Martin-Quinn scores are the leading statistical measure of where U.S. Supreme Court justices sit on a liberal-to-conservative scale, calculated separately for each term a justice serves. Negative numbers indicate more liberal voting; positive numbers indicate more conservative voting. The scores come from a Bayesian model developed by political scientists Andrew D. Martin and Kevin M. Quinn and published in Political Analysis in 2002.1Cambridge University Press. Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953-1999 Because the model estimates a fresh score for every term, it can track how a justice’s position moves across a career rather than fixing them in place at the moment of appointment.
How to Read the Scale
A justice with a negative Martin-Quinn score voted, on average, with the liberal side of the Court that term. A positive score reflects conservative voting. The farther from zero, the more consistently a justice sided with one wing in closely divided cases.
The extremes in the modern dataset give a sense of range. Justice William O. Douglas registered -7.893 in his final term in 1974, the most liberal score in the modern era.2Cardozo Law Review. A Bolt of Lightning: Measuring the Impact of Modern Transitions on the Supreme Court Justice Clarence Thomas has sat near the top of the conservative range for his entire tenure; his career average is 3.604, compared with -2.831 for Thurgood Marshall, whom he replaced in 1991. That transition produced the single largest ideological swing from any justice replacement in the modern database.
The mathematical direction of the scale is a convention, not a natural feature of the data. The raw model could just as easily flip signs; Martin and Quinn used “semi-informative prior distributions” to lock negative to liberal and positive to conservative so the numbers read intuitively.3Washington University in St. Louis. Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953-1999
How the Scores Are Built
The inputs are votes. The model uses the Supreme Court Database, originally compiled by political scientist Harold Spaeth, and looks only at non-unanimous cases decided with written opinions or by an equally divided vote. Unanimous decisions are excluded because they reveal little about ideological disagreement. Each vote is coded simply as a vote to reverse or affirm the lower court.
From those votes, the model infers an “ideal point” for each justice in each term, using item response theory. The method borrows from the same family of statistics used to score standardized tests, where a student’s ability is inferred from the pattern of questions answered correctly. Here, the cases play the role of test questions and the justices play the role of test-takers.3Washington University in St. Louis. Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953-1999
Why the Scores Change From Year to Year
Earlier measures of judicial ideology assigned each justice a single number for an entire career. Martin and Quinn built their model to allow that number to move. In their framework, a justice’s ideal point in any given term is drawn from a probability distribution centered on the prior term’s ideal point, a structure known as a random walk. The result is a smooth trajectory that can shift gradually but not jump wildly from one year to the next.
Solving the model requires Markov chain Monte Carlo (MCMC) simulation, a computational technique that generates many random samples to approximate the probability distributions behind the numbers. The output for each justice in each term is both a point estimate and a measure of uncertainty around it. The recommended summary statistic is the posterior mean of the ideal point.
What the Scores Reveal About Drift
One of the main findings to emerge from Martin-Quinn scores is that justices routinely move ideologically over the course of their careers. The data undercut what the creators described as the “conventional view” that justices remain locked-in representatives of the president who appointed them. Virtually every justice serving since the 1930s shifted left or right, and some reversed direction more than once.4Washington University in St. Louis. Ideological Drift Among Supreme Court Justices: Who, When, and How Important
Justice Harry Blackmun is the clearest example. Nixon appointed him in 1970, and he started as one of the Court’s most conservative members. Over more than two decades he moved steadily leftward, eventually becoming one of its most reliable liberal votes. In 1994 he announced, “From this day forward, I no longer shall tinker with the machinery of death,” repudiating the death penalty he had long supported. Justices Hugo Black and William O. Douglas moved the other way at the end of their careers: both grew more supportive of civil liberties claims for most of their tenures, then saw that support taper off.
Drift matters for case outcomes. Research using the scores has argued that if Justice Sandra Day O’Connor’s preferences had remained where they were at her appointment, she likely would not have provided the fifth vote to uphold the affirmative action program in Grutter v. Bollinger in 2003. A broader study of drift patterns found that justices on average begin as centrists and tend to drift slightly liberal after roughly 15 years on the bench.5National Center for Biotechnology Information. Functional Data Analysis of Supreme Court Ideological Drift
The Current Court by the Numbers
Because the Court decides cases by majority vote, the justice whose ideal point falls in the middle of the nine effectively controls close outcomes. Martin-Quinn scores identify that median justice in every term.6Martin-Quinn Scores. Martin-Quinn Scores Project
In the 2021 term, the first full term of the current 6-3 split, the scores quantified a wide gap between the two wings. Sotomayor sat at -4.170, Breyer at -2.033, and Kagan at -1.835. The nearest conservative was Roberts at 0.682, followed by Kavanaugh at 0.791, Gorsuch at 1.019, and Barrett at 1.318. Alito (2.458) and Thomas (2.949) anchored the right.7Washington University in St. Louis. 2021 Term Data Report Republican-appointed justices voted in the majority 86 percent of the time that term, Democratic appointees 62 percent, a 24-point gap that was the largest in the 2010 through 2021 period.
