In the continuing drama of American law the judge still holds the center of the stage, down in front of the footlights. …Much of our finest intelligence is engaged in studying what judges do and say and in guessing at their inmost sensations.
—John P. Dawson 1
Judges indeed play a central role in the American legal system—more so than in most others. But the behavior of American judges, and in particular the determinants of their decisions, are not well understood, including by lawyers, law professors, and even many judges (we'll explain that paradox in due course). In part this is because judges in our system are permitted to be, and most are, quite secretive.2
Indirect methods must be employed to understand their behavior. Beginning more than half a century ago but accelerating in recent decades, social scientists—political scientists in particular, but also economists and psychologists, and, increasingly, academic lawyers knowledgeable about social science—have used ever more sophisticated theoretical concepts and quantitative tools to penetrate self-serving judicial rhetoric, go beyond judges' limited self-understanding, and place the study of judicial behavior on a scientific basis. Yet this literature is not well known to the legal community,3
apart from the small sliver that without pretension to doing social science research takes an interest, whether sympathetic or critical, in what social scientists might have to say about judges.4
That separation is unfortunate, because many conventional legal scholars, abetted by judges, have promoted an unrealistic but influential theory of judicial behavior in which careerism and ideology play no role in judicial decisions, while some social scientists and some journalists have sponsored the opposite but also unrealistic conception of judges as merely politicians in robes.
The theory of judicial behavior in which careerism and ideology play no role in judicial decisions we call "legalism," though the more common term is "formalism." In its simplest form, judges are said merely to apply law that is given to them to the facts; their task is mechanical, at best a form of engineering (but not "social engineering"), and involves no exercise of discretion. In a more complex version, judges (especially Supreme Court Justices) apply to cases an intellectual system—a methodology— adopted on politically neutral grounds to generate objective decisions. These systems are called by such names as "originalism," "textualism," "the Constitution in exile," "the Constitution as common law," "the living Constitution," "active liberty"—the list goes on and on.
Some academic lawyers have sponsored the opposite conception of judges—that their legalist pretensions are mere rhetoric, designed to conceal the political character of their rulings. About those systems these academic critics might say with Kierkegaard that "in relation to their systems most systematizers are like a man who has built a vast palace while he himself lives nearby in a barn; they themselves do not live in the vast systematic edifice."5 The skeptical theory of judicial behavior was taken to an extreme by such "realists" as Fred Rodell of Yale and by the votaries of "critical legal studies," a revival of Rodell-style extreme legal realism that flourished in the 1970s and is now defunct, although some of its avatars, such as feminist legal theory and critical race theory, continue to have a following among academic lawyers.
Properly understood, legal realism is an attempt to be realistic about judicial behavior, and an attempt with a distinguished pedigree (Jeremy Bentham, Oliver Wendell Holmes, Benjamin Cardozo, Learned Hand, moderate legal-realist scholars such as Karl Llewellyn, and other luminaries6).
But the school of legal realism lacked both an articulated model of judicial behavior and the data and empirical methodology required to test such a model. These tools are now available, and we build with them in this book, hoping to augment the existing social-scientific literature on judicial behavior.
