1 HETEROGENEITY IN AMICI

Experienced advocates demonstrate greater success at the Court (McAtee and McGuire Reference McAtee and McGuire2007; McGuire Reference McGuire1994, Reference McGuire1995, Reference McGuire1998; Szmer, Johnson, and Sarver Reference Szmer, Johnson and Sarver2007). Scholars speculate that experience is a proxy for a latent skill for advocacy, a stronger understanding of the Court's norms, and preferences (and how to satisfy them), or the possession of a strong reputation for reliability and credibility that the justices trust. Whatever the precise causal mechanism, it is obvious that experienced advocates perform considerably better. Corley (Reference Corley2008) shows that the Court's written opinions draw their language more heavily from higher quality/more experienced advocates. These differences show that written advocacy on the behalf of the litigants affects justice behavior. Further, it shows that there exists heterogeneity in the effectiveness of litigant attorneys to influence the justices, motivating the search for similar heterogeneity in the influence of amicus curiae.Footnote ³

Additionally, it is established that the solicitor general's briefs have a potent impact on the Court's decision-making, irrespective of how many other briefs have been filed in a given case (Caldeira and Wright Reference Caldeira and Wright1988; Corley Reference Corley2008; Kearney and Merrill Reference Kearney and Merrill2000; McGuire Reference McGuire1998).Footnote ⁴ Whether it is due to the lofty reputation of the solicitor general's office, respect for the executive branch in a separation of powers system, or simply the vast experience of attorneys advocating on behalf of the federal government (McGuire Reference McGuire1998), these briefs have a powerful impact on the justices’ votes. Taken together, the findings regarding both the party's counsel and the Solicitor General as amicus suggest that not all actors advocating at the Court should be treated as having an equal impact. Given this fact, the logic for treating nonsolicitor general amici as if the impact of any given group's brief is interchangeable with any others’ is unclear, yet this constant effect is implicitly assumed in research that measures the influence of amici as a raw count of the number of briefs supporting the liberal or conservative position in a given case. While all advocates, both for the parties and the amici, seek to influence the Court, not all are equally capable of doing so.

Generally, efforts by actors to exert influence on the decisions of others increase significantly when the persuader is perceived as an expert. In the case of interactions between justices and interests before the Court, expert power effectively influences the decision making if two conditions are met: (1) the justice thinks that the interest group knows the facts, and (2) the justice has confidence that the group is telling the truth, i.e., the group is credible (French and Raven Reference French, Raven and Cartwright1959, 163). Thus, groups that exhibit both higher informational resources and greater credibility should be more likely to persuade the Court and emerge victorious on the merits. Given these conditions, it is clear why the solicitor general's amicus briefs exhibit such a dominant effect on the Court: the solicitor general's office has greater informational resources and enjoys higher credibility with the Court than the average filer. But the solicitor general is unique; the thousands of other filers do not share the privileged institutional role of the solicitor general's office. As such, while other filers must vary in their ability to exert expert influence, conventional metrics thus far incorporated in the literature, i.e., counts, cannot explain variance in groups’ power.

Further, exploring and leveraging heterogeneity in groups sheds light on a fundamental question in judicial politics: How do judges on collegial courts make their choices? That is, this study may also lend some support to one of two contrasting theories of judicial decision making, the legal and attitudinal models (Baum Reference Baum1997; Epstein and Kobylka Reference Epstein and Kobylka1992; Richards and Kritzer Reference Richards and Kritzer2002; Segal and Spaeth Reference Segal and Spaeth1993, Reference Segal and Spaeth2002). Attitudinal and legal models of judicial decision making offer strikingly different predictions about the relative influence of amici (Collins Reference Collins2008). Attitudinal models predict that briefs and other forms of advocacy should have a nonconstant effect on the justices’ behavior (e.g., Segal and Spaeth Reference Segal and Spaeth2002). For instance, motivated reasoning may influence liberal (conservative) justices to support litigants supported by prominent liberal (conservative) groups. Justices might use the presence of these groups as a heuristic to suggest that their preferred position is also preferred by society.

