How the Voting Figures Are Made
The rules behind every Voting Record panel on this site — what the labels mean, what is counted, and where the data comes from.
Every figure comes from the Supreme Court Database — the standard scholarly record of the Court’s decisions — using its classification of each vote, its case counts, and nothing invented here; where its categories stop meaning what they say, this site stops using them.
What “liberal” and “conservative” mean here
They are the Supreme Court Database’s terms of art, not ours. The database defines them issue by issue: liberal means pro-union in labor cases, anti-business in economic cases, pro-federal-power in federalism cases, and — in its own odd but deliberate usage — pro-United-States in tax cases. Criminal procedure, civil rights, First Amendment, due process, privacy and attorneys share a single combined definition running to seventeen clauses, so the label carries less precision in those areas than a tidy percentage suggests. The full definitions are on the database’s codebook page.
The coding is Harold Spaeth’s and his successors’, and reasonable scholars contest it: a vote can be recorded as conservative for reasons that have nothing to do with a justice’s politics. This site reports the database’s judgment; it does not add a judgment of its own on top.
What is counted
One vote per justice per case. The database publishes a justice-centered release organised exactly that way, so no case has to be resolved by a rule of this site’s own. Cases the database declines to code ideologically are excluded, as are votes in an equally divided Court.
The gold tick on each bar — and the “contemporaries” figure beside each headline number — is the rate for the other justices sitting on the same cases, not for the era in general. That baseline is what makes a dissent rate readable across two centuries: dissenting was rare practice before about 1941 and routine after, so a bare rate would mean different things depending only on when a justice sat.
A subject bar is drawn only where a justice sat on at least 30 cases in that area, so a short tenure shows fewer bars. Interstate relations, private action and the database’s miscellaneous category are left out: it codes almost nothing in them. And for the justices of the eighteenth and nineteenth centuries the subject bars are withheld entirely — the database’s categories were designed for the modern Court and, by its own account, fitted to the early decisions as best it could afterwards. The overall figures rest only on whether each vote was coded liberal or conservative, which is the same judgment in every era. The subject bars would not be.
The two releases, and the 1946 join
The database comes in two files: a legacy release running from the 1791 term through 1945, and a modern one beginning in 1946. For the eight justices whose service crosses the join, the figures pool both files, so they describe the whole tenure rather than the part that happened to fall after 1946 — those profiles say so.
Rankings are different. “Most to least liberal” is computed on the 1946-onward record only, because the justices being ranked against have no record before then. A justice who left the Court before 1946 is not “too short” to rank — he is outside the period, and his profile says that instead.
Source
The Supreme Court Database, justice-centered: 2025 Release 01 for the 1946 term onward, and Legacy Release 07 for the 1791 through 1945 terms. Harold J. Spaeth, Lee Epstein, Michael J. Nelson, Andrew D. Martin, et al. The same database, under the same rules, powers the career rankings, the drift series, the agreement matrices and the bench-by-bench orderings; each of those pages carries the notes specific to its own computation.
Dates of birth, death and service are checked against the Federal Judicial Center’s biographical directory; career data on Stepping Stones comes from the Epstein Justices Database; the reading shelves cite their sources work by work in the bibliography.