Wednesday, September 09, 2009

One More Time, The Washington Monthly College Rankings are Shoddy!!!!!!!

Near the end of a discussion in today's New York Times of a new book on college graduation rates and the failure of schools to graduate low income students in a reasonable time David Leonhardt writes
Washington Monthly magazine has published a new college ranking based in part on graduation rates. (Kudos to Penn State, among others.)
This is very frustrating because the book, Crossing the Finish Line, by William Bowen ,an economist and former Princeton president, Michael McPherson ,an economist and former Macalester College president, and Matthew Chingos, a doctoral candidate, by all appearances is based on careful data analysis, while the Washington Monthly rankings are clearly a shoddily executed marketing devices.

So it bears going over one more time how shoddy the Washington Monthly ranking is.



First, the the magazine had to chuck last years ranking completely because they used an incorrect number of the percentage of students at Texas A&M with Pell grants amongst other problems. Here is their explaination.
In compiling this year’s rankings we established that we had made data-entry errors in our 2007 rankings. Texas A&M University was given the wrong Pell Grant number, so it placed first in the university ranking, when it should have been slightly lower down. In addition, some colleges were assigned the wrong research expenditure figures, throwing off the overall college rankings. We deeply regret these past errors and have changed our procedures for compiling the annual rankings so that they do not recur.
Here is their ranking from last year. You can see that the Pell grant percentage given to Texas A&M last year is 49%, but the correct number is likely to be closer to this years number of 19%.

The Pell grant number enters the ranking in two ways. First, it used directly schools with higher percentages are ranked higher. Second, it is used in a regression model, along with SAT scores to get a predicted graduation rate for a school. The difference between the actual gradauation rate and the predicted gradation rate is then used as a second measure to rank schools. The larger this differential the higher the rank of a school. The idea being that this is some sort of proxy for how well these schools serve low income students.

The details of the fitted model are not available, but it is generally the case that the larger the percentage of students with Pell grants the lower the graduation rate. Hence an inflated Pell grant number would lead to a deflated predicted graduation rate and an inflated differential. This would result in a higher rank for the school with the inflated Pell grant number. It would also effect the fit model by making the slope associated with the Pell grant percentage less negative. The result is that other schools would have their predicted graduation rates inflated with the overall effect of deflating their graduation differential and their ranks.

After last years embarressment, you would have thought that the folks at the Washington Monthly would have been extra careful this year. But they weren't. This year they used the wrong Pell grant number for Penn State. (By the way, they did last year too.) The number they used, 25%, is the system-wide number. The perecentage at University Park is not availible for this year or, at least, I haven't found it yet. But the number has historically been around 15% and it is likely to be in the same ballpark this year.

Are there other such mistakes this year? I haven't gone through the whole ranking, but I did catch one more error. It also the 30% Pell grant number for Ohio State is inflated and should be closer to 17%.

Penn State shunts its financially disadvantaged students to its Commonwealth campus system where the Pell grant percentage can be as high as 63% and the graduation rates as low as 34%. Given this, it is hard to see how Penn State could rank high on any measure of social mobility.

It is particulary grating to see Leonhardt giving Penn State an undeserved kudos based on the the shoddy Washington Monthly ranking.

Technorati Tags: , , , ,



Powered by ScribeFire.

1 comment:

Media Mentions said...

Since Washington colleges are in fact doing pretty well academically (http://www.pressdisplay.com/pressdisplay/showlink.aspx?bookmarkid=YJ5OYYQS6MA1&preview=article&linkid=ac39bfae-04a2-414b-8c5c-20236f057ae6&pdaffid=ZVFwBG5jk4Kvl9OaBJc5%2bg%3d%3d), I think it's time we focus on the students elsewhere as well, don't you?

Sincerely,
MediaMentions