It is often taken as axiomatic that an independent judiciary is crucial to a balanced government of separated powers. A truly independent judiciary is free from the “passions” of short-term majorities and can act to protect the interests of electoral and ethnic minorities. In effect, an independent judiciary, although potentially anti-majoritarian, stands ready to protect and defend bedrock democratic principles such as free speech and equal participation in government. The justices of the United States Supreme Court, sitting at the peak of the federal judicial hierarchy, must be responsible for vindicating these principles, even against threats from the government.
Article III of the Constitution creates a judiciary that appears equal to these responsibilities. By its provisions, the judiciary enjoys significant independence compared with historical precursors or with parallel contemporary institutions. Article III prescribes that the justices of the Supreme Court and those holding all other inferior judicial positions created by Congress “shall hold their Offices during good Behavior, and shall, at stated Times, receive for their Services a Compensation which shall not be diminished during their Continuance in Office” (United States Constitution, Article III, Section 1). That is, the Supreme Court and all other Article III federal courts have lifetime tenure and face no realistic threat of retaliation for unpopular rulings or opinions from Congress, the president or the public.
While there has been research and debate as to the extent of judicial independence enjoyed by Article III courts (Rosenberg 1992; but see Segal and Spaeth 2002, chapter 10), there is little debate that the federal courts in the United States enjoy a high degree of independence, especially compared to courts of many other countries (Howard and Carey 2002). However, there is one possible exception to this state of judicial independence—the judge sitting as a recess appointee.
Article II gives the executive the power to “fill up all Vacancies that may happen during the Recess of the Senate, by granting Commissions which shall expire at the End of their next Session.” Included in these appointees is the federal judiciary. Thus, the president can use the recess power to bypass the requirement that the Senate confirm appointees. The original purpose of the Recess Appointment Clause is uncontroversial—it allows the executive to keep the operations of government running even when the Senate is not in session. However, presidents now make recess appointments even during relatively brief intrasession breaks of the Senate.
In this chapter, we examine the situation of Supreme Court justices. We explore this issue by examining judicial recess appointees who have later been confirmed by the Senate to full time Article III judicial positions. Specifically we compare the behavior of three recess-appointed Supreme Court justices during their temporary appointment tenure with a similar period following Senate confirmation.
Judicial independence is just one facet of a larger issue concerning the role of the judiciary within the constitutional system of separated powers. Each branch has overlapping responsibilities and power subject to control and review by the other branches of government. Although this was undoubtedly a deliberate design by the framers to prevent the aggregation of power by any single branch of government, the separation of powers system leads to battles between the courts, Congress, and the executive. The indispensible actor in the exercise of the courts’ responsibilities as a coordinate branch under the Constitution is the United States Supreme Court.
Substantial literature has grown around inter-branch politics, producing studies of the branches as political institutions and analyzing how these institutions constrain, define, and mold political behavior. Scholars working within a rational choice framework often view these interactions as a game with a series of sequential moves by each branch of government and have shown that the policy choice of a governmental actor or institution results from not just from the policy preference of the actor but also a calculation as to the reactions of other actors and institutions (Vanberg 2000; Epstein and Knight 1998; Ferejohn and Shipan 1990; Eskridge 1991; Rogers 2001; Segal 1997).
Walter Murphy’s Elements of Judicial Strategy (1964) is widely recognized as an important early examination of strategic Supreme Court decision making. Murphy details the strategy of the opinion writing process, and judicial anticipation of how other political actors would respond to their decisions. In addition to anecdotal evidence, Murphy utilized Court memos and draft opinions in order to explore the various ways justices attempted to actualize their policy preferences given the complicated and intricate web of constraints presented by the American system.
Building on this and other works, Epstein and Knight (1998) contend that strategic considerations lie at the core of judicial voting. One part of their analysis focused on two pairs of famous cases with similar facts yet opposite outcomes. The first pair they examined was the Civil War habeas cases of Ex Parte Milligan (1866) and Ex Parte McCardle (1869). The second pair was the cold war congressional subpoena cases of Watkins v. United States (1957) and Barenblatt v. United States (1959). In both pairs of cases the Court supported the right of the individual in the first case, but was challenged by Congress and backed the government’s position in the second. Following Milligan, Congress stripped the courts of jurisdiction to hear these types of cases, while after Watkins, Congress threatened similar jurisdiction stripping legislation. Epstein and Knight argue that the justices changed from their sincere preference of supporting individual rights to backing the government because of the pressure from Congress that threatened the independence of the Court. These examples, chosen for their expository clarity, also highlight the normative issues raised by circumstances that threaten the independence of the judiciary.
With the exception of John Tyler and William Henry Harrison (who died in office after only 30 days) all presidents in the eighteenth and nineteenth centuries used the recess appointment power. For example, George Washington made nine recess appointments, including the appointment of three judges to the Federal District Court during the recess of the First Congress. Presidents have made over three hundred recess appointments to the federal judiciary and fifteen Supreme Court justices have initially been seated through the recess appointment power. We list these justices in Table 3.1.
