Seasonal Affective Disorder: Clerk Training and the Success of Supreme Court Certiorari Petitions

Abstract

We investigate why the Supreme Court grants a smaller percentage of cases at the first conference of each term compared to other conferences. According to received wisdom, Supreme Court law clerks are overly cautious at the beginning of their tenure because they receive only a brief amount of training. Reputational concerns motivate clerks to provide fewer recommendations to grant review in cert. pool memos written over the summer months. Using a random sample of petitions from the Blackmun Archives, we code case characteristics, clerk recommendation, and the Court's decision on cert. Nearest neighbor matching suggests clerks are 36 percent less likely to recommend grants in their early cert. pool memos. Because of this temporal discrepancy, petitions arriving over the summer have a 16 percent worse chance of being granted by the Court. This seasonal variation in access to the Court's docket imposes a legally irrelevant burden on litigants who have little control over the timing of their appeal.

“This Broth of the Certiorari Process”¹

Members of the Supreme Court bar have complained that petitions for certiorari review filed in the summer months have a much lower grant rate than at other times during the term. When the justices meet for the first time for the upcoming October Term, sometime at the end of September, they dispose of all certiorari petitions that have arrived during the summer months. Former Supreme Court clerk, Edward Lazarus (2005: 29) has characterized this meeting, appropriately called the Long Conference, as a “single marathon session.” On average, during the Roberts Court era the Court has disposed of more than 1,800 petitions at the Long Conference, and the proportion of grants to denials appears lower than at any other point during a term. In their qualitative analysis, Cordray and Cordray (2004: 210) attribute this difference in grant rates to the “great mass of petitions” considered at the Long Conference exerting a “numbing effect on the Justices.”

We endeavor to find a more empirically satisfying explanation of this behavior by examining the actors whose recommendation often makes or breaks a petition–Supreme Court law clerks. For each petition, the justices rely on one clerk's memo, which summarizes elements of the case and provides a recommendation as to whether the Court should grant review. The law clerks who handle the massive number of summer cert. petitions are new to their posts, having received only a brief amount of training from the outgoing clerks. Another former clerk, Jeffrey Fisher (as cited in Wolf 2013), described the behavioral incentives clerks initially encounter: “New law clerks know that the way to play it safe is almost always to recommend a denial.” Boskey (2012) traces the clerk's lack of confidence in making grant recommendations–what we refer to below as grant-averse behavior–back to clerks working for Chief Justice Stone in the 1940's, long before the creation of the cert. pool. If these anecdotes are true, they raise serious normative questions about the fairness of the Supreme Court's procedures. Given the already low probability of receiving certiorari review, access to the Court's docket should not depend on what month a cert. petition reaches the Court's hands. In this article, we provide a theoretical advancement in the understanding of the cert. pool and document its meaningful and normatively troubling consequences for some litigants.

The study of law clerks has explored various avenues of influence from authoring opinions and preparing justices for oral argument (Ward and Weiden 2006), to the ideological congruity between clerks and justices (Baum and Ditslear 2010; Ditslear and Baum 2001; but see Kromphardt 2015), to the influence of clerk ideology on the decisions of the justice who hired her (Brenner and Palmer 1990; Peppers and Zorn 2008). Another line of literature examines the clerks’ role in reviewing cert. petitions. The quantitative studies that followed Perry's (1991) account of certiorari behavior argued that a justice votes to grant or deny review based largely on her own policy preferences and the expected results at the merits stage (Boucher and Segal 1995; Brenner, Whitmeyer, and Spaeth 2006; Caldeira, Wright, and Zorn 1999). Except for being informed by the clerk's summary of legal issues (see Brenner 2000), the clerk's recommendation is epiphenomenal to these earlier studies. Indeed, Stras (2006) found that agreement among clerks and justices on whether to grant certiorari is above 98 percent when considering all cases on the Court's docket. However, Black and Boyd (2012) have found clerk recommendations influence the Court's certiorari behavior under certain strategic and ideological conditions.

Our contribution to the study of Supreme Court certiorari decision-making is twofold. We conceptualize the cert. pool as a two-stage process: the clerk arrives at a recommendation; then the justices, based in part on the clerk recommendation, determine the final disposition of the case. While prior studies have examined whether and how cert. pool memos influence the Court's decision (Black and Boyd 2012; Black, Boyd, and Bryan 2014), and how clerks arrive at their recommendation (Bryan 2012), we connect these processes to provide a richer, more nuanced exploration of the procedural complexities and permutations that flavor the broth of the Supreme Court's agenda-setting.

More importantly, we assess whether the dynamics of the cert. pool operate in the same fashion across an entire term, especially with respect to the Long Conference. No prior quantitative study has examined whether clerk recommendation patterns vary over the course of their tenure. Litigants have ninety days after a final judgment is entered in the court below to submit a request for further review. While the litigants can control certain aspects of litigation, they cannot control when a lower court enters a final judgment, which creates a narrow window for filing a petition for certiorari with the Supreme Court. If this window opens and closes during the summer months, our analysis suggests these litigants face an arbitrary and legally irrelevant disadvantage that is empirically attributable to the clerks’ initial hesitation to recommend grants of certiorari.

