LEGAL REGIMES AND DOCTRINES IN
SUPREME COURT DECISION MAKING

Mark J. Richards
Department of Political Science
University of Wisconsin -- Madison
1050 Bascom Mall
Madison, WI 53706
richards@polisci.wisc.edu

and

Herbert M. Kritzer
Department of Political Science
University of Wisconsin -- Madison
1050 Bascom Mall
Madison, WI 53706
kritzer@polisci.wisc.edu

Paper prepared for presentation at the 1998 annual meeting of the Southern Political Science Association, Sheraton Colony Square Hotel, Atlanta, Georgia, October 28-31, 1998. The authors would like to thank Jeffrey Segal, Kevin McGuire, and Joseph Ignagni for generously sharing data with us; the analyses and conclusions presented in this paper are the responsibility of the authors.

INTRODUCTION

In this paper we address a multi-faceted research puzzle: is there a role for law in systematic, empirically supportable models of Supreme Court decision making? Proponents of the attitudinal explanation suggest that "the attitudinal model is a complete and adequate model of the Supreme Court's decisions on the merits" and that there is no systematic evidence supporting legal explanations (Segal and Spaeth 1994, 11). The unabashed nature of the claims for the attitudinal model leaves many political scientists very troubled (Brenner and Stier 1996; Brisbin 1996; Knight and Epstein 1996; Lawrence 1994; Songer and Lindquist 1996). Where is law in all of this? The courts, including the Supreme Court, are supposed to be guided by the law. If the attitudinal model does offer a complete explanation, then why is it that attitudinal variables do not explain all of the variance in the Court's decisions? It is possible that legal variables must be introduced in order to achieve a more complete explanation. For example, consider the Court's free speech decision making. If the attitudinal explanation is complete, then why is it that after 1972, when Chief Justice Warren Burger had taken over the leadership of the Court from the liberal Chief Justice Earl Warren and the membership of the Court became more conservative, the Court was more likely to uphold the rights of speakers in cases where government regulations of expression were not narrowly tailored to serve government interests? In order to answer this question, we argue that it is necessary to look beyond a purely attitudinal explanation and consider what we label "legal regimes."

LEGAL REGIMES

The term "regime" is most often associated with a group of leaders, and hence one might be tempted to view "legal regimes" as synonymous with particular configurations of personnel on the Supreme Court. However, we introduce the modifier "legal" to indicate that regimes can be defined by law as well as by people. More specifically, we argue that major precedents can serve to structure the way the Supreme Court justices evaluate key elements of cases in arriving at decisions and such precedents serve to define legal regimes. As with regimes generally, legal regimes change over time.

Legal regimes structure Supreme Court decision making by establishing which case facts which are relevant for decision making, and/or by setting the level of scrutiny or balancing the justices are to employ in assessing case facts. A legal regime can be based on a single precedent, or a number of chronologically proximate precedents. The key to identifying legal regimes is change: only by identifying different legal regimes for a given area of law can we show empirically that such regimes exist. This demonstration is straightforward: the influences of case elements on the justices' decisions should vary across regimes. Thus, to identify legal regimes we start not with justices but with cases and doctrine. The point in time at which a legal regime is established serves as a chronological break point for such empirical analysis.

We have identified two different general types of legal regimes. The first we label "doctrine-based" because underlying it is a legal doctrine. By legal doctrine we mean a specific formulation of the analytic framework that serves to structure Court decisions in a particular area by identifying relevant case facts and indicating how those facts should be weighed. The second type of legal regime we label a "standard-based" legal regime because it does not involve an analytic framework but simply establishes rules regarding things such as level of scrutiny, or whether one side or another in an issue is to be afforded some sort of preference. Clearly, doctrine-based legal regimes are likely to incorporate rules regarding standards of review; thus, one could argue that some legal regimes imply considerably more change than others.

In a paper prepared for last spring's Midwest Political Science Association Meeting (Kritzer, Pickerill, and Richards 1998), we presented an analysis of changes in decision making patterns on the Supreme Court in political free speech cases. We used the case of Brandenberg v. Ohio (395 U.S. 444, 1969) as the demarcation of two regimes. We compared the influence of a variety of factual and party variables, and judicial attitudes, on justices' votes in cases before and after Brandenberg. The results supported the argument that the factors that influenced decision making differed in the two sets of cases. For example, before Brandenberg, the justices did not appear to make significant distinctions among types of speakers while after Brandenberg such distinctions were drawn. Another change was that after Brandenberg the justices were significantly more likely to defer to support the government's position if the case was brought by federal or state government. Lastly, and not surprisingly given the substance of Brandenberg, the issue of whether there appeared to be a likely and imminent risk of lawless action changed sharply in its influence before and after Brandenberg. The analyses carried out in our earlier paper made it clear that the change could not simply be attributed to changing personnel on the Court, but reflected genuine changes in factors influencing the justices' votes.