As we can see, the Recess Appointment Clause has been invoked to appoint justices with some regularity, particularly in the early days of the Republic. George Washington exercised its authority three times to appoint justices to the Supreme Court, although confirmation failed in the case of John Rutledge. John Adams, Thomas Jefferson, and John Quincy Adams all appointed Supreme Court justices while the Senate was out of session. Several justices usually ranked as great or near great received their initial appointments through the recess power. Included among these are Justices Oliver Wendell Holmes, the first Justice John Harlan, Earl Warren, and William Brennan (Pederson and Provizer 2003).
After the recess appointment of Oliver Wendell Holmes by Theodore Roosevelt—an appointment Roosevelt would come to regret1 (Abraham 1999)—the practice of appointing Supreme Court justices temporarily without Senate action entered a moratorium lasting for 51 years (Stanford Law Review 1957). However, President Dwight D. Eisenhower broke with this unwritten protocol and appointed three justices to the United States Supreme Court, to wit, Chief Justice Earl Warren and Associate Justices William Brennan and Potter Stewart.
Table 3.1: Supreme Court Recess Appointments
Judicial recess appointments are not innovative or untraditional uses of the president’s constitutional authority, but they are controversial and, within the context of modern congressional-executive relations, problematic. Controversies over judicial recess appointments are not an entirely new phenomenon. One of President Washington’s recess appointments was of John Rutledge to chief justice of the United States, igniting a controversy, which, according to Curtis (1984), contributed to his rejection by the Senate.
Uses of the recess appointment power in recent decades have triggered substantial debate about the propriety and wisdom of the exercise of the recess appointment power. Following President Eisenhower’s third placement of a Supreme Court justice during a recess, the Senate passed a resolution opposing the practice.2 The resolution stated that recess appointments to the Supreme Court should be used only “under unusual and urgent circumstances” and the corresponding report from the Judiciary Committee asserted that such appointments obstructed the Senate’s “solemn constitutional task” of providing advice and consent.3 Despite the institutional prerogatives asserted, however, the resolution passed on a primarily partisan vote, 48 to 37 (Fisher 2005).
While the United States Senate’s primary concern with the recess appointment of Supreme Court justices might be protecting institutional prerogatives and power, a recess appointment suggests other normative concerns about judicial voting and behavior. Specifically, there is the potential for an adverse impact on judicial independence. Judicial independence is an essential component of our legal system. Its goal is impartial, “law-based” decision making by judges, decisions made without regard for the political preferences of members of the other branches. “The judiciary,” Gerald Rosenberg writes, “is independent to the extent its decision-making is free from domination by the preferences of elected officials” (Rosenberg 1992, 371). Others define judicial independence as “a condition in which judges are entirely free from negative consequences for their decisions on the bench,” (Baum 2003) or “the ability of the individual judge to make decisions based on the facts and the law without undue influence or interference” (Zemans 1999).
Increasingly, unpopular decisions have led to calls for greater political control over the judiciary, leading to a debate over the proper balance between judicial independence and judicial accountability. Recently United States District Court Judge Harold Baer generated both a suggestion from the president that he consider resigning and a presidential candidate’s call for his impeachment when he granted a suppression motion (Van Natta, Jr. 1996). In response, Chief Judge John O. Newman of the Second Circuit Court of Appeals and three former chief judges issued a joint statement, asserting that “when a judge is threatened with a call for resignation or impeachment because of disagreement with a ruling, the entire process of orderly resolution of legal disputes is undermined” (Van Natta, Jr. quoting Newman 1996).
In part, judicial independence is a function of institutional structures. The Constitution creates a federal judiciary as a separate and co-equal branch, a branch given the power of judicial review. Article III of the United States Constitution preserves the independence of judges in their decision making process by guaranteeing a lifetime appointment and salaries that may not be diminished. In theory, this independence, plus the power of judicial review, allows the judiciary to “stand as the ultimate guardians of our fundamental rights” (Horsky 1958, 1111). Thus, judicial independence allows judges to uphold minority rights when they are under attack. However, Ferejohn (1999) suggests that the independence of the judiciary is threatened if Congress and the president are ideologically unified and the judiciary is comprised of judges with a different ideology. In addition, Rosenberg (1999) examines nine periods of intense Court-curbing activity along with the Supreme Court decisions during those periods. He finds that in only three of the nine periods was the Supreme Court clearly independent of congressional preferences, rejecting the notion that independence is a status quo.
In fact, judicial independence lay at the heart of two court cases challenging recess appointments, one by President Jimmy Carter and the most recent recess appointment of William Pryor by George W. Bush. On January 1, 1981, President Jimmy Carter recess appointed Walter Heen to the U.S. District Court for the District of Hawaii. Judge Heen then presided over the trial and conviction of Janet Woodley on drug charges. In United States v. Woodley (1982) the Ninth Circuit analyzed the inherent tension between the President’s recess appointment power under Article II which gives the Executive the power to “fill up all Vacancies that may happen during the Recess of the Senate, by granting Commissions which shall expire at the End of their next Session” and the attributes of judicial independence incorporated into Article III. It held that “only those judges enjoying Article III protections may exercise the judicial power of the United States” and vacated the lower court decision. According to the court, “[a] judge receiving his commission under the Recess Appointment Clause may be called upon to make politically charged decisions while his nomination awaits approval by popularly elected officials. Such a judge will scarcely be oblivious to the effect his decision may have on the vote of these officials.” (726 F. 2nd 1328, 1330) When a recess appointee hears a case, he or she does not have a permanent appointment and whether the appointee receives tenure is still contingent upon renomination by the president and confirmation by the Senate (Mayton 2004).
