industries for which this alternative procedure was followed in 1972.

OCR for page 391
Data Adequacyfor Productivity Analysis TABLE 1 Source of Deflators Used in Construction of the 1972 Census Production Indexes for the Primary Paper Industries (percentage distribution) BUS Next Higher Unit Producer Level of sac Value Prices Aggregation Total 26 52 43 5 1 00 2611 90 5 5 100 2621 58 38 3 100 2631 89 10 1 100 2661 0 100 0 100 395 led to these choices of deflators. As described below, these choices are in sharp contrast with the price measures used in constructing the BUS and BEA output indexes. The next step in the computation is to correct for inventory change, which includes adjustments for intra-year price changes. No account is taken of any possible change in the use of the LIFO or FIFO methods of in- ventory valuation, so that inventory change may be poorly estimated. At this point, the deflated-dollar values in the base and given years for an industry are converted to index form; this is the index of production with value of shipments weights shown in Census publications. For higher levels of industry aggregation, the sums of the deflated values in the base and given years are used to derive indexes with value of shipments weights. BUREAU OF ECONOMIC ANALYSIS The Bureau of Economic Analysis prepares estimates of constant-dollar output (measured as value of shipments plus miscellaneous receipts) as an intermediate step in the computation of eonstant-dollar gross product originating by major industry division (Chapter 41. In this section we are concerned with their current- and eonstant-dollar measures of out- put. Their measures of intermediate inputs will be discussed in the section on materials. BEA Method for Measuring Output Shipments in current dollars are available on a 5-digit product class basis from the Census of Manufactures (CM) and the Annual Survey of

OCR for page 391
396 PAPERS Manufacturers (ASM). In the ASM, these are collected only on a wherever- made basis (that is, regardless of whether or not the plant producing the product specializes in that product), but they do exclude pulp that is both produced and consumed in a paper or paperboard mill. BUS producer price data are used to deflate the product class data. BES special product class indexes are usually used, but sometimes single 7-digit commodity producer price indexes are used as alternatives, using relative quantity weights derived from the latest available census. Current- and- constant-dollar shipments are aggregated to the 4-digit level. The constant-dollar total is then divided into the current-dollar total, to produce a shipments deflator with current year, wherever-made quantity weights. This shipments deflator is then applied to total industry shipments and miscellaneous receipts to produce constant-dollar output. Note that these 4-digit output estimates are on an establishment basis; that is, they reflect the output of plants specializing in the industry (thus, a paper mill that sells small amounts of pulp will have all of its output classified in the paper mill industry). Biases Introduced by "Wherever-Made'' Weights As BEA recognizes, the use of price indexes with wherever-made weights (at the S-digit product class level) to deflate establishment-based industry output data can produce biased results when the coverage or the specialization ratio is low (see Office of Business Economics 1966, p. 101. A special procedure is described in the Office of Business Economics (OBE) document for situations of this kind. (The Office of Business Economics was the predecessor to the Bureau of Economic Analysis.) This procedure is not followed in the case of pulp mills, which has a low coverage ratio, because the industry is small and its specialization ratio is high. However, the two major product classes of the pulp industry have very different coverage ratios. Almost all of"special alpha and dissolving pulp" (sly 26111) is produced by pulp mills, whereas coverage of "other pulp" sac 26112) was only 44 percent in 1972. As a result, sac 26111 has a weight of 26 percent in sac 2611 on a wherever-made basis but a weight of 45 percent on an establishment basis. Although in most years the two price indexes have shown only slight differences, years of rapid price movements can produce substantial bias. Table 2 shows the price indexes used by BEA for the two groups from 1972 to 1975 and compares the results obtained by using wherever-made weights with estimates obtained using current-year establishment weights. It is possible to estimate current-year weights on a 5-digit basis because

