Portal de Periódicos da CAPES

Choosing When to Choose: Explaining the Duration of Presidential Supreme Court Nomination Decisions

2014; Taylor & Francis; Volume: 41; Issue: 1 Linguagem: Inglês

10.1080/07343469.2013.851129

1944-1053

Charles R. Shipan, Brooke Thomas Allen, Andrew Bargen,

American Constitutional Law and Politics

Abstract Although the selection of Supreme Court nominees is of tremendous importance, the amount of time it takes presidents to select nominees varies dramatically across nominations. We argue that the timing of nominations is a function of the political constraints the president faces, and we focus in particular on partisan or ideological disagreement between the president and the Senate. We test our argument using all Supreme Court nominations since 1882, and find that disagreement increases this duration. In addition, we control for a variety of other factors that can affect timing. We find that a newly elected president, a nominee's personal characteristics, and Senate rule changes also increase the length of the nomination stage. On the other hand, this stage becomes shorter as the Senate nears the end of each session and as the presidency's institutional capacity has increased over time. Notes 1. Throughout the article we distinguish between nominations, confirmations, and appointments. The nomination stage begins when a vacancy is announced and ends when the president selects a nominee; the confirmation stage begins with the announcement of a nominee and, for the Supreme Court, usually ends with a vote in the Senate (although it could also end with the withdrawal of a nominee); and the entire is process is the appointment process. 2. Ward (Citation2003) finds evidence for all of the factors mentioned previously—strategic, partisan, personal, and institutional—but also finds that partisan factors have been especially prominent in recent years. 3. Presidents also face geographic constraints when making appointments to the lower federal courts. Scott and Garrett (Citation2011), for example, found that presidents do occasionally change state representation on the circuit courts, but for the most part these changes do not appear to be politically motivated. 4. A host of studies, some quantitative (e.g., Nemacheck Citation2001, Citation2007; Nemacheck and Caulfield Citation2004; Watson and Stookey Citation1995), some primarily theoretical or richly descriptive (e.g., Yalof Citation1999; Goldman Citation1997; Maltese; Citation1995; Massaro Citation1990), have examined the president's choice of a nominee. Although none of these studies conduct empirical analyses of the duration of the nomination stage, most clearly acknowledge the vital importance of timing in the appointment process, refer repeatedly to delays during this stage, and are replete with examples of why presidents sometimes speed up the process of choosing a nominee and other times slow down. In addition, studies of departure from the Court (e.g., Atkinson Citation1999; Ward Citation2003) present information and arguments that are clearly relevant for the nomination stage, even if their main focus is on the prior stage when sitting justices are departing the Court. 5. Krehbiel (1998) makes a strong argument that the filibuster pivot plays a key role when Congress and the president engage in lawmaking. However, when modeling the Supreme Court appointment process, Krehbiel (2007) utilizes the median of the Senate, and not the filibuster pivot, as the key senatorial player in the game. We concur with Krehbiel's use of the median. Although nominees to the Court can, in principle, be filibustered, and although filibusters have occurred on nominations to the lower courts, in reality filibusters of Supreme Court nominations rarely have been considered and even less frequently have been utilized. In fact, prior to minority parties in the Senate raising the possibility of the filibuster over the past several years, there were only two cases in history in which the filibuster played a role: the Fortas nomination for Chief Justice in 1968, and the nomination of Stanley Matthews in 1881. The Fortas case, it should be noted, was based at least in part on ethics charges, not ideology, and thus would fall outside of our model. And the Matthews case was prior to the starting date for our analysis; and in any case, there are doubts that it was actually a filibuster. For an alternative perspective, see Johnson and Roberts (Citation2004, Citation2005). Not surprisingly, given that the filibuster has not played a central role in Supreme Court nominations, substituting the filibuster pivot for the median in our empirical tests produces insignificant results. 6. The reader is referred to the Moraski and Shipan (Citation1999)article for full details of the model. The basic technology builds on Ferejohn and Shipan 1990, which first incorporated the courts into spatial models of the separation of powers game. 7. This is in contrast to the first two regimes. In the first regime in the model, for example, the Court median will shift the same distance regardless of whether the president chooses someone who is either near his ideal point or has an identical ideal point. When the Senate is far from the president and on the opposite side of the Court, on the other hand, if the president inaccurately assesses a nominee's ideology, it could actually have the effect of moving the Court median away from his ideal point. Hence, he will spend more time assessing a nominee under these conditions. 