03 April 2018

The NSF GRFP problem continues

This morning, a fine scientist congratulated two undergraduates in her lab about winning National Science Foundation (NSF) Graduate Research Fellowship Program (GRFP) awards. I thought, “Huh. They’re out? And two seems like a lot from one lab.”

A few years ago, Terry McGlynn wrote an important blog post about how tilted the playing field is for the NSF GRFP awards. He compared awards to Harvard students (with about 7,000 undergraduates) to the more than 20 campuses in the California State University system (over 400,000, according to a check of Wikipedia).

The NSF is good about making it easy to find a list of all 2,000 awards in this program. I went looking for the same comparison of one Ivy League university to an entire state’s system. Embarrassingly, I screwed up the calculation on the first pass, not realizing that several California State universities don’t say “California State” in their name, unlike the University of Texas institutions.

Harvard got 43, and all of California State get 50 (thanks to Terry for counting here and here).

Cal Poly Pomona 4
Cal Poly SLO 5
CSUCI 1
CSUDH 1
CSU Fresno 1
CSU Fullerton 8
CSULB 2
CSULA 1
Sac State 1
CSUSB 1
CSUN 5
CSUSM 3
SDSU 6
SFSU 6
SJSU 3
Humboldt State 2

So one lab in Harvard alone equaled the entire combined output of eight different California State universities (separately, not combined).

If this sort of pattern intrigues you, you must for to Natalie Telis’s post where she digs down into the numbers. Not just this year’s, but over 28,000 awardees worth of data, from 2011 to 2017. It’s bloody brilliant. One of her first points is, “The most expensive undergraduate schools have an extreme excess of (NSF GRFP) recipients.” She also makes some comments on Twitter about this.

I can’t wait to see what she finds for 2018 data.

Matt Cover did some similar things the previous year, and found no relationship between institutional enrollment and number of grants.

External links

NSF Graduate Fellowships are a part of the problem
The price of a GRFP, part 1
Matt Cover thread from 2017

No comments: