A perspective on the U.S.-Canada manufacturing productivity gap

Monthly Labor Review

Volumn 124, Issue 2, 2001, Pages 31-48

  Eldridge, Lucy P ^a   Sherwood, Mark K ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0040152588     PISSN: 00981818     EISSN: None     Source Type: Journal    
DOI: None     Document Type: Review

References (66)

1. "Investigating the Canada-US Productivity Gap: BLS Methods and Data," presented at the Centre for the Study of Living Standards Conference on the Canada-US Manufacturing Productivity Gap, Ottawa, January 21-22, 2000. Conference papers can be obtained on the Internet at http://www.csls.ca, under the heading "recent events." More current data are included in the present article, as well as substantive revisions in the two countries' series that have taken place since January 2000.
2. The list is not exhaustive but includes issues thought to be most important. It should be noted that there are limitations associated with the quality of the particular data available for estimating the variables used in productivity measurement, as well as with the practical procedures that must be used when data are not available. It is not feasible to review all of the limitations. For example, accurate price deflators necessary to determine the real input of purchased services are rare. Because this problem is not unique to either country, we did not attempt to explore the issue.
3. A standardized system being adopted, at least in part, by many countries is the System of National Accounts 1993, which has been undertaken under the joint responsibility of the United Nations, the International Monetary Fund, the Commission of the European Communities, the Organization for Economic Co-operation and Development, and the World Bank. See System of National Accounts 1993, Series F, Number 2 (United Nations Statistics Division, 1993); for more information, visit the United Nations Statistics Division website, on the Internet at http://www.un.org/Depts/unsd/.
4. Standard Industrial Classification Manual: 1987 (Executive Office of the President, Office of Management and Budget, 1987).
5. More information on Canada's 1980 Standard Industrial Classification system can be found on the Internet at http://www.statcan.ca/english/Subjects/Standard/standard_classifications.htm.
6. In 1997, Canada completed a major revision to the national accounts. This historical revision did not result in any changes to the classification of the manufacturing sector or to two-digit industries within manufacturing; however, the numbering structure for industries within manufacturing was modified. Both countries, along with Mexico, are in the process of converting to a common industrial classification known as the North American Industrial Classification System (NAICS). At this time, manufacturing data on a NAICS basis are scheduled to be released in Spring 2001 for Canada and not before 2004 for the United States.
7. There are several other components of Canadian manufacturing that are not in U.S. manufacturing that we were unable to adjust for. These include portions of photo finishing (C-SIC80 2821), vehicle engine repair services (C-SIC80 3081,3261 ), and clay products (C-SIC80 3511, 3512, 3591).
8. Output and hours for the Canadian logging industry (C-SIC80 w-5) were obtained from Statistics Canada. Output and hours for the five-digit US logging industry were obtained from the BLS Division of Industry Productivity Studies. The consistency in aggregation theorem was used to construct the adjusted series. (See W. E. Diewert, "Superlative Index Numbers and Consistency in Aggreagation," Econometrica, July 1978, pp. 883-900.) Under the 1997 North American Standard Industrial Classification System (NAICS), the logging industry will not be included in the manufacturing sector.
9. For more information, visit BEA's website, on the Internet at http:// www.beadoc.gov/.
10. Sherlene K. S. Lum, Brian C. Moyer, and Robert E. Yuskavage, "Improved Estimates of Gross Product by Industry for 1947-98," Survey of Current Business, June 2000, pp. 24-54. Note that prior to the June 2000 revisions, the U.S. data relied upon information from the 1987 input-output table. This raised concern at the Conference regarding comparability with Canadian data. The authors addressed that point in a revised paper prepared for the Conference volume.
11. Statistics Canada is the Canadian national statistical agency; the agency publishes statistics organized into three broad subject matter areas: demographic and social, socioeconomic, and economic. For more information, visit the official Statistics Canada website, on the Internet at http://www.statcan.ca/ start.html.
