Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
4 Studies of the Impact of Technological Change on Employment, Skills, and Ear sings: A Critical Review This chapter reviews a number of studies that analyze the influence of new technologies on jobs, worker skills, and earnings. As in other areas of our inquiry, the extensive empirical literature covering these topics often is inconclusive and supers from methodological weaknesses. Nonetheless, several important conclusions emerge from the discussion that follows. First, new technology will not bring massive unemploy- ment; few studies predict large employment losses from such changes. Neither does it appear that, as a result of technological change, the skills required to get a job or to keep a job in the future will be substantially different from what they are today. Finally, technological change and productivity growth are associated with growth in real earnings. A1- though technological change in the U.S. economy has been cited by some as contributing to lower earnings growth and a more unequal distribution of income, there is little evidence to suggest that technol- ogy, as opposed to slow economic growth, has been responsible for these trends. THE EMPLOYMENT EFFECTS OF TECHNOLOGICAL CHANGE As we noted in Chapter 2, a number of factors interact to influence how technological change affects the level of employment in an industry or sector: · the speed with which a product or process innovation is adopted; 86
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 87 · for a product innovation, the size and rate of growth of the domestic and international markets for the new product; · for a process innovation, the size of any reductions in labor require- ments per unit of output (i.e., increases in labor productivity); · the magnitude of reductions in output prices resulting from labor productivity increases, movement down the "learning curve" (cost reductions associated with more extensive use of the new process technology), and subsequent refinements of the technology; · the size of the increase in domestic and international demand for the product in response to price reductions resulting from the adoption of a new process technology; · interindustry effects (e.g., expansion or contraction in another indus- try in response to changes in the cost of a key input); and · the. ~.ff~.~.t~ of t~c~hnolo~ical chance on wages in the industry or ~ ~ A ~ C ~ sector. These variables exert offsetting influences on the demand for labor within sectors, and they operate with varying lags. A complete accounting of all of their effects is impossible. The studies considered below, in a survey that is meant to be illustrative rather than exhaustive, all ignore one or more of the variables in this list. The range of influences considered within each study, as well as the level of aggregation at which each is conducted, varies considerably. As the number of sectors or technologies expands, however, the data requirements rapidly become overwhelming. To circumvent this difficulty, the studies cited here focus on the impacts of a single technology in many industries or on the effects of technological change within a single industry-with the exceptions of the U.S. Bureau of Labor Statistics (1986b) forecasts of employment and the policy-oriented studies by the Temporary National Economic Com- mittee (1941) and the National Commission on Automation, Technology, and Economic Progress (1966~. Policy-Oriented Studies A perception that technological change had played a role in the Great Depression led to the publication of studies of the economic effects of technological change by the Congress (U.S. House of Representatives, Committee on Labor, 1936), the National Resources Committee (1937), and the Temporary National Economic Committee (19411. Many of these 'For comprehensive surveys of this large and rapidly expanding literature, see Blair (1974), Brooks and Schneider (1985), Fechter (1974), Freeman and Soete (1985), and Kaplinsky (1987).
88 TECHNOLOG Y AND EMPLO YMENT studies reached pessimistic conclusions. The following comments from the report of the Temporary National Economic Committee are repre- sentative: . . . there is unmistakable evidence of a change in kind as well as severity of unemployment in the last depression. This change is characterized by the widespread use of electrical power and mass production methods which have shown a capacity to increase industrial activity on the upturn of the business cycle without a corresponding ability to absorb unemployed labor. (p. xvi) With the return of full employment during World War II and sustained prosperity during the remainder of the 1940s, the conclusions of these studies had little discernible impact on policy or economic research. But high (by comparison with prior years) U.S. unemployment rates during the late l950s and early 1960s,2 coupled with rates of economic growth that fell behind those of Western European nations, fueled a resurgence of the debate over the employment consequences of automation. Pessimism and concern about the consequences of technological change were reflected in such work as Michael (1962), and this concern contributed to the formation of the National Commission on Technology, Automation, and Economic Progress in 1964 (see Critchlow, 19871. The tone of this commission's report, however, contrasted with the pessimistic views that had spurred its development. The commission strongly endorsed the importance of technological change in raising living standards and improving the quality of worklife but acknowledged that its benefits were not costless. Moreover, despite its endorsement of the benefits of technology, the commission echoed the reports of the 1930s in expressing concern over a "glut of productivity." The historically unprecedented productivity growth rates of the postwar period were expected to continue, and the commission argued that increases in output per worker (i.e., labor productivity) would reduce the demand for labor if they were not offset by growth in the demand for output. Aggregate demand, the commission warned, had to be maintained at a level that ensured sufficient jobs for the growing work force. Although it recommended additional assistance for the technologically displaced, the commission concluded that if macroeconomic policy were properly managed, the probability of massive technological unemploy- ment was low because expanding aggregate demand could ensure more jobs, even in the face of an expanding work force and growing labor productivity. Such optimism rested on the apparent triumph in the early 2Annual unemployment averaged 5.8 percent during 1958-1962, well above the average rate of 4 percent that prevailed during 195~1957 (President's Council of Economic Advisers, 1987, Table B-35).
