Future skill shortages in the U.S. economy? Sorting out the evidence
- August 31, 2011
- Policy memo by David Neumark, Hans Johnson, and Marisol Cuellar Mejia*
The impending retirement of the baby boom cohort represents the first time in the history of the United States that such a large and well-educated group of workers will exit the labor force. Recent research arrives at conflicting evidence regarding future skill shortage in the U.S. economy. Our projections suggest little likelihood of large scale skill shortages emerging by the end of this decade. But skill shortages are more likely in states with large and growing immigrant populations, and more broadly in the longer term as more of the baby boomers retire.
Among the labor force challenges posed by the impending retirement of the baby boom cohort is the potential for the emergence of skill shortages. The boomers are well-educated, having come into adulthood as the nation was rapidly expanding postsecondary educational opportunities. In earlier decades, younger workers replacing older workers were both much more educated and much more numerous. But the baby boomers are nearly as educated as current younger cohorts (Figure 1) and are large in number. Thus, their retirement will slow the growth of skill levels in the workforce, leading to shortages if skill demands continue to increase.
Carnevale et al. (2010) recently projected large shortages by the end of this decade: “By 2018, the postsecondary system will have produced 3 million fewer college graduates than demanded by the labor market” (p. 16). But Harrington and Sum (2010) criticize these projections, instead seeing over-education or “mal-employment” – college workers in jobs that do not require college degrees – as “perhaps the most pressing problem facing college graduates in the nation today …”
Our projections of skill supplies and demands for the U.S. economy stake out a middle ground. We foresee rising demand for highly-educated workers. But in the near term this rising demand will by and large be met by rising education levels among the U.S. population, suggesting little risk of a substantial workforce skills gap. At the same time, there are greater risks of skill shortages in states with large and growing, and less-educated, immigrant populations. And over the longer-term, as more baby boomers retire, there is greater risk of substantial skill shortages nationwide.
Projections of Skill Supplies and Demands through 2018
Our demand projections rest on U.S. Bureau of Labor Statistics (BLS) projections of employment growth by occupation to 2018 (Lacey and Wright, 2009). To project the education requirements of future jobs, we could also rely on the BLS, which classifies occupations by educational requirements. However, using data from the American Community Survey (ACS), we find substantial labor market returns, within occupations, to educational levels beyond those that the BLS deems “required.”** We therefore instead use empirical evidence on employment practices to estimate and project workforce skills needs, starting with the baseline education distribution of workers by occupation in 2008 and applying recent trend growth in education distributions within occupation (using ACS and Decennial Census data). Applying these estimates to the occupational projections, we obtain projected skill (education) demands.
To project supply, we construct new population forecasts that take account of nativity (unlike U.S. Census Bureau population projections), and we project population by education, and labor force participation. Three important factors underlie our projections: first, that young adults will continue to experience improvements in educational attainment compared to the preceding cohorts; second, that there will be continued upgrading of educational attainment levels of older workers; and third, that labor force participation rates will continue to rise for more highly-educated older adults, and that past patterns in retirement will prevail for the baby boom as it reaches retirement ages.
In Table 1, we compare our preferred educational attainment projections (supply) with the employment projections (demand), in levels and shares. These projections do not point to significant impending shortages of skilled workers in the United States through 2018, as the projected demand and supply shares by education are quite similar. We do see projected shortages for people with an Associate’s degree (356,000), and some excess supply of less-educated workers (those with some college or a high school degree or less). Our comparisons are based on projected total labor force supply of workers, and do not include forecasts of unemployment. If we adjust the 2018 supply projections for unemployment rates by education category as observed in 2008, then the projected shortage of workers with an Associate’s degree or higher expands to around 800,000, still far less than Carnevale et al. predict.
We have explored numerous explanations for why Carnevale et al. (2010) project much more substantial skill shortages. The difference is primarily attributable to the data they use (the Current Population Survey, or CPS). In particular, the CPS data show a higher share with college degrees at the 2008 baseline, and faster growth of these shares over time, both of which lead to considerably higher projected demand for workers with college degrees in 2018. But the CPS data appear problematic. First, the CPS data appear to overstate the share with Associate’s degrees, because the CPS equates occupational or vocational programs with college degrees, whereas the ACS data do not. The CPS data also show much faster growth rates in the share with Bachelor’s degrees or higher.
