II. From Full Consideration to the Discuss List

During the hearings for the Judges’ Bill of 1925, Justice Van Devanter stated that, as a matter of course, each justice read and prepared a memorandum on each petition for certiorari, and then the full Court considered each case in conference. And for a few years after the passage of the Judges’ Bill, the Court actually did give all petitions at least perfunctory discussion at conference (see generally Hartnett Reference Hartnett2000). But by the mid-1930s, Chief Justice Hughes had established a procedure for “special listing” of cases that the conference would not consider and that would be denied without discussion. The exact date of the change is not known, but references to cases being systematically denied without discussion exist as early as 1931, and it is clear that special listing was systematically implemented by 1935 (Provine Reference Provine1980, 26–30; Caldeira and Wright Reference Caldeira and Wright1990, 810; Ward and Weiden Reference Ward and Weiden2007, 111–12).Footnote ³ Presumably, as the justices accumulated more experience under the new regime of the Judges’ Bill and as writs of certiorari became the dominant part of the Court’s workload, the impetus for the change was Hughes and his colleagues’ recognition that a significant proportion of petitions coming to the Court were nonmeritorious.

From the mid-1930s forward, then, the chief justice and his clerks would make up and circulate this special list of cases, also known as the “dead list,” soon after the briefs and records for cases circulated to the associate justices; then, as the individual chambers reviewed cases, any of the associate justices could remove a case from the list.Footnote ⁴ Since the 1950s (the exact year is unclear), the Court has reversed the formal procedure, and now the chief justice circulates a “discuss list,” the cases to be taken up, instead of a “special list” or “dead list,” the cases to be denied (Stevens Reference Stevens1983). Cases not on the discuss list are denied without discussion. In what follows, we use discuss list to refer to both the special and discuss lists when the distinction is not important.Footnote ⁵

III. Theory and Hypotheses

Our initial theoretical expectations derive from Caldeira and Wright (Reference Caldeira and Wright1990), who divide the set of covariates expected to influence the makeup of the discuss list into those of high quality (i.e., providing valuable, reliable information about the merits of a petition) and those of low quality (i.e., providing only weak or noisy information about the merits of a petition). High-quality indicators include the presence of actual or “square” conflict (i.e., conflict among the circuits or any other conflict described in Supreme Court Rule 10), the United States as petitioner, the ideological direction of the lower-court decision, and the presence of amicus briefs. Low-quality indicators include the presence of a dissent in the court below (usually a three-judge panel), an intermediate reversal (whether the court below the Supreme Court reversed the court below it), issue area, and alleged conflict.

The classification of indicators into these two categories is relatively straightforward, but the role played by the ideological direction of the lower-court decision at the discuss-list stage deserves some comment. On the one hand, no political scientist will be surprised that ideology plays a significant role at the agenda-setting stage (e.g., Caldeira, Wright, and Zorn Reference Caldeira, Wright and Zorn1999).Footnote ⁶ On the other hand, ideological “manipulation” of the discuss list by the chief justice is normatively disfavored. The discuss list’s stated purpose is to eliminate frivolous petitions—a “housekeeping function,” as it were. Justice Douglas is typical in calling it an “efficiency device” that eliminates “patently frivolous cases that [do not] present any substantial question, that [are] not even worthy of conference discussion.”Footnote ⁷ That the justices and clerks interviewed in Perry (Reference Perry1991, 85–91) for the most part denied that Burger engaged in ideological manipulation of the discuss list (cf. Caldeira and Wright Reference Caldeira and Wright1990, 826) suggests the hold of this normative ideal.

Normative claims aside, prior research suggests that, empirically, at the discuss-list stage, each covariate, whether high quality or not, should have some effect (Caldeira and Wright Reference Caldeira and Wright1990).Footnote ⁸ Here, however, we are primarily interested in a different question: Do Hughes, Warren, and Burger put differential weight on indicators in selecting cases for discussion?

