Bryan Caplan  

Educational Signaling and Statistical Discrimination

Gridlock at the Worst Time... A Libertarian One-Liner...
The tacit assumption of signaling models of education is that employers engage in statistical discrimination.  Instead of looking into each applicant's soul and finding his true marginal productivity, employers rely heavily on more-or-less accurate stereotypes about "high school drop-outs," "English majors from Harvard," "community college graduates," etc.  As long as the stereotype leads to profitable hiring decisions, there's not much reason for employers to wonder, "Are Harvard grads more productive because Harvard taught them job skills, or because people who graduate from Harvard were highly productive all along?"

The real-world importance of statistical discrimination is hard to dispute.  Even people like me who think that markets thwart most taste-based discrimination usually admit that statistical discrimination is much more resilient to market pressure.  Indeed, as I explain to my labor students:

Unlike taste-based discrimination, statistical discrimination can survive and thrive in markets.  If group differences are real, and it is costly to judge case-by-case, then people who don't discriminate lose money.

If the typical economist read my notes on discrimination, his main complaint would be that markets do less to undermine discrimination than I imagine.  Why then are most economists so quick to dismiss the signaling model of education as "implausible"?  If you can believe that employers conserve on information costs by using their stereotypes about race and gender, why can't you believe that employers conserve on information costs by using their stereotypes about education?

Comments and Sharing

COMMENTS (10 to date)
Steve writes:

There is an insurance analog: underwriting specific risks is costly, so rates are set based on statistical discrimination (plus some underwriting, presumably the optimal amount given its cost). Of course, government has banned statistical discrimination in insurance for certain statistics (race, neighborhood, and even gender in the case of ERISA-qualifying pensions).

Fazal Majid writes:
“It is difficult to get a man to understand something when his livelihood depends on not understanding it. — Upton Sinclair”
Economists dismiss signaling in the educational market because they have no upside in antagonizing rich institutions like Harvard that could be the source of cushy gigs in the future.
Hyena writes:

So... this is really only an issue for Harvard or?

Frankly, every time this discussion comes up--and I've pointed this out before--the centerpiece of it is a bouquet of highly selective, brand name schools. This makes me think that you're aware that employers do know schools which, despite not being terribly selective, are known to produce good graduates.

This country is littered with engineering, law and professional schools like that which generally serve a regional or industry-specific market.

Scott Wentland writes:

This is your strongest economic case for "education as signaling."

In your book, you should think about making this argument as your primary "go to" argument. There's a lot to draw from in the statistical discrimination literature here, and this line of reasoning will be more convincing to other economists.

Gaspard writes:

Maybe you should ask some actual employers about what drives their hiring decisions - in my field graduate recruitment uses high school GPA and their own SAT-style tests. Of course these correlate with the prestige of the university attended, but if they're not actually weighting university prestige heavily in their models, then you may just be empirically wrong.

Alex K. writes:

It seems you just argued --convincingly-- that statistical discrimination in regards to education exists.

I agree with that.

But I don't see how you deduce from that that education is _mostly_ about signaling. (Although I am not familiar enough with the details of your argument to know the qualifications and exceptions you allow for your thesis)

Lo Statuz writes:

Does any signaling model lead to any specific, testable predictions?

Milton Recht writes:

Any signaling model of education has to explain the very low comparative unemployment rate for Asian high school dropouts.

Blacks, Hispanics and Whites all show downward sloping unemployment rates as education levels increase. Asians show a relatively flat, low unemployment rate for all educational levels.

Asian high school dropouts have an unemployment rate that is 40 percent of Blacks and 60 percent of Hispanics and Whites' high school dropout unemployment rates.

See unemployment chart towards end of article on the Chronicle of Education, "Education Pays, but How Much?" by Beckie Supiano at the following web link:

Furthermore, there are two types of high school dropouts. Those that upon entering high school are expected to graduate and those that upon entering are expected to dropout.

The expected to dropout group obviously have predictive negative characteristics of a high school dropout, alcohol and drug abuse and/or behavioral problems, which make them potentially and likely poor workers.

The expected to graduate group dropouts for non-predictive reasons, such as an unexpected loss of emotional or economic parental support, would make good, cheap workers. One would think the employment market would find a way to identify this subgroup. Whether Asians are the market response to signaling, I do not know since I have not seen data on Asian high school dropout characteristics. Asians generally have the lowest dropout rate of the ethnic groups.

Also, although I cannot find a link, European dropouts who come to the US do very well finding jobs that pay well. The reason often stated for the success of European high school dropouts is that in Europe they do not dropout due to drug use, alcohol abuse, teen pregnancy or behavioral problems as is common in the US for high school dropouts.

Brian Clendinen writes:

I find it intresting Wall Street Journals recent school rankings has mostly public schools.Granted the Journal admits their method favors large institutions but I think this list is way more valid than any other college ranking out there.

I think the degree, industry, and size of a firm all determine how much signaling is used. I know that is a lot to look at but it is like pay for females. One needs to look at a lot of other mitigating factors to really understand why there is pay discrepancies. Signaling in employment practices is no different.

However, I think in your book you need to investigate how HR policies and the attempt for equality( aka diversity), legal requirements, and stopping good old boys network effects credential requirements. In large corporations, HR has a huge amount of power and dictates pay by credentials/years of experience unless you are in the top 1% thus can get a high level manager to overrule HR. I think much of it is due to discrimination practices and legal requirements. I believe large corporations will often run HR for the whole business to fulfill legal requirements per states which have the most strict labor laws.

Bryan from your previous post and based on my experiences I don’t think you understand the internal hiring politics and how it has huge implications on the rationality behind hiring. Nor due you understand how headcount planning is actually done but more importantly used and how this influences wages and hiring and layoff decisions which executives make decisions on. Manpower planning ignores most labor specialization for simplicity sake, this along with HR policies has huge implications on the hiring decisions of hiring managers. I actually think you are right about your signaling theory. I just think you only understand a very small part of why in practice the underlying structure of the job market gives rise to signaling.

For decades in finance the risk model taught and used by academia was not how professional finance personal actually thought and managed risk. Now there is some new risk models which are fairly close to how banks and wall street manage risk. I believe the same can be said for employment theory in economics. Until economics properly models how managers make decisions and model the major limitations and complexities, I think any theory including signaling will be of little use to private industry and policy makers.

ShadowMasamune writes:

Hard to believe that economists dismiss something that I and several other state university graduates live everyday they are unemployed. While signalling is not the problem; the way human resources utilizes signalling is. For example: I have relevant coursework for management consulting in emerging markets; however, someone from Harvard having no relevant coursework can emerge into a top-50 management consulting firm much easier (as I have browsed business analysts at Mckinsey & Company with no relevant degree on LinkedIn).

For your book: I'd study post-graduate entry-level opportunities at management consulting firms. I guarantee that you will find few state university graduates and several Harvard, Columbia, and NYU graduates. I have been looking for a career with such employers for 5 months; this is what I have been up against (and I have an MA).

Comments for this entry have been closed
Return to top