By advancing the construct of legal regimes we do not in any way reject the argument that justices are influenced significantly, perhaps even primarily, by attitudes. Nor do we reject the argument that justices are constrained by institutional and strategic concerns such as the need to obtain agreement from four additional colleagues to secure a majority in favor of the justice's position (Epstein and Knight 1998). Rather, legal regimes provide a construct that can incorporate law systematically, and in a way that can be empirically tested, into the dominant social science models of Supreme Court decision making. Specifically, if legal regimes exist, and points of regime breaks can be identified, then statistical analyses should show systematic differences in the factors influencing justices' decisions before and after the regime break.

The purpose of this paper is to extend our analyses of what we now label legal regimes. As part of this, we develop in some detail the connection between legal regimes and legal doctrine. However, before turning to that explication, we present several analyses based on data collected by other scholars that add further credence to the core argument that differing legal regimes can be identified and their impacts demonstrated empirically. We also present two failed analyses that show that we are not simply picking up what might be labeled random change (i.e., could something resembling legal regimes appear to exist if one simply randomly split a series of cases into two chronologically distinct groups). The largest portion of what follows relies upon original data on free speech cases, broadly defined, which were collected by the first author. As part of the analysis of free speech cases, we develop in more detail the linkage between legal doctrine and legal regimes.

ANALYTIC STRATEGY AND STATISTICAL METHOD

The core hypothesis that we derive from the legal regimes model is that the factors influencing justices' decisions should vary across legal regimes. Thus, if we can identify decisions (or possibly sets of decisions) that represent a change in the legal regime governing a specific set of cases, the results of statistical models predicting decisions before and after the regime change should differ in significant and meaningful ways. Thus, testing the model of legal regimes involves the following steps:

1. Identifying areas of law for which we can hypothesize that there have been differing legal regimes governing decision making over time.
2. Identifying the point (or period) of change from one regime to another for each area of law identified in step 1.
3. Identifying the factors that would be expected to account for decision making in each area of law (e.g., justices' attitudes, parties, situational elements, other factual elements).
4. Estimating statistical models across, before, and after the regime changes, and comparing the statistical results to ascertain whether the statistical results do in fact differ across regimes.
5. Ruling out the major alternative explanation that over time change can be explained by personnel (and hence attitudinal) change.

The key statistical test of change is a variant of the well-known Chow test (Hanushek and Jackson 1977, 127-129) for comparing whether statistical results differ across sets of data. Our analysis relies upon logistic regression, and the computation of the test of change is based on features of logistic regression.⁽¹⁾ As discussed above, in the paper that we presented at last spring's Midwest Political Science Association annual meeting (Kritzer, Pickerill, and Richards 1998), we applied precisely this strategy to one specific area, free speech cases involving advocacy of illegal action.

In this paper, we extend our analysis of legal regimes by expanding substantially our analysis of free speech cases to include the full range of such cases, and to add two additional areas of law: search and seizure, and obscenity. The first of these relies upon original data collection by the first author of this paper; the other two rely upon data collected by other scholars and generously shared with the authors.⁽²⁾ The basis of the regime change for each of these will be discussed in the section reporting the results for each set of data.

One possible challenge to our analysis is that virtually any long series of Supreme Court decisions could be arbitrarily split into two time periods and the results of statistical analyses of the impact of the decision making factors would differ between those two periods. After reporting our primary analysis we report the results of two analyses that give us confidence that our results do not simply reflect what might be called inevitable overtime change. One of these analyses arose from an initial error on our part involving the search and seizure data. The other involves an analysis of free exercise cases based on data collected by Joseph Ignagni (1993).⁽³⁾ These data are useful for this purpose because the period covered was selected by Ignagni specifically because "the basic principles and tests" used by the Court during that period remained essentially unchanged (Ignagni 1993, 513)--in our terminology, a single legal regime governed the Court's decision making during the period included in the data set. As a test of our argument, we will use another major decision that had some free exercise elements but was not regime-changing for this area of jurisprudence.

HYPOTHESES AND RESULTS: OBSCENITY LEGAL REGIME

That obscenity cases have long been an area of difficulty for the Supreme Court is best captured by Justice Potter Stewart's oft quoted observation:

I shall not today attempt further to define the kinds of material I understand to be embraced within that shorthand description; and perhaps I could never succeed in intelligibly doing so. But I know it when I see it, and the motion picture involved in this case is not that (Jabobellis v. Ohio, 378 U.S. 184, 197n2, 1964).

However, this by no means makes it impossible to predict Supreme Court decision making in this area, and certainly one would expect the attitudinal model to predict that conservative justices were more inclined to support government efforts to suppress materials deemed to be pornographic. There have been at least two scholarly analyses of these cases.

