Standing the Test of Time: The Breadth of Majority Coalitions and the fate of U.S. Supreme Court Precedents

Abstract

Should a strategic Justice assemble a broader coalition for the majority opinion than is necessary, even if that means accommodating changes that move the opinion away from the author’s ideal holding? If the author’s objective is to durably move the law to his or her ideal holding, the conventional answer is no, because there is a cost and no corresponding benefit. We consider whether attracting a broad majority coalition can placate future courts. Controlling for the size of the coalition, we find that cases with ideologically narrow coalitions are more likely to be treated negatively by later courts. Specifically, adding either ideological breadth or a new member to the majority coalition results in an opinion that is less likely to be overruled, criticized, or questioned by a later court. Our findings contradict the conventional wisdom regarding the coalition-building strategy of a rational and strategic opinion author, establishing that the author has an incentive to go beyond the four most ideologically proximate Justices in building a majority coalition. And because of later interpreters’ negative reactions to narrow coalitions, the law ends up being less ideological than the Justices themselves.

1. INTRODUCTION

Does a strategic author of a U.S. Supreme Court majority opinion have a clear incentive to push for an opinion as close as possible to his or her vision of the law? Consider a Justice who, consistent with past findings, has a primary goal of instantiating his or her vision of the best understanding of the law (Epstein & Knight 1998; Rohde 1972a, b; Murphy 1964). This Justice has been assigned the majority opinion in a case and the four colleagues most ideologically similar to the Justice have joined the draft majority opinion. A few other colleagues have written memoranda, each stating an unwillingness to join the draft opinion as written because he or she disagrees with some aspect of it, but a willingness to join if said aspect were changed. Accepting those changes would move the opinion further from the author’s ideal holding (i.e., from the holding that he or she thinks the Court should adopt).² Should the author implement some or all of those changes and thereby gain the others’ adherence to the opinion?

One possible answer is no: once a majority has joined an opinion, the strategic author should stop there. And in choosing which colleagues to accommodate, the strategic author should select those who will move the opinion the least distance from his or her understanding of the law. The Justice who accommodates colleagues more than necessary for a majority is bearing a cost (moving away from his or her ideal holding) with no corresponding benefit. This is the conventional wisdom about how strategic Justices should act, and there is empirical evidence that it is how many of them do act (Maltzman, Spriggs, & Wahlbeck 2000; Rohde & Spaeth 1976; Riker 1962). Maltzman, Spriggs, & Wahlbeck (2000, 95) provide anecdotal evidence of this behavior, quoting Thurgood Marshall in a conference memorandum: “Since seven of us agree, my current plan is not to make the change suggested in the Chief’s ultimatum.” Also, Maltzman, Spriggs, & Wahlbeck (2000, 121) find that “[o]nce at least a majority of justices have joined the majority opinion, authors are less likely to accommodate responsively. Before the author receives majority support for an opinion, he or she has a 90.1% chance of circulating an additional draft. After receiving enough support to set binding precedent, the author is only 68.6% likely to continue his or her coalition-building efforts.” Justice Rehnquist (1987) noted that “[t]he willingness to accommodate on the part of the author of the opinion is often directly proportional to the number of votes supporting the majority result at conference; if there were only five justices at conference voting to affirm the decision of the lower court, and one of those five wishes significant changes to be made in the draft, the opinion writer is under considerable pressure to work out something that will satisfy the critic, in order to obtain the five votes for the opinion.” More pithily, Rehnquist stated that once he had a majority for an opinion, he was not inclined to accommodate other colleagues in order to have a bigger majority because that would be “leaving cash on the table” (Friedman 2011).

That said, Justices sometimes accommodate suggested changes even when they already have a majority, and some Justices are more likely to accommodate in this situation than others are (Maltzman, Spriggs, & Wahlbeck 2000).³ What should we make of Justices who go beyond a bare majority in building the opinion coalition? Are they simply failing to act strategically, or do they place less value on moving the law toward their ideal holding? We posit that there is a strategic basis underlying coalition building beyond a bare majority. Not only do Justices prefer to render an opinion that reflects their preferences; they want their rulings to have staying power. Thus, an opinion author will prefer a ruling that is some distance away from his or her ideal holding if moving this distance increases the chances that precedent established by the opinion will survive the test of time and scrutiny.

In a careful statistical analysis covering all cases decided in the 1953–2001 terms, we show that majority opinion coalitions composed of many and ideologically diverse Justices establish precedents that are less likely to be overruled or negatively cited in the future. Our research design does not permit us to distinguish whether subsequent courts are taking a cue from the ideological breadth/size of the majority coalition or are less critical of the substantive content of an opinion that was constructed to attract a broad majority. The key point is that either adding more ideological breadth with a given number of Justices or adding more Justices leads to an opinion that is likely to receive less negative treatment. This produces a clear incentive for the majority opinion author to add ideological breadth and, separately, to add members to the majority coalition in order to reduce the likelihood of negative treatment of the opinion in the future.

