A Test of Stare Decisis

Chapter 5
of
Gay Rights and American Law
Daniel R. Pinello
Cambridge University Press
2003

            The doctrine of stare decisis is central to traditional jurisprudential explanations of decision making in American courts. In theory, adjudged cases furnish examples or authority for analogous future fact patterns or similar questions of law. Stare decisis provides security and certainty, mandating established legal principles under which rights may accrue, be recognized, and be followed. The doctrine facilitates a predictable legal system.  Moreover, precedent undergirds the Langdellian case method, dominant for more than a century in American legal education (Stevens 1983; LaPiana 1994).

            Despite an abundance of theory, however, comparatively little purposeful empirical investigation of stare decisis has been achieved. Early studies by law professors and economists used judicial citation analysis to scrutinize precedential effect (Merryman 1954, 1977; Landes and Posner 1976),  but the technique inadequately captured the phenomenon (Klein 1976; Landes, Lessig, and Solimine 1998: 271-76). Early quantitative research in political science was chiefly descriptive (Ulmer 1959; Schmidhauser 1962; Danelski 1986; Banks 1992).  Not until the 1990s did law professors and economists (Cohen 1991; Schuck and Elliott 1991; Merrill 1992; Cohen and Spitzer 1994; Cross and Tiller 1998; Kerr 1998; Sisk, Heise, and Morriss 1998; Lim 2000) and political scientists (Pacelle and Baum 1992; Songer, Segal, and Cameron 1994; Brenner and Spaeth 1995; Brenner and Stier 1996; Knight and Epstein 1996; Segal and Spaeth 1996; Songer and Lindquist 1996; Gerber and Park 1997; Lindquist and Pybas 1998; Banks 1999; Brent 1999; Spaeth and Segal 1999; Spriggs and Hansford 2001) systematically address precedential impact.

            This research produced conflicting results, regarding both hierarchical (or “vertical”) stare decisis – where higher court decisions bind lower courts – and the collegial (or “horizontal”) variety  – where courts are obliged to follow their own prior decisions (Caminker 1994; Bergman 1996: 982-983; Camp 1997: 1662).

The Legal Literature

            Four empirical articles by law professors and economists published in legal periodicals focused on vertical stare decisis (Schuck and Elliott 1991; Cohen and Spitzer 1994; Cross and Tiller 1998; Kerr 1998), two on horizontal (Merrill 1992; Lim 2000), and two on federal district judges’ deference to other, nonbinding district court decisions (Cohen 1991; Sisk et al. 1998). The hierarchical studies investigated lower court submission to Chevron U.S.A., Inc. v. Natural Resources Defense Council, Inc. (1984), which required federal courts to defer to permissible agency interpretations of ambiguous statutes. One of the collegial articles explored the high court’s adherence to the same precedent (Merrill 1992). Indeed, Chevron was a treasure trove for law professors, begetting an academic cottage industry without which the legal literature evidently would be bereft of empirical fodder for precedential-impact research. 

            The four articles scrutinizing vertical stare decisis exhibit the contradictory findings of the literature. Schuck and Elliott (1991) performed the first consequential hierarchical study and identified a short-term stare decisis effect. They analyzed lower court review of agency statutory interpretations during six-month periods before and after Chevron and found a substantial increase in the rate of affirmances of agency action and significant decreases in the frequencies of remands and reversals after the 1984 precedent, although the effects had weakened somewhat by 1988. The authors concluded that Chevron did indeed cause lower courts to defer more to agencies, as the Supreme Court mandated. Cohen and Spitzer (1994) documented that the appellate court affirmance rate of agency interpretations decreased from the mid–70 percent range in 1983-87 to the mid–60 percent range in 1988-90, discrediting Chevron’s long-term impact. Cross and Tiller’s (1998) investigation of 170 agency reviews by the District of Columbia Circuit between 1991 and 1995 led to the conclusion that judges’ partisanship far better explained case outcome than obedience to precedent. In contrast, Kerr (1998), examining 223 published circuit court decisions applying Chevron in 1995 and 1996, found that 38 percent of courts engaging in two-step analyses resolved the interpretive issue at step one, and 62 percent at step two, while 58 percent of the step one cases reversed the agency, but only 11 percent of the step two cases did – suggesting that Chevron had its intended effect through the mid-1990s.

            Merrill (1992), a horizontal stare decisis study, found that before Chevron, the Supreme Court deferred to agencies 75 percent of the time, while afterward the overall high court deference rate was 70 percent, and that for the cases in which the Court cited the Chevron two-step procedure, it deferred only 59 percent of the time – hardly frequencies manifesting the Court’s unswerving fidelity to its own policy.

            The confusion of outcome among these appellate court studies may be attributable in part to a serious methodological flaw: ecological fallacy, or drawing conclusions about individuals based solely on the observation of groups (Robinson 1950).  Schuck and Elliott, for example, introduced their conclusions about Chevron’s precedential impact as follows:

In recent years, scholars associated with the Critical Legal Studies movement have emphasized that legal doctrine is “indeterminate” in the sense that it does not constrain judges in their resolution of disputes. These “law skeptics” argue that because legal rules are “sufficiently ambiguous or internally contradictory to justify any result we can imagine,” judges can use it to rationalize decisions reached on other, usually undisclosed grounds.

