The Real Education Crisis: Are 35% of all College Degrees in New England Unnecessary?

By Anthony P. Carnevale, Nicole Smith and Jeff Strohl

The notion of the “college labor market” as a fixed set of occupations is remarkably static. In contrast, we assume that job and skill requirements are dynamic.

(This lively debate over future demand of college-educated workers will continue in our Forum.)

Northeastern University economists Paul E. Harrington and Andrew M. Sum argue that in our recent report Help Wanted, we “radically overstate the size of the college labor market.”  This overcount, they claim, has nothing to do with the recession. “Even in times of near full employment,” Harrington and Sum argue that “substantial shares” of college-educated workers are “overeducated,” or “malemployed.” Harrington and Sum argue that we so overstate demand for college labor that we “ignore perhaps the most pressing problem facing college graduates today: malemployment, arguing the reality is that more and more college graduates are stuck in low-wage, low-skill jobs. This assertion contradicts the best available data on the hiring and pay practices of American employers. The evidence on earnings and college degrees is unequivocal: Employers continue to demand better-educated employees, and are willing to pay more to get them.

Harrington and Sum rely on official national and state Bureau of Labor Statistics (BLS) data, which implies that that New England is producing about 35% more college degrees than are actually required for current and future jobs. If true, their empirical assessment of “substantial shares” of “malemployed” people with college degrees in New England includes about 1.1 million people. [1] However, the earnings data raise serious questions about the quality of particular state and national government education data that undergirds their analysis.

If Harrington and Sum and the national and state BLS data are correct, “overeducation” and “malemployment” are rampant in every one of the New England states:

  • Connecticut has 248,062 unnecessary degrees;
  • Maine has 79,738 unnecessary degrees;
  • Massachusetts has 531,669 unnecessary degrees [2];
  • New Hampshire has 119,705 unnecessary degrees;
  • Rhode Island has 70,904 unnecessary degrees; and
  • Vermont has 51,026 unnecessary degrees.

Harrington and Sum have a point on “malemployment.” We agree there is some mismatch between college curricula and career opportunities. As they demonstrate, there are bartenders with bachelor’s degrees even in good times. However, they take the argument about over-qualification too far. The bartenders with bachelor’s degrees (and similar stories) are a testament to our failure to connect college programs to career pathways, but they do not signal overproduction of college degrees in general.

To the contrary, since the 1980s we have been underproducing college talent, and the college wage premium is the proof. Degree production in the 1980s flattened out after baby boomers reached college graduation age, and has remained flat ever since, at slightly above 40% of the labor force. Over the past decades, employers have responded to scarcity in college talent by raising college wages relative to the wages of workers with no more than high school diplomas. Yet in spite of the growing economic advantage of college degrees, the overproduction, over-qualification or “malemployed” school of thought still has a strong following. Harrington and Sum are not alone in their view that Americans get more college than is good for them. [3]


The overproduction argument is always in the public dialogue, but gets more traction in hard times when even the most highly educated are unemployed or underemployed. The Great Recession of 2007, like recessions before it, has many people publicly wondering whether college is a safe investment. Hard times always inspire stories like bartenders with bachelor’s degrees, as well as the ever-popular cab drivers with PhD’s and janitors with advanced degrees. With many college graduates unsuccessful in finding work, the temptation to reject postsecondary education as a viable economic option grows more tempting, especially among working families where college costs are always a stretch. Since we project a continuing slow recovery through 2016, the over-qualification and “malemployment” argument will likely get even more traction.

Media stories on the value of college follow the business cycle, and the bad advice gets more pointed as the recession deepens. The prominent conservative economist Richard Vedder thinks we need only a small fraction of the college talent we now produce. Charles Murray believes the vast majority of Americans are not innately intelligent enough for real college curricula. A few months ago, The New York Times suggested “Plan B: Skip College,” while the Washington Post ran “Parents Crunch the Numbers and Wonder, Is College Still Worth It?” Even the Chronicle of Higher Education has succumbed, recently running “Here’s Your Diploma. Now Here’s Your Mop,” a story about a college graduate working as a janitor that implies a college degree may not be worthwhile in today’s economic climate.

