In this space last Friday, Anthony Carnevale strongly and lengthily argued  that “college is still worth it.” He implicitly criticized those, including me , who rely on U.S. Bureau of Labor Statistics (BLS) data showing that the number of college graduates exceeds the number of available jobs that require a college degree. While he says many things, he has two main points. First, “There’s just one problem with the official BLS statistics: they’re wrong.” Second, he notes that “the most persuasive evidence that the BLS numbers are wrong are earnings data which show employers across the country pay a ‘wage premium’ for college graduates….”
I will argue that the BLS data are, in fact, pretty good, and that while Carnevale is factually correct about the earnings data, his interpretation of it is, at the minimum, misleading. Moreover, I will further argue that what is involved here is a classic application of what economists over the age of 50 call “Say’s Law” (i.e., the theory suggesting that supply creates its own demand; economists under 50 are largely ignorant of it because they have no knowledge of the evolution of their own discipline, reflecting the general abandonment of thorough teaching of the history of economic thought).
Furthermore, I will argue that diplomas are a highly expensive and inefficient screening device used by employers who are afraid to test potential employee skills owing to a most unfortunate Supreme Court decision and related legislation. Finally, I will assert that Carnevale and others who argue “college has a high payoff” are comparing apples with oranges -- i.e., they are making totally inappropriate comparisons that lead to skewed conclusions.
An even-handed interpretation of the data is that college is “worth it” for some significant number of young people, but is a far more problematic investment for others. The call by President Obama, the Lumina and Gates Foundations, and many higher education advocates to rapidly and radically increase the number of college graduates is fundamentally off-base.
The BLS Data
Carnevale essentially argues that the BLS data are pretty bad, mainly because earnings data show that employers pay workers with college degrees a wage premium, which would be irrational if the education associated with a college degree were not valuable for the job in question. Indeed, “a better approach” to this question, according to Carnevale, would be “to track actual earnings of college graduates to determine the demand for postsecondary education.” Additionally, Carnevale accurately notes that there are some variations within skills required within some of the BLS occupational categories, and it is possible that for some jobs a college degree would be necessary or highly desirable, while for others it would not be.
However, Carnevale’s overall description of the BLS data and its system for categorizing education requirements is far from accurate. For instance, Carnevale blandly declares that “[t]he official BLS data assign an education level to an occupation based on the lowest level of education attainment necessary to access the occupation.” Actually, as the BLS makes quite clear on its website , it “assign[s] what [its] research suggests was the most significant source of education or training” for each occupation. (Apparently Carnevale confuses a proposed change to the BLS category system -- and one which isn’t going to be implemented  at that -- with the system currently in place.) Furthermore, the BLS also noted that its data do in some cases understate educational requirements, but in others they overstate it, suggesting that, in the aggregate, the BLS data can be viewed as reasonably sound.
Another problem with Carnevale’s critique of the BLS data is that, in reality, the BLS dataset is arguably superior to that developed by Carnevale and his colleagues. This point was made a couple of months ago by Paul E. Harrington and Andrew M. Sum  when they observed that taking Carnevale’s approach “assumes a world where no under-employment or mal-employment of college graduates exists.” On the other hand, the BLS dataset is robust enough to account for underemployment, albeit perhaps imperfectly.
Carnevale’s criticism of the BLS data is nothing, however, compared to that of Cliff Adelman of the Institute for Higher Education policy, who, in commenting on Carnevale’s article, said “the base data are bizarre,” claiming the statistics misrepresented the numbers in some occupations (he focused on solar panel installers and like occupations) by a huge magnitude. Adelman simply misread the data -- badly -- as Carnevale himself has indicated in a response.
Before turning to Carnevale’s earnings-based argument, I want to comment on his remark about “statistical outliers such as bartenders, cab drivers and janitors with B.A.s and graduate degrees.… These kinds of mismatches between degrees and low-skilled jobs…are relatively small in number and don’t matter much….”
Hogwash. The BLS tells us that for waiters and waitresses alone, there are more than one-third of a million who hold B.A. degrees or more -- not an inconsequential number. And the BLS data would indicate that, in total, about 17 million college graduates have jobs that do not require a college degree. Not only is that 11 percent or so of the total labor force, hardly a “relatively small” number, but, more relevantly, it constitutes well over 30 percent of the working college graduates in the U.S. -- a number of mammoth proportions.
Carnevale argues that a large portion of these persons are short-term in these jobs, and that they typically move on into more appropriate jobs later on. I am the first to admit the turnover rate of waiters is greater than that of physicians, but so what? If roughly one-third of college graduates in general are in jobs not requiring college-level training, that is far more than frictional unemployment -- workers temporarily taking a low paying job while awaiting more permanent employment. And that certainly is not simply a function of the recession, although that phenomenon no doubt has aggravated the problem.
The Earnings Data
Carnevale is absolutely correct that college graduates on average earn more than those with lesser formal educational certifications. And I would agree that on average college graduates have a higher productivity per worker (justifying the higher pay) than those merely possessing high school diplomas. Therefore, for many, going to college is a good personal investment decision.
