Who represents us

Since I’ve been alive, our government has felt static and immovable. The same parties, platforms, and personalities have dominated the national stage for decades and resisted younger voices and new ideas. Our population and demographics have evolved dramatically over the past half century, and yet the composition and demographics of our leadership remains frozen in time.

Below is a visualization of both chambers,1 from 1789 through today.2 Compare each Congress with the country it represented by age,3 gender, race,4 or party.5 This is a companion infographic to complement my post on why representative democracy is worth fighting for, even if we’ve never had one.

1789–1791

1st Congress

1:11K ratio of representatives to voters

House
Senate

    Who counts

    The US population view counts everyone: women, enslaved people, and everyone else the founders’ infamous affordances left out, the same way I counted state populations in the essay.6 The who could vote view narrows it to the people the law let vote at the time: white men at the founding, Black men after the 15th Amendment in 1870, women after the 19th in 1920, Native Americans after 1924, and Asian immigrants once the bar on naturalization fell in 1952.7 Either way, it tells a similar story.

    I haven’t tried to subtract everyone who was eligible on paper and kept from the ballot anyway: by poll taxes, literacy tests, and outright terror, most of all under Jim Crow until the Voting Rights Act of 1965. That suppression isn’t a caveat to the data, it’s part of the point. A Congress that looks nothing like the people it governs is a function of the system working as designed.

    The number in the corner

    The big number is how many people each seat in the House speaks for: everyone, or only those who could vote, depending on what you’re comparing with. In 1790, each of the 65 members of the 1st US Congress represented around 60,000 constituents on average, of whom only about 11,000 could vote for them.8 The House kept growing with the country until 1913, when it stopped at 435 seats, and it hasn’t grown since. Today each representative speaks for more than three-quarters of a million people. It’s impossible for one human to sufficiently represent the interests of that many people, and watching that number climb is the clearest argument I know for expanding the House.

    1. Each dot is a voting seat as its Congress convened, held by its first occupant, so a member who filled a vacancy later isn’t drawn and a seat counts once. Who held which seat comes from Voteview through the 100th Congress (1987–1989), the political-science standard for historical rosters, and from the congress-legislators project after that, which also supplies every member’s birthday, gender, and party. Non-voting delegates from DC, Puerto Rico, and the territories are left out, a reminder that more than four million Americans have no vote in Congress at all. Party labels are grouped into families: “Republican” before 1830 means Jefferson’s party, not Lincoln’s. ↩︎

    2. The 120th Congress is drawn as elected, before it convenes on January 3, 2027, from Wikipedia’s list of its members as of November 9, 2026. Fifteen races were still uncalled then, including recounts in Florida’s 27th district and December runoffs in Louisiana’s 5th and 6th, so each uncalled seat is drawn for the candidate leading it, pulsing, with the leader’s name in italics. New members’ ages, genders, and races are from their Wikipedia articles, checked by hand. It will be replaced by the official roster once the new Congress is sworn in. ↩︎

    3. Each member’s age on the day their Congress began; a few early members’ birthdays are unknown. The country’s bar counts everyone, children too, for the US population, and only those old enough to vote (21, and 18 from 1971) for who could vote, from the Census Bureau’s population by age, 1900 to 2002 and Age and Sex Composition: 2020. The census didn’t publish ages in these groups before 1900. Median ages for 1820 to 1930 are from Historical Statistics of the United States, series A 143. From the 93rd Congress (1973) the age tab groups everyone by generation instead, using Pew Research Center’s birth years, with the country’s generations estimated from its age groups. ↩︎

    4. Neither dataset records race, so it comes from the House historian’s lists of Black, Hispanic, and Asian and Pacific Islander members, and Wikipedia’s list of Native Americans in Congress. Middle Eastern and North African members, whom the census counts as white, are counted separately here, from the ancestry Wikipedia records for them: Lebanese, Syrian, Palestinian, Iranian, Egyptian, Assyrian, Israeli, and others, following the federal government’s 2024 definition, which leaves out Armenian ancestry. Everyone else is counted as white. A member in more than one group is counted as two or more races, unless one of them is Hispanic, which is counted first, the way the census counts it. Each member’s race is recorded with its source and can be corrected by hand. ↩︎

    5. The line under each Congress’s name gives the group it most overrepresents and the one it most underrepresents, by percentage points of seats across both chambers, with each one’s share of the House and of the Senate, then how many representatives and senators would have to change for each chamber to match exactly: the index of dissimilarity between that chamber and the country, times its seats. Against who could vote, it compares with the electorate, voting age included. Against the US population, it compares race and gender with every American, and age with adults 25 and over, the youngest who can serve, since no Congress could close a gap made of children. In years the census folded a group into white (Hispanic Americans before 1970, and Middle Eastern and North African Americans before 1980), its members in Congress are counted the same way, so the comparison is like with like. ↩︎

    6. Population is interpolated between censuses: Gibson and Jung’s historical census statistics on population by race from 1790 to 1990, the 2000 and 2010 counts, the 2020 redistricting data, and the Census Bureau’s 2024 estimate after that. The share of women is from the Census Bureau’s population characteristics, 1900 to 2002, and before 1900 from Historical Statistics of the United States, series A 91. The census didn’t count Hispanic Americans until 1970 or people of more than one race until 2000, and it left out Native Americans living outside the settler population until 1890, so the early bars undercount all three. The census only counted Middle Eastern and North African Americans directly from 2020 (about 3.5 million); from 1980 to 2010 the page uses the people reporting Arab, Iranian, or Israeli ancestry (1990 and 2000, with 1980 and 2010 estimated from Arab ancestry alone), and before 1980, when the census first asked about ancestry, they’re within white. ↩︎

    7. These are the national milestones, so they leave out the states that moved first: women could vote in several western states before 1920, and free Black men in a few northern states before the Civil War. They also leave out the property requirements most states kept until the 1820s and 1830s, which narrowed the founding electorate further still. ↩︎

    8. Eligible voters are those of voting age in the groups the law let vote, assuming every group shares the country’s age structure. From 1900 the voting-age share comes from the census’s age groups; before that, when the census didn’t publish them, it’s estimated from the median age, using how the two moved together from 1900 to 1960. ↩︎

    Backlinks