The issuance of Supreme Court opinions is only the beginning of a process: the Supreme Court writes the opinions, but subsequent courts apply them. These later actors’ decisions help to determine the influence and staying power of Supreme Court decisions.⁴ And the data in this article indicate that later citers respond better to decisions with broader coalitions. Opinions with narrower coalitions encounter greater resistance. Opinion authors who fail to accommodate their differently minded colleagues are likely to have a smaller influence on the decisions of future courts.

Our findings offer two related insights into the behavior of Justices on the U.S. Supreme Court. First, we identify a clear strategic incentive for Justices to pursue an ideologically broad opinion coalition. This provides an explanation for a behavior—authors accommodating colleagues beyond a simple majority—that previously appeared nonstrategic and difficult to explain with reference to Justices in pursuit of opinions close to their ideal holdings. Second, our findings suggest moderation in equating ideologically contentious holdings with precedents that survive the scrutiny of future courts. Though major cases may be decided with ideological 5–4 splits, those opinions that garner support from narrow coalitions are much more likely to be overruled or negatively cited by future courts.

Our assumption that Justices prefer for their precedents to be influential is well established in the literature (Wahlbeck 1998, 614). In Section 2 of this article, we discuss how this preference underlies our hypothesis that opinions issued by broad coalitions will experience, on average, a lower incidence of negative treatment by subsequent courts. In Section 3, we present an empirical test of that hypothesis. In Section 4, we discuss the significance of our findings. We conclude in Section 5.

2. BACKGROUND AND THEORY

After oral argument in a case, the members of the Supreme Court have a conference at which each Justice votes and summarizes his or her reasoning. In some cases, the Justices who voted in the majority at their conference are in sufficient agreement among themselves, and in sufficient disagreement with those who voted in the minority at conference, that after a draft majority opinion is circulated, there is little potential for movement: those in the conference majority agree to the draft majority opinion with few changes and those in the conference minority do not circulate memoranda requesting changes. But many cases are considerably more fluid. The Justice assigned a majority opinion might circulate a draft among his or her colleagues and receive memoranda in response indicating that Justice A joins without changes, Justice B would join with modest changes, Justice C would join with different changes, etc. Maltzman, Spriggs, & Wahlbeck (2000, 61) show that “Justice Brennan’s circulation records indicate that in 535 of the cases decided during the Burger Court (23.4 percent), a justice tried to change the majority opinion by sending a note to the majority opinion author recommending a specific change in the opinion.” Additionally, Maltzman, Spriggs, & Wahlbeck (2000) break down Justices’ responses to a majority opinion among the categories of Join Majority, Wait, Suggestion, Threat, Will Write, Circulate/Join Concur, and Circulate/Join Dissent, and find that in 24.1 percent of cases a Justice made a suggestion or threat seeking a change in the majority opinion. The result of these various memoranda will often be a tradeoff for the drafting Justice between the closeness of the majority opinion to his or her draft opinion and the size and composition of the majority (Maltzman, Spriggs, & Wahlbeck 2000; Maltzman & Wahlbeck 1996; Scalia 1994; Brenner 1980).

Justices frequently face such choices, and for a Justice who seeks to pull the law toward his or her ideal holding, the obvious strategy might seem to be to amass a bare majority that is closest to his or her desired position and go no further. Having more than five Justices forces the opinion writer to move further from his or her ideal holding. As to where Justices seek to pull opinions, a debate has long raged among scholars. Law professors have tended to attribute the major role to legal considerations such as text, canons of construction, stare decisis, etc., with the political preferences of the Justices having a smaller influence. Some political scientists have argued that this is wishful thinking and have stressed the role of political preferences. This perspective is embodied in the attitudinal model, which characterizes Justices as solely pursuing their ideological objectives (George 1998; Segal & Spaeth 1993, 2002; Rohde & Spaeth 1976). Empirical studies have found support for the attitudinal model (Hagle & Spaeth 1992, 1993; Segal & Cover 1989). And there has been considerable cohesion among Justices across a wide range of issues that have clear ideological valence and do not obviously cohere as a jurisprudential matter—e.g., the Establishment Clause, affirmative action, abortion, and, perhaps most (in)famously, the constitutionality of the counting methods authorized by the Florida Supreme Court in the 2000 presidential election.

This establishes an empirical puzzle. If the ideological content of the opinion is all that matters, then why do majority opinion authors often work to build support beyond a simple majority? We see an explanation of this apparent paradox in the role of opinions in establishing precedent.

