Explaining the (Non)Occurrence of Equal Divisions on the U.S. Supreme Court

The Infrequency of Evenly Divided Courts

Reynolds and Young (1983) provide what we believe to be the first comprehensive history of equal divisions on the Court. As they note, the Court first articulated the rule for dealing with equal divisions in The Antelope (1825). The case involved the standard for determining the rightful owners of slaves who were suspected of being transported through piracy. Chief Justice Marshall ruled that the African slave trade could only be prohibited where it violated the laws of both nations engaging in the trade but refused to articulate a legal reasoning in the case because “the Court is equally divided on [the issue] and consequently no principle is settled.” As is the case in judgments of the Court, when the Court evenly divides, its decision has no precedential value. However, unlike a judgment of the Court in which either party can win, equally divided cases simply “affirm the judgment below” and are announced without opinion. Evenly divided cases then provide for a winner in the case immediately before the Court (the respondent) but do not provide broader legal guidance (Reynolds & Young, 1983).

The Court has been careful to articulate that in the same way that a refusal to hear a case (i.e., the denial of certiorari) is neither a tacit validation of the lower court’s legal reasoning nor an affirmance by an equally divided court. For instance, in the per curium opinion in Ohio Ex. Rel. Eaton vs. Price (1960), the Court wrote, “four of the Justices participating are of opinion that the judgment should be affirmed, while we four think it should be reversed. Accordingly, the judgment is without force as precedent.” The per curium in Price went on to explain the reasoning for announcing the judgment without opinion, writing as follows:

Usual practice of not expressing opinions upon an equal division has the salutary force of preventing the identification of the Justices holding the differing views as to the issue, and this may well enable the next case presenting it to be approached with less commitment.

Perhaps because of the steep opportunity costs associated with an even division—justices go through the entire decision-making process without actually producing an opinion—they are rare. Between the Court’s 1946 and 2011 terms, nearly one fifth of the Court’s decisions were made with an even number of justices participating. Yet, in only 4.7% of these cases did the Court end up with an equally divided vote. ³ In addition to being low in an absolute sense, this rate is also significantly lower than what we might expect to naturally occur. On a court with eight remaining justices, for example, if vote outcomes were simply random, then we would expect to observe a tie in 20% of cases where it is possible (i.e., one out of five possible voting splits). Even a more conservative approach still predicts that a tie should emerge in 11% of the cases—more than double what is actually observed. ⁴

In Figure 1, we illustrate the occurrence of both even-member courts and tie votes across time. Here, the x-axis depicts each Court term. On the y-axis, we portray the number of cases where there was an even number of justices that did (light gray) and did not (dark gray) result in a tie vote. As the figure makes clear, while there is substantial term-by-term variation in the number of cases where the Court is “at risk” of ending up in a tie, as a general matter, the frequency with which it does deadlock in a divided vote is quite low. Unsurprisingly, we tend to observe significantly higher numbers of “at risk” cases in terms that immediately follow a personnel change on the Court. For example, 2005, the term following the appointment of Chief Justice John Roberts, observed the highest number of “at risk” cases since 1991—the term in which Justice Thomas joined the bench. The data also reveal what appears to be a longitudinal trend, with the number of cases decided by an even-membered court dropping off somewhat precipitously under the leadership of Chief Justice Rehnquist. This is consistent with existing accounts, which suggest that Rehnquist urged his colleagues to avoid discretionary recusal whenever possible (Black & Epstein, 2005).

OPEN IN VIEWER

Figure 1.

Frequency of even-membered Courts and tied votes, by term (1946-2007).

In short, equally divided courts appear far less often than one might expect based on the Court’s voting patterns. Despite uncovering this interesting and important aspect of the Court’s decision making, Black and Epstein (2005) fail to address the question their study naturally begs: Why do so few even-membered Courts result in a tie vote? Indeed, they concede,

Whatever the explanation for the lack of evenly divided cases, this may be a more interesting and even more pressing matter for future exploration . . . [These findings] alone commend greater attention to the question of why more ties do not occur. (Black & Epstein, 2005, p. 97)

The next section provides such an analysis.

Explaining Evenly Divided Courts

More than 50 years of scholarship have been devoted to understanding the goals that motivate justices, what actions they take to achieve those goals, and the effect of those goals on American jurisprudence. However, less is understood on how justices prioritize those goals when they are in conflict.