Timothy Hagle examined case outcomes for 107 cases (Hagle 1991). His model included variables describing the type of medium (print, movies, live performance, etc.), whether the case came from a southern state, whether the material was offered in a commercial fashion (i.e., whether there was a charge for access to the material), and the lower court decision. Hagle's model was able to improve the prediction of the outcome by 58% over and above what could be done by what amounted to random guessing. The other analysis was conducted by Kevin McGuire (McGuire 1990). He also looked at case outcomes, but his analysis included 85 cases, and his predictive variables were more comprehensive:

• Level of government bringing the case
• Medium (film, magazine, book, other)
• Degree of regulation the government sought to impose (1=nonprior restraining order, 2=seizure, 3=licensing scheme, 4=prior restraint through injunction)
• Government's claimed impacts (erosion of morals, violation of community standards, pandering, antisocial effects)
• Defendant's claims (First Amendment violation, privacy violation, literary value, due process violation, lack of knowledge that materials in question were legally obscene)
• Involvement of amicus curiae
• Court membership

Our analysis relied upon a slightly updated version McGuire's data. This data set contained a total of 96 cases covering the period 1957 (Kingsley Books, Inc., et al. v. Brown, 354 U.S. 436) to early 1988 (Virginia v . American Booksellers Association, 484 U.S. 383). Because our analysis is based on votes rather than cases, we had available a total of 770 individual votes by the justices.

As the case demarcating a regime change in obscenity law, we chose Miller v. California 413 U.S. 15 (1973). In Miller the Court sought to change the direction of obscenity jurisprudence by first modifying the definition of what was not protected by the First Amendment (a definition that still holds today), and by reducing the Court's role in determining what was obscene by dictating that local community standards rather than national community standards should govern the determination of whether the work "appeals to the prurient interest." Miller refined the Roth v. U.S. (354 U.S. 476, 1957) "contemporary community standards" requirement to mean local, not national, standards. After Miller, the standard is whether the average person, applying contemporary local community standards, would find that the work, taken as a whole, appeals to the prurient interest. The second element of the Miller test that must be met before a work can be considered obscene is that it must depict or describe in a patently offensive way sexual conduct specifically defined by state law. The third element of the Miller definition of obscenity involves the question of whether the work in question "as a whole, lacks serious literary, artistic, political or scientific value." Miller is actually less speech protective than Memoirs v. Massachusetts (383 US 413, 1966) because it changes the Memoirs case's requirement that in order for the work to be constitutionally curtailed the work be "utterly without redeeming social value" to an easier standard for the state to meet. Under Miller, the work need only be "lacking serious literary, artistic, political, or scientific value" in order for the state to be able to constitutionally curtail it. The Miller regime remains as the dominant approach to obscenity. The Miller decision has not been overruled and constitutes "a revised standard" according to first amendment scholars Shiffrin and Choper (1996).

We expect the Miller case to represent an important break in Supreme Court jurisprudence on the issue of obscenity. The changes could show up in a variety of ways: factors that were statistically significant before Miller cease to be after Miller, and vice versa; the direction of influence of variables can change; the degree of influence of variables can change; and some types of cases might simply not be present in one period or the other (with "type of case" being defined in terms of the presence or absence of variation in a particular variable).⁽⁴⁾

Table 1 shows our analysis of McGuire's data at the justice level for before and after the Miller decision (Miller itself is included in the after set of cases). In addition to the variables defined in McGuire's analysis, we have included a measure of the justices' attitudes.⁽⁵⁾ The dependent variable is coded so that positive coefficients indicate an increase in support for the defendant (i.e., the "liberal" position). The overall test of significance for differences in the impact of variables before and after Miller yields a chi square of 50.67 with 14 degrees of freedom. If anything, this understates the change because two of the variables are constant in the post-Miller period (the presence of claims that the materials threaten to erode community values or that the materials constituted pandering). One challenge to these results might be that the change actually reflects changing personnel: some justices were on the Court only before Miller and some only after, and the factors predicting decisions differ not because of Miller but because of personnel. To test this, we reestimated the models including only the 513 votes of justices on the Court at the time of Miller; the chi square test value for change was unchanged (50.91, 14 degrees of freedom).

An examination of the coefficients in Table 1 show a number of noteworthy differences before and after Miller, in addition to the lack of variation in two variables post-Miller:

• One variable, whether there was a claim of privacy violation, is significant before Miller, but not after: before Miller, the presence of a privacy claim decreased the likelihood that a justice would support the defendant's position.
• Both the government's claim that community standards are violated and the defendant's claim that due process was violated are significant after Miller but not before. A community standards claim by the government increases the likelihood that the justice will support the government's position (consistent with Miller) while a due process violation claim by the defendant increases the likelihood that a justice will support the defendant.
• The impact of amicus briefs reverses direction. This variable is coded so that it represents the difference in the number of amicus briefs supporting the defendant and supporting the government, with positive values representing predominant support for the defendant. Before Miller, the justices tend to vote against the weight of the amicus briefs; after Miller their votes tended to go in the direction of the weight of amicus briefs.
• Both before and after Miller, liberal justices tended to support the defendant and conservative justices the government. However, the impact of the justices' attitudes increased significantly after Miller. In a sense, attitudes became more important under the Miller standard than they were previously!