There is an important difference between traditional legal realism and the concept of realism that shapes our analysis. The traditional realism was, like legalism, a jurisprudential theory—a theory about the legitimacy and character of particular judicial outcomes. It was not a theory of how judicial behavior is shaped by incentives and constraints. Legal realists, notably Jerome Frank, offered conjectures about judicial psychology, but realist analysis was largely limited to the influence of a judge's ideology, or, what was believed to be correlated with it, social class, on judicial votes. There is more influencing a judge than class and ideology, and the "more" includes, but importantly is not exhausted in, legalist reasoning. This is brought out in a striking passage by Llewellyn, the foremost academic legal realist:
Those phases of human make-up which build habit in the individual and institutions in the group …[are] laziness as to the reworking of a problem once solved; the time and energy saved by routine, especially under any pressure of business; the values of routine as a curb on arbitrariness and as a prop of weakness, inexperience and instability; the social values of predictability; the power of whatever exists to produce expectations and the power of expectations to become normative. The force of precedent in the law is heightened by an additional factor: that curious, almost universal sense of justice which urges that all men are properly to be treated alike in like circumstances. As the social system varies we meet in.finite variations as to what men or treatments or circumstances are to be classed as "like"; but the pressure to accept the views of the time and place remains.7
These are shrewd observations. They are paralleled from outside the legal profession by remarks to us (which we have edited slightly) by the economist Andrei Shleifer, in correspondence:
Consider common law judges who face few prospects of promotion, are not so high in the judicial hierarchy that they are making law, and cannot be fired or voted out. These judges face almost no incentives. They need to move cases through, and they need to be not so utterly random that they get overturned very much. But these are not enormously strict constraints. So what consequences follow? I think that in this context just about any external or internal motivation can prove decisive. Of course, one motivation might be to try to figure out what the law is and to apply it, but this is only one of them. Judicial politics is one that the legal literature focuses on; abuse of lawyers to humor oneself is another. But one needs to take a much broader view: that just about anything can move these judges when incentives are so weak. Right/left politics is only one source of judicial bias. Other sources might be much more important. Judges can be pro-dog or anti-dog. More importantly, judges can be pro-government or anti-government. If many judges are former prosecutors or believe that "prosecutors don't get names out of a phone book," this might be a more significant bias than right/left politics. Judges may have their own philosophies on all kinds of matters, in other words. There may be a very important interaction between the judicial branch and the media. If judges are really not particularly constrained, presumably they will try to get attention, be liked, be popular. Since media intermediates a lot of information, they might choose to do what the media want.
Building on such insights, we aim in this book to present a realistic model of judicial behavior that is sufficiently simple and definite to be testable empirically, and then to test it.
Chapter 1 presents the model. Using concepts from labor economics, it models the judge as a participant in a labor market—the judicial labor market—and defends the model against legalist objections. We explain that a judge conceived of as a participant in a labor market can be understood as being motivated and constrained, as other workers are, by costs and benefits both pecuniary and nonpecuniary, but mainly the latter: nonpecuniary costs such as effort, criticism, and workplace tensions, nonpecuniary benefits such as leisure, esteem, influence, self-expression, celebrity (that is, being a public figure), and opportunities for appointment to a higher court; and constrained also by professional and institutional rules and expectations and by a "production function"—the tools and methods that the worker uses in his job and how he uses them. We rebut the formidable challenge that Judge Harry Edwards of the U.S. Court of Appeals for the District of Columbia Circuit has mounted to the realist approach to judicial behavior.
We call ours "the" realist approach, but other such approaches are possible and are found in the scholarly literature, approaches that emphasize the influence on judging of personal characteristics, such as race and sex, religious upbringing, temperament, cognitive limitations, education—even law-school grades.8
Those approaches are fruitful, but in order to make our analysis manageable we employ a simpler model, though we do from time to time touch on personal-identity characteristics, which influence ideology and through ideology judicial outcomes.
Ours is strictly a positive analysis. We do not ask how judges should decide cases but how they do decide them—more broadly, how they do their judicial work (which is not limited to casting votes in cases). That doesn't mean our book is of purely academic interest. The better that judges are understood, the more effective lawyers will be both in litigating cases and, as important, in predicting the outcome of cases, thus enabling litigation to be avoided or cases settled at an early stage. And judges who understand their motivations and those of other judges are likely to be more effective judges. A realistic understanding of judges should also improve legal education and enable the design of realistic proposals both for continuing judicial education and for judicial reform. We have therefore worked hard to make a book largely of statistical analysis intelligible to lawyers and judges who have no familiarity with such analysis. To that end, we follow this general introduction with a technical introduction that defines statistical terms and explains the statistical methodologies employed in the book. It is very important to us that lawyers and judges who are unfamiliar with statistical analysis be able to understand the entire book, because our findings are of greater potential importance to them than to any other segment of our intended audience.