Baum (1997) argues that justices are not single-minded pursuers of policy goals. Rather, their goals exist in a hierarchy. While justices obviously care about deciding the law “correctly” (which can encompass a commitment to policy preferences, precedent, and legal doctrine), they are also concerned with deciding the law clearly. Justices understand that their position atop the federal judicial hierarchy gives them the additional obligation of sending clear signals to the courts below, the elected branches, and the American people on what the law means. Accordingly, justices are often more concerned with deciding the law clearly than they are with deciding it correctly.

If justices seek to avoid equal divisions, it shows a potential willingness to forgo policy preferences for the sake of the structure as well as the content of the law—a desire to sacrifice for legal clarity. The heart of our theory rests on the assumption that justices should be especially willing to make this sacrifice when legal clarity is of heightened importance. We turn here to case factors—legal and otherwise—that should trigger such a willingness. In short, we hypothesize below that justices recognize signals from the lower courts and the other branches that indicate that certain cases are in particular need of resolution and legal clarity.

First, abundant evidence suggests that the Supreme Court is deeply concerned with resolving lower court conflict. Indeed, Rule 10 of the Supreme Court states that the presence of lower court conflict is a key factor in the Court’s decision to grant certiorari. Not only does the presence of conflict increase the likelihood that the Court will grant a case (Caldeira & Wright, 1988; Ulmer, 1984), but it can also induce justices to suppress policy-minded behavior when casting their agenda-setting votes (Black & Owens, 2009). Conflict in the lower courts is consistently cited by the justices themselves as one of the key problems of equal division (see, for example, Leahy, 2010; Rehnquist, 1972); the obligation to resolve dissensus between the lower courts is one of the key reasons the justices are so committed to legal clarity. Because a tie vote perpetuates the existence of lower court conflict, we expect the following legal conflict hypothesis:

The Court also receives information from the elected branches that clarity in the law is important in particular cases. The clearest of these signals comes from the Office of the Solicitor General (OSG). An abundance of literature suggests that the Court pays significant attention to OSG recommendations, both in terms of helping the Court shape the content of its agenda (Black & Owens, 2009; Caldeira & Wright, 1988, Perry, 1991) and even when deciding a case in a particular manner (Black & Owens, 2012; Wohlfarth, 2009). Whether because it speaks for the preferences of the executive branch, or because it is a trusted repeat player, the OSG’s elective involvement is an indication that resolution of the legal issue is considered important by the federal government. Accordingly, we expect the following executive involvement hypothesis:

Unlike the president, Congress does not have an official representative to the Court. Because of this, it must communicate using more subtle signals. For instance, while members of Congress cannot directly lobby the justices, they can and frequently do get involved in a case as amicus participants. During the 104th session of Congress (1995-1996), for example, just over a third of the members signed at least one brief; in earlier sessions, a majority of legislators signed briefs as well (Spill Solberg & Heberlig, 2004).

Likewise, Congress can signal its interest in a case if it is directly involved as a petitioner, as it determined the case was important enough to warrant a challenge of an adverse decision below, or if the case involves a constitutional challenge to a piece of federal legislation or involves the legislative veto. Each of these factors sends strong signals to the Court that settling the rule of law in those cases might be more important than the content of the law. Thus, we suggest the following legislative involvement hypothesis:

While institutional obligations and constraints will reduce the likelihood that the Court deadlocks in a tied vote, it is also important to account for the policy component of the Court’s decision making. One critical dimension of policy is the opinion-writing process itself. To produce a lasting precedent, an opinion writer must ultimately produce a draft that can garner the support of at least five of her colleagues. The ease with which this can be accomplished, of course, is a function of the policy preferences of justices on the Court. On a Court where justices have relatively consistent preferences, there will be a larger degree of overlap in terms of policy outcomes that are acceptable. This, in turn, makes it easier to build a consensus. By contrast, when the Court is ideologically splintered, the overlap is smaller, and building a majority coalition will be more difficult (Maltzman et al., 2000). Accordingly, we expect the following ideological heterogeneity hypothesis:

