Executive Branch Socialization and Deference on the U.S. Supreme Court

Abstract

Are Supreme Court justices with prior experience in the executive branch more likely to defer to the president in separation of powers cases? While previous research has suggested that such background may signal judicial policy preferences but does not shape them, I argue here that institutional socialization may indeed increase future judicial deference to the president. Using an original data set of executive power cases decided between 1942 and 2007, I model justice-votes to test this hypothesis. I uncover three noteworthy findings: (1) a clear correlation between prior executive branch experience and support for the executive branch, (2) the degree of this support intensifies as executive branch tenure increases, a finding congruent with a socialization hypothesis, and (3) contrary to received wisdom, executive powers cases possess a clear ideological dimension, in line with the expectations of the attitudinal model.

In 2005, on the well-respected legal blog Opinio Juris, law professor Julian Ku reflected on the likelihood that then Judge Roberts would be a strong supporter of executive power once on the Supreme Court. After noting that Roberts had clerked for former Chief Justice Rehnquist, also a supporter of a robust executive branch, Ku stated that “like Jackson, who served as Attorney-General for FDR, and Rehnquist, who served as an Assistant Attorney General for Nixon, Roberts' main government experience has been in the executive branch as associate White House Counsel and Deputy Solicitor General” (Ku 2005). The implication of this statement was clear: as a former member of the executive branch, Judge Roberts was expected to be more deferential to the president in cases involving executive power.

The notion that background affects behavior might seem an obvious truth. When it comes to judicial decision-making, however, particularly for hard cases at the appellate court level, the study of social and background characteristics as systematic correlates for behavior has attenuated, thought to have fallen short on both theoretical (Sisk, Heise, and Morriss 1998) and empirical grounds (Heise 2002). The successor to the social background model has undoubtedly been the attitudinal model, which posits that such cases are mainly resolved according to the ideological policy preferences of the judges who hear them (Segal and Spaeth 1993). However, though ideological attitudes are the single best extra-legal predictor of judicial decision-making, a great deal of variance remains unexplained. Moving to fill this gap, legal scholars have provided persuasive arguments as to the role that legal doctrine (Bailey and Maltzman 2008; Richards and Kritzer 2002), strategic interaction (Epstein and Knight 1998), the desire for comity (Hettinger, Lindquist, and Martinek 2007), and even the need for approval (Baum 2006) play in explaining judicial decision-making when the law is unclear.

While not returning to their place of prominence, social background models remain useful, primarily in improving predictions of judicial decision-making where a reasonable connection can be drawn between the case area and the background in question. These studies have not only used better methods and reduced incomparability to uncover noteworthy correlations between social background factors and decision outcomes (Brudney, Schiavoni, and Merritt 1999; Schneider 2002; Sisk, Heise, and Morriss 1998), but have even successfully tested competing causal explanations (Boyd, Epstein, and Martin 2010).

In this study, I follow this more recent vein of social background studies, contending that executive branch experience has predictive power in explaining separation of powers outcomes on the Supreme Court. Specifically, I build on work by Michael Dorf, who found that in the post-Warren Court era, Republican Supreme Court appointees with executive branch experience were more likely to be consistent conservatives than those who lacked such experience (Dorf 2007). However, his analysis—which only examines recent Republican nominees and does not control for other variables—does not formally assess why this correlation holds. Does service in a presidential administration primarily signal strongly held, previously developed preferences, or does such service have an independent socialization effect? Dorf argues that a signal and recruitment theory seems more likely than a socialization process, given that attorneys without clear partisan beliefs are unlikely to be hired by a presidential administration, and that such attorneys do not demonstrate a uniformly “pro-government” pattern of decision-making.

However, I contend that executive branch experience may indeed have socialization effects when the legal issue at hand concerns presidential power, an area where ideological and partisan signals may be weaker, and where future nominees may not possess strong preexisting policy preferences. I submit that this deference to the executive can be explained by theories drawn from organizational sociology, in which members of a particular organization come to adopt its norms and goals, as well as join personal networks that influence decision-making and affect the interpretation of new information (Chao et al. 1994). As such, I expect justices with such experience to be more supportive of the executive branch and the president than those without, and for that level of support to increase as executive branch tenure lengthens and, presumably, socialization effects increase.

Testing this hypothesis on an original dataset of Supreme Court separation of powers cases ranging from 1942 to 2007, I find clear evidence that prior executive branch experience correlates with an increased likelihood of supporting the president's preferred position. Furthermore, I find that a longer tenure within the executive branch intensifies this support, a finding which supports a socialization hypothesis. Finally, and contrary to much of the received wisdom on separation of powers disputes, I find that these cases contain a clear ideological division, with conservative justices being much more likely to support the president after controlling for other factors. That executive branch experience affects decision outcomes even when ideological divisions are present only strengthens the main hypothesis.

The Social Background Model and Executive Branch Experience

Fifty years ago, the notion that variation in judicial decision-making could be driven by career or social background was an uncontroversial, though largely untested hypothesis. The hypothesis itself is simple and seems intuitively correct: “shared social and political traits reflect similar socialization processes and life experiences, which in turn produce similar attitudes and ultimately behavior (votes)” (Gryski, Main, and Dixon 1986: 528). This model's increasing popularity in the second half of the twentieth century was likely driven by a combination of factors, including the rebirth of legal realist theories of judicial behavior, the creation of empirical methods and datasets which made possible the quantitative examination of such relationships, and an increasingly diverse state and federal bench, which augmented scholarly interest on whether personal characteristics such as gender or race systematically affected judicial outcomes (Uhlman 1978; Walker and Barrow 1985), sentencing (Spohn 1990; Welch, Combs, and Gruhl 1988), or legal reasoning (Sisk, Heise, and Morriss 1998). Alongside the examination of constitutive characteristics such as race, gender, and region, scholars also examined the effect of prior career choices, in particular testing for the potential impact of prior judicial or prosecutorial experience (Nagel 1962; Tate 1981; Tate and Handberg 1991; Ulmer 1973).

