Rankings games, continued

In January I noted a new, annual, for-profit ranking system of Ph.D.-granting departments that relies entirely on productivity measures and not at all on reputation measures. The new round of rankings has been published. Political science is here; Chronicle subscription may be required. The top ten:

1. Harvard Government
2. Harvard Kennedy School
3. Berkeley
4. Michigan
5. Stanford
6. USC
7. UNC Chapel Hill
8. Columbia
9. Yale
10. Wisconsin-Madison

Compared that with last year's list:

Wash U
Harvard
Yale
SUNY Stony Brook
UIUC
U Kansas
U Maryland College Park
Princeton
UCSB
UVA

The only change in the formula I can find is that it seems the weighting of books:articles has been reduced from 5:1 to 3:1 in the social sciences. (I think that's been done.) But that's very strange. There's no way Wash U should drop from #1 to below the top 10 as a result of increasing the relative weight of articles. SUNY Stony Brook and UIUC should also be helped by that change, not hurt by it.

The new rankings look a bit more like what one would expect than the old ones did, though they're still not the same results as one would get with a reputation measure. But that suggests that the ostensibly objective measure has been tweaked to better fit preexisting intuitions (the way US News changed its formula after it reached the implausible result that Caltech was the best university in the US)-- which would seem to undermine the rationale for universities to pay large sums of money for the proprietary objective data being collected.