New Project: Using Supreme Court Data to Find Trends & Predict Case Outcomes
NOTE: this is part 1 of a 2-part series exploring the Supreme Court. The purposes of this mini-series are to identify trends in the Supreme Court using case data, and to consider whether cases can be predicted using variables that are understandable by the common person (such as myself). This is part 1, using exploratory data analysis to explore emerging trends with America’s highest court. Part 2 used a variety of logistic regression models, and can be found here. All errors are my own. To get the R code utilized to do this analysis, reach out to me at ahegedus@hegedusanalytics.com. The data can be found at the database here, though I can also provide it on request.
Edit: Updated 5/23/24 clarifying how year extracted.
Introduction and Problem
Every year, the Supreme Court receives more than 7,000 petitions of appeal on a previous ruling of a court case. In total, the Court claims to review anywhere from 100-150 cases in a year. However, data in the Supreme Court database indicates a decline in the number of cases the Court hears every year, having last exceeded 100 in 1997 (1). While the pandemic has almost certainly accounted for some decline from 2020 to 2022, there is a clear downward trend starting from 1982.
Cases heard by the Supreme Court in every year from 1947-2022. Years extracted from case id, as date argued blank if no oral argument.
While getting a case in front of the Supreme Court is a difficult endeavor, it becomes increasingly difficult when the court hears less cases almost every year. Accordingly, it is worth considering whether we can accurately predict the outcomes of Supreme Court cases through machine learning methods. Doing so would lead to an increased understanding of how the Supreme Court looks at cases, and could lead to more efficiency as to whether to bring a case there in the first place.
The Roberts courts: Compared to themselves and to previous courts
Justice John Roberts has held the Chief Justice chair since 2005. Through 2022, there are 1,348 cases that his courts have heard that appear in the dataset. As previously stated, his courts have continued a trend of seeing less cases, though this is an issue that came before him as Chief Justice. Interestingly, however, Roberts has presided over more unanimous decisions on a rate basis than other Chief Justices of the modern era (2).
When looking at his courts, some courts naturally received more cases than others due to turnover in the justices (3). The following table shows the number of cases each Roberts court heard, the number of unanimous decisions, and the ratio between the two. Excluding the 1st (lack of cases) and 10th (lack of cases and ongoing), we see that the Roberts Courts have had some variety in their unanimous decision rates, ranging anywhere from about 37.5% to about 50% of the time.
| Natural Court | Total Cases Heard | Unanimous Decisions | Ratio |
|---|---|---|---|
| Roberts 1 | 26 | 19 | 0.730769230769231 |
| Roberts 2 | 293 | 114 | 0.389078498293515 |
| Roberts 3 | 92 | 42 | 0.456521739130435 |
| Roberts 4 | 408 | 202 | 0.495098039215686 |
| Roberts 5 | 88 | 44 | 0.5 |
| Roberts 6 | 121 | 58 | 0.479338842975207 |
| Roberts 7 | 132 | 52 | 0.393939393939394 |
| Roberts 9 | 130 | 49 | 0.376923076923077 |
| Roberts 10 | 58 | 29 | 0.5 |
Also worth noting is that the Roberts courts do tend to skew conservative, with about 52% of their decisions going in that direction through 2022, compared with around 45% of decisions going a liberal direction (4). This puts his courts in the middle of both metrics amongst modern Supreme Court Chief Justices.
Finally, due to judicial review, established in part in Marbury v. Madison in relation to laws made by Congress, the Supreme Court is able to declare actions unconstitutional. In the dataset, this includes Acts of Congress, State or Territorial laws, and local ordinances. As expected, the proportion of cases declared unconstitutional does vary by year (5). The overall proportional use of judicial review to declare something unconstitutional is lower than the 1980’s and 1990’s, though the waves of its use have seemed to increase in recent years.
Proportion of cases in which the challenged action is declared unconstitutional.
Overall, we can see that the Supreme Court is hearing less cases than it has previously, that the Roberts courts are middle of the pack in terms of modern era conservative and liberal decisions, though there can be some variance by court. We also see an uptick in their use of judicial review to declare actions unconstitutional, though this can also come down to a variety of factors.
In part 2, I will use multiple logistic regression methods to try predicting Supreme Court cases. Thank you for reading!
Citations linked on relevant lines. Thanks to:
Dr. Jeff Summerlin-Long for inspiring the idea
Dr. Christopher Handy and Dr. Andrii Babii for their teaching of these techniques
Dr. Rui Sun for teaching me to code initially (and for teaching modeling techniques)
The Mercer University Economics and Mathematics Departments
The University of North Carolina at Chapel Hill Economics and Public Policy departments
Notes:
(1) I am not necessarily disputing the claim of the Supreme Court accepting over 100 cases per year, as there may be other explanations that I would not understand as a non-attorney why the data does not match up to that. I do maintain that the number of cases heard by the Supreme Court overall per year has declined. There are 1059 cases in the dataset that did not have a “date argued,” though looking at case id, it appears many of these were in the early days of the modern Supreme Court.
(2) “Unanimous decisions,” in the context of this blog post, mean that there were no dissenting votes.
(3) A “natural court” is a period in which the justices serving on the court are the same. When Roberts first joined the Supreme Court, his first court would be Roberts 1, after Justice Sandra Day O’Connor retired and was replaced by Justice Samuel Alito would be Roberts 2, etc. Oddly, there is no Roberts 8 in the data, and I am admittedly unsure as to why.
(4) Some decisions were unclear in terms of direction, which means the ratios will not add to 100%.
(5) For some cases, declaring an action unconstitutional may not be an option.