OCR for page 391
Data Adequacy for Productivity Analysis TABLE 2 Comparison of Pulp Price Indexes (1972 = 100.0) Special Alpha and Dissolving Other BEA Pulp 26111 Pulp 26112 Index Industry Weights 1972 100.0 100.0 100.0 100.0 1973 115.1 115.3 115.3 115.2 1974 160.9 199.0 189.0 183.8 1975 207.4 256.2 244.1 239.6 SOURCE: All data, except "Industry Weights," are from BEA un- published worksheets; "Industry Weights" is based on CM and ASM data on the assumption that all of sac 26111 is produced in sac 2611 and that sac 26112 is equal to the residual production of the pulp mill industry (see text). 397 almost 100 percent of product class sac 26111 has historically been pro- duced in sac 2611 establishments; production of sac 26112 can therefore be approximated as a residual. As can be seen, the BEA index is biased up- ward for 1974 and 1975. It can be shown that this bias will not be removed by aggregation, but the aggregate bias will be smaller than the bias for an individual industry. Comparison of BEA and CM BEA iS responsible for the publication of official estimates of GNP on a timely basis, and also updates and revises these estimates to provide a basis for historical analysis. As time passes, an approximation that is useful as an up-to-date estimate of year-to-year changes may systemati- cally diverge from a more carefully prepared and analytically sophis- ticated measure. For this reason, BES and FOB output indexes are adjusted to Census output data benchmarks as they become available. This is not true of the BEA data, partly because revisions of the gross product originating estimates are not coordinated with revisions of the rest of the national accounts. The question thus arises whether the Census estimates and the BEA estimates in fact diverge and, if so, whether one is superior to the other. To answer the latter question first, the Census output indexes appear to be superior to the BEA estimates on two grounds. First, they avoid the bias of using wherever-made product class weights and deflate on an establish

OCR for page 391
398 PAPERS TABLE 3 Comparison of BEA and Census Shipments Deflators, 1963- 1967 and 1967-1972, Primary Paper Industries 1963-1967 1967-1972 BEA Census BEA Census (1963 = 100.0, (1967= 100.0. Industry cross weighted) base year weights) sac 2611 106.8 104.7 111.6 111.9 sac 2621 109.2 105.4 116.6 107.3 sac 2631 102.7 101.4 104.8 109.0 sac 2661 96.4 97.7 117.8 113.5 SOURCE: BEA data from BEA worksheets. Census data from CM, Vol. IV, 1967 and 1972. ment basis. Second, the interagency task force may choose unit-value price deflators or special price deflators where these appear to be more ap- propriate than BES price indexes. For the primary paper industries the task force has typically leaned heavily on unit values. Two comparisons are made of the BEA and the CM estimates. The first is at the sac 4-digit level, for each of four industries (sac 2611, 2621, 2631, and 2661~. BEA data for shipments (and miscellaneous receipts) do not conform with Census output indexes because at this level of aggregation the Census indexes are corrected for changes in inventory, while BEA data are not. Therefore we compare the Census and BEA deflators for shipments. As can be seen in Table 3, the Census figures generally show a lower rate of price increase and hence a higher real value of output per current dollar shipped. This is particularly true for sac 2621, the industry with the highest value of shipments (more than half the total of the four industries combined in 1967~. Table 4 shows a similar bias at the 2-digit level. The Census output in- dex for sac 26 is roughly 3 percentage points higher than the BEA value of production (expressed as an index) in each period. These differences seem substantial enough to call into question the acceptability for historical analysis of BEA worksheet data on shipments and materials at both the 4-digit and 2-digit levels. BUS OUTPUT INDEXES This measure, covering the combined four industries of our study, is published for 1939 and from 1947 to date, together with indexes of employment and employee-hours (for all employees, production workers,

