Professor Heather Williams, chair of the Athletic Committee for 2008-2009, kindly replied to my request to know more about the details of their statistical analysis (pdf).

As chair of the committee, I am responsible for the statistical analyses. The design of the analyses was the responsibility of the committee, and several of us are statistically competent. The analyses themselves were run by the Provost’s office, and although you are certainly free to ask Chris Winters about them, I would imagine that he’ll be referring all questions to me. The report will not describe all of the results in detail, but will provide enough information about how the analyses were set up and the results to let the readers understand exactly what we did.

A quick summary, we ran a number of simple linear models (and logistic regressions, where appropriate to the question) that included a variety of variables that seemed likely to have affects on academic performance: reader rating, gender, class (fr., so., etc.), a proxy for socioeconomic status, and sport category (high/low profile), as well as a number of interaction terms.

I’m sure that, as an economist who is into numerical analyses, you would prefer to have more information, but I wouldn’t be comfortable about giving any additional details beyond this summary of how the analyses were done. The more details we give, the more people want to know, and it’s important not to make it possible for anyone (including us) to gain too much information about individual athletes.

Perfectly reasonable! I still think that the Committee ought to share the regression results (which are available in an Appendix on file in the office of the Dean of the Faculty) with the entire community, but this is a minor quibble. The Report is high quality and represents the best of the Williams tradition of faculty governance and transparency. Kudos to all involved!

For those interested, below is the e-mail that I sent to Professor Williams.

Good luck with the write-up. I thought that Michael MacDonald did a
great job on the last go-around. Would you mind telling me who is
handling the statistics on the report? I would like to follow up with


PS. Perhaps you have already covered this in the report, but the
article I read in the Record did not make clear whether or not you
were doing the sort of adjustments that should be done. I wrote a bit
about this on the last report:

CGCL, Day 3: A hard look at varsity athletics

The analog in the current discussion (and, again, I would be happy to
hear that the Record got this wrong and that you have already done
this analysis) concerns the claim that female athletes do as well as
female non-athletes. There are two issues:

1) Critics of the College’s tips policy are not concerned with female
athletes in general, they care about tipped athletes in particular.
How have the 25 (?) tipped female athletes per class performed
relative to their peers?

2) Although it is interesting to compare female athletes to
non-athletes, the more relevant comparison is between female athletes
and female students who are “matched” to the female athletes on all
relevant criteria except for their athletic participation. I have
heard (but do not know for sure) that female athletes are less likely
to be an under-represented minority and more likely to have attended a
prep school. I believe (but do not know for sure) that such students
have lower GPAs than other students. If that is the case, then we need
to compare female athletes (as a group) to a group with similar
demographics in terms of race and secondary schooling.

No need to waste time answering these questions just for me, but I do
hope that a discussion is included in the report.

Print  •  Email