As we noted in our introduction, the Ninth Circuit then reheard the case en banc and reversed. This was not the first time the idea was broached that unconfirmed judges might be subject to unacceptable political pressures. At Justice Stewart’s confirmation hearing, the dissent to his confirmation argued that the right to a trial before a life-appointed judge is just as much a constitutional right as the right of trial by jury in criminal cases (Mayton 2004). In fact, after the Stewart confirmation hearings, the Senate Judiciary Committee submitted a resolution to the Senate, discouraging recess appointments to the Supreme Court.4 Specifically, there was concern about the propriety of a judge hearing cases prior to confirmation.
A litigant “may for the rest of his life wonder” whether his case was influenced, consciously, or unconsciously, by the temporary judge’s knowledge that he must please the President, so that he does withdraw the judge’s nomination, and the Senate, so that it confirms him. The judge might bend to these pressures or he might do the opposite; he “might rear back and bend the other way to prove that we was subservient to neither branch (Mayton 2004, 538).
Of course these concerns lead to a simple, yet important question. Do judicial recess appointees behave differently during the recess appointment than they do as fully independent judges? Are the complaints of litigants and the fears of court observers that diminishing the constitutional protections of judges’ independence compromises the exercise of judicial power justified?
There is evidence that judicial independence is comparatively diminished without the protections given by the United States Constitution. For example, Salzberger and Fenn (1999) show that judges on the English Court of Appeals who consistently take antigovernment positions are less likely to be promoted to the Judicial Committee of the House of Lords, the highest judicial venue in England, than lord justices of appeal who are less antigovernment. Similarly, Ramseyer (1994) and Ramseyer and Rasmusen (1997) show that antigovernment judges in Japan experience less successful and more unpleasant careers than do pro-government judges.
In the United States the other branches have control over jurisdiction, appointment, enforcement of court rulings, appropriations for the operation of the courts, and can impeach judges. Thus, there are many ways in which Congress and the president can exert influence over the federal judiciary. For example, legislation restricting judicial review of habeas corpus petitions, sentencing guidelines that include mandatory minimums that severely limit judicial discretion and the Civil Justice Reform Act of 1990 that imposed oversight by district level committees and mandated alternative resolution programs has been adopted (Zemans 1999).
At the state level, the opportunities for influence are even greater given the different election systems. In a series of articles, Hall (1987, 1992) and Brace and Hall (1995, 1997) have shown state high court vote choice can depend on party competition, electoral consequences, and citizen ideology. In short, judges who must run in retention elections and for reelection exhibit different voting patterns than appointed judges who thus have greater independence from the consequences of their decisions.
Thus, the process by which judges are appointed and retained can have an impact on judicial independence and accountability. When presidents and Senators subject judicial nominees to ideological “litmus tests,” they are routinely accused of undermining the nominees’ independence, while the presidents and Senators assert that asking about a nominee’s “judicial philosophy” is a way to gauge whether the nominees are acceptable (Geyh 2003).
Similar patterns exist for recess appointees. Jefferson B. Fordham, dean of the University of Pennsylvania Law School, remarked that a recess appointee “is serving under the overhang of Senate consideration of a nomination, which is not in harmony with the constitutional policy of judiciary independence.”5 Herz (2005, 450) argues that “circumstances put the recess appointee in something of the same position as a law professor on a ‘look-see visit’; his or her job becomes one extended interview. These circumstances are utterly at odds with the commitment to judicial independence reflected in Article III’s good behavior clause and salary protections.”
While recess appointed justices will not have to confront the electorate, existing research on court nominations, judicial behavior, and judicial voting provides insight that buttresses the assertion noted above. Assuming that a recess appointee wants confirmation, the “electorate” that the nominee would care about most is the Senate. Research has shown that the median ideology of the Senate (Moraski and Shipan 1999) or even the partisan make-up of the Senate (Epstein and Segal 2005) can be critical to confirmation of the nominee as the president and Senate clash over the nominee, each seeking an ideological advantage in the separation of powers struggles (Yates and Whitford 1998). The more liberal or Democratic the Senate the greater is the likelihood that the recess appointee would vote in a more liberal direction. One might also see a liberal or conservative recess appointee retreat to the ideological center.
Ideological voting is not the only way a recess-appointed justice could assuage Senate concerns. He would also likely avoid controversial rulings or behavior that would lead to unwanted attention. This could include avoiding separate opinion writing or controversial votes, such as overturning precedent or state or federal law, or ruling against the state or federal government. We also anticipate that the recess appointee would also be more likely to vote with the majority during the period of their recess appointment.
Research on other courts confirms some of these suspicions. For example, research on appellate court behavior shows that district court judges sitting by designation on the appellate court are less likely to issue separate opinions or otherwise engage in more dissensus oriented behavior (Hettinger et al. 2006). In addition, a recess appointee might not want to alienate a senator by joining a decision to limit state or federal authority.