Conversely, theories recognizing that legal factors play a role alongside judicial ideology predict that additional argumentation should lead justices to support positions they otherwise might not (e.g., Epstein and Kobylka Reference Epstein and Kobylka1992). Influence stems from the informational character of legal advocacy like amicus briefs. As Collins (Reference Collins2008, 83) notes: “. . . for the justices to reach a decision they believe to be correct, they must be persuaded as to which outcome is the correct one.” Persuasion is the heart of this legal account of judicial decision making. The justices should favor the advocacy that lays out novel legal arguments, but not extralegal information, like social science data, in the manner most persuasive to the justices. Further, groups that are recognized as experts by the Court are considerably more likely to be taken seriously, and thus influence the justices. In our case, then, observing justices voting contrary to their attitudinal predispositions due to amicus involvement would suggest support for the legal model. Of course, for such a finding to be fully supportive of the legal model, we would have to be sure that the content of persuasive briefs was legal, not extralegal, in character, which is beyond the scope of this study.

What is missing from prior work is the notion that groups differ in their ability to engage in effective, persuasive advocacy. Extending Collins’ (Reference Collins2008) argument, we claim that the influence of amici is not fully captured in the number of briefs supporting one position or the other. Rather, well-connected and powerful groups are considerably more likely to have the experience, name recognition, and resources necessary to be recognized as experts in a given policy area by the Court, and thus exert a greater influence on the final decision on the merits. Prior work posits that one mechanism by which experienced, connected litigants succeed at greater rates than other litigants is enhanced credibility with the justices. Since the stakes in any one case are necessarily lower for a repeat player, the repeat player has an incentive to maintain credibility with the justices by providing accurate, novel information in order to be taken seriously in future cases (Galanter Reference Galanter1974; Lazarus Reference Lazarus2007; McGuire Reference McGuire1995, Reference McGuire1998). Given that amici must compete to even be read by the law clerks, it seems natural that particularly powerful interest groups would be especially likely to influence outcomes, given their credibility. As such, the influential groups in the network of amici coming before the Court are considerably more likely to be associated with the winning party on the merits than isolated, peripheral groups, ceteris paribus. The central actors in the amicus network are groups whom many other groups desire to be affiliated with. We argue that the justices and clerks will also recognize these groups as the pivotal actors and respond accordingly.

Indeed, the act of producing an amicus brief is not an isolated procedure. Typically, groups cosign briefs, indicating their joint advocacy for the policy position of the brief. This sharing of opinion, information, and resources is absent from many contemporary accounts of group involvement at the Court. Hula (Reference Hula1999) argues that interest groups form coalitions to pursue their strategic goals at reduced costs, shape public debate by influencing a broader platform, gather information, and receive symbolic benefits. Groups with common policy interests and shared ideological values can cosign briefs together, sharing resources and signaling to the Court the wide constellation of interests supporting a particular outcome.Footnote ⁵

Amici who are central to the network of groups before the Court over time, establishing many connections between disparate groups, should be considerably more likely to exert expert power under French and Raven's (Reference French, Raven and Cartwright1959) classic definition. Friedkin (Reference Friedkin1993) demonstrates that actors with higher centrality in a given network are significantly more likely to influence the decisions of other actors. Friedkin's findings are in an educational context, far removed from the formal structure of the Supreme Court. However, network theory and prior work strongly suggest that in a network of actors with heterogeneous capabilities, centrality is a key variable associated with greater influence on the decision making of political actors. Thus we test below whether groups that hold more influential positions in the interest group network have a stronger impact on judicial decision making. We expect that groups who are the most central are more likely to influence the Court's decision-making behavior.

2 JUDICIAL DECISION MAKING

There is an important and extensive literature on explaining individual judicial voting behavior (e.g., Rohde and Spaeth Reference Rohde and Spaeth1976; Segal and Spaeth Reference Segal and Spaeth1993, Reference Segal and Spaeth2002; Sunstein et al. Reference Sunstein, Schkade, Ellman and Sawicki2006). We reexamine this classic question in the judicial politics literature, but with a unique focus on the influence of interest groups based on their network positions. The pathways for group influence on justice voting typically follow one of two possibilities: the affected groups hypothesis or the information hypothesis.

