Bryan Caplan  

Tyler on Signaling: Even the New and Improved Version Doesn't Cut It

Education, Assortative Mating,... The Plunder Economy...

I'm frankly puzzled by Tyler's latest attack on the signaling model of education. Not only does he merely repeat an argument that I previously answered; but he fails to tell readers about a new and improved version of his argument that he told me last week. Now I'm stuck explaining his original argument, his improved argument, and what's wrong with both. Oh well, here goes:

1. Tyler's original argument. Gary Becker said it first, but Tyler independently discovered it:

How can ?? years of schooling be needed to signal your quality, if your employer often knows your quality within months?

2. My original answer.

Sure, employers eventually figure out how productive a worker is IF they hire him. But interviewing is expensive, and so is getting rid of disappointing workers. So it still makes sense to use credentials to make interviewing and hiring decisions: You save valuable time, and reduce the chance of hiring unproductive workers.

3. Tyler's improved argument from last week's lunch. If (as Tyler claims to know from experience) employers know your true productivity after 3 months, why doesn't the typical firm enter the worker quality certification industry? You work at a reduced or negative wage to get an employer to give you a chance, the employer treats you like a normal worker, and after three months, it certifies your quality for the whole labor market. You save years of your life, and the firm shares information it would have gotten anyway.

4. My replies to Tyler's improved argument.

First, if workers knew their whole future depended on their employer's three-month review, almost everyone would try hard, so you wouldn't learn much about workers' conscientiousness (one of the main traits that I believe that education signals). As I said earlier, "But why does school have to go on for years? Simple: Even a lazy weirdo can pretend to be hard-working and conformist for a few months."

Second, it would be a lot harder for employers to discover worker quality in three months if they didn't use educational credentials to pre-screen their hires. It's relatively easy now because employers already cut off a huge left tail of the distribution of workers. In other words, the inframarginal cost of judgment worker quality may be low, but that doesn't mean that accepting more workers and rating them is cheap.

Third, hiring bad workers and seeing how well they do is likely to be very expensive. Employers could give workers fake work to reduce the cost, but at minimum new workers will waste the valuable time of experienced workers. (And before Tyler can say, "Aha! So schooling is a more efficient way to ascertain quality!" let me remind the reader about how heavily we heavily subsidize schooling).

Fourth, even if Tyler can figure out worker quality in three months, it doesn't mean the typical employer has this ability. Tyler has many unusual abilities.

Fifth, and perhaps most importantly, Tyler's idea is weird - and weirdness is one of the qualities that we go to school to prove that we don't have. Let's put it this way: Suppose you announce that your firm is going to start selling worker certification on the side. A bunch of workers show up who want to skip college. Some may be smart, creative people; but most of them are going to be slackers and weirdos. If there are enough weirdos in your pool, and enough error in your grades, merely accepting your offer will brand workers as undesirable to the outside market.

If this seems odd, let me point out a much simpler way to get around signaling that I've previously discussed. You apply, without credentials, for a job that normally requires a college degree. You credibily offer the employer a money-back guarantee. Do you think you'll get the job? It's possible; but it's also quite likely that the employer will think, "This guy isn't normal; who knows what kind of damage he'll do?" In fact, the most likely scenario is that you won't get an interview, even if you attach your money-back guarantee to your resume. Whoever reads the resumes will laugh, then put your application in the trash.

Let me end with a challenge to all the economists who dismiss the signaling model of education on theoretical grounds: Almost all economists already admit that labor markets, competitive though they are, adjust slowly in the face of unemployment. Why is it so hard to believe that labor markets suffer from another major problem, too? If you can believe that labor markets are not evolving into spot markets, why can't you believe that labor markets are stuck in a socially inefficient signaling equilibrium?

Comments and Sharing

COMMENTS (16 to date)
BGC writes:

I agree with Bryan - I didn't use to believe it, but on investigation I think he is right and education _is_ mostly signalling

See my comment on the adjascent post

I think that one reason why the signalling theory is denied, is that it cuts at the roots of the ruling class ethos - which basically hates genetic/ herditary explanations of social phenomena, such as IQ or inherited personality.