The opinions issued by the U.S. Supreme Court contribute to the body of precedents that subsequent courts draw upon. Studies have shown that subsequent courts, particularly lower courts, are influenced by Supreme Court precedents to varying degrees and do not respond equally to every opinion issued (Benesh & Reddick 2002; Spriggs & Hansford 2001; Songer, Segal, & Cameron 1994; Canon 1973). Studies have also found that in terms of the ideological composition of the majority coalition, the Supreme Court and lower courts respond negatively to opinions issued by majority coalitions to which they are ideologically opposed. Past opinions issued by majority coalitions that diverge ideologically from the current court fare worse in the citing behavior of the current court. In a study of all orally argued cases decided during the 1946–1995 terms, Spriggs & Hansford (2000) find that the likelihood a precedent will be overruled is inversely related to the ideological proximity of the median Justice in the majority coalition that established the precedent to the median Justice on the current court. In other words, the Supreme Court is more likely to overrule precedents established by coalitions of Justices that are ideologically distant from the current Court’s center. Additionally, Songer, Segal, & Cameron (1994) find that in search and seizure cases, appellate panels that are, on average, ideologically opposed to the Supreme Court’s ruling are less likely to follow precedent. Hence, in terms of establishing durable law, the opinion author needs to consider future ideological opposition as well as contemporaneous conflicts with his or her colleagues.

The influence of a ruling is short-lived if it is quickly overruled or regularly rejected as inapplicable. If an opinion with a broader coalition is less likely to be negatively treated by later courts, then making the concessions that attract a broader majority coalition can be seen as a tradeoff between a large, short-term shift in Court doctrine and a longer-lived, more moderate influence on the law. In resolving this tradeoff, a strategic opinion author may seek to make concessions and broaden the opinion coalition in order to increase the influence of the opinion. This raises the question of whether opinions established by broad coalitions are more likely to stand the test of time than those arising from narrow coalitions, and we now turn to that question.

3. EMPIRICAL ANALYSIS

Central to our claim that opinion authors have a strategic, forward-looking motive to establish an ideologically broad coalition is the proposition that breadth improves an opinion’s treatment in subsequent cases. Indeed, the absence of such a relationship between breadth and treatment would disconfirm our hypothesis that this strategic incentive exists. We use data from Shepard’s Citations to study the subsequent treatment of opinions issued by the Supreme Court (Spriggs & Hansford 2000). The analyses of Corley (2009), Spriggs & Hansford (2001), and Benesh & Reddick (2002) serve as the basis for our empirical models. We extend these specifications to account for the ideological breadth, in terms of Martin–Quinn scores, of the majority opinion coalition (Jacobi & Sag 2009; Martin, Quinn, & Epstein 2005; Martin & Quinn 2002). ⁵

Following Corley (2009), we use the coding in the Shepard’s database to measure the subsequent treatment of Supreme Court majority opinions. Shepard’s tracks (and has tracked for decades) the subsequent history of judicial opinions. This history includes every reference to an opinion in a later opinion, law review article, brief, treatise, statute, etc. Shepard’s also examines every published state and federal court opinion to see which earlier opinions it cites. Shepard’s analyzes the citations contained in the main opinion and any concurrences.⁶ If a given citation refers to a case but has no substantive reaction to it, Shepard’s does not put the citation into a substantive category.⁷Shepard’s divides all the remaining citations (that is, all citations that contain substantive responses to the cited case) into the following main categories: Overruled, Criticized, Questioned, Limited, Distinguished, Explained, Harmonized, Paralleled, or Followed (Spriggs & Hansford 2000).⁸Shepard’s characterizes the first four of these categories as “strong” negative treatment (with Overruling of course being the strongest), and if a cited opinion has been subject to such negative treatment, Shepard’s lists a “warning” with that cited opinion. Shepard’s treats Distinguished as mildly negative, Paralleled and Followed as positive, and Explained or Harmonized as neutral (Spriggs & Hansford 2000).

Spriggs & Hansford (2000) undertook a careful study to measure the reliability of Shepard’s Citations. They took a random stratified sample of Supreme Court cases citing earlier Supreme Court cases, yielding 602 citing opinions, and they coded all the citing opinions according to the coding rules in the Shepard’s training manual.⁹ They found high levels of agreement between their coding and Shepard’s coding. Notably for our purposes, the negative categories of Overruled, Limited, Questioned, and Criticized were particularly reliable, with agreement between their and Shepard’s coding of 99.5 percent, 99.0 percent, 96.4 percent, and 94.5 percent, respectively (Spriggs & Hansford 2000). The mildly negative category Distinguished had 87.9 percent agreement. The positive and neutral categories were less reliable, although they still had fairly high levels of agreement, ranging from 79.1 percent to 94.7 percent (Spriggs & Hansford 2000).¹⁰

Note that the negative categories represent a spectrum of responses. Only the Supreme Court can overrule one of its own cases, so Overruled represents the Supreme Court’s rejection of its own work.¹¹ Criticized, Questioned, and Limited, in contrast, include negative treatments from all courts. And, as noted above, these are not bare citations; in each case, the citing court indicates disapproval. Specifically, Shepard’s Citations (2008) defines Criticized as “disagree[ing] with the reasoning/result” of a case, Questioned as “question[ing] the continuing validity or precedential value” of a case, and Limited as “restrict[ing] the application of the case you are Shepardizing.”

