Patterns of advanced technology adoption and manufacturing performance.
CES researchers used the same data set in earlier studies of advanced technologies and plant performance, e.g., Dunne and Schmitz (1995). In addition, Doms, Dunne, and Roberts (1994) supplement these data with data from the 1991 Standard Statistical Establishment List (SSEL). Doms, Dunne, and Troske (1994) add data from the Worker-Employer Characteristics Database (WECD). The WECD matches employee data from the 1990 Census of Population to establishment-level data (from the 1987 Census of Manufactures) on their presumed workplaces. McGuckin, Streitwieser, and Doms (1996) used the 1993 Survey of Manufacturing Technology (which was similar to the 1988 survey) and linked it to data from the 1992 Census of Manufactures and the 1988 Survey of Manufacturing Technology. Because these studies are closely related to the present study, we briefly review them below (see Alexander 1994).
Doms, Dunne, and Roberts (1994) find that plants that adopted more of the SMT technologies experienced higher rates of employment growth and lower closure rates than otherwise similar plants using fewer technologies. The analysis controls for variables that may be associated with employment growth and plant survival, such as productivity and capital-labor ratio.(2)
Dunne and Schmitz (1995) find that the "most technology intensive" plants (i. e., those plants that used six or more of the seventeen SMT technologies) paid wage premiums of about 16 percent to their production workers and 8 percent to their nonproduction workers, compared with otherwise similar plants. Their regression analysis suggests that the technologies explain up to 60 percent of the estimated wage premium paid by large plants. They also find that the most technology-intensive plants employed relatively more nonproduction workers than otherwise similar plants.
Some evidence suggests that the technologies are complements to human capital but still have a positive association with wages. In particular, Doms, Dunne, and Troske (1994) find that including data on workers' education and occupation in regressions similar to those estimated by Dunne and Schmitz diminished but did not eliminate the positive and statistically significant association between technology adoption and wages. Their study also analyzes the association between advanced technology adoption and skilled-worker employment shares.
McGuckin, Streitwieser, and Doms (1996) found that regression analysis of labor productivity levels using the 1993 Survey of Manufacturing Technology data yielded results "remarkably similar" to those based on analysis of 1988 data. The authors found only a modest net increase in adoptions of the seventeen Survey of Manufacturing Technologies between 1988 and 1993, most of which were concentrated in computer-aided design and associated technologies and in local area networks (see also U.S. Bureau of the Census 1994, p.8). In contrast, they found that plants surveyed in both the 1988 and 1993 surveys experienced an average of four gross changes (adoptions or abandonments of technologies) for an average net increase of 0.5 in the number of technologies, suggesting a high degree of "churning" in technology adoption. The authors found only a small, positive correlation between more extensive technology adoption during the 1988-93 period and labor productivity growth,(3) although they did find that labor productivity growth was highly associated with more extensive adoption of technologies in 1988, controlling for each plant's location in the 1988 labor productivity distribution. The authors do not explore how patterns of adoption of technologies combinations had changed over time and how such changes may be associated with labor productivity growth.
Several studies on the relationship between the performance of Canadian manufacturing plants and advanced technology adoption have been written by researchers at Statistics Canada, based on data collected in surveys similar to those conducted by the U.S. Bureau of the Census. Baldwin and Diverty (1995), using the 1989 Canadian Survey of Manufacturing Technology, found results similar to those found using U.S. data (Dunne 1991 and 1994). Using regression analysis, Baldwin and Diverty found that technology adoptions are more extensive in larger plants, in plants that engage in R&D, and in certain industries, but there is no clear-cut relationship between the extent of technology adoption and plant age or diversification. Baldwin and Diverty did not attempt to examine how plant characteristics and activities were associated with the adoption of specific technology combinations.
Baldwin, Diverty, and Sabourin (1995) estimated the relationship between advanced technology adoption and plant performance by linking plant-level data from the 1989 Canadian Survey of Manufacturing Technology to panel data from the Canadian Census of Manufactures. Rather than using the number of technologies adopted as the measure of technology use, the authors looked at the performance of plants that adopt at least one technology from a functional group (design and engineering, fabrication and assembly, automated materials handling, inspection and communications, manufacturing information systems, and integration and control) and also plants that adopt from various combinations of functional groups. Baldwin, Diverty, and Sabourin found that advanced technology adopters, especially those that adopt inspection and programmable control technologies, grow faster, had higher labor productivity, and paid higher wages than nonadopters. While the performance gap is narrow between plants that adopt only fabrication and assembly technologies (e.g., numerically controlled tools) and nonadopters of advanced technologies, plants that use fabrication and assembly technologies in combination with programmable control and integrated control technologies perform much better. Unlike studies performed at the U.S. Bureau of the Census, however, the authors did not control for other plant characteristics that may be associated with plant performance.