OCR for page 391
Data Adequacy for Productivity Analysis TABLE 4 BEA and Census Output Indexes, 1963-1967 and 1967-1972, sac 26 1963- 1967 (1963 = 100.0, cross weighted) 1967-1972 (1967= 100.0, end year weights) BEA 120.7 119.2 Census 123.7 122.1 SOURCES: BEA data from BEA worksheets. Census data from CM, Vol. IV, 1967 and 1972. 399 and nonproduction workers) and corresponding productivity measures (see, for example, Bureau of Labor Statistics 1977a). The output measures are derived from data collected by the Census and published in C1R. The weighting system is fairly complex, involving unit values, deflated value series, and employee-hours per physical unit of pro- duction. The following summary was drawn from the Technical Note for this group of industries (Bureau of Labor Statistics 1977b) and from in- spection of worksheets provided by sES. The method of computation was changed in 1967; the description is therefore divided into two parts, cor- responding to the periods to which each procedure is applicable. 1939-1967 Production indexes for pulp and for paper, paperboard, and building paper and board were computed using physical quantities weighted by unit values for base (CM) years. The quantity data, drawn from C1R, are collected on a product (wherever made) basis. The two indexes were com- bined using base-year production-worker hour weights; these weights were derived from the most recent CM, and represented (1) the hours spent on pulp production in all four industries (tic 2611, 2621, 2631, and 2661), relative to total production-worker hours in the four industries and (2) all other production-worker hours relative to total hours. 1 967-Forward The value of pulp shipments, wherever made, is taken from ASM for two primary product classes and is deflated by a price index. The latter is base-year weighted, using sES producer price indexes and CM value of shipments weights, for 7-digit primary products. The two primary product

OCR for page 391
400 PAPERS (5-digit) groups are combined with base-year employee-hour weights, derived from the most recent CM. The resulting pulp production index is further adjusted for coverage to the pulp mills (sir 2611) industry and for inventory change. The coverage adjustment is carried out annually by computing for sac 2611 the ratio of industry shipments to shipments wherever made, both taken from ASM. The inventory change adjustment is also based on annual ASM data. For paper, paperboard, and building paper and board, production in physical units is taken from CIR at the 7-digit level. These quantities are weighted with base-year unit values (CM years), aggregated to the 5-digit level, and converted to index numbers. The 5-digit level index numbers are then combined using employee-hour weights. Finally, the pulp index, on an industry basis, is combined with the in- dex of other products using base-year (CM) employee-hour weights. Employee-Hour Weights The BES output index is used together with labor input data to measure changes in labor productivity. Unit labor requirements are used as weights whenever available in order to eliminate changes in the produc- tivity measure due to changes in the mix of products whose unit values dif- fer but whose unit labor requirements are the same. Such a shift of output among products will change a labor productivity measure that uses unit- value weights, but not one that uses unit labor requirement weights (see Chapter 41. Prom 1967 to 1972, a development of this type occurred in the paper- board mills industry. Output rose 31.0 percent when measured with unit- value weights between these years, but rose only 25.7 percent when measured with employee-hour weights; as a result, output per production- worker hour rose 27.8 percent on the unit-value basis, but only 22.5 per- cent on the labor-hour basis. Comparison with CM Indexes When preliminary versions of the CM indexes have been prepared, they are compared with the BES and FRB indexes, and an attempt is made to recon- cile any major differences. As a result, the BUS and CM indexes tend to agree quite closely. For the 1967-1972 change, the BES index rose 22.3 percent, while a combination of the CM indexes (to obtain the same in- dustrial grouping) rose 21.2 percent with employee-hour weights and 23.1 percent with value-added weights. Similar results were found for earlier CM comparisons (Table 5~. It is interesting to note that the BUS indexes

