NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Research Council (US) Center for Education. Research on Future Skill Demands: A Workshop Summary. Washington (DC): National Academies Press (US); 2008.

Cover of Research on Future Skill Demands

Research on Future Skill Demands: A Workshop Summary.

Show details

2Labor Market Trends: A Loss of Middle-Class Jobs?

To provide a framework for sessions focusing on skill demands in knowledge work and service work, two presenters provided an overview of broad trends in the labor market, which include rapid growth in high-wage and low-wage jobs.


Bureau of Labor Statistics (BLS) assistant commissioner for labor projections Dixie Sommers presented an overview of the BLS projections for the period 2004-2014 (Sommers, 2007). She explained that the BLS methodology begins with projections of labor force growth, which are combined with econometric models to project aggregate economic growth. From this, BLS derives final demand and output in major industry sectors. Next, BLS analysts translate output in each industry sector into occupational employment in that industry sector. Every two years, the BLS releases updated projections, and it regularly evaluates its projections after a ten-year projection period has ended to see how accurate they were. Sommers emphasized that the model assumes a long-run full employment economy.

Sommers outlined four broad trends in the BLS projections to 2014 (Saunders, 2005): (1) slower labor force growth than in previous decades; (2) an aging population and labor force; (3) a continuing shift of employment to service-producing industries;1 and (4) strong growth in labor productivity. Overall, the national economy is projected to create 19 million new jobs and to generate an additional 34 million job openings due to retirements and job turnover.

Sommers said that future workforce skill demands may be derived from BLS projections in alternative ways. One approach is to look at the types of jobs that will exist in the future, since different types of jobs require different skills. Another approach is to look at wages, although wages are not a perfect measure of skills. Finally, one can look at the education and training requirements of jobs.

Turning to the first approach, Sommers presented a table showing how 10 large job families (referred to as “occupational clusters”) are projected to grow or decline from 2004 to 2014 (Table 2-1). Among these 10 clusters, the two largest in 2004—professional and related occupations and service occupations—accounted for 19.6 percent and 19.0 percent, respectively, of all jobs.2 These two clusters are projected to grow the fastest among the eight clusters over the decade, accounting for 21.0 percent and 20 percent of all jobs, respectively, in 2014. By comparison, the smallest cluster in 2004—farming, fishing, and forestry occupations—accounted for only 0.7 percent of all jobs and is projected to decline to 0.6 percent of all jobs in 2014 (Hecker, 2005).

TABLE 2-1 Bureau of Labor Statistics Employment Projections 2004-2014, by Major Occupational Group (numbers in thousands).


TABLE 2-1 Bureau of Labor Statistics Employment Projections 2004-2014, by Major Occupational Group (numbers in thousands).

The professional and related cluster and the service cluster are at opposite ends of the labor market in terms of education and wages. Most jobs in the professional and related cluster—such as health care practitioners and technicians; education, training, and library professionals; and computer and mathematical science professionals—require at least a bachelor’s degree and pay high wages. In contrast, most jobs in the service cluster—such as those in food preparation and serving and in health care support—require no more than a high school diploma and pay low wages. Sommers noted that wage trends reflect these occupational trends. The largest number of new jobs created from 2004 to 2014 will be in the top-wage quartile, and the second-largest number of new jobs will be in the bottom-wage quartile, with fewer new jobs created in the middle quartiles.

Sommers reported that BLS has tried alternative approaches to address the challenge of analyzing the education and training requirements of jobs. In the past, BLS analysts tried to identify the education and/or training usually needed to become fully qualified in an occupation. Using this approach, they found that most job openings in 2014 will require short-term or moderate-term on-the-job training, while smaller numbers will require a bachelor’s degree, a bachelor’s degree with related work experience, an associate degree or postsecondary vocational award, or a master’s degree. More recently, she said, BLS has used the education and training of people currently employed in various jobs as a measure of the education and training required. Using this method, BLS projects that in the year 2014, nearly half (47.1 percent) of all jobs will be filled primarily by individuals with no more than a high school diploma, another 28.4 percent will be filled primarily by individuals with some college education and about one-fourth (25.7) percent will be filled primarily by individuals with at least a bachelor’s degree (Hecker, 2005).


Sam Leiken (Council on Competitiveness) argued that forecasting future skill demands may not be as straightforward as Sommers’ presentation of the BLS projections suggests. He noted that rapid technological change may have profound impacts on the labor market. Leiken expressed agreement with Peter Cappelli (2003) that the labor market will adjust to alleviate any potential labor shortage. At the same time, however, Leiken asked what the nation should do while waiting for labor markets to clear. He noted that the Council on Competitiveness is grappling with several policy questions related to the labor market, including what to do to help the majority of the future workforce that is already at work. He said that many workers do not manage transitions to new jobs well, particularly those laid off in manufacturing industries who must make a transition to the service sector. State and local employment systems for displaced and employed workers are not well linked to each other or to the education system. Referring to Finegold’s presentation, Leiken said he was “not so sanguine” about the nation’s ability to make lifelong learning a matter of habit. He also asked how to help young people and adults to be wise consumers of the array of educational and career options that are available.