            This claim, law skeptics acknowledge, is at bottom an empirical one, and it has implications that can be tested. At least in its strong form, the claim implies that doctrinal changes such as Chevron should not affect the pattern of results reached in subsequent cases by lower courts, provided that the judges’ personal and political predilections remain essentially unchanged during the period under study. (1991: 1029; footnotes omitted; emphasis added)

Clearly, the authors were concerned here with how attributes of judges affected outcomes in cases, but their unit of analysis was the case, not the judge. Had their empirical investigation focused on judges, the findings may have varied significantly. Likewise, the units of analysis in Merrill (1992) and Cohen and Spitzer (1994) were cases, not judges.

            The other horizontal stare decisis study, by an economist, is the most statistically sophisticated. Lim (2000) used logit regression and linear probability models constructed from factor analysis to demonstrate that Supreme Court justices’ votes in precedent-setting cases very much influenced their actions in later decisions. Yet Lim’s article suffers from methodological limitations, too. For example, his data set is confined to just thirteen judges over an eight-year span, and as Supreme Court case studies, his models do not incorporate important institutional or environmental variables. Moreover, the paper fails to report proportional reductions in error for the logit models, while the adjusted R2 values for the linear probability analysis are often modest at best. These shortcomings hinder the generalizability of, and confidence in, Lim’s findings and conclusions.

            With regard to the empirical examination of federal district judges’ deference to other district court decisions, Cohen (1991) conducted a study of 196 judges adjudicating the constitutionality of the federal Sentencing Guidelines and found none of his precedent variables was statistically significant, while Sisk et al. (1998: 1497), investigating 291 district court votes on the same issue, did conclude that “judges appear to find the written opinions of other judges valuable as persuasive precedent when they confront difficult issues and there is no binding higher authority” – again, contradictory results.

The Political Science Literature

            Five published studies by political scientists examined vertical stare decisis. Gruhl (1980) found that 93 percent of libel decisions by U.S. Courts of Appeals between 1964 and 1974 fully complied with Supreme Court mandates. Using lower federal court reactions to fourteen randomly selected Supreme Court decisions between 1950 and 1975, Johnson (1987) found inferior courts did heed high court precedent. Pacelle and Baum (1992) studied cases remanded by the Supreme Court during its 1965 through 1974 terms and documented that the high court’s authority significantly shaped the behavior of lower court judges. However, Songer et al. (1994), tracking how Supreme Court search-and-seizure precedents fared in federal circuit courts, reached mixed conclusions about the impact of stare decisis. Although Songer and Sheehan (1990) confirmed New York Times v. Sullivan’s (1964) very substantial impact in federal libel law, those authors discovered no impact at all of Miranda v. Arizona (1966) in the federal criminal procedure context. A sixth article, examining Supreme Court cases in which justices were faced with overturning their own precedent from lower courts on which they served, determined that the most important information for anticipating justices’ behavior was their policy preferences (Gerber and Park 1997).

            Most of the published precedential-impact research in political science focused on horizontal stare decisis. Segal and Spaeth surveyed a 40 percent random sample of landmark, nonunanimous U.S. Supreme Court decisions between 1953 and 1995 and checked the frequency with which dissenting justices in precedent-setting cases followed, in progeny, the majority opinions with which they disagreed. The researchers found justices voted their preferences, and against precedent, 90.8 percent of the time, concluding:

While we have no argument with Cardozo’s claim that stare decisis should be the rule and not the exception, the empirical results are to the contrary. Potter Stewart and Lewis Powell are the sole exceptions among modern Supreme Court justices who virtually never subjugate their preferences to the norms of stare decisis. (1996: 987)

            In contrast, Knight and Epstein believe that precedent acts as a meaningful constraint on judicial decision making, just as a norm of consensus impeded the filing of separate opinions during the Marshall Court era. Counting attorneys’ use of authorities in case briefs, justices’ appeals to precedent during conference, their invocation of precedent in opinions, and the Court’s infrequent alterations of stare decisis, the authors concluded that justices strategically modify their positions in deference to the stare decisis norm in order to produce outcomes close to policy preferences. Knight and Epstein view the Segal and Spaeth research design as myopic:

[J]ust as one would be unable to make claims about the operation of the no-dissent norm by looking only at the content of Marshall Court votes, one would be unable to demonstrate the importance of precedent by merely considering dissents cast by justices in the progeny of important cases. (1996: 1019)

            Brenner and Stier (1996) tracked how the four center justices of the Warren Court conformed to the precedent of thirty-six major decisions from which they dissented, used selection rules for progeny cases different from Segal and Spaeth, and also included memorandum and per curiam decisions as progeny. The research discovered the four center justices complied with precedent 47 percent of the time.

            Songer and Lindquist (1996) criticize Segal and Spaeth as misunderstanding the nature of the choices justices face when addressing precedent, oversimplifying the process of stare decisis, and using an unrepresentative sample of Supreme Court decisions. The authors recoded some of the Segal and Spaeth data and included summary decisions associated with their landmark cases, determining that support for preferences dropped to 68.8 percent, compared with the earlier finding of 89.4 percent (Segal and Spaeth 1994).

