For What It's Worth (nothing to do with Buffalo Springfield)

Earlier I noted the problems in attempting a study comparing scholarship of elite and non elite law school grads who now teach at mid level schools. The law schools I selected as elite were Harvard, Yale, Michigan, Columbia, Yale, Stanford, California-Berk., Chicago, Duke, Virginia NYU, Cornell, Penn. This was complied from Brian Leiter's web site listing the schools that supply most law professors. I ranked undergrad school based on USN&WR.

Each faculty member was ranked by levels of "eliticity" like this:
No elite credentials -0
Undergrad elite --1
LLM from elite -2
Undergrad and LLM from elite --3
JD only from elite--4
Undergrad and JD from elite --5
Undergrad, JD,and Post grad from elite -- 6

The number of westlaw listed publications since 1990 while the faculty member was at his or her current school was determined and divided by the number of years at that school to get westlaw hits per year.

A simple correlation was run and found to be .15 without culling out any outliers. By outlier in this context I mean people who are at the extremes and able to move the numbers single handedly. Including them in the analysis may create an inaccurate impression. I do not know of an official test for identifying an outlier. I did look closely at the outliers and without exception they were the people who had racked up far more hits from "introductions" colloquiums, symposiums, transcribed panel discussions, and short book reviews than their colleagues. Whether appropriate or not, I took out some of the extreme outliers and the correlation fell to .036.

That seems tells you that the outliers were more often from the elite ranks. There are interpretations here -- they are more productive, they benefit from better networks, they are more self-promotional or aggressive -- or all of the above.

The problem is that outliers can go both ways. Having eliminated people with a great number of hits, would it make sense to take those out with no hits? I did that too and ended up with a correlation of .11. If my interpretation is right, then the elites must be found at both ends and the upper end can pull up the lower end by virtue of the fact that you cannot write less than 0 but there is no limit on how much you can write.

Not being a statistician gives one a great deal of freedom to do things that are probably not acceptable. So I ran a little regression using one independent variable -- eliticity. The R-square was minuscule and the coefficient on eliticity was very far from significant.

Given all the judgments and arbitrary line drawing I did, I would not suggest that anyone go to the bank with the results but, for what it's worth, it did not appeal that hiring at mid level law schools should be guided by elite credentials, or at most they should be a tiny factor at the margin.