As the twentieth century drew to a close, however, the scholarly focus on career background and socialization effects faded from prominence. In terms of theory, these studies often relied on relationships that were overly broad or crude, such as suggesting that minorities will side with “underdogs” across a wide range of case types. More importantly, these studies fell short in their results, often failing to find significant or substantial differences (Ashenfelter, Eisenberg, and Schwab 1995; Heise 2002). Finally, the social background model was overshadowed by the attitudinal model (Segal and Spaeth 2002), which elevated ideology above race, gender, religion, or region as the foremost extra-legal factor behind judicial-decision making at the appellate level.

Nevertheless, as the Sotomayor nomination showed in regards to her Hispanic origins, her gender, and even her status as a former prosecutor, the social background hypothesis continues to capture the imagination of pundits, publics, and politicians. Moreover, while the influence of social background models has undoubtedly diminished, they remain an important part of judicial scholarship. In particular, more recent social background studies have examined narrower slices of legal issues, a choice which not only reduces incomparability, but also allows scholars to generate more plausible theoretical relationships between independent and dependent variables. For example, these more narrowly focused studies have found that women decide cases differently than men in sex discrimination cases (Boyd, Epstein, and Martin 2010; Gryski, Main, and Dixon 1986), that judges with “elite” backgrounds rule differently than other judges on tax and NLRB adjudications (Brudney, Schiavoni, and Merritt 1999; Schneider 2002, 2005), and that judges with prior criminal defense experience had significantly different reactions to the adoption of the federal sentencing guidelines than other judges (Sisk, Heise, and Morriss 1998). As these studies show, persistent attempts to fine-tune both theoretical relationships and methodological approaches have borne fruit.

Here I examine a career background trait for Supreme Court justices that is often discussed but rarely systematically examined: prior service in the federal executive branch. The working assumption by most observers, as seen in the quote by Professor Ku, is that justices with such experience will be more likely than their fellows to defer to the president in executive power disputes. However, to date, prior executive branch experience has been used as either a control variable in separation of power studies (Ducat and Dudley 1989; Yates and Whitford 1998)—included based on an intuition regarding its likely impact rather than on any prior findings or well explicated theories—or as a predictor not of judicial votes per se, but of future ideological consistency (Baum 2006; Dorf 2007). None of these studies, note, explicated or tested a relationship between such experience and particular outcome patterns. Given that about half of Supreme Court justices since FDR have had such experience, the salience and substantive importance of executive powers cases such as Clinton v. Jones (1997) or Hamdi v. Rumsfeld (2004), and the increasing reality that the Court, rather than Congress, often serves as the final effective check on the expansion of unilateral presidential power, such a hypothesis deserves formulation and testing.

Signaling and Socialization in Separation of Powers Decisions

My study builds on recent work on executive branch experience by Baum (2006) and Dorf (2007), who examine whether executive branch experience might help solve the riddle of why some justices—like Blackmun—“drift” in their ideological leanings over time while others—like Burger—remain largely consistent (Epstein, Martin, Quinn, et al. 2007). Baum (2006) reports that Republican Supreme Court nominees new to the Washington D.C. area were more likely to drift leftwards on rights and liberties cases—after accounting for issue change by term—than were appointees who had worked in the District of Columbia prior to their nomination (143). He tentatively suggests that this finding may be driven by the pull of a liberal social environment, which he thinks would more deeply impact justices new to the Washington area.

However, Baum's “D.C. Republicans” shared another characteristic that his “D.C. Outsiders” lacked: common experience working in a Republican presidential administration. Expanding on this intuition, Michael Dorf argues that it is previous executive branch experience in the president's party, rather than D.C. insider status per se, that provides the best signal of future ideological consistency (Dorf 2007). Dorf's analysis is straightforward: using the United States Supreme Court database, he tabulates the percentage of liberal votes made by each justice appointed by a Republican president since Nixon, and aligns the justices from most liberal to most conservative. Under both Burger and Rehnquist, Dorf finds Republican appointees with prior executive branch experience are uniformly more conservative than those who lack it. Moreover, when Dorf makes justice by justice comparisons in several different issue areas (such as federalism, civil rights, etc.), his executive branch hypothesis remains fairly robust, notwithstanding exceptions such as Scalia's more liberal than expected stance on Sixth Amendment issues or O'Connor's more conservative than expected stance towards federalism cases. Using Martin-Quinn scores (Martin and Quinn 2002), another measure of judicial ideology based on judicial votes, I replicate this contrast in Figure 1, comparing the scores of Republican appointees to the Supreme Court (since 1970) who possess prior executive branch experience with those who lack it.

Figure 1. Martin-Quinn Scores for Republican Appointees to the Supreme Court, 1970–2007, Sorted by Prior Executive Branch Experience.