On positive versus negative citations, not all of the findings in this article apply to the central positive treatment Followed (the only other positive treatment, Paralleled, is too rare to yield reliable results). We do not focus on Followed for two reasons. First, overruling, criticizing, questioning, or limiting a case is a stronger signal than is following a case. That is, choosing to overrule, criticize, question, or limit a case is a more negative reaction than following a case is a positive one. Second, Spriggs & Hansford found that Followed was less reliable than the negative signals (81.9 percent agreement between Spriggs & Hansford’s blind recoding and Shepard’s).

To test our hypothesis that opinions issued by broad opinion coalitions will fare better in subsequent cases, we estimate five regression models predicting negative treatment according to the Shepard’s classifications. We assemble a dataset of all cases decided during the 1953–2008 terms. In four of the models, the dependent variable is the count of the respective citation instances that occurred before the end of the 2008 term for the citation categories Criticized, Questioned, Limited, and Distinguished. Overruled, a binary indicator of whether an opinion was overruled by the end of the 2008 term, is the dependent variable in the fifth model. Note that there is a single observation for each case in the dataset. We aggregate over citing cases, giving the number or indicator of the respective citation type up to the 2008 term. It may, at first glance, appear that we should standardize our dependent variables by the number of times a case has been cited. Many cases are never cited, however, which leads to a zero denominator in this measure. Our central theoretical results are unchanged if we include in the regressions the number of total citations, but due to concerns about endogeneity, we elect to report the models without total citations on the right-hand side.

Our dependent variables are summarized in Table 1.¹² We see that the strong negative treatments occur with considerable rarity relative to the moderately negative Distinguished. Most of the opinions covered by our dataset are never cited in a strong negative manner. Overruling is particularly rare. Approximately 3 percent of Supreme Court cases are overruled. Figure 1 disaggregates the incidence of negative treatment by the term the opinion was written. If negative citation were simply a constant risk to any opinion, then these plots would depict a strictly decreasing likelihood of negative treatment as decisions become more recent. As more time passes, given a constant risk of negative citation at any given point in time, the likelihood an opinion has been negatively cited should simply increase with the age of the opinion. This is not the case. We see in Figure 1 that there is much interesting variation in the risk of negative treatment. Indeed, cases decided in the late 1980s and early 1990s are at least as likely to have received at least one criticizing and/or questioning citation as cases decided in the 1960s. We now turn to the specification of models for these outcomes.

Our primary independent variable of interest is the breadth of the Ideological Interval constituted by the majority opinion coalition. To identify this variable we first construct the group of Justices that authored/signed the majority/plurality opinion in a case using the Opinion and Vote variables in the Supreme Court Database (SCDB, http://scdb.wustl.edu). We then identify the Ideological Interval of a case to be the difference between the maximum and minimum Martin–Quinn scores of the Justices in the majority opinion coalition, using the scores for the term in which the case was decided. We expect that this variable will have a negative effect on all of the dependent variables—the broader the opinion coalition, the lower the incidence of negative citation.¹³ We also account for a number of potential confounding covariates, following the specifications in Spriggs & Hansford (2001), Benesh & Reddick (2002), and Corley (2009). We detail these variables below.

We also include the variable Majority Size, the number of Justices in the majority opinion coalition. Both Spriggs & Hansford (2001) and Benesh & Reddick (2002) find that opinions issued by unanimous courts are treated better by subsequent courts. Also, Corley (2009) finds that plurality opinions are less likely to be complied with by lower courts. We use a more informative measure in the count because of the obvious potential pitfall of collinearity with our primary independent variable of interest. Larger opinion coalitions will tend to have broader ideological intervals. Indeed, the Pearson’s correlation coefficient between Ideological Interval and Majority Size is 0.49. If we omit or incompletely account for the effect of the size of the coalition, we may inaccurately attribute the effect of coalition size to the ideological breadth of the coalition. This omission would be particularly problematic since it is probably the case that Justices are more likely to join opinions that are, based on the quality of legal reasoning or other factors, less likely to be negatively treated. A model that does not account for the size of the majority runs the risk of attributing this strategic anticipation to the effect of Ideological Interval. We expect Majority Size to have a negative effect on the incidence of negative citation.¹⁴

