« ELS and not be 'laughably beside the point' | Main | Case Studies & Empirical Scholarship »

24 April 2006


Brandon Bartels

I very much agree with Sara concerning the importance of recognizing the traditional tautological issue when employing measures of policy preferences.

However, while recognizing Sara’s concerns, I think a viable defense can be made for using various measures based on votes, including Martin-Quinn scores (which Martin and Quinn discuss in their working paper) and scores based on justices’ past behavior. I would be curious to hear what Sara and others think.

[By the way, at Midwest, I expressed reservations regarding Wilson’s use of scores that ranged over a justice’s entire career, favoring instead a “lesser of two evils” measure that used justices’ liberalism scores for terms previous to the one under examination. Alternatively, one could use percent liberal in the previous term.]

Consider first a preferences measure that uses the proportion of liberal votes cast by each justice in the previous term for an analysis at the choice level (that is, analyzing justices’ choices in a given set of cases over a given period of time). First, using proportion liberal for the PREVIOUS term reduces the tautology issue to a degree. Also, this vote-based measure would tap a justice’s average tendency to favor liberal or conservative policy. So it’s not necessarily a case of “using votes to explain votes.” Instead, the measure represents our best guess of a justice’s ideological leanings toward legal policy based his or her most recent tendency to favor liberal or conservative policy. Moreover, vote choice models require a preferences measure that can validly ascertain how much more liberal or conservative Justice A is compared to Justice B, and vote-based measures communicate these ideological differences between justices with greater validity than, e.g., Segal-Cover scores.

I think Martin and Quinn make some excellent points in their working paper (which Sara referenced) on whether their scores can be used as independent variables in vote choice models. Martin and Quinn essentially conclude that while the tautology issue is always a concern in a theoretical sense, practically speaking, under certain conditions, it does not represent a huge concern. [Of course, people should read the Martin-Quinn paper for further details] I also think it’s important to note that the tautology issue is one of a number concerns we should think about when employing preferences measures in vote choice models. Other important issues are (1) making valid comparisons about ideological differences between justices, which I talked about in the previous paragraph, (2) the issue of inter-justice comparability over time, and (3) accounting for the possibility that justices’ preferences may change over time. Martin-Quinn scores rank highly on all three of these additional criteria.

Sean Wilson

No one, I imagine, disputes that judging involves value choices and that law is not algebra. So if values display themselves in judging by definition, it is logical to collect these values as they manefest themselves in the work product through some objective accouting process. (As to the ultimate cause of the bias, let's put that aside for a moment).

Also, it must be remembered that it is not methodologically problematic to use career ratings in a logit model to explain discreet segments of voting data that, together, comprises the summary statistic. This is nothing other than an appeal to a "moving average." Of course, it may become problematic if the modeler simply uses a base rate to explain every set of votes analyzed, and each time calls the base rate "ideology." That is clearly tautological. But if one uses career ratings as a proxy for the extent of liberalism and conservatism possessed by justices, period, the model is not explaining votes with votes, but is merely making an empirically-based value assignment that is tested with moving-average logic.

If researchers don't like this, they can still use the moving average as a descriptive phenomenon in need of further explanation. That is, what causes the value assignments to move in the manner they do? Why are the values punctuated for some areas of voting (search and seizure) but not in others (core political speech)?

Hence, there is utility in using career numbers even if they are not endogenous.

The comments to this entry are closed.


April 2018

Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30          

Site Meter

Creative Commons License

  • Creative Commons License
Blog powered by Typepad