            In their “first falsifiable, systematic test of the influence of stare decisis on the behavior of U.S. Supreme Court justices,” Spaeth and Segal (1999: 315), in the book-length sequel to their 1996 journal article and the most ambitious empirical examination of stare decisis to date, argue that precedent is key when compelling justices to vote for case outcomes they otherwise would not support. Indeed, scholars testing stare decisis should not be concerned primarily with precedent’s preaching to the choir – or to congregants generally. Judges already adhering or indifferent to policies that precedents promote are not the attentive audience for gauging stare decisis’s coercive power. Rather, precedent is consequential when it converts atheists (to strain the metaphor). Judge Jerome Frank recognized this long ago, writing that “[s]tare decisis has no bite when it means merely that a court adheres to a precedent that it considers correct. It is significant only when a court feels constrained to stick to a former ruling although the Court has come to regard it as unwise or unjust” (United States ex rel. Fong Foo v. Shaughnessy 1955: 719).

            Thus, researchers’ methodological challenge is specifying the jurists constituting the relevant yardstick for measuring the full import of stare decisis: the precedential atheists. In fact, much of the precedential impact literature is irreparably flawed because it fails to disaggregate judges into precedential congregants and indifferents, on the one hand, and atheists, on the other. Without the distinction, scholars cannot make meaningful estimates of stare decisis’s constraining effect.

            Spaeth and Segal met the challenge again by investigating, now throughout the Court’s history, how dissenting justices in both major and minor precedent-setting decisions voted in progeny cases. Their inquiry continued an insistence that justices vote twice, once in dissent and again in progeny cases. The book indubitably is a pathbreaking achievement whose primary – and substantial – methodological virtue is its caution: Justices’ issue positions are clear once they have voted publicly, and there is no need for conjecture about how they would vote with or without precedent. Indeed, any other vote-based research design testing precedential impact necessarily will be less scrupulous than Spaeth and Segal’s approach. Certainly, the least attractive surrogates involve speculating about how judges would vote under the sway of precedent. That option is untenable, begetting too much uncertainty. Rather, alternative empirical strategies minimally must require votes subjected to the influence of stare decisis and then ascertain precedential atheists by other means.

            Spaeth and Segal’s admirable methodological clarity and circumspection come with a considerable price. For example, their analysis disregards the clout of unanimously decided precedents, the ones most likely to command respect in progeny cases. Chief Justice William Rehnquist, writing for the Court in Payne v. Tennessee (1991), argued that stare decisis carries less force for decisions rendered by close margins. Moreover, scholars’ use of only nonunanimous opinions has been identified as a research-design moderator distorting findings of variable strength for the broad range of judicial action. 

            The title of the book by Spaeth and Segal, Majority Rule or Minority Will, should more properly be Institutional Precedent versus Personal Precedent. The psychological dynamic of justices’ adherence to personal precedent prevents the generalizability of the results.  Publicly switching policy positions requires political actors with abundant humility, a character trait typically in short supply on the high bench. The Spaeth and Segal research blueprint in fact focuses on the least psychologically prepared group of jurists likely to bend to precedent’s bidding. The authors require that justices not only acquiesce to policies they oppose but also to contradict themselves conspicuously.

            The research design conceived a crucial case study, that is, if precedent works here, then it must work everywhere. Spaeth and Segal discovered that justices later deferred to majority positions from which they dissented only 11.9 percent of the time and concluded that “the justices are rarely influenced by stare decisis” (1999: 287-88). But as noted at the end of the last chapter, a failure of proof for a crucial case study hypothesis does not confirm its obverse, that is, if precedent does not work here, then it does not work anywhere. Hence, Spaeth and Segal’s finding does not warrant an inference that precedent is impotent under less exacting – and more frequently occurring – conditions. The book mistakenly conflates the imperatives of its ingenious research technique with the universe of latent judicial obedience to stare decisis.

            Spaeth and Segal in fact test the resolve of justices only under singular conditions. Clearly, for example, a group of jurists more pliable to precedent’s siren call are those opposed to precedential policies but joining the bench after they were rendered. There, judges do not have to humble themselves publicly by abandoning vanquished positions, and thus, submission to stare decisis may be more pervasive than Spaeth and Segal’s divining rod detects.

            Inconspicuous compliance with stare decisis, furthermore, promises a vastly larger usable-vote pool than Spaeth and Segal’s circumscribing methodology permits. Approximately 214,000 votes by justices in signed opinions of the Court have been cast since its creation.  Unaware of the actual number decided under the authority of stare decisis, I adopt, for argument’s sake, the same proportion of precedent-controlled cases for the Supreme Court found among court-of-last-resort decisions in the present study: 17.4 percent. Relying on this premise, I estimate a total of 37,236 precedent-bound high court votes, which indicates that the 2,425 qualifying for Spaeth and Segal’s inspection (1999: 287) represent roughly 6.5 percent of the Court’s stare decisis-constrained output.  Surely, even the most hostile adversaries of the legal model (Segal and Spaeth 1993) cannot believe stare decisis is potentially decisive at best less than 7 percent of the time. If that were true, their 300-plus-page book would be the quintessential quixotic quest.