The affected groups hypothesis states that amicus briefs are influential because they send a signal to the Court of how many groups and individuals are potentially affected by the decision (Collins Reference Collins2004; Spriggs and Wahlbeck Reference Spriggs and Wahlbeck1997). It is measured by the number of groups that file briefs. Contrarily, the information hypothesis says that amicus curiae briefs are effective because they provide the Court with added information, which is typically measured as the number of briefs filed (Collins Reference Collins2008).

We propose an alternative hypothesis to the study of the effect of amicus briefs on judicial decision making: the influence hypothesis. For an amicus brief to influence a justice, two conditions must be met: first, the brief must attract the attention of the justice and her clerks enough to merit a close reading; second, the brief must contain novel, high-quality information and advocacy which does not replicate the arguments of the litigants’ briefs (Lynch Reference Lynch2004).

The influence hypothesis posits that groups’ statuses should matter to justices. High status groups likely satisfy the conditions of the influence hypothesis for two reasons. First, while law clerks give at least a cursory skimming to most amicus briefs (Lynch Reference Lynch2004, 45), not all briefs are read by all justices (Lynch Reference Lynch2004; Rozen Reference Rozen2011). Groups with high status earn a reputation for presenting high-quality briefs that provide useful new information to the justices, and thus earn a closer initial screening by law clerks (Lynch Reference Lynch2004, 50). Thus, a brief from a high status group is given more attention, all else equal, than a brief from a low status group. Second, amicus briefs are expensive to produce (Caldeira and Wright Reference Caldeira and Wright1988; Lynch Reference Lynch2004), and are only useful to the justices when they provide novel advocacy, social science data, and/or factual information not contained in the litigants’ briefs (Lynch Reference Lynch2004). A high status group, given its greater financial and other resources, can retain the best counsel and support the necessary research needed to produce high quality amicus briefs. Why would groups go to the expense and trouble of producing these briefs, and why would justices welcome these briefs (as evidenced by the fact that no limits have been placed on amici despite the huge growth in briefs), if the final decisions of the Court did not at times incorporate the views of high quality amicus briefs?

Given that clerks report that many briefs are poorly written and are not particularly informative (Lynch Reference Lynch2004), it is natural to posit that groups with lower status may be much more likely to produce low quality briefs which are not read closely or perhaps even considered by the justices. The influence hypothesis therefore suggests that litigants supported by briefs from high status groups should be more likely to prevail on the merits, as these briefs are both more likely to be closely read by the justices, and more likely to contain novel advocacy which impacts justices’ decision making.

We believe that a different signal and interpretation is given to a brief depending on who signed the brief. Contrary to the information and affected hypotheses, all cosigners are not equal.Footnote ⁶ For example, does it matter that the National Wildlife Federation, one of the most powerful groups by a variety of network measures, signed a brief in a case, versus the Beer Institute, one of the least powerful groups? The influence hypothesis implies that we can gain a better understanding of amicus influence by looking at who is signing the brief, not just how many signed.

The motivation for our work is aptly captured by Collins's (Reference Collins2008) review of the literature. He points out that a “growing body of scholarship, predicated on Galanter's (Reference Galanter1974) seminal study of repeat players, indicates that the quality of argumentation and the status of litigants (and perhaps amici) can shape judicial choice (Johnson, Wahlbeck, and Spriggs II Reference Johnson, Wahlbeck and Spriggs II2006; McAtee and McGuire Reference McAtee and McGuire2007; McGuire Reference McGuire1998)” (Collins Reference Collins2008, 183). Collins goes on to posit that “as applied to amicus briefs, this implies that Courts might be particularly attentive to the arguments advanced by highly experienced advocates, giving those briefs favorable attention (Samuels Reference Samuels2004).” We test whether groups that hold more powerful positions in the amicus network have an impact in the various models of judicial decision making that previous scholars have laid out. We build on Collins (Reference Collins2008) insights by bringing evidence to bear with our measure of interest group power that comes from network theory and methodology. As such we begin to address important questions posited by judicial scholars that have heretofore been unanswered, such as whether and how certain interests are more influential than others before the Supreme Court.