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 89 1960s of policies for the management of aggregate demand. By the time the commission's report was released in 1966, however, the economic and political outlook had changed dramatically. Concern over the impacts of new technology had declined, in part because the U.S. unemployment rate was only 3.8 percent, having fallen from 5.2 percent in 1964 in response to an expansionary fiscal policy (President's Council of Eco- nomic Advisers, 19874. This improvement in the economic environment, as well as the escalating U.S. involvement in the Vietnam conflict, meant that the commission's policy recommendations were largely ignored by the Johnson administration. During the 1970s the employment conse- quences of technological change received little attention, but the subject returned to a position of prominence in public debate in the 1980s. Studies of Individual Firms, Industries, or Occupations A recent survey by Flynn (1985) analyzed almost 200 case studies of the employment effects of process innovations during 1940-1982. The technological advances considered by Flynn were evenly divided be- tween those affecting the automation of production or distribution and those affecting office automation. Process innovations in skill-intensive manufacturing processes often eliminated high-skill jobs and generated low-skill jobs. The opposite was true, however, for the adoption of data- and word-processing technologies in offices, which eliminated low-skill jobs and created high-skill jobs. Flynn concluded that the net effect of process innovations on employment was indeterminate and depended heavily on conditions within individual industries or firms. Hunt and Hunt (1986) surveyed the effects of technological change on clerical employment. The authors criticized several other studies of this topic for overlooking the often slow pace of technological change and diffusion, the output-expanding impacts of reductions in the price of such clerical or secretarial activities as text editing, and the effect of expanding aggregate demand. They argued that these flaws led the studies to overstate the job-displacing impact of technological change on clerical workers: The forecasts of declining clerical employment are based on over-optimistic expectations of technological improvements or exaggerated productivity claims on behalf of existing technology. In our opinion, current office technology offers significant improvements in product quality and modest improvements in produc- tivity. There is as yet no empirical evidence of an office productivity revolution that will displace significant numbers of clerical workers. (p. 65) Osterman (1986) also studied the impact of information technologies on
90 TECHNOLOG Y AND EMPLO YMENT office and clerical employment in several industries and found that displacement was partly offset by an expansion in the demand for automated activities or functions. Although the adoption of computers initially reduced the employment of clerks and managers in these indus- tries during 1972-1978, displacement typically was followed in a few years by increases in clerical and managerial employment. The timing of the employment-displacing and employment-expanding effects of technological change in Osterman's study suggests the empirical problems that result from differences in the rates and timing of produc- tivity growth, cost reduction, and output and employment growth. According to Osterman, the increases in employment that followed the introduction of computers generally were insufficient to overcome the employment losses. Over a longer period, however, the net employment losses might well have been smaller or nonexistent. Osterman also did not consider the employment effects of new jobs created elsewhere within the firms adopting computers. Nevertheless, differences in the timing of employment displacement and creation mean that the workers who are initially displaced may not be the individuals who are subsequently hired. Significant displacement problems thus may develop even in the face of expanding employment opportunities. In their recent analysis of office automation, Roessner et al. (1985) present conclusions that contrast sharply with those of the National Research Council's Panel on Technology and Women's Employment. In its 1986 report the panel concluded that "massive job loss is unlikely to occur" (p. 125) within clerical and office occupations as a result of technological change. The Roessner team, on the other hand, projected that office automation could displace as much as 40 percent of 1980 clerical employment within the financial services and insurance industries by the year 2000. To reach this conclusion the authors surveyed experts on likely improvements in office automation technologies and applied these forecasts to a functional taxonomy of clerical tasks. They assumed that the functional composition of typical clerical tasks and duties would be unaffected by technological change during 1980-2000. They also assumed that technological change and diffusion would be rapid and minimized or dismissed the possibility that the enhanced productivity of clerical workers might increase the demand for clerical services. Finally, the study ignored the employment implications of product innovations that result from office automation technologies, even though executives within the financial services industry, among other sectors, have cited such innovations as important sources of employment growth. Denny and Fuss (1983) investigated the effects of automation on occupational groups within Bell Canada, using data on four separate occupations and a direct measure of the rate of technological change
STUDIES OF THEIMPACTOF TECHNOLOGICAL CHANGE 91 (based on the share of direct distance dialing in total telephone traffic). Technological change in Bell Canada during 1952-1972 increased the amount of capital and reduced the amount of labor per unit of output, with the laborsaving effects felt most strongly in the least skilled occupations. The study found, however, that net employment growth within these occupations was positive because output growth more than offset the impact on employment of reductions in labor requirements per unit of output. The Denny-Fuss study did not deal with the potentially employment-creating effects of these innovations on other industries or on occupations within Bell Canada beyond the four considered. Levy et al. (1984) analyzed the interactions among technological change, growth in productivity, and growth in output and employment in a number of industries. They assessed the effects on output growth and employment of labor productivity growth resulting from technological change and increases in production plant scale during 1960-1980 in five manufacturing and mining industries (steel, aluminum, automobiles, coal mining, and iron mining). Within all these industries, technological change led to the substitution of capital for labor and to increases in labor productivity (although steel exhibited a very low rate of technological change), a finding similar to that of Denny and Fuss. An important improvement in this analysis, however, is the Levy team's consideration of the effect of productivity growth on the demand for the output of these industries. By lowering prices and increasing the demand for industry output, labor productivity growth supported employment growth that offset much or all of the reduction in labor demand associated with the productivity-increasing impact of technological change. In three of the five industries (coal mining, iron mining, and aluminum production), the output-enhancing effect of technological change increased total employ- ment; in the other two (steel, where technological change was minimal, and automobiles), demand growth was insufficient to offset the impact of reductions in the labor required per unit of output. Studies by Ayres and Miller (1983) and Hunt and Hunt (1983) consid- ered the impact of robots on manufacturing employment. Ayres and Miller concluded that current robotics technologies could displace 1.5 million jobs in current manufacturing and as many as 4 million by 2005. Hunt and Hunt, on the other hand, estimated that total employment displacement by 1990 would amount to only 68,000-134,000 jobs well below levels of normal turnover within the manufacturing work force. (Turnover in U.S. manufacturing averages more than 20 percent per year, based on data from 1976-1980 cited in Levy et al., 1984.) One reason for these divergent estimates is Ayres and Miller's assumption that diffusion and technological improvement within robotics would be rapid. Ayres and Miller's study focused on the technological "frontier" and
92 TECHNOLOGY AND EMPLOYMENT considered the number of jobs that potentially could be performed by robots in 2005. The alternative approach, developing a model that incorporates adoption costs and diffusion rates, places greater emphasis on the length of time needed for investment in and adoption of the new technology. The Ayres-Miller study also surveyed a small and narrow sample of firms and industries (16 firms, almost all of which were in the automotive industry and therefore contained a high proportion of jobs that could be performed by currently available robots). The empirical basis for the less dramatic dis- placement estimates of Hunt and Hunt's 1983 work, which made explicit assumptions about rates of adoption of robotics technologies and employed a broader data base for estimates of employment effects, seems stronger than that for Ayres and Miller's predictions. Neither study considered the employment implications of the potential growth in output resulting from the positive effects of robots on manufacturing productivity growth, although Hunt and Hunt compared their displacement estimates with the BLS estimates of employment growth in affected occupations through 1995. The contrasting results of these studies, like those of the studies by Roessner et al. (1985) and the National Research Council's Panel on Women's Employment and Technology (1986), illustrate the sensitivity of empirical estimates of the employment impacts of technological change to detailed assumptions concerning diffusion rates, technological improvement, and the organization of manufacturing and office production processes. Yet prediction of these vanables, which is necessary for forecasts of employment impacts, is extremely difficult and frequently incorrect; therefore, the forecasts based on such assumptions are often unreliable. An important collection of sectoral studies of employment and techno- logical change in Great Britain recently has been completed by the Science Policy Research Unit of the University of Sussex. Known as the TEMPO (Technological Trends and Employment) project, these studies (Clark, 1985; Freeman, 1985a; Guy, 1984; Smith, 1986; Soete, 1985) analyzed recent trends in technological change, productivity growth, and employment in 17 British manufacturing and service industries. The project focused on sectoral studies because of the evidence that the impacts of technological change on productivity and employment growth varied greatly across sectors.3 The TEMPO studies also undertook forecasts of employment through the late l990s. The analytic framework used by most of the studies (with the exceptions of Ray, 1985, and the studies of the services sector in 3"A broad macroeconomic approach was therefore deemed to be inadequate for assessing the specific employment effects of technological change. It was felt that only in-depth studies of each of the main sectors would encompass the full range and variety of technical change" (Guy, 1984, p. vii).
