Inter-Institutional and Cross-Temporal Ideal Points for Supreme Court Justices, Presidents and Members of Congress
Michael A. Bailey, Georgetown University
Updated: May 2021
These Supreme Court, presidential and congressional ideal points extend the ideology estimates in Bailey (2013, 2007) to cover years 1950 to 2020. They are comparable over time and can be directly compared across institutions.
The estimates achieve comparability by incorporating the following bridging and other observations.
- Actors in one institution taking positions on votes in another institution
- 42,415 observations of members of Congress taking positions on Supreme Court cases via amicus briefs, public statements, roll call votes and Tweets;
- 1,320 observations of presidents taking positions congressional roll calls;
- 681 observations of presidents taking positions on Supreme Court cases.
- Actors taking positions on votes from an earlier period. For example, we have 2,380 observations of justices taking positions in their opinions on previous cases.
- Votes having an ideological relationship. For example, the ideological cut point of the 1991 Civil Rights Act is to the left of the 1964 Civil Rights Act because the 1991 measure extended the earlier measure.
- 277 votes in Congress that could be coded as being more/less conservative than another vote;
- 360 cases in the Supreme Court that could be coded as more/less conservative than another case.
- The data includes votes that are occur within Congress and the Supreme Court.
- 689,460 observations of members of Congress voting on topics related to those that dominate the court docket, including civil rights, race, abortion, speech and Senate confirmations of judicial nominees. These include 175 roll calls that were on identical legislation across the House and Senate (such as conference reports).
- 84,115 votes by justices on Supreme Court cases.
If you use the data, please use the following citations:
- Bailey, Michael A. 2021. Bridge Ideal Points. https://michaelbailey.georgetown.domains/data/
- Bailey, Michael A. 2013. Is Today’s Court the Most Conservative in Sixty Years? Challenges and Opportunities in Measuring Judicial Preferences. Journal of Politics 75 (3): 821–834.
If you spot errors or anomalies, have questions, have observations of inter-institutional position taking you would like to suggest or would like your research included in the applications listed below, please contact me at Michael.Bailey@georgetown.edu.
The figure at the top of the page shows ideal points of House, Senate and Supreme Court medians and presidents over time. The figures below show ideal points of justices over time, by period.
How do these scores differ from Martin and Quinn Scores?
The Bridge Ideal Points for justices correlate reasonably highly with Martin and Quinn scores and often produce similar conclusions when used in analyses.
There are two differences, however. First, Bridge Ideal Points are designed to be intertemporally comparable. Martin and Quinn scores assume the ideological cutpoints of Supreme Court cases are drawn from a common distribution across years, which makes them less comparable across time when the case load changes. For example, the figure below (from Bailey 2013) shows that Martin and Quinn scores in the early 1970s suggest the court moved dramatically to the right, reaching a conservative peak around the time it legalized abortion and banned the death penalty. The Bridge Ideal Points suggest that while the Court did move to the right, it was still relatively liberal but voting on a caseload that was quite liberal.
Second, Bridge Ideal Points for justices are comparable to Bridge Ideal Points for presidents and members of Congress, while Martin and Quinn scores only cover the Supreme Court.
How do these scores relate to Nominate?
Nominate scores are widely used to measure congressional ideology. Depending on the time period, voting in Congress is best explained by one or two dimensions.
A major challenge when studying judicial politics is that neither of the two Nominate dimensions characterizes judicially-relevant political preferences in the post-war period.
In the 1950 to 1980 period, congressional politics were two dimensional: the first dimension largely reflected party loyalty while the second dimension reflected divides over race (and, to a lesser extent, crime and culture). First dimension Nominate scores in this period do not at all reflect political divides on race and related topics. Nominate second dimension ideology scores reflect judicially relevant politics from 1950 to 1980, but then become much less useful after 1980 or so.
The figure below (from Bailey 2021) illustrates the challenges of using Nominate for judicial politics by plotting Nominate ideal points for selected senators. The clearest issue is that segregationists such as Eastland and Russell are moderate on the first dimension; it is simply untenable to view them as moderates on topics like race, crime and culture that dominated the court’s docket in the 1960s.