By the 2022–2023 term, Justice Kavanaugh was the median at 0.524, and Alito had overtaken Thomas as the Court’s most conservative member for the first time.8Axios. Supreme Court Justices Political Ideology Chart Across the October 2022 through October 2024 terms, Roberts, Kavanaugh, and Barrett clustered near the center, while Sotomayor sat 4.648 units from the median and Thomas 2.613 units out on the other side.9Legalytics. Measuring the Justices: An Portrait
Where the Measure Falls Short
Martin-Quinn scores have drawn sustained criticism, and reading them carefully means knowing what they do not do.
The deepest problem is cross-era comparability. Because ideology is inferred from voting coalitions, the meaning of a score depends on the cases the Court happened to hear. If the Court’s agenda shifts over the decades, the yardstick shifts with it, so comparing a 2020 score to a 1970 score in any absolute sense is risky. Political scientist Michael Bailey of Georgetown University built an alternative, “Bridge Ideal Points,” specifically to address this limitation by using thousands of bridging observations in which justices, members of Congress, or presidents took positions on the same legal questions across different eras.10Georgetown University. Bridge Ideal Points Bailey’s estimates sometimes diverge from Martin-Quinn. Where the Martin-Quinn scores suggest the Court hit a conservative peak during the 1970s (the era of Roe v. Wade and Furman v. Georgia), Bailey’s bridge estimates place that Court as relatively liberal compared with later periods.11Georgetown University. Is Today’s Court the Most Conservative in Sixty Years?
Legal scholar Wade Farnsworth, writing in the Northwestern University Law Review, raised a related objection: the original method codes only whether a justice voted to reverse or affirm, not whether that vote was ideologically liberal or conservative. A reversal of a liberal lower court ruling and a reversal of a conservative one look identical in the raw data.12Northwestern Law Review. The Use and Limits of Martin-Quinn Scores to Assess Supreme Court Justices
The scores also compress all of judicial ideology onto a single dimension, collapsing distinctions between, say, a justice who is conservative on criminal law but libertarian on free speech. They create an endogeneity problem in some research designs, since using scores derived from voting behavior to predict voting behavior risks circular reasoning.13Cambridge University Press. Measuring Judicial Ideology Through Text And they only work for courts that decide cases as a single body, which makes extension to lower courts (where judges sit in rotating panels) difficult.
One quirk worth knowing about the current numbers: on a Court where the same three justices are routinely in the minority, the model infers their positions as farther from the winning coalition’s center. Analysts call this “minority inflation,” and it partly explains why the liberal justices’ scores look so distant from the median on today’s Court.
Alternative Measures
Several other approaches to judicial ideology exist, each with trade-offs. Segal-Cover scores rely on pre-confirmation newspaper editorials to estimate a justice’s ideology at the time of appointment, but they are static and cannot capture drift. Appointment-based indicators (coding judges by the party of the appointing president) correlate with voting on politically divisive issues but ignore wide variation among appointees of the same president. DIME scores, based on campaign contributions by judges or their law clerks, capture ideological variation dynamically but through an indirect proxy. Newer text-based methods such as “Wordshoal” analyze the language of judicial opinions, which may help extend measurement to judges whose voting coalitions are harder to observe.
A 2019 study by Rok Spruk and Mitja Kovac replicated the Martin-Quinn methodology with a modification, coding the ideological direction of each vote rather than just the reversal-or-affirmance binary, and extended the estimates back to 1793. Their results showed a moderate-to-strong correlation with the original scores and generally confirmed the dynamic trends Martin and Quinn had identified.14ScienceDirect. Replicating and Extending Martin-Quinn Scores
Where to Find the Data
The official Martin-Quinn dataset is hosted jointly through Washington University in St. Louis and Emory University. Scores are available for every justice who served from the October 1937 term through the October 2024 term, which is the most recent release as of December 2025.15Martin-Quinn Scores. Martin-Quinn Scores – Measures The data comes in CSV, Stata, Excel, and SPSS formats and includes both justice-level ideal point estimates and court-level data identifying the median justice in each term. Replication code and the underlying case data are available for download, and researchers citing the scores are asked to credit the original 2002 Political Analysis paper.16Martin-Quinn Scores. Martin-Quinn Scores – Replication