Estimating the Ideal Points of Organized Interests in Legal Policy Space

Abstract

Scholars have been limited in the development and testing of theory regarding the incidence and impact of organized interest advocacy at the U.S. Supreme Court due to a critical measurement issue - the inability to properly locate these interests in the legal policy space in which the Court operates. We treat the positions articulated by organized interests in their amicus curiae briefs as “votes” in Court cases, allowing us to use an IRT model to estimate the locations of both the 600 most active organized interests and the justices in the same legal policy space. The resulting ideal point estimates yield substantive implications (e.g., the distribution of organized interest ideal points is slightly to the left of the justices) and lend themselves to a number of future applications to important questions involving judicial politics in the United States.

Keywords:

• Supreme Court
• organized interests
• interest groups
• amicus curiae
• ideal points

Acknowledgments

This paper is based upon work supported by the National Science Foundation under Grant No. SES-1351922. The amicus curiae data and ideal point estimates discussed in this article can be found at https://amicispace.ucmerced.edu/.

Notes

1 Measures of the ideological positions of organized interests have largely been limited to the subjective coding of interests’ self-advertised policy positions (see Hansford 2004; Manzi and Hall 2017). These measures suffer from subjectivity, incompleteness (many interests active at the Court do not advertise their policy positions), coarseness (i.e., they classify interests as either liberal or conservative), and noncomparability with measures of judicial ideology.

2 For reasons of theory, practicality, and convention, we assume that the underlying policy space is unidimensional. Spatial theories of courts and judging are typically based on the assumption that there is a single, fundamental dimension that can reasonably represent legal policy space (e.g., Hammond, Bonneau, and Sheehan 2005; Owens 2010). On the practical side, for many of the interests there will not be enough “votes” to relax this assumption and allow for a second dimension. Finally, with the exception of Lauderdale and Clark (2012), work on ideal point estimation for justices (e.g., Clark and Lauderdale 2010; Martin and Quinn 2002), judges (e.g., Epstein et al. 2007), interest groups (e.g., Bonica 2013), legislators (Bailey 2007), and agencies (e.g., Clinton et al. 2012) typically assumes unidimensionality. Below, however, we empirically examine the plausibility of this assumption.

3 There is also evidence of amicus curiae briefs influencing the Court at the certiorari (Caldeira and Wright 1988) and opinion-writing (Collins, Corley, and Hamner 2015; Spriggs and Wahlbeck 1997) stages of the decision process, though this is not of as much concern here as our estimates are based on positions taken on the merits.

4 There are also missing votes for all the justices in the data in the sense that Justice Scalia, for example, did not vote in any of the cases prior to his appointment in 1986. This form of missingness is ignored in all IRT models of justice ideal points and we likewise ignore it here. Importantly, this form of missingness is not determined by any sort of indifference-generated abstention process.

5 Hansford, Depaoli, and Canelo (2019) use this approach to estimate the location of U.S. solicitors general in the Court’s legal policy space.

6 An alternative approach to the missingness issue is to focus on one organized interest at a time and only include cases in which the interest filed a brief. This approach is used by Fischman (2015) in his examination of the National Association of Criminal Defense Lawyers and U.S. Chamber of Commerce.

7 While for the reasons articulated above we view the Rosas, Shomer, and Haptonstahl (2015) model as superior for our application, we also estimated the ideal points of the organized interests and justices using a traditional IRT model in which it is assumed that missing votes are missing-at-random. The two sets of ideal points correlate at r = .742.

8 Harlan, Douglas, Marshall, Brennan, Frankfurter, Fortas, Rehnquist, Scalia, and Thomas have prior means of 1.0, −3.0, −2.0, −2.0, 1.0, −1.0, 2.0, 2.5, and 2.5, respectively. Their prior variances are set to 0.1. All other justices have diffuse priors with the prior mean set at 0 and the prior variance set at 1.0.

9 For the 600 organized interests we include in our analysis (i.e., the 600 most active interests at the Court during the 1953–2013 terms), the average number of amicus brief “votes” per Court term is 0.79. Only one organized interest – the ACLU – averages more than 10 votes per term. Only eight other interests average five or more votes per term. Thus even using a low minimum vote threshold of five votes per term would lead us to estimate dynamic ideal points for 591 fewer organized interests and to ignore the non-random nature of missing votes.

10 Nine of the groups in her data have scores for all ten years in her time frame and plotting these scores over time reveals that these positions are quite static (see Figure A1 in the Online Appendix).

11 See http://scdb.wustl.edu/.

12 To identify the position expressed by a brief, we consult the cover page, summary of argument, and concluding statement. The cover page provides sufficient information to code the position for the vast majority of briefs. Some briefs have cover pages that do not provide a clear statement about the position or explicitly claim to support neither party. Many of these briefs, though, articulate a clear position in the summary of the argument and/or concluding statement and we rely on this information when the cover page is unclear. We are ultimately able to identify the position taken in 97.4% of the briefs in our data. The remaining 2.6% of the briefs cannot be coded as clearly advocating for reversal or affirmance, even after examining the summary of argument and concluding statement. We treat these ambiguous or neutral briefs as equivalent to abstentions.

13 The setting of 10 “votes” as the minimum is necessarily somewhat arbitrary. Efforts to estimate organized interest ideal points in other contexts do not establish a clear precedent for us to follow, as different studies use quite different cut-offs for the minimum number of votes. For example, Bonica (2013) uses a minimum of 30 votes while Crosson, Furnas, and Lorenz (2020) set their minimum at five. Our use of a 10-vote threshold thus falls within the range defined by recent scholarship. An alternative approach would be to include all the interests, regardless of their number of votes, and utilize Bailey’s (2001) random effects ideal point model for small numbers of votes. The potential downside to this approach, though, is that it is not clear how to then handle the missing data issue for the organized interests.

14 The United Steelworkers of America appears in both our data and Bonica’s, but we do not include it here as it has two very different CFscores (one of which is a substantial outlier).

15 Figures A2 and A3 of the Online Appendix provide the distributions of ideal point estimates and comparisons with the issue-pooled estimates, respectively. Figures A4–A7 present issue-specific ideal point estimates for select interests and justices.