DEAR DATA

How we made the 2026
If Only One Group Voted map

Last updated November 8, 2026

These maps show which party each group of voters favored in each state in the 2026 midterms. This soon after an election, exit polling is the only source for that. It gets thin when a group makes up a small share of a state's voters, so some of the colors carry a big margin of error.

We wanted a full map for every group on screen, and we wanted to show as many groups as we could without misleading anybody. That meant filling gaps with estimates. About 79% of the colors come directly from the poll and we estimated the other 21%. The grid further down rates all of them.

The data

The 2026 numbers come from The Voter Poll, which SSRS ran this year for AP, ABC, CBS, CNN, Fox, MS NOW and NBC. It reached about 100,000 voters, most of them by phone or online in the days before the election and about 25,000 at polling places in 15 states. State samples on our map go from 467 interviews in Hawaii up to 5,145 in Texas.

Every color on the map is a 2026 result or our estimate of one. The only older numbers that went into it come from AP VoteCast's 2024 survey, a separate poll, and they helped with 127 of the 188 estimates. Vote counts are AP's, as published by NBC News.

Picking a race for each state

All of a state's colors come from one race. We used the Senate race if the state had one. Ten states had a governor's race and no Senate race, and in Idaho, Nebraska and South Dakota the Republican's main opponent was an independent, so those 13 use the governor's race. When the poll's rounded numbers showed a tie between the Democrat and the Republican, we settled it with the unrounded ones.

Six states had no race we could use. Indiana, Missouri, North Dakota, Utah and Washington had no Senate or governor's race this year. Montana had no governor's race, and an independent finished well ahead of the Democrat in its Senate race. Every group in those six states is an estimate, based on the state's total US House vote.

Filling in the gaps

The poll won't publish a number for a group with fewer than 100 interviews in a state. In Wyoming it reached 23 Latino voters and no Black voters. Between that rule and the six states above, we had 188 colors to estimate.

A group's vote usually rises and falls with its state's overall result, so for each group we drew that line through the states the poll did report and read the missing states off it. If AP VoteCast had the group in that state in 2024, we averaged in that number, adjusted for how far the state and the group have moved since.

To test this we hid each of the 712 numbers the poll did publish, one at a time, and predicted it the same way. We got the party right 93% of the time and were off by about 6 points on average. Predictions with a clear lead went 429 for 429, and close ones were right 81% of the time. In the test's smallest groups, those under 250 interviews, we got the party right 90% of the time. We can't test groups too small to publish, and they probably do worse.

Reading the maps

These maps are a thought experiment, and they only count people who voted. Every state is a different race, which is how a Republican like Phil Scott can turn Vermont red on some of them. A color can't show the size of a lead either, so a state decided by 1 point looks like one decided by 50.

How sure we are

Each of the 900 colors (18 groups in 50 states) got one of three ratings.

High490 colors, 54%

The poll measured this group in this state and the lead is bigger than the margin of error. Black voters in Georgia are a good example, with 1,143 interviews and Democrats ahead by 87, give or take about 4.

Medium340 colors, 38%

Some of these the poll measured, but the lead falls inside the margin of error (222 colors). Democrats won men in Texas by 4, give or take about 6, even with 2,258 interviews. The other 118 are estimates with a wide lead. The poll reached no Black voters in Wyoming, but Black voters backed Democrats by about 70 points or more in all 21 states it reported.

Low70 colors, 8%

These are estimates with a small lead, given how far off our estimates can be, or estimates where our two methods pull in opposite directions. For women in Montana, one method said Democrats by 2 and the other said Republicans by 6. We'd call that color a coin flip.

The margins of error are our own math, because the poll's data doesn't include them at this level. We treated each group's sample as half its real size to allow for survey weighting. A stricter or looser allowance would put the High count between about 450 and 530.

Groups that aren't on the map

Black voters are reported in about as few states (21). They stayed in because they were so lopsided everywhere the poll found them that the missing colors aren't in much doubt. We skipped Democrats, Republicans, liberals and conservatives because each came out one color in every state reported.

Mistakes we've found

Every state and group

HighPolled, clear lead
MediumPolled but close, or estimated with a wide lead
LowEstimated, too close to call
MenWomenWhite menWhite womenBlack votersLatino voters18 to 2930 to 4445 to 6465 and olderSuburbanRuralWhite, degreeWhite, no degree$100K and upCatholicsProtestantsIndependentsALAKAZARCACOCTDEFLGAHIIDILINIAKSKYLAMEMDMAMIMNMSMOMTNENVNHNJNMNYNCNDOHOKORPARISCSDTNTXUTVTVAWAWVWIWY

Group by group

Every group is an estimate in the six states with no usable race (IN, MO, MT, ND, UT and WA), so those aren't repeated below.

GroupStates measuredAlso estimated in
Men44None
Women44None
White men43HI
White women43HI
Black voters21AK AR CO CT DE HI IA ID KS KY MA ME NE NH NM OK OR RI SD VT WI WV WY
Latino voters23AL AR DE HI ID KY LA MA MD ME MS OK OR RI SC SD TN VA VT WV WY
18 to 2929AL AR DE HI ID LA MA MS NM RI SC TN VT WV WY
30 to 4441DE HI WY
45 to 6444None
65 and older44None
Suburban voters41VT WV WY
Rural and small-town voters44None
White, college degree44None
White, no degree43HI
Income $100,000 or more44None
Catholics34AL AR HI ID MS OK TN VT WV WY
Protestants and other Christians43VT
Independents43SD

Sources