In the post-War era, three Supreme Court justices have received recess appointments, Chief Justice Earl Warren, Justice William Brennan, and Justice Potter Stewart, each appointed in October of 1953, 1956, and 1958, respectively. Each sat on the bench for five to seven months before receiving confirmation by the Senate. Before Chief Justice Warren, the most recent recess appointment to the Supreme Court was of Oliver Wendell Holmes in 1902, well before the earliest point at which case-vote level data for the Supreme Court are readily available. In order to isolate the possible effects of sitting via temporary appointment, we wish to compare the behavior of these justices under this condition with an appropriate benchmark, similar behavior without the condition.
Warren was a dominant political figure in California politics with his political positions moving from traditional Republican views to a prominent position as a leader of the progressive wing of the Republican Party (Schwartz 1996). He obtained his law degree from the University of California Berkeley’s newly opened law school in 1914 and after service in the army quickly moved up the political ladder in California politics, from district attorney to attorney general to governor of the state, in the process was nominated by both the Democratic and Republican party in 1946 and easily beating Franklin D. Roosevelt’s son James in the 1950 gubernatorial election.
An aspirant for the presidential nomination in both 1948 and 1952, Warren reportedly threw his support to Eisenhower over that of Senator Robert A. Taft of Ohio and for that was promised the first vacancy to occur on the United States Supreme Court (Schwartz 1996, 438), although Abraham dismisses this story and argues that Warren’s appointment to the Court was as much to remove him from California politics than any deal to secure the presidency for Eisenhower (Abraham 1999, 192-193).
With the sudden death of Chief Justice Vinson in September of 1953 a vacancy opened up on the Supreme Court and Eisenhower used the recess power to quickly fill the position less than one month later with Earl Warren. Although eventually confirmed to the position by a unanimous voice vote in March, 1954, Warren’s recess nomination led to some initial opposition by Senator William Langer of North Dakota, who apparently, was opposed to the nomination of anyone not from North Dakota (Abraham 1999, 194) and some Southern Democrats concerned over Warren’s liberal views. Thus Warren was on the bench for several months while awaiting confirmation.
Brennan’s appointment, like that of Earl Warren, also had the appearance of political considerations. Brennan, the son of working class Irish immigrants grew up in a large Catholic family in Newark, New Jersey. His father became active in local Democratic politics and eventually ran and won office to the Newark Board of Commissioners with Brennan helping out in the campaign.
Despite his father’s desire to see his son in a business career, William Brennan opted to enroll in Harvard Law School after graduating from the University of Pennsylvania. Following graduation and after a stint in the Army during World War II, Brennan practiced law before being named to the Superior Court of New Jersey by Republican Governor Alfred Driscoll in 1949, following successful efforts to help reform the New Jersey court system and establishing a pattern of Republican political support for his judicial aspirations. For example, soon thereafter, Brennan became an ally, trusted lieutenant, and eventually right-hand man to the Republican chief justice of the New Jersey Supreme Court, Arthur Vanderbilt. By 1952 with Vanderbilt’s backing, Brennan was appointed to the New Jersey high court.
Following Sherman Minton’s death in April 1956, Eisenhower had an opening for an Associate Justice on the United States Supreme Court. Eisenhower sought to quickly fill the position and had political considerations in mind as he approached reelection for his second term. Brennan, a northeastern Catholic Democrat, offered several politically appealing traits to Eisenhower’s reelection efforts and met the state criteria of relative youth, judicial experience, and high state bar standing (Abraham 1999). Vanderbilt strongly supported Brennan’s nomination. Thus Eisenhower nominated Brennan during an October Senate recess, shortly before the presidential election in November, and formally nominated him with the congressional session of January 1957, and was confirmed two months later on a voice vote with the only dissent coming from Senator Joseph McCarthy (Wermiel 1996).
Stewart was the fifth and last nomination of President Eisenhower. Stewart fits the traditional mold of a politically connected nominee. Born into a well-to-do family prominent in both law and Ohio Republican politics, Stewart enjoyed an affluent childhood and privileged education that included a prestigious prep school, The Hotchkiss School in Lakeville, Connecticut. While the Great Depression brought financial hardship to his family, Stewart was able to secure scholarship aid to finish prep school and enroll at Yale. Following a year of study abroad, Stewart entered Yale Law School, eventually becoming editor of the Yale Law Journal.
Following corporate practice and a stint in the Navy during World War II, Stewart entered into politics and supported Eisenhower’s 1952 presidential bid over that of Ohio native son, Robert Taft. Upon the recommendation of Ohio Republican Senator John W. Bricker, Stewart was appointed by President Eisenhower to the Sixth Circuit Court of Appeals in 1954. At the time he was the youngest judge on the Court.
When fellow Ohioan Harold H. Burton retired on October 13, 1958, due to ill health, Eisenhower immediately filled the vacancy with the recess appointment of Potter Stewart. Although Stewart’s qualifications were never an issue, the nomination did run into some trouble with Southern Democrats concerned over Stewart’s obvious commitment to civil rights and desegregation. Several southern senators then were able to delay the confirmation vote for several months. The judiciary committee eventually reported favorably on the nomination by a twelve to three vote, with three Southern Democratic senators dissenting. The full Senate confirmed Stewart by a vote of seventy to seventeen in May of 1959.
