This project will develop an original latent variable estimator to scale Supreme Court opinions in a "case space." Systematic study of the determinants and consequences of judicially-created policy have been hampered by the lack of systematic, quantitative measures of legal doctrine. The project will develop a theoretical model of the location of cases and opinions in an underlying "case space" as a function of citations to past cases. Taking advantage of the common law method used in American jurisprudence, the factual components of cases and the legal components of the various opinions written in the course of disposing of the cases will be systematically estimated. Original data will be collected on factual and legal links - citations - among cases to which the estimator will be applied. The method and data will allow scholars to better address the theoretical questions that dominate the literature. For example, the method will allow for new tests of theories of bargaining on collegial courts, theories of the influence of extra-judicial actors on the creation of legal policy, and theories of the effects of Supreme Court appointments.
The project will have broad implications for the study of democratic political institutions. First and most directly, the research will shed light onto the systematic patterns in judicial policy making by providing a quantitative measure of judicial policy. Second, the project will allow future researchers to answer many of the outstanding questions about the constraints under which courts decide cases and the path along which the law develops. For example, a significant question in political science concerns the extent to which judicial decisions are effective protections for minority rights and the extent to which the courts are constrained by majoritarian politics. With a systematic measure of the policies created by judicial opinions, better assessments can be made of how judges shape the law and the extent to which lower courts and extra-judicial officials comply with the courts' decisions. It also will be possible to gain better insight into the forces that drive judicial policy.
." (Throughout this report, I use "we" to refer to myself and Benjamin Lauderdale, who was a graduate student when I applied for this grant but has been a full collaborator throughout the project.) The funding we requested was to collect data on citations between pairs of Supreme Court opinions. Through the period of the grant, the project grew and morphed in a variety of ways, as we discovered more efficient methods for collecting our data and additional aspects of the project we did not anticipate. In this report, we summarize how we spent the funds from our grant and the products we have produced. The efforts on which we spent the funds from this grant can be divided into three categories. 1. While we collected a great deal of data manually (as originally planned), we were able to augment our manual data collection with a number of automated techniques. First, though, we hired a team of research assistants from the Emory Law School and Emory College who worked from March 2011 through September 2012, manually coding citations between pairs of Supreme Court opinions. Specifically, we had them code all citations to non-majority opinions and were able to complete coding of 10,146 spanning the Court's docket from 1946 through 1987. We ave supplemented, and continue to supplement, that process with two automated coding systems. First, we hired a graduate research assistant from Emory University to write computer code to read through Supreme Court opinions, identify citations, and count the number of instances the cited case is cited in the new opinion (a piece of information we had not anticipated collecting, but something that was at least suggested by one of the reviewers on out proposal). This source of data led directly to a new project, described below. Second, we have recently joined forces with a computational linguist from Carnegie Melon, who is able to use the data we collected from the 10,146 cases as well as the automated counting of citations from all opinions to identify citations and generate model-based estimates of the content of the opinion. Essentially, the manually-coded data collected with this grant will be used to train an algorithm to continue the process moving forward. We anticipate a paper growing out of that project (see below). 2. The funds from our grant enabled us to pursue a number of tasks related to the development of our measurement models. First, during the Summer 2011, we devoted a least one entire month to developing a new model for systematically measuring the doctrinal connections among opinions. Using the data collected on the number of citations to each precedent in each new case (described above), we developed a statistical model, which we call the genealogy of law. Second, during the Spring and Summer 2012, we similarly devoted a great deal of time to developing a new model of Supreme Court justice preferences, which relies on information about the similarity between cases. We are currently working on an extension to that paper (described below). This theoretical work required considerable effort, including diagnostic work. 3. We requested funds to present our research at conferences. As described below, we have presented a number of papers fro this grant. Our efforts on this project are still on-going. However, there are several direct products that we can identify at this time. 1. We have presented papers from this project at the following conferences: 2011 European Political Science Association Meeting (Dublin, Ireland), 2011 Political Methodology Meeting (Princeton University), and the 2011 American Political Association Meeting (Seattle, WA). We have also submitted proposals to present on-going research at the 2012 Midwest Political Science Association Meeting (Chicago, IL), 2012 European Political Science Association Meeting (Berlin, Germany), and 2012 Political Methodology Meeting (Duke University). 2. We have completed two papers to date. The first, "Measuring Issue- and Time-Variation in Supreme Court Justice Preferences" is currently under review. The second, "The Genealogy of Law" is also under currently review. We are also in the process of developing a paper that will use the manually-collected data to develop a algorithm for predicting the nature of citations in Supreme Court opinions and another paper that uses the text of Supreme Court opinions in conjunction wit roll call votes to estimate more substantively-connected estimates of Supreme Court justice preferences. Following NSF protocols, the protocols of the journals to which we have submitted, and our own professional standards, the data one which these papers are based and the code that develops the models, which was collected and funded by the NSF, will all be publicly posted upon publication of these papers.