In the following section, we provide a brief background on the certiorari process and the clerks’ role therein. While we examine how certiorari works in practice, we also examine its theoretical importance to the study of judicial decision-making. We explore the balance between justices’ personal preferences, limitations placed on those preferences by institutional rules and context, and the importance of legal factors as developed by previous research. In the final section, we reconsider the role of the law clerk in the certiorari process, using the Blackmun Archive to create a new dataset of certiorari decision making from the 1987–1993 terms.² Nearest neighbor matching suggests clerks are 36 percent less likely to recommend grants in their early cert. pool memos. Consequently, petitions arriving over the summer have a 16 percent worse chance of being granted by the Court. We conclude by considering the ripple effect that the lack of Long Conference grants of certiorari has on the Court's docket management.

The Supreme Court Agenda-Setting Process

The study of clerk influence at the agenda-setting stage is complicated by the presence of the “cert. pool,” an institutional arrangement designed to conserve resources and enhance the Court's ability to manage its agenda efficiently.³ Currently, all but one justice⁴ participate in the cert. pool. The process tasks one randomly assigned clerk to write a memo evaluating a petition for certiorari. That memo is distributed to all justices participating in the pool, not simply to the justice who hired her. Pool memos follow a relatively simple template. Clerks will include a brief summary of the legal issues, material facts, and the reasoning of the court below. Further, clerks record the names of judges in the court below who heard the case and who wrote the corresponding opinions. The memo concludes with the clerk's own analysis of the case, summaries of the parties’ arguments, and her recommendation on granting the petition. The primary concern of the pool clerk in writing the memo is to justify her recommendation for or against granting certiorari review. The Supreme Court's Rule 10 states that “a review on writ of certiorari is not a matter of right, but of judicial discretion.” Rule 10 goes on to define “certworthiness,” those characteristics that may influence the Court to grant a petition for certiorari review.

Petitions arrive at the Supreme Court all year long, even when the Court is out of session. The Court's June recess marks an important cut-off for cert. petitions. When a petition arrives following the Court's June recess, the Clerk of Court assigns a docket number corresponding to the upcoming Supreme Court term beginning on the first Monday in October. Once the clerk dockets the petition and the opposing parties have submitted their certiorari briefs, the Court then receives the lower court record, and any interested parties granted permission may file pre-certiorari amicus briefs. Once all the briefs are filed, the intake clerks “distribute” the case to the justices’ chambers. A case's distribution date determines what Distribution List it will be placed on for consideration at a regularly scheduled Conference. The Court announces its case distribution schedule at the beginning of each term, identifying the approximate dates of each conference and which cases with specific distribution dates it will consider at those conferences.⁵

When the intake clerks distribute cases to each chamber, review of those cases by the justices and their law clerks begins. Review of petitions across the nine chambers follows two distinct processes: review within nonpool chambers, and review within chambers participating in the cert. pool. It is this second process that concerns us here. The Clerk of Court randomly assigns petitions to chambers, and each chamber randomly assigns pool memo writing responsibilities among its clerks. In the past, clerks could take significant time to craft pool memos, but Chief Justice Rehnquist eventually imposed a two-week submission deadline on all pool clerks. In the timeframe analyzed in this article, the Chief Justice had yet to impose the two-week deadline, meaning that a clerk could allow memo writing responsibilities to languish until the end of the summer. The Chief Justice's chambers distribute completed memos to the other chambers participating in the pool. These memos undergo a separate mark up in each chamber, allowing a justice's own clerks to comment on the recommendation made by the clerk who authored the pool memo (Ward and Weiden 2006: 147–49).

New law clerks arrive in July after the Court has taken its summer recess. They undergo a few days of training and supervision by the clerk they replace in a justice's chambers. At the end of that week, the old law clerk moves on and the new law clerk is left to sift through the mound of petitions she has been assigned, oftentimes without guidance from her justice. Edward Lazarus (2005: 27), who had clerked in the U.S. Court of Appeals for the Nineth Circuit before accepting a clerkship in Justice Blackmun's chambers in 1988, recalled this orientation period as “a giant blur, a jumble of shorthand explanations of procedures I couldn't quite grasp, mixed with a number of ‘don't worry, you'll figure it out as you go along.’” Perry (1991: 78) recounts one former clerk's comments on training: “Previous clerks gave you certain pointers, but you didn't get all that detailed of information from them, and frankly, it wasn't all that consistent information.”

When the justices return from their recess, they dispose of all the accumulated petitions at the Long Conference. A small portion of the cases considered at the Long Conference are holdovers from prior conferences with cert. pool memos authored by outgoing clerks. The vast majority of the pool memos assisting the justices at the Long Conference were written during the first month or two of a new clerk's service to the Court. From the justices’ perspective, the Rule of Four primarily defines the certiorari process.⁶ At conference, the justices enjoy a range of options, including granting or denying the request for certiorari review, requesting the Solicitor General's view, or rescheduling discussion of the petition for a later conference. The Court and the parties expend considerable resources hearing a case on the merits. That fact, combined with the limited amount of oral argument days the Supreme Court schedules per term, means that the Court is cautious in granting certiorari.