HYPOTHESES AND RESULTS: SEARCH AND SEIZURE LEGAL REGIME

Another change in legal regimes occurred in 1983-84 in the area of search and seizure. Defining a legal regime change here is a bit more difficult than in the obscenity arena, in part because of the wide variety of issues that arise in these cases. Taken together, six important precedents constitute a legal regime that breaks with the Court's prior approaches to fourth amendment search and seizure jurisprudence by changing the how the justices weigh the elements of cases. This regime reflects a shift to a less libertarian view of the fourth amendment protections against unreasonable government searches and seizures.

Generally, probable cause is required for a search and seizure to pass constitutional muster. Aguilar v. Texas (378 U.S. 108, 1964) established two independent requirements for probable cause. Police must explain how informants know what they know, and why the police believe that the information is accurate. Prior to 1983, a failure on the part of the government to meet either one of these requirements meant that the search lacked probable cause and would most likely be held unconstitutional. However, the 1983 Illinois v. Gates (462 U.S. 213) decision began the move toward a new legal regime by holding that the two requirements were no longer independent. In a closely related 1984 decision (Massachusetts. v. Upton, 466 U.S. 727), the Court held that the "totality of the circumstances" is enough to justify a finding of probable cause. This leads to the expectation that after 1984, the Court would be more likely to accept lower court findings of probable cause because the Court would no longer apply the stringent Aguilar test in evaluating the determinations of the police and the lower courts.

The 1984 search and seizure legal regime was also shaped by two 1984 companion cases, United States v. Leon (468 U.S. 902) and Massachusetts v. Sheppard (468 U.S. 981). Previously, the Court had established that evidence which was obtained through searches based on defective warrants lacking probable cause was to be excluded from consideration in a defendant's conviction because the means of obtaining such evidence violated the fourth amendment. This is the exclusionary rule. Leon and Sheppard established a good faith exception to the exclusionary rule. The exclusionary rule no longer applies to situations where the police, acting in good faith, seized evidence in the context of reliance on a warrant which lacked probable cause.

Two additional 1984 cases also carved out exceptions to the exclusionary rule. Nix v. Williams (467 U.S. 431) and Segura v. U.S. (468 U.S. 796) created the inevitable discovery exception to exclusionary rule. If the police would have inevitably discovered the illegally obtained evidence through other independent sources, it does not need to be excluded. The good faith and inevitable discovery exceptions to the exclusionary rule indicate that based on the 1984 legal regime, the Court will be generally less sympathetic to the claims of the accused in search and seizure cases. For example, if the police performed a full search rather than a partial search based on a defective warrant which lacked the probable cause necessary to justify a full search, this would not necessarily lead to exclusion of the evidence and a decision against the government. Based on the precedents of the legal regime, the police could argue that they were acting in good faith or would have inevitably discovered the evidence.

Jeffrey Segal pioneered the statistical analysis of search and seizure cases (Segal 1984; 1986). His work basically defines the factors to be considered whenever one looks at these cases. His most recent analysis of search and seizure, done in collaboration with Harold Spaeth, is included in The Supreme Court and the Attitudinal Model (1993, 214-231). That analysis shows both the impact of factual elements and the strong impact of justices' attitudes (measured in the same way as in the analysis presented in the previous section). Our question is whether that analysis can be improved by taking into account legal regimes. Here we rely upon the same set of 196 cases included in the Segal and Spaeth analysis which covers the period 1963 through 1990. We also include the same variables used by Segal and Spaeth:

• object of the search (house, business, person, car, other)
• whether it was a full or partial search
• whether a warrant was obtained
• whether the lower court determined that the officer had probable cause
• whether the search was incident to a lawful arrest
• whether the search followed but was not incident to a lawful arrest
• whether the search followed an unlawful arrest
• whether the search fell under one of the accepted exceptions to the requirement for a warrant
• the justice's attitude

Segal and Spaeth's analysis included a total of 1,550 votes. We were able to include a total of 1,680 by including the votes of three justices (Black, Douglas, and Souter) omitted by Segal and Spaeth because of the absence of a measure of those justices' attitudes.⁽⁶⁾ In keeping with Segal and Spaeth's analyses, we coded the dependent variable so that a positive coefficient value indicated support for legality of the search.

We use Leon as our split point for legal regime because it and the Sheppard companion case are the last cases in the series of six precedents, and because the good faith exception reflects our general argument that the new regime shapes how the justices weigh the various factors included in Segal and Spaeth's analysis. This split gives us a total of 1,466 votes before Leon and 214 votes after and including Leon (and Sheppard).