Of course, not all cases are equally important to the justices. A wealth of scholarship demonstrates that justices are more committed to their ideological positions in salient cases (see, for example, Johnson, Spriggs, & Wahlbeck, 2005; Maltzman et al., 2000). Because avoidance of an equal division requires the justices, to an extent, to sacrifice their own policy preferences for the sake of legal consistency and institutional legitimacy, we expect the following salience hypothesis:

Finally, while recusals are the primary reason for even-membered Courts, they are also the least predictable. Justices have the discretion to recuse or not recuse at any time for any reason. Thus, justices cannot plan around recusals, either by organizing the docket to avoid granting cases that should be decided with a full slate of members or by holding over important cases until the bench is at full strength. By contrast, when justices leave the bench permanently, their colleagues are better able to plan for an even-membered Court. They both know who has left the bench and can likely anticipate how long the absence will last and the probable ideological predisposition of their future colleague. There is abundant anecdotal evidence that justices hold over important or controversial cases when faced with a permanent vacancy (see, for example, Woodward & Armstrong, 1979). We anticipate that this ability to strategically plan ahead allows justices to better avoid even divisions and therefore expect the following permanent vacancy hypothesis:

In summary, we argue that the Court strives to balance between institutional and policy considerations when attempting to avoid tie votes. The presence of institutional factors such as legal conflict and the interest of other branches of government act to depress the likelihood of a tie vote because legal clarity is more important, while ideological forces tend to push the Court toward an even split because the justices place more weight on legal accuracy. We next turn to describing the data and measures we deploy to test these hypotheses.

Data and Measurement

To test the above-mentioned hypotheses, we started by combining the Spaeth Database “allcourt” and “vinwar” files and examining all orally argued cases heard between the Court’s 1946 and 2007 terms. The year 1946 represents the first term included in a multiuser database and 2007 corresponds to the most recent term available when we began this research project. Because a tie vote can only occur in cases with an even number of justices, we then excluded any and all cases that were decided by an odd number of justices. This left us with a total of 1,242 cases that were “at risk” of observing an evenly divided court. Our dependent variable is whether the Court, when at risk of an even division, actually ended up doing so (1 = yes; 0 = no). The initial coding of this variable was done by using the “dec_type” variable in the Spaeth databases. ⁵ We then conducted two reliability checks. First, we searched the Lawyers’ Edition of Supreme Court opinions available on LexisNexis for the phrase “affirmed by an equally divided court” to capture any cases that were not included in Spaeth. Second, we read the text of each opinion identified as an evenly divided court to ensure that such a division actually occurred. ⁶ All told, 97 of our initial 1,242 observations satisfied this criterion (about 8%).

Turning to our independent variables, Hypothesis 1 argues that the Court will be less likely to reach a tie vote when there is conflict among two or more circuits on an important legal question. Because evenly divided courts (i.e., the “1s” in our data) do not produce a written opinion, we cannot use the most common approach of measuring conflict, which is to use the database’s variable of whether the Court’s final written opinion noted the existence of legal conflict. Instead, we turned to the parties’ initial certiorari briefs or jurisdictional statements. We started by determining whether the petitioner alleged the existence of legal conflict. If she did, then we examined the respondent’s brief in opposition to determine whether the respondent contested this allegation. We take this step to control for the potential that a petitioner has exaggerated the extent of a conflict in order to attract the Court’s attention at the agenda-setting stage. If the respondent conceded or otherwise failed to refute the conflict claim, we code Legal Conflict as 1. ⁷

With Hypothesis 2, we argue that the Court will be less likely to reach a tie vote in cases where the United States has expressed a strong preference toward having the case resolved. Accordingly, we code U.S. Case Interest as 1 if the OSG, the executive branch’s primary judicial representative, participated in the case as the petitioner or submitted any type of amicus brief in a case. ⁸

To operationalize our Legislative Involvement Hypothesis, we started with data from Epstein, Segal, Spaeth, and Walker (2007), which we updated using LexisNexis to cover the 1997-2007 terms, to determine whether any member of Congress participated in a case. If one did, we coded the variable as 1. We also coded the variable as 1 if any of the following conditions were met: The House or Senate was the petitioner in a case, the case involved judicial review of a federal statute, or the case presented an issue involving the legislature as an institution (e.g., the legislative veto). We code these additional elements to capture other cases that are likely important to Congress. Our variable, Congressional Interest, takes on a value of 1 in just about 12% of our observations and 0 in the other 88%. ⁹