We specify several hypotheses about several political and institutional factors that we expect to have an impact on the behavior of these justices before confirmation and afterward. Our first set of hypotheses relate to how likely the justices are to cast liberal votes in given circumstances.
Hypothesis 1: A recess appointee will be less likely to cast liberal votes before Senate confirmation.
Hypothesis 2: A recess appointee will be more likely to cast liberal votes when the Senate is controlled by the Democratic Party before Senate confirmation.
Hypothesis 3: A recess appointee will be less likely to cast liberal votes in cases that are publicly salient before Senate confirmation.
Hypothesis 4: A recess appointee will be more likely to cast liberal votes as the number of other justices in a liberal majority coalition increases before Senate confirmation.
We stipulate a second set of hypotheses concerning the propensity of the justices to cast votes to affirm based on several conditions of the cases and the institutional context, pre- and post-confirmation.
Hypothesis 5: A recess appointee will be less likely to cast votes to affirm when the United States appears as the petitioner/appellant before Senate confirmation.
Hypothesis 6: A recess appointee will be less likely to cast votes to affirm when a state appears as the petitioner/appellant before Senate confirmation.
Hypothesis 7: A recess appointee will be less likely to cast votes to affirm when the United States appears as the respondent/appellant before Senate confirmation.
Hypothesis 8: A recess appointee will be less likely to cast votes to affirm when a state appears as the respondent/appellant before Senate confirmation.
Hypothesis 9: A recess appointee will be more likely to cast votes to affirm when the decision the Court is reviewing affirmed the judgment of the Court below before Senate confirmation.
Hypothesis 10: A recess appointee will be more likely to cast votes to affirm as the number of other justices in an affirm majority coalition increases before Senate confirmation.
President Dwight Eisenhower appointed each of these justices, so there is no variation in appointing president among our subjects. However, all three of Eisenhower’s high court recess appointments went on to successful confirmation after spending five to seven months as an active participant in cases. Each justice’s continued tenure in the Court depended on the nomination and support of the president and the consent of the Senate during these months. Because each of these justices spent a considerable portion of their first terms as recess appointees, we can compare their behavior during their recess periods with their behavior post-confirmation.
Using the Spaeth Supreme Court Database, converted to justice as the unit of analysis by Sarah Benesh and Chris Zorn, we gathered data on each of the justices from the date they arrived on the bench to the date they were confirmed by the Senate. We also collected data from a similar number of case-votes of the same justices in the months immediately following their confirmation. In addition to case factors, the database contains indicators of the directions of the justices’ votes in ideological terms—liberal or conservative—and in terms of the disposition for the lower court’s decision—affirm or reverse.
We considered and rejected contrasting the justices’ pre-confirmation behavior with the remainder of their careers on the bench. While this would make use of as much information as is available about their performance while enjoying life tenure, we cannot be certain that votes made years after the recess period are truly comparable. Changes in the institution of the Court, the dynamics of small-group decision-making, the political context, and the legal agenda facing the justices could all change, confounding recess effects. Furthermore, several studies indicate that many justices demonstrate ideological change over time on the bench, which manifests itself in behavioral changes (Epstein and Knight 1998; Martin and Quinn 2002). Finally, many court scholars find that certain justices are prone to acclimation effects, changes in their behavior as they adjust to their new role and setting, although such periods of adjustment are thought to take a full term or two before noticeable differences appear (Hagle 1993; Hurwitz and Stefko 2004). This issue in particular cautions against casting too far from the recess period for comparable observations. Votes were considered “before” or “after” confirmation depending on the report date for the vote, rather than the conference date, because for several cases exact conference date information was not available and justices are always free to change their votes until the report date (Howard 1968).
For Warren and Brennan, the inclusion of the remainder of their initial terms on the bench produced as many or more post-confirmation votes. Justice Stewart, however, was not confirmed until rather late in his first term of service. In the 1958 term, Stewart cast 75 votes before confirmation and only 46 votes following. To estimate pre- and post-confirmation behavior for Stewart with comparable precision, we included several decision days from the 1959 term, adding 25 votes to nearly equalize the two sets (the latest votes for Stewart were reported January 10th, 1960). The periods and number of votes for each justice are summarized in Table 3.2.
Our data come from three justices appointed by a Republican president, albeit one who by reputation did not make ideology a high priority for judicial appointments (Epstein and Segal 2005). Two of those appointments faced a very closely divided Senate, only one controlled by the president’s party, while the third justice’s nomination (Stewart) was considered by a newly elected and substantial Democratic majority. To gain a better understanding of the extent to which being a recess appointee affects the behavior of the justices, we analyze the impact of political and case factors on their voting before and after their confirmation.