12. See "Gross Domestic Product by Industry," Catalogue no. 15-001-XPB, (Statistics Canada, October 1997) pp. vii-xxvi.
13. In both countries deflation is conducted at the most detailed level possible. In the United States, current dollar gross output is deflated primarily using BLS producer price indexes. Notable exceptions are price indexes constructed by BEA for computers, telephone-switching equipment, selected semiconductor products, and government purchases. Intermediate purchases are deflated with prices based largely on those used to prepare constant dollar gross output and imported intermediate inputs are deflated mainly using BLS import price indexes. In Canada, current dollar gross output and intermediate purchases are deflated with producer price indexes. In general, the two countries use similar deflation techniques to measure real output and inputs.
14. The System of National Accounts 1993 recommends the use of basic prices or modified basic prices to value gross output. Prior to 2000, Statistics Canada valued output at factor costs.
15. Kishori Lal, "Value Added by Industry: A Problem of International Comparison," OECD Meeting of National Accounts Experts, Paris, (Statistics Canada, September 21-24, 1999).
16. Prior to 1987, subsidies in U.S. manufacturing never accounted for more than .05 percent of nominal gross output in manufacturing.
17. BEA implements the aggregation procedure at a product level while Statistics Canada implements it at a higher industry level. Because the adoption of the new aggregation formula by the Canadians led to small overall differences at the manufacturing level, it is presumed that implementing the method at a more detailed level would also have a small impact on the aggregate manufacturing sector.
18. Robert E. Yuskavage, "Improved Estimates of Gross Product by Industry, 1959-94," Survey of Current Business, August 1996, pp. 133-155, especially p. 138.
19. Ibid., pp. 138-39.
20. For any given interval between base periods, Canada constructed a fixed-weight Laspeyres output index. Since 1981, the base year of this index was updated every 5 years (1981, 1986, and 1992). See Kishori Lal, "Remaining Differences Between the 1997 Canadian System of National Accounts and the 1993 International System of National Accounts," OECD Meeting of National Accounts Experts, Paris, France, September 22-25, 1998 (Statistics Canada), p 23.
21. Software enters the national accounts as purchases of software and as own-account software production. Own-account software refers to software produced by a business or government for its own use. Purchased software is produced by the computer programming and the prepackaged software industries (US-SIC87 7371 and 7372) in the service sector.
22. For a detailed discussion of the revision, see Brent R. Moulton, Robert P. Parker and Eugene P. Seskin, "A Preview of the 1999 Comprehensive Revision of the National Income and Product Accounts: Definitional and Classificational Changes," Survey of Current Business, August 1999, pp. 7-20.
23. System of National Accounts 1993 (United Nations Statistics Division, 1993); see also note 3.
24. Nominal GPO is measured as the sum of the distribution by industry of the components of gross domestic income (GDI). The new treatment of software as fixed investment increased proprietor's income and profit within GDI by eliminating the deductions for software purchases and by adding the value of the production of own account software. These effects were only partially offset by the deduction of the consumption of fixed capital on both purchased software and own-account software production. Thus, this change increased nominal GDI. To calculate real GPO, nominal intermediate purchases are constructed as gross output less GDI. The new treatment of software increased both gross output and GDI; however, the impact on GDI was greater, and thus the result was decreased nominal intermediate purchases. Consequently, real GPO was revised upward slightly.
25. See Lum, Moyer, and Yusgavage, "Improved Estimates of Gross Domestic Product, 1947-98," June 2000.
26. At the industry level, software accounts for slightly larger shares in some industries: 2.7 percent in printing, publishing and allied industries (USSIC87 27) and in industrial and commercial machinery and computer equipment (US-SIC87 35); 2.2 percent in measuring, analyzing, and controlling instruments, photographic, medical and optical goods, watches and clocks (US-SIC87 38); and 2 percent in electronic and other electrical equipment and components, except computer equipment (US-SIC87 36).
27. Consistency in aggregation theorem; see Diewert, "Superlative Index Numbers," July 1978.
28. Statistics Canada.
29. This specification is used in Information Technology in the Service Society, a report prepared by the National Research Council (Washington, DC, National Academy Press, 1994); see p. 23, note 1.
30. See Robert E. Yuskavage, "Gross Product by Industry Price Measures, 1977-96," Survey of Current Business, March 1998, pp. 17-25, especially p. 22. In the United States, within the US-SIC87 two-digit industry electronic and other electrical equipment and components, except computer equipment, the prices of more than half the value of the products' shipments fell somewhat or greatly between 1992 and 1996. However, most did not fall to the extent the information technology items we reviewed fell. Further, by 1996, the information technology products that we reviewed in this industry made up about a third of the output of the electronics industry.