The New York Times and other prominent newspapers were printing the same kind of stories in the early 1980s during the last severe recession. The Times ran headlines like “The Underemployed: Working for Survival Instead of Careers.” The Washington Post even ran the college graduate-to-janitor story back in 1981: “When Lyman Crump graduated with a liberal arts degree he was confident his future rested in an office somewhere. But after working a year as a file clerk, Crump, 31, took a higher-paying job as a janitor.”

These ideas of “overeducation” were popular among labor economists in the 1970s and 1980s. It was in the context of stagflation in the 1970s and early 1980s that the big think books and articles were written about over-qualification and “malemployment.”  In that era, Harvard economist Richard Freeman wrote the Overeducated American and University of Pennsylvania sociologist Ivar Berg wrote Education and Jobs: The Great Training Robbery. It was also in the 1970s  that Frederick Harbison, the seminal author mentioned in Harrington and Sum’s critique of our work, coined the term “malemployment” as he worried over the ability of the economy to employ the full talents of the baby boom in other than low-wage, low-skill jobs. The BLS chimed in as well. As late as 1984, as the “bidding war” for college talent was underway, the New York Times quoted BLS Associate Commissioner Ronald Kutscher as saying “We are going to be turning out about 200,000 to 300,000 too many college graduates a year in the 80’s. … the supply far exceeds the demand.”

A boom in college jobs and earnings

Their central premise proved embarrassing when the boom in college jobs and earnings came in the 1980s—a boom that has continued with no sign of stopping, although it has slowed in recessions and picked up again in recoveries. After the early 1980s, the surplus of baby boomer college grads quickly became a shortage and spawned the most rapid and highest college wage premium in history.

The source of error in the dire predictions in the 1970s and early 1980s was too strong a focus on demography and not enough focus on the “upskilling” that would come with the knowledge economy that was replacing the industrial economy. At 74%, the college wage premium still dominates our labor markets and is the major cause of growing income inequality. Moreover, most of the growth in college degree requirements and earnings came in occupations where college degrees weren’t deemed “necessary” in the official data—occupations like insurance agents and financial analysts.  Going forward, both the demography and economic change favor increased demand for college degrees. The baby boom that reduced the college wage premium in 1970s by surging into the workforce will be surging out of the labor force over the next two decades. This most highly educated generation includes more than 40 million workers, each with roughly 40 years of experience. The retirement of college-educated baby boomers will only increase the demand for degrees to make up for their lost educational attainment as well as their experience.

That’s why we use the actual earnings of college to determine the demand for postsecondary education in the labor market instead of relying on the BLS’s subjective and static designations of college and non-college occupations. We reason that if the wages of people with postsecondary education are high and/or rising relative to people without postsecondary education within an occupation, there is an advantage that postsecondary education confers.  People with postsecondary education in these occupations, therefore, are not overeducated, because they see a real return to their educational investment—while all degrees may not produce equal returns, in virtually all cases, that return is far greater than the cost of obtaining the degrees. [4]

The difference between Harrington and Sum and BLS and our method is that we believe that we should not define the “college labor market” as “essentially a set of occupations,” in keeping with the elite, traditional, white-collar and professional jobs. The notion of the “college labor market” as a fixed set of occupations is remarkably static. In contrast, we assume that job and skill requirements are dynamic. Technology and other economic forces are constantly updating the skill requirements in jobs.  We view a college job as any job that brings substantial, positive earnings returns to a college degree, irrespective of occupation—whether an individual is an insurance agent or a rocket scientist. In contrast, Harrington and Sum argue that the economy and employment is like a game of musical chairs where opportunity is limited by a very small and fixed set of college jobs, many fewer than the economy’s college graduations.

Most importantly, what Harrington and Sum miss by defining occupations as either college occupations or non-college occupations is the shift toward increased postsecondary requirements that occurs even within occupations that are not deemed college jobs at a given point in time.  Their conclusions don’t coincide with the consensus among labor economists—that there has been a consistent shift towards increased postsecondary requirements on the job across a growing share of occupations that previously did not require two year or four year college degrees. [5] The increasing demand for college degrees among managers, healthcare workers, and office workers are examples in the white-collar world.  The increasing degree requirements among computer and information systems workers, production workers who become degreed technicians, the growth in healthcare technicians, and increased degrees among workers in utilities and transportation are examples in the blue-collar and pink-collar worlds.