Why this should be, is a very important topic - but maybe for another time...

Grant writes:

I think its a number of different things:
-Signaling intelligence and general ability
-Signaling work ethic (this is huge, since its really impossible to test and certify work ethic in the same way MIT would)
-Signaling socio-economic status (for mating, socializing, etc)
-Signaling knowledge to a much lesser degree
-Many people like the "college experience"
-Teaches students how to work

Add to this the irrational importance most people place on universities and the huge subsidies and charities which go towards them, and I don't think you end up with an efficient educational system. What I'm wondering is, if we didn't have those subsidies, would our education system be any different? Would our attitude towards education be different?

I don't think its so much that labor markets are inefficient as it is that its just really hard to match up a good employee with a good employer. This is why placement agencies charge so much. I don't think all employers would toss out a 'weird' resume like Bryan states, since in some cases a 'trial-run' employee can be hired and fired without significant opportunity costs.

I propose a way to improve the efficiency of labor markets: An incentive-compatible mechanism whereby employers actually honestly report their experiences of a past employee. Currently, a job candidate's references have no incentive to honestly report that person's performance. They gain nothing by honesty, except for possibly offending the person if word gets back to them. If we could think of a way to get employers to honestly rate past employees, I think that could replace educational signaling to some degree.

Pinkston writes:

The literature on asymmetric information between employers with respect to worker productivity is also relevant here. The firm that hires the worker for those first few months observes a lot more about that worker's productivity than outside firms do. Since outside firms can then learn about the worker's productivity by observing the actions of the current employer, there can then be an incentive to stragically under-promote workers in order to avoid having the wages of high productivity workers bid up by the market.

Acad Ronin writes:

The typical MBA program has some 20+/- instructors' ratings on the students' transcripts, and those ratings are each based on some 1.5 to 3 months of observation. You can fool all of some of the time, and some of us...

So the MBA program gives the employer some 20 low-quality 3-month observations, rather than one high-quality observation. Furthermore, the student can show by their course selection and performance in the courses their numeracy, literacy, and presentation skills.

Bob Knaus writes:

We already have trial programs in many industries. They are called temp workers. If you like a temp and want to make him permanent, you pay the temp-to-perm fee to the agency.

My own experience with that is telling. I'm a high school dropout, but built up a small PC support company in Miami from 1981-95 which peaked at about a dozen employees. For personal reasons, I had to move to Seattle, and I was pretty much broke. So I went to work for a temp agency, doing network support.

I was called in to a firm to do a network diagram, and ended up writing a sales proposal for one of their clients. The manager was so impressed he made me a great employment offer... but did not want to pay the agency fee. I was hard enough up at the time that I actually had to go home and think about that overnight. Fortunately my sense of ethics recovered. The next day I called the manager up and told him I could not work for someone of his ethics, and called my agency rep to let him know what had gone down.

It was a bad scene all around. Normally it works better than that.

Shortly thereafter, I got a job as a management consultant on the strength of my skills, and have never regretted my lack of education since.

liberty writes:

Like Bob Knaus, I'm a high school dropout, self-taught in the software industry. Every job I applied for of the past 10 years (a lot) "required" a college degree. Most of them didn't actually care- they just wanted someone with the necessary skill. However, my current job normally cares.

So, "You apply, without credentials, for a job that normally requires a college degree. You credibily offer the employer a money-back guarantee." -- well, no. You take the job as a contractor, first, for a 10 week trial period. No money back, just a limited risk contract.

After 10 weeks, you get a full time employee position. Their idea not mine, but I would have suggested it if they hadn't offered it first.

Mike writes:

Bob Knaus writes:

We already have trial programs in many industries. They are called temp workers. If you like a temp and want to make him permanent, you pay the temp-to-perm fee to the agency.