To measure the ideological influence on the Court’s propensity to result in a tied vote, we include Ideological Heterogeneity, which we code in three basic steps. First, for each case, we identified the remaining justices who participated in that decision. Second, using data from Martin and Quinn (2002) we estimated their policy preferences using their ideal point for the term in which the case was decided. Finally, we calculated the interquartile range for the justices’ scores. ¹⁰ Substantively, then, our measure taps into the extent to which the policy preferences of the remaining justices are tightly clustered versus diffused. The intuition here is that it will be more difficult to amass a coalition of five votes when the Court is ideologically divided than when it is not (Maltzman et al., 2000). Higher values correspond to a more diffused court whereas lower values suggest the Court is more homogeneous.

To measure political salience, we follow the general tack of Epstein and Segal (2000) and code Case Salience as the number of stories—excluding “round up” summaries—written about a given case in the New York Times from the day the case was accepted for review up to (but not including) the day of the decision. That is, we do not include any coverage by the Times about the decision itself, since so doing would constitute a noncontemporaneous measure of case importance (Black, Sorenson, & Johnson, 2013).

To operationalize Hypothesis 6, we merely identified if the Court was operating at less than full strength because a justice had retired or died and her replacement had not yet been appointed rather than as a result of a recusal, illness, or other temporary vacancy. ¹¹ Therefore, Open Seat Vacancy equals one when the Court was waiting for a new member to be appointed and zero when the vacancy was the result of a recusal or nonparticipating member.

Finally, to address the potential confounding effect case complexity might have with our variables of interest, we control for case complexity by coding the total number of amicus curiae briefs present in a case (Collins, 2008). ¹²

Method and Results

Because our dependent variable is dichotomous, we estimate a logistic regression model. ¹³ Parameter estimates for our model are reported below in Table 1. As hypothesized, we find that the Court is less likely to evenly divide when the case presents the Court with legal conflict. Similarly, when the Solicitor General is involved in the case—evidence of interest from the executive branch—the Court is also less likely to end up deadlocked in a tie. We also find, as expected, that the Court is more likely to evenly divide when the case is salient as opposed to when it is not. In addition, the Court is less likely to result in a tie when operating at full strength due to a retirement or death rather than a recusal. This, too, is consistent with our initial hypothesis.

OPEN IN VIEWER

Table 1.

Logistic Regression Model of Whether an Even-Membered Supreme Court Results in a Tie Vote.

Variable	Coefficient	Robust SE
U.S. case interest	−0.558**	0.236
Case salience	0.277**	0.110
Legal conflict	−0.546*	0.277
Congressional interest	−0.689	0.456
Ideological heterogeneity	0.159	0.136
Case complexity	−0.008	0.022
Open seat vacancy	−1.574**	0.430
Constant	−2.200**	0.258
Observations	1,242
Log likelihood	−321.340

*

p < .05. **p < .01, one-tailed tests.

However, several of our hypotheses were not supported. We had expected more heterogeneous courts to be more likely to result in a tie. Although our coefficient is in the expected direction its p value exceeds conventional levels of significance (p = .12, one-tailed test). ¹⁴ We also find no evidence that justices are any more (or less) likely to avoid an equal division when they observe interest by members of Congress. However, it is worth noting that this null finding is largely consistent with existing work on the topic. Indeed, Spill Solberg and Heberlig (2004), in their comprehensive analysis of congressional amici, conclude, “congressional amicus curiae briefs are not efficacious signals of the merits of policy or statements of institutional prerogatives but low cost methods to reveal members’ positions to all those within earshot” (p. 604). Because the signals are less powerful and much more indirect, it seems logical that the Court would be less responsive to them than it is to signals from the Solicitor General (SG). Finally, we fail to document any systematic relationship between a case’s complexity and the Court’s likelihood of reaching an equally divided outcome. ¹⁵

To further examine the substantive nature of our results, we turn to predicted probabilities, which we present in Figure 2. Along the x-axis, we show the estimated probability of observing a tie. The y-axis provides a series of counterfactuals from our data. The dashed horizontal line provides a baseline comparison for the likelihood of observing an evenly divided court, holding all other variables at their mean or modal values (as appropriate)—approximately .12.

OPEN IN VIEWER

Figure 2.

Predicted probability that the Court evenly divides given different variables of interest.