OCR for page 391
Data Adequacyfor Productivity Analysis TABLE 5 Comparison of Census of Manufactures In- dexes of Production with sibs Indexes, Employee-Hour Weights and Value-Added Weights for Selected Years, 1958-1972, Pulp, Paper, and Board Mills Combined Census of Manufactures Employee- Value Hour Added Year Weights Weights BES 1972 (1967 = 100) 121.2 122.9 122.3 1967(1963 = 100) 121.3 121.4 121.1 1963 (1958 = 100) 128.2 130.5 130.3 SOURCE: 1972 CM indexes derived from 1972 Census of Manufactures, Volume IV; 1967 CM indexes from 1967 Census of Manufactures, In- dexes of Production (Special Report); 1958 CM indexes from 1963 Cen- sus of Manufactures, Volume IV; BES indexes from Productivity Trends for Selected Industries, 1977 Edition, p. 66. 401 agree more closely with the CM value-added indexes than with the CM employee-hour indexes in two of the three comparisons. Price Deflators From 1967 onward, the value of pulp shipments is deflated using BUS price indexes. For paper, paperboard, and building paper and board products, unit-value weighted quantity indexes are prepared; this is equivalent to deflating the value of production with unit-value price indexes (con- structed with current-year quantity weights). The BUS procedure differs considerably from the method used by the Census Bureau, where 90 percent of pulp shipments are deflated with unit values, and 38 percent of paper, 10 percent of paperboard, and 100 per- cent of building paper and board shipments are deflated using BUS pro- ducer price indexes. FEDERAL RESERVE BOARD INDEXES OF INDUSTRIAL PRODUCTION Federal Reserve Board (FRB) Indexes of Industrial Production are calculated for pulp mills (SIC 2611), paper mills (2621), paperboard mills (2631), and building paper and board mills (2661~. At this level of detail,

OCR for page 391
412 TABLE 9 Prices for Pulpwood Consumed PAPERS Prices ($ per cord) 1967 1972 Price Index (1967= 100.0) Value (million $) 1967 1972 Southern pine 261123.34 27.73118.861.546.0 262121.60 26.54122.9130.4186.9 263122.84 27.93122.3260.1431.4 266119.81 24.20122.25.96.9 Softwood chips 261116.48 28.29171.748.473.2 262120.64 22.89110.996.4144.4 263118.51 22.71122.7115.0240.6 266117.72 16.7494.43.57.0 Southern mixed hardwood 261119.90 25.57128.522.732.4 262118.59 23.00123.732.959.9 263118.11 22.29123.159.081.0 2661S 19.69NAS1.4 Total20.61 25.29123.4a835.81,311,1 122.7b Total, all wood reported20.89 25.15120.4b1,120.61,737.0 S = does not meet publication standard. UMarshall-Edgeworth, cross weighted index. bBased on average price per cord, unweighted. SOURCE: 1972 Census of Manufactures. NA = not available. puts, whereas integrated woodpulp is an intermediate product within the production process and appears neither in input nor in output. In the FRB framework, on the other hand, all pulp mills, whether integrated or not, are considered a separate industry and integrated pulp is conceptually both an input and an output. CIR publishes estimates of consumption of woodpulp at paper and board mills, which should be very close to the total for the primary paper industries. This pulp is divided into own pulp (transfers and integrated) and purchased pulp. As shown in Table 10, there are major disagreements between CIR and CM. There are two reasons why CM may show smaller figures than CIR. One is firms for which no data are collected in detail; these represented

OCR for page 391
Data Adequacyfor Productivity Analysis TABLE 10 Comparison of Current Industrial Reports and Census of Manufactures Data on Woodpulp Consumed, Selected Years, 1963-1972 (thousands of short tons) 1963 1967 1972 Total Wood Pulp CM CIR Ratio, CM/ CIR Market Wood Pulp CM CIR Ratio, CM/CIR Own pulpa CM CIR Ratio, CM/ CIR 28,223 32,545 43,750 30,220 36,994 47,347 0.93 0.88 0.92 3,081 3,253 4,581 3,380 3,910 4,839 0.91 0.83 0.95 25,142 29,292 39,169 26,840 33,084 42,500 0.94 0.89 0.92 aIncludes interplant transfers. SOURCE: Census of Manufactures, 1967 and 1972; Current Industrial Reports, M26A, selected years. 413 roughly 1 percent of consumption in 1972 and 3 percent in 1967 and 1963. The other is data suppressions and omissions. As can be seen in Table 11, data availability has become progressively worse for pulp and building paper and board mills. A change in Census classification of pulp mills, to exclude nearly all integrated pulp mills, means that pulp consumption in pulp mills (sac 2611) in 1972 was undoubtedly very minor in that year (probably less than 0.1 percent of all pulp consumption in the primary paper industries). Building paper and board probably accounted for about 5 percent of pulp consumption in that year. These two factors ex- plain most of the discrepancies in 1963 and 1972, but not in 1967. From 1963 to 1967 the small growth of total pulp consumed in the paperboard industry is quite striking. Industry output rose over 24 per- cent, yet pulp consumed rose only 4 percent. On the other hand, from 1967 to 1972 pulp consumed rose by roughly 80 percent, while output rose