He is onto something but misses the most obvious trial program. It's called summer internships. If a company is considering finding future corporate leadership it isn't going to be from a temp agency its going to be in an MBA program. There are always counter examples but in the main, future executive suite candidates are found in graduate business programs. Candidates for the corporate executive suite wouldn't be seeking the door to the corporation via a temp agency because really good candidates would have already been found. Summer internships are really a disguised mating dance but I'm not sure the internees realize it.

ram writes:

This has all the feel of a false dichotomy created by the need to publish papers with clear-cut mathematical models. Let me offer these two cents (but pennies should be abolished, so...):

1. Formal education reveals natural type(s) (with noise) and improves human capital. Total human capital is the output of these inputs.
2. The resulting signal is a convex combination of natural talent (one kind of type), "natural" work ethos (another kind of type), and school-acquired knowledge (factual and procedural, improving talent and work ethos). That is, one can ace school as a slacking genius, as a hard working intelligent person, or as a point between these extremes.
3. The processes of revealing type and of increasing human capital are highly correlated at the school level. It's more efficient to acquire both at the same time than just one at a time.
4. Formal education is not the only source of signals for type nor the only way to improve human capital, but is just one of a spectrum of trade-offs.
5. Within education, there are different paths and each represents a different combination of revealing natural type and acquiring human capital. (A PhD in Econ is not the same as an MBA and neither as a JD.)

6. Not all jobs are equal and different jobs will get different informational benefits from the education signal.
7. Piecing out additional info to separate type from acquired skills requires ancillary signals (tests, interviews, etc.). Thus, different jobs will required different combos of screening processes.
8. Reducing the noise in the signal is of different importance to different jobs. So different jobs will require more or less robust signals, each representing a different trade-off.
9. And let's not forget agency issues. The signal finally acquired very likely affects the prior distribution of beliefs the employer will have on type in the presence of moral hazard. The higher the hazard, the more informative the signal should be (e.g. for tenure).

A lot of this fight arises from using one word, "signal", to explain such tightly woven dimensions as natural type and acquisition of human capital to result in a "final" type. And then to see the demand side as homogenous in its need for THE signal or THE human capital parameter.

Self-promoting example: I wasn't born knowing econometrics; school taught me that. My type might have been favorable (compared to, say, playing soccer), but I wouldn't have the skill without academic training. Similarly, my capacity to work hard was in my type, but honed through the requirements of making it to a PhD.

Now I have a signal, a PhD in Econ, which says something about both my acquired skill (econometrics) and a lot about a (masochistic) work ethos and about a floor for my level of intelligence/analytical skills, whether natural or acquired.

For whom is this a valuable signal? It might be a cost-effective screening for the hiring process of some industries, perhaps for private sector consulting, but not sufficient for a tenure-track position in Harvard (but the consulting jobs require me to reveal other types of info through interviews, case studies, etc). And the PhD certainly is meaningless for a job as a sales clerk in a high-fashion boutique.

frankcross writes:

I would add one additional argument. For many jobs, the initial months are not reflective of the goal sought by employers. New lawyers, for example, may be stuck in document discovery, which has little relationship to screening for partners. Legal education may be signaling for skills that are crucial later on in the job but are not discovered by the work assigned to first year associates

Lord writes:

Some signaling, some skill acquisition. Labor markets are incredibly inefficient. I have never seen employers spend one tenth of the amount on firing that they spend on hiring. They are irrationally fearful of making the wrong decision. Hiring is treated like making a lifetime association, but firing is usually based on quarterly profits. There is dread fear of interpersonal relations, particularly firing, that causes hiring to be extraordinarily costly, to be delayed, deferred, and denied as much as can be. Then, despite all those costs, all that care, dedication, and effort, they realize was, is, and always will be only about money. Employers go to great lengths to avoid or delay hiring such as interns, temps, contractors. These affect the companies and the hires, by selection as well as by action. I would never do any of these because I know the companies that do are not ones I would ever care to work for. Companies get the workforces they desire. If they complain about how hard it is find employees, it is generally for a reason.