The results of this analysis are shown in Table 2. The test of significance for differences in the coefficients before and after Leon yielded a highly significant chi square of 77.83 (11 degrees of freedom).⁽⁷⁾ We have to be a bit cautious in discussing certain types of before-after differences. Specifically, the difference in numbers of cases could readily account for why variables are significant before Leon, but not significant after. Consequently, we focus our attention specifically on differences in coefficients that the statistical analysis shows to statistically significant. The variables with such differences are indicated in bold in Table 2. These include:

• Object of the search: the justices appear to be much less likely to uphold searches of homes, business, and persons after Leon than before, everything else being equal.
• Before Leon, full searches (as opposed to partial searches) influenced the justices to strike down searches. After Leon, this variable does not influence the justices.
• Before Leon, the lower court finding probable cause was not statistically significant; after Leon it is, and the coefficient goes from essentially zero (0.05) to 1.20. After Leon, the odds of the justices' upholding the search increase by a factor of three if the lower court found probable cause; before Leon, a finding of probable cause had essentially no impact on the odds of the search being upheld.
• Before Leon, a search after but not incident, to a lawful arrest was more likely to be upheld; after Leon, such a search was much less likely to be upheld.
• Before Leon, the presence of one of the accepted exceptions to the warrant requirement increased significantly the likelihood that the search would be upheld. After Leon, there is no such influence (if anything, the presence of one of the exceptions might decrease the likelihood that the search would be upheld).
• The influence of the justices' attitudes increased sharply after Leon, with a conservative justice even more likely than before to uphold the search and a liberal justice more likely to find it unlawful.

The results of this analysis again support both the attitudinal model and our notion of legal regimes. As in Segal and Spaeth's analysis, the justices' attitudes are the best single predictor of justice's votes in search and seizure cases. However, as in the obscenity analysis, the magnitude of the influence of those attitudes and the influence of a variety of other factors depends on the legal regime.

FAILED ANALYSES AS FURTHER CONFIRMATION OF LEGAL REGIMES

We want to report briefly two analyses the increase our confidence in the existence of legal regimes. The first of these was our first attempt at the search and seizure analysis. Because of a misunderstanding of how cases were identified in Segal's data set, we ended up splitting the data not with the 1984 Leon case as we had intended but rather in mid-1973. When we ran the analysis with this incorrect split, the before-after difference was not significant.⁽⁸⁾

Second was an analysis of free exercise cases. We relied upon data originally collected by Joseph Ignagni. As noted previously, Ignagni's choice of period to include in his data set was intended to reflect a single set of controlling doctrines which is what we are calling a legal regime. The period covered by his set of 57 cases is from 1961 (Braunfield v. Brown, 366 U.S. 599) through 1990 (Jimmy Swaggart Ministries v. Board of Equalization of California, 493 U.S. 378). His own analysis (Ignagni 1993) includes:

• whether a marginal religious group was involved
• type of issue (employment, education, tax, other)
• whether the issue involved general government activities such as health, education, safety, etc.
• whether the law or practice involved seemed to treat all citizens in equal terms
• whether any religious practice or activity had a long history and tradition
• what position the Solicitor General took, if any, whether as a party of an friend of the court

One of the most important religion cases of the period covered by Ignagni's data was Lemon v. Kurtzman, 403 U.S. 602 (1971). While this case deals primarily with the Establishment Clause of the First Amendment, it does touch on free exercise, and Ignagni did include it in his set of free exercise cases. Given its prominence, we decided to see if we could generate results consistent with our legal regime hypothesis if we split the free exercise cases at Lemon. In the resulting analysis we had 124 votes before Lemon and 366 votes including and after Lemon. The test for change before and after Lemon yielded an insignificant chi square statistic of 7.21 (5 d.f.). Thus, as we expected, Lemon cannot be used to demarcate legal regimes in free exercise cases. We would expect, however, that if we had available a set of establishment clause cases that extended both before and after Lemon, we would find identifiable legal regimes in the area of jurisprudence with Lemon serving as the demarcation.⁽⁹⁾

DOCTRINE AS LEGAL REGIME

The analyses reported above relied on data generously provided by other authors. Although these data are useful for illustrating that legal regimes can shape how justices evaluate the elements of cases, a thorough analysis of a doctrine-based legal regimes requires a more nuanced empirical approach than those existing data permit. Recall that a doctrine-based legal regime establishes for the justices both the doctrinal case facts that should be relevant for their decisions and how the justices should balance or weigh those facts. To analyze the doctrinal variant of legal regimes, we rely on data collected expressly for this purpose. It is to this aspect of our project that we now turn

Doctrine

Doctrine is a standard that frames decision making in a particular area of law. Many precedents can exist in an area of law, but at times the Court establishes a doctrine: a specific formulation of the analytic standard which should structure Court decisions in a particular area by identifying relevant case facts and indicating how those facts should be weighed. A doctrine is a decisional standard established in case law that is expected to be more influential than other precedents on judicial decisions. A doctrine is established as a standard which should apply to the adjudication of all cases within a substantive area, so it is less likely to be distinguished as inapplicable to cases within its area than is ordinary precedent. A doctrine is a standard in the sense that it is the rule or test that must be met in order for an action of the state to be judged as constitutional. As an analytic standard, a doctrine identifies case facts that are relevant to how the members of the Court vote, and requires balancing or analysis of the relevant case facts. For example, the doctrine for content-based regulations asks the justices to consider whether the state's regulation is justified by a compelling government interest, and whether the regulation is narrowly tailored to achieve that interest. A doctrine is not merely a standard, but also indicates support for a normative position in a substantive area of law. These analytic standards are based on foundational normative principles.