            The Spaeth and Segal analysis, dependent on dissenting justices’ remaining on the bench, also inflicts a severely limiting temporal constraint on the progeny votes collected. The mean judicial tenure for all justices (through Lewis Powell) completing their service is sixteen years, four months.  The Spaeth and Segal approach, then, allows about sixteen years, on average, for progeny cases to appear – presuming, of course, that justices dissent in precedent-setting cases at the beginning of high court careers, a sizable postulate itself. As a result, Spaeth and Segal miss progeny cases decided, on average, more than sixteen years after precedent-setting decisions. This is important, considering how long precedents last. The mean number of years between overruled and overruling cases for the more than 200 overruled decisions in Table 2-14 (“Supreme Court Decisions Overruled by Subsequent Decisions, 1789-1990") of Epstein, Segal, Spaeth, and Walker (1994) is 28.7 years. Similarly, Brenner and Spaeth (1995) calculated the mean age of overruled decisions between 1946 and 1992 to be 29.4 years. Thus, from the perspective of overruled cases, the Spaeth and Segal research design overlooks, on average, about thirteen years’ worth of progeny (i.e., more than 40 percent). However, overruling delineates the minimum span of precedential life. Since the overwhelming majority of precedents are never overruled (Ulmer 1959; Knight and Epstein 1996), the prospect for multigenerational progeny is patent. Accordingly, Spaeth and Segal tap only a peripheral portion of progenial potential.

            Of course, the authors reasonably might retort to these methodological criticisms: “Yes, that’s all true, but how otherwise do we pinpoint precedential converts never publicly coming out of the atheist closet?” This chapter answers the question, and avoids the pitfalls detailed above, by offering a different empirical strategy to test stare decisis.

            I note again the challenge in analyzing precedential potency: identifying objective indicators demonstrating that judges, faced with controlling precedent compelling a particular policy outcome, indeed follow that authority, especially when personal attitudes, or environmental or institutional forces, militate against the result. To do so, researchers must measure how judges would vote in the absence of precedent in order to match that assessment with how they actually vote under its putative authority. No meaningful empirical appraisal of stare decisis is possible without reliable yardsticks for gauging judicial votes in precedent-free environments.

            Scanning adjudicated cases alone, nonetheless, usually provides only one side of a comparison. A control is necessary to measure impact, and excessive speculation about counterfactuals is not a sufficient proxy. Yet opportunities to observe, in the same policy domain, adequate numbers of both court decisions resolved within the presumed constraint of precedent and those without it are rare.

            This book is such an occasion.

Lesbian and Gay Rights Claims and Precedent

            Lesbian and gay rights cases offer an excellent opportunity for empirical study of precedential leverage. Through the 1960s, American society – and particularly its courts – treated homosexuals as pariahs. In January 1966, for example, Time magazine published an unsigned essay, “The Homosexual in America,” that concluded:

[Homosexuality] is a pathetic little second-rate substitute for reality, a pitiable flight from life. As such it deserves fairness, compassion, understanding and when possible, treatment. But it deserves no encouragement, no glamorization, no rationalization, no fake status as minority martyrdom, no sophistry about simple differences in taste – and above all, no pretense that it is anything but a pernicious sickness. (1966, January 21, p. 41)

            A 1965 Harris public opinion survey rated homosexuals the third most dangerous group (after communists and atheists) in the United States; a 1966 CBS poll found that only one in five respondents would support legalizing homosexual behavior between consenting adults; and a 1969 Harris poll found that 63 percent of respondents considered homosexuals to be harmful to American life (Weinberg and Williams 1974: 19-20). The federal government prohibited employment of gays in the civil service until 1975 (Lewis 1997). Symptomatic of society’s fearful attitude toward lesbians and gay men, the American Psychiatric Association classified homosexuality as a mental disorder until 1973 (Marcus 1993: 172-73).

            Homophobia also was the legal norm in the 1960s:

The homosexual in 1961 was smothered by law. She or he risked arrest and possible police brutalization for dancing with someone of the same sex, cross-dressing, propositioning another adult homosexual, possessing a homophile publication, writing about homosexuality without disapproval, displaying pictures of two people of the same sex in intimate positions, operating a lesbian or gay bar, or actually having oral or anal sex with another adult homosexual. The last was a serious felony in all states but one, and in most jurisdictions also carried with it possible indefinite incarceration as a sexual psychopath. Misdemeanor arrests for sex-related vagrancy or disorderly conduct offenses meant that the homosexual might have her or his name published in the local newspaper, would probably lose her or his job, and in several states would have to register as a sex offender. If the homosexual were not a citizen, she or he would likely be deported. If the homosexual were a professional – teacher, lawyer, doctor, mortician, beautician – she or he could lose the certification needed to practice that profession. If the charged homosexual were a member of the armed forces, she or he might be court-martialed and would likely be dishonorably discharged and lose all veterans’ benefits. (Eskridge 1999: 98) 

            As a result of gays’ outcast status, a federal judge at the end of the 1960s could assert prosaically, without fear of appellate or political censure, that “[a]ny schoolboy knows that a homosexual act is immoral, indecent, lewd, and obscene” (Schlegel v. United States 1969: 1378).