Chapter 2 seeks to introduce the legal profession, including law professors and judges, to a substantial social-scientific literature about judges of which much of the profession is ignorant. We show that it is a literature richer in data than in theory and that it is heavily focused on the U.S. Supreme Court and to a lesser extent on the lower federal courts, though recently it has begun paying increased attention to state and foreign courts as well. We review the history of the empirical studies that relate to our model of the judge, noting a fourfold division: studies of individual decision-making (the effect of a judge's ideology, life experiences, personality, etc., on his judicial votes); of collegial and other group effects (such as panel composition, dissent aversion, and opinion assignment by presiding judges); of hierarchical effects (the relations between lower and higher courts); and of interactions between the judiciary and other branches of government. This last division of the literature plays almost no role in our analysis, however.
A particular focus of chapter 2 is on the methodological problems that empirical studies of judicial behavior have encountered and tried to overcome. These problems primarily involve determining the ideology of judges and the ideological character of judicial votes, with the further complication that a judge's ideology when he is appointed (what we call his ex ante ideology) may differ from the ideology reflected in his judicial votes (his ex post ideology). Both are important, and the methods of determining ex ante ideology may not be appropriate for determining ex post ideology, and vice versa. The various methods, as we explain, all have their strengths and weaknesses, and we use all to a greater or lesser extent.
The remaining chapters present the results of original empirical re-search conducted by us. Data availability and a desire to keep the book to a manageable length have persuaded us to limit our analyses to the federal judiciary, and specifically to the Article III federal judiciary, consisting of Supreme Court Justices, court of appeals judges, and district judges. There are other federal judicial officers, primarily federal magistrate judges and bankruptcy judges (all bankruptcy judges are federal), but they are not appointed by the President or confirmed by the Senate, and have fixed terms of office rather than life tenure. Their appointment is authorized by Article I of the Constitution rather than by Article III, the latter being the provision authorizing a federal judiciary consisting of judges who are appointed for life by the President subject to Senate confirmation.
Chapters 3 through 8 propose and test a variety of hypotheses concerning judicial behavior. A major focus is on the relative weight of ideology and legalistic analysis in decision-making in the different tiers. But we also measure more mundane, though we think very important, influences on judicial behavior, such as effort aversion (which includes but is not exhausted in leisure preference), that are predicted by our labor-economics model of judicial incentives and constraints. effort aversion includes both reluctance to work "too" hard—that is leisure preference—and reluctance to quarrel with colleagues (conflict aversion). Both are aspects of the "quiet life" that is especially valued by persons in jobs that offer little upward mobility—and in the case of a federal judgeship involve virtually no downward mobility— and few opportunities for increasing their pecuniary income other than by quitting and, in the case of judges, going into the private practice of law.
Regarding the appellate tiers—the Supreme Court and the courts of appeals—we explore in addition a variety of social or collegial phenomena that may influence judicial votes, such as conformity (an aspect of conflict aversion—the desire to "go along"), group polarization (the tendency of a faction to become more extreme as it grows relative to other factions— the opposite of the conformity effect), and what we call political polarization (the tendency, distinct from group polarization, for an increase in the size of one factional group to produce more extreme voting by all the factions). Relations with coworkers are an important influence on workplace behavior—and remember that we're treating courts as workplaces.
In the case of the Supreme Court (chapter 3), we find that our evaluation of ideology, utilizing several measures, tracks pretty closely what "everyone knows" is the ideological rank order of the Justices who served between the 1937 and 2009 terms, the period to which we limit our study. Consistent with many other studies, we find that Justices appointed by Republican Presidents vote more conservatively on average than Justices appointed by Democratic ones, with the difference being most pronounced in civil rights cases and least pronounced in federalism, privacy, and judicial power cases. These findings support the realist understanding that many cases that reach the Supreme Court tend both to be highly charged politically and to be indeterminate from a legalist standpoint, forcing the Justices back on their priors—which often have an ideological component—to resolve the case.