Glen writes:

I understand and agree with the signaling model. (Well, to be accurate, I think signaling accounts for some large fraction of the value of a degree, though less than 100%.) But I don't understand why you (Bryan) think that it's necessarily inefficient. Do you think all signaling models are inefficient? When information is hidden and costly to obtain, we need some kind of signaling. And all the reasons you give for why the employer can't learn what he needs to learn at low cost are reasons to think education may be useful as a sorting mechanism.

If your answer is solely that education is heavily subsidized, then I might agree; maybe we have an inflated signaling market. But for the most part, I think the public subsidies are meant to disrupt a pre-existing signaling mechanism. Before public subsidies, the existing universities credibly signaled high productivity. Adding lots of new universities of questionable quality injected noise into the signal, as now it's not as obvious whether any given degree actually means anything.

dearieme writes:

Only once have I checked an applicant's claimed academic credentials. He'd lied twice. He went on to become a Member of the European Parliament.

Bruce K. Britton writes:

The federal government, and most state governments, have a component of a trial program, in which new employees can be fired easily up to when 6 months have passed, but after 6 months it is almost impossible to fire. The result is that many employees behave themselves for 6 months, are retained, and then immediately go completely slack, as anyone who has worked for any government agency has probably seen. But of course governments also originally hire based on qualifications, including education, so this also shows that even education is a noisy signal.

Nathan Smith writes:

Great post, Bryan. Thanks. I would add that I think education as investment in capacity to enjoy is underrated. For a given income, someone who is smart enough to be able to enjoy a good book can get more utility than someone who can't. Also, there's the social class effect of going to college, which isn't just about income or vanity: if you want to make friends with interesting people, it helps to be able to hold your own in conversation with them. College-educated is not synonymous with interesting but the overlap may be large (or perceived as large).

You give arguments against arguments against the signalling argument; but you don't offer any positive arguments against the competing human capital account. Do you disbelieve the human capital account and prefer the signalling account, or are you just arguing against settling the question prematurely? If you disbelieve the human capital account, why? (Maybe I missed that post... but skimming the "Economics of Education" index, I don't see any titles that are obviously about that.)

The reason I see to doubt the human capital account is intuitive: it just seems obvious that what most people do on the job doesn't have much to do with what they learned in college. But there may be subtler forms of human capital. Ability to learn is an (increasingly) important trait in many jobs; college may train that, even if the particular things you learn aren't relevant. It doesn't seem at all implausible that the conscientious and normalcy you cite as important things that college signals are actually produced, at least to some extent, by the experience of attending college. (Particularly if "normal" means "like a college-educated person," what better way to become normal than to hang out with college-educated people and do what they do for a while?) Also, humility. I think a lot of people would have much more unrealistically high ideas of themselves if they hadn't been clobbered by a few tests.

Ajay writes:

I must say, for once I agree with Lord, in that labor markets are incredibly inefficient and employers get the workforce they deserve. However, the bigger problem is that the education market that feeds workers into the labor market is not just inefficient, it's broken, largely because it's primarily publicly funded. The main point that commenters are missing is that nobody is saying education is currently 100% signaling, only that it is primarily so, and that whatever skills are taught, that learning process is done incredibly badly, both in the choice of subjects and then the way they're taught. This is why people often say that they never use what they learned in college once they're in the workplace, as Nathan has mentioned.

Michael writes:

A couple of points:

I agree that temp workers are a way around the problem in line with the (improved) Cowen argument. Also internships are a good tool. In Germany apprenticeships are typically unpaid chances to prove to your boss that you are worthy -- it is really hard to get rid of employees there.

Another point is that college serves alternative purposes. It is a chance to grow up. I buy into the thesis that adolescence is now longer than it was in previous generations. As a gap between childhood and adulthood, this is simply a time to make mistakes in an artificially soft environment. Getting caught drinking in the dorms is a way to learn social limits in a way that does not jeopardize your future career. Think of all the low-income people who end up in juvenille detention centers (which is harder to recover from). I don't agree with this use of college, but it has become the accepted norm. While I think that adolescence has grown, I think it is artificially higher due to some obvious institutional failures.

Comments for this entry have been closed
Return to top