31. See Bruce T. Grimm, "Price Indexes for Selected Semiconductors, 1974-96," Survey of Current Business, February 1998, pp. 8-24; see p. 9 for a brief discussion of the issues associated with the measurement of output and prices for high-tech goods.
32. Andrew A. Wyckoff, "The Impact of Computer Prices on International Comparisons of Labour Productivity," in Economics of Innovation and New Technology, vol. 3 (Amsterdam, Overseas Publishers Association, 1995), pp. 277-93, especially page 282.
33. See, for example, Ernst R. Berndt and Zvi Griliches, "Price Indexes for Microcomputers," in Murray F. Foss, Marilyn E. Manser, and Allan H. Young, eds., Price Measurements and Their Uses, National Bureau of Economic Research Studies in Income and Wealth, no. 57 (Chicago, University of Chicago Press, 1993), pp. 63-93, especially table 2.1.
34. The price data are not strictly comparable because the U.S. implicit price series is a chained annual-weighted price index; the Canadian implicit price index is not.
35. Computers, storage devices, terminals, and peripherals (US-SIC87 3571, 3572, 3575, and 3577) in the United States and electronic computing and peripheral equipment (C-SIC80 3361) in Canada. In 1992 and 1996, respectively, these industries accounted for 2.1 percent and 2.6 percent of manufacturing nominal shipments in the United States. The comparable figures for Canada were 1.2 percent and 1.6 percent.
36. The U.S. price index is calculated by weighting the respective industry price changes by shipments shares.
37. BEA constructed a hedonic price index for personal computers in 1987.
38. Telephone switching equipment (US-SIC87 36611) in the United States accounted for about 41 percent of telephone and telegraph apparatus (US-SIC87 3661 ) in 1992. We do not have information on the relative importance of telephone switching equipment in Canada, but telephone equipment accounted for 90 percent of telecommunications equipment (C-SIC80 3351) in 1992 (Statistics Canada). In 1992 and 1996, the shares of nominal shipments for telephone and telegraph apparatus within the manufacturing sector in the United States were 0.7 percent and 0.9 percent, respectively. The comparable values in Canada for C-SIC80 3351 were 1.3 percent and 1.2 percent.
39. After 1996, the BEA price series were linked to the appropriate BLS producer price index series. The producer price index series are based upon a matched model technique.
40. To estimate the possible magnitude of the impact from the different techniques, we accepted the BEA choice of deflator for estimating telecommunications equipment growth. We replaced the real telecommunications equipment outputs and inputs in the Canadian real value added series with values constructed with an exchange-rate adjusted BEA price series. The results for the 1992-96 period indicated that the productivity gap between the United States and Canada would be slightly reduced (0.1 percent). Telephone switching equipment, where the different techniques are employed, accounts for less than half of the telephone and telegraph apparatus industry in the United States. We would expect the impact, therefore, to be even smaller than 0.1 percent.
41. In the United States, the industry group is semiconductors and related devices (US-SIC87 3674). In Canada, two components of the electronic parts and equipment industry (C-SIC80 3352), semiconductors and integrated circuits, are equivalent to the U.S. industry and will be referred to as semiconductors. Semiconductors account for 72 percent of the output in C-SIC80 3352 in 1996 (Statistics Canada). In 1992 and 1996, the shares of nominal output of semiconductors within the manufacturing sector in the United States were 1.1 percent and 1.9 percent, respectively. The comparable shares in Canada for C-SIC80 3352 were 1.2 percent and 1.4 percent.
42. The industry is US-SIC87 36741, integrated micro circuits (including semiconductor networks, microprocessors, and MOS memories).
43. The hedonic method was used to adjust for missing prices for matched models for 32 percent of one type of microprocessor and 7 percent of a second type of microprocessor. See Bruce T. Grimm, "Price Indexes for Selected Semiconductors, 1974-96," p.22.
44. For 1996, the BEA implicit price index for shipments of industry US-SIC87 3674 is 32.6; the BLS producer price index for that industry is 86.2 (both with 1992 = 100.0).