For the most part, researchers analyzing the relationships between technology use and plant performance take the number of technologies adopted by a plant as the primary measure of a plant's technological sophistication. This method, however, obscures the diversity in technology adoption patterns and precludes analysis of the effects of specific technology combinations on plant performance. In contrast, this paper develops a richer analysis of the relationships between specific technologies and technology combinations and various measures of plant performance, including the rate of job creation and the levels and rates of change in productivity and earnings.
DIVERSITY IN ADOPTION PATTERNS
An important finding of our study is that technology adoption patterns of manufacturing plants exhibit enormous diversity, even within the same industry or the same production process. The most frequently used stand-alone technologies or technology combinations are computer-aided design, numerical control tools, and the combination of these two technologies. Each of these two stand-alone technologies or their combination is adopted only by less than 400 plants, each representing about 2 to 4 percent of the overall sample of approximately 10,000 plants. Plants that use any one of the thirteen most frequently adopted technology combinations together account for about 19 percent of all sampled plants. On the other hand, about 18 percent of plants adopt rather unique technology combinations that are found only in one or two plants. Moreover, differences among industries and types of operation (i. e., fabrication, assembly, or both in combination) cannot explain this enormous diversity of technology adoption patterns. Diversity in patterns of technology adoption is also found within each major industry group and within each type of operation (fabrication, assembly, or both).
In spite of the diversity of technology adoption patterns, there are clear linkages among the most frequently adopted technologies and technology combinations. The pattern of combination in some cases suggests a "natural" progression or technology "ladder" that plants follow when they expand their acquisition of technologies. The five most frequently adopted stand-alone (single) technologies are: computer-aided design; numerically controlled machines; programmable logic controllers; computers used for control on the factory floor and intercompany computer networks. They form the bases for most frequently adopted pairs of technologies. In fact, the three most frequently adopted pairs of technology combinations - computer-aided design and numerically controlled machines, numerically controlled machines and programmable logic controllers, and programmable logic controllers and computers used for control on the factory floor - are composed of four of the five most frequently adopted stand-alone technologies. Furthermore, the most frequently adopted combination of three technologies is computer-aided design, numerically controlled machines, and computer-aided manufacturing. This combination is one of the most frequently adopted pairs of technologies mentioned (i.e., computer-aided design and numerically controlled machines) and computer-aided manufacturing technology. Finally, the most frequently adopted combination of four technologies is the combination of computer-aided design, numerically controlled machines, programmable logic controllers, and computers used for control on the factory floor, which is the combination of two of the frequently adopted pairs of technologies - the pair of computer-aided design and numerically controlled machines, and the pair of programmable logic controllers and computers used for control on the factory floor.
TECHNOLOGY COMBINATIONS AND PLANT PERFORMANCE
Specific technology combinations generally have differing degrees of association with plant performance. Our analysis shows that simply using the number of technologies adopted by the plant as a measure of technological sophistication obscures significant differences in the relationships between specific technology combinations and plant performance. As the regression results summarized below show, this finding holds true for all measures of plant performance examined, including job growth, labor productivity, and earnings.(4)
More than 80 percent of technology categories adopted are associated with plants with relatively higher rates of job growth and higher levels of labor productivity than plants that do not adopt any of the surveyed technologies. The relationships between technology adoption and the rate of productivity growth tend to be positive but often weak. These findings are consistent with those reported in earlier studies.
Technology and Job Growth
Adoptions of most technology combinations are positively associated with overall job growth. Table 2 shows the eleven technologies and technology combinations adopted by establishments with the highest rates of employment growth between 1982 and 1987. These plants have a much different rate of job growth than plants that did not adopt any of the surveyed technologies. Local area network technologies, either combined with computer-aided design for the exchange of technical data or adopted alone for factory use, are associated with about a 25 percentage point faster rate of job growth from 1982 to 1987 than plants adopting none of the surveyed technologies. Other more complex technology combinations, many of which include local area networks technologies, are also associated with higher rates of job growth.