As we noted above, Spriggs & Hansford (2001) find that the Supreme Court is more likely to overrule opinions issued by majority coalitions with ideological medians far from the Court’s current ideological median. To account for this we construct the variable Ideological Divergence. We first identify the median Martin–Quinn score among the Justices in the voting majority on each case. Ideological Divergence is then measured as the average absolute difference between the median on the case and the median on the Court over the terms between the decision and 2008. So, if the median Justice in the voting majority in case A, decided in the 2006 term, has a Martin–Quinn score of 1, and the median Martin–Quinn score on the Court in terms 2006, 2007, and 2008 are −2, −1, and 2, respectively, then Ideological Divergence for case A is the average of {|1 − (−2)|, |1 − (−1)|, |1 − 2|} = 2. We expect that this variable will increase the incidence of negative citation.

We also consider the effect of the salience of a case. It is possible that more salient cases are considered more often by subsequent courts and therefore cited negatively more often. Following Collins (2004), we use the number of amicus curiae briefs filed in a case to measure the salience of a case. We expect this variable, Amicus Briefs, to have a positive influence on the frequency of negative treatment.¹⁵ This is an important effect to control for given the possibility that organizations and individuals file briefs in cases anticipated to be controversial (in particular, those likely to produce ideological 5–4 splits). Also, this will control for the possibility that ideologically broad and large coalitions arise in noncontroversial cases. If the Amicus Briefs variable were omitted, this could confound our findings with regard to the size and breadth of the majority.

Spriggs & Hansford (2001) find that cases addressing statutory issues, as opposed to questions of constitutionality, are less likely to be overruled. We include Statutory, an indicator that assumes the value 1 if the case addressed a statutory issue according to the lawType variable in SCDB. It is expected that this variable will have a negative effect on the incidence of negative treatment.

Benesh & Reddick (2002) find that lower courts are significantly more likely to adhere to the Supreme Court’s reversal of a precedent if the members of the Court that established the original precedent are no longer on the Court. This tendency to adhere to a reversal of a precedent can be interpreted as a negative treatment of that precedent by the lower court. We reason that the incidence of negative treatment will be higher for opinions existing through many terms where the members of the opinion’s majority coalition are no longer on the Court. We include the variable Justices Remaining, which is the mean proportion of the original majority opinion coalition that remains on the Court in the terms following the decision, up to the 2008 term. In order to construct this measure for each case, we first construct a vector that records the proportion of the justices in the majority opinion coalition who remain on the Court, for each term from the one after the opinion was issued until 2008. We then take the arithmetic mean of this vector for each case. We expect that this measure will exercise a negative influence on the incidence of negative citation.

Both Spriggs & Hansford (2001) and Benesh & Reddick (2002) find that more complex opinions, in terms of the number of legal issues and provisions they address, are more likely to be negatively treated by subsequent courts. We use the number of legal issues and/or provisions addressed in an opinion to constitute our measure of Complexity. This measure is equivalent to the number of rows occupied by a case in the “Legal Provisions” version of the case-organized SCDB.

We also include a measure of uncertainty on the Court regarding the legal justification of its ruling. Spriggs & Hansford (2001) find that the risk of overruling increases with the number of concurring opinions authored in the case. It is particularly important that we account for this effect, because the presence of acceptable alternative legal arguments could cause both the breadth of the majority coalition and the likelihood of future negative treatment, thus confounding the primary relationship of interest in our study. We use the Opinion and Vote variables in SCDB to create the variable Concurrences—the number of concurring opinions authored in each case. It is expected that the incidence of negative treatment will increase with the number of concurring opinions authored.

It is reasonable to expect that a case will accumulate more negative citations the longer it is exposed to scrutiny. To account for exposure we include a quadratic in the number of terms since the case was decided (until 2008). This function is represented by the Terms (number of terms) and Terms² (number of terms squared) variables in the models. It is expected that the number of negative citations will increase with time exposed.¹⁶

We carefully select the specific estimators to conform closely to the properties of our dependent variable. When modeling count data, it is important to account for the facts that (i) the dependent variable can only assume nonnegative integer values and (ii) the mean and variance may not be equal (i.e., there may be under- or overdispersion), as is the case with the Poisson distribution (King 1989). For the count variables, we use the generalized Poisson distribution proposed by Consul (1989), which properly accounts for these two properties.¹⁷ For Overruled, we use a regression model designed for binary rare events, since the incidence in our sample is only 3 percent. Similar to logit, the “cloglog” model, a generalized linear model using a complementary log-log link function, is more appropriate for skewed binary data (i.e., rare events) (Buckley & Westerland 2004). We use the cloglog model for Overruled.¹⁸

