Sources

Everything on this site comes from public records and published surveys. This page lists each source, what we take from it, how often we check it, and where to find the original.

Voting records

House of Representatives

Source: the Office of the Clerk of the U.S. House, which publishes an official record of every roll-call vote: clerk.house.gov/Votes.

What we use: for each vote, the question, the result, the date, the bill, and how each representative voted (Yea, Nay, Present or Not Voting).

How often: checked every 2 hours. Each vote is checked before it is stored: every member's vote must add up to the Clerk's published totals.

Senate

Official source: the U.S. Senate's roll-call records: senate.gov.

Source we use for now: Voteview, a long-running research project at UCLA that republishes every Senate and House roll call. The Senate's website doesn't currently allow automated downloads from our server, so we load Senate votes from Voteview instead. Voteview can be days to weeks behind the official record; the footer of every page shows how current each chamber is.

How often: checked daily. If we load a vote from the Senate's own record later, it replaces the Voteview copy.

Paired and announced votes: a senator who is absent can "pair" with one on the other side or announce a position. These aren't counted in the official tally, so we treat them as not voting, as the Senate does.

Members of Congress

Source: congress-legislators, a public dataset of every member's terms, seats, parties and official IDs, maintained by volunteers and widely used by news organizations and researchers.

How often: checked daily, so new members appear before their first vote.

How we use it: every vote is matched to the seat the member held on the day of the vote, so redrawn districts, special elections and moves between the House and Senate are handled correctly.

Bills

Bill names and links go to Congress.gov, the official site for federal legislation, run by the Library of Congress.

What we use: for each bill with a recorded vote, its official title, its policy area and the subject tags the Library of Congress assigns (for example, "Immigration status and procedures"), from the Congress.gov API. Search uses them, so a search for a topic finds bills whose names don't mention it.

How often: checked daily for bills with new votes, and every bill again each month, since subject tags are added over time.

Surveys of what people want

We compare votes with what people said in independent, published surveys. We use more than one survey organization, and we show each one separately, never combined into a single score, so you can see where they agree.

We only compare a vote with a survey question when both are about the same proposal. See Methods for how votes and questions are matched.

Cooperative Election Study

Who runs it: a consortium of universities led by Harvard University and Tufts University, fielded by YouGov. cces.gov.harvard.edu

What it is: one of the largest academic surveys of U.S. adults, run every year since 2006. Each year it asks whether people support specific actions Congress has voted on.

What we use: each person's state, House district and answers, with the survey's own weights. House districts have only a few dozen respondents each, so we estimate each district's opinion with help from statewide answers; see Methods.

Data: published openly on Harvard Dataverse.

Loaded on this site:

Program for Public Consultation

Who runs it: the Program for Public Consultation at the University of Maryland's School of Public Policy. publicconsultation.org

What it is: surveys in which people are given brief background on a proposal before saying whether they favor or oppose it. Its 2026 "Common Ground in the Battlegrounds" series surveyed 11 competitive states and 28 competitive House districts, with samples of several hundred to 2,000 people in each. Respondents come from online panels and are weighted to Census figures.

What we use: the published results for each state and district, taken from the questionnaires on vop.org, with each sample's size and margin of error. Only members from the states and districts it surveyed get these comparisons.

Loaded on this site:

Democracy Fund + UCLA Nationscape

Who runs it: the Democracy Fund Voter Study Group and UCLA. voterstudygroup.org/data/nationscape

What it is: a survey of about 6,250 people a week from July 2019 to February 2021, nearly half a million interviews in all, conducted online and weighted to Census figures. Respondents said whether they agree or disagree with specific policy proposals.

What we use: the answers to the proposals Congress actually voted on, from the published data files, pooled across the weeks each question was asked. Its size allows an estimate for nearly every congressional district. Districts are as drawn for the 116th Congress (2019-2020), which the 117th also used everywhere except North Carolina, so North Carolina's representatives are compared only on votes from 2019-2020.

Loaded on this site:

State ballot measures

What they are: questions a state's voters decided themselves on the ballot, such as a constitutional amendment. These aren't surveys: they are certified counts of votes actually cast, so there is no sample and no margin of error.

What we use: each measure's certified statewide Yes and No totals, as compiled by Ballotpedia, with a link to each measure's own page. A statewide result is compared only with members who represent the whole state: senators, and the representative of a state with a single House seat. Representatives of states with several districts aren't compared, because a statewide result doesn't say what their own district voted.

How they're shown: no ballot measure asks exactly what a bill in Congress does, so every comparison is marked as a related vote and none is counted in a member's score. For example, several states amended their constitutions in 2024 to say only U.S. citizens may vote; the SAVE America Act in Congress would also require documents proving citizenship to register. Measures with no matching vote in Congress (for example on abortion or the minimum wage) are left out.

Loaded on this site:

Population

Source: the Census Bureau's American Community Survey, 1-year estimates of total population for every state and House district, from api.census.gov.

How we use it: to show how many people each member represents, and how many of them a survey reached. Each year's figures are matched to the district lines in use at the time.

How often: a new year's figures are released each September and loaded automatically.

Considered and not used

• America's Political Pulse (Polarization Research Lab): a well-run weekly survey, but it asks about attitudes, not specific bills, so its answers can't be compared with votes.

Updates and corrections

Every downloaded file is kept unchanged, so any figure can be rebuilt and checked, and the code that does it is public. If you find an error, please tell us at corrections (at) uscongressvsvoters.org.