While this replication (like Dorf's original analysis) does not control for other variables, the contrast in Figure 1 is striking. Clearly, the heuristic that Republican-appointed justices with prior executive branch experience are more likely to be consistent conservatives on the Supreme Court is one worth considering.¹

When he considers the cause of this correlation, Dorf believes that such experience signals, rather than shapes, ideological preferences. In theory, he writes, greater ideological consistency could result from institutional socialization. However, he finds a signaling or recruitment hypothesis, in which ideologically consistent legal conservatives are more likely to work for Republican presidential administrations in the first place, to be more likely. Dorf supports his intuition by noting that he does not see a similar effect for Democratic appointees, and that there is little evidence that former government lawyers take “pro-government” positions across the board. In other words, Dorf argues, Justice Alito is not a consistent conservative vote on the Court because he previously worked in the Reagan administration; rather, he was selected to work in the Reagan administration because he was already a consistent conservative.

Dorf's arguments are reasonable, given that presidential administrations might indeed eschew moderates for ideologues in hiring. However, his argument that executive branch socialization can be rejected as a causal factor overlooks the possibility of potential socialization effects on a narrower slice of decisions. Specifically, separation of powers issues present a stronger theoretical case for the causal role of career socialization. First, unlike rights and liberties, economics, or even federalism cases, separation of powers disputes are not thought to consistently break down along standard ideological lines. In the United States Supreme Court database, for example, the ideological direction of decision in separation of powers cases such as Morrison v. Olson (1988) are coded “unspecifiable” (Spaeth et al. 2010). Such cases are thought to raise institutional rather than ideological cleavages, creating Madisonian conflicts that may sit perpendicular to the standard attitudinal axis. To take one recent example, while the policy preferences of the George W. Bush and Barack Obama administrations have little overlap, they have been fairly congruent in how they view the power of the presidency in foreign affairs. It may be that the uneasy fit between standard ideological dimensions and separation of powers battles leaves more room for other factors to affect decision-making.

Second and more importantly, there are specific sociological reasons to expect that justices with previous service in the executive branch will have greater deference towards it once on the Court. Given that the study of judicial socialization has largely attenuated among political scientists, I turn to organizational sociology for insight. A key tenant of this literature states that individuals who join a particular organization come to adopt the values, goals, and behaviors shared by other members (Chao et al. 1994). These studies demonstrate that in a relatively short period of time, members of an organization come to trust its aims, support its goals, and develop a “relational or affective bond with the organization” (Thomas and Anderson 1998: 760). Though not all of the content areas (such as adopting slang or learning organizational politics) in organizational socialization theory are obviously applicable to judicial decision-making, at least two—personal networks and organizational values—appear theoretically relevant.

First, the personal networks content area relates to the creation of “successful and satisfying work relationships” (Chao et al. 1994: 731) with other members of the executive branch. When the president selects a nominee with prior executive branch experience, he chooses someone either within his own administration or from a recent administration from his own party. In either case, justices may find it more difficult to rule against an administration composed of many of the same people with which he or she had previously worked, especially when the legal dispute directly implicates the power or standing of the organization in which they mutually served. To be sure, the pressure these bonds might place on decision-making would presumably operate at the margins, but we should not overlook that judges are neither Delphic oracles nor preference maximizing machines, but instead human beings with the same psychological needs and flaws as everyone else.

Second and more important is the organizational goals and values content area. Here research suggests that individuals tend to not only adopt the stated goals and values of the organization, but also the “unwritten, informal, tacit goals and values espoused by members who are in powerful or controlling positions” (Fisher 1986). In one such study, Kenneth Meier and Lloyd Nigro found that despite radically different social backgrounds, agency affiliation among employees in the executive branch came to outweigh all other social background variables as an explanation of political attitudes (Meier and Nigro 1976). Similarly, Robert Huckfeldt and John Sprague have shown that contextual interaction with members of particular social networks structures information relevant to political beliefs and encourages members to adopt the beliefs of the relevant majority (Huckfeldt and Sprague 1987). The executive branch can reasonably be viewed as a network of this sort.

What organizational goals and values might we expect to be socialized within the executive branch? For most issues, one would expect that such experience would reinforce preexisting ideological preferences, as presidential administrations generally hire individuals who already share their ideological worldview. For these cases, Dorf may be correct in suggesting that socialization effects have limited impact. In executive power disputes, by contrast, preferences regarding the proper balance of institutional power do not as readily divide across either ideological or partisan lines, and socialization effects may have more room to operate. In particular, I expect that time spent within the executive branch will socialize its members to support rulings that enhance the power of the institution where they once worked, leading to a more generic support for the presidency in constitutional issues such as war powers, executive privilege, or standing to sue.

My hypothesis is not novel: prior studies on executive power decision-making on the Supreme Court include such experience as a control variable for this very purpose (Ducat and Dudley 1989; Yates and Whitford 1998). In these studies, though, the inclusion of executive branch experience as a control variable was based on intuition regarding its likely impact on presidential support. Both of these studies found that the effect of prior executive branch experience on support for the president was positive and substantial, but not statistically significant. The lack of significance, however, may have been driven by low levels of statistical power, as each study had less than 300 observations. A study which (1) controls for other variables, making use of improvements in the literature, and (2) has more data can shed further light on the relationship between executive experience and judicial-decision making.

Hypotheses

This study has two parts. First, I test for a correlation between prior executive branch experience and judicial deference towards the president. As such, I offer my first hypothesis:

Hypothesis One: Prior executive branch experience will correlate with an increased likelihood of support for the president's position in separation of power decisions involving the executive branch.

This step is not merely a prerequisite for further investigation—predictive value is useful even when explanation is lacking, and such a correlation may have significant policy implications for both the nomination process and the long-term power of the presidency even if we cannot say for certain why it occurs.

