Bryan Caplan  

Bartender, Cashier, Cook, Janitor, Security Guard, Waiter

How Staggering is the Briggs-T... Saez You: Income Distribution...
The human capital and signaling stories can both explain the existence of malemployment.  But malemployment research still provides some of the most compelling evidence in favor of the signaling model.  The latest draft of my The Case Against Education explains why:

By itself, malemployment is compatible with the human capital model.  How?  Graduates are "malemployed" because they failed to acquire marketable job skills in school.  This could mean that malemployed graduates failed to learn and retain the curriculum; recall that on the National Assessment of Adult Literacy, over 50% of high-school grads and almost 20% of college grads have less than Intermediate literacy and numeracy.  Or it could mean that malemployed graduates learn an irrelevant curriculum; recall that over 40% of high school coursework and over 40% of college majors score "Low" in usefulness.  When a B.A. bartender asks, "Why oh why can't I get a better job?," the human capital model bluntly answers, "Because despite your credentials, you didn't learn how to do a better job."

The signaling model weaves a rather contrary tale.  Malemployment reflects an arms race in the labor market - workers' never-ending struggle to outshine each other.  Rising education automatically sparks credential inflation: The amount of education you need to convince employers to hire you automatically rises too.  In an everyone-has-a-B.A. dystopia, an aspiring janitor might need a master's in Janitorial Studies to land a job scrubbing toilets.[i]  When a B.A. bartender asks, "Why oh why can't I get a better job?," the signaling model answers, "Because too many competing workers have even more impressive credentials than you do."

If both human capital and signaling allow for malemployment, why raise the issue?  Because the two stories diverge on one crucial point: Does the labor market reward workers for education they do not use?  Human capital says no; signaling says yes.  Take bartenders with B.A.s.  On the plausible assumption that college does not transform students into better bartenders, the human capital model predicts that B.A.s will fail to raise bartenders' income.  The signaling model, in contrast, predicts the opposite: Bartenders with B.A.s will outearn bartenders without B.A.s.  Why?  Because bars, like all businesses, want intelligent, conscientious, conformist workers - and a B.A. signals these very traits.  So given a choice, bars favor applicants with B.A.s despite the irrelevance of the academic curriculum to the job.


To weigh the power of human capital versus signaling, however, we must zero in on occupations with little or no plausible connection to traditional academic curricula.  Despite many debatable cases, there are common occupations that workers clearly don't learn in school.  Almost no one goes to high school to become a bartender, cashier, cook, janitor, security guard, or waiter.  No one goes to a four-year college to prepare for such jobs.  Yet as Table 4.6 shows, the labor market comfortably rewards bartenders, cashiers, cooks, janitors, security guards, and waiters for both high school diplomas and college degrees.


H.S. Premium

College Premium













security guard






High school premium = [(median earnings for high school graduates)/(median earnings for high school drop-outs)] -1.

College premium = [(median earnings for college graduates)/(median earnings for high school graduates)] -1.

Source: Supplementary data for The College Payoff, supplied by author Stephen Rose.

None of these occupations are weird outliers.  True, most bartenders, cashiers, cooks, janitors, security guards, and waiters lack college degrees.  Yet in the modern economy, all are common jobs for college grads.  More work as cashiers (48th most common job for college grads) or waiters (50th) than mechanical engineers (51st).  More work as security guards (67th) or janitors (72nd) than network and computer systems administrators (75th).  More work as cooks (94th) and bartenders (99th) than librarians (104th).  I selected Table 4.6's occupations to minimize controversy.  Human capital purists could insist that college provides useful training for electricians, real estate agents, or secretaries.  But even the staunchest fans of human capital theory struggle to say, "College prepares the next generation of cashiers and janitors for their careers" without smirking.

[i] I borrow this example from Vedder et al, p.28: "We jokingly predict that colleges will offer a master's degree in Janitorial Studies within a decade or two and anyone seeking employment as a janitor will discover no one will hire unless proof of possession of such a degree is presented."

COMMENTS (22 to date)
Will May writes:

"Because bars, like all businesses, want intelligent, conscientious, conformist workers - and a B.A. signals these very traits."

What would lead you to believe that the premium results from the degree (signaling), rather than the traits themselves? Sounds like it would be pretty hard to disentangle the two different effects.

To get really rigorous here, you'd have to actually measure intelligence, conscientiousness, and conformism, and then compare the outcomes of psychologically equal people with and without degrees. Or supplement your data in some other way.

I don't doubt that signaling plays an important role in income premiums, but I do doubt that this table of statistics measures those signaling effects. It probably overstates them by a lot, because it doesn't account for pay premiums due to the underlying psychological characteristics (the premium that would occur even without signaling).

Will May writes:

But either way, that's not consistent with the human capital model, so the point still stands, at least partially.