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 93 Smith, 1986) is discussed in Clark and Patel (19841; it relied on estimates of investment and the rate of growth of the capital stock to compute measures of growth in "best-practice" and "average" productivity for both labor and capital. ("Best-practice" productivity is the level attain- able with the latest process technologies; "average" represents the actual level of measured productivity.) Estimation of these trends relied heavily on imperfect data on the value of the capital stock in industrial categories that are highly aggregated; the estimates also incorporated strong assump- tions concerning the rate of growth in the productivity of new technolo- gies. The methodology demands considerable data, and the absence of an aggregate analytic framework precludes the examination of interindustry effects of the type that are salient within input-output analysis (see below). Nonetheless, by emphasizing the roles of capital formation and diffusion in the growth of sectoral productivity and employment, this methodology makes an important contribution. For many of the sectors in the TEMPO series, projected employment growth was low or even negative; these predictions were affected by the inability of the researchers to take into account interindustry linkages, by the low rate of growth of the British economy during the late 1970s and 1980s, and by the extensive penetration of many British markets by imports. Many of the studies examined capital productivity trends and reached conclusions resembling those of Baily (19861; in a number of British industries during the 1970s and early 1980s, the measured produc- tivity gains from additional investments in physical capital appear to have declined somewhat for reasons that are not well understood. Freeman (1985b), for example, suggested that the radical nature of many new technologies made it difficult for British firms to exploit their productive potential rapidly.4 As this survey of the empirical literature suggests, few case studies are able to consider the complex effects of technological change on employ- ment beyond the confines of a single firm, industry, or occupation. A study of the effects of robotics on assembly line workers, for example, may estimate the worker displacement that occurs due to one aspect of this technological change, but it cannot assess all the employment impacts of the new technology. Such an assessment requires additional informa- tion on the number of jobs created in designing, manufacturing, and servicing robotics machinery, as well as data on the effects on prices, 4Freeman (1985b) stressed the ". . . tendency to diminishing returns with incremental innovations and economies of scale in the older electro-mechanical plant and equipment of the 1960s . . . ," and the ". . . failure to exploit the full productivity potential of the revolutionary new technologies, based on computerization, because of the piecemeal pattern of implementation and the lack of necessary skills" (p. 77).
94 TECHNOLOG Y AND EMPLO YMENT demand (how much will demand for the product increase if the price is reduced?), and, consequently, employment in all industries affected by the robotics technology. A broad analytic perspective is needed to capture interactions among firms, industries, and occupations, as well as changes over time in these effects. Despite their value as descriptions of potential employment impacts, sectoral studies cannot incorporate these complex interactions and should not be relied on for forecasts of the aggregate employment impacts of technological change. Aggregate Analyses Input-output analysis can incorporate the interactions among indus- tries that are essential to determining the total employment effects of technological change. The expanded scope of such an analysis, however, creates extensive data requirements. Input-output analysis requires the estimation of "input-output coefficients," which describe the amount of labor and each industry's output needed to produce the outputs of all other industries in the analysis.5 The effect of technological change on these coefficients must be estimated, and final demand for each good must be projected in forecasts of the employment impact of new technologies. The input-output coefficients in many cases are invariant with respect to price: doubling the cost of an input need not affect the amount of that input consumed by an industry. Thus, most forms of input-output analysis can account only for changes in interindustry relationships that are based on the technologically driven substitution of one input (e.g., capital) for another (e.g., labor). Recent applications of input-output methodology to the analysis of the employment effects of new technologies largely ignore changes in final demand and in the demand for inputs that result from changes in price. This means that there is no link between growth in labor productivity and growth in demand within a specific industry. Because input-output analysis typically projects the existing matrix of output and input require- ments forward in time, predictions based on this methodology also have difficulty incorporating the employment effects of product innovation. Howell (1985) used an input-output framework to forecast the employ- ment effects of industrial robots. His methodology required projections of the use of robots in each of 86 industrial sectors in 1990, as well as estimates of input-output coefficients that measure the robotics indus- try's consumption of the output of other industries in the production of robots. Howell considered the employment consequences of six different 5Leontief and Duchin (1985) used an 89-industry input-output table.
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 95 estimates of the number of installed robots in 1990, ranging from 72,000 to 285,000.6 Howell's analysis did not consider the increases in employment that might result from reductions in price and increases in demand associated with the diffusion of robotics technology. For example, the introduction of robots within an industry might lower prices and increase demand for the industry's output, but Howell's methodology largely ignores the employment effects of such potential growth in output. Using a methodology that may overstate the potential employment displacement due to robots, Howell concluded that the net number of jobs displaced by robots by 1990 would range from 168,000 (assuming slow diffusion of robotics) to 718,000 (for the most rapid assumed diffusion rate). The latter figure is only 0.7 percent of total U.S. employment and 3.7 percent of manufacturing employment in 1986; it accounts for an even smaller share of total projected 1990 employment. Leontief and Duchin (1985) undertook the most extensive input-output study of the effects of computer technology on employment. Their study concluded that the widespread use of this technology would reduce employment in the year 2000 to approximately 8-12 percent below the levels that would be needed to produce this output with an unchanged technology. The Leontief-Duchin study illuminates the serious limitations of aggregate studies of technological change. The study's assumptions concerning the quantity of labor displaced as a result of computer diffusion were based on limited evidence from case studies. Rates of diffusion and technological change were assumed to be rapid, but the authors did not allow for any output- and employment-expanding erects of reductions in the costs of clerical and other functions as a result of technological change and productivity growth. A more serious defect of the Leontief-Duchin study is that it combined an economy-wide analysis of employment impacts with the assumption that advances would occur in only one technology. For example, as a result of assuming that no technological change beyond that in computers would occur within the agricultural sector and that demand for total output would grow, the authors projected employment gains in farming by the year 2000. Such an outcome is open to considerable question. The Leontief-Duchin projections of reduced employment by the year 2000 were criticized by the National Research Council's Panel on Technology and Women's Employment (1986), which concluded that "there is insufficient evidence to support the . . . negative outlook of the Leontief- Duchin study" (p. 1111. This panel concurs in that assessment. 6In 1986, according to the Robotic Industries Association (1986), the U.S. stock of industrial robots was slightly more than 25,000.