OCR for page 391
414 PAPERS TABLE 11 Woodpulp Consumption in Primary Paper Industries, Selected Years, 1963-1967 (thousand short tons) 1963 1967 1972 Pulp (sac 2611) Total pulp consumed Transfers Integrated Market Paper (sac 2621) Total pulp consumed 14,926.8 Transfers 1,973.1 Integrated 10,323.8 Market 2,629.9 Paperboard (sac 2631) Total pulp consumed Transfers Integrated Market Building paper and board (sac 2661) Total Transfers Integrated Market 160.9 NA 11.0 149.9 NA 12,028.2 300.8 11,307.6 419.8 1,166.9 NA 14.7 1,130.5 31.7 S NA NA D 334.9 S D 18,811.4 2,523.5 13,308.2 2,979.7 12,479.5 505.0 11,701.1 273.4 S S 920.1 21,312.5 2,742.2 14,244.0 4,326.3 22,439.5 339.2 21,845.4 254.9 NA D NA D NA = not available. D = withheld to avoid disclosure. S = data failed to meet Census publication standards. SOURCE: Census of Manufactures, 1967 and 1972. only 31 percent. It therefore appears that the 1967 pulp consumption by the paperboard industry is understated by more than 20 percent. This error is primarily in the integrated pulp data and therefore need not affect the accuracy of the input measures on an establishment basis. There is a substantial discrepancy in the data for market woodpulp, but the size of the reporting error, if any, cannot be determined with precision because of the missing data for pulp mills and building paper and board mills. BLS Woodpulp Producer Price Index The BUS producer price index for woodpulp is based on six kinds of wood- pulp. This index would appear to be a superior basis for deflation than CM

OCR for page 391
Data Adequacy for Productivity Analysis TABLE 12 Comparison of Woodpulp Price Indexes (earlier year = 100) Price Index 1958-1963 1963-1967 1967-1972 PPI pulp Census output Unit-value relative Census materials consumed Unit-value relative 87.3 104.7 91.7 106.8 88.0 103.6 .5 110.9 111.6 SOURCE: BES Wholesale Price Indexes, Annual, 1964, 1968, 1973; Cen- sus of Manufactures, 1963, 1967, 1972. 415 unit-value relatives for pulp consumed, for which pulp is divided only into transfers and purchases (values of integrated pulp are not collected). The interagency committee that decides which deflators to use in con- structing Census output indexes has consistently chosen CM unit-value relatives over the PA in measuring pulp output. This is because the Gil uses list prices rather than transaction prices. The unit-value relatives for pulp output are relatively free of changes in product mix. It is interesting to note that unit-value relatives for pulp consumption are closer to the unit-value relatives for pulp output than is the apt (Table 12~. This suggests that the unit-value relatives in this case are to be pre- ferred to the PA for deflating CM inputs. However, the PA appears to be sufficiently reliable for annual series. Conclusions on Primary Data Materials input data from the Census of Manufactures seem to have some errors and omissions. Because of the existence of CIR statistics on wood and pulp consumption, it would appear that these errors and omissions could be corrected through more active cross checking and data verif~ca- tion. It would also be useful for the reporting form on which the data are collected, as well as the published table, to include subtotals of the major categories of inputs consumed, such as pulpwood, woodpulp, all other fibers, chemicals, and fillers. For purposes of deflating CM inputs, the unit-value relatives obtained from CM appear superior, at least for wood and pulp, to the Pat. However, there does not appear to be a good substitute for the PA for annual infor- mation. Data from the CiR on wood and pulp cannot be used for in- dividual 4-digit industries but may be useful in checking the

