But doesn't your assumption depend upon elite colleges acting independently of each other in admissions, when we know historically that they colluded in an anti-competition conspiracy?
The elder Bush's Administration sued Ivy League colleges for collusion: holding an annual conference at which they decided who gets which admits. The Clinton Administration later dropped the suit, but the colleges announced they wouldn't do it anymore.
So if the admit cohort is overrepresented in the top 1% of incomes, but shows no difference in mean incomes, for what ranges of incomes are the reject cohort overrepresented relative to the admits? Are the admits overrepresented at the low end of the spectrum as well?
What interests me, in next questions is what's the explanation for how Ivy League attendance channels graduates to these futures? Do they do so by providing a higher level of education, or by a credential that opens doors? More so, if the latter, is this a credential that is exercised once, at graduation, or repeatedly over a lifetime?
The lack in better performance below the 1% mark is suggestive of the latter, since a better education would probably have value to all students. But that's only probable since there are scenarios that could pose still predict that outcome.
Within the credential based hypothesis, you might start to be able to distinguish the one-time vs. lifetime hypothesis' by looking at the intermediary steps (elite graduate schools for example). If it was a one-time credential, then you'd expect a regression to the mean where say, elite graduate law school attendees are equally likely to take prestigious post-graduate positions regardless of undergraduate experience.
Some caution is appropriate for the above example though, since a elite-graduate school education is surely composed of some credentials and some education in terms of it's future contribution, and there may be alternatives for one (non-elite graduate schools may provide an equivalent education the education without the equivalent credential). But you might be able to start diminishing the risks of that effect by looking at not just stage-to-stage transitions, but inter-stage performance differences.. e.g graduation rates. Or in the business domain, there may be performance data there (though unless it's something like a publicly traded companies performance.. the data is probably not publicly available).
There's some obvious reasons for interest here. The education hypothesis will always be interesting. If it ever was proven true it would justify investments. That said, the evidence seems to be falling more on the other side, and the investment is already very high.
The two different credential hypothesis are very important too though. A one-time credential is somewhat lamentable, since it would mean equally capable, but non-credentialed persons would be excluded, but it's also understandable considering the difficulty of evaluation. It's not possible to evaluate all individuals, and so credentials as door openers are somewhat a fact of life. They hopefully have some predictive power.
What would be more lamentable is a lifetime credential, because it's somewhat incredulous that there's anything unique done at these schools that can't be learned later in life, and that another 5 years of work after finding opening the same door the hard way vs. the credential way, shouldn't erase as a factor.
The other aspect is, that lifetime credentials offer a somewhat easier resolution. Asking those who hire and offer future opportunities to ignore a recently acquired credential as a screening step, is unrealistic. But, simple awareness of the illogical nature of placing undue confidence in a lifetime credential is more realistic, since there is always going to be more recent data, which should be more relevant.
At the extreme edge, it would be incredibly lazy to make your decision about appointing a CEO on the basis of their college diploma. If a board made that decision without thoroughly understanding the experiences in between to the point that this is less than a footnote, well, I think that's a board that should be voted out immediately.
David writes: "In the paper we also look directly at whether the waitlist is balanced on applicant characteristics. The answer is mostly yes, with two exceptions. First, waitlist admits have slightly lower test scores and academic ratings than waitlist rejects. Second, waitlisted legacies and applicants from the top 1% of family income are more likely to be admitted. That turns out not to matter very much, both because legacies do slightly worse than non-legacies and because we get the same (if anything, slightly bigger) results when we exclude legacy and high-income applicants from the analysis. All the gory details are in the paper."
But don't you also in Table 4 show HUGE differences in nonacademic credentials of waitlist admits vs not?
Great stuff here. If you have the data, it would be interesting to have some kind of alternative to Figure 12A for the distribution of earnings. Even just showing it statically age 33 would be interesting. I would imagine fatter tails for the waitlist admits than rejects would be consistent with the results you have. Since you're doing controls and stuff, not sure if you have to use quantile regression or something to get it.