96 TECHNOLOG Y AND EMPLO YMENT The most recent aggregate projections of employment that incorporate the effects of technological change were prepared by BLS for 378 industries and 562 occupations in 1995. In the past the BLS projections have proved quite accurate in tracking changes in employment, although they tend to understate employment growth in the fastest growing occupations and employment decreases in declining occupations (Carey, 1980; Carey and Kasunic, 1982; Rumberger and Levin, 19841. For its 1986 forecasts, the bureau used a system of five interconnected economic models to project growth in the labor force, the level of aggregate economic activity, each industry's output of goods and services, and each industry's demand for labor. The industry demand projections were based on historical relationships between growth in GNP and growth in the output of individual industries.7 In recent years the bureau has disclosed much of its methodology for measuring and incorporating technological change within these projec- tions, and further disclosure and discussion of these methods are highly desirable. Technological change was incorporated into the BLS projec- tions through assumptions about the rates of development and diffusion of new technologies and their direct impacts on occupational structure and input-output coefficients. BLS also allowed for changes in output de- mand resulting from productivity increases and changes in production processes within industries. The bureau was conservative in making technology adjustments in these models. Nevertheless, hundreds of adjustments were made in the most recent revision of the projections (Hansen, 19841. The 1995 employment projections issued by BLS forecast growth in virtually all of its more than 350 occupations with at least 25,000 workers. Some categories, however, were projected to decline in absolute terms as a result of technological change. Information technologies affected a number of the 11 occupations posting absolute declines in projected employment as a result of technological change (Table 4-11. The declining occupations fall into four groups: office workers involved primarily in data-entry tasks; communications workers who are displaced by declines in the service requirements of telecommunications equipment; truck and tractor operators affected by increases in warehouse automation; and 'The moderate-growth scenario used by BLS as the basis for industrial and occupational employment projections assumes strong productivity and investment growth, a declining unemployment rate (6 percent in 1995), and a real annual rate of GNP growth of 2.9 percent between 1984 and 1995; using those assumptions, the bureau forecast that employment in the U.S. economy will increase by 16 million. For a detailed discussion of these assumptions and those underlying the alternative low- and high-growth scenarios, see U.S. Bureau of Labor Statistics (1986b).
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 97 TABLE 4-1 Occupations with 25,000 or More Workers Forecast by the U.S. Bureau of Labor Statistics to Experience Net Employment Declines Due in Part to Technological Change: Moderate Growth Scenario, 1984-1995 Occupation 1984 Employment (000) Net Employment Decline (000), 1984-1995 Percentage Share of Decline Stenographers 239-96 40 Industrial truck and tractor operators 389-46 12 Postal service clerks 317-27 9 Station installers and repairers, telephone 111- 19 17 Stock clerks 788-16 2 Statistical clerks 93-12 13 Payroll and timekeeping clerks 207- 11 5 Central office telephone operators 77-9 11 Order filers, wholesale and retail sales 226-7 3 Service station attendants 303-6 2 Directory assistance telephone operators 32-2 7 Total decline -251 SOURCE: U.S. Bureau of Labor Statistics (1986b). service station attendants displaced by the use of information technology in self-service gas stations. In total, BLS predicts that there will be 251,000 fewer employment opportunities in these occupations by 1995. This is equivalent to 1.6 percent of total projected employment growth during 1984-1995. A recent retrospective analysis of the employment effects of techno- logical change used input-output methodology to decompose the growth in employment during 1972-1984 into changes resulting from growth in final demand and those resulting from technological advance in 79 industries (Young and Lawson, 1986~. The authors computed the changes in employment that would have resulted if the 1984 output had been produced with 1972 technology. This calculation yields the change in employment that is attributable solely to growth in demand. The "constant technology" employment change is subtracted from the actual 1972-1984 employment change. The difference provides an esti- mate of the effect of technological change on employment during 1972-1984.
98 TECHNOLOG Y AND EMPLO YMENT The Young-Lawson methodology did not require the forecasts of rates of diffusion and technological progress that were so salient in the BLS, Roessner et al., and Leontief-Duchin studies. Like the Roessner and Leontief-Duchin studies, however, the Young-Lawson study was unable to determine the employment impact of output growth that resulted from price reductions stemming from productivity growth. As a result, Young and Lawson attributed many of the employment-increasing effects of technological change to changes in final demand rather than to the expansionary impact of new technology. Technology-related effects on employment in this analysis incorporate only the labor-displacing eBects of changes in process technology, a procedure that ignores Levy et al.'s (1984) "output enhancement" effect. In addition, Young and Lawson could not isolate the reductions in industry demand for domestic inputs that resulted from increases in imports of foreign inputs. This effect, which had negative employment implications, was categorized by Young and Lawson as technology based, although ironically it may reflect the absence of technological change within U.S. industry. Young and Lawson found that technological change during 1972-1984 reduced labor requirements per unit of output in 65 of 79 industries. Changes in final demand during this same period affected some or all of the decline in labor demand in 73 of the 79 industries.8 In 44 of the 79 industries, the laborsaving effects of new technologies were more than offset by growth in final demand; that is, total employment expanded. Even this assessment of the employment effects of technological change, which overstates the job displacement consequences of it, suggests that sectoral employment displacement often is more than offset by output growth. Summary Process and product innovations affect prices, wages, incomes, and international trade Dows. Unfortunately for analysts, all of these forces operate simultaneously and interact. Forecasting methodologies for the assessment of the sectoral or aggregate employment impacts of techno- logical change remain primitive, and therefore any results must be viewed with skepticism and caution. Although this brief survey of recent empirical studies should lessen concerns about massive technological unemployment, technological dis- placement remains a potentially serious problem for some workers in the Sixty-one of the 65 industries in which technological change had a laborsaving impact also experienced increases in labor demand as a result of changes in final demand.