8. More specifically, this stage will take longer when the Court is located between the president and the Senate, which of course occurs with higher probability when the president and Senate are distant from each other. 9. Yalof's (Citation1999) account of the appointment process contains numerous examples of these sorts of activities (e.g., President Ford calling Senate Judiciary Committee chairman James Eastland, Nixon's consultation with Eastland and other Senate leaders about his choices, etc.). See also Abraham 1999; Maltese Citation1995; and Watson and Stookey 1995. 10. We recognize that there may be countervailing pressures—in particular, Court appointments have become increasingly prominent and salient, drawing ever expanding levels of attention from the media and interest groups. These influences may provide an incentive for a president to move more slowly. We maintain that these countervailing pressures have been outstripped by the president's dramatically increased capabilities. 11. We began in 1882 since so many of the nominations prior to this date had missing data. We should note that when we extend our dataset back to 1865, the results are unchanged. 12. Some might argue that we should include a variable that accounts for how the seat opened (i.e., whether the vacancy was due to a death, which accounts for one-third of our cases). It might be expected, for example, that if a sitting justice dies, this catches the president by surprise, especially when compared to cases where the sitting justice announces his or her retirement. We are not persuaded. Our claim, as we make in this section, is that by identifying the date on which a justice announces his or her intent to retire, we are capturing the time when a president first learns of the opening. It is possible, of course, that presidents would have gotten a hint that an announcement is coming; but similarly, they often have a good sense of when a justice is ailing and nearing death. Furthermore, although such an opening on the Court might catch a president by surprise, it is also possible that, as Garrett and Rutkus (Citation2010, 17)) note, “[w]hen Justices die unexpectedly, Presidents can be eager to bring the Court back to full strength as soon as possible.” Thus, it is not clear that the nature of the opening should have any systematic effect. The data back us up on this point. Including a variable that measures whether the seat opened due to a death, as opposed to a resignation or retirement, produces an insignificant finding. 13. There may, of course, be cases in which justices secretly inform the White House of their intent to retire before making it public, and presidents may therefore spend time working on a nomination prior to an official announcement. While we acknowledge this possibility, it is more likely to happen when the justice shares a party or ideology with the president. We discuss a control for this in footnote 22 (along with many robustness checks in other footnotes), and note here that it does not affect the results of our main theoretical tests. In addition, presidents are more likely to prepare in advance when they have a higher institutional capacity, which we control for in our models with the Year Counter variable. More generally, while there is undoubtedly some error in this measure, it is most likely random; and if the official announcement date is irrelevant, that will only make it more difficult to find significant results. 14. All of the missing dates, as one might expect, come from observations for justices who served in the 1920s or earlier. 15. Because scores are not available for presidents for much of the time period we examine, we use the value for the median member of the president's party in the Senate as a substitute (e.g., McCarty and Poole Citation1995; McCarty and Razaghian Citation1999; Shipan and Shannon Citation2003). Substituting the president's ideology score, when available, for the majority party median does not affect the results. 16. If the vacancy occurs when the Senate is not in session, we use the end of the upcoming session. 17. We include this variable primarily as a control, and not as a key part of our explanation. But because we want to make sure that our key explanatory variables—namely, the ones reflecting the differences between the President and Senate—are not affected by this specific measure, we also tested many other measures. Including a squared version of the counter variable produces results essentially the same as what we report. A logged version is not significant, but also does not affect our other findings. Ragsdale and Theis (Citation1997) note that the presidency became especially institutionalized in the 1950s, which fits with Moe's (Citation1985) assessment of the presidency. It was also during the 1950s that Eisenhower first began to utilize the Congressional Liaison Office. Substituting a dummy variable that takes on a value of 1 starting in 1950 produces results very similar to those we report. Interestingly, although some observers have argued that the appointment process changed in the aftermath of the Bork nomination (e.g., Freund Citation1988; Epstein et al. 2006), a post-Bork dummy variable is not significant, consistent with the findings of other scholars (Guliuzza, Reagan, and Barrett Citation1994). Finally, including a dummy variable for each presidential administration or including a presidential counter variable (where we count by administrations rather than years) does not affect the significance of our main independent variables. 