The Talent Bubble

I've argued before that the last decade has seen tuition increases and discount rate increases driven in part by a red-queen's race for the most talented students. The best applications are often seen to be those with high SAT, high high school GPA, and extras like co-curricular activities. Competition for these is fierce, and institutions with the highest endowments, or otherwise can offer the best aid packages, are in a commanding position. The Internet facilitates multiple applications, and the price (from a college's point of view) gets bid up as in an auction.

I was part of a conversation today with an enrollment professional who put the proportion of second-generation African-American applicants at 15%. For an HBCU, this means the pool of "good" applications (in the standard recruiting definition) is tiny. It becomes expensive to create attractive packages for these students. There will be pressure to sacrifice need-based aid in order to buy talent.

This is a lousy business model in the short term. With a decade-long perspective, it's attractive to have a growing pool of successful alumni, but you can bankrupt yourself in the process. It's a zero-sum game--there are only so many really good applications. But is that really true?

In my study of an admission matrix at one institution, the student enrolled at the bottom end (provisionally) succeeded about half the time. That is, there's a 50% chance that an applicant that looks lousy on paper is going to defy expectations on the upside. This isn't really surprising when you consider that predictions of first-year GPA based on grades and SAT aren't very good. This is especially true for low social-capital applications, such as first-generation students.

I've argued that we overprice high SATs at the cost of under-pricing some of the low SATs. If we could tell which low SAT students would succeed, this would be a gold mine for any institution. I actually wrote to ETS years ago and suggested that they develop and alternative instrument, but never heard back.

Here's how it would work. In addition to high school GPA (and SAT if you absolutely have to have it), find other indicators of success. Things like high school attendence records ought to be useful, but there are surely surveys that can be developed that would help. The CIRP is very helpful, for example, in post-facto analysis of attrition. A few attitude and behaviour question slipped into the application form might be enough to get started. The danger is that applicants figure out what combination of responses will help them, and 'game' the responses. There may be some way around that.

If you crack open that puzzle, you find yourself with the 85% of the African-American students (or 65% for caucasian) who are not being bid for aggressively. Of these, perhaps only 20% are of interest to you, but if you can zoom in on that 20% you've done yourself a real favor: found good students who don't cost as much as the high SAT crowd.

This is a project I'll be engaged in soon. I'll start with the app form and add some questions of a the type identified from an analysis of CIRP responses compared to college GPAs. Once we've identified a few indicators of success, we'll focus a few questions on those topics.