That said, establishing a correlation does little to test between competing theories. Assuming Hypothesis One is supported, how could one distinguish between socialization and signaling? Ideal research methods—such as panel data which surveys the legal policy preferences of future Supreme Court nominees before and after their executive branch service—are not feasible. At least one study has circumvented this difficulty by controlling for other known factors which might impact the dependent variable (specifically, the potential impact of socialization effects on ideological reactions to government spending) (Garand, Parkhurst, and Seoud 1991), but such an approach is similarly implausible, given our lack of survey data on the policy positions of administration officials or federal judges.

As such, one can only create indirect tests of socialization effects. One such test involves examining whether support for the president's legal position is affected by the duration of service. Though socialization effects are thought to develop relatively quickly, they are also believed to accumulate over time (Chao et al. 1994). Given this, if a positive correlation between executive branch experience and support for the president was driven by socialization, rather than recruitment, one might expect the probability of a future pro-presidential vote to increase alongside the length of service within the executive branch. This leads to my second hypothesis:

Hypothesis Two: The probability of a pro-presidential vote in separation of powers decisions will increase as the length of prior service within the executive branch increases.

Moreover, socialization effects are enduring but not necessarily permanent. Once someone leaves a particular organization, these effects may decay over time (Chao et al. 1994), or perhaps be displaced by judicial socialization (Carp and Wheeler 1972). If increased support for the pro-presidential position in separation of powers cases is driven by socialization effects, then justices who have more recently been part of the executive branch may be more likely to support the president than those whose served decades ago (Segal, Timpone, and Howard 2000). Specifically, one might expect tenure on the Court to ameliorate or even replace executive branch socialization effects as the years go by. As such, I offer my third hypothesis:

Hypothesis Three: Among justices with executive branch experience, the probability of a pro-presidential vote in separation of powers decisions will decrease as their judicial tenure increases.

As these latter two hypotheses do not directly address the influence of specific socialization content areas, their tests will provide an admittedly imperfect answer as to whether socialization effects drive deference to the president. That said, rejecting the null for hypothesis each will advance the viability of a socialization hypothesis, while a failure to find the expected correlations will instead provide support for recruitment and signaling theories.

Data and Model Variables

To test these hypotheses, I employ an original dataset of Supreme Court separation of powers cases involving the executive branch, ranging from 1942 to 2007. I selected these cases using the following methods. First, I searched the United States Supreme Court database (Spaeth et al. 2010), selecting any case that its coders identify as raising legal issues from Article I, Section 7; Article II; or Amendments XII, XX, XXII, or XXV. Second, I searched the United States Supreme Court Digest in the general topics of “United States,” “constitutional law,” or “war,” examining cases which reference the executive branch.² Third, I drew cases using a keyword search of Supreme Court separation of powers cases in the LEXIS legal database. Here I selected cases if one or more of the following criteria were present: (1) the president or vice-president was a party to the suit, (2) a cabinet secretary was a party to the suit, or (3) the contours of a specific presidential power (e.g., the appointments clause) were under direct consideration.³ In the third criterion, I generally limited my selection of cases to decisions where the Solicitor General filed an amicus brief advancing the legal position of the president, using this as a signal that the executive branch deemed this case sufficiently important to its political or institutional interests.⁴ After eliminating duplicate cases and those which did not meet these criteria, I compiled a total of 107 cases containing 919 separate justice-votes.⁵ The unit of analysis was the justice-vote; the dependent variable was the direction of that vote, coded 1 if the vote favored the president's preferred legal position, and 0 if against.⁶

Executive Experience

The primary independent variable is whether the justice who casts the vote for or against the president possessed prior experience in the federal executive branch.⁷ I coded this variable using the U.S. Supreme Court Justices Database (Epstein et al. 2010), cross-checked against the Federal Judicial Center's Biographical Database.⁸ For the test of Hypothesis One this variable is dichotomous, coded as 1 when the justice was employed in at least one executive branch position and 0 otherwise. The breakdown of this experience for each justice in the dataset can be seen in Table 1.

Table 1. Executive Branch Positions Held by Supreme Court Justices, in Chronological Order with Title and Years of Service

* Fortas also held a fifth position as the U.S. Undersecretary of the Interior for one year; Jackson had multiple overlapping positions during his seven years of service, including Special Counsel to the SEC while serving at Treasury.

I expect the coefficient for prior experience to correlate positively with increased support for the executive branch's preferred decision outcome.

For the test of Hypothesis Two, I generate two additional experience variables: a linear term, which counts the number of total years a particular justice served in the executive branch prior to joining the Court (rounded up), and an alternative specification which takes the square-root of that term. I provide this alternative measure because the rate at which socialization effects accumulate might slow over time, rather than be additive. I expect one or both of these terms to correlate with an increased likelihood to support the executive branch. For the test of Hypothesis Three, I count the number of years the justice has served on the Court at the time of the vote, and create an interaction term between this measure of judicial tenure and the dichotomous measure of executive experience. With this measure, I can examine both the impact of tenure per se as well as assess whether there is a difference in how tenure affects former executive branch employees versus those who lack such experience.