OCR for page 391
416 PAPERS reasonableness of the PPI by aggregating the four primary paper in- dustries, particularly for the pulpwood index, which is more likely to go astray. BEA UNPUBLISHED WORKSHEET ESTIMATES OF MATERIALS INPUT The Bureau of Economic Analysis prepares estimates of current- and constant-dollar materials inputs for each 2-digit input-output industry group (ISP). These estimates appear as intermediate results in the com- putation of constant dollar gross product originating by industry. BEA Data Sources Current-dollar totals of materials consumed are available annually for each SIC 4-digit industry from the Annual Survey of Manufactures. These are aggregated by BEA to ISP 2-digit groups. "' The ISP 2-digit groups are more disaggregated than the sac 2-digit groups. sac 26, paper and allied products, is composed of asp 24 and 25; ISP 25 is the same as sac 265, paperboard containers and boxes, and asp 24 represents the rest of the in- dustry groups, SIC 261-4 and 266, the primary paper industries and con- verted paper and paperboard products, excluding containers and boxes. Real materials consumed is obtained through use of BUS producer price indexes, which are aggregated for each asp group using weights from the most recently available input-output (I-O) table. The I-O table becomes available after a lag of 6 or 7 years. As a consequence, 1967 weights are still being used in 1978. The 1972 I-O workfiles will become available over the course of the next year, and 1972 weights will eventually be incor- porated. The input deflators are, in principle, current year weighted, as are the output deflators constructed by BEA. For the output deflators, current- year weighting is made possible by the existence of 5-digit shipments data from the Annual Survey of Manufactures. Similar detail is not available for materials inputs. For years prior to 1967, current-year weights are ap- proximated by straight-line interpolations between I-O tables available for census years. For 1967 and later years (until the data from the 1972 I-O table can be incorporated), 1967 weights are used. Detailed purchases of materials consumed are obtained from BEA worktables created during preparation of the 1967 I-O tables. These con ~The lisp classification is derived from the 1958 input-output system, which contained 85 sectors, including 51 in manufacturing.

OCR for page 391
Data Adequacyfor Productivity Analysis 417 stitute input categories that are sometimes as small as 7-digit product codes but often as large as 4-digit industries. The amount of detail varies considerably from industry to industry. asp 25 has 29 specified inputs in BEA worksheets, and is dominated by one, paperboard, which accounted for nearly 70 percent of specified inputs in 1967. asp 24 has 65 specified in- puts of which the largest is paper; this accounted for nearly one fourth of specified inputs in 1967. Each specified input is matched to a BES producer price index. Like input-output data, the BEA inputs are also in producer prices. The pro- ducer price valuations are constructed from CM materials consumed (which are in purchaser prices) by subtracting transportation costs, wholesale trade margins, and taxes, calculated by the Census inter- industry division. BEA includes transportation costs as separate com- ponents of materials input, but omits wholesale trade costs. ESTIMATES OF REAL VALUE ADDED We have argued that unit value relatives are superior to producer prices for deflating materials input for the primary paper industries. Here we present our estimates of real materials input and real value added, deflated with unit values, and compare them with BEA estimates. Deflating Materials Input Energy and nonenergy materials were separately deflated. For nonenergy materials, we took all items from CM'S Table 7A for which price and o~uan- tity data are shown for either 1958 and 1967 or 1967 and 1972 at the 4-digit industry level. For each of those industries, we constructed a price- weighted quantity index (Marshall-Edgeworth). This provided a quantity index for detailed inputs that was extended to all nonenergy materials by assuming that the implicit deflator for the detailed inputs is the same for all primary paper industries. The procedure was similar for energy materials, except that detailed energy inputs were collected in 1971 and 1974, but not in 1972. The 1972 energy index was constructed by interpolation between 1971 and 1974 us- ing the PA for fuels and power. "All materials" is the sum of energy and nonenergy materials. As can be seen from Table 13, from 1958 to 1967, energy input in- creased more rapidly than nonenergy materials input, but between 1967 and 1972 they grew at approximately the same rate. The deceleration of energy consumption coincided with a sharp increase in energy prices after 1970.

