New York Times opinion columnist David Leonhardt linked to EphBlog in his column a few weeks ago about the enfolding admissions bribery scandal. Thanks! Alas, Leonhardt is one of the more clueless education reporters, a not particularly clued-in breed, as we have documented again and again and again. Is a link from him something to be proud of? Either way, his article merits 10 days of discussion. Day 4.

With the last three days of discussion as background, we can now go through the article line-by-line.

Getting a peek inside the college-admissions process isn’t easy.

What nonsense! There are a dozen or more excellent books on the college admissions process. (EphBlog recommends The Gatekeepers and A is for Admission.) We provide a detailed summary of how admissions works at Williams, updated each year for your convenience. There are scores of academic articles.

But a team of academic researchers managed to do so several years ago.

Leonhardt is such a hack! What does the word “several” imply to you, dear reader? Three? Six? Eight? Try 14! And that is just when the book was published. The underlying information is from the (impressive!) College and Beyond database, constructed in the mid 1990s. Leonhardt uses the phrase “several years ago” to describe a study conducted with data more than 20 years old!

It helped, no doubt, that two of the researchers were former college presidents — William Bowen of Princeton and Eugene Tobin of Hamilton.

Bowen has made a nice post-presidencies career of writing books with suspect empirics and minimal replicability. Nice work if you can get it.

The researchers were given access to anonymous admissions records at 19 elite colleges and then analyzed how admissions offices treated different groups of students. Low-income students, for example, were no more likely to be admitted than otherwise similar students with virtually identical academic records. So-called legacy students — those whose parents attended the same schools — received substantial boosts. So did underrepresented minorities.

Much of this was probably true in the 1990s. But Leonhardt is passing it off as being true today when, on many dimensions, things are vastly different. First, legacy advantage matters much less today than it did in the 1990s, for reasons that we have explored ad nauseum. Second, there has been a big push in favor of low-income students. Third, note how Leonhardt pretends that legacies and URMs both received “substantial boosts,” when, in fact, the boost for URMs was much bigger in the 90s than that for legacies, and that is even more so today.

The average legacy at Williams has a higher SAT than the average non-legacy. The average African-American at Williams has an SAT score 200+ points lower than the average non-African-American student. These two things are not comparable!

But the biggest boost went to recruited athletes: An athlete was about 30 percentage points more likely to be admitted than a nonathlete with the same academic record.

I realize that the academic research uses terminology like this, but it no longer captures how admissions works, to the extent it ever did.

Assume that Williams has a 20% admissions rate. The only way for that “30 percentage points more” formulation to make sense is if someone (who?) puts together 200 athletes that they (who?) want to come to Williams. This list of 200 goes to Admissions, and 100 are accepted. This 50% acceptance rate is, indeed, 30% more than the 20% baseline, but the theoretical process which allows that statement to make sense is not how Williams (or Yale) athletic admissions work. This is how they work.

Summary: If you are on the coach’s list, you are (almost) 100% going to get in. Any statistical model which does not account for that process will produce nonsense numbers.

Print  •  Email