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 99 American economy of the 1980s and 1990s. Differences in the timing of the labor-displacing and labor-enhancing effects of technological change, coupled with the fact that jobs may be lost in industries, regions, or occupations that are very different from the ones in which job creation occurs, make it difficult for those workers and managers who must change their occupations and learn new skills. Indeed, the entire issue of skills- those a worker needs to get a job and those necessary to do a job-is central to the debate about the effect of technological change on employ- ment. SKILL REQUIREMENTS AND TECHNOLOGICAL CHANGE Recent poll data (Cambridge Reports, Inc., 1986) confirm that the public remains concerned that technological change will make the skills of the average worker obsolete and will raise the level of skill required for good jobs (defined in terms of wages or advancement prospects) beyond the reach of those workers entering the labor force or changing jobs. Two types of worker skills are relevant to our discussion of the erects of technological change on skill requirements. Basic skills typically are acquired by workers prior to entering the labor force. They consist of literacy, problem-solving, numerical reasoning, and written communica- tions competencies. Within this economy, workers acquire basic skills through the public and adult educational systems rather than through employers. U.S. employers, however, historically have been willing to provide the specific occupational or job-related skills required by their employees that is, those skills necessary to perform a specific job or function. Basic cognitive skills and job-related skills appear to be complemen- tary; the real contribution of basic skills to productivity lies in helping workers learn what they need to do their current and future jobs (Bishop, 1984; COSEPUP Panel on Secondary School Education for the Changing Workplace, 19841. Most empirical studies of the impact of technological change on skills focus on job-related skills, which in many instances are acquired through on-thejob training (see Chapter 71.9 A substantial body of literature on the skill impacts of technological change has reached few consistent conclusions. There are a number of 9Computer skills are one example of job-related skills. Goldstein and Fraser (1985) studied the sources of computer training and concluded that most workers were able to obtain such training on the job. The amount of training required for computer users was modest; no more than 5 percent of computer users actually required extensive computer training.
100 TECHNOLOGY AND EMPLO YMENT reasons for this, many of which also apply to the literature on the employment impacts of technological change: · The methodologies and data used in studies of technological change and skills are weak and imprecise. · There is little agreement on the definition (and therefore the mea- surement) of job-related skills. Analysts often disagree as to whether skills are an attribute of individuals in which case they would be related to educational attainment and would be portable among jobs-or whether skills are highly specific to firms or occupations and only loosely related to educational attainment. · To study skills, it is also necessary to study occupations; but data on the U.S. occupational structure are unreliable for comparisons over time because of extensive revisions in the classifications in successive pub- lished dictionaries of occupational titles and categories (see Spenner, 1985, for a detailed discussion). · Case studies of the impacts on skills of specific technologies or of technological change within a specific industry rarely consider a lengthy period of time; thus, they are unable to trace changes in skill require- ments as a technology, industry, or production process passes through different stages of its development or diffusion (see Chapter 21. · The skill effects of technological change are sensitive to the ways in which new technologies are implemented in the workplace. Managers have considerable discretion in such implementation, which may affect skill requirements (Adler, 19841. Thus, identical innovations introduced in different firms can alter skill requirements in different ways (Spenner, 1985~. The literature on technology and skills includes aggregate studies, which examine changes in skills within a large number of occupations and industries, and case studies, which focus on a single occupation, firm, or industry. Aggregate studies obtain greater coverage but at the cost of overlooking certain types of skill change. They generally report greater stability in skill requirements because they aggregate industry- or firm-specific variations. Spenner's review of 11 aggregate studies (1986) is consistent with this description; he found no evidence to support claims of significant upgrading or downgrading in aggregate skill requirements as a result of technological change. Levin and Rumberger (1986) used educational attainment as a proxy for skill requirements in their analysis of the overall skill implications of the BLS employment forecast for 1982 through 1995 and reached similar conclusions. Indeed, these scholars found that the educational requirements of projected 1995 jobs were virtually identical to those needed for 1982 jobs.
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 101 Case Studies of Manufacturing Case studies sacrifice the breadth of the aggregate approach but offer a detailed understanding of skill changes within a single occupation, firm, or industry. These studies generally show more change and volatility in skill requirements (Spenner, 19851. Flynn's 1985 survey of process automation, which was discussed earlier, also considered the impacts of technological change on skill requirements. According to Flynn, the auto- mation of high-skill jobs shifted their content from the direct operation of a machine to monitoring the operations of a different machine that was more nearly self-controlling. Do these changes in job content represent a reduction of work skills or a shift from higher-level mechanical to higher-level mental skills? Flynn argued that the changes resulting from new technology reduced skill requirements; although process automation often increased the level of worker responsibility, the use of technologically sophisticated equipment made operator positions less demanding. Hirschhorn (1984) reviewed similar evidence on changes in job content but reached the opposite conclusion, arguing that the shift toward increased responsibility required higher-order mental skills to ensure quick and appropriate responses to mechanical breakdowns. (The Three Mile Island and Bhopal accidents are examples of breakdowns in complex systems in which initial operator responses aggravated the problem.) These conflicting interpretations reflect the problems of defining and measuring skills. Flynn found that in addition to transforming the content of many jobs, automation created new jobs at both ends of the skill spectrum. New, lower-skill jobs required less judgment, skill, and discretion than the previous higher-level craft and operative positions. The adoption of computerized process control systems, on the other hand, created new jobs for computer programmers and systems analysts. These systems also required more advanced technical knowledge for supervisory, maintenance, and technician tasks. The evidence summarized by Flynn, however, does not include estimates of the proportion of high- and low-skill jobs created as a result of technological change. Complementing Flynn's work is the recent report by the National Research Council's Committee on the Effective Implementation of Ad- vanced Manufacturing Technology (19861. The committee found that the introduction of automated manufacturing technologies reduced the num- ber of job classifications while broadening the scope of activities within each classification. The new groupings typically involved a broader range of skills, reflecting greater numbers of machines, an expansion of the range of operations for which a worker was responsible, or the rotation of workers through different jobs. Finally, Baran (1986) notes that the effects