(Unless managers are paying higher wages with an expectation that they'll promote these workers in the future, and the workers will use their college human capital in the future positions. Then all bets are off.)

Enial Cattesi writes:

I have a small difficulty about the table.

Should I understand that in a bar with two bartenders, one with a high-school diploma and one with a college degree, the college graduate will likely have a 62% higher income than his colleague?

What kind of bars employ college graduate and which employ high-school graduate? In other words what is the ratio high-school graduates/college graduates in a bar?

Also, colleges are not all the same. What's the income difference for a Harvard graduate compared to a GMU graduate for example? Which are the most likely to see malemployed, like working in a bar?

Himanshu Sanguri writes:

In my opinion, colleges and universities foster the graduates to join the professional world. It will not be pragmatic to try and establish a one to one mapping of degree curriculum with the work done in future jobs. Moreover, there are many vocational training institutions that provides a gamut of short term courses and trainings on specialized fields like call centers, hotel managemnt, security services, plumbing, electricians and likes. I think we should factor them in our analysis.

Glen S. McGhee writes:

I also had problems with the Table 4.6, which shows that "the labor market comfortably rewards bartenders, cashiers, cooks, janitors, security guards, and waiters for both high school diplomas and college degrees".

First, it is a table of someone else's data. We don't know anything about these numbers, so we can't make any valid inferences about what the labor market is or is not doing. (Just what is the social topology of the hiring process? Where? When? Who is getting hired, who is NOT getting hired? What are size are these businesses? Are they using HR? or just word of mouth?)

More importantly, the causal connection between degrees/diplomas and wages has to be plausible. It is not. Small businesses can hire their friends and friends of friends on the basis of weak ties (Granovetter's work on hiring), with no connection to credentials.

How do we know they even assessed credentials? The figures seem to assume that they did, and in the same way for each job candidate. How do we know this from the Table? We don't.

The route to untangling malemployment lies in understanding the complexities of the hiring process which leads to outcomes shown, if in fact they are meaningful.

Bostonian writes:

Does Caplan think college-educated bartenders earn more than those without a diploma because employers are willing to pay more for degreed bartenders or because the degreed bartenders are smarter and more skillful and show that they are more productive? It does take brains to know how to make a wide range of drinks and stay on top of trends.

Hadur writes:


I am not Caplan, nor am I an economist, but I am sympathetic to the signaling model of education. I believe the response to your question is that employers are looking for a particular set of skills in an employee - intelligence, conformity, etc. In an ideal world, the employer would analyze each job applicant for these skills, through intensive testing, lengthy interviews, etc. But in the real world, employers cannot administer this battery of tests, so they rely in part on proxies for those traits. And having a BA is a proxy for having those traits, because it takes those traits to get into a school and graduate from it.

The more prestigious your school, the more rigorous your degree program, the more of a proxy it is. Somebody who graduates from CalTech must surely be intelligent and hard-working. Soembody who graduates from Harvard must surely be intelligent, but may be lazy (it's hard to get in, easy to graduate). Somebody who graduates from GMU...

Enial Cattesi writes:


Somebody who graduates from CalTech must surely be intelligent and hard-working.

And fantasizes about working in a bar or as a cashier. You missed the point of the post: that signaling has better predictive power than human capital theory.

Peter St. Onge writes:

Next experiment is when employers don't know credentials of their bartenders or cashiers, which are often filled informally. I've run bars and restaurants and I didn't know education of most employees. I hired on personality, competence and, unintentionally, connections via recommendations.

If the gap sustains, which I guess it would, it would be a strike against signalling but would argue for human capital formed pre-college.

Peter St. Onge writes:

I should say "formed ex-college" not "pre-college." Some of that capital may be recent vintage, such as social networks.

Hazel Meade writes:

It's the credit assignment problem (for all the reinforcement learning (all two of them) nerds out there).

People study something in college and the rewards aren't going to come until years and years in the future. So how do they decide which courses are going to provide the most future benefit? We deal with the same problem in machine learning, and it's the problem of how to assign credit for rewards or losses to past actions, when you have a long string of past actions that contributed.

In a purely private market, there would be little money available to borrow to study subjects that didn't provide marketable skills. So students would get an immediate signal from the market: "warning! This decision is likely to lead to low rewards!" This would come in the form of high interest rates and stringent terms. The lending market does the job of determining likely future rewards and assigning them backwards to the present.

Unfortunately, we have a student loan program where you can basically get money to study whatever you want. So the student has no direct financial incentive to study the hard (but rewarding) subjects. Indeed, the student may be incentivized to take Mickey Mouse courses and get straight A's because the "credentials" are easier to get than the marketable skills. The future is discounted. In the present you can choose the easy class that you can get an A in, or you can choose the hard class and maybe get a B or a C.
The student loan system creates this problem that students aren't getting effective signals about what are the bad long-term decisions.

