economics personal-finance education twins

Are Expensive/Selective Colleges Worth It?

[ I think Bryan Caplan's book explores whether you should go to college in great detail, so this page is about whether you should go to the most selective/expensive/prestigious college. ]

The Effect of College Selectivity on Earnings

We can't just interpret the difference in average income between two school's graduates as entirely caused by the school. As Bryan Caplan discusses in depth, ability bias (i.e. selection effects) are typically a large part of these outcome differences.

Still, if we control for a lot of variables beforehand, these estimates might be more useful. In particular, I think such studies could help us place an upper bound on the effect of college selectivity on income.

Several studies have found that, after adjusting for characteristics people have before going to college, there is no correlation between college selectivity and later income, and this is true for both sexes and even at quite selective colleges Dale, S. B., & Krueger, A. B. (2002). Dale, S., & Krueger, A. B. (2011) Dale, S. B., & Krueger, A. B. (2014). Moreover, the confidence intervals tend to be fairly tight. For instance, Dale, S. B., & Krueger, A. B. (2014) finds that a 100-point increase in your college's average SAT score is associated with only a 0.5% increase mens' incomes with a standard error of 1.1%. While these studies did find positive effects for particular demographics (low-income students, hispanics, and blacks), the overall conclusion is that college selectivity has minimal impact on earnings in adulthood. That being said, all three of the studies have the same authors and draw (at least partly) on the same dataset.

That being said, Caplan calls these papers "outliers" and instead concludes.

Punch line: moving from the bottom to the top quartile raises male earnings by about 12% and female earnings by about 8%.

He takes this from College quality and wages in the United States and How robust is the evidence on the effects of college quality?, which he calls the "most impressive research". These studies are both based on the NLSY 1979 dataset, which means the subjects were in college in the late 70s and early 80s. Both measure college quality with average faculty salary, average SAT score, and freshman retention rate. Both control for IQ (ASVAB), race, age, region of birth, and other characteristics. In these 5 correlational studies, incomes were assessed during these students 30s.

They generally find that adding these controls reduces the correlation between college selectivity quality and future income by ~35% (see Tables 2, 3 and 4 from College quality and wages in the United States and Table 6 from How robust is the evidence on the effects of college quality?)

Despite these controls, the obvious weaknesses in both these studies are (1) there are undoubtedly additional confounders that can't be controlled for, biasing the estimates upwards, and (2) the measures of college quality are imperfect, biasing the estimates downwards.

Moreover, it seems intuitively likely that the college you attend has ceiling effects: a genius going to a community college will have a hard time demonstrating their abilities are well above the typical community college summa cum laude. Since most students probably go to (roughly) the most selective college that admits them, the above papers are unlikely to pick up on these effects.

It's like how even though twin studies show ~0 effect of shared environment on IQ, it doesn't therefore follow that poisoning you kids with leads won't make them dumber. The statistical conclusions are only valid within what is typical in your sample.

Speaking of twins, several studies have used twins as a partial control for ability bias, thereby controlling for both genetics and shared environment. I've found three studies:

  • Attending a school with an average SAT score 100 points higher is associated with a 5% higher chance of graduating Ova and out.
  • School selectivity does not predict higher incomes in Japan Nakamuro
  • Behrman finds more selective schools are associated with higher incomes later in life:

As I found here, it looks like the link between education and income is small abroad but large in the US.

One enlightening thing to consider is how these estimates compare to naive estimates College Salary Report. In the following table, I set Mankato State to 100 and give the estimated causal impact and then the naive population ratio difference:

College Causal Estimate Early Career Mid Career
Ann Arbor109125123
U Penn130140157

From the above it looks like ~30% of the naive estimates are due to ability bias.

In conclusion, while some correlational studies find the entire association between college selectivity and adult income is due to ability bias, others find that only ~35% of the association is. Probably a fair guess is that roughly half of the association is ability bias.


Harvard costs $47,927 more per year than UW Madison Cost of Attendance Harvard at a Glance. The risk-adjusted discount rate is around 7%, so if we use the first year out of college as our "neutral point", that implies a 4-year degree at Harvard cost $227,689 more than one at Madison in time-discounted dollars.