Labor economist David Autor (Massachusetts Institute of Technology) argued that the labor market is growing increasingly divided (Autor, Katz, and Kearney, 2006; Autor, 2007a). First, employment and wages are growing at the top and bottom of the labor market—a process he called “polarization” of the labor market. Autor said that the rise in earnings inequality over the period 1963-2005 is well known, but few people realize that, during the 1990s, employment and wages grew at the bottom as well as the top, although much more rapidly at the top.

Second, Autor argued that computerization and outsourcing of work to other nations are both likely contributors to this polarization (Autor, Levy, and Murnane, 2003b; Levy and Murnane, 2004). The share of workers using a computer at work grew from 24 percent in 1984 to 56 percent in 2003, as the cost of computing continued its historic pattern of rapid decline. Because computers excel at routine tasks with specified rules—such as providing automated flight check-in at airports and other clerical tasks—computers often substitute for humans in carrying out such tasks. However, computers are not as good as humans at two types of nonroutine tasks: abstract tasks and manual tasks. When humans carry out abstract tasks (e.g., solving novel problems, developing and testing hypotheses, managing others), they often use computers to complement their skills. Because manual tasks often require adapting to changing physical and social environments (e.g., driving a truck through traffic, serving a meal), these tasks cannot be specified with rules and carried out by computers. Autor said that computers neither complement nor substitute for human skills in carrying out manual tasks (see Table 2-2).

TABLE 2-2 Potential Impacts of Computerization on Three Major Task Categories.


TABLE 2-2 Potential Impacts of Computerization on Three Major Task Categories.

Next Autor presented evidence in support of his contention that computers have indeed reduced demand for routine tasks and jobs over the past four decades (Autor, Levy, and Murnane, 2003b). He presented a figure depicting trends in the task composition of U.S. jobs from 1960 to 2002 (Figure 2-1). Compared with 1960, jobs requiring high levels of abstract tasks have increased, jobs comprised mostly of routine tasks have decreased, and jobs including many manual tasks initially decreased but then leveled off.

FIGURE 2-1 Trends in job tasks


FIGURE 2-1 Trends in job tasks. SOURCE: Autor (2007a). Reprinted with permission.

In the future, Autor said, there will be many high-education professional and managerial jobs (involving abstract tasks) and low-education service jobs (involving manual tasks), with fewer jobs involving routine tasks and paying middle-class wages. The same pattern of rapid growth in occupations at the high and low ends of the labor market is apparent in the United Kingdom (Goos and Manning, 2007) and Germany (Dustmann, Ludsteck, and Schoenberg, 2007). Autor said that service occupations (as defined by BLS, see Appendix B) will be increasingly important in the future, because they are difficult to automate and difficult to move offshore. These occupations do not require high levels of formal schooling but involve “natural” human skills—such as locomotion, visual recognition, and spoken language—that are in plentiful supply. In conclusion, Autor predicted further development of a “barbell”-shaped economy. He said that abstract analytical and problem-solving skills will be crucial, but not everyone will have an analytical job. The future economy will be not only a “knowledge” economy, but also a service economy.


Janis Houston (Personnel Decisions Research Institutes) responded to Autor from her perspective as a corporate consultant. For the past 25 years, she said, she has worked with colleagues helping public and private organizations to analyze the competencies, skills, and abilities required for a single job or small groups of jobs, primarily for the purpose of developing selection and promotion criteria. Houston said she and her colleagues also do “job competency modeling,” in which they identify the competencies that are most critical to have, across jobs, for a company to compete in the future. She agreed with Autor that abstract problem-solving skills, cognitive adaptability, and flexibility are increasingly important. However, despite his “compelling case” for polarization, she said she continues to see demand for mid-level skills used in carrying out routine tasks (Houston, 2007).

Based on a random sample of findings from both job analysis and job competency modeling in several large firms (including American Express, IBM, Boeing, and many large telecommunications and life insurance companies),3 Houston said that she has observed growing demand for the following broad competencies:

  1. Creative problem solving.
  2. Complex communication skills, including knowing “the appropriate channels for getting things done,” negotiating, influencing without authority, and team-building skills.
  3. Adaptability (Pulakos, Arad, Donovan, and Plamondon, 2000). Houston said she has seen a “vast increase” in the importance of adaptability.
  4. Self-management. With increased remote work, self-management and complex communication skills are more important, including knowing “when it is appropriate to make a phone call instead of an e-mail.”
  5. Self-development.
  6. Systems thinking.

She has observed less demand for mathematics ability, which is in line with Autor’s conclusions about declining demand for routine tasks, including routine mathematical calculations.