We offer several theoretical expectations for differences among chief justices. Two hypotheses are quite straightforward. First, it is well understood that civil liberties issues (including criminal procedure and civil rights) concerned the Court little before Vinson’s chief justiceship and that they were a particular emphasis for the Warren Court (Schwartz Reference Schwartz1983, 215–61; Urofsky Reference Urofsky1999; Belknap Reference Belknap2007, 150–86). Insofar as we can assume that the chief justice’s agenda setting played a part in these differences, we hypothesize that the presence of civil liberties as the primary issue in a case will have a greater effect on Warren than on Hughes and, to a lesser extent, on Burger. Second, as we note in table 1 and as Caldeira and Wright (Reference Caldeira and Wright1988) observe, the number of amicus briefs on certiorari has increased over time; this leads us to hypothesize that the signaling value of amicus briefs will decrease over the course of the period we cover. The idea is that when amicus briefs are rare, they are “news” and thus informative, but as they become more numerous and as participants realize the impact they have on the Court, they become ordinary and thus less informative.Footnote ⁹ Therefore, we hypothesize that the presence of amicus briefs will have the greatest effect on Hughes’s discuss list, when amicus briefs were rare, and the smallest effect on Burger’s list, when these briefs were more plentiful.

Table 1.

Relative Frequencies of Each Covariate, for OT 1939, 1968, and 1982

Note.—For example, 9% of non-IFP cert petitions in OT 1939 were filed by the United States.

Historical and anecdotal accounts of Hughes, Warren, and Burger as chief justices inform other expectations. Chief Justice Hughes had a reputation for excellent administrative skills, mastery of the Court’s caseload, and skills as a leader across the board, and he also had the advantage of his service as an associate justice and as a lawyer practicing before the Court (McElwain Reference McElwain1949; Ross Reference Ross2007; Danelski Reference Danelski, Ward and Danelski2016). His legal acumen was held in unanimously high regard: typically, he is described as presenting cases “comprehensively, precisely, and impartially” (Goldstein Reference Goldstein2011, 724). Of course, during his tenure as chief justice, the Court’s caseload was small in comparison to the post–World War II period, and there were relatively few in forma pauperis (IFP) petitions; also, until his final two terms, he continued to preside over a relatively consensual group. Yet he and his associates had small staffs, did most of their own work, and had to deal with briefs of tremendously variable quality, arguments from lawyers of heterogeneous skills, and a carryover of diversity cases even after Erie v. Tompkins (304 U.S. 64 [1938]).

In contrast to Hughes, Warren came to the Court as an administrator. His days as a practicing lawyer were in the 1930s, and, accordingly, he could not rely on his intellectual ability or experience to establish mastery of the cases and issues coming to the Court (Schwartz Reference Schwartz1983; Belknap Reference Belknap2007). He also faced a larger caseload, a rising tide of IFPs (see, e.g., Boskey Reference Boskey1946), a stable of strong and contentious associate justices, a bar of mixed quality, and a more controversial mix of cases. To his advantage, he was well organized, had a larger staff than did Hughes (see, e.g., Peppers Reference Peppers2006, 145–205), had the advantage of a well-established routine of handling the caseload inherited from Hughes and Vinson, and, like Hughes, could draw on his moral authority and reputation for fairness.

In contrast to and unlike Hughes and Warren, Burger came to the Court with long experience as a federal appellate judge and with an interest in, and clear ideas about, how to administer the Court. He also had the advantage by the mid-1970s of a much larger staff than his predecessors. To his disadvantage, he faced an exploding caseload, of both paid and unpaid cases; a set of strong-minded, confident, and outspoken associates; and more and more politically controversial and contentious issues. And he followed a beloved chief. He quickly established a reputation for pushing his own agenda and at times for incompetence; generally, he was not regarded as trustworthy. Specifically, Burger’s normatively dubious ideological machinations in conference (strategic passing, misrecording of votes, manipulation of opinion assignment) have been widely reported (e.g., Woodward and Armstrong Reference Woodward and Armstrong1979; Schwartz Reference Schwartz1990; Johnson, Spriggs, and Wahlbeck Reference Johnson, Spriggs and Wahlbeck2005; Goldstein Reference Goldstein2011). Indeed, there is at least an oblique suggestion by a law clerk that Burger engaged in similar behavior when composing the discuss list (Perry Reference Perry1991, 86). Accordingly, we have grounds to hypothesize that ideological compatibility with the decision below will affect Burger’s decision to slate a case for discussion to a greater degree than it will affect Hughes’s decision or perhaps Warren’s decision.

