Timothy R. Johnson and Jason M. Roberts
costs” view of Supreme Court nominations discussed earlier. Despite consider- able evidence that the effectiveness of presidential speeches declines along with presidential approval (Ragsdale 1984), these data suggest that presidents may try to make up for the lack of effectiveness by increasing quantity or, in the case of President Clinton, choosing noncontroversial nominees. Finally, the model reveals no discernible effect from nominee qualifications or years remaining in the president’s term on the frequency of presidential public statements.
The Impact of Going Public
While we have provided both statistical and substantive evidence that presi- dents go public to support their Supreme Court nominees, we do not yet know whether presidents can actually help their nominees do better by going public after the nomination has been sent to the Senate. To test our final hypothesis we regress the difference between the predicted number of votes against confirma- tion and the actual number of no votes each nominee received in the Senate (Segal and Spaeth 2002, 216) on the president’s use of public statements (the depend- ent variable in Table 2).29 We expect a positive relationship between these vari- ables because if going public is an effective strategy for presidents we would expect the Senate to cast fewer no votes than predicted. We also included a vari- able to measure the time that elapses between nomination and Senate action, as well as a variable for the year of nomination. Our intuition for the former vari- able is that longer confirmation battles often arise because something unforeseen in the nominee’s background might cause more “no” votes than predicted by Segal and Spaeth’s model. For the latter we expect more political, and therefore more controversial, nominations in recent years. Thus, we expect a negative sign on both of these variables (meaning more no votes than actually is predicted).30
29 The predicted no votes are derived from Segal and Spaeth’s (2002, 216) model of individual senator’s confirmation votes, where the independent variables used to predict “no” votes are nominee ideology (Segal/Cover scores), nominee qualifications, ideological distance between nominee and each Senator, presidential strength in the Senate (Senate party control), presidential approval, and the extent of interest group activity for and against a nominee. The actual no votes are also taken from Segal and Spaeth. The dependent variable is the difference between the two. For example, the obser- vation on our impact variable for Justice Thomas is 8 (56 predicted no votes minus 48 actual no votes). Thus, a positive sign on this variable indicates fewer no votes than predicted, while a nega- tive sign means more no votes than predicted. Note that we use an OLS model here, instead of a GEC or another count model, because we have negative values in our dependent variable. Thus, these other models cannot be used.
30 Certainly there may be some endogeneity in this process—meaning that the dependent variable (difficulty in the Senate) might be affecting the main independent variable (the president’s propen- sity to go public). We control for this to a large extent with the variable that measures time from nom- ination to a final confirmation vote is taken. As argued above this variable acts as a proxy for the level of contention encompassed by the confirmation process. Additionally, using the predicted number of no votes as part of the dependent measure allows us to distinguish between ex ante support/opposi- tion for a nominee and support induced through presidential persuasion or opposition as a result of revelations occurring during the confirmation process. Thus, while this is a difficult problem with which to deal, we are confident that we are capturing what is actually happening in this process.