45. Consumption may be from semiconductors produced in the domestic manufacturing sector or from imports.
46. Real semiconductor output growth was removed from real value added and real semiconductor input growth was added to real value added. Then real gross output and inputs of semiconductors were re-estimated with an exchange-rate-adjusted BEA price index. The adjusted real output growth was added and the adjusted input growth was subtracted to arrive back at a measure of real value added for Canadian manufacturing. Information on the procedures used to adjust the Canadian data is available from the authors upon request. The consistency in aggregation theorem was used to make this adjustment; see Diewert, "Superlative Index Numbers," July 1978.
47. Unpaid family workers are negligible in U.S. manufacturing and are not included in the BLS estimates of U.S. employment and hours. The Canadians use the term "other-than-paid" for the concept of unpaid workers.
48. In this survey, jobs rather than persons are counted. Hours of labor input are treated as homogeneous units; no distinction is made among workers with different skill levels or wages. For more information on the Current Employment Statistics (CES) program, see BLS Handbook of Methods, Bulletin 2490 (Bureau of Labor Statistics, April 1997), pp. 15-31.
49. See Employee Compensation in the Private Nonfarm Economy, 1977, Summary 80-5 (Bureau of Labor Statistics, April 1980).
50. The direction and magnitude of any bias created by this estimation technique is not clear. Data from the CPS suggest that nonproduction workers may work more hours than production workers, in which case the estimation technique may understate hours. (See Marilyn Manser, "U.S. Labor Market Data and Issues in Comparing Goods and Services,"prepared for the 10th meeting of the Voorburg Group on Service Statistics, September 11-15, 1995.) However, the fact that production workers tend to work more paid overtime relative to nonproduction workers may cause the use of production worker hours in the estimation process to offset the issue of unpaid hours. For a bias to exist, there would have to be some change in these relationships.
51. For nonproduction workers, the ratio hours-worked to hours-paid is only available for durable and nondurable groups within manufacturing. Therefore, industry data on hours paid are adjusted with a ratio for the appropriate group. For information on the Hours at Work Survey, see Kent Kunze, "A New BLS Survey Measures the Ratio of Hours Worked to Hours Paid," Monthly Labor Review, June 1984; see also BLS Handbook of Methods, Bulletin 2490 (Bureau of Labor Statistics, April 1997), pp. 89-102.
52. For nonproduction workers, the ratio hours-worked to hours-paid is only available for durable and nondurable groups within manufacturing. Therefore, industry data on hours paid are adjusted with a ratio for the appropriate group. For information on the Hours at Work Survey, see Kent Kunze, "A New BLS Survey Measures the Ratio of Hours Worked to Hours Paid," Monthly Labor Review, June 1984; see also BLS Handbook of Methods, Bulletin 2490 (Bureau of Labor Statistics, April 1997), pp. 89-102.
53. For more information on the Current Population Survey (CPS), see BLS Handbook of Methods, Bulletin 2490, pp. 4-14.
54. The Canadian Labour Force Survey (LFS) underwent a comprehensive revision in January 2000. See "Improvements in 2000 to the LFS," on the Internet at www.statcan.ca/english/concepts/method.htm.
55. Labor force survey data are adjusted to a jobs basis. For completeness, estimates for Armed Forces, the Territories, and Indian reserves are added to the LFS data. Industry data are adjusted to remove own-account construction and noncommercial activities.
56. Katharine G. Abraham, James R. Spletzer, and Jay C. Stewart, "Divergent Trends in Alternative Wage Series," in John Haltiwanger, Marilyn E. Manser, and Robert Topel, eds., Labor Statistics Measurement Issues, National Bureau of Economic Research Studies in Income and Wealth, no. 60 (Chicago, University of Chicago Press, 1998), pp. 293-324.
57. The CPS collects data on a per-person basis, while the CES collects data on a per-jobs basis. Using data on multiple jobholders in manufacturing, we constructed an adjusted CPS series that converts the hours data to a jobs basis. This is consistent with efforts employed by Statistics Canada to convert the LFS data to a jobs basis. For 1998, this adjustment reduces the CPS average weekly hours for wage and salary workers from 42.2 to 41.5. Growth rates for the adjusted CPS series do not differ greatly from those of the published CPS.
