I SETTING THE STAGE
The Debate over What Law Has to Do with What Judges Do
1 Whatâs Law Got to Do with It
Thoughts from âthe Realm of Political Scienceâ
Jeffrey A. Segal
CHIEF JUSTICE ROBERTS has declared himself to be a believer in precedent, a follower of the rule of law, an umpire calling balls and strikes:
Somebody asked me, you know, âAre you going to be on the side of the little guy?â And you obviously want to give an immediate answer, but, as you reflect on it, if the Constitution says that the little guy should win, the little guyâs going to win in court before me. But if the Constitution says that the big guy should win, well, then the big guyâs going to win, because my obligation is to the Constitution. Thatâs the oath.â (Roberts 2005)
While this may have been mere show, it wasnât mere show just for the Judiciary Committee. In a recent speech at the University of Arizonaâs Rehnquist Center, Roberts declared that the shift to a Supreme Court filled exclusively with former appellate judges took constitutional law out of âthe realm of political scienceâ and onto âthe more solid grounds of legal arguments. What are the texts of the statutes involved? What precedents control?â(Liptak 2009).
Letâs leave aside for the moment the fact that Robertsâs assertions are empirically falseâjustices who served on lower appellate courts are not more likely to abide by precedent, and are not less likely to vote ideologically than are judges without appellate court experience (Epstein et al. 2009). What is the realm of political science? A quick answer is that political science examinations of judicial decision-making have focused on four partially overlapping models of such behavior: the legal model, the historical institutional model, the attitudinal model, and the strategic model.
The Legal Model
The legal model in its various forms holds that judges are motivated to establish an accurate, clear, and consistent interpretation of the law (Baum 1997). They do so by applying the facts of the case to relevant statutory and constitutional text, the intent of those who established those texts, and the precedents established by previous courts interpreting those texts (Segal and Spaeth 2002). Within this general framework, three broad operationalizations of the legal model exist. The most extreme version of the legal model claims determinate answers to legal questions. While Frank Cross correctly claims that âmost contemporary scholars no longer adhere to the strict determinate formalist modelâ (Cross 1997, 255), it is still possible to read that there are âinternally correct answers to all legal questionsâ (Markovits 1998, 1), or that âany extreme thesis that the law is always or usually indeterminate is untenableâ (Greenawalt 1992, 11). From the political science perspective, while the charge of âformalismâ is readily hurled at those testing the legal model (for example, Gillman 2001), only one article of which I am aware tests a determinate model of judicial decision-making (Kort 1963).
More realistically, political scientists, borrowing conceptually from Dworkin (1978), have typically examined whether legal factors have a gravitational force on judicial decisions. Even within this framework, scholars differ on how strong the force of law is. Dworkinâs notion of a strong gravitational force requires judges to find the correct answer to legal questions and disallows any âindependent forceâ of a judgeâs convictions âjust because they are hisâ (Dworkin 1978, 118).1 Political scientists who have adopted a gravitational perspective on the influence of law nevertheless expect that policy preferences will play a substantial role in judicial decisions (Songer and Lindquist 1996; Spaeth and Segal 1999).
Beyond the gravitational models are postpositive legal models. According to Howard Gillman:
In the version of the argument that might be called âpost-positivist,â legalists make claims, not about the predictable behavior of judges, but about their state of mindâwhether they are basing their decisions on honest judgments about the meaning of law. What is post-positivist about this version is the assumption that a legal state of mind does not necessarily mean obedience to conspicuous rules; instead, it means a sense of obligation to make the best decision possible in light of oneâs general training and sense of professional obligation. On this view, decisions are considered legally motivated if they represent a judgeâs sincere belief that their decision represents their best understanding of what the law requires. Burton [Steven J., Judging in Good Faith (Cambridge University Press, 1992), xiâxii, 44] has persuasively argued that this notion of âjudging in good faithâ is all we can expect of judges (Gillman 2001, 486).
Under the postpositivist approach, virtually any decision can be consistent with the legal model. And any decision is consistent with the model as long as the judge has sincerely convinced herself that the decision is legally appropriate. The most basic problem with this approach is clear: the model is not falsifiable in terms of which decisions judges actually make. By accepted standards of scientific research, the model cannot provide a valid explanation of what judges actually do (see, for example, the discussion in Daubert v. Merrell Dow Pharmaceuticals, Inc., 1993, 593). The fact that law might be nuanced (Burbank, this volume) does not relieve its empirical supporters of the need to provide falsifiable tests of its influence.
It is difficult to know whether judges sincerely believe they are judging in good faith. The extensive psychological literature on motivated reasoning suggests that plausible arguments are all that are needed to create an overlap between prior views and a subjective belief in correct results (for example, Braman 2006; Baumeister and Newman 1994; Kunda 1990), thus suggesting that good faith will not be all that difficult to come by.
Modeling Law
Any hope of modeling law from a political science perspective is likely to require a gravitational approach. A deterministic account will almost certainly fail, particularly at the appellate level, as there is very little evidence that this is how judges rule. Alternatively, a postpositive account will never fail, as it offers no predictions about what a judge will do (Gillman 2001, 486). The gravitational approach, on the other hand, allows for comparative statics. It is also consistent with statistical models that allow for error terms, or unexplained variance.