We find that the ideology of some Justices changes during their time on the Court. They may become more liberal or more conservative than when they were appointed (we follow the previous literature on judicial behavior in calling this "ideological drift") or relative to the ideology of the President who appointed them. The second phenomenon, which we dub "ideological divergence," shows that Justices aren't merely agents of the President who appointed them—which is no surprise, because once appointed they are free from any control by him.
The chapter includes a separate analysis of the Court's unanimous decisions, a subject that the existing literature has rather neglected. Unanimity in an ideologically divided court, as the Supreme Court is and has been, has been thought to undermine the theory that the Court's decisions are ideologically driven. We interpret the substantial percentage of unanimous decisions—more than 30 percent—to mean that the ideological stakes in many of the Court's cases are small enough to be overcome by mild dissent aversion of Justices; and yet discernible traces of ideological influence do remain in those decisions. We look for group effects in the Court—and do not find any. Members of a liberal or conservative minority on the Court are no more likely to go along with the majority if the majority grows. That is, we find no conformity effect, common in many workplaces, where a worker who bucks the office consensus may find himself ostracized. We attribute its absence to the absence of strong dissent aversion in the Court, as well as to the Justices' secure tenure and (with exceptions, none recent) lack of ambition for higher office, which encourages greater independence than in an ordinary workplace. We find no group polarization or political polarization effects, either; these absences may also be related to the Justices' secure tenure.
In analyzing the courts of appeals in chapter 4, we find that the judges of these courts are less ideological than Supreme Court Justices on aver-age, but not that ideology plays a negligible role in their decisions. Be-cause the ideological stakes are greater in Supreme Court than in court of appeals decisions (the courts of appeals have a mandatory jurisdiction, and as a result can't insist on a diet of ideologically charged cases and anyway don't have the last word in such cases, or indeed in any cases that the Supreme Court might decide to hear if the losing party petitioned for certiorari), and because the courts of appeals have heavier workloads than the Supreme Court, we expect effort aversion to play a greater role. effort aversion further reduces the influence of ideology by placing a premium on unanimity, which reduces effort (no dissents, and no revision of majority opinions to counter dissents) both in time and in strained relations among judges. We thus attribute panel composition effects— the well-documented tendency of the ideological direction of a court of appeals judge's votes to be influenced by the opposing ideology of another judge, or other judges, on the panel—in part to dissent aversion, as explained at length in chapter 6.
We also find a conformity effect, which is the other side of dissent aversion, in the courts of appeals. As a result, there is a double effect when a judge appointed by a President of one party is replaced by a judge appointed by a President of the other party (the party of the appointing President being a proxy for the appointee's ideology) and the newly appointed judge is part of the majority bloc on the court. If, for example, in a court in which the majority consists of judges appointed by Republican Presidents, a more conservative judge replaces a less conservative one, the majority bloc will vote more conservatively, of course— but so will the minority.
Also in chapter 4 we use a new method of assessing the ex ante ideology of federal judges (the ideology they bring with them to the court, be-fore they begin to hear cases) and compare it with their ex post ideology (the ideological direction of their judicial votes). We apply the method both to court of appeals judges and, for the sake of comparison, to Supreme Court Justices as well. We find substantially greater divergence between ex ante and ex post ideology in the courts of appeals and attribute it to the greater influence of legalistic considerations in those courts. The weightier such considerations are, the less likely a judge of conservative inclinations is to vote consistently conservatively or a judge of liberal inclinations consistently liberally.
Federal district courts, the subject of chapter 5, are the trial level of the federal judiciary, though nowadays the vast majority of their cases are decided without a trial and often with.out any significant judicial proceedings, being settled or abandoned early in the litigation. Cases in these courts, unlike cases in the courts of appeals, are unfiltered, whereas the courts of appeals hear only cases that a party who lost in the district court has chosen to appeal. District courts therefore hear a higher percentage of cases that can be disposed of readily by the application of legalist concepts than courts of appeals do. Indeed, a high percentage of cases filed in the district courts lack even arguable merit.