Table 2 Technology and Technology Combinations Relative Employment Growth (Percent) Local Area Network for Factory Use 26.6 Computer Aided Design & Local Area Network for Exchange of Technical Data 23.0 Computer Aided Design, Numerical Controlled Tools, Programmable Logic Controllers and Factory Floor Computers 14.7 Programmable Logic Controllers 12.0 Numerically Controlled Tools & Programmable Logic 10.6 Seven or More Technologies 9.4 Computer Aided Design & Numerically Controlled Tools 7.5 Computer Aided Design & Its Use in Procurement -19.3
Employment at plants that adopted a combination of computer-aided design, numerically controlled machines, programmable logic controllers, and computers used for control on the factory floor grew nearly 15 percentage points faster than plants that did not use any of the surveyed technologies between 1982 and 1987. In fact, even plants that adopted programmable logic controllers technology alone or in combination with numerically controlled machines exhibited a more than 10 percentage point faster rate of employment growth. However, plants that adopted the combination of computer-aided design and digital representation of computer-aided design output used in procurement activities experienced slower job growth by nearly 20 percentage points. Plants that adopted this particular technology combination experienced a very fast rate of productivity growth, but a substantial decline of production worker jobs, and a much slower increase of nonproduction worker earnings during 1982-87.
Technology and Productivity Levels
Plants that adopted advanced technologies generally had a relatively high level of productivity. Table 3 shows the twelve technologies or technology combinations adopted by establishments with the highest levels of productivity in 1987. The table indicates that the combination of computer-aided design and computer-aided design output used for procurement and the combination of technologies of local area networks for technical data and for factory use, intercompany computer networks, programmable logic controllers, and computers used for control on the factory floor are associated with nearly 50 percent higher productivity levels than similar plants that adopted none of surveyed technologies. These plants also experienced a much higher rate of productivity growth over the 1982 to 1987 period. Plants that adopted intercompany computer networks, computer-aided design with programmable logic controllers or numerically controlled machines, and the combination of numerically controlled machines, programmable logic controllers, and computers used for control on the factory floor are associated with 20 to 25 percent higher productivity than otherwise similar plants that did not use any of the surveyed technologies.
Jobs and Earnings of Production and Nonproduction Workers
There are sharp differences in how technology combinations are associated with growth on employment of production and nonproduction workers. Most technology combinations are positively associated with employment growth for both production and nonproduction workers, but the associations are stronger for production workers than for nonproduction workers.
Production worker employment growth is 35 percentage points higher in plants that adopted local area networks on the factory floor than in plants that adopted none of the technologies. It is 32 percentage points higher in plants that adopted the combination of computer-aided design and local area networks for technical data. Nonproduction worker employment growth is weakly associated with most technology combinations, with some significant exceptions. For example, plants that adopted flexible manufacturing cell technology had a much lower rate of employment growth for nonproduction workers, relative to plants that adopt none of the technologies, suggesting that this technology is a substitute for less-skilled nonproduction workers. (Nonproduction workers in these plants increased their earnings relatively faster than workers in other plants.)
Table 3 Technology and Technology Combinations Relative Productivity Levels - 1987 (Percent) Computer Aided Design and Its Use in Procurement 47.4 All Networks, Programmable Logic Controllers & Factory Floor Computers 47.3 Computer Aided Design and Programmable Logic Controllers 25.3 Numerically Controlled Tools, Programm- able Logic Controllers & Factory Floor Computers 24.3 Computer Aided Design & Numerically Controlled Tools 23.4 Intercompany Computer Network 21.6 Seven or More Technologies 18.8
Sharp differences also are evident in how technology combinations are associated with earnings of production and nonproduction workers. About 60 to 80 percent of the technology categories are associated with higher earnings levels for both production and nonproduction workers, but production workers benefited the most. For example, production workers at plants that adopted intercompany computer networks, the combination of computer-aided design and programmable logic controllers, and the combination of numerically controlled machines, programmable logic controllers, and computers used for control on the factory floor, earned 10 percent more than their counterparts at plants that adopted none of the technologies. Nonproduction worker earnings are only weakly related to the adoption of any of the surveyed technologies. Nonproduction workers at some plants, particularly those that adopted technologies other than computer-aided design, earned significantly less than their counterparts at similar plants that adopted none of the surveyed technologies. This is plausible because relatively fewer highly paid nonproduction workers in managerial or professional occupations, such as design engineers, work at plants that did not adopt computer-aided design.
IMPACT OF TECHNOLOGY AND PLANT CHARACTERISTICS
The relationships between technologies and plant performance are better for plants that integrate fabrication with assembly operations than for plants that engage in only fabrication or only assembly. Plants that engage in both fabrication and assembly appear to make better use of advanced technologies than those that engage in only one or the other function. For example, plants with integrated operations that adopted a complex technology combination (involving seven or more technologies) generally had higher labor productivity and production worker earnings levels in 1987 and higher production worker earnings growth over the 1982-87 period. These findings support the hypothesis that improving management by integrating different functions complements the adoption of advanced technologies. Plants that engage in both fabrication and assembly are also apt to integrate other functions, e.g., marketing, design, and manufacturing. So the finding may also reflect this integration.