            Very few court decisions before 1970 addressed legal claims by self-identified and self-affirming lesbians and gay men (Cain 1993: 1564-79). For example, a Westlaw database search of state court cases between 1960 and 1969 for the keywords “homosexual,” “lesbian,” and their cognates retrieved relevant decisions almost exclusively in four categories: revocation and suspension of liquor licenses of bars and restaurants catering to homosexuals; revocation of homosexuals’ teaching licenses; determinations of whether homosexual solicitation or other activity constituted gross indecency, disorderly or lewd conduct, or comparable offenses; and prosecutions for possession and distribution of obscene materials with lesbian and gay content – in other words, disputes typically with closeted lesbians and gay men in defensive posture. All but three  appellate decisions were resolved against the interests of the gay people involved. State appellate court treatment of homosexuals during the 1960s is best illustrated by the disposition in a rare child visitation case: A Los Angeles trial court granted a gay father visitation rights with his children on the condition that he

“immediately quit his present residence [with another man] and take up residence in the home of his parents”; [and] obtain psychiatric treatment and continue with such “care until further order of the Court.” The order also required that the paternal grandmother accompany the children “at all times” during [the father’s] custody of them. (Evans v. Evans 1960: 568)

            When the mother remarried and moved to Salt Lake City with the children, the California Court of Appeal ordered the gay father to continue making child support payments, even though the children’s relocation to Utah effectively terminated his visitation with them.

            In essence, American law in the 1960s concerning what today would be considered lesbian and gay rights claims – for example, child custody, visitation, adoption, and foster care by lesbian or gay parents or their domestic partners; health insurance and other benefits for domestic partners; same-sex marriage or its approximation; the constitutionality of consensual sodomy laws; and discrimination in public accommodations, housing, and the workplace – was a tabula rasa. Doubtless, legal precedents existed regarding child custody and visitation, marriage, invidious discrimination in public accommodations and the workplace, and so forth. And, doubtless, homosexuals were parties to lawsuits (albeit reluctantly in the kinds of cases outlined above). But courts in the 1960s and before did not apply precedents created with heterosexual litigants to lesbians and gay men because they were perceived as deviants unworthy of equal legal treatment (Pacelle 1996: 200; Wald 2000: 9). Moreover, although homosexuals and their advocates argued that civil rights precedents from the African American and other minority contexts should be applied to homosexuals, that thesis required courts’ acknowledging lesbians and gay men as a legitimate minority, which was universally denied in the 1960s.  Indeed, even through much of the 1970s and early ’80s, appellate cases on lesbian and gay rights claims virtually ignored civil rights jurisprudence since judges apparently were unwilling to sully those precedents with even the suggestion that homosexuals were in any way comparably circumstanced to African Americans and other minorities.

            Further, even in the rare instances that legal standards such as “best interests of the child” were invoked in cases with lesbian and gay litigants, courts’ next sentences would be variants of “and the best interests of the child require as close to total separation from these immoral, sexual-predator monsters as possible.” Again, the Evans and Schlegel cases epitomized contemporary judicial attitudes toward homosexuals. Accordingly, the 1960s and before were to lesbians and gay men what much of the nineteenth century was to people of color in this country. Homosexual rights was an oxymoron in American law. 

            The lesbian and gay rights social movement changed that. New York’s Stonewall Rebellion in the summer of 1969 touched off mass protests by newly self-identified, out-of-the-closet lesbians and gay men (Marotta 1981; Duberman 1993; Button, Rienzo, and Wald 1997: 25; Keen and Goldberg 1998: 91-94). As a result, beginning in the 1970s, American courts started resolving disputes with self-proclaimed and self-confident lesbians and gay men as parties to lawsuits, and a pool of gay civil rights precedent slowly took form.  That process continues today, with the mean of the annual number of appellate court lesbian and gay rights decisions increasing from 11.8 for the ten years beginning 1981 to 27.3 for the following decade (with 36 in 2000). Hence, because of homosexuals’ thirty-year metamorphosis from pariahs to ordinary litigants, lesbian and gay rights cases offer a policy area inviting comparison of judicial decision making in precedent-free and precedent-bound environments.

            In addition, lesbian and gay rights, like abortion, is an issue domain prompting strongly held positions, at both the mass and the elite levels (Haeberle 1999; Schroedel 1999). Judges thus are not apt to be indifferent to lesbian and gay rights claims and are more disposed to vote their attitudes there than in other, less controversial areas, simply because the topic is so emotionally charged (Rand National Defense Research Institute 1993; Sherrill 1993, 1996; Gamble 1997).

            Yet, as lesbian and gay rights precedents accumulated, nonlegal factors such as attitudinal attributes should not have had the free rein to influence decisions in the late 1980s and '90s as they did in the 1970s and early '80s, if stare decisis in fact plays a decisive role in adjudication.