Ideological Distance and Judicial Ideology

As discussed above, it is unclear whether separation of powers outcomes divide along the standard liberal-conservative ideological axis. To examine this question, as well as to control for any potential ideological effects, I include both the absolute ideology of the voting justice, as well as the ideological distance between the voting justice and the sitting president. I do so because we do not know, for example, whether liberal justices are less likely than conservative justices to side with the executive regardless of who is president, whether it is ideological congruence that affects deference to the executive, both, or neither. To account for this, I first control for ideological distance with Judicial Common Space (JCS) scores, which place federal judges, the president, and members of Congress in a uniform one-dimensional ideological space (Epstein, Martin, Segal, et al. 2007). Specifically, I generate an ideological distance variable for each vote, which takes the absolute value of the difference between the JCS score of the president and the JCS score of the voting justice (calculated in the term in which the case was decided). As the ideological distance between the justice and the sitting president increases, I expect the likelihood of a pro-presidential vote to decrease.⁹ To measure the absolute ideology of the voting justice, I employ Segal-Cover scores (Segal et al. 1995), which are calculated using editorial assessments of the justices' ideology prior to their nomination (thus avoiding endogeneity problems) and range from zero (most conservative) to one (most liberal).¹⁰

Strategic Deference

One potential confounding factor in the model is that deference to the executive branch could potentially arise not from socialization effects or ideological congruence, but from what the judicial decision-making literature refers to as strategic concerns. In a complex institutional system, justices might sometimes vote against their true preferences in order to protect their institutional legitimacy or avoid popular backlash. To take a relevant hypothetical, justices may be reluctant to limit the power of a popular president during wartime. The evidence on whether judicial voting is systematically affected by the preferences of the other branches is mixed: some studies have found evidence that the Court behaves strategically in constitutional cases (Bergara, Richman, and Spiller 2003; Epstein and Knight 1998; Epstein, Knight, and Martin 2001; Harvey and Friedman 2005, 2009), while others have found little evidence of such behavior (Brudney, Schiavoni, and Merritt 1999; Segal 1997; Segal and Westerland 2005; Spriggs II and Hansford 2001). Nevertheless, it seems quite reasonable to control for factual scenarios where such strategic deference would be more likely.

To begin, one can control for deference towards the Solicitor General (S.G.). It is well-established that the Court is more likely to side with the position of the S.G. when he or she argues the case or files an amicus brief (Caldeira and Wright 1988; Deen, Ignagni, and Meernik 2003; McGuire 1998; Segal 1988). Such success is generally viewed as either resulting from the unique quality of the legal information the S.G. can provide, or more generically because the S.G. is the ultimate “repeat player,” especially skilled at presenting his or her argument. Regardless of the cause, justices may decide cases differently depending on the degree of the S.G.'s involvement. In these cases, participation by the S.G.'s office is almost universal. However, there remains considerable variation regarding whether the S.G. directly argues the case, or whether a subordinate takes the reins. I therefore include a dummy variable coded as 1 if the S.G. argues the case, and 0 otherwise, hypothesizing that the Court will be more likely to defer when the S.G. is directly involved.

I also consider the political strength of the president. If the Court defers to the president to avoid public or institutional backlash, it should be more likely to do so when the president is politically potent. I account for the strength of the executive branch in three ways. First, I hypothesize that the Court may be more likely to defer to the executive branch when the Court is an ideological outlier relative to the other branches. If the Court is much more liberal than both Congress and the White House, for example, the justices may feel more uncomfortable ruling against executive power claims. To account for this possibility, I create an “ideological outlier” variable using the JCS median scores for the Court, House, and Senate, as well as for the president. If the median member of the Court has either the most liberal or most conservative JCS score of the four in a particular term, I code the variable as the absolute value between the Court's score and the score of the branch closet to the Court. If, by contrast, the Court's JCS median is bounded on either side by another branch, I code the variable as zero. I expect that when the Court is an ideological outlier—and thus faces the possibility of coming into conflict with not just the president but Congress—it should be more likely to defer to the executive branch.

Second, I consider how wartime might affect judicial behavior. In previous studies, the classification of a case as foreign or domestic was thought to affect the likelihood of deference (Ducat and Dudley 1989; Yates and Whitford 1998). Simply put, the Court is believed to show greater deference to the president in foreign affairs than in purely domestic matters. However, the coding for “foreign” and “domestic” in these studies was neither formally defined nor sufficiently reliable. Take, for example, Youngstown Sheet & Tube Co. v. Sawyer (1952), a case which many constitutional casebooks reference as an example of the Supreme Court limiting presidential power in the domestic sphere. While the exercise of presidential power within the borders of the United States may indeed have alarmed the justices more than executive action whose scope was wholly overseas, not all the justices viewed the controversy as domestic, separate from the war effort. To be sure, Truman tried to nationalize the domestic steel industry, but he did so (nominally, at least) to ensure adequate steel production for the Korean War. Coding a case as foreign or domestic depending on how the majority treats the case introduces significant uncertainty in the coding scheme. Admittedly, some cases in the data, such as Morrison v. Olson (1988) or Clinton v. Jones (1997) seem unambiguously domestic. In general, however, this distinction lacks reliability; moreover, so many cases in the data are potentially classifiable as foreign that the measure would lack sufficient variation.

In this study I take a different approach, relying on a previous examination of how war and crisis affect the Court's willingness to protect civil liberties (Epstein et al. 2005). The rationale behind this approach is primarily a historical one, namely that when the nation is at war or in the midst of a foreign policy crisis, the Court is less likely to stand in the way of actions desired by the executive branch. I look at two distinct scenarios. First, the decision itself may take place during a war or international crisis. Second, a case may have its legal genesis from actions taken by the executive branch during a prior period of war and crisis, but be decided after the conflict has ended. In either situation, I hypothesize that the Court will be more willing to defer to the president's position. Relying on Epstein et al. (2005) for my classifications,¹¹ I code the former dummy variable (during) as 1 when the case is decided during a period of war or international crisis, and 0 otherwise. I code the latter dummy variable (from) as 1 if the central legal conflict of the case relates to crisis or wartime actions by the executive branch, and zero otherwise.