Although we have stated that doctrines declare which facts are relevant for adjudication, the term "facts" is a misnomer. The doctrines we discuss are not focused on concrete, objective facts such as whether a speaker punched a police officer. Rather, they use abstract terms such as incitement to imminent lawless action or compelling government interest. These abstract terms require interpretation.⁽¹⁰⁾ We label these interpretive aspects of cases doctrinal facts in order to keep in mind that these are not facts as such but instead are interpretive components of cases. However, we retain the use of the word "facts" to indicate these interpretive components are the parallel of objective facts in other fact-pattern analyses of Supreme Court decision making, and to minimize the use of the clumsy phrase "interpretive components."

Consistent with our legal regime approach, we theorize that the influences of doctrinal facts on the justices' votes should change after a new doctrine is established. Doctrines inform the justices which doctrinal facts are relevant for decision making. Prior to the adoption of doctrines, justices may not even consider that such doctrinal facts should matter. In addition, doctrines tell the justices how to weigh doctrinal facts. After a doctrine is created, justices will have a better sense of how important particular doctrinal facts should be in their decision making.

To ensure that our model tests doctrines rather than simple precedents, we established criteria for deciding whether a particular case qualifies as establishing doctrine. Each of the following four criteria must be met for a case to be described aS establishing doctrine. A failure to meet any one of them would mean that a case had not established doctrine. Taken together, all four constitute sufficient conditions for doctrine. The justification for each of the four criteria is the principle that doctrine should be more authoritative than ordinary precedent. Doctrine should be authoritative because doctrine is expected to structure the justices' decision making. Doctrines apply to broad areas of law, and because they apply broadly, they identify what the relevant doctrinal facts for adjudication are. These doctrinal facts structure the justices' votes. Even if justices vote attitudinally, their attitudes must operate relative to facts (Segal and Spaeth, 1994). Facts do not exist a vacuum; they must come from somewhere. Doctrines are a key source. This is why doctrines are expected to be authoritative.

By contrast, many precedents are so particular to unique factual situations that they offer an inadequate guide to decision making in a broad area of law. When a justice thinks that a particular precedent lacks authority, it is unlikely that the justice would be influenced in his or her decision making by that precedent unless it was consistent with his or her ideology. The broad, authoritative character of doctrine is what makes it different from ordinary precedent. Each of the four criteria we present is based on the principle that doctrine should be more authoritative and influential than ordinary precedent. A doctrine should establish a standard for an area of cases, receive scholarly recognition for establishing a standard, be adopted by at least a five member majority of the Court in the case in which the doctrine originates, and not be overturned or substantially limited.

Doctrinal Criteria

First, a doctrine should present a formulation of an analytic standard in an area of law. This is the standard criterion. A precedent that establishes only a narrow holding that is particular to the facts of one case fails to establish a standard. In some areas of law, doctrines are not established because the case law is fractured into multiple precedents rather than a broad doctrinal theory. A doctrinal standard is formulated to identify the relevant case facts that should structure decisions in an area of law. As noted above, we label these case facts doctrinal facts, and a standard must identify doctrinal facts. To qualify as an analytic standard, a precedent must establish a test for how the justices should weigh or analyze doctrinal facts.

Secondly, a doctrine should be recognized by at least one scholar in the appropriate area of law as a case that formulates an analytic standard, not a mere precedential rule. For this research, the appropriate area of law is free expression. If a scholar has recognized the doctrinal character of a particular precedent, this provides more evidence of its status as a standard. This is the scholarly recognition criterion. Scholarly recognition is also useful because it is a criterion that is external to the Court.

Third, a doctrine should be adopted by at least a five-member majority of the nine-member Court. This is the five-member majority criterion. Anything adopted by less than a five-member majority would not be expected to be binding on future Court decisions. For example, if a standard was mentioned in a concurrence, but not in the main majority opinion, it would not be a doctrine. Likewise, if the Court could not establish a five-member majority but instead had a four-member plurality that agreed on an analytic standard, another member who wrote a separate concurrence that agreed with the plurality result but did not agree with the analytic standard, and four dissenters who also rejected the analytic standard, doctrine would not be established.

Finally, any precedent, including a doctrine, can be overturned, significantly modified, or substantially limited in practice. If this happens to a doctrine, then it loses its status as authoritative doctrine. This is the not overturned criterion.

When all four criteria are met, we recognize that a case constitutes a doctrine. We next present the doctrines we have identified for free expression law.⁽¹¹⁾

Free Speech and Free Press Doctrine

Before we can be certain that the various doctrines qualify as legitimate doctrines rather than mere ordinary precedents, we must apply the tests for doctrinal status elaborated above. In this section, we summarize the doctrines we have identified by applying the four doctrinal criteria to the body of free expression law.⁽¹²⁾ The key doctrines are the threshold doctrine, the two track doctrine, and the less protected categories doctrine.