The standard explanation for those shifts within the economics literature is a concept called “skill-biased technology change.”  The core mechanism behind this is that information technology automates repetitive tasks, increasing the relative value of non-repetitive tasks in individual occupations.   The relentless engine of technological change, spurred onward by global competition, drives up skill requirements and demand for postsecondary education and training within occupations—all occupations, not just “professional, technical, managerial and high-level sales occupations.” There is no indication that the economic trend has suddenly reversed itself, and the demographic effects of baby boom retirement are clear.

Moreover, our method is careful to minimize counting statistical outliers like those ever-present bartenders, cab drivers and janitors with BA’s and graduate degrees. As we point out in our Help Wanted study, these kinds of mismatches between degrees and low-skilled jobs are relatively small in number and don’t matter much in an economy of almost 150 million jobs. In addition, we have to account for the fact that most bartenders with bachelor’s degrees will eventually move on to better-paying jobs. Many workers are just passing through low-wage, low-skill jobs as part of their natural career progression and are not indicative of career-long effects of college degrees. Over a 10-year period, each cashier job has 13 incumbents who permanently leave the occupation; among medical doctors, that replacement rate is only one. People rarely leave jobs that require college because they have the best earnings, benefits and working conditions. There are many more brain surgeons who used to be cashiers than there are cashiers who used to be brain surgeons.

In addition, these kinds of non-college jobs are greatly over-represented in the official data because so many of them are part-time.  Although low-wage, low-skill jobs make up 20% of all jobs in a single year, they only make up 14% of the hours worked in a single year. Jobs that require a BA or better make up 30% of  all jobs, but 75% of them are full-time, full-year jobs, compared with 64% of jobs that require a high school diploma or less.

The BLS method: flaws and misinterpretation

Bartenders with bachelor’s degrees aside, in the final analysis, Harrington and Sum rest their empirical case on an appeal to a higher authority above reproach: the U.S. Bureau of Labor Statistics. Harrington and Sum write:

Could BLS, the most objective, impartial and certainly data rich observer of American labor markets so objectively underestimate the demand for college graduates for such a relatively short time horizon? Our answer to this is no!”

We beg to differ.

We have high regard for BLS, and believe that the national and state level BLS occupational and employment data are unimpeachable. However, the BLS educational data is an offhand by-product of its employment and occupational data and is of substantially lower quality.

To a large extent, the poor quality of data that connects education to labor markets is a natural function of institutional silos. Labor departments at the federal and state level produce good employment, earnings and occupational data but are weak on its link with education. Education departments are strong on educational data but not its linkages with occupational and labor market data. Since no agency has responsibility for linking education and employment data, the connection is done badly and does not square with the broader economic literature that has shown skyrocketing returns to college degrees since the 1980s. That is why we set out in Help Wanted to link degrees and jobs both historically and over the near future.  Our report includes our results.

Because of the silos that separate official data on jobs form the official data on college degree production, the quality of data that links education to careers gets very little scrutiny  Every state and the vast majority of social scientists use the BLS education data uncritically. Similarly, Harrington and Sum accept the BLS data as gospel. In this regard, they are not alone. BLS’s deserved reputation on employment data gives its undeserved credibility to its static and misleading metrics on education requirements in labor markets. Very few ever look closely enough to see the huge discrepancies between the BLS and Census data on educational demand, or read the fine print, indicating that the Bureau does not claim to project educational demand. [6]

The Census data allow us to assess the BLS method. The Census Bureau actually counts college workers and their earnings in jobs. As time passes and the census data catch up with the BLS projections, we can determine if the BLS projections were accurate. To get to the punch line: The BLS projections always underpredict college demand.The BLS estimate the numbers of college degrees required and the census data report the actual numbers of college degrees employers hired; their conclusions are dramatically different. When we compare the BLS projections for 2006 and the actual count of people in the labor force with degrees in 2006, we see that the BLS undercounted the true count of postsecondary-educated workers by 17 million  in 2006, or roughly 30%, and by 22 million, or 40% in 2008. Our alternative method missed by 4%.