102 TECHNOLOG Y AND EMPLO YMENT of product innovation and redesign (e.g., substituting microelectronics for electromechanical components in office equipment) on skills are also considerable. Case Studies of Office Automation Numerous case studies of office automation have analyzed the impact of a single group of technologies information and computer technolo- gies on skill requirements. A review of these studies suggests that the impact of these technologies on job skill requirements has changed as the technologies have developed, a phenomenon consistent with the discus- sion in Chapter 2 of the skill-intensive characteristics of new technologies in the early stages of their development. As a result of this characteristic of technological development, successive studies of these technologies have reached different conclusions. Studies of office automation in the 1950s and 1960s (summarized in Flynn, 1985) found that office and "back-office" (transactions processing, recordkeeping, data entry, and other functions involving little or no customer contact) automation created new employment opportunities for skilled computer programmers, systems analysts, computer maintenance engineers, and other administrative and managerial personnel. At the same time, back-office automation eliminated low-skill clerical positions but created positions for low-skill data-entry workers. Many of these early case studies reported "job enlargement" for clerical positions as the personnel occupying them became less specialized and absorbed new, computer-related tasks. As computer technology developed, however, the number of higher-skill opportunities for clericals appeared to decline. Many case studies during the 1970s (see Baran, 1986) reported that automation fragmented and standardized clerical work, requiring lower- level and narrower skills. The most recent set of case studies suggests that a new wave of computer technology, supporting the movement of office automation out of the back office and into desktop and distributed data processing, may be reversing these tendencies toward reductions in skill requirements. Baran (1986) reports that the introduction of minicomputers, personal computers, and higher-level programming languages has restructured office work. The insurance clerical worker of the future, for example, is likely to have a computerized workstation. Because of increased desktop computing power, this worker will be responsible for a wider range of tasks including rating, underwriting, issuing new policies, and policy updating and renewal. Continued advances in data-processing and office automation technologies have also changed the skills required for many of the support personnel employed in data-processing departments. Consistent with the argument that
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 103 skill requirements change over the life cycle of a technology, many of the operating tasks assigned to engineers in the 1950s shifted to technicians in the 1960s; in the 1980s, they appear to be shifting to clerical employees (Flynn, 1985). Like the literature and evidence on the employment impacts of tech- nological change, the empirical evidence of technology's effects on skills is too fragmentary and mixed to support confident predictions of aggre- gate skill impacts. Despite this uncertainty, however, the evidence suggests that the skill requirements for entry into future jobs will not be radically upgraded from those of current jobs. Many of the computer- based technologies examined by this panel are now being developed with more powerful software and user-friendly interfaces, which should reduce the device-specific skills needed to operate them. As more such "intelli- gence" is embedded in hardware or software, users will require less training for particular equipment. Consequently, the workplace of the future will place a greater premium on a strong foundation in basic skills for career advancement and for changing jobs but should not require massive investments in computer literacy for all entrants or employees. Even more than in the analysis of the employment impacts of techno- logical change, the evidence on skill impacts has led us to stress the considerable uncertainties that pervade the issue. In examining educa- tional and other policy responses to the challenges of technological change, it behooves policymakers and others to avoid planning based on inflexible commitments to a single (and almost certainly flawed) vision of the skill and vocational requirements of the workplace of the future. Nonetheless, education and training that improve the basic and job- related skills of American workers are important contributors to U.S. competitiveness and living standards. Continued investment in the train- ing of professional scientists and engineers to sustain the development and adoption of new technologies is also critical. THE EFFECTS OF TECHNOLOGICAL CHANGE ON THE LEVEL OF EARNINGS Technological change and its effects on earnings have long been topics of debate among economists and other analysts. Poll data (Cambridge Reports, Inc., 1986) suggest that a large segment of the U.S. public views technological change as a force that may erode wages, leading (among other things) to a polarized wage and income distribution. This section reviews the evidence on the impact of technological change on the level of wages and considers the relationship between technological change and the distribution of earnings (i.e., salaries and wages) and income within the United States.
104 TECHNOLOGY AND EMPLO YMENT 20 18 ~ . ~^ In · v o ._ - E - tn lo: A: o Cal so 14 12 10. 8 64 4 Hourly Output _~ _~ .~ D ~ ~ Hourly Compensation O- 1 1 1 1 1 1 1 1950 1955 1960 1965 1970 1975 1980 1985 YEAR FIGURE 4-1 Real output per hour and real employee compensation, 195~1985. SOURCE: U.S. Bureau of Labor Statistics, Office of Productivity and Technology. Developed by the bureau from U.S. Department of Commerce, Bureau of Economic Analysis, national income and product account data. Growth in Real Earnings During the Postwar Period A widely accepted measure of real earnings growth is average real compensation (wages and salaries plus employee benefits) per hour in the nonfarm business sector. Figure 4-1 plots trends in this quantity and in labor productivity over time, revealing a close relationship between the two.~° The share of labor in total output has remained fairly stable throughout the postwar period in the U.S. economy, in contrast to its behavior in Western European economies in which, according to Bruno and Sachs (1985), this share fluctuates. Increases in U.S. real compensa- tion therefore depend on growth in labor productivity; far from supporting erosion in real earnings, technological change, by increasing labor pro- ductivity, is associated with increases in them. The stagnation in U.S. real '°Figure 4-1 plots real nonfarm output per hour and real compensation per hour. Both series are deflated (converted into constant dollars) by the implicit nonfarm output deflator.
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 105 earnings that has occurred since 1973 reflects lagging labor productivity growth. Improvements in real earnings within this economy depend on renewed productivity growth, which in turn requires more rapid genera- tion and adoption of new technologies. The Impact on Compensation of Worker Movement Among Sectors The growth rate of average compensation within the overall economy can be broken down into a weighted average of the growth rates of earnings within each sector and a composition erect, reflecting the impact on average compensation of shifts in the shares of total employ- ment accounted for by sectors with different levels of average earnings. Costrell (1987) broke down growth in real hourly compensation into compositional edects and changes in real compensation growth within sectors (the 12 sectors of Table 3-7) and obtained striking results. The effect on real compensation growth of changes in employment shares was modest and, if anything, positive (approximately 1-2 percent) prior to 1979. During 1979-1985, however, the impact on average compensa- tion of changes in employment shares became negative and increased in size (to more than 10 percent), consistent with the findings of Bluestone and Harrison (1986) for 1979-1984. Yet real compensation growth within sectors remained positive during 1979-1985, increasing by almost 7 percent. Costrell identified the decline in the share of durables manu- facturing employment and growth in the share of services employment during 1979-1985 as the major contributors to the negative compensa- tion impact of intersectoral employment shifts. This finding is qualified, however, by the small number of workers that have actually moved from the manufacturing to the nonmanufacturing sector; absolute levels of manufacturing employment have not fallen below the levels of the 1960s. Although these estimates are based on aggregate data and represent average compensation losses, they are broadly consistent with survey data on earnings losses among displaced manufacturing workers (Podgursky, 1987~. Relatively well-paid, unionized, blue-collar workers in durables manufacturing have made disproportionate contributions to the displaced worker population, as was noted in Chapter 3, and many (but not all) of these workers have found new jobs outside of manufacturing that pay substantially lower wages than their previous jobs. "More than 40 percent of the displaced workers formerly employed in durables manufacturing and more than 30 percent of those previously employed in nondurables manufacturing found jobs in wholesale and retail trade or services, according to data from the 1984 survey of displaced workers summarized in Podgursky (1987).