Justices on the Supreme Court are expected to be relatively free of external influences, free to carry out the judicial function, address fundamental questions of law and policy, and confront the political branches when necessary despite controversy. We suspect, however, that when justices are recess appointees, they will be less inclined to take actions or positions that expose them to public attention or criticism from within the executive or legislative branches. To what factors, we ask, do justices respond when sitting temporarily that they would feel free to disregard with the security of life tenure? In order to address this question, we specified several maximum likelihood models of justices’ votes, comparing the influences of various factors before and after the justices are confirmed.
Table 3.2: Eisenhower Supreme Court Justices Seated via Recess Appointment
Descriptive statistics for the variables used in our models appear in Tables 3.3 and 3.4. We analyze the justices’ tendencies toward liberal voting and their choices to affirm or reverse. Other variables we use include the size of the liberal voting majority in the case, if the case is decided liberally, whether the federal government or a state appear in the case as petitioners or respondents, and the ideological direction of the lower court, as well as an indicator of whether or not the Democrats controlled the Senate at the time. Table 3.3 breaks down a set of case and contextual factor indicators across the confirmation line. Because most of the pre-confirmation votes come from relatively early in the terms and most of the post-confirmation votes come from their last months, we observe some differences. Publicly salient cases, measured by coverage in the New York Times, are more common late in the term and are thus more frequent after confirmation. We cannot discount the possibility that announcement of some of these cases were delayed until after recessed justices were confirmed. Most of the other variables are distributed similarly across samples, differing only by small fractions of their sizes.
The next table, 3.4, presents the same variables across justices, or terms. Again, while we find some differences across terms, none that generate reliability or validity concerns. Media coverage appears to be constant, but the size of the successful liberal voting blocs increase from Warren’s initial term to Brennan and Stewart’s. We detect some differences in the appearances of government litigants from term to term, as the United States is seen more often as respondent and less as petitioner in 1956 than three terms earlier. The trend for states is consistently upward. States are more frequently petitioners and respondents in cases late in the decade. None of the differences are extraordinary, however.
Table 3.3: Descriptive Statistics – Pooled, Before, and After Confirmation
Table 3.4: Descriptive Statistics – by Justice/Term
Our first analysis is of liberal-conservative voting, as coded in the Spaeth Database. The dependent variable, coded one for a liberal vote and zero for a conservative vote, is dichotomous, thus we use a logit model. We fashioned hypotheses about the expected direction of various influences on Supreme Court justices before and after confirmation, but what we are really interested in is whether the influences of various factors on justices’ votes are different from each other before and after confirmation. Thus, our approach is to specify a model of justices’ votes that captures the potential impact of external influences or undesirable attention and test whether the coefficients estimated from that specification differs across groups of votes: those made during the recess appointment period and those made after Senate confirmation.
A standard approach to testing whether the same coefficients govern the relationships between a set of covariates and the outcome across two different groups of observations is to conduct a Chow test (Chow 1960). The F-distributed Chow test is often referred to as a test of the “pooling assumption” or a test of “structural change,” and is equivalent to an F-test of a fully interactive model pooling both sets of data. Unfortunately, as Allison (1999) notes, in discrete choice models such as logit, the validity of hypothesis tests using variables interacted across different groups of data relies on the assumption that the residual variation in data is equivalent between groups. Further studies have revealed that if this assumption is violated, naïve hypothesis tests of such interactive effects and the Chow test can give misleading results, perhaps even contrary to the actual relationships in the data (Hoetker 2007).
Heterogeneous choice models allow the variance of discrete choice models to vary as a function of a set of specified covariates, which can be indicators of group membership or other characteristics of the observations (Williams 2006). In political science, the heteroskedastic probit model (Alvarez and Brehm 1995) is probably the most widely familiar variant of this family of models. Specifying a variance function can make outcome coefficients robust in the face of non-constant residual variation, as is the case here, or because variance is itself of substantive interest (as was the case in Alvarez and Brehm’s application.) In the models we present, we find significantly non-constant residual variation across confirmation status and present results from heterogeneous logit models that take account of those differences. We do not report the coefficients for the variance function as it is not the focus of our interest.
We wish to capture the conditional effects of these variables—their influences on vote choices before confirmation and following confirmation. Thus, we specify conditional models, estimating separate coefficients for the variables in each condition. We employ two complementary strategies: a “multiplicative interaction” model and a “regime” model. The former is the traditional “interactive” model widely used in social sciences. Variables whose effects are thought to depend on the value of another variable are multiplied by that variable and the resulting term added to the specification. The un-multiplied terms are referred to as the “constituent” terms while the new terms are the “interactions.” The second approach is useful when the condition upon which other effects are thought to depend is discrete, as it is here. We produce the “regime” model by multiplying the constituent terms by two variables indicating whether a vote was made before or after confirmation and enter those two sets of covariates into the model in place of the constituent terms. Thus, in the regime model, one set of covariates take their observed values for the before confirmation votes and are zero otherwise, while the other set take their observed values only after confirmation. The regime model allows us to estimate separate marginal effects on the outcome for the covariates across conditions (before and after confirmation) without the complications of the traditional interaction model (Brambor, Clark, and Golder 2006, 69).
As a basis for comparison, we estimated a logit model pooling the pre- and post-confirmation votes, allowing only the constant to vary across these conditions. The standard errors for this and the two subsequent models are heteroskedasticity-robust, clustered on the individual justices. The results appear in the first column of Table 3.5.