The regression results are given in Table 2.¹⁹ The estimates are supportive of our central hypothesis. In four of the five models, the effect of the breadth of the Ideological Interval is negative, and the effect is statistically significant at the 0.05 level (one-tailed) in three of the models. The effect is negative but not significant in the model for Distinguished. In the Limited model, the effect is positive, but not significant. Following the findings of Ward, Greenhill, & Bakke (2010) that p-values alone should not serve as a basis for concluding that a variable contributes to the fit of a statistical model, we use the Akaike Information Criterion (AIC) (Akaike 1974) to identify the contribution of Ideological Interval to the fit of each model. We do this by estimating each model with and without Ideological Interval and record the Change in AIC as the difference in AIC that occurs from removing Ideological Interval (i.e., AIC (without Ideological Interval) − AIC (with Ideological Interval)). This measure is computed such that a positive value of Change in AIC indicates model improvement from adding Ideological Interval. Inferences based on p-values and model fit align in our analysis. In the models for Overruled, Criticized, and Questioned, the AIC is lower when Ideological Interval is in the model. The values of Change in AIC for Limited and Distinguished indicate that Ideological Interval should not be in the model. Overall, our hypothesis that Ideological Interval reduces the incidence of negative treatment of a precedent is supported by the analysis.

Evaluations of statistical significance and improvements in model fit do not convey the substantive significance of the effect of ideological breadth. Figure 2 presents the expected values of the negative citation variables, conditioned on the value of Ideological Interval given majority sizes of five, six, and seven Justices, for the models in which Ideological Interval has a significant negative effect. Numerically, these effects may seem moderate, but they are substantively large. Because all forms of negative treatment constitute relatively rare events, the difference between an ideologically narrow and ideologically broad opinion coalition may well determine whether an opinion is ever treated negatively, and it would be alarming to see a particularly large magnitude of the effects of ideological breadth since large magnitudes of negative treatment are particularly uncommon.²⁰ In Figure 2a, the most damaging form of negative citation, overruling, is considerably affected by Ideological Interval. Regardless of the size of the majority coalition, the probability of an overruling citation, albeit low in all instances, is 50–70 percent lower when the majority coalition is ideologically broad than when the coalition is ideologically narrow. In Figure 2b, we see that moving from the minimum Ideological Interval (<1) to the maximum (approximately 10) reduces the expected number of criticizing citations by approximately 0.1–0.2. In terms of questioning citations (Figure 2c), we see that given a majority coalition with six Justices, an opinion issued by a very narrow majority coalition is expected to be questioned approximately 30 percent more than the average opinion, whereas cases with a very broad coalition are expected to be questioned approximately 15 percent less than the average opinion. Additionally, it is informative to note that the effect of coalition breadth is on par with that of coalition size. Specifically, the change in negative treatment when moving from line to line (i.e., the effect of coalition size) is comparable to that when moving along any given line (i.e., the effect of coalition breadth). This reinforces the notion that majority opinion authors can prevent negative treatment of opinions by attracting an ideologically broad coalition of Justices.

In terms of the control variables, many of the effects align with our expectations. Most notably, Majority Size, the number of Justices in the majority coalition, has a negative and significant effect on the incidence of negative citation in each model. Concurrences, the number of concurring opinions authored in each case, increases the incidence of negative citation in each model and is statistically significant in every model but the one for Overruled. Also, more complex cases, as measured by the number of legal provisions and issues addressed by an opinion, experience a higher incidence of negative citation. We basically have null findings with regard to the ideological divergence between the median of the opinion coalition and future courts. The effect is positive, as expected, in every model, but only significant in the model for Distinguished. Also, in terms of the salience of cases measured by Amicus Briefs, we find that the more salient a case, the greater the instance of negative treatment in each model—an effect that is significant in three of the five models. We have mixed results with regard to Statutory cases and Justices Remaining, with some nonsignificant, positive-significant, and negative-significant findings for these two variables.

4. DISCUSSION

The data indicate that cases with narrow Supreme Court coalitions are more likely to be treated negatively by later courts. This is true on two different metrics of narrowness: the ideological breadth represented by the Justices who join the majority opinion and the number of Justices who join the majority opinion. Adding either ideological breadth or a new member to the majority coalition results in an opinion that is less likely to be overruled, criticized, or questioned.

Our data include a single observation for each cited case, aggregating over potential citing opinions. It would be useful to evaluate our hypotheses in the dyadic setting, such as that in Westerland et al. (2010), testing whether individual citing cases are more likely to negatively cite opinions established by an ideologically narrow coalition. However, in the current analysis, we do much to account for the aggregated effects of potential citing opinions. First, through the quadratic function of the number of terms an opinion has existed, we account for increased exposure over time. Second, through our measure of Ideological Divergence, we account for variance in the overall ideological hostility to which an opinion is subject over time. Third, through our measure of Justices Remaining, we account for the potential effects of turnover on the likelihood of citing cases negatively treating the opinions. In sum, though the dyadic analysis would be a worthy pursuit for future research, there is no reason to believe that our results are confounded by aggregation over the citing cases.

