The Supreme Court Project attempts to build on and extend empirical work explaining the United States Supreme Court’s decision making and its review of the 13 United States Courts of Appeals. We have started our inquiry by looking at the question of how to measure the Supreme Court’s reversals of the United States Courts of Appeals (Phase I) and have recently extended that analysis to build an empirical model that explains and predicts Supreme Court Justice voting and Court decisions (Phase II).  As discussed below, we have just published an article setting forth our current model.

Phase I

Initially, the Project examines whether there is a better way to track how well the Courts of Appeals do before the Supreme Court, conventionally measured through reversal and affirmance rates. Traditionally, the reversal rate for a circuit court is an outcome-based, score-card test: For any given term of the Supreme Court, a reversal rate can be established by dividing the number of appeals where the Court reversed the circuit by the total number of appeals decided by the Supreme Court for that circuit. While straightforward, this traditional method suffers from two fundamental weaknesses: (1) it measures outcomes rather than legal reasoning; and (2) it fails to account for decisions implicitly reversed or affirmed by the Supreme Court, even though not directly before the Supreme Court on appeal.

Rule 10(a) of the Supreme Court Rules explains that a reason for granting certiorari is that a Court of Appeals “has entered a decision in conflict with the decisions of another United States court of appeals.” Our analysis confirms this as, for the six terms of the Roberts Court (2005-2011), nearly half of the Supreme Court’s annual merit decisions address substantive legal issues for which there is a circuit split. For each of these cases, the Supreme Court ruled on a legal issue addressed not just in the case on appeal, but also one or more “shadow decisions” (i.e., court of appeals decisions that have ruled on a legal issue that comprises the circuit split). Any measure of reversals and affirmances that does not account for these shadow decisions is therefore incomplete and potentially misleading.

We, therefore, have established a “full” measure of the circuit courts’ reversal and affirmance rates, taking into account both decisions on direct appeal and these shadow decisions. This measure, we believe, provides a more robust and more accurate view of the relationship between the United States Courts of Appeals and the United States Supreme Court.

We discuss our preliminary results in an article, “Towards a Better Measure and Understanding of U.S. Supreme Court Review of Courts of Appeals Decisions,” published in BNA Weekly. Generally, we conclude:

  • the Supreme Court reverses the Courts of Appeals less often than is commonly thought;
  • the full measure of reversals identifies different Courts of Appeals as least reversed than those identified under the traditional method;
  • in resolving circuit splits, the Supreme Court affirms the majority approach of the circuit approximately 50% of the time;
  • our analysis permits us to construct a concordance table showing the degree to which the various Courts of Appeals agree with each other, much like the concordance tables showing agreement among justices’ voting.

Overall, we believe this analysis, which takes into account both cases directly on appeal and shadow decisions, provides a more robust and more accurate view of the Supreme Court’s review of the Courts of Appeals.

For those interested in more detailed Tables and Charts regarding this analysis, they are available here.

Phase II

The second phase of our work has focused on an empirical investigation of the determinants of US Supreme Court Justices’ voting patterns.  In particular, we have sought to extend the landmark work of Washington University Supreme Court Forecasting Project[1], as well as recent work by Judge Richard Posner and Professors Lee Epstein and William Landes[2], to predict Justices’ votes and assess what variables are statistically significantly correlated with their votes.

Specifically, we have just published an article entitled “An Econometric Investigation of the Determinants of US Supreme Court Decisions” in Volume 83.4 of the Tennessee Law Review that describes the model and results.  Click here for the article.

In the article, Mr. Summers, along with Michael J. Newman (a former Hangley Aronchick colleague) and Michael Cliff, PhD (Vice President, Analysis Group) examine the decisions of the first eight terms of the Roberts US Supreme Court (October 2005 through 2012) to evaluate the variables that may explain the more than 4,150 votes by the ten Justices who sat on the Court during this period.  The model focuses on five categories of variables that may be correlated with Justices’ votes and Supreme Court decisions: characteristics of (a) the court of appeals judge who authored the decision reviewed, (b) the advocates before the Supreme Court, and (c) the Justices themselves, in addition to (d) the characteristics of the case and (e) a set of control variables.

Our multivariate logistic regression model accurately predicts 70% of the Court’s decisions and from 70 to 78% of the Justices’ individual votes.

For a full discussion of the study and results, please read the article.  Highlights of our results include the following:

  1. Advocacy experience before the Supreme Court matters

Clients and advocates with cases before the Supreme Court will be particularly interested in our finding that relatively experienced Supreme Court oral advocates are more likely to succeed by the Court against less experienced adversaries.  We found that the 20% most active advocates in a year had a significantly greater likelihood of success when they opposed an advocate not among the 20% most active.  Also, those advocates with a better batting average before the Supreme Court were more likely to prevail against advocates with a lesser record.

