EXPERIMENT · CENSUS ONLY

Only the census

What if all you knew about each district was who lives there? No polls, no markets, no past votes in the district: income, schooling, age, family and origin from the Census Bureau, and nothing else.

93.6%of House seats called right, 2016–2024

22census variables per district

0polls

How it works How it did in 2016–2024 →

What the census says about November 3

239Democratic House seats, median

75%chance of a Democratic House majority

80% of simulations between 200 and 283 seats. Senate: 50 Democratic seats (45–55), 47% chance of a majority.

01 House

The House, district by district

Every district on the November maps, scored from its census profile alone, plus a national mood taken from history: the 2024 House popular vote moved by the average midterm swing against the president's party.

Each bar is a number of Democratic seats; its height, how often it came up in 20,000 simulations. The shaded band holds 80% of them.

239Democratic seats, median

75%chance of a Democratic majority

D+4.8assumed national vote (±7.4)

Why the national mood is so uncertain, and where it comes from: the national mood →

Loading district boundaries…

Texas · new 2026 lines

TX-09

CensusD+6.9

Bachelor's degree or higher14.1%
White adults without a degree22.0%
White (non-Hispanic)23.1%
Black8.8%
Hispanic64.4%
Married adults49.1%
Median age33.7 years
Commute by public transit0.9%
People per square mile512
Median household income$67,864

this district average district bar: lowest to highest districtOpen the race →

What pushes it, in the simplified model

White (non-Hispanic)D+20.5
Bachelor's degree or higherR+17.7
Married adultsD+2.9
HispanicR+1.4
BlackR+0.8
Commute by public transitR+0.6
People per square mileD+0.4
Median ageD+0.3

02 The Senate

The same census, state by state

A state is the sum of its districts, so the Senate needs no new model: each state's margin is the average of the census margins of its districts on the November maps. The error per state is not assumed, it is measured on every Senate race since 2014, and it comes out larger than for a House district.

Each bar is a number of Democratic seats; its height, how often it came up in 20,000 simulations. The shaded band holds 80% of them. A 50–50 Senate is Republican: the Vice President breaks ties.

50Democratic Senate seats, median (45–55)

47%chance of a Democratic majority (51 seats)

±10.9points of error per state, against ±8.0 for a House district

Not on the ballot: 34 Democratic and 31 Republican seats. Same national mood as the House (D+4.8), drawn once per simulation for every state.

MAD+40.5100%
RID+29.599%
ILD+23.796%
NJD+20.994%
DED+19.593%
COD+18.992%
NMD+18.792%
ORD+18.592%
NHD+17.891%
MED+17.290%
VAD+14.286%
MND+10.278%
MID+7.271%
GAD+6.268%
AKD+0.953%
NCD+0.552%
FL^*R+0.648%
TXR+1.645%
OH^*R+1.944%
SCR+10.222%
IAR+10.621%
KSR+11.719%
LAR+13.316%
MTR+13.715%
MSR+15.911%
NER+18.78%
ALR+19.17%
TNR+23.04%
KYR+26.92%
IDR+29.71%
SDR+30.51%
OKR+32.31%
ARR+32.71%
WVR+35.10%
WYR+38.70%

Census margin and Democratic win chance in each state with a 2026 Senate race. * Special election (Florida, Ohio). Margins beyond ±40 points are capped.

Demographics set the board. Campaigns play on the edges.

Compare with the polls See MP-26, the poll-based forecast →

NextHow it works