Attention must be paid

A few days ago, the Public Editor of the New York Times reported on a study done by Jack Fischer, a student at Binghamton University.  The basic conclusion of the study was that public universities got less coverage in the Times than private universities, controlling for other factors that could be expected to affect coverage.  In reading the story, I was struck by the list of other factors:  "The Times gives more coverage to colleges and universities that are highly ranked, larger, more liberal, and closer to New York.  That's probably not surprising, but after controlling for those factors....."  Size, distance from New York, and rankings [presumably US News] seem like good controls, but liberalism?*  If you think of the control variables as representing the coverage that different institutions "deserve," it certainly doesn't belong.  If you just think of other things that might affect coverage, I could see it as making some difference, but not as likely to be among the top factors.  However, there is something going on:  if you regress the logarithm of the number of articles on rankings, number of students, distance from New York (all logged), public/private, and the political rating, the t-ratio for college liberalism is 4.6.

So I added some other control variables.  The obvious one was research quality.  Presumably top research universities will be disproportionately represented in stories that report research finding or the opinions of "experts."    There are good, although somewhat dated, measures of perceived research quality from a survey of graduate departments conducted by the National Research Council in the mid-1990s.  H. J. Newton, a professor of statistics at Texas A & M, computed university averages, which are given here.**

Next I looked at the residuals from a regression of articles on rankings, size, distance from NYC, public/private, and graduate program rating.  The largest positive residuals were for American, NYU, Georgetown, Fordham, and Berkeley.  The largest negative ones were for Rochester, Worcester Polytechnic, Stevens Institute of Technology, Renssalaer Polytechnic, Purdue, and the SUNY College of Environmental Science and Forestry.

The residuals suggest two things:  first, that distance from New York isn't an adequate measure of the Times's geographical focus; second, that it gives less attention to technical universities.  So I added two variables:  one for location in the northeast corridor (Washington to Boston) or on the West coast, and the other for technical universities.  I defined both narrowly to minimize the chance of biasing things towards my expectations.  For example, my definition of technical schools included only places called "Polytechnic," "Institute of Technology," or mentioning a specific application (mines or forestry).  The estimated effects of both variables on coverage are highly significant, and when you include the political rating the estimate is considerably smaller than it had been and no longer statistically significant.  The reason that the controls matter is that the universities which are rated as more liberal tend to be on the coasts or to have highly-ranked graduate programs.

This doesn't overturn Fischer's basic finding:  with the new controls, the public/private difference is smaller but still statistically significant.



*The rating is from an organization called College Prowler.  I couldn't find much information on it, but it seems to be derived from surveys of students who signed up on their website.
**The NRC did another round of ranking in 2010, but they changed the methodology, and the resulting rankings were widely and deservedly regarded as essentially worthless.