To determine the impact of law is not much different than determining the impact of other social phenomena. Simply put, judgesâ decisions should changeânot deterministically, but at the marginsâas law changes, holding alternative phenomena constant. This is easiest to see in the case of vertical stare decisis. While the strategic implications of such models can be quite complex (Cameron, Segal, and Songer 2000), the essence of testing this requires little more than determining how lower court decisions change as higher court decisions change, after controlling for other relevant factors such as case characteristics and the preferences of lower court judges.
Text
When scholars make causal inferences that justices are influenced by text, too often they reach these conclusions from nothing more than the official writings of the justice in question (Karsten 1997; Kahn 1999). But as Harold Spaeth (1964) demonstrated decades ago, Frankfurter was able to convince scholars that he was a restraintist, despite ample evidence to the contrary, because he so often wrote about the need to defer in those cases in which he chose to do so. Similarly, the fact that Scalia readily claims to be constrained by text doesnât necessarily mean that he really is. Even Brennan, after all, was willing to make use of textual arguments in his opinions, at least when they helped him reach liberal results (Phelps and Gates 1991).
Sometimes, doctrinal-based scholarship makes a plausible case for reliance on text. This is especially true when the decisions appear to contradict the justiceâs broad policy values. Itâs impossible to imagine that anything other than a lack of any constitutional command could have led Black to vote to uphold Connecticutâs silly birth control law. Similarly, Scaliaâs dedication to the confrontation clause (for example, Maryland v. Craig, 1990) lends an air of plausibility to his self-proclaimed originalism (Scalia 1997; for empirical support, see Barkow 2006).
How then can we systematically assess, through a priori measures and falsifiable tests, text as a potential explanation for the justicesâ behavior?2 The fundamental problem in testing textâat least from a gravitational perspectiveâconsists of validly measuring this factor. One possibility is to measure not law, but âthe more solid ground of legal argumentsâ (Roberts 2009). This can be accomplished by examining the briefs filed by petitioners and respondents in cases heard before the Court. Under the adversary system of justice, one can rely on litigants to present the best possible set of arguments in their favor. If a plausible case can be made that a legal argument supports a party, that party will almost undoubtedly make that claim. While counsel will vary in quality and thus might miss some obvious claims, the principle that justices should not decide issues sua sponte largely limits the justices to those claims that are made by counsel. Thus, we can measure whether one or both parties make a text based claim, whether the other side disputes that claim, or whether the other side makes a counterclaim (Segal and Howard 2002).
These are certainly not ideal measures, but if Scalia is a textualist, he should be more willing to support a liberal litigant who makes an undisputed textual claim than a liberal litigant who makes a disputed textual claim, or, ceteris paribus, one who makes no textual claim at all.
Intent
Perhaps the biggest difference between political science and law school examinations of law involves the difference between the normative scholarship that largely drives law school scholarshipâthat is, writings that seek to answer what law should beâand the positive scholarship that typically drives political science scholarshipâthat is, writings that try to explain the influences on law (Friedman 2004). Thus, legal academics carefully examine many different flavors of framersâ or legislative intent, finding profound normative differences between, say, semantic originalism and original intentions (Solum 2008). Positive political scientists have paid less attention to intent, in part because of paltry evidence that intent actually influences judicial decisions (Segal and Howard 2002), and in part because the concept of legislative or framersâ intent does not make sense.
âIntentâ is a concept that applies to individuals, not to groups. As Kenneth Shepsle (1992) appropriately noted, âCongress is a âthey,â not an âit.ââ Thus, legislative intent is an oxymoron (ibid., 244). The damage that the plural nature of legislatures causes to the concept of legislative intent is less severe if legislation is single-dimensional rather than multidimensional. With but a single dimension, it might be possible to argue that the intent of a law depends on the author of the legislation, committee reports on what the bill is intended to accomplish, the public positions of those who supported it, or as McNollGast have argued, on the pivotal legislatorsâthe moderates who sit between the ardent proponents and ardent opponentsâwhose assent was necessary to secure passage of the law (McCubbins, Noll, and Weingast 1992).
This prescription falls apart when legislation is multidimensional. This can happen because an omnibus bill contains multiple issues or because preferences over a bill dealing with just one issue are not single peaked. Such multi-peaked preferences can occur over a single issue in a variety of circumstances: a legislator might prefer that we either expand a war or pull out rather than continue on a current course of stalemate; or, say, prefer that we allow either a wide variety of religious holidays or none rather than just those from mainstream religions.
Under these circumstances, group decisions, such as those made by legislatures, can cycle. Consider the preference ordering in Table 1.1. Legislator 1 prefers giving mainstream religious holidays off (M), followed by all religious holidays (A), followed by no religious holidays (N). In other words, Legislator 1âs preference function is MPAPN. Legislator 2 prefers to give all religious holidays off, followed by none, followed by mainstream: APNPM. Legislator 3âs preference is to give no religious holidays, then mainstream, then allâthat is, NPMPA. With these alternatives, giving all holidays off defeats giving no holidays off, giving no holidays off defeats giving mainstream holidays off, but giving mainstream holidays off defeats giving all holidays off.
TABLE 1.1
Hypothetical Choices of Three Groups among Three Alternatives
While we can make useful but disputable arguments as to legislative intent when preferences are transitive, to speak of legislative intent when preferences a...