The ninety-day window to file a cert. petition⁷ leaves litigants some limited flexibility. If a lower court ruling was handed down in midsummer, savvy litigants, perhaps those represented by a member of the Supreme Court bar, could file a petition for certiorari after the Long Conference.⁸ At the same time, petitioners cannot control when counsel for respondent file a brief in opposition. Petitioners must then decide whether to file a reply brief, delaying consideration of the petition by another ten days. Thus, while attorneys may have some discretion that allows them to strategically plan when to file a petition for certiorari and avoid the Long Conference, the likelihood of this strategy's success is subject to a complex web of rules for meeting filing, briefing and case handling deadlines.

Summary statistics presented in Table 1 compare the workload and grant rates of the Court at the Long Conference in recent terms to the combined workload and grant rates of the nine other sittings held throughout the term. Across the five terms listed in Table 1, on average the Court disposed of 5,946 cases during the nine sittings following the Long Conference compared to an average of 1,849 at the Long Conference alone. Cordray and Cordray (2004: 237) report similar summary statistics for selected terms ranging from 1992 to 2002 in their otherwise qualitative account. Perhaps of greater concern than the sheer number of petitions disposed of at a single conference is the disparity in grant rates between the Long Conference (0.591 percent) and conferences held during all other sittings (1.1 percent). A difference of means t-test indicates this disparity is statistically significant at the p < 0.001 level. The data in Table 1 suggest the trend in recent terms is for the Court to docket only a fraction of its agenda at the Long Conference.

Table 1. Supreme Court Agenda-Setting at the Long Conference

These descriptive statistics suggest the possibility of two different certiorari patterns, one at the beginning of the term and one thereafter. This offers no comfort to the litigants, who have very little control over the timing of their appeal. Cordray and Cordray (2004: 211) have described this fluctuation in grant rates as “disturbing,” and Felix Frankfurter (as cited in Cordray and Cordray 2004: 196) cautioned, “[T]he whole operation of the device of certiorari will be seriously affected if selection is determined not by the intrinsic importance of legal issues but by the arbitrary exactions of the size of the docket.” We proceed to investigate whether clerks are partially responsible for what appears to be a seasonal variation in access to justice.

A Theory of Clerk Decision Making on Certiorari

The anecdotal evidence drawn from clerk and practitioner experiences suggests two contradictory hypotheses. One story posits that clerks are “grant averse.” Because they have received only minimal training, they are overly cautious and recommend fewer grants for review in cert. pool memos to the justices written during the summer. Former clerks (Wolf 2013) have spoken of the uncertainty in their institutional role during this period and their desire to preserve their personal reputation with their justice and the Court as a whole. Justice Stevens remarked (as cited in Ward and Weiden 2006: 144) that “you stick your neck out [as a clerk] when you recommend to grant a case. The risk-averse thing to do is to recommend not to take a case. I think it accounts for the lessening of the docket.” The clerks recognize justices will examine cases for the presence of certain cert-worthy characteristics over others. If the clerk recommends a grant for a case which does not have some or any of these identifiable cert-worthy characteristics, justices may discount her recommendation and trust her less in future cases.

Lucas Powe (as cited in Peppers 2006: 116) described his early experience clerking with Justice William O. Douglas, “A not atypical introduction to WOD's style was a summer memorandum informing the law clerk that other employment might better suit his inadequate legal talents.” Without supervision from their justices, clerks express consternation about when, and if, to recommend grants. Ward and Weiden (2006) argue the cert. pool creates intense pressure on the clerks to deny certiorari throughout the term, a pressure grounded in the fact that justices scrutinize grant recommendations more closely and clerks seek to preserve their reputation by not enduring the embarrassment of having a recommended case dismissed as improvidently granted.

The second story suggests precisely the opposite effect. Cordray and Cordray (2004: 228) discounted the role of clerks in explaining the certiorari grant inequity at the Long Conference because, in their view, the inexperience of the clerks would result in more clerk grant recommendations, rather than fewer. They suggest clerks need time to adjust from their prior clerkship on the Court of Appeals, an error correction court, to their new post on a court designed to ensure uniformity of law. Likewise, Perry (1991) proposes that the role of the clerk changes as she moves from a clerkship in the chambers of a circuit court judge to that in the chambers of a Supreme Court justice. He quotes (80) one former clerk's attitudes toward certiorari: “Of course you do have a bit of a change of perspective from the summer when you first start and then as you get into the term. I mean at first you think everything is interesting and certworthy … But then as you go on, you realize what the justices think is certworthy and what they don't.” Another clerk (78) told Perry: “I was used to the Court of Appeals mode of decision making, which is where they basically resolve everything. It took me quite a while to figure out that cert. was not like this.”