58. The SEPH collects average hours paid, rather than hours worked. Trends in SEPH data are a reasonable approximation for this comparison if it is reasonable to assume that the ratio of hours worked to hours paid did not fluctuate dramatically over the 1988-98 period.
59. Statistics Canada.
60. Growth of the adjusted Canadian average hours series is virtually unchanged for the 1988-98 period. The adjusted Canadian series is 0.4 percent higher for 1988-92, and 0.2 percent lower for 1992-98 than the original series. Growth of the adjusted U.S. average hours series is 0.1 percent lower for 1988-98, 0.3 percent lower for 1988-92, and 0.1 percent higher for 1992-98 than the published series.
61. Because Statistics Canada does not use data from the Annual Survey of Manufacturers without such benchmarking, we presume that there are limitations associated with our establishment-based comparisons.
62. The growth of the adjusted Canadian employment series is revised upward 0.2 percent for 1988-98 and 0.4 percent for 1988-92, and is unchanged for 1992-98. The growth rate of the adjusted U.S. employment series does not change for the 1988-98 period, and is revised upward 0.1 percent for 1988-92 and downward 0.2 percent for 1992-98.
63. The growth of the adjusted Canadian series for total hours is revised upward 0.1 percent for 1988-98 and 0.9 percent for 1988-92, and revised downward 0.3 percent for 1992-98. The growth rate of the adjusted U.S. series is revised downward 0.2 percent for the periods 1988-98 and 1988-92, and revised downward by 0.1 percent for the 1992-98 period.
64. These results indicate a fairly limited impact on the productivity gap due to the two countries' uses of different data sources for the labor hours measures for the entire 1988-98 period. However, we found that these results are sensitive to the time frame selected for analysis. For example, if the Canadian data for aggregate hours were constructed on an establishment basis, the 1992 to 1996 growth in hours would be 1.1 percent rather than the published 1.7 percent. In this case, the choice of the source for hours affects the measured gap in a substantial way.
65. As discussed earlier, the aggregate manufacturing sectors in the United States and Canada are very similar, with the major exception of the treatment of logging. However, within the manufacturing industries, there are additional classification differences that must be taken into account. Using data supplied by Statistics Canada and BEA, we constructed 19 comparable two-digit industries. These reflect the US-SIC87, with the exception of textile mill products (US-SIC87 22) and apparel and other finished products made from fabrics and similar material (US-SIC87 23) that we were forced to combine. To make these adjustments, the consistency in aggregation theorem (Diewert, "Superlative Index Numbers," July 1978) was employed for both countries. We were unable to match industries perfectly, but made every effort to adjust components of industries that were significant in size, and for which data could be obtained. An industry concordance was constructed based upon four-digit industry descriptions. Where these descriptions were vague, the NAICS concordance was used as a bridge to map US-SIC87 to C-SIC80. We also reviewed the publication produced jointly by the United States and Canada which presented industry concordances. See International Concordance between the Industrial Classifications of the United Nations and Canada, the European Union, and the United States (Bureau of the Census, Statistics Canada, and EUROSTAT, undated). Of the 19 two-digit industries that are compared in this paper, 8 industries were considered one-to-one matches (tobacco; furniture; printing and publishing; chemicals; leather; stone and clay; primary metals; and transportation equipment). Food and rubber and plastics required aggregation of Canadian industries, while textiles and apparel required aggregation of both U.S. and Canadian industries. For the remaining industries, we obtained data from Statistics Canada at approximately the three-digit level to move components among industries. The following adjustments were made: asphalt roofing (Canada L-51) was moved from the paper industry to the petroleum industry; machine shop (Canada L-69) was moved from fabricated metals to machinery; office equipment (Canada L-86), which includes computers, was moved from electronics to machinery and computers; logging (Canada L-3) was added to lumber; and scientific instruments (Canada W-154 and 155) was separated out of miscellaneous to become the instruments industry.
66. 1 is the average hours share for periods 1 and 0. Because industry contributions can be constructed in different ways, we also estimated industry productivity shares as output-share weighted productivity growth and as output-share-weighted output growth less labor-share-weighted labor growth -these alternatives resulted in the same industries contributing significantly to the aggregate measures.