We examine whether who wrote and signed the brief matters; that is, whether interest groups with more power make a difference in judicial behavior.Footnote ⁷ Are justices more likely to join or author an opinion on the side of a powerful group, and, if so, under what conditions? Who writes may matter because justices and clerks pay uneven attention to briefs from different groups and power may be a heuristic for taking a brief more seriously and considering it more fully.Footnote ⁸ Network centrality indicates which groups to pay attention to when studying briefs.

When might the presence of a single, highly influential group as a cosigner make a difference on the votes of the justices? If one litigant is supported by many more briefs overall than their opponent, extant literature demonstrates that the side with many more briefs is more likely to prevail, without having to consider the types of organizations filing these briefs (Collins Reference Collins2007, Reference Collins2008; Kearney and Merrill Reference Kearney and Merrill2000). But these large disparities are rare empirically. In the far more common situation, where the number of briefs on either side is well balanced, many voices compete for the justice's attention, and many groups in society signal their interest in the case to the justices. Thus, when the count of briefs on either side is balanced (the modal situation), a brief by a dominant group is more likely to be noticed and read carefully by the justices, which may exert a greater influence, compared to the briefs of less powerful organizations. It is under these conditions that organizations are most likely to be influential.

Subsequently we explore whether it is the content of the brief or the signaling of which outcome political interests in society support that causes changes in behavior. While it may be a combination of the two, if it is a signal about political interests, then the relationship between interest group power and voting should be negative for some justices and we should expect judicial ideology to moderate the effect of amicus briefs. For example, for conservative justices the appearance of a highly central liberal interest group may signal that they should vote against the side supported by the liberal interest group, if briefs serve to inform the justices of which groups in society support which litigant. Indeed, Bailey, Kamoie and Maltzman (Reference Bailey, Kamoie and Maltzman2005) shows that justices’ probability of supporting the solicitor general's (SG) position decreases as the ideological distance between the SG and the justice increases. Their findings suggest that a prominent actor taking an ideological stance in a case can serve as a negative heuristic for justices. We investigate whether this pattern holds more generally for powerful groups.

3 JUSTICE VOTE AND AMICI POWER DATA

We analyze whether amici characteristics have an effect on justice votes, controlling for a standard slate of covariates that have also been shown to correlate with justice voting behavior (Collins Reference Collins2007, Reference Collins2008; Kearney and Merrill Reference Kearney and Merrill2000). We build our data set from the traditional data on U.S. Supreme Court behavior, the Spaeth Database, later utilized and amended by Collins (Reference Collins2007). Specifically, we use the justice centered data format, which contains information on the cases, party resources, and how each justice voted. We control for justice ideology, lower court decision direction, litigant resources, and the presence of the solicitor general as amicus, as in prior literature.

Table 1 presents a basic description of the key independent variables. The first set of variables correspond to the justice centered data. Of the well-known covariates of judicial decision making, one of the strongest relationships involves a justice's ideology (Segal and Spaeth Reference Segal and Spaeth2002). Justices who are already ideologically inclined to rule in a certain way are unlikely to be influenced by amici activity urging the decision they were already inclined to make. As such, it is the interaction of our measures of network influence with justice ideology, as measured by Martin and Quinn (Reference Martin and Quinn2002), which is a focal point of interest here. If groups have an influence, then the marginal effect of briefs should be positive when the justice's baseline proclivity to support the group's side is low. We see from the table that the central tendency of ideology on the Court is 6.27. The range is from 0 to 10.64, where higher values signify a more conservative justice.

TABLE 1.