Moreover, legalist concepts such as standing, ripeness, and mootness, and doctrines that allow early dismissal of weak cases, are designed in part to reduce judicial workloads (for the concepts have largely been created by judges), and workloads are heavier in trial courts than in appellate courts. So we expect those concepts to play a greater role in decision-making in the district courts than in the courts of appeals or the Supreme Court. And so we find, and conclude that there is indeed less ideology in district court decision-making than in that of either of the higher levels of the federal judiciary.
We relate this phenomenon to the concept in organization theory of "management by exception," whereby the simplest questions are answered at the bottom of a decisional hierarchy and only the most difficult at the top. In the case of the federal judiciary, this implies that the role of ideology in decisions will increase as cases rise through the judicial hierarchy from the district court to the Supreme Court. Because both the district courts and the courts of appeals have a mandatory rather than a discretionary jurisdiction, they decide (the district courts particularly) a great many one-sided cases. The Supreme Court decides more evenly balanced—typically because they are more indeterminate legalistically— cases than the courts of appeals because the Court's decisional capacity is so limited in relation to the number of decisions by those courts and also by state courts, which decide many cases that involve a federal issue that the losing party might want to raise in the U.S. Supreme Court. Also, legalistically indeterminate cases tend to give rise to conflicts between federal courts of appeals, and only the Supreme Court can resolve those conflicts.
The contrast between the federal appellate tiers is especially pronounced in criminal cases. Most criminal appeals are by defendants rather than by the government, are subsidized, and are so lacking in merit that even liberal court of appeals judges usually vote to a.rm. The Supreme Court hears criminal appeals only if they have substantial merit. Thus we expect and find that the fraction of conservative votes (a vote to uphold a criminal conviction is scored as a conservative vote) is substantially higher in courts of appeals than in the Supreme Court, even in the case of court of appeals judges appointed by Democratic Presidents.
Yet mention of criminal cases suggests a possible objection to the proposition that the ideological influence in judicial decisions increases as one moves from the lowest to the highest rung of the judicial hierarchy. Sentencing is traditionally an area of almost untrammeled exercise of discretion by district judges—a discretion that was, it is true, curtailed by the Sentencing Reform Act of 1987 but that has largely been restored by the Supreme Court. Appellate review of many other types of ruling by district judges is also supposed to be deferential.
To the extent that appellate judges defer to district court rulings, we might expect district judges to have the scope and inclination to engage in ideological decision-making. We test this hypothesis in chapter 5 but find only limited support for it, even with regard to sentencing, where the decision-making autonomy of the district judge is at its highest. We attribute this negative finding to the fact that many appellate judges accord only quite limited discretion to district court rulings, and this in turn triggers reversal aversion by district judges—they cannot be confident that the appellate court will defer to an ideologically colored exercise of discretion by them.
The last three chapters, like the section of chapter 5 that discusses the judicial hierarchy, deal with topics that cut across two or even all three tiers of the federal judiciary, although data limit our analysis in chapter 7 to the Supreme Court.
Chapter 6 analyzes the phenomenon of dissents in both the Supreme Court and the courts of appeals. We stress the existence and significance of a reluctance by some judges to dissent publicly even when they disagree with their colleagues' decision. Dissent aversion is an example of a behavior that neither legalism nor legal realism in its jurisprudential sense can explain, but that a richer model of judicial incentives and constraints—our model, emphasizing effort aversion—can.
We find pronounced dissent aversion in the courts of appeals but not in the Supreme Court. We trace this difference to differences in the size of the appellate panel (nine Justices in the Supreme Court unless there are lighter caseload without any diminution in pay) and of age on the propensity to dissent, assuming that effort aversion increases with age.
Chapter 7 examines the goals of Supreme Court Justices in question-ing lawyers at oral argument. We suggest that those goals are strategic rather than just a matter of trying to get at the truth in order to facilitate legalist decision-making. We relate patterns of questioning at argument to limitations of judicial deliberation, to personality, and, counterintuitively, to leisure preference. Regarding personality, we consider extroversion a characteristic likely to impel a Justice to ask many questions at oral argument and/or to participate in many public events.