18. Not surprisingly, a Cox regression produces very similar results. In the results that we report in Table 2, we use robust standard errors. If we instead use regular standard errors, all the coefficients (except Senate Floor Debate and Female) remain significant. 19. We also considered several alternatives to these hypotheses and measures. First, we considered whether it might be the number of days to the next Senate recess, rather than to the end of session, that would matter. A variable for the number of days to recess, however, whether included in addition to, or substituted instead of, Days to End of Session, was not significant. Similarly, as Garrett and Rutkus (Citation2010) have pointed out, several appointments occurred during congressional recesses. It might be expected that the president will act more quickly to announce a nominee when a vacancy occurs during a recess. A dummy variable indicating recess vacancies was significant, but surprisingly it was positive. Finally, it might be expected that former senators would have less controversy surrounding their nomination and therefore might move more quickly through the nomination process. However, such a conclusion was not supported by the data. A dummy variable indicting service in the Senate was did produce a negative coefficient, but it did not reach a level of statistical significance. 20. These findings comport with those presented in Garrett and Rutkus's (Citation2010) study, which showed that presidents were quicker to announce nominations after 1980 than they were in years prior. Replacing our Year Counter with a dummy variable indicating post-1980 nominations produces similar results, with the duration being significantly shorter in recent decades. 21. Wars and other exogenous events that force themselves onto the president's agenda could also act as additional constraints and slow down a nomination. We tested for this possibility by including dummy variables to denote times of war and other crises involving the executive branch (such as international affairs and impeachments). These variables were not significant in any of the models and did not affect our independent variables of interest. 22. The simple fact that an administration has already gone through the nomination process, regardless of whether the previous nominee was a success or failure, could also decrease the time to nomination. We tested for this possibility by including a dummy variable for all first-time nominations. The results were in the opposite direction than expected and insignificant. A president may also be able to build on the work of a previous administration. If a new president succeeds a former president of the same party, the successor administration could inherit a shortlist, thereby accelerating the nomination stage. A dummy variable indicating whether the two administrations were of the same party was negative and significant, but did not affect other results. It is also reasonable to expect a possible collusion effect—if the outgoing justice and the current president are of the same ideological leaning, then the justice might give the president advance notice of his or her upcoming retirement. We tested for this possibility by including a dummy variable that equaled one if the current president and the president who nominated the outgoing justice were of the same party. The results were not significant. Finally, we also tested whether presidential mandates affect the length of the nomination stage, with mixed results. A dummy variable if a president claimed a mandate in any policy area (Conley Citation2001) was significant and negative, as expected, and did not affect our other results. Using the president's margin of victory, however, which arguably better captures the idea that he is advantaged in dealing with the Senate, was insignificant, counter to the finding of DeRouen, Peake, and Ward (2005) Date for DeRouen, Peake, and Ward is 2005 in references. 23. Here we measured New Chief Justice according to whether the nominee is new to the Court. If we instead include both new nominees to the Court and also sitting associates who are nominated to the Chief Justice slot, the coefficient remains negative and insignificant. John Roberts was a particularly difficult case to code. Roberts was initially nominated in July of 2010 as an Associate Justice to replace Sandra Day O’Connor. However, President Bush withdrew this nomination and re-nominated Roberts to the Chief Justice position following the death of William Rehnquist in September. Roberts was confirmed later that month. Since Roberts was initially nominated to the Associate Justice position, we use the earlier date and do not count him as a New Chief Justice appointment. However, coding the case as a 1 does not change the results. 24. The Minority variable controlled for all racial minority nominees, including Sonia Sotomayor, Clarence Thomas, and Thurgood Marshall. The Female variable controlled for all five women who have been nominated to the Court: O’Connor, Ginsberg, Miers, Sotomayor, and Kagan. While these measures are somewhat problematic due to the small n, there is no way around this problem and the inclusion of these variables is theoretically and (at least in the case of the Female variable) statistically justified.