Let's compare the average Harvard graduate to the average UW Madison graduate. The former starts earning $17,300 more early on and this grows to $42,700 by mid career PayScale. Suppose this gap grows linearly from your 1st to 20th year working and then stays steady for another 20 years before you retire. The time-discounted value of this difference in earnings is $426,515.

If you think the entire income gap is caused by college quality, going to Harvard is a no-brainer.

If you think the "12%" conclusion above is true, then the reasoning goes something like

  1. Harvard is ranked #2 by US News; Madison is ranked #42
  2. There are ~4000 colleges in the US. Assuming a normal distribution in quality and using Laplace soothing, this suggests the z-score difference between Harvard and Madison is ~0.9. The z-score difference between the average top-quartile and bottom-quartile school is 2.3
  3. So the income gain from going to Harvard is about (0.9/2.3) * 12% = 4.7%.
  4. The average income of Harvard graduates is $76,400 early on and $147,700 mid-career, and 4.7% of those numbers are $3591 and $6942, respectively.
  5. This suggests only about ~18% of the income gap between Harvard and Madison alumni is caused by the college they went to, so the time-discounted value of going to Harvard is ~$78,352.
  6. This is much smaller than the cost of $227,689, suggesting there's no reason to go to Harvard over Madison.

Obviously, if you think the effect on earnings is ~0, the you should definitely prefer Madison.

Conversely, "Table 4" from above suggests the effect may be 2-3x as large, bringing the expected benefit up to $157k - $235k. At those levels, Harvard is roughly equal to Madison from a time-discounted dollar perspective.

Obviously, if you buy the "0 effect" studies, Harvard is a terrible deal.

I generally agree with the conventional wisdom that it makes sense to go to college if you're reasonably likely to graduate. I agree geniuses shouldn't go to community college. But I think it's more likely than not that attending Harvard instead of a good state school is a bad financial idea if you're not getting significant financial aid.


todo Marmaros

See also

Chapter 5 of Bryan Caplan's The Case Against Education:

What then is the return for graduating from a top school instead of a bottom school? Since research is mixed, Figure 5.9 builds on low, middle, and high estimates of the quality premium. The low estimate is no quality premium. The middle estimate is that top schools lead to 5% higher—and bottom schools to 5% lower—compensation.89 The high estimate is +10% for top schools and −10% for bottom schools.

Dale, S. B., & Krueger, A. B. (2002). Estimating the payoff to attending a more selective college: An application of selection on observables and unobservables. The Quarterly Journal of Economics, 117(4), 1491-1527. Dale, S., & Krueger, A. B. (2011). Estimating the return to college selectivity over the career using administrative earnings data (No. w17159). National Bureau of Economic Research. Dale, S. B., & Krueger, A. B. (2014). Estimating the effects of college characteristics over the career using administrative earnings data. Journal of human resources, 49(2), 323-358. Cost of Attendance. University of Wisconsin-Madison. Harvard at a Glance. Harvard. The Best Universities For a Bachelor’s Degree. (2020). PayScale. Marmaros, D., & Sacerdote, B. (2002). Peer and social networks in job search. European economic review, 46(4-5), 870-879. Black, D., Smith, J., & Daniel, K. (2005). College quality and wages in the United States. German Economic Review, 6(3), 415-443. Black, D. A., & Smith, J. A. (2004). How robust is the evidence on the effects of college quality? Evidence from matching. Journal of econometrics, 121(1-2), 99-124. Smith, J. (2013). Ova and out: Using twins to estimate the educational returns to attending a selective college. Economics of Education Review, 36, 166-180. Nakamuro, M., & Inui, T. (2013). The returns to college quality in Japan: Does your college choice affect your earnings? (No. 306). ESRI Discussion Paper Series. Behrman, J. R., Rosenzweig, M. R., & Taubman, P. (1996). College choice and wages: Estimates using data on female twins. The Review of Economics and Statistics, 672-685. Lindahl, L., & Regnér, H. (2005). College choice and subsequent earnings: Results using Swedish sibling data. Scandinavian Journal of Economics, 107(3), 437-457. College Salary Report: The Best Universities For a Bachelor’s Degree. (2020). payscale.