- Category
- Civic data and legislative transparency
- Audience
- Voters, journalists, students, and researchers
- Coverage
- 12,249 members, 1789 to present
Congress already publishes everything. Nobody can read it.
Congress Atlas takes the records Congress keeps on itself and turns them into plain answers: how divided the chambers are, whether anything is passing, who is getting things done, and how the public has felt about it since 1974.
The raw material is genuinely public and genuinely unreadable. It arrives as bulk XML, fixed-width vote matrices, and codes like [R-TX-11], so the most complete legislative record in the world functions in practice as a closed one. Four free sources feed the pipeline — Voteview, GovInfo BILLSTATUS, congress-legislators, and Gallup — and the heaviest is a 700 MB matrix of 26.4 million individual votes, collapsed into per-vote party splits and per-member records so two centuries of behaviour can be queried instantly.
The site answers five questions directly and shows its working on each. The measures are symmetric, treating both parties identically, and wherever a number reflects a judgement rather than a fact in the record it says so and displays the arithmetic. It is not a scorecard, not a prediction engine, and it does not tell anyone how to vote.
The most-cited finding is not the one people expect. The party averages did move apart, but the larger change is that the members in between stopped existing — zero House members now sit in territory the other party also occupies, against 203 in 1969. The Senate finding is almost as stark and far less discussed: it now takes 19 votes a Congress on passing a bill and 219 on confirming appointments, where in 1989 those numbers were 83 and 24.
Why the most documented legislature on earth is illegible
The format is the barrier
Bulk XML, fixed-width vote matrices, and internal identifier codes are public in the strictest sense and inaccessible in every practical one. Publication without legibility is not transparency.
'More polarised than ever' is asserted, never evidenced
The claim is made constantly and almost never accompanied by a number, a scale, or a source — which leaves it arguable as a vibe when it is in fact measurable, and measurably true.
Raw law counts flatter the institution
Roughly 30% of the laws passed in the last completed Congress renamed a post office, awarded a medal, or declared a day of recognition. Counting them alongside substantive legislation makes Congress look considerably busier than it is.
Filing is free, so success rates lie
A member can introduce as many bills as they like, and one filed 606 nearly identical bills in a single Congress. Any metric shaped like bills-passed-over-bills-filed falls for reasons that have nothing to do with gridlock.
The Senate's job changed without an announcement
The chamber has largely stopped voting on bills and started spending its floor time on confirmations and cloture. The shift is visible in the record and almost entirely absent from public understanding of what the Senate does.
'Do something' is not an actionable demand
Diffuse frustration cannot be acted on by a constituent or answered by a member. A specific figure about a specific representative can be.
Analysis usually arrives as advocacy
Most tools in this space are built to make one side look worse. A measure that is not symmetric between the parties cannot be trusted by the half of the country it indicts.
How we built it
Take only free, public, primary sources
Four of them: Voteview for every recorded vote since 1789 and the ideology scores derived from them, GovInfo BILLSTATUS for bills and sponsorship, congress-legislators for biography and committee rosters, and Gallup for fifty years of the identical approval question. Nothing scraped from behind a paywall.
Reduce 26.4 million votes into something queryable
The heaviest input is a 700 MB matrix covering every member on every roll call since 1789. The pipeline collapses it into per-vote party splits and per-member records, so a page that reads two centuries of behaviour still returns instantly.
Lead with the answer, then show the arithmetic
Each question opens with a one-line answer in plain English — 'Yes, and this is the widest gap ever recorded' — followed by the figures, the scale they sit on, and the source line that produced them.
Separate substantive from ceremonial everywhere
Post office renamings, medals, and commemorative days are counted, labelled, and kept apart from substantive legislation in every chart and table, rather than being quietly folded into a headline number.
Rank by distance travelled, not volume filed
Member scores reflect how far a member's bills move through committee, chamber, and signature — because filing costs nothing, and a leaderboard built on filings rewards the member who files 606 near-identical bills.
Declare the gap in the data, repeatedly
Complete machine-readable bill records begin in 2013. The site says so on every page that depends on them and refuses to compare today's output with the 1960s, even though the comparison would be more dramatic.
Publish reforms with their strongest counterargument
Six proposals from the House modernisation committee, the Bipartisan Policy Center, Brookings and AEI, each tied to a specific figure on the site, each paired with the best case against it and a rating for how contested it is. A list of reforms without its counterarguments is advocacy dressed as analysis.
End every section with a specific ask
Rather than a call to action, each reform closes with a question a constituent can actually put to their member — whether they have signed a discharge petition, and what their standard is for doing so.
Capabilities
Five plain answers
An overview that answers polarization, output, individual productivity, partisan share of laws, and public approval in one screen each, with the source under every figure.
Member lookup across two centuries
Search any of the 12,249 people who have ever served and see how they voted, how often they broke with their own party, what they wrote, and which of their bills became law.
Chamber leaderboards
House and Senate rankings by score, bills that became law, bills written, or support attracted from the other party — filterable by Congress back to the 113th.
The House and the Senate as separate institutions
Each chamber gets its own page, because the thing that changed in the Senate — floor time moving from legislation to confirmations — has no House equivalent.
Two centuries of division on one scale
Every member placed by their own voting record, with a slider that moves through history and shows the shared ground between the parties opening and then closing to nothing.