Because qualitative and historical scholarship holds Hughes’s skills as a lawyer in particularly high regard but assesses the abilities of Warren and (particularly) Burger less positively, two further hypotheses are suggested. We hypothesize that the presence of actual conflict will affect Hughes’s selection of cases the most and Burger’s the least—not only because Hughes had a sharper eye but also because he likely valued the presence of conflict more than did the others. Relatedly, we hypothesize that alleged conflict will affect Burger the most and Hughes the least.

To derive expectations about dissent below and intermediate reversal (i.e., reversal in the court below), we consider the relative prevalence of these indicators among petitions to the Court, across the three terms (see table 1). Both of these indicators are considered “low cost” (i.e., easily discerned) and “low value” (Caldeira and Wright Reference Caldeira and Wright1990). Still, their value may vary across terms as a function of their relative prevalence—a signal that is relatively rare may have greater probative value. We see, however, that the relative prevalence of these indicators is very similar across terms. Accordingly, we propose that the effect of dissent below and intermediate reversal will be similar across terms.

Finally, we turn to the United States as petitioner. Each of the three chiefs were—somewhat roughly speaking in the case of Hughes—ideologically compatible with the contemporaneous presidential administration in the term we consider.Footnote ¹⁰ Existing studies have consistently indicated that the petitions by the United States are much more likely to be granted and—in the few terms for which we have evidence—discussed. Thus, we hypothesize no significant differences in effects for a US petitioner on Hughes’s, Warren’s, or Burger’s discuss list.Footnote ¹¹

IV. Data and Measures

For the 1939 term, for information on the relatively objective, descriptive features of cases, we relied on petitions and briefs in opposition in Supreme Court Briefs and Records (Hein), lower-court opinions, and United States Law Week. For data on subjective indicators, we read and coded the “cert memoranda” in the papers of William O. Douglas (Library of Congress), primarily, and, in the absence of them, those in the papers of Stanley F. Reed (University of Kentucky).

For the 1968 term, we gathered our data from Records and Briefs of the Supreme Court on microfiche (issue area, presence of amici curiae, the United States as a petitioner or respondent, reversals between or dissents in the lower courts), United States Law Week (dates of actions, resolutions of cases), and, for conflict and allegations of conflict, Marshall’s certiorari memoranda. The data for the OT 1982 are from, and as described in, Caldeira and Wright (Reference Caldeira and Wright1990).

We concisely describe the operationalization of our variables (see Caldeira and Wright [Reference Caldeira and Wright1990] and Caldeira and Lempert [Reference Caldeira and Lempert2017] for additional details). Our outcome variable, discuss, is whether the chief justice did (= 1) or did not (= 0) set a case for discussion. US petitioner is coded 1 if the “United States,” a federal agency, or its representative (in an official capacity) is one of the petitioners. Intermediate reversal is coded 1 if the court immediately below (nearly always either a federal court of appeal or state supreme court) reversed the lower court (usually a trial court, less often an agency). Actual conflict is coded 1 if the case involved a square conflict between two or more cases occurring in different circuits or state supreme courts, or between the lower court and the Supreme Court (i.e., the conflicts enumerated in Supreme Court Rule 10 [formerly Rule 38]). For OT 1939 and 1968, we relied on assessments in certiorari memoranda (primarily those of Douglas and Marshall); for OT 1982, we relied on the New York University Law Review’s Supreme Court Project (NYU Law Review Supreme Court Project 1984a, 1984b) to code this variable. Alleged conflict is coded 1 if the clerk (OT 1939 and 1968) or the NYU Law Review’s Supreme Court Project (OT 1982) notes that a petitioner has alleged a conflict enumerated in Rule 10.

