Rich, Amy, and Susan have done a great service by writing their study. My comments here focus on a specific aspect of the paper: the issue of causality. The paper itself makes a number of explicitly causal claims, referring (for example) to the "most influential causal factors" that might determine ratings, and to the construction of "a series of causal models" (p. 11).
My narrow point is that models of which they speak aren't really causal models at all; they are straightforward regression-type models, of the sort social scientists do a lot of. That isn't a criticism of what they've done, but rather just an observation, albeit one that suggests a simple change in approach.
Manipulability theories of causation (perhaps most famously set out in Holland's (1986) paper) -- upon which most of the current approaches to causal inference are based -- certainly have their critics. But they have a natural application to the central question of the paper: What ABA rating would nominee X have received had s/he been appointed by a Democratic president, rather than a Republican (or vice-versa)?
This (to me) suggests that the paper is an obvious candidate for a matching-based analysis of the influence of (e.g.) party on ABA ratings. Such an approach is (a) very easy to do, as a practical matter, (b) allows one to control for other potentially confounding factors, and best of all (c) provides a direct answer to the question posed above that can be interpreted in causal terms. These approaches are increasingly widely used in empirical legal studies, and seem (to me) to hold particular promise in addressing the authors' central question.