Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
8 The Quality of Data on Technological Change, Its Employment Effects, and Adjustment Mechanisms An important part of the charge to this panel called for a review of the adequacy of the available data on the impacts of technological advances on employment, productivity, and economic growth in the United States. For many of the questions of concern to this panel, the data are sufficient to support informed conclusions. In a number of other areas, however, especially areas of interest for further research or policymak- ing, these data are seriously deficient. The problems with the evidence are both conceptual and empirical. In many areas the only measures of technological change (e.g., productivity trends) are indirect, respond- ing to many influences other than technology, or they are direct measures that capture only a part of the processes of innovation and adoption (e.g., patents). In other areas, insufficient public investment in data collection and analysis means that the relevant data are of poor quality. This chapter surveys the deficiencies in the public data on techno- logical change and its economic impacts (in terms of employment, productivity, and output growth) and suggests a research strategy to illuminate the effects of new technology on the worker, the firm, and the workplace. In addition, we briefly discuss potential improvements in the evaluation of programs to aid worker adjustment to technologi- cal change. The broader issue is an important one; although data do not drive the policy formation process, their absence surely leaves this process less informed, less effective, and potentially counterpro- ductive. 160
QUALITY OF DATA 161 DATA ON TECHNOLOGY AND ECONOMIC PERFORMANCE Elsewhere in this report, we have criticized the use of case studies and other methodologies to predict the employment and skill impacts of technological change. Although these methodologies do offer important insights into the processes and ejects of such change, the consumer of the findings of such studies must be wary of treating them as applicable to the entire economy or even to a sector of the economy. Moreover, the large bands of uncertainty that underlie virtually all estimates of impacts must be acknowledged, both by researchers and by those who would apply these findings. A key reason for caution in interpreting and generalizing the results of sectoral or case studies is the disjunction between the aggregate and detailed levels of analysis of the employment and economic ejects of technology. Virtually all of the data on the ejects of technological change have been compiled at the individual industry, firm, or even production establishment level; statements or conclusions about overall trends, impacts, and rates of change, however, require aggregate data. Unfortu- nately, sectoral and industry studies do not aggregate well, and the transition from the detailed or sectoral to the aggregate level of analysis cannot be made with most data. Data on the aggregate economic impact of technological change in manufacturing are sparse, and their quality may be declining, due in part to reductions in federal programs of data collection and analysis. Spending under the fiscal year 1987 budget for several key federal statistical agencies has been stagnant or has declined In real terms since fiscal year 1980 (Slater, 19861.~ Budget cutbacks have produced significant deterioration in data on innovation and economic change in several specific programs. The line-of-business data of the Federal Trade Com- mission have been an important source of information about R&D and other measures of economic performance (e.g., sales and net revenues) at the level of individual product lines within large U.S. businesses. These data allow a researcher to account explicitly for the multiproduct char- acter of the modern manufacturing and service sector firm. For example, Measured in 1982 dollars, appropriations for Census Bureau programs excluding the Survey of Income and Program Participation (which analyzes household participation in federal assistance programs, including Social Security, Medicaid, and food stamps) have declined from $62 million in fiscal year 1980 to $58.6 million in fiscal year 1987; appropria- tions for the Bureau of Economic Analysis have remained constant, at $18.6 million, and appropriations for the BLS (excluding the project charged with revising the consumer price index) have declined from $121.1 million in fiscal year 1980 to $119.9 million in fiscal year 1987.
162 TECHNOLOGYAND EMPLOYMENT the line-of-business data allow one to take into account the fact that a firm like General Electric produces a vast array of goods and services (e.g., financial services), instead of assuming that its products are only those of the electrical equipment industry. Unfortunately, these data are no longer being collected, and research on the existing data base has been sharply reduced. Still another significant data collection effort within the federal govern- ment has been abandoned recently. In the 1970s a unit of the Patent Office, the Office of Technology Assessment and Forecasting, began to compile and analyze machine-readable time series data on U.S. patenting activity, assessing overall trends and analyzing patenting activity in specific economic sectors and technologies. Although they are an imper- fect measure of innovation, patents capture an important part of the overall innovation process. This unit, however, has been disbanded; its data analysis and dissemination activities have been reduced consider- ably. Any analysis of the impact of technological change is also hampered by the incompatibility of many of the data bases supported by federal agencies. As Courtenay Slater (1986), former chief economist of the U.S. Department of Commerce, has noted, the lack of a central statistical agency in the federal government has encouraged the prolif- eration of incompatible statistical series and surveys. For example, the organizational and analytic categories used by the Census Bureau are only partly compatible with the data structure developed by the BLS, which in turn is compatible with only a few of the data bases of the Bureau of Economic Analysis of the Commerce Department. There are potentially great returns to better coordination among these publicly funded data collection and organization efforts. Slater has recommended that statistical agencies (mainly the Bureaus of Labor Statistics, the Census, and Economic Analysis) be allowed and encouraged to ex- change with appropriate safeguards to ensure privacy confidential information about businesses and individuals. Such a policy also might require coordination and agreement among federal statistical agencies on the specific topics of interest in a coordinated data collection and publication effort. Another deficiency in our knowledge stems from the fact that data on technological change within the United States cover only the generation of new technology. As we noted in Chapter 2, adoption (i.e., diffusion) is crucial to a technology's economic impact; but virtually no data on adoption are collected by public statistical agencies. Devel- oping statistical series on the diffusion of innovations within the manufacturing and services sectors would be a useful investment of public funds.