In addition, there is evidence that the justices anticipate that the clerks may not have socialized into the different institutional role played by the Supreme Court. Chief Justice Rehnquist (as cited in Mauro 2004) counseled law clerks against selecting cases because of personal interest in outcomes to avoid introducing bias into pool memos. The purpose of random clerk assignments was “to avoid any temptation on the part of law clerks to select for themselves pool memos in cases with respect to which they might not be as neutral as is desirable.” This memo reflects the attitude expressed by one clerk (as cited in Perry 1991: 159): “[I]t takes a while to learn the Supreme Court is not there to right all wrongs in cases.” Thus, the second story would suggest that clerks are looking to recommend more grants during the early days of their clerkship.

Given these conflicting accounts, there is no compelling theoretical basis for hypothesizing one way or the other. We side with the grant-averse prediction because it comports with the literature on justice-clerk agreement on certiorari (Black and Boyd 2012; Stras 2006), clerk reputational concerns (Ward and Weiden 2006), and the small grant rate at the Long Conference. Thus, we hypothesize the probability of a clerk recommending a grant of certiorari is smaller during the period between their arrival at the Supreme Court and the Long Conference than at other times during their tenure as clerk.

The relevant literature on certiorari (Black and Boyd 2012; Brenner and Palmer 1990; Caldeira, Wright, and Zorn 1999; Peppers and Zorn 2008) and the major competing theories of judicial decision-making–attitudinal (Segal and Spaeth 2002), strategic (Knight and Epstein 1997), and institutional (Maltzman, Spriggs, and Wahlbeck 1999)–provide for the remaining ideological, strategic, and legal hypotheses for inclusion in this stage of our analysis. While Bryan (2012) finds no evidence of ideological influence on the clerk's recommendation, we hypothesize that ideological distance serves as a cue to clerks when making a recommendation to grant or deny certiorari. When ideological distance between the clerk and a lower court judge is at its greatest, the differences in policy positions should also be larger. Black and Owens (2012) find the Supreme Court is more likely to review cases decided by ideologically distant lower court judges, so it stands to reason that their clerks will be similarly inclined. The other justices may be aware of this facet of a cert. pool memo and react accordingly. Black and Boyd (2012) find clerk influence in their cert. pool memos has a limited ideological component. In cases of medium cert-worthiness, which constituted 27 percent of their sample, the probability of a justice voting in agreement with the clerk's recommendation declines by 20 percent where an ideologically distant clerk recommends granting review. In cases of low or high cert-worthiness, however, ideological distance between clerk and justice had no statistically significant explanatory value.

Strategic variables implicate a decision-maker's ability to maximize their utility, given the preferences of other decision makers. While no direct negotiation or bargaining takes place over the clerk's recommendation, there is good reason to conceive of the clerk's behavior in the cert. pool as strategic because of concern for preserving personal reputation. If clerks behave strategically in this sense, the probability that a pool clerk will recommend granting certiorari increases in the presence of certain cert-worthy characteristics: when the court below has reversed a decision of the trial court (Ulmer 1984), when amicus curiae file briefs at the pre-certiorari stage, when the case involves an issue of high salience (Provine 1980; Tanenhaus et al. 1963), and when a dissent is filed in the court below (Black and Boyd 2012). We hypothesize clerks will be more likely to recommend a grant when the federal government is the petitioning party and less likely to recommend a grant when the federal government is the respondent (Provine 1980). These variables are among a set that justices themselves use as informational cues, and social scientists have traditionally considered as important indicators of cert-worthiness, but do not necessarily appear in the Court's own rules pertaining to grants of certiorari.

Legal variables refer to specific legal attributes of cases that either preclude the Court from hearing and deciding a case, or relate directly to the Court's role as steward of the federal system and arbiter of the Constitution for the other coequal branches. Bryan (2012) finds cert-worthiness overwhelming explains a clerk's recommendation to grant review. As we noted above, the Supreme Court's Rule 10 states a writ of certiorari involves the use of judicial discretion when cases exhibit certain characteristics. Although the Court acknowledges the Rule 10 criteria are neither dispositive nor a complete list of reasons it might grant review, we draw on the Court's own rules to identify legal variables. We hypothesize that the probability a pool clerk will recommend granting certiorari review increases with the presence of two cases raising substantially the same legal question decided by two or more circuit courts resulting in a difference in the law across jurisdictions (Black and Owens 2009) and when the court below has altered precedent of the Supreme Court (Ulmer 1983, 1984). We also predict a pool clerk will be more likely to recommend a grant when a case raises a unique legal question for which there is no authoritative Court precedent, another Rule 10 criterion.