Supreme Court first amendment free speech and free press doctrine can be expressed by three basic doctrines. First, cases must meet the threshold of first amendment protection. Cases in which free expression is not actually abridged or there is no government action do not invoke the protection of the first amendment. A second fundamental doctrine applies to nearly every case that raises a free expression claim. Tribe's (1988) interpretation of general free expression doctrine, his "two track" analysis, suggests that in all such cases, the Court asks whether the regulation in question is a content-based regulation aimed at the communicative impact or viewpoint of the expression, or a content-neutral time, place or manner regulation. According to the Court, content-based regulations merit the most rigorous scrutiny and are unlikely to be sustained as constitutional. Content-based, or track one, regulations must be narrowly tailored to serve a compelling government interest. A challenged regulation is not narrowly tailored if there is another regulation the government could have used which would be less restrictive of expression than the challenged regulation but would still achieve the government interest. Expression governed by content-neutral incidental or time, place or manner regulations receives less constitutional protection than expression regulated by content-based regulations. These regulations must be narrowly tailored to serve significant government interests. The two track doctrine was formally established in the Chicago Police Department v. Mosley (408 U.S. 92, 1972) and Grayned v. Rockford (408 U.S. 104, 1972) companion cases, where the Court struck down an attempt to prohibit all picketing except for labor picketing outside of schools as a content-based regulation which was not narrowly tailored. However, the Court also upheld a regulation of noisy picketing outside of schools as a narrowly tailored content-neutral regulation. Finally, certain regulations of expression receive less rigorous scrutiny because the Court has recognized that there are specific justifications for regulating these types of expression. These less protected categories include commercial speech, obscenity, broadcast media expression, expression in non-public forums, speech in a private forum against the will of the owner of the property, and libel against private figures. All of these doctrines meet the four criteria that we have established for a case to qualify as a doctrine.

Other Independent Variables

The general form of our argument is a factor (doctrine) which appears at a point in time causes changes (in votes) after that point. One danger inherent in this type of causal argument is that there are other factors that actually explain the changes over time. This is the logical fallacy post hoc, ergo propter hoc. To avoid this fallacy, three requirements must be met. First, there must be actual change over time associated with the factor (doctrine). Second, there must be a plausible reason why the factor would actually cause change. Finally, other potential explanations must be taken into consideration. In this section, we briefly delineate the other independent variables that we consider.

The ideologies of the justices on the Supreme Court are their preferences for the outcomes of cases in general issue areas. The attitudinal or ideological (we use the two words interchangeably) component of our free expression model has two dimensions: first, a general attitude regarding support for the individual versus the government; and second, attitudes toward groups or individuals which may influence support for the free expression rights of those groups or individuals.

The first, most general dimension of support for individuals versus the government is essentially the attitudinal model as presented by Segal and Spaeth, with one important modification. Segal and Spaeth see the attitudes of justices as varying along the dimension of liberal and conservative, although they really focus on whether the justices' attitudes support the individual or the government. Our modification emphasizes that the actual dimension of ideology considered is support for individuals versus government. There are a variety of connotations associated with the terms liberal and conservative that limit their usefulness here. Segal and Spaeth associate liberal with support for the individual and conservative with support for the government, but an important aspect of both liberalism and conservativism has been the idea of limiting the role of government in the lives of individuals.

The second dimension of ideology refers to the attitudes that the justices have towards particular groups of speakers and individual speakers, such as racists, racial minorities, feminists, military protesters, socialists, corporations, and members of religious groups. If the justices are biased in favor of or against the ideas presented by particular speakers, this may influence the willingness of the justices to protect their civil liberties. This theoretical point is supported by three distinct groups of scholars. First, public opinion scholars, observing that people were inconsistent in their willingness to protect the civil liberties of members of different groups, defined political tolerance as "a willingness to permit the expression of those ideas or interests that one opposes" (Sullivan, Piereson and Marcus 1979, 784). Although we do not adopt this definition for the purposes of our research, it makes the useful point that people's attitudes toward certain ideas or people who advocate those ideas may shape their willingness to protect the expression rights of the individuals and groups advocating those ideas. Although first amendment principles such as content-neutrality are supposed to make such considerations irrelevant, the justices are only human and may be influenced by such attitudes. Secondly, some Critical Legal Studies scholars have suggested that the attitudes of Supreme Court justices regarding whether to support the free expression rights of dissidents such as communists and war protesters has wavered according to historical events, shifts in societal consciousness and power relations, and has varied from justice to justice (Kairys 1990). Finally, some Critical Race Theorists have suggested that racial attitudes may also influence free expression decision making (Matsuda et. al. 1993).

In addition, the level of government involved in the case may matter. The justices may be more inclined to defer to the federal government than to other levels of government in free speech cases. As McGuire (1990) argued regarding obscenity cases, the presence of the solicitor general arguing for the federal government and other factors may lead to more support among the justices for the federal government.