The legislative funnel
Where bills die, stage by stage: filed, survived committee, passed one chamber, passed both, signed. In the 119th, 15,924 filed and 104 signed.
Substantive versus commemorative output
Laws per Congress split into substantive and ceremonial, so the raw count cannot be used to make a session look more productive than it was.
Senate floor-time analysis
Votes on passing bills, votes on confirming nominations, and cloture motions tracked against their 1989 baselines — 19 against 83, 219 against 24, 284 against 24.
Fifty years of public approval
Every year of Gallup's identical question since 1974, by year and by decade, with the 2001 spike and the post-2005 slide visible on one line.
Subject-area breakdown
Which policy areas absorb the most bills and which of them actually produce law — 1,800 health bills filed in the 119th, one enacted.
Contested reform proposals
Six outside proposals, each tied to a figure on the site, each with the strongest case against it and an explicit rating of how contested it is.
Traceable arithmetic throughout
Every derived measure is shown as arithmetic a reader can check, and every source line names the record or poll the figure came from.
The visual language
Congress Atlas is designed as a reference document, not a dashboard. An Instrument Serif display sets the questions, IBM Plex Sans carries the answers at reading length, and IBM Plex Mono is reserved for source lines and structural labels — so a reader can tell at a glance which text is a claim and which is a citation. The ground is warm paper; the one accent is brass.
Colors
Record Ink
#171D24
Primary type, the hero field, and chart marks
Paper
#F7F5F0
Base field, chosen warm so long tables stay readable
Brass
#96731F
The single accent — source lines, active nav, primary action
Slate
#38454B
Secondary copy and table values
Muted Slate
#5C6668
Captions, metadata, and column headers
Rule
#D6D3CA
Table rules, card borders, and dividers
Panel
#E8E5DD
Raised surfaces and alternating table rows
Typography
Instrument Serif
Sets the questions and section headings. A newspaper serif rather than a product face, because every page is framed as a question a reader came in with.
IBM Plex Sans
Carries the answers, the explanatory prose, and table content. Chosen for a large x-height that survives dense tabular layout.
IBM Plex Mono
Reserved for source lines, eyebrows, and structural labels. The monospace register is the visual tell that a line is provenance rather than argument.
Principles
- 01
Symmetric or nothing
Every measure treats both parties identically. A tool that indicts one side is only usable by the half of the country that already agreed.
- 02
Show the arithmetic
Where a number reflects a judgement rather than a fact in the record, the site says so and displays the calculation line by line.
- 03
Declare the gap
Complete bill records begin in 2013, so no comparison reaches further back than that — stated on every page that depends on it, including the ones where reaching back would be more striking.
- 04
Never let ceremony inflate the count
Substantive and commemorative laws are separated everywhere, because the combined figure is the one most often quoted and the least informative.
- 05
Specific beats loud
'Do something' cannot be acted on. 'Your representative wrote 125 bills and 3 became law' can, which is the whole point of making the record legible.
- 06
Publish the counterargument
Every reform proposal ships with its strongest objection and a rating of how contested it is, and listing one is explicitly not an endorsement.
Questions
A public-record explainer that turns Congress's own data — every recorded vote since 1789, every bill since 2013, and fifty years of Gallup polling — into plain answers about how divided the chambers are, what is passing, who is getting things done, and how the public feels about it.
Four free public sources, none of them the project's own: Voteview for roll-call votes and the ideology scores derived from them, GovInfo BILLSTATUS from the Government Publishing Office for bills and sponsorship, the open-source congress-legislators dataset for biography and committee rosters, and Gallup for congressional approval since 1974.
The measures are symmetric — they treat both parties identically — and wherever a figure reflects a judgement rather than a fact in the record, the site says so and shows the arithmetic. It is not a scorecard, not a prediction engine, and it does not tell anyone how to vote.
Each member is placed on a scale from −1 to +1 derived from how they actually vote. 0.91 is the distance between the average House Democrat at -0.39 and the average Republican at 0.52, on a scale where 2.0 is the maximum possible. It is the widest gap ever recorded.
Complete machine-readable bill records begin with the 113th Congress. Voting records reach back to 1789 only because academics rebuilt them. The site refuses to compare modern legislative output against the 1960s because the underlying records do not support it.
Commemorative laws rename post offices, award medals, or declare days of recognition. They made up 81 of the 274 laws passed in the 118th Congress — about 30%. Congress Atlas counts them separately everywhere so the raw total cannot be used to overstate output.
By how far their bills travel through the process — committee, chamber, signature — rather than by how many they file. Filing costs nothing, and one member filed 606 nearly identical bills in a single Congress, so a volume-based ranking measures the wrong thing.
No. They come from the House Select Committee on the Modernization of Congress, the Bipartisan Policy Center, Brookings and AEI, and each one is published alongside its strongest counterargument and a rating of how contested it is. Listing a proposal is explicitly not an endorsement of it.
Interested in Congress Atlas?
Tell us how you would like to be involved — investing, building, advising, or simply following along — and we will be in touch.
Register Your Interest
Work with Vital Ventures
We study, support, and build the foundational systems that shape how societies function. If you are working on something in that direction, we would like to hear about it.