  1. Amicus support, particularly from the Solicitor General’s Office, correlates with success before the Court

Of particular interest as the Trump Administration nominates an Attorney General and Solicitor General, the study shows that a party before the Supreme Court enjoys a very substantial (10.6 percentage points) greater likelihood of success if the SG’s Office is on its side and submits a supporting brief.  This result suggests that who heads the SG’s office will be very influential in the decisions of the Supreme Court.  However, one caveat:  Because the statistical analysis can show only that SG support correlates with a party’s success, not that the support caused success, this finding results from the SG’s Office being strategically effective in identifying the positions that would otherwise prevail before the Supreme Court, and not from the SG’s Office being extraordinarily persuasive or influential.

  1. Ideology matters in at least two ways

As the county looks ahead to the debate over the confirmation of a replacement for Justice Antonin Scalia, the study confirmed what is well accepted – ideology matters.  The study shows this in two ways.

First, we examined the match between the ideological direction of the court of appeals decision under review (i.e., is it liberal or conservative leaning) with the party affiliation of the President who appointed the Justice voting in the case.  Our analysis shows that a Republican (Democrat) appointed Justice is much less (15.2 percentage points) likely to vote to reverse a conservative (liberal) leaning decision.

Second, we looked at whether a Justice appointed by a President from one political party was less likely to reverse the decision written by a court of appeals judge who was appointed by a President from the same party.  Our analysis shows that when the party of the President who appointed the Justice and judge are the same, Justices vote to reverse the lower court 5 percentage points less than show the parties differ.  Among other things, this suggests that opinions written by the fifty-five court of appeals judges appointed by President Obama will fare less well in front of the as-yet-Trump-appointed Justice than they would have had Chief Judge Merrick Garland been confirmed.

  1. Who wrote the Court of Appeals decision under review makes a difference

The study also looked at whether various characteristics of the court of appeals judge who wrote the decision under review correlated with Justices’ votes to affirm or reverse the decision.  Counterintuitively, we found that Justices were more likely to reverse the decisions written by longer tenured members of the court of appeals than those written by newer judges to the court of appeals.  (Specifically, a judge on the court of appeals for 9 years was nearly 3.8 percentage points more likely to be reversed than a recently appointed judge.)  A possible rationale for this is that court of appeals judges are more cautious earlier in their career, perhaps “auditioning” for elevation to the Supreme Court.

Other characteristics that might serve as a proxy for the “competence” of the court of appeals judge – whether the judge was a former federal law clerk or his ABA quality rating at the time of confirmation – were found not to be statistically significant indication of reversal.

Additional Articles


  • John S. Summers is a shareholder at Hangley Aronchick Segal Pudlin & Schiller. He received a BA from Wesleyan University in 1980 and a JD from the University of Pennsylvania in 1984.
  • Michael J. Newman is a former Hangley Aronchick Segal Pudlin & Schiller associate. He received a BA from the University of Pennsylvania in 2002 and a JD from Columbia Law School in 2006. He is a member of the Supreme Court of the United States Bar.
  • Michael Cliff is Vice President at Analysis Group. He has a BS from Virginia Tech and a PhD in finance from the University of North Carolina at Chapel Hill.
  • David Klein is an Analyst at Analysis Group. He has a BA in Economics and Political Economy from Washington University in St. Louis.

Research Assistants

  • Sharon Weiss is an assistant at Hangley Aronchick Segal Pudlin & Schiller.
  • Danielle Acker Susanj earned her law degree from the University of Pennsylvania Law School in 2013. She attended Wheaton College and earned a BA in history.
  • Jonathan Conigliari was also a member of the University of Pennsylvania Law School Class of 2013. He received a BA in French and an MA in French studies from New York University.
  • Sarah Gignoux-Wolfsohn is working toward her PhD at Northeastern University. She earned her BA in biology and French, with honors in biology, from Wesleyan University.
  • Gabrielle J. Niu earned her MA in East Asian languages and cultures from the University of Pennsylvania in 2012. She is a graduate of Bowdoin College, where she majored in Asian studies and minored in chemistry.
  • Ben Jackal is a member of Temple University School of Law Class of 2014. He attended Cornell University and earned a BA in history.
  • Colleen Daniels graduated in 2015 from the University of the Arts in Philadelphia, where she was a Crafts major with a concentration in woodworking. She also had a minor in creative writing.
  • Ellen Boyer graduated from Temple University in 2013 with a Bachelors in English and Political Science.
  • David Huppert is a 2015 graduate of George Washington University, where he earned a BS in Economics and a minor in English.


[1] Theodore W. Ruger et al., Essay, The Supreme Court Forecasting Project: Legal and Political Science Approaches to Predicting Supreme Court Decisionmaking, 104 COLUM. L. REV. 1150 (2004) (discussing the results of a statistical model used to predict the outcome of Supreme Court decisions)

[2] Lee Epstein, William M. Landes & Richard A. Posner, The Behavior of Federal Judges:  A Theoretical and Empirical Study of Rational Choice  (2013)