Nicholas Kristof recently wrote a NYT column about how pundits should be graded by how accurate their predictions are. But most of my predictions are really boring and depressing. So, none of the other pundits would touch making predictions about topics I'm interested in.
Everybody else likes to make predictions about tournament-like phenomenon that are inherently interesting because they are hard to predict: e.g., Who's going to win the NCAA basketball championship? Who's going to win the next election? Will the stock market go up or down? So, most pundits aren't much more accurate than random, but they are popular.
In contrast, I hate being wrong, so I like to make predictions about things that have a track record. I've been following social science stats for 37 years now the way Bill James follows baseball stats. (National Review published a letter from me commenting on Ernst Van Den Haag's review of Christopher Jencks' Inequality: A Reassement of the Effect of Family and Schooling in America when I was a freshman in high school.) Over that time, not all that much has changed.
For example, after I posted LSAT scores of law school applicants by race, a reader asked about MCAT scores of medical school applicants by race.
So, I said, to myself: I bet the MCAT scores will be a lot like the LSATs, except for the Asian non-verbal boost that will show up more in the MCATs than the LSATs.
From the American Association of Medical Colleges, here are the 2007 MCAT scores and GPAs for both applicants and matriculants (i.e., people who were accepted and enrolled). I spent some time converting them from raw mean scores (which you can find at the link above) to where each group would fall on the white percentile ranking. In other words, the mean score for whites is set to always be the 50th percentile. So, the mean score of the Mexican-American applicants on the MCAT's Verbal Reasoning subtest would fall at the 21st percentile of white applicants, the African-American mean would fall at the 10th percentile, and the Asian-American at the 36th percentile. (This is based on the means and the standard deviations, so it assumes normality. Your mileage may vary, but this should be close enough.)
|2007 Med School||Mex-Am||Af-Am||Asian-Am||White|
|% of total applicants||2.4%||7.4%||19.8%||57.0%|
|Percentile if White|
|MCAT Verbal Reason||21%||10%||36%||50%|
|MCAT Physical Sci||25%||14%||61%||50%|
|MCAT Biological Sci||25%||10%||54%||50%|
|% of total matriculants||2.5%||6.4%||19.9%||59.9%|
|Percentile if White|
Obviously, there's a lot of affirmative action in the med school racket: the acceptance rate (43%) for Mexican-Americans is virtually the same as for non-Hispanic Whites (44%) even though Mexican Americans average around the 26th percentile of the white distribution in MCATs and college GPA. And 36% of blacks get accepted compared to 44% of whites even though blacks scores and grades are down around the 17th percentile of the white distribution.
In fact, the AAMC posts offical grids showing how much easier it is to get into medical school for Non-Asian Minorities (NAMs) than for overall applicants. For example, 32.4% of "self-identified" NAMs get accepted to medical school with 3.00 to 3.19 GPAs and MCATs of 21-23, while only 13.4% of overall applicants get in with the same credentials. For applicants with 3.40 to 3.59 GPAs and 24-26 on the MCAT, 67.1 of NAMs get in versus 27.5% of the overall applicants (and somewhat less for Whites/Asians, of course).
In summary, it's probably a good idea to get a second opinion.
There just aren't many high-scoring NAMS. Only 45 of the 1,682 applicants who scored 39-45 on the MCAT (the highest bracket) were NAMs. In contrast, NAMs made up 1,731 of the 2,705 in the lowest scoring bracket of 5-14.
Asians are accepted at a 42% rate versus 44% for whites, which sounds about right because their test scores are almost exactly the same and their grades are slightly worse.
One interesting note is that white applicants to med school have slightly better grades in college than Asians, both in science and non-science courses. I suspect that whites who apply to med school tend to be individuals who want to be doctors, while Asian-Americans who apply to med school tend more than whites to be individuals whose parents want them to be doctors. Hence, the much higher percentage of Asians who apply to medical school, and their slightly less impressive college grades.
Also of potential interest: Law School Admissions Test (LSAT) scores by race