Bryan Caplan  

Owen, Sawhill, and "the" Return to Education

Dwight Lee on Ebenezer Scrooge... Happy Hayek Day: Is it Investm...
Yesterday I attended a roundtable discussion on the new Brookings brief, "Should Everyone Go to College?"  The authors, Stephanie Owen and Isabel Sawhill, succinctly document a point that labor economists routinely neglect: there ain't no such thing as "the" return to education.  This doesn't merely mean that some people earn a bigger than average return, while others earn a smaller than average return.  The key point is that education's private payoff predictably varies.

Some of the predictors are common sense: More selective schools and more vocational majors pay more.  Other predictors are a little more surprising: Adjusting for selectivity, public colleges have a higher return than private colleges because private tuition is so much higher.  Low-ranked private colleges offer the worst of both worlds - a low education premium with high tuition:


I'm especially pleased by the fact that Owen and Sawhill take a hard-headed look at drop-out rates.  I've long argued that calculating your return to education while ignoring drop-outs is like calculating your return on bonds ignoring defaults.  In both cases, even a small probability of losing your shirt can drastically slash your expected return.  And for most colleges, the drop-out rate is enormous.  How many junk bonds have default rates as awful as those for "competitive" colleges?

Given the power of the sheepskin effect, moreover, you can't just cavalierly say, "Students who finish half the program still get half the benefit at half the cost."  No, they don't.  Drop-outs get less than half the job market benefit of the degree - and unless they perfectly time their premature exit, they pay more more than half the cost of the degree in tuition and foregone wages.

My main complaint about Owen and Sawhill is that (except for brief asides on drop-out rates), they ignore ability bias.  Part of the reason why selective schools and vocational majors appear so lucrative is that their students are exceptionally smart, hard-working, and conformist before they even arrive on campus.  When researchers adjust for ability, they find that returns to education are markedly more uniform than they appear on the surface. 

Defenders of education might be tempted to rejoice, but they should not.  When you adjust for ability, you also find that returns to education are markedly lower than they appear on the surface.  Ignoring preexisting ability doesn't just lead us to overestimate the earnings gap between engineers and liberal arts majors; it also leads us to overestimate the earnings gap between between engineers, liberal arts majors, and people who skipped college entirely.

Comments and Sharing

COMMENTS (10 to date)
Glen S. McGhee writes:

"I've long argued that calculating your return to education while ignoring drop-outs is like calculating your return on bonds ignoring defaults."

I like this way of putting it -- but could you expand this a bit to include a "rough job market" too? Thus: "calculating your return to education while ignoring [rough job markets] is like calculating your return on bonds ignoring

This externality (am I using the word correctly?) impacts all the predictors, because not all programs at selective/non-selective correlate with the average ROE for that school. Do we know the standard deviation?

Mike W writes:

"I've long argued that calculating your return to education while ignoring drop-outs is like calculating your return on bonds ignoring defaults."

"Your" who? The taxpayers? Doesn't really apply to private schools. The individual (as in the comparison to "your return on bonds")? Certainly dropping out would lower an individual's return but, students don't generally plan to dropout when they make the decision to go to college. Individual schools? Maybe, if the data for the denominator can be determined (the administration's College Scorecard is attempting to do this) but those "low-ranked private colleges" will not voluntarily provide that information.

Noah Yetter writes:

This post is solid gold. Kudos.

Glen S. McGhee writes:

Does the model account for this? How is this predicted?

Notice evidence of credential stratification at the upper end, post-grad degrees. This is the only exception. Everyone else is headed down-hill.

MingoV writes:
... calculating your return to education while ignoring drop-outs is like calculating your return on bonds ignoring defaults...

The analogy fails except for below average students.

The mean high school drop-out rate in the USA is approximately 25%. However, the drop-out rate for those who were average to above average 8th graders is only a few percent. Similarly, college students who graduated in the top third of their high school class have low drop-out rates.

For below average 8th grade students, the high school drop-out rate is around 50%. For below average high school students who attend college, the drop-out rate exceeds 60%. This group should strongly consider careers that don't require a college degree.

Brad D writes:

The authors also suggest that one social benefit to college is increased political participation. I hardly see this as a social benefit.

People who desire to become politically active are generally those who lust for power; those who clamor for more federal intervention (home and abroad); and those who seek rents.

Glen S. McGhee writes:

Mingo, How does the analogy fail? What is your example supposed to show?

It's probably a mistake to look at degrees -- like looking down the wrong end of a telescope -- instead of occupations/earnings.

This is from the Brookings report cited (page 4 and 5):

//When we dig even deeper, we see that just as not all college degrees are equal, neither are all high school diplomas. Anthony Carnevale and his colleagues at the Georgetown Center on Education and the Workforce use similar methodology to the Census calculations but disaggregate even further, estimating median lifetime earnings for all education levels by occupation. They find that 14 percent of people with a high school diploma make at least as much as those with a bachelor’s degree, and 17 percent of people with a bachelor’s degree make more than those with a professional degree. The authors argue that much of this finding is explained by occupation. In every occupation category, more educated workers earn more.

But, for example, someone working in a STEM job with only a high school diploma can expect to make more over a lifetime than someone with a bachelor’s degree working in education, community service and arts, sales and office work, health support, blue collar jobs, or personal services.//

All this is perfectly true. Yet there is no economic theory (at least that I know of) for this, because it is sociological in nature: the process of searching for jobs, non-search paths to jobs, job availability, hiring practices, transportation, location of jobs. These are the things that we need to be looking at, not degree data, simply because it is available, and job success data not so much.

Jack P. writes:

It would be nice to see some back-of-the-envelope estimates of the return to education, adjusting by field, dropout probability, and endogeneity (ability bias).

Eyeballing the figures, say ability bias adjustments lead to a (causal) return to education that is only about 6 percent regardless of where you go (ability bias nullifying the effect of going to a better university).

So 6 percent, times the probability of graduating, gives you between 2 percent and 5.5 percent, but again the graduation rates are affected by ability bias.

So the true return would be between 2 percent and 5.5 percent. Not exactly blue-chip.

Anon 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.]

Stephanie Owen writes:

Readers of this blog might be interested in the follow-up piece we posted today, which addresses ability bias and some other responses we've received about the original policy brief:

Comments for this entry have been closed
Return to top