            I test precedential power in lesbian and gay rights jurisprudence in two ways. First, I reconsider the stare decisis components of Chapter 3's models. The book’s database permits a level of comparative analysis, incorporating environmental and institutional dynamics, not possible in a U.S. Supreme Court case study such as that of Spaeth and Segal (1999). 

            Second, since more than two-thirds (69.8 percent) of the votes in the lesbian and gay rights data here were cast without the influence of binding precedent, I identify precedential atheists in the 30.2 percent decided under the authority of stare decisis by estimating the probabilities of how precedent-bound votes would be cast in precedent-free environments and then inspect how the probability-delineated atheists actually voted. In other words, the precedent-free votes serve as benchmarks for how judicial attitudes, case characteristics, environmental factors, institutional forces, time, and amicus participation affected judicial dispositions of lesbian and gay rights claims, independent of stare decisis. My methodology relaxes the Spaeth and Segal two-judicial-vote requirement to one – that done in the ambit of binding precedent – and consequently expands the parameters of the usable-vote pool.

Test One

            Chapter 3's models demonstrate that stare decisis was a powerful force in the adjudication of lesbian and gay rights claims. Positive precedent was statistically significant in six of seven layers of analysis, three at the .001 level. Its estimates were uniformly positive, and its impact statistics ranged from .236 to .374.  Among 38 other variables and interaction terms, just one (minority) was more consistently significant than positive precedent, and only in the singular subpopulation of opinion writers (where attitudinal factors understandably held greatest sway) was minority’s impact ever larger than that of positive precedent. At the same time, negative precedent was significant in two levels of analysis. Hence, stare decisis undeniably is a powerful force in judicial decision making. The only issue is whether its influence can be further explicated.

            The intermediate appellate court model permits more nuanced stare decisis specification because precedents there come from the same courts or higher ones. Also, intermediate appellate court judges sometimes cite decisions from other districts in their states as authority (although, as observed in Chapter 3, sister-district cases usually are not controlling). Accordingly, six stare decisis variables in the intermediate appellate court model here substitute for Chapter 3's two. Descriptions of the new variables and their coding rules appear in Appendix 5.1. All cases in the study decided under the influence of stare decisis are listed in Appendix 5.2, together with relevant controlling precedents. 

            Appendix 5.3 exhibits the intermediate appellate court regression analysis with the expanded stare decisis variables. Four of the six are statistically significant there: negative precedent, same court; negative precedent, higher court; positive precedent, higher court; and negative precedent, other district. In addition, positive precedent, same court is on the verge of significance (p = .051; impact = +.166). That five of six precedent variables are consequential (with four signed as predicted) is not surprising because Table 3.4 demonstrated intermediate courts to be the most responsive to the legal model. In addition, the positive signs of both higher court estimates intimate that hierarchical stare decisis constrained prospectively conservative intermediate appellate court votes, while the estimate for negative precedent, other district implies that liberal votes were responsive even to nonbinding sister-district decisions.

            Equally important, the dominant impact statistics of the positive-precedent variables throughout Chapter 3 and here, coupled with the comparative inability of negative precedents to defeat lesbian and gay rights claims, suggest that stare decisis encumbered conservative votes more reliably than liberal ones. Further, same-court negative precedents having dependably negative weight, while higher court negative precedents do not, hints that liberal court-of-last-resort justices responded differently to stare decisis than did liberal intermediate appellate court judges. These clues to differential precedential impact are explored more conclusively in the next section.

Test Two

            I use the beta coefficients from the precedent-free versions of Chapter 3’s five principal models to estimate vote probabilities for both precedent-free and precedent-bound votes in each model, thereby predicting how all votes would be cast in precedent-free environments. I identify precedential atheists among the precedent-bound votes from those probability estimates by comparing judges’ probabilities with how they actually voted in the ambit of stare decisis. Following earlier precedential-impact research (Brenner and Stier 1996: 1037) and Chapter 3's practice, I divide the probability scale into thirds.

            Table 5.1 displays frequency distributions for the six actual-vote subpopulations of each model determined by outcome and either the absence of precedent or the presence of negative or positive precedent: (1) actual votes against lesbian and gay rights claims without the influence of precedent; (2) votes against claims with negative precedent; (3) votes against claims controlled by positive precedent; (4) votes in favor of claims without precedent; (5) votes favoring claims with negative precedent; and (6) votes favoring claims with positive precedent. Further, actual votes are classified according to their probability estimates into four categories: suppressed, not suppressed, inconsistent, and consistent. For example, the first row of Table 5.1 presents the 482 votes cast in all 398 decisions and opinions against lesbian and gay rights claims without controlling precedent. Some 204 votes have probability estimates under .333, qualifying them as conservative and therefore consistent votes, since conservative votes by definition opposed lesbian and gay rights claims. Likewise, 216 votes have probability estimates between, or equal to, .333 and .667, designating them as center and also as consistent since center votes by definition were not predisposed to particular outcomes. However, the 62 liberal votes, with probability estimates above .667, are inconsistent, since, ceteris paribus, they should have favored the claims, not opposed them. The second row of Table 5.1 arrays the 188 votes against lesbian and gay claims with controlling negative precedent. Again, the conservative and center votes are consistent, but now stare decisis suppressed the 20 liberal votes of precedential atheists not abiding by predispositions to support the lesbian and gay claims. In the third row, displaying votes against claims with controlling positive precedent, stare decisis did not suppress the 16 conservative votes of precedential atheists there, since they were cast according to predilections to oppose the claims and against the positive precedent. The center and liberal votes are inconsistent since the former was indifferent to outcome and thus should have followed the positive precedent, while the latter favored claims regardless of positive precedent. The other rows and model subpopulations in Table 5.1 follow similar analyses.