Anonymous writes:

[Comment removed pending confirmation of email address. Email the to request restoring this comment. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

Mr. Econotarian writes:

These anecdotes from "College-educated workers are taking jobs that don't require degrees" are informative:

With the generous tips of New Yorkers and his pick of shifts, he pulls in around $80,000 a year as a bartender. It's more than he was offered for various office jobs he considered when he arrived in the city, even though he's highly educated.

"I have a master's in economics and I'm bartending in New York," he said with a shrug. "It's a good way to make money."


Saim Montakim has a bachelor's degree in accounting but drives a New York City taxicab. It's strenuous work, but he can make $200 on a good day. On a bad day, he barely can pay the rent for the taxi and the cost of gas.

He's getting a master's degree in human resource management and hoping that his dual degrees will help him find a better job. The accounting jobs he was offered upon graduation were all entry-level, paying $10 an hour. He prefers driving a taxi to being stuck in an office from 9 to 5 at that wage, he says.


Mariah Arcuri paid off all her debts before starting her job in a lab, which required a college degree.

She worked as a bartender in New York, earning about $90,000 a year. She paid for her college education, her graduate school and her wedding with savings from tending bar.

But because she wanted to spend time with her husband, Arcuri eventually stopped bartending and got back her nights and weekends. She now works in a lab and makes only about two-thirds of what she did as a bartender, despite her master's degree in biochemistry.

Bostonian writes:

Hazel Meade correctly notes that federally guaranteed student loans obscure information about which majors provide marketable skills, since the interest rate does not vary by major. The Obama administration is pushing "income-based" repayment, which will worsen this problem and even further subsidize lower-paying majors.

Arthur_500 writes:

You seem to say that college does not teach people how to do a job. this is consistent with the elitism of academia in which they claim that college is where you learn how to think.

Excuse me while I roll on the floor laughing.

But we have teachers who go through a regimentation in order to become a teacher. We have accountants that must have five years of college to become an accountant.

We had the requirement for a Masters' Degree in order to become a Licensed Practical Nurse (doctor-lite) but now we require a doctorate! What is necessary to achieve a doctorate in Nursing sufficient to become an LPN? Statistics and management classes. Hmm, how does that make you a better nurse?

It appears that in the quest for respectability we require college education for a specific trade. However, we do not require the college to actually teach anything useful to support that respectability.

I would say we have malemployment with those who have failed to learn to think and failed to take the ethereal studies of college and relate them to the real world. Is this a problem of the student, the school or both?

Glen writes:

I like Peter's experiment, but in the context of this debate, I think he is confusing "human capital" with "social capital" and/or "cultural capital" when he draws his conclusion. The former is increased by education, according to the human capital proponents.

Cultural capital is different.

Peter wrote:
"Next experiment is when employers don't know credentials of their bartenders or cashiers, which are often filled informally. I've run bars and restaurants and I didn't know education of most employees. I hired on personality, competence and, unintentionally, connections via recommendations.

If the gap sustains, which I guess it would, it would be a strike against signaling but would argue for human capital formed pre-college."

I think the rounding of the personality alluded to here is better understood as cultural capital or social capital. I always think of human capital as more technical in nature, having to do with some saleable skill or expertise (at least since Ivar Berg's work).

Roger Sweeny writes:

"In an ideal world, the employer would analyze each job applicant for these skills, through intensive testing, lengthy interviews, etc."

In America, since the 70s, it has been illegal for employers to give any test where 1) the test has a racially "disparate impact" e.g. whites on average score higher than blacks, and 2) the employer has not proved that the test is a pretty much perfect predictor of job performance. In practice, that burden of proof is just about impossible to meet, and since most any test will have some racially disparate impact, employers don't try to develop any.

One reason college diplomas are such a sought after signalling device is that many other ways of signalling are illegal.

Hadur writes:

@ Roger Sweeney:

But even if those tests were fully legal, how many businesses would have the time to do it? Signaling saves time.

JKB writes:

Seems to me that job premium chart needs to be sorted for age, years of service, # of years since first job, etc. Probably sex and race as well.

It does appear there is a signal using high school graduate for fiscal reliability given the high premiums over non-grads for those jobs that either handle money or involve trust. Or to put it more bluntly, criminal history.

It would also be of interest to know if those with college listed such on their application or were able to hide it by having a job history during college. I know many who've been told their college knocked them out of selection because the employer didn't think they'd stick around.

MingoV writes:

Budding bartenders, cooks, waiters, and security guards might benefit from 3-6 month trade school programs. If such programs were used, would the trade school premiums exceed the college premiums?

isomorphismes writes:

Brian, have you thought of any other potential explanations besides these two famous ones? Maybe you've already written a post on it.

isomorphismes writes:

Thanks for the summary on the malemployment article Brian.

Comments for this entry have been closed
Return to top