We verified that these data differences explain the differences between our demand projections and theirs. Moreover, in both data sets the shares in each education category in 2008 appear anomalous, whereas education trends through 2007 were much more similar. We therefore redid the demand-side forecasts using data from 2000-2007 (rather than 2000-2008) to estimate the within-occupation trends in education, in which case the entire difference between the projections was attributable to the different baseline educational distribution in the CPS. Finally, because Carnevale et al. use a supply projection from a completely different source, it seemed likely that using CPS on both the demand and supply sides of the market should substantially reduce the sensitivity of the projections to the definition of education, which we verified: using CPS data on both sides of the market leads to much milder projections of skill shortages than the dramatic shortages that Carnevale et al. project.
The Harrington and Sum criticism of the projections in Carnevale et al. (2010) could equally well be directed at our projections. They argue against using observed educational distributions within occupations to measure educational requirements, and instead believe that the BLS determinations of skill requirements are accurate. Although they do not develop projections, our full paper (Neumark et al., 2011) shows that if we project skill demands based on BLS skill requirements, we project massive oversupply of skilled workers. So Harrington and Sum are right that conclusions about skill shortages depend critically on how one measures skill requirements.
However, the argument that BLS skill requirements are accurate is belied by the evidence that there are substantial economic returns, within occupations, to education levels above those “required” according to the BLS. Although Harrington and Sum (forthcoming) present some evidence that appears to suggest the opposite, their evidence is based on earnings regressions that omit occupation controls, leading to spurious evidence of lower estimated returns to education for those in occupations that use less-educated workers. For example, Harrington and Sum (2010) tell the “story” of bartenders (college degree is not required) and compensation and benefits managers (college degree required). The question is not whether bartenders earn less than compensation and benefits managers, but whether the return to education within for bartenders is less than the return to education for compensation and benefits managers. We therefore conclude that the Harrington and Sum critique of using observed rather than “required” education to capture skill demands is unfounded.
Skill shortages in some states?
Although we do not project significant near-term skill shortages nationwide, the situation could differ in states with large and growing Hispanic immigrant shares in which older adults nearing retirement ages are notably better educated than young adults. States that fit this profile include California, Texas, Florida, Arizona, Colorado, New Mexico, and Nevada. The importance of these demographic changes is illustrated by a simple exercise where we project supply for the nation, but substituting California’s ethnic composition in 2018 for that of the entire United States. In other words, we ask the question, would there be a national skill shortage if the country had California’s demographic mix? The answer is yes: we project a deficit of 3.1 million workers with an Associate’s degree or higher. Of course, domestic migration could ameliorate some of these shortages.
Skill shortages in the longer term?
The longer-term perspective is also less sanguine. Our projections extend to 2018 because the BLS occupation projections end there. But the majority of boomers (two of every three) will be younger than age 65 in 2018, whereas by 2030 all of the boomers will have passed age 65. We expect that projections of the U.S. economy to 2030 would show a continuation of greater rates of growth in industries and occupations that employ highly-educated workers, consistent with the long-standing trend in the United States, and accentuated by the increased demand for health care as the baby boomers enter old age. Yet the size of the baby boom cohort coupled with its high education levels imply that the replacement of older with younger cohorts will not lead to rising education levels to anything like the extent to which it did in the past (Figure 1). It is plausible, then, that general skill shortages would be much more evident in projections extended out a couple more decades.
Carnevale, Anthony P., Nicole Smith, and Jeff Strohl. 2010. Help Wanted: Projections of Jobs and Education Requirements Through 2018. Washington, DC: Center on Education and the Workforce, Georgetown University.
Harrington, Paul E., and Andrew M. Sum. 2010. “College Labor Shortages in 2018?” New England Journal of Higher Education, November. http://www.nebhe.org/thejournal/college-labor-shortages-in-2018/.
Harrington, Paul E., and Andrew M. Sum. “Recent Projections of Labor Shortages Through 2018: From Great Recession to Labor Shortages? A Critical Look at the Evidence.” Forthcoming in Monthly Labor Review.
Lacey, T. Alan, and Benjamin Wright. 2009. “Occupational Employment Projections to 2018.” Monthly Labor Review, Vol. 132, No. 11, November, pp. 82-123.
Neumark, David, Hans P. Johnson, and Marisol Cuellar Mejia. 2011. “Future Skill Shortages in the U.S. Economy?” NBER Working Paper No. 17213.
*The research on the projections was supported by the Gates Foundation and the AARP Foundation. The views expressed are the authors’, and do not reflect the views of PPIC or the AARP or Gates Foundations.
**For additional details on this analysis, and the other analyses described below, see Neumark et al. (2011).