Civil liberties is coded 1 if the primary substantive issue area of the case involves civil liberties (including criminal procedure and civil rights). Compatible ideology with decision below equals 1 if the decision below is liberal (following generally the criteria in Spaeth [Reference Spaeth2007]) in 1939 and 1968 and if it is conservative in 1982; in other words, we consider Hughes and Warren attitudinally liberal and Burger conservative. Obviously, we recognize that this categorization of Hughes is contestable, but in OT 1939 Hughes voted on the merits to reverse 39 of 65 conservative lower-court decisions and to affirm 44 of 61 liberal lower-court decisions (Spaeth et al. Reference Spaeth, Epstein, Segal, Ruger, Martin and Benesh2017). And as we describe in what follows, our interpretation of relevant results is robust to categorization of Hughes as conservative. We omit observations in which the lower-court decision has no discernible ideological implication. If amicus curiae briefs in support of or opposed to certiorari appear in a case, we code amicus present as 1, and 0 otherwise. Finally, dissent below is coded 1 if, in the court immediately below the Supreme Court, one or more judges dissented.

V. Method

Our primary question is how the effect of each (potential) influence on the makeup of the discuss list varies across the terms 1939, 1968, and 1982. For example, does the presence of the United States as a petitioner influence Hughes more than Warren or Burger? Thus, generically, we are interested in comparing effect sizes across groups. With binary outcomes, the calculation of effect sizes to be used in such a comparison involves subtle and more difficult methodological choices than commonly recognized. The two prominent approaches to calculating effect sizes after modeling binary outcomes are the “representative-case” method and the “observed-value” method (Hanmer and Kalkan Reference Hanmer and Ozan Kalkan2013). The former calculates effect sizes by holding “control” covariates constant at some representative or interesting set of values; the latter sets “control” covariates at their in-sample observed values and calculates an effect size by averaging over each observation in the sample (see, e.g., Hanmer and Kalkan Reference Hanmer and Ozan Kalkan2013; Long and Mustillo Reference Long and Mustillo2019). We argue in what follows for a different approach to calculating effect size differences across groups that bases such calculations on the comparison of units with similar “baseline probabilities” of a positive outcome. An informal sketch of our argument follows; specifics and the information about the Stata program that implements our recommendation are in the appendix.

To compare effect sizes across groups of some variable Z on a binary outcome Y, it is typical to calculate the change in $\mathrm{Pr}(y=1)$ as z goes from some baseline value to some reference value for each group and then calculate the size and significance of the difference between these two effects. Because models for binary outcomes (in our case, a logit) are nonlinear, the effect size depends not only on the estimated coefficients (as in ordinary least squares) but also on the values that the other independent variables X take on. Thus, an important, but perhaps undertheorized, question is how to set x for a given substantive inquiry. Define $\mathrm{Pr}(y=1|z=\mathrm{baseline value},x,\mathrm{group}=g)$ as the baseline probability for group g. Observe that both common approaches to setting x for calculation of effect sizes—the representative-case and the observed-value approaches—may entail calculations that are based on very different baseline probabilities across groups, if x is fixed across groups (as it is in all applications of which we are aware). The central issue we address is that effect size comparisons based on substantially different baseline probabilities may be strongly influenced by the nonlinear functional form of the logit model (i.e., where on the logit curve the baseline probability for each group is located). We argue that for many substantive inquiries, a more appropriate comparison is to set the values of the control covariates x so that baseline probabilities are as similar as possible across groups.

Consider our application. It is unsurprising that (even conditional on covariates) the baseline probability that the chief selects a case for discussion is greater in 1939 than in 1968, which is greater than in 1982. The increasing number of petitions facing a chief, combined with a relatively small number of petitions granted and heard, ensures that cases with comparable attributes will have different probabilities of being granted (Caldeira and Lempert Reference Caldeira and Lempert2017) and discussed. Whether one considers a representative case or averages over observations via the observed-value approach, baseline probabilities of a positive outcome will be significantly different across terms. If one is interested in comparing the effects of a covariate across terms, we argue, the appropriate comparison is based not on cases with identical attributes and very different baseline probabilities of being discussed but rather on cases with very similar baseline probabilities of being discussed, even if their attributes differ. Thus, as we do in what follows, one might compare how the presence of actual conflict changes the probability of making the discuss list for a case with an approximately 50% chance of being discussed ceteris paribus, for Hughes, Warren, and Burger.