OCR for page 391
418 PAPERS TABLE 13 Real Output, Materials Consumed, and Value Added in Primary Paper Industries, Deflated by Unit Values, 1958, 1967, 1972 (indexes, 1967 = 100, cross weights) 1958 1967 1972 Energy 63.3 100.0 119.4 Nonenergy materials 67.2 100.0 118.0 All materials 66.8 100.0 118.2 Output 63.6 100.0 121.7 Value added 60.5 100.0 125.5 Annual Rates of Changes (percent) 1958- 1967 1967-1972 Materials 4.6 3.4 Output 5.2 4.0 Value added 5.7 4.6 SOURCE: Census of Manufactures, 1958, 1963, 1967, 1972. The output measure in Table 13 is the CM index (1958-1963 and 1963-1967 indexes were linked to derive the 1958-1967 number). Since the input indexes were constructed by a similar procedure, they are com- parable with the output index. The materials index for the primary paper industries grows less rapidly for both 1958-1967 and 1967-1972 than the CM output index. As a conse- quence, value added grows more rapidly in both periods than output, the difference averaging 0.6 percent annually. Comparison with BED Value Added BEA does not estimate materials input and value added at the level of detail of the primary paper industries. The BEA'S smallest categories are asp 24 and 25. asp 24 comprises the primary paper industries plus sac 264, while asp 25 is sac 265. To make comparisons, we have constructed measures of energy, nonenergy materials, output and value added for sac 264 and 265 that parallel those used in the previous section, and have aggregated them with the results for the primary paper industries to produce totals for asp 24

OCR for page 391
Data Adequacy for Productivity Analysis TABLE 14 Comparison of BEA and Authors' Estimates of Real Materials, Output, and Value Added in Paper and Allied Products Industry Group (sac 26), 1958, 1967, 1972 (indexes, 1967 = 100.0) Industry Group, Measure, and Source of Estimate 1958 1967 1972 lisp 24 (sac 26, except 265) Materials BEA 61.5 100.0 108.5 Authors 64.2 100.0 122.1 Output BEA 63.6 100.0 119.8 CM 61.9 100.0 121.8 Value Added BEA 66.6 100.0 135.7 Authors 59.4 100.0 121.5 asp 25 (sac 265) Materials BEA 58.0 100.0 121.7 Authors 61.0 100.0 126.6 Output BEA 62.0 100.0 117.9 CM 61.6 100.0 124.1 Value Added BEA 66.8 100.0 113.6 Authors 62.4 100.0 121.1 sac 26 Materials BEA 60.5 100.0 112.1 Authors 63.2 100.0 123.4 Output BEA 63.2 100.0 119.2 CM 61.8 100.0 122.5 Value Added BEA 66.8 100.0 128.9 Authors 60.2 100.0 121.4 SOURCE: BEA workfiles; Census of Manufactures 1958, 1963, 1967, and 1972. 419 and 25 and SIC 26. These are shown in Table 14 together with sEA measures from worksheets. The results are perverse. Over the periods shown, the BEA estimates of output are invariably better estimates of real value added (accepting our estimates as the standard) than are the BEA estimates of value added. At these levels of aggregation, our figures show relatively small differences