If a narrower coalition means a greater likelihood of an opinion being treated negatively by later courts, the incentives facing opinion writers become more complex. Accommodating an ideologically distant colleague may entail a significant move from the author’s ideal holding. That is, ordinarily we would expect that accommodating an ideologically distant colleague would entail a greater change in the draft opinion than accommodating an ideologically close colleague would.²¹ Similarly, accommodating a potentially dissenting or nonjoining concurring Justice who is not needed for purposes of forming a majority likely means moving the opinion away from the drafting Justice’s ideal position. But the data show that bringing in potential dissenters, potential nonjoining concurrers, or ideologically distant colleagues is associated with a significant benefit—fewer negative treatments, meaning the long-term influence of a precedent. Note that we do not present any evidence that majority opinion authors choose to compose broadly appealing opinions in order to avoid future negative treatment. Rather, we demonstrate that opinion authors have an incentive to attract a broad ideological coalition in order to avoid future negative treatment of the opinion.

We note a possible pitfall in drawing inferences about the direction of causality with regard to the number of Justices joining the majority. This does not necessarily mean that reducing the number of dissenters or nonjoining concurrers in a given case will reduce the negative treatments of that case. It could be that Supreme Court Justices are good judges of quality: more Justices dissent, or concur without joining the majority opinion, when they think the quality of the majority opinion is low, and they are correct in their judgments.

This reasoning does not explain the negative correlation between ideological breadth and negative treatments.²² In addition, one of its assumptions fails because adding potential dissenters/nonjoining concurrers involves a change in the majority opinion (i.e., the change comes before the joining). The relevant universe is Justices who are potential dissenters/nonjoining concurrers. By hypothesis, we are talking about a situation in which a Justice has circulated a draft opinion and has been able to garner the support of enough Justices (with whatever changes they may have requested as a condition of their joining) to form a majority of the Court. The question involves any Justices who have not joined the draft opinion and thus are potential dissenters/nonjoining concurrers.²³ One or more Justices may be convinced that the case should be affirmed rather than reversed (or vice versa) and thus unwilling to join any conceivable opinion that the majority would also adopt. But there may be potential dissenters/nonjoining concurrers who would be willing to join the majority opinion under the right circumstances. Such circumstances are almost always (if not always) specified changes to the majority opinion: Justices offer to join the draft opinion if changes are made to it (Hammond, Bonneau, & Sheehan 2005; Maltzman, Spriggs, & Wahlbeck 2000).²⁴ That is, just as some who already joined the majority may have done so because the drafter accommodated their concerns, so, too, may the drafter accommodate the concerns of those who are not necessary to constitute a majority. Maltzman & Wahlbeck (1996, 583) give one example: in Wardius v. Oregon, 412 U.S. 470 (1973), Justice Rehnquist voted at Conference to affirm and was in the minority. After Justice Marshall circulated a draft majority, Justice Rehnquist stated: “I voted to affirm in this case at Conference, but before writing a dissent to Thurgood’s proposed opinion I think I will wait to see if anything narrower… is written.”²⁵ Marshall circulated a new draft opinion that was narrower than the previous version (Marshall cut a discussion addressing the state’s exclusionary rule). As Maltzman, Spriggs, & Wahlbeck (2000, 99) note, “Marshall’s decision to limit the breadth of his opinion was rewarded when Rehnquist joined his opinion to reverse. … Rehnquist was willing to abandon his preferred outcome in order to ensure that the majority opinion was narrowly drawn.”

It may also be that later courts use the number of dissenters/nonjoining concurrers (and the identity of those in the majority) as proxies for value and quality. In such a situation, we would expect that a simple vote switch from dissenter/nonjoining concurrer to the majority opinion would be correlated with a decrease in negative treatments. From the standpoint of the majority drafter, it does not matter much whether the changes requested by a potential dissenter/nonjoining concurrer are making the majority less likely to receive negative treatments (i.e., the changes improve the quality of the opinion in the eyes of citing courts) or the mere addition of another person to the majority is doing all the work (i.e., changes requested by the potential dissenter/nonjoining concurrer do not improve the quality of the opinion in the eyes of citing courts, or perhaps even decrease it, but the addition of a member to the majority is such a strongly positive signal that citing courts are less likely to treat the opinion negatively on that basis alone). The reason it does not matter to the majority writer is that potential dissenters/nonjoining concurrers do ask for changes as a price for their joining. The two are a package. So whether it is the substance of the changes or the addition of a Justice does not matter, because the two cannot be separated.