In chapter 8 we study the behavior of federal judges who have a realistic prospect of promotion. Desire for promotion is a significant motivating factor in many workplaces, and federal district judges are not infrequently (though not routinely) promoted to courts of appeals, while court of appeals judges are sometimes, though rarely, promoted to the Supreme Court; at this writing, eight of the nine Supreme Court Justices are former federal court of appeals judges, although they are a tiny minority of all court of appeals judges. We ask whether aspirants for promotion within the judiciary alter their judicial behavior in order to improve their promotion prospects—in other words whether they "audition" for appointment to a higher court. Comparing the voting behavior of court of appeals judges when they are realistically in the promotion pool and thus potential auditioners with their behavior after age eliminates their prospects for promotion and with the behavior of court of appeals judges who were never in the pool, we find that potential auditioners do tend, though only on average, to alter their behavior in order to improve their prospects for appointment to the Supreme Court. We conduct a similar analysis of district judges' auditioning for promotion to the court of appeals, with similar though weaker results. Auditioning behavior by judges is important evidence of the role of self-interest in the judicial utility function.
The conclusion sums up briefly but also suggests a number of promising topics for future study by ourselves or others.
We seek in this book to make four contributions. First, we present and test empirically a more realistic theory of judicial behavior than either the traditional legalist theory— still beloved of the legal professoriat and of the judiciary in its public-relations mode—or the dominant theory in political science, which exaggerates the ideological component in judicial behavior. Although the bulk of our analysis consists of hypothesis testing by means of statistical analysis, the hypotheses themselves are generated by theoretical reflection on likely judicial behavior; the book thus contains a good deal of interstitial qualitative analysis.
Second, we revise some of the coding of U.S. Supreme Court and (especially) federal court of appeals decisions in the U.S. Supreme Court Database (which covers the 1946 to 2010 terms)9
and the U.S. Appeals Court Database (1925 to 2002).10
These are the principal databases that have been used in previous studies, and we use them extensively as well. We have also updated a dataset created by Sunstein and his colleagues11
on the votes of court of appeals judges in ideologically sensitive subject-matter areas, and we use this dataset in our study, as we do a database of judicial sentencing in chapter 5 and four databases that we have created ourselves: an ex ante ideology measure that we use in chapters 3 and 4, a dismissal dataset that we use in chapter 5, a database of individual court of appeals judges' dissent rates that we use in chapter 6, and a judges' background dataset that we use in chapter 8.
Third, we make extensive use of regression analysis, rather than just simple correlation (though we use simple correlation extensively as well), to isolate variables that our model predicts should influence judicial behavior in specific ways. Claims about judicial behavior based on simple correlation are commonly made in law journals and reported in the media, but studies that do not control for potentially relevant variables, as regression analysis enables us to do, often produce unsound results.12
The structure of the federal judiciary facilitates cross-sectional statistical analysis. At both the court of appeals and the district court level, though of course not at the Supreme Court level, the structure is regional (except for the Federal Circuit, which we exclude from our analysis because its case mixture is so different from that of the other courts of appeals). Circuits and districts, while differing from one another in caseload and number of judges and the ideological composition of the court, exhibit sufficient similarities (in mode of appointment and terms of employment, for example) to minimize noise in the data. And because the output of federal judges is contained in public documents almost all of which are online, modern electronic search techniques make it feasible to create and analyze large databases of judicial activity.
The technical introduction that follows this general introduction ex-plains the rudiments of regression analysis, a type of statistical analysis that we employ extensively. But as it also sets forth some terminological conventions that we employ in the book, even our statistically sophisticated readers will want to read it (it is very short).
A fourth point, related to the comprehensive character of our model of judicial behavior, is that the scope of our analysis is broader than that found in most of the existing literature. That literature is concerned al-most exclusively with either votes cast by individual judges or case out-comes, in the Supreme Court and to a lesser extent in the courts of appeals. We too test hypotheses related to voting in these courts, but we study judicial behavior in the federal district courts as well, plus such other dimensions of judicial behavior as opinion length, citations, dissents, and questions asked by Supreme Court Justices at oral argument.