Every Court, Left to Right

Fourteen benches, 1801 to 2024, each justice measured against the colleagues who decided the same cases — and the seven Courts where that produces no order at all.

The short answer

For most of the Court’s history you cannot rank its justices at all. Not because the records are missing — every vote is here — but because on the Marshall, Taney, Chase, Waite, Fuller, White and Taft Courts, running from 1801 to 1929, not one justice is distinguishable from the justice standing next to him in the order. Across 81 such comparisons, zero clear the measurement error. An order first appears at the Hughes Court and never goes away.

What is being measured, and what it cannot do

Every figure here is a gap: a justice’s liberal-vote share minus the share for the other justices deciding the same cases. Same docket, same bench, same day. That is the one comparison that survives two centuries of a changing Court, because it never reaches outside the room.

What it cannot do is travel between rooms. A gap can only be as large as the Court’s willingness to divide allows: if eight justices in ten join every opinion, nobody can sit far from the rest. That is not a subtlety, it is the largest effect in the data. The Fuller Court’s justices have a spread of 0.6 points; the Rehnquist Court’s, 16.1. Ranking them in one table would say more about 1888 than about anybody on the bench.

So each Court is ranked only against itself — and where even that produces nothing readable, the page says so instead of printing a list.

Spread of the bench
Decided unanimously
Marshall1801
1.0 pts
93.0%
Taney1837
0.8 pts
82.8%
Chase1864
0.9 pts
87.1%
Waite1873
0.7 pts
90.5%
Fuller1888
0.6 pts
84.0%
White1910
1.5 pts
86.5%
Taft1921
1.5 pts
87.3%
Hughes1929
3.8 pts
80.9%
Stone1941
6.7 pts
51.6%
Vinson1946
14.4 pts
31.4%
Warren1953
13.1 pts
36.1%
Burger1969
15.9 pts
37.2%
Rehnquist1986
16.1 pts
41.7%
Roberts2005
13.4 pts
45.4%

Left: the standard deviation of the justices’ gaps within each Court — how far apart they sat. Right: the share of that Court’s decisions with no dissent at all. The two are the same fact seen twice.

How the bars work

Every bar below is drawn on the same scale, −35 to +35 points, so the Courts can be compared with one another even where their justices cannot. The bar is the gap; the bracket around it is twice the sampling error in that one figure, estimated by resampling that justice’s cases. Where the bracket swallows the bar, that figure is not telling you much, and it is drawn that way on purpose.

One caution about reading the brackets against each other: two justices are counted here as being in a readable order when the difference between their gaps exceeds twice their combined error, which is a lower bar than requiring their brackets not to touch. So a pair whose brackets overlap slightly may still be genuinely ordered.

Where a Court does print a numbered order, it also prints how many of that order’s steps clear that test — and on every one of them it is a minority. The numerals are a reading aid for the shape of a bench, not a claim that each place in the list is separately earned. The Courts with no numerals are the ones where not a single step clears it.

The fourteen Courts

The Marshall Court

OT1801–OT1836

1,280 cases · 93.0% unanimous · spread 1.0 pts · 9 of 16 justices listed

No order can be read from this Court. Sort these 9 justices by their gap and you get 8 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 10 of its 36 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The Taney Court

OT1837–OT1863

1,670 cases · 82.8% unanimous · spread 0.8 pts · 13 of 20 justices listed

No order can be read from this Court. Sort these 13 justices by their gap and you get 12 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 2 of its 78 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The Chase Court

OT1864–OT1873

1,371 cases · 87.1% unanimous · spread 0.9 pts · 11 of 14 justices listed

No order can be read from this Court. Sort these 11 justices by their gap and you get 10 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 8 of its 55 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The Waite Court

OT1873–OT1888

3,650 cases · 90.5% unanimous · spread 0.7 pts · 14 of 15 justices listed

No order can be read from this Court. Sort these 14 justices by their gap and you get 13 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 19 of its 91 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The Fuller Court

OT1888–OT1910

4,986 cases · 84.0% unanimous · spread 0.6 pts · 18 of 21 justices listed

No order can be read from this Court. Sort these 18 justices by their gap and you get 17 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 29 of its 153 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The White Court

OT1910–OT1920

2,547 cases · 86.5% unanimous · spread 1.5 pts · 12 of 13 justices listed

No order can be read from this Court. Sort these 12 justices by their gap and you get 11 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 34 of its 66 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The Taft Court

OT1921–OT1929

1,618 cases · 87.3% unanimous · spread 1.5 pts · 11 of 13 justices listed

No order can be read from this Court. Sort these 11 justices by their gap and you get 10 steps from each one to the next — and not a single one of those steps is larger than twice the combined error in the two figures it separates. Not one. There is no sequence here to print. That is not to say the bench was uniform: 30 of its 55 possible pairings do differ by more than twice their combined error. Those differences simply never line up into a sequence. The justices below are therefore listed by seniority, and the bars are drawn only so this Court’s scale can be compared with the ones further down.

