Tuesday, May 09, 2006

All the rest is vanity 

Give a guy a dataset and SPSS, and he'll regress anything:

James Felton, a professor of finance and law at Central Michigan University, and colleagues looked at ratings for nearly 7,000 faculty members from 370 institutions in the United States and Canada, and his verdict is: the hotter and easier professors are, the more likely they�ll get rated as a good teacher.

As far as students � or whoever is rating professors on the open Rate My Professor site � are concerned, nothing predicts a quality instructor like hotness.

Felton found a positive correlation of 0.64 (0.00 means there is no correlaton whatsoever, and 1.0 describes a perfectly linear relationship) between the �hotness� and �quality� � quality is a composite of �helpfulness� and �clarity� � ratings on the site.

...The 102 professors ranked as least attractive in the sample had an average quality rating of 2.14, and an average easiness rating of 2.20. Meanwhile the 99 �hottest� profs had an average quality score of 4.43, and an easiness rating of 3.5.

There also appears to be a bias in how students rank various deparments: Faculty in engineering and sciences were both not hot and low quality; the top departments for quality were languages, sociology and political science, and they were all in the top six for hotness.

Two observations: First, there's reverse causality -- if the professor grades easily, they may be seen as more attractive and offering a high-quality class. ("Oh, of course it's a great class. I learned so much, as you can see from my 'A'".) Second, easy is subject to diminishing returns and is relative. When you see it even in the engineering schools, it might be time to surrender -- then nobody will get the benefit over anyone else.

I find the most interesting statistic in the article to be the relatively low correlation between hotness and easiness. Could it be that older faculty are substituting the activity they control for the appearance they cannot, in order to maintain contact and likability?