Both intercepts are significant and negative, suggesting that the justices did not tend to cast liberal votes, conditional on the other factors. Senate control by the Democrats and public salience are both not statistically significant, but the proportion of the Court joining a liberal majority is significant and appears substantively large. Perhaps unsurprisingly, these justices were more likely to cast liberal votes when their colleagues did so.
The first conditional model is reported in the second column of results. The pre-confirmation intercept remains significant and negative, although the intercept for post-confirmation votes is not significant and much smaller. With other factors accounted for across conditions, it seems that the justices were not generally inclined against liberal votes after they were confirmed. The liberal majority size coefficient is virtually unchanged from the pooled model, but public salience, operationalized by appearance of the case on the front page of the New York Times, is significant and negative. Before confirmation, recess-appointed justices are less likely to cast liberal votes when the Court’s decision is highly newsworthy.
Table 3.5: Heterogeneous Logit Models of Liberal Vote
In this model, the interactive terms represent the departures from the estimated constituent effects for votes cast after confirmation rather than marginal effects. Interpreting the estimated effects of changes in the variables on liberal voting requires taking account of both estimated effects, and assessment of whether those effects are statistically significant requires calculation of compound standard errors taking account of the covariance of the estimated effects. If anything, the coefficients and test statistics of interaction terms might reveal whether the marginal effects of the constituent terms are significantly conditional on each other, although perhaps not even that (Brambor et al. 2006, 74).
The third column of results contains the coefficients of the “regime” specification. In this model, the estimated effects of the post-confirmation variables are marginal effects, rather than interactive departures from the constituent effects. There are observable differences between each of the coefficients and these differences are very revealing. The intercepts are noticeably distinct, significantly so, and the pre-confirmation baseline is substantially more negative, indicating that the justices were considerably more likely to cast liberal votes after they were confirmed. The variable indicating a Senate controlled by the Democrats is not significant in the pre-confirmation specification, but the post-confirmation effect is significant and negative, although its substantive impact is not great. Consistent with the interaction results, the coefficients for public salience are profoundly dissimilar. Pre-confirmation, in cases salience to the public, measured by being reported on the front page of the New York Times, are significantly less likely to prompt a liberal vote. Post-confirmation, however, this effect is insignificant and vanishingly small.
To demonstrate the impact of public salience on liberal voting before and after the justices receive permanent tenure, Figure 3.1 shows the changes in the predicted probability of a liberal vote for non-salient and salient cases across confirmation status.
All other variables in the model are fixed at their central values, median or mean. The leftmost bar shows the likelihood of a liberal vote before confirmation for a “typical” non-salient case, which is above .6, or 60 percent. A salient case, pre-confirmation, has a substantially lower probability of eliciting a liberal vote, less than .2. The two bars corresponding to vote probabilities post-confirmation, for salient and non-salient cases, are practically identical, however.
The size of the voting majority in liberally decided cases is also statistically significant pre-confirmation, but insignificant following confirmation. To demonstrate this effect visually, Figure 3.2 graphs the predicted likelihood of a liberal vote across a range of values for the size of the majority in a case decided liberally. Before confirmation, the predicted probability increases from about 65 percent when the number of liberal votes cast by the other justices is equal to 4 (some cases in our dataset are decided by a less-than-full Court, meaning that a case receiving only four liberal votes could be decided liberally even without the vote of the recess-appointed justice). As the size of the liberal majority increases, however, the likelihood of the recess-appointed justice casting a liberal vote increases as well, approaching 100 percent when all of the justice’s fellows vote. In contrast, recess appointed justices after confirmation are virtually unaffected by the number of other justices on the Court casting liberal votes. In addition to our analysis of liberal voting, we also conducted an examination of voting to affirm cases before the Court by the justices receiving recess appointments. The results of this analysis appear in Table 3.6. As before, the estimates are adjusted for non-constant residual variation.
Table 3.6: Heterogeneous Logit Models of Affirm Vote
Following the same pattern of our previous analysis, the first column of Table 3.6 contains the results of a model pooling all of the votes of our three justices. Only one of the covariates has a statistically significant effect, although the separate intercept for votes cast after confirmation is significant and negative. Justices Warren, Brennan, and Stewart were more likely to vote to reverse after they were confirmed. Also, the number of justices in majority coalitions voting to affirm a case is also significantly related to the justices’ choice to cast an affirm vote.
The second column presents the results of a model interacting the affirm vote specification with the indicator of post-confirmation votes. Among the constituent terms, the presence of the U.S. as a respondent is significant and positive. The interpretation of this finding is that cases in which the United States is defending a favorable decision in the lower court, the justices were observably more likely to cast a vote to affirm the lower court’s decision. In other words, while the justices were awaiting action on their nominations by the political branches, they were significantly less likely to vote against the United States when appearing before the Court against an appeal. Also, the justices were significantly more likely to vote to affirm when their brethren did so. The multiplicative interactive terms for each of these variables are also significant and negative. Moreover, the size of the coefficients is roughly equal to the size of the constituent terms. The suggestion of these results is that effects present before confirmation largely disappear following Senate approval.