Data Collection and Variables

Using the Digital Archive of the Papers of Harry A. Blackmun, we created a simple random sample of 2,000 cases from the paid docket spanning the 1987–1993 October Terms.⁹ In designing the sample, we were able to avoid coding any in forma pauperis (IFP) petitions because these cases are assigned a distinctly higher docket number in each term, usually beginning with 5,001. Our sample only includes petitions seeking certiorari review originating from the 11 circuit courts and the Court of Appeals for the District of Columbia. Petitions not meeting these criteria were discarded without replacement, yielding a final sample of 1,179.

We excluded IFP petition because the Court grants review to less than one percent of these cases per term. These cases often raise narrow legal issues, most often involving an inmate filing for habeas review. We excluded cases originating from the Court of Appeals for the Federal Circuit because the vast majority of those judges do not have a Judicial Common Space (JCS) ideology score (Epstein et al. 2007), which is needed to construct an ideological distance measure. For the same reason, we excluded cases petitioned from state supreme courts.

For each case, we coded both the clerk's cert. pool memo and the docket sheet, which lists the decision of the Court to grant or deny certiorari review. For the pool memos, we code for measures of cert-worthiness identified by the Court in its rules, informal norms, and case characteristics, and the clerk's recommendations. The most common form of certiorari behavior for justices and clerks alike is to deny a petition (or recommend a denial). We code denials as 0 and any positive recommendation from the clerks as 1. The strongest positive recommendation from the clerks is a grant recommendation, but clerks sometimes recommend consulting the views of the solicitor general or calling for a response brief. These other options are sometimes accompanied with the additional recommendation “…with a view to grant.” Finally, clerks may also recommend a GVR–grant, vacate, and remand.

We coded any and all positive recommendations with the same value because clerks face a threshold decision. When evaluating a given cert. petition, the clerk must determine whether the combination of ideological, strategic, and legal calculations are sufficient to overcome some deep institutional presumptions against granting review: cases are fungible, the Court's resources are limited, and the Court's role is not to correct errors (Perry 1991). Any cert. petition that overcomes this threshold has succeeded, even if the clerk's recommendation falls short of an unequivocal grant. When coding docket sheets, we employ the same dichotomy as at the clerk stage: a 0 represents a denial, while any positive action from the Court is coded as a 1.

As an alternative measure of clerk behavior, we coded clerk recommendations on an ordinal scale. Any denial recommendation was coded as zero. A recommendation to hold the case for future consideration was coded as 1. Some cases included both a primary and secondary clerk recommendation. Cases with a primary recommendation of denial alongside a positive secondary recommendation (such as to call for the view of the solicitor general) were also coded as 1. A recommendation to call for a response brief or call for the view of the solicitor general was coded as a 2. Recommendations to grant a case or GVR (grant, vacate, and remand) were coded as 3.

In addition, we recorded the date on which the cert. pool memo was circulated to the justices, as well as the date of the conference the justices considered the petition. These variables inform when in a clerk's tenure the memo was written, which is connected to our interest in the Long Conference. We coded the name of the clerk authoring the memo, and the justice for whom he or she works.¹⁰ Finally, we recorded the names of the judges on the lower court that wrote opinions in each case. Based on new research on this issue (Kromphardt 2015), we assume law clerks will take the ideology of the judge they clerked for the year before they came to the Supreme Court. The most common prior experience for Supreme Court clerks is with a federal appellate judge, though a few clerks worked with federal district court judges or state supreme court justices.¹¹ The measure of ideological distance between the clerk and the lower court opinion author is the absolute value of the difference of these two JCS scores. We define salient legal issues as First Amendment, privacy, and civil rights cases, as defined by the coding protocols of the Supreme Court Database. Black and Boyd (2012) also controlled for these types of cases. We look to the clerk's memo for information concerning the cert-worthiness of each petition. For example, if the clerk alleges the existence of a circuit split, we code that variable as a 1. Further information about our coding protocols is available in Table 2.

Table 2. Variable Descriptions

Supreme Court agenda-setting is a multistage process.¹² In the first stage, clerks construct a recommendation on certiorari.¹³ Thus, in our first stage analysis we treat the clerk recommendation as a dependent variable. In stage two, when the justices vote in conference to grant or deny certiorari, we utilize the clerk recommendation as an independent variable to examine behavior at the Long Conference. We recognize a potential drawback to employing the Court as the unit of analysis: three justices in our sample–Brennan, Marshall, and Stevens–were not members of the cert. pool. Thus, the signal provided by a clerk's recommendation is not uniformly distributed across the court. All three of these justices occupy the ideological left of the Court, which could further complicate matters to the degree that ideology structures agenda-setting behavior.

Nonetheless, the Court remains the appropriate level of analysis for three reasons. First, the research question central to this analysis involves predictions about the number of cases docketed by the Court, not the behavior of the justices themselves. Second, agenda-setting behavior is strategic and not simply in terms of being forward-looking to the merits stage (Caldeira et al. 1999). Other norms and practices, such as the Rule of Four, Join-3, the discuss list, discussion order, and voting order structure the agenda-setting game (Cross and Lindquist 2006; Hall 1989). Unless political scientists could observe conference discussions directly, any model of certiorari votes will be incomplete to a significant degree. Third, even if all justices were members of the cert. pool, the justices receive other signals not captured in the Blackmun Archives. As previously mentioned, when a pool justice receives a memo written by a clerk from another pool chamber, the justice assigns one of her own clerks to provide additional information or advice.

