Reading the Statistical Tea Leaves on Justice Kennedy at Oral Argument in NAMUDNO

Much has been made of Justice Kennedy’s comments at the NAMUDNO oral argument, leading supporters of Section 5 to expect the worst and prepare a legislative response. But the question has been raised about how much to read into comments at oral argument.
Along comes a fascinating paper on oral argument questions and the supreme court by Epstein, Landes, and (Richard) Posner. The paper is so rich, and well worth reading, but I want to pull out a snippet about Justice Kennedy’s partial exceptionalism, which might give a bit of comfort to supporters of NAMUDNO–though the trends are mixed. This is from pages 16-17 of the pdf (footnotes omitted and my emphases added):

    Even after we control for the direction of the lower-court decision and the participation of the United States, the number of questions and the total words in question still provide a reasonable predictor of most Justices’ votes. Regardless of the way in which we measure questions, the variables are always significant (p ≤ .05) and have the expected sign for Breyer, Ginsburg, Scalia, Souter, and, with one exception, Stevens. For Roberts, questions to and words in questions to the petitioner are more predictive of his vote than are questions to the respondent; for Alito the reverse holds. The one exception is Kennedy; none of his question variables produces a statistically significant coefficient.
    This is consistent, as explained earlier, with his being the swing Justice. Kennedy’s and Thomas’s questions (or in Thomas’s case the absence of questions) to the petitioner and respondent have no significant effects on their votes; but might their votes be influenced by the questions of other Justices? The question is explored in the second part of Table 12. All eight regression coefficients have the expected signs–that is, Kennedy and Thomas tend to vote for the outcome signaled by the questions asked by the other Justices–and seven are statistically significant. Unsurprisingly, the effects tend to be smaller than the corresponding effects of a Justice’s own questions.
    We also considered whether Kennedy’s and Thomas’s votes are more responsive to questioning by other conservative Justices than to questioning by liberal Justices. The answer is yes, but most of the regression coefficients (not shown in Table 12) are not statistically significant. Both Thomas and Kennedy are less likely to vote for the petitioner the more questions conservatives ask the petitioner (Thomas, significantly so); and Thomas and Kennedy are more likely to vote for the petitioner the more questions conservatives ask the respondent (although the regression coefficients are not statistically significant). Kennedy but not Thomas is significantly less likely to vote for the petitioner the more questions the liberals ask the petitioner’s lawyer. Kennedy’s votes are not affected by the number of questions the liberals ask the respondent’s lawyer. Thomas’s votes are affected, but the influence runs counter to our other findings: the more questions the liberal Justices ask to the respondent, the more likely Thomas is to vote in his favor (p ≤ .10).
    The ideological variable in Table 12 has a significant positive effect on the voting of the five conservative Justices and a negative effect on the voting of the four liberal Justices. For the latter group, however, the negative coefficient is significant only for Breyer. The implication is that the conservative Justices, plus Breyer, have a stronger political commitment than the liberal Justices, other than Breyer. Petitioners won about 69 percent of the cases in the period covered by our study, but when the petitioner was appealing a liberal lower-court decision the figure exceeded 90 percent for Scalia, Roberts, and Thomas and 75 percent for Alito and Kennedy. The corresponding figure for the liberal Justices is less than 50 percent, but it is statistically significant only for Breyer.


Comments are closed.