The action the government takes against the speaker may also have an influence. The justices may be less sympathetic to criminal prosecutions of speakers, denial of opportunities for expression and disciplinary actions against lawyers than they would be to civil suits, regulations, and losses of government benefits or employment. There are not strong doctrinal justifications for why these factors would influence the justices, but they are worth evaluating in order to discount alternative explanations for doctrinal change.

For notes on the data sources and coding of variables, see Appendix B.

FREE EXPRESSION HYPOTHESES AND RESULTS

In general, we expect to see significant changes in doctrinal variables before and after the Grayned doctrine is established. For example, we expect that government regulations of expression that are not narrowly tailored will influence the justices to vote against the government after Grayned. Prior to Grayned, this doctrinal fact should not significantly influence the votes of the justices. For a specific list of variables and hypotheses, see appendix C. For a flow chart of the doctrinal variables we have generated based on the Court's free expression doctrines, see Figure 1.

Table 3 reports the results of three models, all of which consider the influence of the two tracks and the threshold doctrinal variable relative to the baseline less protected category. The first model is a model that covers all cases; the second model looks at the cases before the track two doctrine was established in Grayned; and the third model assesses the cases after Grayned. The overall model predicts 67.80 percent of the cases correctly, with a proportional reduction in error of 23.97, and a pseudo R-squared of 0.13. The sets of doctrine, action, government and identity variables are all significant at the p<.01 level. Looking at individual variables, the attitudes of the justices are a strong predictor of the justices' votes. The coefficient of the Grayned variable for the influence of the establishment of the two track doctrine is significant and shows the expected negative direction. As hypothesized, the adoption of the speech protective two track doctrine influences the justices to vote for the rights of speakers. Similar to other fact-pattern models of Supreme Court decision making,⁽¹³⁾ the variables in this model adequately explain the justices' votes. However, the more interesting question is whether the influences of the variables are conditioned by Grayned.

Turning to the before and after Grayned conditional models in Table 3, all of the sets of variables are significant at the p<.01 level both before and after, with the exception of the government set of variables after, which is significant at p=.016. The Chow test indicates significant differences for the before and after models (chi square 97.83, 20 d.f., p<.01). The before Grayned model predicts 72.08 percent of the cases correctly, with a proportional reduction in error of 18.65, and a pseudo R-squared of 0.13. The after Grayned model predicts 66.34 percent of the cases correctly, with a proportional reduction in error of 29.27 percent and a pseudo R-squared of 0.13.

Among the doctrinal variables, two stand out as being conditioned by Grayned in a manner consistent with the doctrine.

The most striking result is that the influence of content-based regulations is insignificant before and significantly negative after Grayned. This indicates that the adoption of the speech protective content-based doctrine influenced the justices to be more supportive of the speakers who were regulated based on the content of their speech, relative to speakers whose expression fell within the less protected categories.

The content-neutral variable is significantly positive before and insignificant after Grayned. The more speech protective track two doctrine made it harder to regulate after it was adopted and made it less probable that the justices would vote for the government in track two cases.

The threshold not met variable remains significant and positive both before and after Grayned, as expected, because the Grayned doctrine does not apply to the threshold issue. When speaker's claims do not meet the threshold of first amendment protection due to a lack of abridgment of speech or a lack of government action, the justices tend to vote against the speakers.

To this point, we have not examined the influence of the narrowly tailored and government interest analytic requirements, which we do in Table 4. Looking at these more detailed variables will facilitate a more detailed analysis of the influence of the two track doctrine.

A number of the non-doctrinal variables change before and after Grayned. We do not argue that the Grayned doctrine itself causes the changes, as these variables are not directly related to doctrine. However, it is likely that as new doctrines lead the justices to focus on specific doctrinal facts, the influence of some non-doctrinal facts will change. These changes include:

• The criminal and disciplinary action variables are insignificant prior but are significantly negative after the adoption of the Grayned doctrine.
• The "other" and military protester identity variables are insignificant prior but significantly positive after Grayned.
• The racial minority, alleged communist and religious speaker variables are significantly negative prior and insignificant in the after period.
• The newspaper variable is significantly negative before and significantly positive after Grayned. The broadcast media variable is significant and positive before. After, it is insignificant.

Several of the non-doctrinal variables are significant in the before and after conditional models. As expected, the influence of the justices' attitudes is significant both before and after. The influence of the federal government remains significant and positive before and after Grayned, as does the deny benefit action variable. The deny expression variable is significantly negative before and after.

Table 4 looks at the influence of narrowly tailored, compelling government interest, and significant government interest requirements. Again, there are three models: a model which considers all cases; a model for all cases before the two track Grayned doctrine was established; and a model for all cases after Grayned. The overall model predicts 70.53 percent of the cases correctly with a 30.41 proportional reduction in error, and a pseudo R-squared of 0.16. The doctrine, action, government and identity sets of variables are all significant at the p<.01 level.