The third column in Table 3.6 follows the same strategy as in Table 3.5, estimating the separate effects of the variables before and after confirmation. As is to be expected, the results for the pre-confirmation votes are the same as for the constituent terms in the second column. The post-confirmation coefficients, however, yield no reliable estimated effects. Thus, while we can infer a relationship between the U.S. respondent and votes to affirm variables and justices’ decisions to cast affirm votes while sitting via recess appointment, no such relationships appear to exist subsequently.
The effects of these two variables are presented graphically in Figures 3.3 and 3.4.
The bar graph in Figure 3.3 shows the probabilities of Warren, Brennan, or Stewart voting to affirm a lower court decision before and after confirmation, varying the position of the United States in the case. Figure 3.4 shows the same information in linear format. For cases in which the United States appeared before the Court as a petitioner, seeking to reverse a lower court decision, the justices’ predicted probability of voting to affirm was about 25 percent during the recess appointment period, while the estimated likelihood of such a vote when the United States appeared to defend the judgment of the court below is over 40 percent. Following confirmation, the effects of the United States’ appearances before the Court were seemingly irrelevant to the justices. Both conditions yield predicted probabilities at almost exactly 50 percent.
The findings of our analyses indicate that there are differences in the voting records of these Supreme Court justices pre- and post-confirmation. These differences also provide some basis for the sense expressed by the Senate that recess appointments to the Supreme Court are unwise. Our study of ideological voting by the three recess-appointed justices of the Eisenhower administration, Warren, Brennan, and Stewart, reveals that the direction of the justices’ votes appeared to be influenced by the salience of the case to the public and the degree of consensus on the Court, but only during their temporary tenures. Following permanent appointment, the estimated effect of these factors disappears.
Given that recess appointees must still receive nomination from the president and be confirmed by the Senate, they lack the independence accorded by the tenure and pay protections of Article III of the United States Constitution. The American system of separated powers is reliant on a vigorous judiciary with the ability and will to check government encroachment on individual liberties and rights, but a recess appointee must be mindful of the Senate, strategically choosing to avoid controversial rulings and behavior that would lead to unwanted attention. Rather than acting sincerely with the protection of life time tenure, the recess appointee must keep “one eye over his shoulder on Congress.”
Each of the coordinate branches has overlapping responsibilities and powers subject to control and review by the other branches of government. Although Congress does have some limited checks on the judiciary post-confirmation, and there is evidence that threat of congressional retaliation against the judiciary can constrain the choices of a judge, our analyses indicate that a recess appointee sits in a more precarious position, and, consequently, is less able to fulfill the responsibilities of the judiciary. Specifically, the justices in our study were significantly less likely during their temporary tenure to cast liberal votes in publicly salient cases or when fewer of their colleagues are not voting similarly. Our indicator of public salience, coverage on the front page of the New York Times, is likely to identify cases in which the justices would need to cast votes protecting the rights of individuals against government violation, a standard definition of a liberal vote. We also discovered that justices were less likely to vote against the government appearing to defend a favorable lower court judgment during their temporary commissions.
Justices sitting via recess appointment are in a different position than confirmed justices. Our results raise serious concerns about the validity of permitting recess-appointed justices to exercise judicial power. Litigants cannot be certain that the decisions in their cases are made without strategic considerations or inappropriate influences by the political branches affecting the votes of the justices if those justices lack permanent appointment. More importantly, the inferences from our analyses question the presumption that the judicial branch is capable of fulfilling the role many statesmen and scholars attribute to it. We assume that life-tenured justices are more insulated from external influences than those who lack it, but even the protections of Article III cannot ensure that sitting judges are entirely immune to being influenced by external pressures. Epstein and Knight (1998) describe several examples of circumstances in which the interests of the Court as a whole, or the justices collectively, arguably retreat from legal positions due to external pressure. They conclude that in these circumstances, justices are not averse to adjusting their decisions to reduce or avoid confrontation or threats to their institutional interests. We can more easily see the basis for justices’ vulnerability to external influence in the admittedly rare recess appointment cases, but this is far from the only circumstance in which justices’ interests are conditional on the favor of other actors.
Additional research on other recess appointees to lower federal courts is necessary to determine if other judges change voting pre- and post-confirmation. Unfortunately, we do not think it possible to combine that information with the analyses presented herein. Given the discretionary docket of the Supreme Court and other institutional factors that separate these justices from all other courts, the behavior of Supreme Court justices is unique among the federal judiciary. However, given the differences pre- and post-confirmation that we have discovered it is sobering to consider the implications that recess appointments cause for judicial independence.
1. Holmes dissented in Northern Securities v. United States (1904) opposing the breaking up a railroad trust favored by the majority. The dissenting vote prompted Roosevelt’s famous tirade of “I could carve out of a banana a judge with more backbone!” (Abraham 1999; Cushman 1996)
2. Senate Resolution 334.
3. 106 Cong. Rec. 12761 (1960); S. Rept. No. 1893, 86th Cong., 2nd Session 1-2 (1960).
4. The resolution was ultimately adopted by the Senate.
5. Testimony before Congress, 106 Cong. Rec. 18131 (1960).