We find enormous variation across manufacturing plants in their patterns of technology adoption and significant differences in the relationships between specific technology combinations and plant performance, regardless of the number of technologies adopted.
Integrated plants that use advanced technologies may forgo fabricating parts at lowest possible cost if designing parts for ease of assembly reaps more than offsetting cost savings. These gains would be reflected in some of our findings showing that advanced technologies are associated with higher levels and growth rates of labor productivity and higher production worker earnings in integrated plants. Our findings on integrated manufacturing are consistent with claims that successful implementation of advanced technologies requires close cooperation between design, fabrication, assembly, inspection, and marketing departments (e.g., see Duimering, et al., 1993 and Womack and Jones, 1994). Further efforts are needed in data collection and research on managerial innovations that complement the technologies studied in this report.
Technologies Surveyed in the 1988 Survey of Manufacturing Technologies
By Major Technology Group
Group 1: Computer Aided Design and Related Technologies
1. Computer-aided design and/or computer-aided engineering (computer aided design)
2. Computer aided design output used to control manufacturing machines (computer-aided manufacturing)
3. Digital representation of computer aided design output used in procurement activities
Group 2: Flexible Manufacturing
4. Flexible manufacturing cells or systems
5. Numerically or computer numerically controlled machines
6. Materials working lasers
Group 3: Robotics
7. Pick and place robots
8. Other robots
Group 4: Automated Materials Handling
9. Automatic storage and retrieval systems
10. Automatic guided vehicle system
Group 5: Automated Sensors
11. Automatic sensor-based inspection and/or test equipment performed on incoming or in process materials
12. Automatic sensor-based inspection and/or test equipment performed on final product
Group 6: Communications Networks
13. Local area network for technical data
14. Local area network for factory use
15. Intercompany computer network linking plant to subcontractors, suppliers, and/or customers
Group 7: Programmable Manufacturing Control
16. Programmable logic controllers
17. Computer(s) used for control on the factory floor
Source: Bureau of the Census, ESA, U.S. Department of Commerce
1 The 1988 Survey of Manufacturing Technology sampled more than 10,000 manufacturing plants with more than twenty employees in the five selected manufacturing industry groups. This entire sample is used for the analysis of technology adoption patterns. The five groups account for more than 40 percent of the U.S. manufacturing contribution to GDP and employment share. Plant performance data used in this study are from the Longitudinal Research Data Base (LRD) compiled from various Censuses of Manufactures and Annual Surveys of Manufactures. Researchers at the Center for Economic Studies matched the 1988 Survey of Manufacturing Technology data to corresponding plant data contained in the LRD for 1982 and 1987 for two subsamples of nearly 7,000 plants. These plants are the focus of the present study.
2 Doms, Duane, and Roberts estimated the equations of employment growth and plant closure rates simultaneously. Like most other related studies, the present study uses a single-equation approach.
3 McGuckin, Streitweiser, and Doms do not investigate changes in the adoptions of specific technologies or technology combinations between 1988 and 1993. Doing so may shed light on why there is a weak association between adoption of more technologies and labor productivity growth, For example, while the prevalence of programmable logic controllers and computers used for control on the factory floor declined slightly between 1988 and 1993, this may mask a high degree of substitution between these two similar technologies that have been converging in recent years (Bedworth, et al., 1991, pp. 401-403). Moreover, some technologies (computer-aided design, computer-aided manufacturing, numerically controlled tools, and automated materials handling) may have been integrated into (or disintegrated from) flexible manufacturing cells. A computer numerically controlled machine retrofit to a numerically controlled machine using a programmable logic controller or computer used for control (Bedworth et al., 1991, p. 461) may be replaced by a newer, more productive computer numerically controlled machine. Finally, technological advance within each technology would not show up in the Survey of Manufacturing.
