This article from the Harvard Crimson garnered my dismay for saying that moderate professors dominate the college campus when I noted from the article that 62% of college profs are liberal.
Turns out, the Harvard Crimson was reporting accurately; please note their pie chart.
But the original study is the one which said moderate professors dominate.
Stephen Balch also noted the issue and wrote:
Perhaps Gross and Simmons were counting on people listening to their patter rather than looking at their data. Numbers put a lot of folks to sleep, but if you keep one eye open, these numbers tell a remarkable tale.
Take, for example, their claim that there is a “moderate” bloc comprising 46.6% of the sample, which is bigger than the 44.1% they classify as “liberal”, and the 9.2% they call “conservative”. Examined more closely, it turns out that this claim depends on a methodological sleight-of-hand. Gross and Simmons produce their “moderates” by taking seven survey-elicited ideological self-designations, “extremely liberal”, “liberal”, “slightly liberal”, “middle-of-the-road”, “slightly conservative”, “conservative”, and “very conservative”, and lumping the two “slightlys” with the “middle-of-the-roaders”. But is this composite category actually made up of “moderates”? When Gross and Simmons report how the seven original categories distribute themselves according to a multi-issue policy scale, it turns out that all but the self-designated “conservatives” and “very conservatives”, fall to the left of the scale’s center. Worse yet, the “slightly liberal”, are actually closer on the scale to the “liberals” and “extreme liberals” than they are to the “middle-of-the-roaders” with whom Gross and Simmons lump them. (On the 1 to 5 scale, the score for the “liberal/extremely liberal” group is 1.4, that of the “slightly liberals” 1.7, that of the “middle of the roaders” 2.2, and that of the “slightly conservative” 2.8. Only the “conservatives and very conservatives” actually fall to the right of the scale’s midpoint at 3.7.)
Something similar to this data gerrymandering also occurs in the analysis of professorial partisanship and voting behavior. The authors report the existence of a sizeable “Independent” bloc representing 35.8% of their sample, with the Democrats constituting 50.3% and the Republicans, 13.9%. But when Gross and Simmons report on the 2004 presidential vote, it becomes evident that whoever these “independent” professors may be, they don’t divide their support with anything like partisan evenhandedness — Kerry’s share of the overall vote standing at 77.6%, with Bush’s at 20.4%.
In both cases, Gross and Simmons proceed as if self-labeling among professors takes place in the same way that it would amid the general public. But that’s far-fetched. A professor only a tad left of center, given the environment in which he moves, might well think himself “slightly conservative”, but it is misleading for a researcher to assume him so in any more global sense. Likewise, a radical academic plaguing both parties’ houses might still define himself as an “independent”.