Table 5.1

Frequency Distributions of Precedent-Free Vote Probabilities by Subpopulation

 Legend for Characterizations of Conservative, Center, and Liberal Votes:

            Table 5.2 summarizes Table 5.1's probability-distribution data  and reveals several noteworthy phenomena. First, as intimated in test one, stare decisis, unsurprisingly, was more influential in intermediate appellate courts than courts of last resort. Controlling precedent governed almost 70 percent of precedential atheists’ votes in the intermediate appellate court pool, whereas only about 30 percent of pertinent court-of-last-resort votes deferred to it. The latter finding comports with Brenner and Stier (1996) and Songer and Lindquist (1996), who concluded that stare decisis constrained Supreme Court justices between about 30 and 45 percent of the time. The discovery also challenges the 12 percent compliance rate of Spaeth and Segal (1999), the most comprehensive study in both the legal and political science literatures separating precedential atheists from judges sympathetic or indifferent to the policies precedents promote. As explained in this chapter’s introduction, their book’s abridged vision of the full gamut of latent precedential impact is the price paid for a supremely scrupulous yet circumscribing methodology.

Table 5.2

Summaries of Precedent-Free Vote-Probability Distributions from Table 5.1



            Table 5.2's second prominent feature is that stare decisis subdued conservative votes at a noticeably higher rate than liberal ones. The smallest proportional difference occurred in the all-decisions model, where the votes of conservative atheists were suppressed at a rate 26.8 percent higher than those of liberal atheists (66.7 vs. 52.6 percent). Courts of last resort had a 33.3 percent proportional difference (36.4 vs. 27.3 percent), while the largest was 54.5 percent in the nonfamily model (69.2 vs. 44.8 percent). Not surprisingly, opinion writers (not displayed in Tables 5.1 and 5.2) were the only conservative precedential atheists not suppressed more frequently than liberal counterparts. The respective suppression rates in that subpopulation were 37.5 and 45.5 percent. Yet only eight relevant conservative votes were in play, compared with 22 liberal ones, thus limiting the comparison’s reliability.

            The substantially higher precedential suppression rates for conservative votes, compared with liberal ones, are not easily explained. Theorists might proffer that stare decisis is innately conservative and therefore bridled even the sympathetic, ideologically conservative forces opposing the lesbian and gay rights claims studied here. But empirical evidence exogenous to this study supporting the theory is hard to find. For example, matching justices’ civil liberties voting scores from Table 6-1 (“Aggregate Liberal Voting of Justices, 1953-1991 Terms”) of Epstein et al. (1994) with their participation rates in precedential alteration from Table 6-8 (“Votes in Support of and Opposition to Decisions Formally Altering Precedent, 1953-1991 Terms”) of the same volume uncovers no significant correlation.  Cross-tabulating the civil liberties scores with corresponding percentages of preferential versus precedential votes from Table 9.1 (“Justices’ Precedential Behavior by Case Type”) of Spaeth and Segal (1999) discloses a modest relationship.  Instead, the relative liberal atheist defiance of stare decisis in my study may be an artifact of improved public opinion on lesbian and gay issues (Yang 1998; Blendon et al. 2000: 26-29; Wilcox and Wolpert 2000). Yet post-1990, Chapter 3's temporal control, was not consistently significant across models and thus provides little support for this interpretation.  In any event, further quantitative precedential impact research on other legal subjects is necessary for conclusive documentation of a conservative-liberal stare decisis fidelity gap.

            Tables 5.3 and 5.4 display the votes of precedential atheists in Tables 5.1 and 5.2 disaggregated by vertical and horizontal stare decisis effects.  Although the total-vote suppression rates in the all-decisions model are nearly identical in the two tables, the gap between conservative and liberal suppression rates is substantially higher for hierarchical stare decisis (71.4 vs. 38.9 percent) than collegial (63.0 vs. 56.0 percent). In other words, conservative precedential atheists honored higher court precedents more faithfully than same-court precedents, while liberal atheists gave less deference to higher authority than to colleagues. The latter finding is counterintuitive and flies in the face of stare decisis theory. Yet some consolation is available comparing Table 5.4's court-of-last-resort data with those for intermediate appellate courts. Liberal high court atheists honored collegial precedent about a quarter of the time, while those on lower courts did so two and a half times more frequently (27.3 vs. 70.8 percent). Indeed, the liberal court-of-last-resort suppression rate is the lowest in Tables 5.2–5.4.