VII. Discussion and Conclusion

Although we have found support for several of our hypotheses, a few things have surprised us. Here, we speculate about potential explanations for these results. First, we found that although the presence of the United States as a petitioner still had a very substantial effect on Burger’s decision to list a case, it carried more weight under Hughes and Warren. One possible explanation is a decrease in the standing of the solicitor general during the early Reagan years; as Caplan (Reference Caplan1988, 264–66) reports, the justices, across the ideological spectrum, perceived a decline in the commitment of the solicitor general’s office to its traditional strategy of litigation, of reliance on the small set of criteria on which the Court itself relies. Burger, in particular, is described objecting to Solicitor General Rex Lee’s and his office’s willingness “to be a spokesman for the ‘reactionary’ Reagan Administration” (265). More generally, we know empirically that the solicitor general has pursued somewhat different strategies of petitioning over time—filing 56 petitions in OT 1939, 30 in OT 1968, and 48 in OT 1982—with varying degrees of success (see, e.g., Cordray and Cordray Reference Cordray and Cordray2010).

Second, compared to Hughes and Burger, Warren was more willing to set cases for discussion on the mere allegation of conflict, perhaps as a result of his more general openness to considering important policy issues in the absence of traditional criteria such as square conflict. For one, the Warren Court devoted a large share of its plenary agenda to cases of state and federal criminal procedures, many of which came to the Court as IFPs. We know, of course, that the justices vary considerably in their propensity to vote to grant certiorari, from Byron R. White, who at least toward the end of his career made a point of dissenting in dozens of cases he thought the Court should hear, to Thurgood Marshall, who in his later years seldom voted for cert, believing that the conservative majority would simply engage in “mischief” if permitted to decide a case on the merits.

Finally, we reported some decrease in the effect of actual conflict from Hughes to Warren to Burger, although none of these differences was statistically significant. The decline in the “punch” of actual conflict may well be a function of the capabilities of the three chief justices, as we hypothesized. Yet, in the hearings on the Judges’ Bill of 1925, the justices testified that the Court would grant certiorari on “conflicts” as a matter of course, and Robertson and Kirkham (Reference Robertson and Kirkham1936) and Stern and Gressman (Reference Stern and Gressman1950) repeated this claim. It remains an open question whether the Court ever granted certiorari on all conflicts, as the justices had warranted, but we know from our data that as early as OT 1939, the Court denied certiorari on a substantial number of square conflicts. Stern (Reference Stern1953), of Stern and Gressman’s Supreme Court Practice fame, offers an alternative explanation. He suggests that the Supreme Court, as time passed after the passage of the Judges’ Bill, has become less likely to grant certiorari in the presence of conflicts that (a) are no longer live (e.g., because resolved by statutory amendment), (b) involve a case that would be decided by the Court on grounds that would not resolve the conflict, and (c) involve a case that is at an interlocutory stage. And it is possible that the Court—and thus the chief justice making up the discuss list—has placed additional emphasis on such exceptions and perhaps added others.

A visitor from the Supreme Court of the 1930s would scarcely recognize the Court of the 1980s, let alone the Court of the 2010s, so radically has it changed. An institution in which justices employed a solitary law clerk; wrote their own opinions; discussed most of the cases; issued short, unadorned opinions; seldom issued dissents; and decided numerous cases—a diverse mixture of common-law, statutory, and constitutional—has morphed into an institution in which the justices act as partners in individual law offices of four clerks; publish long, scholarly opinions; and decide on a small, rarefied set of cases. And yet the decision-making calculus of the chief justice in setting the discuss list evinces marked stability across time, with square conflicts, the United States as a petitioner, and amicus briefs as high-impact and relatively consistent forces. Nonetheless, as we hypothesized, Hughes, Warren, and Burger differed in the weights they placed on features of cases and in the extent to which political ideology shaped their decisions. As such, we conclude that both institutions and individuals play an important role at the initial agenda-setting stage. We have been able to show this through the first systematic comparative account of how chief justices select the set of cases for discussion. And we have presented a method of calculating second differences that is broadly relevant for research involving comparison of effect sizes across groups.