Summary Statistics

In addition to controlling ideology at the justice level, we include several case-level control variables. The influence of the solicitor general as amicus on the votes of the justices is well documented (Collins Reference Collins2004, Reference Collins2008; McGuire Reference McGuire1998; Nicholson and Collins Reference Nicholson and Collins2008); as such we control for the presence of the solicitor general as either a liberal or conservative amicus. The solicitor general participated in about 9% of cases as a liberal amicus, and about 8% of cases as a conservative amicus. Further, since the ability of the litigants themselves to employ highly experienced counsel varies, and since counsel experience also has a clear influence on the justices’ votes (Johnson, Wahlbeck, and Spriggs II Reference Johnson, Wahlbeck and Spriggs II2006; McGuire Reference McGuire1995, Reference McGuire1998), we control for the resources of the liberal and conservative litigants. We use the litigant resource typology of Sheehan, Mishler, and Songer (Reference Sheehan, Mishler and Songer1992), which ranges from 1 to 10.Footnote ⁹ Table 1 shows that the typical resources of conservative litigants are considerably higher than that of liberal litigants. About 46% of liberal litigants are individuals, compared to 16% of conservative litigants. Finally, it is well established that the Court tends to accept cases that it intends to reverse (Caldeira and Wright Reference Caldeira and Wright1988), so we include the ideological direction of the lower court decision. This variable is equal to 1 if the lower court ruled in a liberal direction. The sign of this variable is expected to be negative, as the Court is anticipated to rule in a conservative direction if it accepts a case that was decided in a liberal direction in the lower courts (Collins Reference Collins2007, 60).

In order to test if particularly influential groups exhibit an effect, we operationalize interest group power in this political arena via a foundational network centrality statistic, and augment the justice vote data with this measure. In constructing the measure we rely on interest group network data from Box-Steffensmeier and Christenson (N.d., Reference Box-Steffensmeier and Christenson2012).Footnote ¹⁰ The data allow us to create a network of amici cosigners from 1930 to the present.Footnote ¹¹ We then use a measure of the relative position of the interest group within the network, or centrality, as a measure of the group's power.

Each group that cosigns a brief, regardless of whether they are the sole signatory or part of a larger group of cosigners, is a node in the network of amici. Whenever two groups cosign a brief together, an edge or tie is created between them and centrality refers to the relative position of the node within the network.Footnote ¹² Consider, for example, the American Civil Liberties Union (ACLU). The ACLU participated as amicus for each decade under analysis here (1946–2001), thus we have a nonzero centrality score for each decade in the justice vote data. Figure 1 illustrates the change in size and shape of the egocentric network of the ACLU for each decade in the data. The ACLU's node is white and its cosigners are gray with the size of the node referring to the relative power of the organization in the egocentric network (i.e., eigenvector centrality). The most obvious feature of the network is the increasing number of cosigners over time. In the 1930s the ACLU cosigned a brief with one organization, by the 1960s that had increased to 39, and by the 1990s it was up to 233, which is smaller than the ACLU's decade high cosigning in the 1980s of 368. This mirrors the larger trend in the network of cosigners, which increased from 194 in the 1930s to 674 in the 1960s to 4,349 in the 1990s. In the full network, however, cosigning has increased each decade with the largest number of cosigners in the 2000s at 5,291.

FIGURE 1.

ACLU Egocentric Networks by Decade

Notes: White nodes refer to the ACLU. Gray nodes refer to other organizations in the ACLU's egocentric network. The size of the node is proportional to the eigenvector centrality of the organization in the egocentric network.

Also apparent from the egocentric graphs are the overtime changes in the network density, which is the sum of the number of cosigned connections in the network divided by the total number of possible connections. Such a measure gives us a rough idea of how well connected all the cosigners in the egocentric network are to each other. This is of particular interest to us because it suggests that the ACLU was a part of different kinds of networks in each decade. While the 1930s through 1950s had very high densities due to a low numerator, since the 1960s the density has remained fairly steady near 0.1. Over the last five decades, the sparsest decade for the ACLU was the 1980s, with a density of 0.086, and the highest the 1990s with 0.141.