But doesn't your assumption depend upon elite colleges acting independently of each other in admissions, when we know historically that they colluded in an anti-competition conspiracy?
The elder Bush's Administration sued Ivy League colleges for collusion: holding an annual conference at which they decided who gets which admits. The Clinton Administration later dropped the suit, but the colleges announced they wouldn't do it anymore.
Can we trust their non-binding declaration?
So if the admit cohort is overrepresented in the top 1% of incomes, but shows no difference in mean incomes, for what ranges of incomes are the reject cohort overrepresented relative to the admits? Are the admits overrepresented at the low end of the spectrum as well?
What interests me, in next questions is what's the explanation for how Ivy League attendance channels graduates to these futures? Do they do so by providing a higher level of education, or by a credential that opens doors? More so, if the latter, is this a credential that is exercised once, at graduation, or repeatedly over a lifetime?
The lack in better performance below the 1% mark is suggestive of the latter, since a better education would probably have value to all students. But that's only probable since there are scenarios that could pose still predict that outcome.
Within the credential based hypothesis, you might start to be able to distinguish the one-time vs. lifetime hypothesis' by looking at the intermediary steps (elite graduate schools for example). If it was a one-time credential, then you'd expect a regression to the mean where say, elite graduate law school attendees are equally likely to take prestigious post-graduate positions regardless of undergraduate experience.
Some caution is appropriate for the above example though, since a elite-graduate school education is surely composed of some credentials and some education in terms of it's future contribution, and there may be alternatives for one (non-elite graduate schools may provide an equivalent education the education without the equivalent credential). But you might be able to start diminishing the risks of that effect by looking at not just stage-to-stage transitions, but inter-stage performance differences.. e.g graduation rates. Or in the business domain, there may be performance data there (though unless it's something like a publicly traded companies performance.. the data is probably not publicly available).
There's some obvious reasons for interest here. The education hypothesis will always be interesting. If it ever was proven true it would justify investments. That said, the evidence seems to be falling more on the other side, and the investment is already very high.
The two different credential hypothesis are very important too though. A one-time credential is somewhat lamentable, since it would mean equally capable, but non-credentialed persons would be excluded, but it's also understandable considering the difficulty of evaluation. It's not possible to evaluate all individuals, and so credentials as door openers are somewhat a fact of life. They hopefully have some predictive power.
What would be more lamentable is a lifetime credential, because it's somewhat incredulous that there's anything unique done at these schools that can't be learned later in life, and that another 5 years of work after finding opening the same door the hard way vs. the credential way, shouldn't erase as a factor.
The other aspect is, that lifetime credentials offer a somewhat easier resolution. Asking those who hire and offer future opportunities to ignore a recently acquired credential as a screening step, is unrealistic. But, simple awareness of the illogical nature of placing undue confidence in a lifetime credential is more realistic, since there is always going to be more recent data, which should be more relevant.
At the extreme edge, it would be incredibly lazy to make your decision about appointing a CEO on the basis of their college diploma. If a board made that decision without thoroughly understanding the experiences in between to the point that this is less than a footnote, well, I think that's a board that should be voted out immediately.
David writes: "In the paper we also look directly at whether the waitlist is balanced on applicant characteristics. The answer is mostly yes, with two exceptions. First, waitlist admits have slightly lower test scores and academic ratings than waitlist rejects. Second, waitlisted legacies and applicants from the top 1% of family income are more likely to be admitted. That turns out not to matter very much, both because legacies do slightly worse than non-legacies and because we get the same (if anything, slightly bigger) results when we exclude legacy and high-income applicants from the analysis. All the gory details are in the paper."
But don't you also in Table 4 show HUGE differences in nonacademic credentials of waitlist admits vs not?
Great stuff here. If you have the data, it would be interesting to have some kind of alternative to Figure 12A for the distribution of earnings. Even just showing it statically age 33 would be interesting. I would imagine fatter tails for the waitlist admits than rejects would be consistent with the results you have. Since you're doing controls and stuff, not sure if you have to use quantile regression or something to get it.