The bottom line is that the BLS predictions didn’t even come close to what actually happened in the economy. The only way to reconcile the BLS projections with what actually happened is to assert, as BLS, Harrington and Sum argue, that BLS is predicting the number of college degrees that employers require, not the actual numbers of college educated workers that employers hire. If this is the case then not only did employers hire these “extra” workers, in 2006 and 2008, but paid them more than 70% wage premiums for postsecondary degrees they didn’t need.   This would be cause for concern—it would mean that in 2008, 22 million workers—or more than a third of all workers with postsecondary education—got an appreciable economic benefit from their degrees that they didn’t earn. It would mean that employers were smart enough to cut back the college wage premium in the 1970s when they experienced an oversupply, courtesy of the baby boom, but the same employers started throwing money at degrees in the 1980s and continue to do so. If Harrington and Sum are correct, crisis abounds, markets don’t work, employers are irrational, and preparing your children for college is naive for all but a very select few.

We hope the dialogue over the measurement of the future demand for postsecondary-educated workers does not end here, but is carried into state agencies.  We need to know why the national and state BLS data show so much difference between what they estimate as the number of required college degrees and the actual counts of college degrees in each state. Intuitively, the difference between what the BLS says is required and the actual number of degrees is overqualification or “malemployment.” If that’s what the BLS believes, it needs to expand on why overqualified workers with college degrees make so much more than workers with high school or less. Eventually, the steady progress in most states to align education and careers will ultimately make the current flaws in our information systems moot, but in the meantime the myth of overeducation is perpetuated in national and state labor market data.

Unfortunately, the myth of overeducation misinforms policymakers looking for places to cut their budgets, and, worst of all, discourages decisions about college-going that are made at kitchen tables all across America. The sensationalist stories, the high unemployment among college graduates, and the misleading official data are unlikely to keep middle- and upper-class youth from going to college.  The real tragedy of these headlines is the message they send to less privileged youth for whom college is not an assumed path. The negative press on college fuels preexisting biases among working families that college is neither accessible nor worth the cost and effort. Moreover, the bad press and worse data strengthen the hand of elitists who argue that college should be the exclusive preserve of those born into the right race, ethnicity and bank account.

[1] Nationwide, a comparison of the BLS and Census data shows 37% or 22 million college degrees that are not required, even though employers pay much higher wages for these unnecessary degrees than they do for high school degrees.  See author’s calculations, CPS, various years.

[2] This count includes associate degrees and higher.

[3] Anyone who knows them or their work knows that Harrington and Sum are not to be associated with another popular view on overqualification that begins with the assertion that the majority of Americans are not smart enough for college. This elitist view on what’s best for other people’s children is most closely associated with Charles Murray and Richard Vedder and many more who believe we are lowering the bar by increasing access to college. To their credit, Harrington and Sum worry that college isn’t good for many students, not that the students aren’t good enough for many colleges. Our argument with Harrington and Sum is one of fact not values.

[4] Our projection method intentionally minimizes the impact of outliers—like bartenders with college degrees. For more information, please see the technical report on our website and Appendix 4 in our report.

[5] There is a deep and long literature on this subject.  It is best and most recently summarized in Claudia Goldin’s and Larry Katz’s book The Race Between Education and Technology (Harvard University Press, 2008).  The empirical essence of Goldin’s and Katz’s and Katz’s narrative is that the rising wage premium for college proves that technology is increasing the demand for college workers faster than we can produce them since the 1980s.

[6] The BLS does not claim to analyze educational demand nor do they project these estimates. The BLS data don’t project skill change at all. Instead they “assign” the most significant education and training requirements for employment in 755 particular occupations. BLS does not track skill or earnings from skill in occupations empirically.  Their educational assignment method is based on the subjective judgment of analysts in consultation with experts and 755 occupations, and requires a lot of subjective judgment and consultation. To some extent, BLS‘s limited efforts are a function of their limited goals. The fine print in the BLS data states at great length that their purpose is to represent the most significant education and training requirement in particular occupations. BLS recognizes assigning a single education level to a job does not accurately reflect what is needed on the job.  As they will tell you if asked, virtually every occupation in the economy comes with a variety of legitimate educational attainment levels.  According to BLS:

Because of the variability of job functions within a given occupation, and because different employers have many different requirements of education and training, workers in the same occupation can have substantially different education and training backgrounds. [BLS, 2009]

Anthony P. Carnevale, Nicole Smith and Jeff Strohl are economists at the Georgetown University Center on Education and the Workforce.


Leave a Reply

  • (will not be published)

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>