Third, I account for presidential strength by controlling for the president's popularity. Anecdotally, one can see greater Supreme Court support for popular presidents—such as FDR's tenure during World War II—and decreased support for less popular presidents—such as Nixon following Watergate. The relevant public opinion literature also suggests that the Court's positions on salient issues are generally congruent with public preferences (McGuire and Stimson 2004; Page and Shapiro 1983). It is unclear whether these effects are driven by judicial responsiveness to public preferences, or whether mass publics and judges are similarly affected by external forces. Regardless, at least one recent study has found that the Court is more likely to curtail civil rights and liberties during salient conflicts when the president's approval ratings are high (Craig 2011). This hypothesis is easily understood—the Court's legitimacy is at greater stake when both the public and elites are aroused to support a popular president, while the Court may have far less trouble opposing an unpopular president, especially if the president's unpopularity is connected to the case in question. While the majority of the cases in this study likely did not capture the public's attention, a sizeable number did involve salient issues such as wartime detention, impeachment, executive privilege, and the powers of special prosecutors. As such, it seems warranted to control for presidential approval. Here I employ Gallup's presidential approval ratings, using the average approval level during the month of oral arguments, the month of the decision date, and all months in-between.¹²

Other Contextual Variables

In a series of articles which refine the concept of partisan regimes and their relationship to the Supreme Court, Clayton and Pickerill argue that American history can be divided into distinct eras in which a particular majority party “regime” (e.g., the New Deal Democrats) generates a legal and policy environment which impacts judicial decision-making (Clayton and Pickerill 2004, 2006). During such eras, the majority regime pulls the political and legal center of gravity in particular directions, such as the broad deference given federal regulation of the economy during the era of the New Deal Democrats. On issues of great interest to the majority regime, then, standard ideological measures may not fully capture policy preferences across time. To put things more simply: Eisenhower appointees may be quite different from Reagan appointees, even as Johnson appointees may be different from Clinton appointees, even if they share similar party labels or are viewed as liberal or conservative in the context of their times.

Most relevant to this study is that the most recent such partisan regime, the New Right Republicans, made support for a strong executive branch a key component of their own partisan beliefs (Zschirnt, Clayton, and Pickerill 2008). Ideological measures notwithstanding, there may be important differences between justices appointed before and after the New Right rose to power. The exact point of transition between the New Deal regime and the New Right is up for debate, but in terms of Supreme Court nominations, there appears to be a clear distinction between justices appointed by Lyndon Johnson and those appointed by Richard Nixon. As such, a nomination cut-point of 1969 seems a reasonable demarcation. Given that Republicans controlled the majority of appointments between 1969 and 2006, and that the New Right's position on executive power may have affected the outlooks of justices appointed by Democrats as well as Republicans, I expect justices appointed in or after 1969 to be more favorable towards the expansion of executive power than those appointed before it (Zschirnt, Clayton, and Pickerill 2008). For this variable, I simply code the dummy variable as 0 for cases decided as before 1968, and 1 for those decided after.

Finally, I control for political salience. In more politically salient cases, justices might feel additional elite and public pressure to side with the president due to “rally ‘round the flag” effects, meaning salience essentially serves as another proxy for presidential strength. Alternatively, salience might inspire greater resistance to presidential power when relevant legal audiences—such as the law professoriate, legal elites, like-minded politicians, interest groups, other judges, etc.—are paying particular attention and oppose the executive branch (Baum 2006). In other words, salience may instead strengthen preexisting ideological tendencies (Unah and Hancock 2006). And if salience has a first-order effect on deference towards the president's position, interaction effects must be explored as well. For example, salient cases decided during war or crisis may lead to increased deference, while salience might have the opposite effect on cases decided during peacetime.

In contemporary quantitative studies of judicial decision-making, the standard measure of salience for Supreme Court cases is whether the decision merits a story on the front page of the New York Times on the day following the decision (Epstein and Segal 2000). I simply code cases as 1 if they do so, and 0 if they do not. Given the uncertain effect of salience in the model, I make no directional hypotheses for this variable.

Results

First, I test Hypothesis One, looking for a correlation between prior executive branch experience and an increased likelihood of supporting the presidentially-preferred position. Since the justice-vote is a dichotomous variable, I employ logistic regression using robust standard errors. Table 2 presents the logit coefficients, standard errors, and significance indicators for the variables in the main model.

Table 2. Binomial Logit Estimates on the Probability of a Pro-Presidential Vote, Selected U.S. Supreme Court Separation of Powers Cases, 1942–2007

Robust standard errors in parentheses; the constant is omitted to conserve space.

* p < 0.05, ^**p < 0.01, ^***p < 0.001.

Most importantly, and in line with Hypothesis One, prior executive branch experience has a moderate, positive impact on the probability of a justice siding with the president. The results also demonstrate the ideological dimension of decision-making in these cases. Support for the president's position decreases as ideological distance between president and justice grows, while a justice's own ideological position has an even stronger impact, with support for the president's position decreasing sharply as justices become more liberal. The fact that executive branch experience clearly impacts support for the president even when ideological factors are in play strengthens the study's primary claim.

The role of strategic variables is more mixed. On the one hand, there appears to be little difference between cases where the SG takes the reins and those where he delegates to his subordinates—perhaps because almost all of these cases are, by definition, of some importance to the executive branch. Similarly, and somewhat surprisingly, cases which are either decided during wartime or crisis, or originate from such periods appear to be treated no differently from those that do not. On the other hand, the Court does attend to some strategic considerations. Justices are more likely to side with the President when he is more popular, as well as when the median justice is an ideological outlier relative to the other branches.