Now we turn to the negative correlation between ideological breadth and strongly negative citations. Recall that this correlation is separate from that involving the number of members of the majority coalition: given majority coalitions of the same size, ideological narrowness is correlated with more negative treatments. In terms of the choices facing the author of a majority opinion, the question is whether, in choosing whom to accommodate and bring into the majority coalition, he or she chooses a Justice who is ideologically more similar. Let’s say, for example, a moderately conservative Justice has circulated a majority opinion that has garnered four votes (including his or her own), and he or she has received a memo from a conservative Justice that would pull the opinion in one direction and a memo from a liberal Justice that would pull it in the opposite direction. The data presented in this article indicate that there is a benefit to accommodating the more ideologically distant colleague (here, the liberal Justice): the resulting opinion will receive fewer negative citations. This finding contradicts the strategic conception put forth by Maltzman, Spriggs, & Wahlbeck (2000) that opinion authors should strive to achieve ideologically narrow, minimum winning coalitions.

Our findings indicate that a majority opinion author who compromises on content to attract ideologically diverse signatories will be rewarded with greater staying power of the opinion. As we noted above, accommodating an ideologically distant colleague is likely to entail a bigger shift from the author’s preferences than accommodating an ideologically close colleague would. Accommodating an ideologically distant colleague may produce opinions that are in some sense better (e.g., better reasoned) or simply attractive to a broader range of later courts (because the opinion had to have broad appeal in order to gain the votes of an ideologically diverse coalition on the Court). Or it may be that later citers are not responding to the substance of the opinion, and instead rely entirely on the ideological breadth of the majority coalition. That is, as with the size of the coalition, later courts may be responding to substantive changes involved in the author’s accommodating an ideologically distant Justice. Also, it is possible that the ideological breadth of the opinion coalition sends a cue of broad acceptability to later citing courts. Later citers may view a broad coalition as an indicator that the content of an opinion is not ideologically offensive to either side of the spectrum. Whatever the internal process through which later citers favor opinions with broader ideological coalitions, our results show that Supreme Court majority opinion authors can render their contributions more resistant to future criticism by attracting the support of ideologically distant colleagues.

As with the number of dissenters, from the standpoint of the author of the majority opinion, later citers’ internal motivations are of little consequence: the question is which Justice(s) will join the author’s majority coalition, and the price of such joining is accommodating requested changes. It makes no difference to the author whether the substance of the changes or the mere fact of having an ideologically distant Justice as part of the coalition is doing the work, because the two go together.

All these findings point in the same direction: opinion writers who can assemble broad, ideologically diverse coalitions are less likely to have opinions that are treated poorly. These findings thus present in a new and different light the choices facing an opinion writer. Some Justices assigned a majority opinion stop accommodating their colleagues once they get to five and others accommodate even when they have more than five. It turns out that this choice appears to have significant implications for the impact of the relevant opinion. That is, the data presented here yield useful information related to questions about the long-term influence of opinions, and thus the tradeoffs facing authors of majority opinions, and the actual influence of ideology on precedent.

5. CONCLUSION

Many scholars have focused on Justices’ votes in cases and have largely ignored the later reception of a given opinion. Considering only the consequences in the present, the seemingly obvious practice for a strategic Justice, and a common practice as an empirical matter, is for a Justice to stop accommodating once he or she garners a majority in support of the majority opinion. The assumption is that there is little or no benefit to adding Justices.²⁶ Our data show that the matter is a good deal more complex, because broader coalitions are associated with opinions that have greater staying power. This enriches (and complicates) the considerations for Justices, particularly the Justice writing the majority opinion.

Cases decided along narrow ideological lines are more likely to be overruled, criticized, and questioned by later courts. Later courts treat cases with less ideologically predictable lineups more favorably than those with more ideologically predictable lineups. Thus, to whatever extent ideology influences Justices’ votes, that gives us information about how to predict votes in particular cases but tells us far less about the development of the law. That is, even if ideology explains a significant amount of voting in cases, it seems to explain much less in terms of the resulting shape of the law. Thus ideological voting has less significance than has previously been supposed.

This also means that insofar as Supreme Court Justices act ideologically (whether intentionally or not), their votes are ideological more than the law ends up being ideological—because of the mediating force of later interpreters. Each Justice can control his or her own vote, but the Justices cannot control how their opinions will be received later. And, as it turns out, later interpreters cast a more skeptical eye on opinions that are decided along ideological lines. Once we take into account what happens to opinions after the Court issues them, we see that the Justices are more ideological than the law is. Ideology, in other words, goes only so far.

REFERENCES

Author notes