Institute on Education and the Economy


NCEE Brief No. 7, January 1990

The Changing Occupational Structure

Thomas Bailey

Developments in markets and changes in technology have changed the content of many jobs. The activities and responsibilities of textile machine technicians, bank clerks, junior accountants, or sewing machine operators are different from what they were ten years ago. But changes in the skill level in the economy can also come about as a result of shifts in the occupational structure. Even if the jobs of accountants and sewing machine operators did not change, the overall skill level (or required educational level) would rise if the number of accountants grew while the number of operators fell.

Indeed, since skill changes within occupations are difficult to measure, much of the controversy about changing skill requirements has involved analyses and projections of changes in the occupational structure. Some analysts point to the large numbers of positions in low-skilled jobs, such as janitorial and low-level clerical occupations, that are projected to be added to the economy over the next decade. Russell Rumberger and Henry Levin (1987) used occupational projections to argue that educational requirements of future job openings would actually decline. In contrast, the Department of Labor's widely-publicized report, Workforce 2000, used the same occupational data to argue that over 50 percent of the jobs created between 1984 and 2000 would require at least some college education (Johnston and Packer, 1987). How can the same data be used to support such different conclusions?

This Brief attempts to resolve these apparent conflicts. It uses the latest occupational projections by the Bureau of Labor Statistics (BLS) to examine trends in occupations in the United States over the last 15 years and projected trends between 1988 and the year 2000. These data suggest that those occupations filled by workers with higher educational levels have grown and are projected to grow faster than occupations dominated by workers with lower educational attainments.

Data on the growth of broad occupational categories between 1976 and 1988 reveal that, as a group, occupations with above-average educational levels (as measured by the share of workers in the occupation with at least some college education) have grown two and a half times the rate of occupations with average or lower-than-average educational levels. Occupations with higher levels of education grew 51 percent; those with lower levels of education grew only 19 percent. Although the higher-level occupations still accounted for less than 40 percent of employment in 1988, more than half of the net employment growth between 1976 and 1988 took place within occupations requiring higher levels of education.

But what about trends between now and the year 2000? Proponents of the argument that skills will fall point to occupational forecasts that some occupations generally considered to require little skill will add large numbers of jobs to the economy over the next decade. Of the ten occupations that are projected to add the most jobs to the economy by the year 2000, seven generally require low skills - these include retail sales people, janitors, waiters, office clerks, nursing aides, truck drivers, and receptionists. The seven lower-skilled occupations are projected to account for about one-fifth of the total job growth between 1988 and 2000.

Although tales of the proliferation of fast-food workers and janitors have had a strong influence on public opinion, the absolute growth in particular occupations is misleading. Large occupational categories with low rates of growth can still add many jobs. And since low-skilled jobs tend to be less differentiated than higher-skilled jobs, lower-skilled jobs tend to be categorized in large groups.

The list of the fastest growing jobs is dominated by occupations characterized by middle-level skills. Paralegals, medical assistants, radiologic technologists, and data processing equipment repairers are all among the top five. But this approach can also be criticized since fast growing occupations can start from extremely low bases and therefore will contribute few actual jobs. The three fastest growing occupations will account for only about 1 percent of the total net job growth by the turn of the century.

These problems can be avoided by looking at the occupational structure as a whole rather than focusing on selected jobs. The trend toward higher skills that has characterized the last 12 years will continue, although projections suggest that the trend may be somewhat weaker. In the past 12 years, the higher-skilled occupations grew two and a half times as fast as the lower skilled. Over the next 12 years, they are expected to grow just under twice as fast.

The accompanying table extends this analysis by presenting data on the implications of these occupational projections for the distribution of education in the year 2000. These projections are based on the assumption that the educational distribution within each occupational group will not change between 1988 and 2000. For example, if 44 percent of the managers in 1988 had a college degree, it is assumed that, in 2000, 44 percent will still be college graduates. Therefore, all of the change in the distribution of education results only from the differential growth rates of the occupation. This is an assumption with a strong bias against change.

Despite this conservative bias, the analysis suggests that the new jobs that are expected to be created over the next decade will have higher educational levels than current jobs. For example, 17 percent of the current jobs are filled by workers who have not completed 12 years of school, while the share for new jobs created by 2000 will be only 13 percent. The discrepancy between current jobs and new jobs for college graduates is even larger. According to this calculation, college graduates hold 22 percent of the current jobs while 30 percent of the new jobs will go to workers who have completed four years of college.

