Bryan Caplan  

Higher Education Bubble?: Getting to Bet

Of Bunk and Economics... Another Education Bet?...
David Henderson nobly expresses interest in my proposed bet:
I might take that bet. But I don't believe on betting on a change in a number without having a good idea of what the number is now. Please tell me the number and link to your source. That way we can agree on the source. Then I'll tell you if I'll take the bet.
More than fair, David. 

I propose that we use the official numbers from the National Center for Education Statistics' Table 212.  2009 is the latest available year of data.  29.6% of 18-24 year-olds were enrolled in 4-year institutions.  I bet that in 2019, that percent will be no more than 10% lower.  Rounding in your favor, I win if the number is 26.7% or more.  If the number is lower, you win.  If the data series is discontinued, the bet is canceled.  Stakes: $100 at even odds.

I'm happy to tinker with the terms.  What do you say, David?

P.S. Thanks to Gordon Dahl for pointing me to the NCES website.

Update: David accepts in the comments.  Another bet I hope to lose!

Comments and Sharing

COMMENTS (16 to date)
David R. Henderson writes:

It’s a deal!

Nathan Smith writes:

I hope David wins this one. Technology needs to overhaul these high cost sectors.

GIVCO writes:

Some traditional 4-year universities (including ones you've heard of) currently grant degrees to students who've only attended their online classes. The diploma & transcripts are identical. How will the National Center for Education categorize them?

Robin Hanson writes:

I'll add another $100 on Bryan's side of the bet. Takers?

Mark Little writes:


I will not take your bet but will offer you the opposite one--that the proportion will be not less that 10% HIGHER.

I believe that your signalling theory is correct, but woefully incomplete. The demand for higher education is driven by a combination of motives, including at least:

1. Education (human capital model)
2. Quality signalling (your model)
3. Credentialing (union card model)
4. Social signalling (status seeking model)
5. Consumption 1 (learning for its own sake model)
6. Consumption 2 (4-year party model).

Motive 1 has declined in relative (not absolute) importance as the proportion of youth attending college has grown. (See Charles Murray.)

Motive 2 is very important but easy to exaggerate.

Motive 3 is even more important than 2. A 4-year degree has become the While Collar Union Card: you can't get a "good" job without it, regardless of how much you know or how good you are, and regardless of how effectively you otherwise signal.

(Note that 3 is clearly distinct from 2. Membership in the Guild--to switch analogies--is certainly a signal, but the Guild initiation fee is an entry price, for the profit of the Guild masters, and not a signal. I'm surprised Arnold doesn't emphasize this, given his general concern regarding credentialing.)

Motive 4 is also very important. As a 4-year degree has become the prerequisite for a middle class career, it has also become the marker for middle class social status.

Motives 5 and 6 are less important, but not to be overlooked.

All of these sources of demand tend to grow, along with technology growth, income growth, and the usual downward diffusion of social status markers. (Soon, as in Alice's caucus race, all must have prizes, all Wobegonean children must be above average, and, in democratic America, all must be at least middle class and so go to to college.)

So, since I've added 4 sources of demand growth to your 2, I'll bet on the high side.

Care to take that bet?

blink writes:

Bryan, what is your point-estimate for the percent enrolled statistic 10 years hence? I'm curious, for example, whether you think it is likely to go up or just not go down very much.

Noah Yetter writes:

Easy money for Bryan, no way I would take that bet. Robin has the right idea piling on Bryan's side.

Bryan Willman writes:

I agree (somewhat sadly) with Mark Little - and would join the "it will go up" side of the pool.

Bman writes:

This bet should probably reflect that the numbers in Table 212 are likely imprecise estimates subject to some level of year-to-year variation ("Data are based on sample surveys of the civilian noninstitutional population...Standard errors appear in parentheses."). To get a crude sense of this, we can consider two plots: rate by year and change in rate by year.

The former plot suggests David may be taking a bum deal by accepting the 2009 number of 29.6% as the starting point. One can clearly see an aberrant drop for 2006-2008 and a sharp increase in 2009. If one smoothes the data (not unreasonable if there are i.i.d. measurement errors which is plausible for independent annual samples), then the smoothed value for 2009 will be quite a bit smaller than 29.6% (something more like 28.9%) and that this could be a more reasonable estimate for 2009 / starting point for the bet.

The latter plot shows that yearly differences of about 0.35% are typical. Note, this is roughly in line with the reported standard errors of about 0.40%. The difference of two independent random variables with standard deviations of this magnitude has a standard deviation of about 0.50-0.57%. Thus, there should be a range around the agreed cutoff value where no money changes hands.

Given that we are considering a bet over a drop of ~3.0% and that rate under consideration has considerably slowed in growth since the mid/late-1990s, both of these effects are non-trivial and, unless they are accounted for, they could result in money changing hands even if the "true" drop is less than ~3.0%. It would be most exciting for the readership if the random component of this bet (which is present to some extent in almost any bet) could be minimized as much as possible!

See you in ten years...

rapscallion writes:

No need to rely on this particular data set. Annual CPS college enrollment data is well defined, stable, and nearly certain to still be around in 10 years.

rapscallion writes:

Actually, this table is based on CPS data anyway.

David R. Henderson writes:

Thanks. I wish I had talked to you before I accepted the bet. But a deal’s a deal.

Ari T writes:

Hundred bucks is pocket money at that income bracket. That's more like status signaling rationality than being rational.

But I guess its better than nothing. :-)

Maximum Liberty writes:

It seems like there is a possible outcome that is both consistent with Professor Caplan's theory and with him losing the bet. In the highly unlikely event that Congress seriously rolled back the availability to student loans, we woudl see exactly the kind of outcome that he predicted: people who would have been students will find some cheaper way of sending the same signal that a four-year degree does now. I'm guessing the Professor is OK with losing the bet if that were to occur.


Nathan Smith writes:

I think the fact that both of the bettors are professors might affect the informational value of the bet. David might just be hedging.

Troy Camplin writes:

What does an education bubble look like? It looks like 1968. Or Occupy Wall Street. Or Egypt:

But if 1968 is what an education bubble bursting looks like, and higher education expanded more and more and more between 1968 and now, and government continues to support higher education, including providing cheap money through low-interest student loans, then I would bet that the number of students in higher education, broadly understood, increases.

Comments for this entry have been closed
Return to top