Table 5.3

Votes of Precedential Atheists Subject to Vertical Stare Decisis

Table 5.4

Votes of Precedential Atheists Subject to Horizontal Stare Decisis



            Table 5.5 segregates the vertical and horizontal effects of state court precedential atheists.  Both the conservative and total-vote suppression rates in the all-decisions model demonstrate greater deference to higher court precedents than to same-court (80.0 vs. 66.7 percent, and 77.3 vs. 59.5 percent, respectively). In contrast, the liberal rates of the intermediate appellate court model are reversed (40.0 vs. 72.2 percent), while too few votes-in-play in the all-decisions model prevent meaningful comparison between liberal-vote levels there. Thus, although the combined federal and state court votes of Tables 5.3 and 5.4 do not manifest a consistently greater potency of hierarchical stare decisis compared with collegial, the dominant state court votes of Table 5.5 do. Accordingly, judicial federalism may be in play here, too.

Table 5.5

Votes of State-Court Precedential Atheists Only

Subject to Vertical Stare Decisis 



Subject to Horizontal Stare Decisis

            Indeed, variation in vertical and horizontal stare decisis effects should be the norm. Subject to both varieties of precedential influence, intermediate appellate courts responded to each at relatively similar rates, between approximately 65 and 75 percent of the time, in Tables 5.3–5.5. Remarkably, conservative and liberal suppression rates are both high in the intermediate appellate court and family models of Table 5.4 and the horizontal intermediate appellate court data of Table 5.5 – aberrations from the general trend of greater liberal defiance. Hence, liberal votes in nonfamily court-of-last-resort decisions were the least pliable to stare decisis.

            Subject only to collegial precedential effects, courts of last resort are the fly in the ointment of stare decisis. Court-of-last-resort suppression rates are by far the lowest of any in Tables 5.3–5.5, again reflecting the high courts’ greater policy-making role and their resultant propensity to follow attitudinal, environmental, and institutional forces more than legal ones.

Caveats

            The chapter’s introduction catalogs pitfalls in Spaeth and Segal’s landmark precedential impact testing scheme. Let me be equally candid about problems with my alternative methodology. To repeat: Any vote-based approach to gauge the potency of stare decisis different from Spaeth and Segal necessarily will be less cautious in ascertaining precedential atheists. Nothing predicts a vote as well as, or better than, the vote itself. By relaxing Spaeth and Segal’s two-judicial-vote requirement to a single precedent-influenced one and relying on precedent-free votes to identify suitable reference points, I inject error into the analysis since my models do not perfectly predict votes, either precedent-free or precedent-bound. In the precedent-free versions, on which the probability estimates of test two rely, votes correctly classified range from 69.50 percent for family cases to 79.72 percent for intermediate appellate court rulings, with proportional reductions in error from 35.06 to 58.21 percent. The inconsistent votes in Tables 5.1 and 5.2 result from this model imperfection and thus measure error in the analysis. Table 5.2's inconsistent-vote percentages are relatively stable across subpopulations, disclosing fairly uniform and modest error rates.

            Moreover, the research method here is not subject to the kinds of criticism lodged by Baum (1994: 4)  because I use fully integrated models of judicial behavior, investigating much more than judicial attitude. Indeed, mine is not a test of the attitudinal model versus the legal model, but rather of all measurable nonprecedential stimuli shaping judicial behavior versus the power of precedent.

            Another problem arises from test two’s focus only on ideologues: No center-vote analysis analogous to Brenner and Stier (1996) is possible because I postulate center votes are fully acquiescent to stare decisis.

            A worry apparently endemic to precedential-impact research is small usable-vote pools. Brisbin (1996: 1006) noted the issue for Segal and Spaeth (1996). My data provide a maximum of 86 votes of precedential atheists for test two (Table 5.2, “All Decisions,” total of “Suppressed” and “Not Suppressed”), 6.0 percent of the total in the study. Yet my usable votes are 19.8 percent of the precedent-bound ones – three times Spaeth and Segal’s estimated 6.5 percent. Their confining research design, then, may have missed up to two-thirds of the evidence relevant to a comprehensive evaluation of justices’ deference to stare decisis.

            Although the results here are highly suggestive, one issue domain alone is not a surrogate for all. I cannot necessarily generalize my findings to every legal topic. However, in light of the highly emotionally charged character of lesbian and gay rights claims in American public policy making, stare decisis’s decisive suppression of conservative votes in this ideologically cloven terrain certainly resembles a crucial case study for precedential impact, much like Spaeth and Segal’s strategy of investigating how frequently dissenting justices later conformed to majority opinions with which they disagreed. That is, if stare decisis arrested attitudinal and other important effects in the lesbian and gay rights policy domain, then it should everywhere else as well.

Conclusion

            In 1980, a legal scholar announced the demise of collegial precedent in constitutional law (Maltz 1980). Almost twenty years later, two eminent political scientists published a similar obituary (Spaeth and Segal 1999). Yet rumors of stare decisis’s death have been greatly exaggerated. Fully integrated models of appellate court treatment of lesbian and gay rights claims reveal that stare decisis is statistically significant with consequential impact statistics. In particular, precedent is alive and well in intermediate appellate courts, was pivotal in courts of last resort over a quarter of the time, and suppressed potentially conservative votes at a much higher rate than liberal ones. Further, vertical stare decisis swayed judicial decision making in state courts more consistently than horizontal stare decisis.