Using these preferences over time creates other problems. If we use the first dimension scores, modern moderate Republicans such as Lincoln Chafee and Susan Collins would be similar to (even more conservative, actually) than segregationists like Russell and Eastland. More conservative Republicans like John McCain would be markedly more conservative than the segregationist senators.
Using the second dimension creates a different – and almost opposite – set of issues. The Nominate second-dimension preferences of moderate Democrats like Sam Nunn and Fritz Hollings look like the segregationist senators. And the second-dimension Nominate scores for Chafee and Collins are very liberal, more liberal than Hubert Humphrey and Barack Obama. McCain would be to the left of Obama when using the second dimension, as unbelievable as the implication that he was to the right of the segregationists on the first dimension.
In addition, these Nominate scores are static. It certainly seems plausible that the Senator Hollings of 1970 would differ from the Senator Hollings of 2000, something ruled out by the assumption of fixed preferences.
The Bridge Ideal Point estimates differ markedly from Nominate scores. The figure below shows the Bridge Ideal Points for selected senators over time. The segregationists are, as they should be, the most conservative. Those who remained in Congress moderated over time.
The Bridge Ideal Points seem to avoid some other issues that arise with the Nominate scores as well. Obama is clearly to the left of McCain. Moderate Democrats such as Hollings and Nunn started as conservative, but moved steadily to the left, a pattern consistent with their eventual support of the 1991 Civil Rights Act, for example. Moderate Republicans like Chafee started quite liberal, but moved to the middle.
The inability of Nominate scores to capture judicially relevant ideology may affect other measures as well. For example, Judicial Common Space scores (JCS) use first dimension Nominate Common Space scores to estimate comparable preferences for Supreme Court justices and lower court judges (Epstein, Martin, Segal and Westerland 2007). The inability of the first dimension Nominate Common Space scores to reflect the politics of the court from 1950 to 1980 or so may make JCS scores less useful in that time period.
Are ideological preferences really one dimensional?
For Congress, we know that there was a very important second dimension in the early part of the period covered (from 1950 to roughly 1980). After that, politics tended to converge to a single dimension.
As discussed above, using the Nominate first dimension for judicial politics from 1950 to 1980 is problematic. The Bridge Ideal Point estimates are limited to cases and votes related to race, crime and culture issues for the entire period, meaning that the preferences reflect preferences on a relatively stable set of issues.
In the Supreme Court, one dimensional models are highly predictive of voting. That is not to say preferences are perfectly one-dimensional. Non-ideological factors matter (Bailey and Maltzman 2011) and Lauderdale and Clark (2012) have shown that preference alignments can vary across substantive legal issues. Nonetheless, for many studies of inter-institutional judicial politics (about nominations or separation of powers theory, for example), a one dimensional characterization is reasonable given that Congress typically can often deal only with general ideological patterns.
Have the ideal point estimates changed since the 2011 release?
The data set has been expanded to include labor cases and some of the coding of specific bridge observations have been updated.
Generally the estimated ideal points in this release are similar to the 2011 release. The largest differences are in the early era (Frankfurter, Vinson and Clark, for example). Two factors should be kept in mind regarding these early ideal points. First, the unidimensionality assumption may be less valid before the court came to focus much more on race, crime and culture after 1953. Before 1953, issues before the court were more likely to involve labor and other issues for which the New Deal coalitions looked different than Warren Court divisions. Second, there is much less information in these early eras, especially relative to Congress. Therefore the scores before 1953 should be taken with more caution (which is true for Martin and Quinn scores as well). For analyses in which the pre-1953 time period plays an important role, it may make sense to estimate preferences for specific issue of interest such as labor law based on a subset of data.
Analyses using Bridge Ideal Points
- Bailey, Michael A., and Forrest Maltzman. 2011. The Constrained Court: Law, Politics and the Decisions Justices Make. Princeton University Press.
- Cottrell, David, Charles R. Shipan, and Richard J. Anderson. 2019. “The Power to Appoint: Presidential Nominations and Change on the Supreme Court.” The Journal of Politics 81 (3): 1057–1068.