Starting on the top of the y-axis, we examined the effect of legal conflict. We had hypothesized that the Court would be, owing to its institutional obligations, less likely to end in a tie vote in cases where conflict is present. Our results suggest the Court does exactly that. As Chief Justice Rehnquist, Senator Leahy, and the Court itself have argued, one of the key duties of the Supreme Court is resolving lower court disputes. The fact that the Court is less likely to evenly divide in cases where there is demonstrated tension in the law is a key indicator that the justices are willing to overcome policy disputes to resolve key legal controversies and articulate consistent and enforceable law. In cases where conflict is present, the Court has only a .08 probability of resulting in a tie vote—a relative decrease of 33%. As our coding procedure, which uses the litigant’s briefs, takes a fairly conservative approach to identifying conflict, we suspect this likely underestimates the true effect size.

While the Court pays significant attention to its obligations to the legal community, our results imply its ears are equally attuned to suggestions from the U.S. government that certain cases are especially in need of legal clarification. Moving down to the second line in our figure, we estimate that the likelihood of a tie vote when the United States is an active participant drops by more than 40% to a value of .07. Equal divisions make federal policy harder to implement and fail to add clarity to the law. As a result, the Court seeks to avoid them when the United States indicates that a case must be resolved. ¹⁶

Our results do demonstrate there are limits to the Court’s ability to avoid an equal division, however. In this vein, the results from our Case Salience variable are informative. For the roughly 84% of the cases in our data where we observe no newspaper articles, the probability of observing an equally divided court is the baseline value—that is, .12. However, we estimate a .16 probability of a tie vote in cases that observed a single newspaper story by the Times—a 33% relative increase. A case with an exceptionally high level of media attention (three articles or roughly the top 1% of our data) would have about a 24% chance of being equally divided, which is double the initial baseline probability. ¹⁷

Finally, we consider the impact of the “type” of the vacancy on the probability that the Court splits. As the bottom line of the plot makes clear, the Court is exceptionally reluctant to end in a tie when the vacancy has been caused by a colleague’s death or retirement. When a justice will not be returning to the bench (i.e., it is not a discretionary recusal), we estimate only a .03 probability of observing a tie vote—a relative decrease of 75%. It is interesting to note that open-seat vacancies are the single strongest predictor of whether the Court will evenly divide.

Conclusion

Justice Brandeis famously wrote, “It is usually more important that a rule of law be settled, than that it be settled right.” Insights from political scientists ranging from Murphy (1964) to Epstein and Knight (1998) have taught us that “right” usually means that the law was settled in accordance with the justice’s most preferred policy outcome. However, there has been significantly less work done to investigate the lengths justices go to reach the more important goal, according to Brandeis, of settling the rule of law—or, to put it another way, making the application of the law clear and consistent.

Our analysis of equal divisions on the Supreme Court examines one possible way the clarity of law can be harmed by ideological goals. In so doing, we also provide an initial answer to the broader topic of what happens when the multiple goals of judges come into conflict. Our results are important not only because even divisions are so legally problematic but also because their (non)occurrence gives us a unique insight into how the Court responds when institutional obligations and policy preferences clash. On this dimension, our results demonstrate that justices may be willing to forgo their own ideological predilections for consistent law with long-term precedential value, especially when other institutions signal that consistent law is necessary.

Understanding equal divisions of the Court is, by itself, important. Justices often use the possibility of a tied vote as a justification for not recusing themselves (Laird vs. Tatum, 1972). This puts them in an ethical dilemma between their obligation to decide the nation’s most pressing legal controversies and avoiding potential conflicts of interest. While our findings indicate that even divisions are rare, and especially rare when legal or institutional concerns are paramount, they also demonstrate that politically salient cases are at a higher risk for division as are cases where justices have taken the perhaps most responsible course of action of acknowledging potential conflicts of interest and stepping back from a case.

Regardless of how infrequent tied votes may be, each one has a broad and lasting impact on the consistent application of law in the United States. As such, it is important for scholars to know not only when and why they occur, but also the potential consequences of trying to avoid even divisions. Our results here indicate that, by and large, the Court sees even divisions as an institutional problem that it has an obligation to prevent. Indeed, our results suggest that while justices are often ideological in their behavior, they are also ultimately good shepherds of the law.