We use eigenvector centrality (Bonacich Reference Bonacich1972) to capture the notion that groups best positioned in the network are pivotal actors before the Court. Eigenvector centrality is one of several possible measures of network power. It is particularly appropriate for our case because it delineates global power in the network from more local power structures. Eigenvector centrality refers to the value of the first eigenvector of the network with respect to the sum of the interest groups to which it is connected in every decade. Formally, the eigenvector centrality of a group is

where a_ij = 1 if the groups have a connection, 0 otherwise, λ is the largest eigenvalue of the adjacency matrix, and x_j is group j’s centrality.

Thus a benefit of eigenvector centrality over degree centrality is that “The centrality of a vertex is proportional to the sum of the centralities of the vertices to which it is connected” (Bonacich Reference Bonacich2007, 556). That is, it weights relationships between nodes according to their centralities, not just the number of connections, which means that both direct and indirect relationships are accounted for with this measure.Footnote ¹³ In the context of our interest group network, measuring power with eigenvector centrality suggests that the quality of ties are an important consideration, not just the number of them. Interest groups that are tied to well connected groups are taken to be more powerful than groups who have a similar number of connections to less connected groups.

The five groups with the highest eigenvector centrality scores for each decade are listed in Table 2. While the top five groups do not represent anywhere near all of the coalitional activity, they suggest substantial variation in the kinds of groups that are most powerful within and across the decades. For example, railroads dominate the 1930s, utilities the 1970s, unions the 1980s, and farm associations the 2000s. Other groups are more consistently prominent, like our egocentric network example, the ACLU. The eigenvector centrality for the ACLU ranges from a high of 0.182 in the 1980s, when it was among the top 1% most central groups, to a low of nearly zero in the first few decades of data. Relative to the 1980s, the ACLU had less power in the 1990s with an eigenvector centrality score of 0.016 and even less in the 2000s at 2.108e ⁻⁷. However, relative to all groups in the data, the ACLU has remained among the most powerful groups, consistently in the top 25% since the 1940s, and in the top 10% since the 1970s.

TABLE 2.

Most Powerful Interest Groups

An appropriate measure of interest group power should therefore take into account this longitudinal variance. We break up the network into decade-long windows in order to account for the fact that some interest groups were extremely powerful in one period but not in the next. Thus our measure of power for each interest group is not a single score across the entire 54 year period in study, but a different score for each decade in which there is a relevant case. This windowing process makes an important distinction: political and economic climates change, and thus several groups are more powerful and more present in some periods than others.

In the analyses that follow, the basic measure of interest group power is presented in terms of liberal and conservative ideological direction on the case. To test the impact of network centrality on Court outcomes, we need to summarize the interest group power in favor of the liberal/conservative side for each case. We do so by identifying the maximum eigenvector centrality for all cosigners on the amicus briefs supporting both parties in cases before the Court between 1946 and 2001.Footnote ¹⁴ The result is a continuous variable of the maximum power of interest groups that signed briefs on each side of a case (see Table 1).

We use the maximum and not an additive or summary measure because our hypothesis is concerned with “who signs,” and not “how many sign.” For example, if there is one brief for the liberal side, cosigned by three highly central groups, and on the conservative side there are 20 briefs, all with a few relatively unconnected cosigners, the sheer volume of briefs could overwhelm the high centrality of a few signers on the other side if an additive index was used. The average reduces that complication, but has the downside of washing away the impact that an extremely well connected group can have. Thus, like Anderson and Habel (Reference Anderson and Habel2008) we chose a measure that is not susceptible to biases created by averaging. Using the highest centrality for a single group on each side of a case tests whether having a highly central group join the brief influences the justices. Given that when groups collaborate on a brief, clerks tend to consider the identities of the cosigners more than the raw number of cosigners (Lynch Reference Lynch2004, 59), focusing on the power of the most central group cosigning a brief, as opposed to a metric which incorporates all cosigners, is the appropriate strategy. The alternative hypothesis suggests that justices will be more likely to vote for the litigant whose amici exhibit the greater power. In the next section we test these hypotheses with the judicial vote and amici network data.