Table 2 also shows that after controlling for ideology, the concept of partisan regimes does not substantially or significantly affect decision outcomes. Conservatives justices do support the president at a higher rate than liberal justices, but this effect does not appear to be conditioned on partisan eras (a point that an examination of interaction effects between partisan eras and the ideological controls confirms), at least using the admittedly crude measure employed here. Finally, Table 2 illustrates a strong role for political salience in the judicial calculus: in salient cases, the justices are considerably less likely to side with the president. This finding could mean several things. Perhaps the default position of most justices is to defer to the executive branch on matters of lesser importance, a dynamic that salience disrupts. In salient cases, by contrast, justices may care more about the outcomes and thus be more likely to act on their personal preferences (Perry 1991). Alternatively, in salient cases, the justices' decision-making calculus may be more affected by input from judicial audiences—many of whom are likely to oppose the presidential position—who are presumably paying greater attention in these cases (Baum 2006). Either way, this finding fits the broader theory that salience reinforces ideological behavior (Unah and Hancock 2006) and decreases individual variability in judicial decision-making (Collins Jr 2008)—though it should be noted that the interaction terms of both ideology variables and the salience measures are not statistically significant.

The main model was also rerun without unanimous decisions. Despite dropping the number of cases by about one-third, the executive experience measure remains positive and statistically significant. However, some changes do occur in this alternate specification. First, neither the ideological outlier measure nor the president's approval rating is positive or significant—perhaps reflecting that when the Court is an outlier or faces a popular president, it is more likely to unanimously side with the President. Second, the partisan regime variable becomes statistically significant with a moderate negative effect, contrary to expectations. There is no obvious explanation for such a finding. The variables of greatest import—namely executive experience and ideology—continue to perform as hypothesized, with larger coefficients for each. Given that extra-legal factors may have more room to operate in cases where the legal outcome is less clear, this finding is unsurprising. Overall, however, the model holds up well whether or not unanimous cases are included. Finally, a search for interaction effects among multiple variables did not reveal any noteworthy findings.¹³

At the very least, then, we have evidence that prior executive branch experience correlates with a higher probability of supporting the president. As such, we can move to testing Hypotheses Two and Three, indirectly examining whether socialization might explain this correlation. The organizational socialization assumptions embedded in Hypothesis Two suggest that support for the president should rise alongside additional years of service in the executive branch. Hypothesis Three, by contrast, states that since socialization effects may decay over time, judicial support for the president's position may decline as socializing effects recede (or more speculatively, are supplanted by judicial socialization).

In Table 3, I present three model specifications, replacing the dichotomous executive experience variable with: (1) a linear variable that counts the number of years (rounded up) a justice spent in the executive branch prior to being nominated, (2) a non-linear counter variable that takes the square-root of those years, and (3) the interaction term between the dichotomous measure of executive experience and the length (in years) of a justice's tenure on the Court at the time of the decision. All other model variables remain unchanged.

Table 3. Impact of Socialization Proxies on the Probability of Pro-Presidential Vote, Selected U.S. Supreme Court Separation of Powers Decisions, 1942–2007

Robust standard errors in parentheses.

* p < 0.05, ^**p < 0.01, ^***p < 0.001.

Model 3 also contains a variable for judicial tenure, which simply counts the number of years a justice has been on the Court as of the year of the decision. Since constituent terms have no meaningful interpretation in an interaction model, I omit this coefficient to conserve space.

These results provide mixed support for the hypothesis that socialization effects drive the correlation between executive branch experience and support for the president. On one hand, additional experience on the executive branch—for both measures, with the square-root measure being a slightly better model fit—clearly correlates with an increased likelihood of voting for the president's preferred position, supporting Hypothesis Two. On the other hand, the model results do not support Hypothesis Three, as the interaction coefficient is small and statistically insignificant.¹⁴ It may be, of course, that these results demonstrate instead that the socialization effects which come from executive branch experience do not decay, instead persisting over time. Regardless, both results remain more suggestive than dispositive, as they necessarily only indirectly measure socialization. At the least, however, the failure to reject Hypothesis Two suggests that the systematic study of career socialization, when confined to narrow areas of law where tighter theoretical justifications can be made, deserves further study.

Making sense of the impact of the model variables can be made clearer by moving from coefficients to predicted probabilities. Using CLARIFY, a software package for Stata that simulates such probabilities (Tomz, Wittenberg, and King 2001), I demonstrate how changes in the (main) model variables affect the probability of a pro-presidential vote. After setting the dummy variables at their modal value and other variables at their mean, I simulate the impact of changes in each variable which reached statistical significance in the main model (i.e., from Table 2). I present these predicted probabilities in Table 4.

Table 4. Effect of Statistically Significant Model Variables on the Predicted Probability of a Justice Making a Pro-Presidential Decision

This table uses CLARIFY (Tomz, Wittenberg and King 2001), a Stata package which simulates the predicted probabilities of the dependent variable (here, the probability of siding with the President) given different values of the independent variables. Here I hold the other main model variables (see Table 2) at their mean and modal values, and simulate the change in probability which results from moving each statistically significant model variable from its minimum to maximum value. CLARIFY's simulation process leads to slightly different estimates for each subsequent replication.

Justices with executive branch experience were on average about nine percent more likely to side with the president's position—a moderate effect (in the model which incorporates years of executive experience, each additional year of service makes a pro-presidential vote about one percent more likely, with the effect declining slightly as it reaches its dataset maximum). Ideological distance between the president and the justice has a similarly sized effect, while the absolute ideology of the justice has a much stronger impact. I better illustrate the importance of executive branch background and ideology in Figure 2, which illustrates the predicted probability of a justice siding with the president in a salient case under eight different conditions. I consider a vote where the justice is either liberal or conservative (in which I enter a Segal-Cover score either one standard deviation above or below the mean, respectively), as well as a vote where the justice is either ideologically close or far to the sitting president (again entering an ideological distance score one standard deviation below or above the mean). Finally, I vary whether the justice possesses executive branch experience.