Can this change be attributed to "credentialing" or to "over-education"? That is, is the increase of the educational level of workers due to an escalation of educational attainment in the population that is not called for by the skill needs of employment? According to this view, if college graduates are available, employers will hire them rather than a high school graduate even if the high school graduate could do the job just as well. The credentialing argument suggests that it is misleading to judge educational needs (or the demand for education) by looking at changes in the average educational levels over time since the supply of educated workers could rise simply because individuals want more education.

While this is possible, the rise in educational levels during the 1980s does not appear to reflect an exogenous rise in the supply of education. If this were true, we should observe a decline in the wage differentials among the different educational classes, yet during the 1980s, the earnings differential between college and high school graduates has risen dramatically. Moreover, the particular analysis presented here cannot be influenced by any credentialing effect (if it does exist). The analysis shows that jobs that are currently filled by workers with higher educational levels are expected to grow faster than those filled by individuals with lower levels of educational attainment. It assumes that over the next decade, employers will continue to fill each type of job with workers who have exactly the same educational level as the incumbents. Since the analysis only uses educational levels at one point (1988), it cannot be influenced by supply-induced changes in educational levels over time.

The analysis presented in the table uses only nine occupational groups. To determine whether the results would change if a more detailed classification were used, the analysis was repeated for 46 occupational categories. The results were basically the same, with only some tiny shifts. For the 46 occupational groups, the analysis showed a strong positive relationship between the educational level of the occupation and the projected rate for that occupation.1

In recent research, Henry Levin and Russell Rumberger (1987) have used an analysis of occupational projections similar to the one presented here to argue that the educational needs of the jobs that will open over the next decade may even be lower than the existing jobs. They used Bureau of Labor Statistics (BLS) occupational projections for the period 1982 to 1995 with 1980 Census data on educational levels and concluded that "the educational requirements of new jobs over this period [1982-1995] will be almost identical to those of existing jobs in 1982." Moreover, they argue that analysts should also look at "replacement jobs - those arising from turnover" as well as new jobs since this will reflect the educational requirements of job openings. If replacement jobs are included, they contend, "then the educational requirements of job openings in the future will actually decline, simply because replacement opportunities are expected to be much higher among jobs with low educational requirements."

Why does their argument differ from the analysis presented here? First, more recent data give a different picture. Second, the discussion of replacement needs is simplistic and misleading.

First, it is possible to check the accuracy of the 1982 projections using current data. In the six years following 1982, the share of the employed population with no more than a high school degree fell from 64 to 57 percent. Thus in 1988, the total employed population had a higher educational level than the people in the new jobs that were projected to be created by 1995; thus the earlier projections underestimated the shrinkage of low-level jobs. Given the rising wage differential between college and high school graduates, this growth cannot be attributed merely to credentialism.

Furthermore, the educational distribution of net new jobs projected in 1982 and those projected in 1988 contrast sharply. The earlier projections foresaw virtually no growth in the share of college graduates, while the latest numbers indicate strong relative growth for this group.

The use of the replacement data gives the impression that the emerging job "opportunities" actually require less education than the current jobs. The replacement data that these authors use are based on an analysis by the Bureau of Labor Statistics that matches data from two successive years of the Current Population Survey (Eck 1984). Many individuals are in the sample for two years and it is therefore possible to compare their activity during those years. "Separations" occur if individuals who are in one occupation during one year are no longer in that occupation the following year, either because they have transferred to another occupation or because they are not working. In the case of occupations with growing employment, replacement needs are equal to separations (total job openings are equal to replacement needs plus growth). In shrinking occupations, replacement needs are equal to separations minus net job decline. Primarily because workers in low-skilled jobs turn over rapidly, separations are much higher for lower-skilled jobs with less-educated incumbents, than for occupations dominated by college graduates. (Fifty percent of all dishwashers but only 1.4 percent of all physicians "separate" each year.)

Whether or not replacements should be included in an analysis of the characteristics of emerging jobs depends on the question being asked. Replacement data give a sense of the nature of job openings. This is of course important for an individual looking for a job. Even in shrinking occupations, jobs open up because of replacement needs. Replacement data might also have educational implications. For example, many more nurses would need to be trained if nursing careers lasted only five years instead of 35.