Methods

Our original formulation of a clerk recommendation is dichotomous. Thus, we employ logistic regression with robust standard errors clustered on each Court term to model the clerk's recommendation on certiorari. The Court's decision to grant to deny certiorari is also dichotomous, and we employ a similar logistic regression analysis on this stage of agenda-setting behavior. Cases can vary dramatically in their cert-worthiness, which can bias the regression estimates of both a clerk's recommendation and the Court's decision on certiorari. To fully understand the effect of the Long Conference, we utilize nearest neighbor matching from the treatment effects package in Stata.

In using this quasiexperimental method, clerk behavior at the Long Conference becomes the treatment group while clerk behavior at other times in the term serves as the control group. Nearest neighbor matching imputes the missing potential outcome for each observation using an average of the outcomes of similar observations that received the treatment (Nichols 2007). A weighted function of the covariates determines observational similarity. The average of the difference between the observed and imputed potential outcomes for each observation is the average treatment effect. In other words, nearest neighbor matching pairs cases at the Long Conference with similar levels of cert-worthiness to cases at other conferences and then compares the difference in clerk recommendation tendencies. Nearest neighbor matching has been used in other studies to untangle the effect of gender (Boyd, Epstein, and Martin 2010) and religion (Blake 2012) on judicial decision making.

It is also possible that in the run up to the Long Conference, clerks not only encounter difficulties in determining their recommendation, but also in correctly diagnosing the cert-worthiness of a case. Because it is difficult for litigants to time their petitions strategically, and the timing of the disposal of cases at the Court of Appeals does not vary by cert-worthiness, we assume levels of cert-worthiness to be randomly distributed across the Court's term. Given this assumption, if the clerks have not learned to recognize cert-worthy characteristics early in their tenure, Long Conference memos should report significantly different levels of cert-worthiness than clerk memos written later in the term.

To determine if the clerks’ ability to diagnose cert-worthiness increases over their tenure, we utilize the same method as Black and Boyd (2012) to generate a single measure of cert-worthiness. That is, we estimated a logistic regression model, reprinted in Table 5 of the Appendix, which uses the Court's decision to grant or deny as the dependent variable and the cert-worthiness variables found in Table 2 as independent variables. We then generated the predicted probability that each petition would be granted review, and these predictions serve our single measure of cert-worthiness. We conducted an analysis of variance to examine whether cert-worthiness varied across the months of a Court's term. The F-statistic of 1.8 is not statistically significant, which suggests that the clerk's ability to identify cert-worthy case characteristics does not vary across their tenure.

Analytical Results

We begin the analysis by examining the determinants of a clerk's recommendation. The first column of Table 3 reports the results of a logistic regression, Model 1. One immediately notes the smaller number of observations in this model compared to the original sample of 2,000. In addition to discarding cases appealed from state courts and the Court of Appeals for the Federal Circuit, we were unable to locate JCS scores for many of the circuit court judges in our sample. As a whole, the model performs well, as indicated by the pseudo-R ² value of 0.367.

Table 3. Logistic Regression Model of Clerk Recommendation to Grant Certiorari

Robust standard errors clustered on Court Term.

^* p < 0.05

The model also provides support for many of our strategic and legal hypotheses. On the strategic dimension, it appears that the clerks like the justices, appreciate the special role between the Court and the Office of Solicitor General, and are more likely to recommend a grant when the federal government petitions for review. The reverse, however, did not occur. That is, when the federal government is a respondent, the clerks are not significantly disinclined to recommend a grant, though the coefficient was in the expected direction. The clerks exhibit sensitivity to signals sent not only by the federal government, but interest groups as well. Clerks are significantly more likely to recommend a grant in cases in which amicus curiae filed one or more certiorari-stage briefs. Lower courts send signals when a court of appeals judge writes a dissent, or when the court of appeals reverses the district court. However, neither of these variables significantly influence the clerk's decision making.

It also appears that clerks understand the different institutional role they play on the Supreme Court, compared to their prior post on a court of appeal. Careful attention to circuit splits and alterations of precedent constitute important tools that serve the Supreme Court's institutional function of maintaining uniformity in federal law. The likelihood of a clerk recommending a grant of certiorari increases significantly when they detect a circuit split and in cases in which they allege the lower court has altered Supreme Court precedent. While Rule 10 defines part of the Court's role as handling legal issues of great importance, clerks are no more likely to recommend a grant in cases raising salient legal issues. Additionally, the unique legal issue variable perfectly predicted a clerk recommending a denial. This nonfinding suggests the clerks subscribe to the notion Perry (1991: 230–34) described as “percolation”–that the Supreme Court need not involve itself in a case until several lower courts have weighed in over time, notwithstanding Rule 10's instruction to the contrary.