106 TECHNOLOG Y AND EMPLO YMENT This evidence suggests that recent structural change in the U.S. economy that is, changes in the employment shares of different sec- tors has contributed to lower earnings growth. Domestic technological change, however, is not the primary factor affecting the displacement of manufacturing workers. In addition, the role of technological change in supporting productivity growth and competitiveness in many U.S. man- ufacturing industries means that new technologies may aid in the stabili- zation, rather than any erosion, of high-wage manufacturing employment. We must also distinguish the impact on average earnings of movements of workers among different industries from the impacts of changes in the occupational structure of the economy. The potential reductions in average U.S. wages caused by employment growth in lower-wage indus- tries have thus far been largely offset by growth in employment in higher-wage occupations. Nevertheless, the wage reductions associated with movement of displaced workers from manufacturing (especially those from durables manufacturing) to nonmanufacturing employment contribute to the high social and individual costs of displacement. TECHNOLOGICAL CHANGE AND THE DISTRIBUTION OF EARNINGS AND INCOME Trends in the distribution of income and earnings within the United States recently have received considerable attention (Blackburn and Bloom, 1985, 1986, 1987; Bluestone and Harrison, 1986; Harrison et al., 1986; Henle and Ryscavage, 1980; Kuttner, 1983; Lawrence, 1984; Levy, 1987; Levy and Michael, 1983, 1985; Medoff, 1984; Rosenthal, 1985~. Some analysts have attributed increased inequality in the distri- bution of income and wages to the growth of service sector employment and the development of "two-tiered" occupational structures within high-technology and service industries, both of which are widely per- ceived to result from technological change (Industrial Union Depart- ment, 19841. This section briefly reviews the evidence concerning changes in household incomes and earnings~3 distributions and dis- cusses explanations for these distributional shifts. It is important to note at the outset that the distribution of income may shift with no corre- sponding change in the distribution of earnings as a result of changes in household structure. Moreover, the distribution of annual earnings also '~Household income is defined as all income received by a household, which in turn is defined to be a housing unit occupied by related or unrelated individuals. Yearnings are typically defined as employment-related wages and salaries, commissions, and tips received by an individual. Both hourly and annual measures of earnings have been used in analyses of the distribution of earnings.
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 107 may change with no corresponding shift in hourly earnings as a result of changes in the shares of full- and part-time employment within the work force. The Distribution of Earnings and Income Nearly all of the analyses to date (most of which examine data from the CPS compiled by BLS) agree that household income inequality- however it is measured has increased during the past two decades. This tendency reverses a previous trend of increasing equality, which appears to have peaked in the late 1960s. The current level of inequality in the U.S. household income distribution is slightly less pronounced than in 1947 (Levy and Michael, 19831. Researchers have used several measures of household income to reach these conclusions. Blackburn and Bloom (1987) analyzed changes in the distribution of total household income (including income derived from sources not related to the occupations of household members-e.g., interest income) and the distribution of household earnings during 1967-1984.~4 Analyzing changes in the number of households in each quintile of this distribution, Blackburn and Bloom found that the distri- butions of both household income and household earnings exhibited increasing inequality during this period. In the case of total household income, increased inequality reflected increases in the number of house- holds classified as "upper middle" or "upper" class, at the expense of middle class households. The distribution of household earnings dis- played a similar trend while also exhibiting growth in the number of households whose total earnings placed them in the "lower" class. Levy and Michael (1983) adjusted household income for taxes paid and food stamps received and found increasing inequality in the distribution of this category of household income. Thurow (1987) used data from the U.S. Bureau of the Census to trace declines during 1969-1985 in the share of total income received by the poorest 60 percent of the population. During this period, the income share of the top 20 percent of all U.S. families increased. '4Blackburn and Bloom divided the distribution into five categories, or quintiles. "Lower class" households received incomes of less than or equal to 60 percent of the median income; "lower middle class" households received incomes greater than 60 percent but less than or equal to 100 percent of the median income. "Middle class" households received incomes greater than 100 percent but less than or equal to 160 percent of the median income; "upper middle class" households received incomes greater than 160 percent of the median income but less than or equal to 225 percent of the median; and "upper class" households received incomes greater than 225 percent of the median income.
108 TECHNOLOG Y AND EMPLO YMENT The evidence on trends toward polarization in the distribution of earnings in the U.S. economy is much weaker than it is for the distribu- tion of income during the past two decades. Blackburn and Bloom (1986) found "no evidence of any trend in the dispersion [i.e., polarization] of annual individual earnings over time" (p. 7) during 1967-1983, based on analyses of data for the "principal earners" in each household. Kosters and Ross (1987) found that the dispersion of average hourly earnings (a measure of the inequality of the distribution of earnings i.e., greater dispersion implies a more unequal distribution) for full-time workers has declined since 1973. Other researchers, however, have detected some tendency toward increased inequality in the earnings distribution. Henle and Ryscavage (1980) detected a modest tendency toward increased inequality in the annual earnings of all workers during 1958-1977, although this polarizing trend peaked during 1968-1973, well before technological or other forms of employment displacement were a major concern.~5 Moreover, Henle and Ryscavage found growing earnings inequality only among part-time workers, a group that largely excludes the "principal earners" analyzed by Blackburn and Bloom. (Interestingly, the distribution of the earnings of women does not exhibit increasing inequality during or after this period.) Blackburn and Bloom (1987), however, suggest that during 1967-1984, the inequality of the annual earnings distribution may have increased among both full- and part-time male workers (although the inequality of the earnings distribution among women declined during this period). Tilly et al. (1987) also found increases in the inequality of annual earnings among male workers during 1979-1984, although their results are influenced by declines in labor force participation among older male workers and by the poor performance of the economy during this period. As in the case of Bluestone and Harrison (1986), the magnitude of the increases in earnings in equality found by Tilly et al. (1987) therefore may be sensitive to the choice of years for analysis. Without additional data and analysis, it is difficult to determine the significance or durability of any trend toward greater inequality in the earnings distribution. Explaining the Trends The trends in the distribution of household income reflect changes in the structure of the American family and increased participation by '5The results of the Henie-Ryscavage analysis conflict with the findings of Harrison et al. (1986), who detected no trend toward increased earnings inequality prior to the late 1970s. Harrison et al. found increasing inequality ~ in the earnings distribution beginning in the late 1970s.