−35level with this Court+35

The Hughes Court

OT1929–OT1940

1,895 cases · 80.9% unanimous · spread 3.8 pts · 15 of 16 justices listed

3 of the 14 steps in this order are larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

The Stone Court

OT1941–OT1945

781 cases · 51.6% unanimous · spread 6.7 pts · 9 of 11 justices listed

2 of the 8 steps in this order are larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

The Vinson Court

OT1946–OT1952

812 cases · 31.4% unanimous · spread 14.4 pts · 11 of 11 justices listed

1 of the 10 steps in this order is larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

The Warren Court

OT1953–OT1968

2,205 cases · 36.1% unanimous · spread 13.1 pts · 16 of 17 justices listed

5 of the 15 steps in this order are larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

The Burger Court

OT1969–OT1985

2,809 cases · 37.2% unanimous · spread 15.9 pts · 13 of 13 justices listed

5 of the 12 steps in this order are larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

The Rehnquist Court

OT1986–OT2004

2,044 cases · 41.7% unanimous · spread 16.1 pts · 13 of 14 justices listed

4 of the 12 steps in this order are larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

The Roberts Court

OT2005–OT2024

1,471 cases · 45.4% unanimous · spread 13.4 pts · 14 of 16 justices listed

2 of the 13 steps in this order are larger than twice the combined error in the two figures it separates; the rest fall inside it. Read the ends as real and the middle as a cluster.

−35level with this Court+35

Method, and its limits

Built from the Supreme Court Database — both releases, the legacy file covering the 1791 through 1945 terms and the modern one from 1946. One vote per justice per case. Cases the database declines to code ideologically are excluded. A Court needs 150 decisions to appear, and a justice 200 coded cases under that Chief to be listed in it, which is why a justice can be missing from a Court he briefly sat on.

The error bars are a case-level bootstrap. A justice’s coded cases are resampled with replacement and both his share and his colleagues’ share recomputed on the resampled set, four hundred times; the bracket is twice the standard deviation of the result. Resampling both together matters, because the two figures come from the same cases and move with each other — an error bar that assumed otherwise would be too narrow, and would let this page claim orderings it has not earned.

The coding is contested, and it is not ours. “Liberal” is the database’s term of art, defined issue by issue: pro-union in labor cases, anti-business in economic cases, pro-federal-power in federalism cases, pro-United-States in tax ones. The full definitions are on the codebook page. The database also says plainly that its modern issue categories were applied to eighteenth- and nineteenth-century decisions after the fact, and its direction coding is thinner the further back you go — a little over half the records in the 1800–1849 terms carry any direction at all. Those cases simply drop out, for every justice equally.

What is not here. The Jay, Rutledge and Ellsworth Courts decided 63 cases between them and are below the floor entirely. Eight justices have fewer than 50 ideologically coded votes in the whole database and have no figures anywhere on this site: Jay, J. Rutledge, Wilson, Blair, Iredell, T. Johnson, Ellsworth and Moore. Every one of them left the Court before 1805. And a justice who sat under two Chiefs appears under both, with separate figures, because they are separate benches: 179 entries below describe 100 people.

The Courts are the database’s, not a calendar’s. Each case is assigned to a Chief Justice by the database, and the term years shown here are October Terms, so a Court’s span will not always match the dates of its Chief’s commission. The first Court below is a good example: Marshall died in July 1835, and the era pages elsewhere on this site date his Court from his service, but the database keeps assigning cases to him through OT1836.

A last caution about the word “unanimous.” It means here that no participating justice was recorded in dissent. Before the 1940s the Court often disposed of cases without any opinion at all, and silence is not the same as agreement. That is precisely why this page treats the early Courts’ flatness as a fact about the institution rather than a fact about the judges.

Questions

Can you rank every Supreme Court justice from most liberal to most conservative?

Not in one list, and the reason is measurable. Within each Court a justice can be compared to the colleagues who decided the same cases. Across Courts the comparison breaks, because the size of any such difference depends on how often that Court divided at all — 93.0% of the Marshall Court's decisions carried no dissent, against 45.4% of the Roberts Court's. On seven of these fourteen Courts, no two justices standing next to each other in the order are even distinguishable from one another.

Which Supreme Court had the widest ideological range?

The Rehnquist Court, where the justices' gaps have a standard deviation of 16.1 points, about 25 times the Fuller Court's 0.6. Its widest pair, T. Marshall and Thomas, sit 46.3 points apart.

When did the Supreme Court start dividing along ideological lines?

A readable order first appears at the Hughes Court (1929–1940) — the earliest on which any two justices adjacent in the order can be told apart at all. But there is no single moment when the Court came apart. The narrowest bench in the whole series is Fuller's, at 0.6 points, and the widest is Rehnquist's at 16.1, about 25 times as wide — yet no Court's bench is more than 2.6 times as spread out as the one before it. It is a climb across the first half of the twentieth century, not a cliff. The institutional change legal historians date to 1940–41, when the Court's practice of publishing disagreement changed, sits inside that climb rather than at its start. What none of this can say is whether the justices before then agreed more, or simply said so less.

The 1946-onward ranking, where one table is defensible, is at The Court from Left to Right. Who sat with whom, in the same years, is at Who Voted With Whom, and who moved is at Who Changed on the Court.