But to a considerable extent, the reason college graduates have higher pay has little to do with what they learned in college per se. Suppose an employer has two applicants, who in personal interviews seemed similar in quality. The employer likely will choose the college graduate over the high school graduate because, on average, college graduates have higher levels of cognitive skills (as measured by IQ tests or similar instruments), are more likely to have relatively high levels of motivation and discipline developed before attending college, have more general knowledge about the world in which we live, etc. Hence such employees are often offered a wage premium, since the anticipated level of performance of the college graduate is perceived to be higher. The diploma serves as a screening device that allows businesses to narrow down the applicant pool quickly and almost without cost to the employer, but with a huge financial cost to the individual earning the diploma (often at least $100,000), and to society at large in the form of public subsidies.
For the past several decades, moreover, the ability of employers to find other means of certifying competence and skills has been severely circumscribed by judicial decisions and laws. In Griggs v. Duke Power (1971), the U.S. Supreme Court essentially outlawed employer testing of prospective workers where the test imparted a “disparate impact” on members of minority groups. Cautious employers have sharply reduced such testing, and are now forced to rely on other measures of competence, namely the possession of a college diploma.
This perception that college is primarily a screening device rather than the source of a true vocationally relevant curriculum is supported by a good deal of data that show college students spend relatively little time in academic studies (e.g., the Time Use Survey  data of the BLS). The most notable recent effort, utilizing detailed data from the National Survey of Student Engagement, is examined in Richard Arum and Josipa Roksa’s new book Academically Adrift: Limited Learning on College Campuses , just released by the University of Chicago Press. “How much are students actually learning in contemporary higher education? The answer for many undergraduates, we have concluded, is not much,” write Arum and Roksa.
Say’s Law and Credential Inflation
The economist Jean Baptiste Say, writing in 1803, formulated his “law of markets,” which can be roughly summarized as: “supply creates its own demand.” In the context of American higher education, colleges have supplied millions of graduates over recent decades -- more than needed to fill jobs that historically have been considered ones requiring the skills associated with a college education. Therefore, employers are flooded with applicants who possess college degrees, and given the inherently better character and intellectual traits that college graduates on average have, the employers demand a diploma for a job. The rise in the supply of diplomas created the demand for them, not the other way around.
Jobs that have not changed much over time, such as serving as a mail carrier or restaurant manager, now have large numbers of college graduates filling them, relative to the past. This is almost certainly mainly a manifestation of what might be termed “credential inflation.” To be sure, the quality of high school graduates may have declined over time somewhat owing to the mediocre state of our public schools, but it is hard to believe this is important in explaining the rise in, say, college-educated mail carriers.
Apples and Oranges: Risk-Taking in Attending College
A huge problem in any analysis such as that performed by Carnevale is that it ignores the vast number of students who enter college and do not complete a degree. While I am the first to admit there are some problems with the underlying IPEDS data used to measure dropout rates, it is probably nonetheless true that at least two out of every five persons entering college full-time fail to graduate within six years. There is a huge risk associated with enrolling in college: you might not graduate. A person considering a $100,000 investment in 1,000 shares of XYZ Corporation common stock at $100 a share is assured that he or she will have those 1,000 shares, although there are some risks associated with the shares declining in value over time. A person making a $100,000 investment in a B.A. degree is not assured that he or she will obtain the investment –i.e., actually graduate in a timely fashion.
For years, economists have written that the rate of return on college investments tends to be high -- 10 percent is an oft-cited estimate, greater than the average investor is likely to earn in alternatives such as stocks, bonds or real estate. Thus the studies have concluded that going to college typically makes sense, independent of any non-pecuniary advantages college offers. Yet these studies have failed to account for the added risk associated with it -- the probability of dropping out.
Two new studies have attempted to correct for this problem, one by Gonzalo Castex and the other by Kartik Athreya and Janice Eberly (both are available for download here ). They suggest that the reported superior rate of return on investing in college disappears when investments are adjusted for risk.
At the individual student level, it is possible to reasonably estimate the risk. A student who was at the top of her class at a top-flight suburban high school, had a composite SAT score of 1500, and plans to attend a private college with relatively low dropout rates is probably going to get a reasonable return on her investment, although even that is no certainty. By contrast, a student who is below average in his graduating class from a mediocre high school, has a combined SAT score of 850, and is considering a college with high dropout rates is very likely not to graduate even in six years, and probably will get a very low return on his college investment. That student might well do much better by going to a certificate program at a career college, learning to be a truck driver, or becoming a barber, for example.
In short, a good maxim is “different strokes for different folks.” A one-size-fits-all solution does not work as long as human beings have vastly different aptitudes, skills, motivations, etc. On balance, we are probably over-invested in higher education, not under-invested. The earnings data reflect less about human capital accumulation imparted to college graduates by their collegiate experiences than the realities of information costs associated with job searches.
With fewer public subsidies of higher education, I suspect much or all of the problem would disappear: College enrollments would reach levels consistent with the needs of our economy and the personal economic welfare of those attending.
Richard Vedder directs the Center for College Affordability and Productivity (CCAP), and is a Distinguished Professor of Economics at Ohio University. He was assisted by Jonathan Robe, Matthew Denhart, and Christopher Matgouranis.