QUALITY OF DATA 163 A final area of concern is the poor quality of the data on the economic behavior and performance of the nonmanufacturing sector, a large and still growing portion of the U.S. economy that now employs more than 70 percent of the U.S. work force (President's Council of Economic Advis- ers, 19874. As Representative David Obey and Senator Paul Sarbanes (1986) noted: A review of our national data collection efforts with regard to the American labor force and American businesses would lead one to believe that we are still a society of blue collar workers primarily engaged in manufacturing. While we continue carefully to count the number of people employed in the textile industry who are engaged in sewing on snaps as opposed to those who stitch sleeves, we have no information on how many Americans now work in computer sales. We do not know how many people make a living writing software or how much they make. We have no definite information on whether the Nation's movement toward a `' service economy" has helped or hurt family income or what kinds of specific skills are required in growth industries. We don't even have detailed information on what the growth industries are or how fast they are growing. (p. 2) The deficiencies in the data on service sector output, employment, and productivity to say nothing of technological change in this sector of the economy (see National Research Council, Committee on National Statis- tics, 1986Ware such that there is genuine uncertainty as to whether the apparent productivity slowdown of the past decade is significant or whether it reflects increasing problems in measuring service sector productivity growth outside of manufacturing. Neither can one distin- guish with precision among productivity, trade output, or employment trends in different service industries because of the high levels of aggregation within these data. For instance, despite rapid growth in international trade in the services sector, the categories of primary interest, such as earnings from the foreign sale of U.S. financial, insur- ance, and consulting services, are lumped together into a "miscella- neous" grouping that is the largest single category of U.S. services exports. The industrial classification scheme used for service sector data collection and analysis is in need of major revisions; such revisions would allow disaggregation of the data and analyses of the trends in more economically meaningful components of the nonmanufacturing sector. Whatever specific actions are taken, however, we feel that the quality of these data are a cause for concern. Continued neglect of the nonmanufacturing sector data base is dangerous in view of the growing importance of this sector to the national economy and to economic policy decisions.
164 TECHNOLOGYAND EMPLOYMENT A STRATEGY FOR SURVEYS OF THE IMPACT OF TECHNOLOGY ON THE WORKPLACE As we noted in Chapter 4 and the previous section of this chapter, researchers need representative data bases that combine data on the rate of technological change and diffusion with data on the changing level and distribution of employment and investment within firms, allowing aggre- gate trends to be monitored with greater precision. In addition, they require better data on the quality and character of the work environment and on workers' reactions to new technology. In this section we briefly discuss strategies for collecting such data. Surveys of Firms Hunt and Hunt (1985) noted in their survey of data sources for the analysis of technological change that the Bureaus of Labor Statistics and the Census have already collected much of the data needed to analyze aggregate trends in employment and technology. Currently, however, there is no way to link these data either with one another or with data on investment and technology. As these researchers put it: "We have occupational data, but it cannot be linked to specific technologies in use. We have demographic data, but it does not possess sufficient occupational or technological content. In the area of the technology itself, we lack even the most rudimentary data with which to address policy concerns" (p. 421. Another crucial data deficiency stems from the fact that technological change in most cases involves substituting capital for labor. To analyze such change, compatible data must be gathered on the evolving demand of firms for inputs of capital and labor. Currently, no such data exist at the firm or establishment level. Responding to these gaps, Hunt and Hunt (1985) proposed that a "Current Firm Survey" of manufacturing and nonmanufacturing firms be undertaken to complement BLS's Current`Population Survey. Like the population survey, the firm survey would be constructed as a sample of various groups within the relevant population- in the case of the estab- lishment survey, firms would be chosen from various industries. The survey would be designed to elicit detailed data on the characteristics of new capital investment in the lines of business of respondent firms; those data in turn would enable researchers to develop an accurate profile of the rates of adoption of new technologies. Similarly detailed data would be obtained on the characteristics of the firm's work force. Firm surveys have problems that are well known to anyone who has ever tried to conduct empirical studies of the behavior of companies over time. The product lines of large modern firms are extremely diverse, which makes
QUALITY OF DATA 165 the development of a balanced cross-industry sample of firms extremely difficult. Moreover, companies vanish and are born with considerable frequency, although an appropriate sampling technique could reduce prob- lems from such changes in the population. Resolving these difficulties would be well worth the time and effort involved, however, because this type of survey could provide valuable product-line data on investment, technologi- cal change, and occupational requirements. Without such data, it is unlikely that significant progress can be made in the study of technology and employment. If the costs of such a survey are judged to be excessive, a less expensive alternative is an expansion of the range of questions asked of firms in the Census of Manufactures administered by the Census Bureau. In addition, as in the population census, a sample of these firms could be asked to provide more detailed and extensive data. A Survey of Workers and Working Conditions A detailed study of the effects of technological change is also con- strained by a dearth of reliable data. One survey-based study of U.S. firms and workers (Mueller et al., 1969) was published 18 