Article
What a roll-call score can tell you
How Voteview’s DW-NOMINATE score places the 119th Congress, and the questions a single number cannot answer.
A roll-call score is a summary of recorded votes. Voteview’s DW-NOMINATE score is the summary Uncanted uses for seated members of the 119th Congress. The first dimension runs, in Voteview’s own terms, from liberal to conservative. Uncanted draws it left to right and multiplies by one hundred, so a score of −0.45 becomes economics −45 on the chart. The second dimension is drawn on the authority axis the same way. The snapshot is October 2026. Members who were not seated then are not in the file. Vacancies are not given ghosts.
The first dimension is good at one thing. It says where a member sits relative to the rest of the chamber on the votes that actually divide the chamber. In this Congress that division still lines up with party. Every Democrat’s first-dimension score is left of every Republican’s. The chart does not hide that. It also does not treat it as the end of the question. Inside the Democratic caucus, and inside the Republican conference, the scores spread. A member can be at the edge of their party or in the middle of it. The member page gives the rank from the left and the distance from the party median so that spread is a number, not a vibe.
The second dimension is weaker, and the pages say so. Voteview’s second dimension is whatever is left after the first dimension has explained the main partisan split. In some eras that leftover had a clear story. In this one it is not a clean measure of borders, speech, policing, faith, or emergency power. Uncanted still plots it. Inventing a prettier social score from news articles would be worse than using a number the public file already computed and labeling the mismatch. A reader who wants a pure authority score will not find one in a roll-call model that was not built to answer that question.
What the score can tell you is comparative. Susan Collins sitting at the moderate edge of the Republican file is a fact about her votes relative to the conference, not a headline. Bernie Sanders sitting at the left edge of the Senate file is the same kind of fact. The nearest three names on a member page are the members with the smallest distance on the two scores. They are useful because they stop a reader from guessing. They are dangerous if the reader decides the three of them are a faction. Neighbors on a scale are not co-sponsors.
What the score cannot tell you is the content of a single bill. A member can have a conservative first dimension and still break with the party on one war, one subsidy, or one procedural vote that never moves the score. NOMINATE is built to fit thousands of votes. One vote is a rounding error inside it. The member pages refuse to list “key votes” the file does not contain. A list of famous bills, copied from memory or from a news recap, would look helpful and would be unverifiable from the source this chart claims to use.
Estimated scores are the honest exception. Some members have not cast enough votes for a stable personal file. The chart marks those scores as estimated, and the written page says to read them as provisional. An estimate is a placeholder so the seat is not missing. It is not a finding about the member’s beliefs. When a real score exists, it should replace the estimate. Until then the page would rather say “provisional” than borrow a paragraph from a campaign website and call it a record.
Party color is drawn on top of the score, not instead of it. The color answers “what do they call themselves?” The score answers “how have they voted with the chamber?” In this snapshot those answers mostly agree, which is itself a fact about 2026 and not a law of politics. Other decades on the presidential and former-member charts show the disagreement clearly. The Congress chart should be allowed to show agreement when the votes agree. Forcing a crossover that is not in the file would be a picture of a wish.
Check the score the way you would check any other secondary source. The bioguide on each member page is the door to congress.gov. Voteview publishes the model. Uncanted publishes the snapshot date, the scale, the rank, and the refusal to invent votes. A roll-call score is a good instrument for “who votes with whom.” It is a bad instrument for “who is a good senator.” The chart only claims the first.