4 The study employs regression analysis to explore the statistical associations between frequently adopted technology combinations and plant performance. The analysis controls for other observable plant characteristics to isolate the association between technology adoptions and plant performance. The dependent variables in 1987 level regressions were expressed as logarithms of the variables, whereas the dependent in the growth rate regressions were expressed as the difference in the logarithms of the 1987 and 1982 level. Two different subsamples of approximately the same size were used in these two types of analysis. The explanatory variables are represented by a set of indicator or dummy variables, each identifying whether a specific combination of technologies was adopted by the plant. Specifically, dummy variables were included for the ten most common technology combinations within each technology count class (i.e., the plants that adopted exactly one, two, three, four, five, or six technologies). In addition, for each technology count class, a dummy variable was included that represented all of the other less-frequently adopted technology combinations. Finally, we included a single dummy variable to indicate whether the plant adopted any combination of seven or more technologies. The regressions also control for other plant characteristics, including size, age, multiunit status, capital/labor ration, and four-digit SIC industry. Thus, the coefficient estimates for the technology dummy variables can be interpreted as the difference in the level of growth rates of the dependent variable associated with the adoption of a technology or technology combination, when compared with a reference group, defined as the plants that did not adopt any of the surveyed technologies.
Alexander, Lewis, Technology, Economic Growth and Employment: New Research from the Department of Commerce, U.S. Department of Commerce, December 1994.
Baldwin, John R. and Brent Diverty, "Advanced Technology Use in Canadian Manufacturing Establishments," Analytical Studies Branch Research Paper Series No. 85 (Ottawa: Statistics Canada, November 1995).
Baldwin, John R., Brent Diverty, and David Sabourin, "Technology Use and Industrial Transformation: Empirical Perspectives," Analytical Studies Branch Research Paper Series No. 75 (Ottawa: Statistics Canada, August 1995).
Bedworth, David D., Mark R. Henderson, and Philip M. Wolfe, Computer-Integrated Design and Manufacturing, (New York: McGraw Hill, 1991).
Doms, Mark, Timothy Duane, and Mark Roberts, "The Role of Technology Use in the Survival and Growth of Manufacturing Plants," U.S. Bureau of the Census, Center for Economic Studies, (April 1994) Mimeo.
Doms, Mark, Timothy Dunne, and Ken Troske, "Workers, Wages, and Technology," U.S. Bureau of the Census, Center for Economic Studies, May 12, 1994, Mimeo.
Duimering, P. Robert, Frank Safayeni, and Lyn Purdy, "Integrated Manufacturing: Redesign the Organization before Implementing Flexible Technology," Sloan Management Review, (Summer 1993)47-56.
Dunne, Timothy, "Technology Usage in U.S. Manufacturing Industries: New Evidence from the Survey of Manufacturing Technology," Center for Economic Studies Working Paper 91-7, (Washington, DC: U.S. Bureau of the Census, November 1991).
Dunne, Timothy, "Plant Age and Technology Use in U.S. Manufacturing Industries," RAND Journal of Economics, (Autumn 1994) 25:488-499.
Dunne, Timothy and James A. Schmitz, Jr., "Wages, Employer Size-Wage Premia and Employment Structure: Their Relationship to Advanced-Technology Usage at U.S. Manufacturing Establishments," Economica, (1995) 62:89-107.
Fagerberg, Jan, "Technology and International Differences in Growth Rates," Journal of Economic Literature, 32:1147-1175, September 1994.
McGuckin, Robert H., Mary L. Streitwieser, and Mark Doms, "The Effect of Technology Use on Productivity Growth," Center for Economic Studies Working Paper 96-2, (Washington, DC: U.S. Bureau of the Census, April 1996).
U.S. Bureau of the Census, Manufacturing Technology: 1988, Current Industrial Report SMT (88)-1, May 1989.
U.S. Bureau of the Census, Manufacturing Technology: Factors Affecting Adoption, 1991, SMT (91)-2, September 1993.
U.S. Bureau of the Census, Manufacturing Technology: Prevalence and Plans for Use 1993, Current Industrial Report SMT (93)-3, November 1994.
Womack, James P. and Daniel T. Jones, "From Lean Production to the Lean Enterprise," Harvard Business Review, (March-April 1994) 93-103.
David N. Beede and Kan H. Young are with the Economics and Statistics Administration, Office of Policy Development, U.S. Department of Commerce, Washington, DC. This paper is a revised version of one originally issued as ESA/OPD 95-3 in Working Papers on Industrial and Economic Performance, Office of Business and Industrial Analysis, Office of Policy Development. A more detailed version was also issued as a working paper (ESA/OPD 96-1). The opinions and conclusions expressed in this paper are those of the authors and do not necessarily represent those of the Department of Commerce or the U.S. Bureau of the Census.
|Printer friendly Cite/link Email Feedback|
|Author:||Beede, David N.; Young, Kan H.|
|Date:||Apr 1, 1998|
|Previous Article:||Economic sanctions: an emerging business menace.|
|Next Article:||The business economist at work: Ontario Teachers' Pension Plan Board.|