years ago, and serves as a model for the proposal outlined below. The survey of workers we propose would be more rigorous than that in the Mueller team's study; it would attempt to resurvey respondents at regular intervals to track changes in the workplace resulting from the use of new technologies and to trace the employment and skill effects of technological change on individuals. The survey also would include a matched sample of employ- ers and workers for a subset of the worker population, an addition that allows tests for biases in the survey responses and supports further analyses of technology's impacts on workers. The proposed survey would include monitoring and research studies at 5-year intervals in which respondents would be asked about their current job and the job they had 5 years earlier (the employer would be asked about the changing mix of jobs over the 5 years). The basic sample would be drawn from adults of working age (16-75 years old). In addition, through a sample of establishments, employers could be asked to provide workplace data for a sample of employees in each establishment, allowing information to be compiled both on individuals and on their interaction with a changing workplace. Follow-up interviews would concentrate on individuals and their experience. The two samples would include some overlap-for example, interviewing workers in the first sample who work for employers interviewed in the second sample. How might such a survey be undertaken? One option is to supplement the regular BLS CPS questionnaire with these inquiries every 5 years. Such a procedure, however, does not allow for tracking particular
166 TECHNOLOGY AND EMPLOYMENT workers. The alternative is a large-scale study in which a sample of workers is followed over a number of decades; such a study would resemble the large panel studies of family income and economic status that have been developed during the past 20 years. Combined with results from the detailed survey of firms proposed earlier, these data would permit an analysis of the effects of technological change on both employ- ers and employees. A similar survey was proposed in a recent report to the U.S. Depart- ment of Labor by a Social Science Research Council (1986) advisory group. This group proposed collecting information from both employees and employers. The quality of worklife was the main concern of this proposed research, but the advisory group's proposed survey could easily be modified to cover the impact of technology on both employment and the nature of work. INFORMATION ON THE EFFECTIVENESS OF WORKER ADJUSTMENT PROGRAMS The panel's charge called for an evaluation of adjustment assistance programs for workers displaced by technological change. Such an evalu- ation is difficult, however, in view of the embryonic state of research and knowledge concerning program design and effectiveness in retraining for displaced workers. As we discussed in Chapter 7, there have been few rigorous evaluations of displaced worker programs of the type being supported by Title III of the JTPA. Those that have been conducted offer limited evidence on the design of successful programs for improving the long-term employment and income prospects of displaced workers. We are encouraged by the U.S. Department of Labor's recent support of rigorous evaluations of displaced worker adjustment programs. With- out additional evaluative research and related policy experiments, the knowledge base from which to develop effective programs cannot be assembled. Evaluations are needed that will provide information on the effectiveness of adjustment program designs that combine job search assistance, basic skills training, and job-related training in different quantities and use varied delivery approaches. Evaluation of these programs also should incorporate analyses of their effectiveness in assisting displaced workers with different characteristics. The requirement for additional research need not and should not preclude the development of new initiatives to aid displaced workers. A pluralistic policy is required that will encourage the development and evaluation of different approaches to worker adjustment, including re- training in job-related and basic skills. In this regard, we concur with the report of the U.S. Department of Labor's Advisory Panel on Job Training
. QUALITY OF DATA 167 Longitudinal Survey Research (1985), which recommended that more experimental programs be conducted using randomly selected experimen- tal and control populations. A useful principle to guide the evaluation of worker adjustment assistance programs would be for the federal or state agency charged with the administration of the program to share responsibility for its evaluation with another agency or advisory group. For example, the evaluation of federal adjustment programs might be shared by the Department of Labor with another agency or outside advisory panel, and a similar procedure could be used by the Department of Education in evaluating the Perkins Act. The Advisory Panel on Job Training Longitudinal Survey Research recommended that the "research process should be monitored by a firm or group with no ax to grind in order to assure adherence to DOL's [the Department of Labor's] policy needs and maintain a quality level that will inspire public confidence in the final research products" (1985, p. 321. An expert panel could be established under the auspices of the National Research Council or another group to oversee the design and implemen- tation of evaluations. Whatever actions are taken in this area should proceed simultaneously with the development of additional innovative programs for worker adjustment. Finally, evaluations are needed of state-level programs in skills im- provement for the employed work force to determine whether this strategy is effective in improving the productivity and technological performance of U.S. firms. To what extent, for example, are these programs simply providing public funds to support training activities that otherwise would have been supported by private firms? Answers to this and other questions are difficult to obtain, but they are essential for determining the returns on investment in such training. As noted in Chapter 7, the number and heterogeneity of these state efforts provide an opportunity for comparisons and evaluations of programs that serve similar populations but use different designs. Such evaluations could also inform modifications of the federal vocational education assistance legis- lation (the Perkins Act) that allows for state support of employer-provided training to employed workers.