OCR for page 391
420 PAPERS between output and value added, while BEA figures show major dif- ferences. It is not possible to arrive at a definitive critique of the BEA calculations in this report because sac 264 and 265 lie outside its scope. However, we have presented a number of arguments that suggest that the BEA figures are not reliable. The figures presented in Table 14 underscore the impor- tance of data reliability. Because of the size of these differences, it would be useful for the ASM to carry additional detail on materials input. In the primary paper in- dustries, materials input could be usefully subdivided into a small number of categories: pulpwood, woodpulp, other fibers, chemicals, fillers, and other. Whether this is true of other industries would be a worthwhile sub- ject for investigation. SUMMARY AND RECOMMENDATIONS 1. There are four measures of output for the primary paper industries, prepared by four agencies of the federal government (FRB, BES, BEA, and Census). These measures are prepared using differing concepts, deflators, and weighting schemes. The principal recommendation of this study is that the four measures be considered successive approximations to the same industry output concept and that the deflators and weighting schemes be standardized in order to achieve this end. Special purpose in- dexes could easily be produced by individual agencies in addition to the standard ones. 2. There are 492 industries (4-digit sect in manufacturing and mining, each with peculiarities that affect output measurement. No one of the four agencies has staff with expertise on even a majority of these industries. A greater degree of cooperation among the agencies would result in the pro- duction of better measures by all four and would aid in resolving the prob- lems we found in each existing index. 3. It is important that careful comparisons be made among the measures produced by the estimating agencies, as well as with measures produced by trade associations and other government agencies. Such comparisons can help to reveal new developments in individual industries that require special estimating procedures and can help to avoid a feeling of insularity with respect to an agency's own measures. 4. The three published measures of industry output (FRB, BUS, and Census) are all gross of intermediate purchases. Variations in the ratio of purchased materials to output can produce misleading movements in

OCR for page 391
Data Adequacy for Productivity Analysis 421 labor productivity measures, as has been shown in this study. These arise from such sources as physical integration of production processes, imports of semimanufactured inputs, and recycling of used products. Measures of real materials inputs would provide the basis for solving this problem, since they would permit productivity analysts either to ad- just the output measure (yielding real value added) or to include materials explicitly as an additional input. Satisfactory official measures of real materials inputs do not exist now for individual industries (4-digit sect or industry groups (2-digit sect, yet we have demonstrated that it is possible to prepare such measures from data published in the quinquennial Census of Manufactures. 5. Small changes in survey questionnaires and publication format would make the Census of Manufactures data much more useful for the measurement of materials input. More extensive verification of the data would also yield large benefits. 6. The Census Bureau should consider the calculation of materials in- put indexes that parallel the CM output indexes. These indexes should, at a minimum, be estimated for the period since 1958. 7. Current data are inadequate for the accurate measurement of real materials input on an annual basis. The most likely vehicle for improved data collection in this area is the Annual Survey of Manufactures. For the primary paper industries, a few subtotals of major input categories would greatly improve our ability to measure changes in materials input. Whether this is true in other industries is a worthwhile question for the Bureau of the Census and other agencies to explore. NOTATION API ASM BEA CIR CM FRB I-O ISP PPI SIC American Paper Institute. Annual Survey of Manufactures. Bureau of Economic Analysis. BLS Bureau of Labor Statistics. Census Bureau of the Census. Current Industrial Reports. Census of Manufactures. Federal Reserve Board. Input-Output Table. Inter-Industry Sales and Purchases. Producer Price Index. Standard Industrial Classification.

OCR for page 391
422 REFERENCES PAPERS Bureau of the Census (1972) Census of Manufactures, Vol. IV, Indexes of Production. Pp. 10-19. Washington, D.C.. U.S. Department of Commerce. Bureau of Labor Statistics (1977a) Productivity Indexes for Selected Industries, 1977Edi- tion. Bulletin 1983. Washington, D.C.: U.S. Department of Labor. Bureau of Labor Statistics (1977b) Paper, paperboard and pulp mills-SIC 2611, 2621, 2631, and 2661. Technical Note. Washington, D.C.: U.S. Department of Labor. Early, John F. (1978) Improving the measurement of producer price change. Monthly Labor Review 101(4):7-15. Office of Business Economics (1966) GNP by Major Industries, Concepts and Methods. Washington, D.C.: U.S. Department of Commerce.