Moderator Peter Cappelli (University of Pennsylvania) kicked off the discussion with a comment on Autor’s category of manual work, which, in Cappelli’s view, requires some judgment. He said that, since the time of Frederick W. Taylor (1911), managers have tried to restructure manual tasks to make them routine, asking Houston if she has observed this pattern. She replied that she has seen mid-level, routine jobs being comput erized and moved offshore, but not manual jobs. In response to another question, Autor said it is important to recognize that the three categories of work he defined as abstract, manual, and routine “are fluid.” There are powerful economic incentives for computer scientists to take tasks in the abstract or manual category and make them routine, he said (see Chapter 5 for further discussion of evolving computer capabilities). For example, automated systems for making an airline reservation using a phone tree may not provide high-quality service and may even waste some of the customer’s time, but they are very inexpensive.

In response to a question, Houston indicated that companies view the six broad competencies she listed as important for lower level employees as well as those at higher levels. Although some of the competencies are likely to be most valuable in professional jobs, adaptability includes interpersonal adaptability and environmental adaptability, which are important in manual jobs. Autor agreed, saying these are “attributes or skills that anyone would want to have,” and that the ability to solve an ill-defined problem is valuable whether one is in a managerial or customer service job.

Based in part on the workshop paper she coauthored (Gatta, Boushey, and Appelbaum, 2007b), labor economist Eileen Appelbaum (Rutgers University) questioned whether Autor assumed that, because service jobs such as child care workers receive low pay, they are unskilled workers performing manual tasks. She said that one “has only to look at France and the Netherlands to get a different picture” of child care workers as professionals engaged in abstract tasks. She suggested that, as the economy is increasingly dominated by service industries, researchers and policy makers should focus less on measuring quantities (a metaphor derived from manufacturing) and more on measuring quality of life (a metaphor more appropriate for services). She said she believes many service occupations require complex communication skills and asked Autor which jobs he includes in the service occupations category (see Appendix B).

Autor responded that, in health care, the BLS cluster of service occupations includes support occupations such as orderlies and nurse’s aides, but not highly educated doctors or nurses. He agreed with Appelbaum that society has a choice about whether service occupations should be more highly skilled and paid. He said that debates about the intrinsic value of something versus its market value are often decided on the basis of supply. Things that may be intrinsically “extremely valuable” may not be highly paid “when they are abundantly supplied.” Autor said that there are many workers who are capable of performing service work in the United States, and they receive low pay. In France, there are far fewer service workers, and they are more highly paid and highly skilled.

Labor economist Harry Holzer (Urban Institute and Georgetown University) cautioned that metaphors of an “hourglass economy” or “barbell economy” may overstate economic trends. He noted that many mid-level jobs are not disappearing—including jobs in technical support, crafts, health care support, and transportation—and will generate large numbers of new openings as the current job-holders retire. Holzer expressed concern that the popularity of the polarization metaphor is leading to a polarized education policy, focusing on college for all and standardized testing. While agreeing that he did not want to suggest that there are no mid-level jobs, Autor nevertheless argued that the current trend raises an important policy question about how to help people move from low-wage to high-wage positions if there are fewer jobs in the middle. He suggested that society would need to “find ways to ensure economic mobility is not further eroded” in the future.

Labor economist Larry Michel (Economic Policy Institute) asked whether his belief that “the jobs of the future are not going to be all that different than the jobs now” is correct. Sommers replied that some of the trends in the most recent BLS forecast to 2014 have been under way since the end of World War II. These trends include continued growth of industry sectors other than manufacturing and continued creation of new jobs due to technological change. Responding to Holzer’s concern that the barbell economy metaphor has been exaggerated, Sommers noted that the economy still has many mid-level jobs. She provided the example of the office and administrative support job cluster, which included nearly 24 million jobs in 2004 and is projected to generate large numbers of replacement openings over the following decade (see Table 2-1).

Reflecting on the session, Cappelli highlighted two points. First, the focus of the policy discussion has shifted from concern about how skills affect firm competitiveness to concerns about how skills affect individual workers’ careers and wages. Second, the adaptability of the labor market—as mentioned by Eric Wanner—makes it very difficult to forecast future skill demands. On the basis of these two observations, Cappelli offered a suggestion. He said that he often works with experts in decision science, who spend little time “fixating on point estimates, but instead focus on the robustness of the estimate or on alternative future scenarios” and suggested that these methods might be helpful to project future skill demands.



BLS classifies industries into three large groups—goods-producing (manufacturing, mining, construction), agricultural (including forestry, fishing, and hunting), and service-producing. The service-producing group is often referred to as the service sector.


A listing of the occupational groups included in these two clusters appears in Appendix B.


Houston cautioned that this sample is not truly representative of all jobs, either in the United States or in the world of work generally.

Copyright © 2008, National Academy of Sciences.
Bookshelf ID: NBK4064


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...