But if we are interested in gauging the general trajectory of the economy and the needed skills, then the replacement data are misleading. An example can make this clear: There are three occupations each with 100 incumbents at the beginning of a period; two require no education and the third requires many years of education. At the end of the period, the high-skilled job has grown by 10 percent and the low-skilled jobs have each shrunk by 10 percent. There are therefore now 110 high-skilled and only 180 low-skilled positions. This would seem to be a clear movement toward a higher-skilled economy.

But replacement data might suggest the opposite conclusion. If all 100 high-skilled workers stayed on the job, but if by the end of the period, 10 percent of the workers in each of the two low-skilled jobs were not working and if another 20 percent of each occupation simply switched to the other low-skilled occupation (for example the dishwashers got jobs as janitors), then, according to the data that Levin and Rumberger use, there would be 40 "replacement opportunities" in the low-skilled occupations, but only 10 new job opportunities in the high-skilled occupation. Thus it would appear that the new opportunities had lower educational requirements than the jobs that existed at the beginning of the period. But these openings in the low skilled positions recorded by replacement data all resulted from churning among the lower-level jobs. Not a single new worker was needed for the low-skilled occupations taken as a whole, while ten new workers were recruited for the high-skilled occupations.

This is not to say that it is not important to consider the implications of the need to replace workers who leave the labor force, but the available replacement data combine so many phenomena that they can give a false impression of the underlying changes in the job structure. Given the much greater frictional changes in the lower-level jobs, replacement data will tend to exaggerate the importance of those low-level positions. It would be important to consider changes in turnover rates and career trajectories over time (for example if the average length of nursing careers changes significantly), but the replacement data used by Levin and Rumberger give no indication of this. Thus if we are trying to understand changes in the underlying occupational structure, given current data, it makes more sense to examine changes in the distribution of occupations entering the economy than to focus on job openings generated by replacement needs.

In sum, the occupational data do not reveal an increase in low-level jobs. Nor are the educational levels of new jobs more or less the same as those for current jobs as Levin and Rumberger argue. At the very least, these data show a steady increase in those occupations that tend to employ more highly-educated workers. Moreover, the techniques used here to evaluate the trends in educational levels have a strong bias against change, based on the assumption that change can only take place among the new jobs. If this assumption were valid, then, since the BLS expects employment to grow by only 20 percent over the next 12 years, in the year 2000 about 80 percent of the jobs would have exactly the same skill requirements as they do now. Yet, the case studies suggest that there are important changes taking place within that 80 percent.

Projected Occupational Growth and Education

Percent of Total Employment in 1988 in Each Occupation
Jobs, 1988
Projected Growth, 1988–2000 (%)
Less Than HS (%)
HS Only (%)
1-3 Years College (%)
4 or More Years College (%)
Exec., Admin., & Manag.
Prof. Specialty Occs.
Tech. & Rel. Support
Marketing & Sales
Administrative Support
Service Occupations
Ag., For., & Fish.
Prec. Prodctn. and Craft
Ops., Fab., & Laborers
Net New Jobs (1988–2000)


1. The educational data provided in the Silvestri and Lukasiewcz article were not sufficiently desegregated to carry out the analysis using the 46 categories. As a result, their data were supplemented by educational data from the 1988 public use sample of the Current Population Survey.


Eck, A. (1984). New Occupational Separation Data Improve Estimates of Job Replacement Needs. Monthly Labor Review.

Johnston, W. B., & Packer, A. (1987). Workforce 2000: Work and Workers for the 21st Century. Indianapolis, IA: Hudson Institute.

Levin, H., & Rumberger, R. (1987). Educational Requirements for New Technologies: Visions, Possibilities, and Current Realities. Educational Policy, 1, 333–54.

Sivestri, G. T., & Lukasiewicz, J. M. (1989). Projections of Occupational Employment: 1988–2000. Monthly Labor Review.

The paper from which this Brief is drawn is available from the National Center on Education and Employment: Changes in the Nature and Structure of Work: Implications for Skill Requirements and Skill Formation. The paper was prepared with support from the National Center for Research in Vocational Education and the Office of Vocational and Adult Education of the U.S. Department of Education.





Institute on Education and the Economy, Teachers College, Columbia University
525 West 120th Street, Box 174, New York, NY 10027
Phone: (212) 678-3091 | Fax: (212) 678-3699 | iee@columbia.edu