The analysis did not support the ideology prediction hypothesis. Though the ideological distance between a clerk and the opinion writer below took the expected positive coefficient, the finding did not achieve statistical significance in any of the analytical models. Given the strength of some of the cert-worthiness variables, it appears that, in many instances, clerks are not left with much discretion to base their recommendation on ideological factors. Perry (1991: 57–59) related ample evidence that the clerks view a pool memo, intended for a diverse set of justices rather than one, as a formalized function. While most clerks strive for impartiality, the clerks of one particular justice (the infamous Justice A) tended to write strategically motivated pool memos, requiring clerks from other chambers in the pool to carefully go back through each of those cases to eliminate bias from the reading of the fact pattern and even interpretation of precedent. The strong finding regarding cert-worthiness and the nonfinding on ideology is similar to those of Bryan (2012).

Finally, the analysis provides support for our temporal prediction. Compared to all other times in their tenure, clerks are significantly less likely to recommend a grant at the Long Conference. We consider whether the clerks increase their recommendation rate over the course of their tenure, perhaps because they increasingly adjust to institutional expectations. Thus, Model 2 of Table 3 omits the Long Conference variable of Model 1 in favor of Months in Clerkship. While the coefficient of Months in Clerkship is positive, the finding is not statistically significant. Measuring cert. pool recommendations as a linear function of time assumes that clerks are likely to increase their grant propensity at the same rate between their 11th and 12th months on the job as they are between their first and second months. This is a strong assumption, considering they receive no feedback from the justices before the Long Conference and then regular feedback after each subsequent conference.

To relax this assumption, we substitute the natural log of Months in Clerkship in Model 3 of Table 3 as the measure of the temporal variable. A log-linear approach to job experience assumes clerks adjust to their role early in the clerkship and do not alter their behavior as much down the road. Unlike the linear measure, the natural log of Months in Clerkship has both a positive and statistically significant coefficient. This finding suggests that, while some role adjustment may take place for the clerks after the Long Conference, they soon learn what the justices expect from them in pool memos. The only other important difference between the findings of Model 3 and Models 1 and 2 is that the government petitioner variable in Model 3 did not achieve statistical significance.

Table 6 in the Appendix reports the results of two alternate modeling strategies for analyzing a clerk's pool memo recommendation: ordered logit and rare events logit. The ordered logit model utilizes the ordinal coding scheme for a recommendation described in the data section. Traditional logistic regression assumes an equal probability between a positive and negative response in the dependent variable, which is a strong assumption when grants of cert are so rare. Rare events logit (King and Zeng 2001) corrects for this potential source of bias, and it has been employed in many prior studies of Supreme Court behavior (e.g., Blake and Hacker 2010; Johnson, Spriggs, and Wahlbeck 2005; Hansford 2004). The rare events logit model performed similarly to the traditional logit analysis, except that the government petitioner variable failed to achieve statistical significance.¹⁴ The ordered logit model did not produce a statistically significant finding for the Long Conference variable, while all the other findings in Table 6 are consistent with those of the traditional logit analysis.

The impact of the clerk's initial risk-averse behavior on the Supreme Court's agenda requires further consideration. First, which justice a clerk serves might conceivably influence grant recommendation rates. Thus, we conducted a logistic regression model of clerk recommendation, reprinted in Table 7 of the Appendix, with dummy variables for clerks from each pool chamber and the single measure of cert-worthiness as a control. Once again, cert-worthiness performs strongly as expected. Further, the model suggests that only clerks from the chambers of Justice White behave differently, and they are more likely to recommend grants than other pool chambers.

Second, it is possible that justices react differently to a grant recommendation from the clerks at the Long Conference versus other times of the year. Figure 1 visualizes the results of a logistic regression model of the Court's decision on certiorari using the same list of cert-worthiness variables as the clerk recommendation model.¹⁵ When clerks recommend a denial, the justices rarely disagree with that diagnosis. However, the justices are slightly more likely to overrule a denial recommendation from the clerks at the Long Conference than during the rest of the Term. This finding suggests the justices are cognizant of clerk's initial risk-averseness and are willing to overrule them by granting some cases at the Long Conference that the clerks had thought unworthy. Conversely, the right side of the figure indicates the justices are slightly more likely to agree with a clerk grant recommendation at the Long Conference than at other conferences.

Figure 1. Court's Response to Clerk Recommendations, By Conference, With 95% CIs.

Since the justices view grant recommendations from the clerks in essentially the same way regardless of the time period, we can proceed to the matching analysis. The first question to which nearest neighbor matching can lend purchase is the difference in clerk behavior at the Long Conference compared to other times during the year. We run nearest neighbor matching with the clerk recommendation as the dependent variable, the multitude of cert-worthiness variables in Table 2 as independent variables, and the Long Conference as the treatment variable. The results, presented on the top row of Table 4, indicate that the clerk's overall grant recommendation rate is 7.7 cases per 100.¹⁶ Holding case characteristics constant, the average treatment effect of the Long Conference is to decrease the grant recommendation rate by 2.8 cases per 100. This result achieves both statistical and substantive significance. In cases of similar cert-worthiness, clerks are 36 percent less likely to recommend a grant because of the timing of the cert. petition.