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 109 women in the labor force, all of which affect household income rather than individual wages. The number of single-parent, low-income house- holds has grown since the 1970s, fattening the lower "tail" of the family income distribution. The number of two-earner households also has grown, which has led to some expansion in the upper reaches of the income distribution. According to Levy and Michael (1983), changes in federal tax and income support policies during the early 1980s also increased household income inequality. Reductions in transfer payment programs benefiting low-income households, combined with large tax reductions in the uppermost income brackets, increased the polarization of the distribution of after-tax income. What factors might account for any observed increases in the disper- sion of earnings? Among the least likely causes are movements of workers from manufacturing into service sector occupations or a change in the structure of the work force within high-technology industry. The current share of manufacturing within total U.S. employment is suffi- ciently small, and the size of the middle-income share of the manufactur- ing work force sufficiently resembles that of nonmanufacturing, that movements of labor out of manufacturing have little effect on recent earnings trends (Lawrence, 1984~.'7 Moreover, the characteristic form of structural change within this economy does not involve a large net outflow of labor from manufacturing into nonmanufacturing employ- ment; rather, it reflects more rapid employment growth in the nonmanufacturing sector than in manufacturing industry. During the past seven decades, employment has been growing in industries in which average wages currently are lower than in manufacturing. At the same time, however, the occupational structure of the U.S. economy has shifted in an opposite direction, with faster growth in higher-skill, higher-wage occupations (Leon, 1982; Singelmann and Tienda, 19851. Partly for this reason the gap in average wages between manufacturing and rapidly growing nonmanufacturing sectors such as business services (which include computer services and consulting) has been shrinking during the past decade (Howe, 19861. Many declining manufacturing industries for example, textiles, apparel, and leather products now pay wages that are low in comparison with those paid by much of the services sector. There is little evidence to suggest that newer high-technology manu- facturing industries have occupational structures that support increases in the inequality of the earnings distribution. Data from the 1980 census '7Blackburn and Bloom (1987) concluded that the impact on the earnings distribution of shifts in the sectoral distribution of employment during 1967-1984 was small.
110 TECHNOLOG Y AND EMPLO YMENT (Lawrence, 1984) suggest that high-technology industries "have smaller shares of lower-class jobs than manufacturing in general has, and almost all of them have larger shares of middle-class jobs" (p. 41. Assertions that technological change will produce a "two-tiered" work force, reducing the skills (and earnings) of a large share of jobs while producing a much smaller number of highly paid jobs for scientists, managers, and engineers thus receive little support from either this evidence on occupational structure in high-technology industries or the evidence on skill requirements discussed above. Demographic trends and slow economic growth, rather than techno- logical change, appear to be the primary causes of any tendency toward earnings polarization. Dooley and Gottschalk (1984) found that in- creases in the inequality of the distribution of weekly and annual earnings for males were attributable to the "baby boom and bust" since World War II, which brought large numbers of workers into the labor force during the late 1970s and early 1980s. This surge in the labor supply, along with low productivity growth and continued growth in the real wages of established workers covered by cost-of-living clauses in labor contracts, resulted in entry-level wages that were lower and that grew more slowly than the wages of older workers. Slow growth in earnings, which contributes to such increases in earnings inequality as are detectable, appears to be concentrated among workers between the ages of 25 and 34 (Lawrence, 1984~. Because of low productivity growth, these new entrants have not experienced the rapid increases in earnings that had characterized previous cohorts. The lower end of the earnings distribution thus appears to have expanded as a result of demographic and productivity trends rather than because of technological or structural changes (Levy and Michael, 1985~. BLS projections of future employment growth do not suggest that the jobs of the future will produce additional polarization in earnings. Indeed, Rosenthal (1985) maintains that these projections "show an increasing proportion of employment in higher than average earnings occupations and a declining proportion in occupations with lower than average earnings, rather than a trend toward bipolarization" (p. 61. Completion of the absorption of the large baby boom cohort into the labor force should reduce earnings inequality somewhat, whereas the expansion of income transfer and entitlement programs could offset trends toward increased inequality in the distribution of income. A resumption of productivity growth also appears to be a major component of the solution to the problem of earnings dispersion; because technolog- ical change supports such growth, it may help to reverse any trends toward the polarization of earnings in the U.S. economy. The evidence suggests that reports of a vanishing middle class due to
STUDIES OF THE IMPACT OF TECHNOLOGICAL CHANGE 111 technological change are exaggerated. The existing and very disturbing tendencies toward increased inequality in the distribution of income reflect changes in government policy, family structure, and labor force participation rather than the effects of technological change. Earnings, rather than household incomes, is the variable that should be most responsive to technologically induced changes in employment opportuni- ties; but the hypothesis that the distribution of earnings has become more unequal receives limited support from the data. The data on the earnings distribution (e.g., Henle and Ryscavage, 1980) also suggest that much of the growth in earnings inequality predates the recent period of concern over technological and structural change in the U.S. economy. Trends in the distribution of earnings also appear to be influenced more by demo- graphic than by technological factors, as well as by slow growth in productivity and in the overall economy. There is certainly cause for concern in the apparent inability of the young workers of the 1980s to experience the earnings growth and employment expansion that was the lot of their predecessors in the l950s and 1960s. To ascribe this circumstance solely to the effects of technological change, however, would be incorrect. The panel believes that the answer to this problem is to be found in policies that will support a resumption of productivity and output growth at the levels of the l950s and 1960s. We further believe that the increased use of new technolo- gies, in conjunction with policies to facilitate adjustment to them, is indispensable to the achievement of this goal.