Figure 2. Predicted Probability of a Pro-Presidential Vote in Salient Cases, Accounting for Ideological Distance from the President, Judicial Ideology, and Executive Experience.

As Figure 2 shows, the effect of executive branch experience is fairly consistent across each of these four scenarios. Whether a justice is liberal or conservative, however, clearly has the largest impact, with liberal justices being much less likely to side with the executive branch than conservatives. This figure clearly confirms that the attitudinal model does, contrary to received wisdom, apply to these separation of powers cases.

Implications

The model results clearly support a correlation between executive branch experience and increased deference to the president. Executive branch experience is not simply a promising predictor of future ideological consistency, but affects outcomes in line with the expectations of the social background model. Moreover, the fact that deference to the president is conditioned by one's duration within the executive branch suggests that institutional service may socialize preferences. Of course, while the empirical support for Hypothesis Two is what one would expect if socialization effects were a causal mechanism, a separate preference dimension for executive support, in which one stays within the executive branch longer because of preexisting preferences for a stronger presidency, remains a plausible, if less likely alternative. When combined with the fact that the results did not support Hypothesis Three, evidence for organizational socialization effects remains more suggestive than conclusive. Moreover, these results must be appropriately circumscribed—the correlation may not hold prior to the dramatic expansion of the executive branch during the 1930s. Given the nature of these findings, as well as the more recent successful uses of social background theory, I suggest that social background theories will continue to inform a set of “mid-level” models, not readily aggregated into a more global theory of judicial behavior for “all cases.” Nevertheless, these results still (1) point to the utility of further examination of socialization effects, (2) improve our ability to predict judicial decisions in separation of powers cases, and (3) refute a long-standing hypothesis that separation of powers cases do not align along a liberal-conservative axis.

In concluding, I discuss two additional implications of these results: one primarily of interest to scholars, the other of interest to any political observer. For the first point, I contend that these results demonstrate the value of paying attention to judges' career background when there are good theoretical reasons to do so. The field of judicial behavior has rightly placed ideology and policy preferences at the heart of contemporary judicial decision-making models. Nevertheless, while such measures remain the best predictors of judicial behavior in general, in narrow legal policy areas where a more rigorous theoretical connection can be drawn, it seems wise to test for social background factors. This study thus joins other work that emphasizes similarly circumscribed connections, in which gender affects sex discrimination cases (Boyd, Epstein, and Martin 2010), elite backgrounds affects the propensity to side with unions in National Labor Relation Boards decisions (Brudney, Schiavoni, and Merritt 1999), and experience as a criminal defense attorney decreased judicial support for the Federal Sentencing Guidelines (Sisk, Heise, and Morriss 1998). Once such relationships come into view, additional research can assess whether socialization effects might be the causal factor, or whether the literature should take more seriously the project of building separate preference dimensions for specific issue areas.

This approach makes sense not only as a means of building more accurate decision models, but also because presidents who nominate and confirm judges and justices may be more concerned with predicting decision-making in a few policy areas of personal concern, rather than divining a nominee's overall ideology. As David Strauss notes, presidents do not give equal attention to all issues when they select for policy preferences, sometimes only desiring congruence on a few key issues (Strauss 2007). In this vein, Strauss says, FDR cared about federal power, not race (but see McMahon 2004), Nixon cared about criminal justice and the pace of desegregation, not abortion, and so on. If presidents mainly desire judicial support for a small set of issues, then understanding what proxies or other variables correlate with outcomes in these narrower areas may be useful.

As for the broader point: we do not have to explain the cause of the relationship between prior executive branch experience and increased deference to the president for the relationship itself to matter. Obviously, when presidents select nominees for their expected positions on separation of powers cases—as President Bush may have done in 2005 after his legal defeats over the Guantanamo Bay detainees—such experience will be directly relevant. Beyond that, however, presidents generally desire to nominate judges and justices who will be ideologically congruent and consistent on the issues they care about, but face significant uncertainty in doing so. I submit that presidents may come to view prior executive branch experience as a good indicator of future ideological consistency, a proxy clear enough to avoid nominating “Souters” but oblique enough to hinder the opposition's rhetorical attacks.¹⁵ For example, as the front-runner to replace Justice Stevens, Elena Kagan had served within both the Clinton and Obama administrations. Such executive branch experience appears to have assuaged the Obama administration (though certainly not all liberals) that Kagan was an ideologically congruent choice and one likely to stay the course, while also failing to provide the administration's opponents with adequate fodder for attacking her specific legal policy preferences. Such a potentially dominant selection strategy—akin to the current confirmation hearing stratagem that nominees say nothing of substance—might lead to a Court that consistently has seven or more justices with such experience, assuming that policy preferences, rather than electoral politics, tend to guide nomination decisions.

This study clearly shows that while ideology matters a great deal, nominees with such experience can still be predicted to be more deferential to the presidency regardless of who is president. If one accepts my specualtion that both parties will increasingly appoint individuals with such experience to maximize the chances of implementing their legal policy preferences, then the end result will be a Court increasingly likely to support the executive branch in separation of powers disputes. In an era where the Court rather than Congress often provides the primary check on the expansion of unilateral presidential power, this possibility deserves further investigation.