Table 4. Nearest-Neighbor Matching of Certiorari Processes

^* p < 0.05

We proceed to examine the effect of this finding on the Court's agenda by creating a counterfactual: what would have happened if the clerk's grant recommendation rate remained constant over time? Our sample included 204 cases from the Long Conference, 189 of them being denial recommendations. Putting the 2.8 percent reduction into real terms means that 5.26 of these 189 denial recommendations would have become grant recommendations in this counterfactual. In view of the fact that certiorari is a two-stage process, however, it would be incorrect to assume that the justices would have agreed with every additional grant recommendation. Thus, we perform a second matching analysis examining the influence of a clerk's grant recommendation at the Long Conference.

This nearest neighbor model uses the Court's cert. decision as the dependent variable, the cert-worthiness variables listed in Table 2 as independent variables, and the clerk recommendation at the treatment variable. This model considers only those cases the justices discuss at the Long Conference. These results are presented in the bottom row of Table 4. In our sample, the Court granted certiorari in 11.8 percent of petitions considered at the Long Conference. Again, holding cert-worthiness constant allows for an isolation of the average treatment effect of a clerk's grant recommendation. At the Long Conference, if a clerk recommends a grant, the justices will follow suit and grant certiorari 70.9 percent of the time.

The justices granted certiorari in 24 Long Conference cases in our sample. If supplied with 5.26 more grant recommendations, the justices likely would have granted 3.73 of these cases. These additional cases would represent a 15.5 percent increase in the size of the Court's Long Conference agenda. Put another way, because of the clerk's risk-averse behavior before the Long Conference, litigants have a 15.5 percent worse chance of getting a case granted if their petition arrives at the Supreme Court over the summer, compared to any other time of year. This finding has a margin of error of 6.9 percent.¹⁷ Thus, in the best case scenario for the litigants, clerk reticence to recommend a grant at the Long Conference only reduces their chances of getting their case before the Court by 8.6 percent. Under the worst case scenario, the clerk's grant-averseness reduces litigant chances at certiorari by 22.4 percent. Either extreme represents a grave administrative issue for the Court when its primary agenda-setting mechanism functions, at least in part, on factors unrelated to the qualities of the cases brought to it.

Conclusion and Implications

Docket management, particularly for a court that has almost absolute discretion over the composition of that docket, poses a set of problems for administration of justice, raising concerns about the unbiasedness of selection procedures impacting overall fairness to the litigants. Our analysis of Supreme Court agenda-setting confirms the suspicion of practitioners: not only do their petitions for certiorari have less chance of being granted review at the Long Conference than at other conferences throughout a term, but we can explain this anomaly as a result of clerk behavior. Although cert. grant rates vary within the terms under study, at no other conference does clerk behavior account for a surge or decline in the grant rate of the Court. Indeed, at no other conference do clerks show the reticence to recommend a grant that they do at the Long Conference.

A simple administrative remedy exists for this anomaly. Implementing a program that provides more and better training for new clerks and direct supervision by the justices during the weeks after the clerks’ arrival at the Court would eliminate one facet of Supreme Court agenda-setting that produces an improper negative impact on the litigants. Such an adjustment would constitute a measured response by the Chief Justice of the United States to the findings of this study and the concerns of the attorneys who practice at the Supreme Court bar. Eliminating the anomaly at the Long Conference, however, would likely affect grant rates across the rest of each term. Rather than creating a dearth of grants at some other point in the term, we contend that a higher grant rate at the Long Conference may result in a more even distribution of grant rates across each term.

Because the Supreme Court's docket has declined from an historical high of nearly 160 cases given full treatment each term to approximately 80 cases in recent terms, the Court has spread oral argument days evenly across each term. To achieve this goal, there are significant spikes in grant rates at conferences throughout each term, typically in late October, November, and June conferences when the justices recognize a need to replenish their supply of cases. Cordray and Cordray (2004: 207) noted that grant rates rise and fall because of the Supreme Court's “conscious concern to fill out its docket as much as possible and to avoid having to cancel argument days in the middle of the term.”

One example of how maintaining a schedule of oral arguments spread evenly throughout a term affects litigants relates to submission of written briefs. To maintain its schedule, the Court must regularly waive its rules for what constitutes adequate time for briefing, thus putting pressure on attorneys to expedite the filing of written briefs. This phenomenon occurs for approximately one-third of cases granted certiorari (Cordray and Cordray 2004: 221). We suggest that a higher grant rate at the Long Conference would eliminate at least part of the need to increase grant rates artificially at other times during a term. This would, in turn, reduce pressure on the Solicitor General's office and the litigants to expedite their brief writing. Thus, a revision to the clerk training procedures would positively impact the institution and address what amounts to a biased and unnecessary variation in the administration of justice.