Federal Elections AnalysisElection analysis
Features / Polling research
Which party might Senate polls make look stronger than the result? Explore FEA’s estimates for 2026, then see what happened in past elections.
Source: FEA calculations ↗
Looking ahead
The map shows one estimate: how much polls may overstate either party in 2026. Hover or focus on a state for a preview; select it to explore its history.
What “D by 3” means here. Polls may make Democrats look 3 percentage points stronger than the final result. “R by 3” means the same for Republicans. This is a polling-error estimate, not a race forecast.
Past misses inform the estimate. FEA combines historical Senate polling errors with a regional starting point. States with little polling have less evidence behind their estimates. Colors deepen up to 7 points; the actual values remain in the tables.
State detail · Election history
Loading state detail…
Under “Polling average” and “Actual result,” D means a Democratic lead and R a Republican lead. The polling miss column shows which party the polls made look too strong. Blue means Democrats; red means Republicans.
Swipe the table sideways to compare the polling average with the result →
| Election | Polling miss | Polling average | Actual result | Polls studied |
|---|
For example: if polls show D by 5 but the result is D by 2, the polling miss is D by 3 — Democrats looked 3 points stronger in the polls than in the result.
The map is FEA’s adjusted estimate for 2026. The historical average below describes past polling errors before that adjustment. These are different measures, so only the 2026 estimate appears on the map.
Regional starting points help when a state has few polls. Evidence weight is a calculation score, not a poll count or a probability. The consistency adjustment is an input to FEA’s calculation. Evidence confidence is a qualitative label, not the chance that a party will win.
| Election | Polling period | Polls in last 21 days | Miss used in calculation | Share of state’s historical weight |
|---|
Most races use polls from the last 21 days before the election. Some use up to 45 days when recent polling is sparse. Extreme misses are limited to 12 points in either direction for the calculation; the main table still shows the actual miss. Weight shares are approximate percentages and may not add to 100 after rounding.
Compare states
The estimated miss shows which party polls may make look stronger in 2026. Select a state to see its past races.
| State | Estimated miss | Evidence confidence | Past races | Polls studied |
|---|---|---|---|---|
| Alabama | D by 4.72 | Very low | 2 | 3 |
| Alaska | R by 0.43 | Very low | 1 | 4 |
| Arizona | R by 0.84 | High | 5 | 62 |
| Arkansas | D by 5.17 | Low | 2 | 9 |
| California | D by 1.49 | Low | 2 | 4 |
| Colorado | R by 0.09 | Medium | 3 | 20 |
| Connecticut | D by 0.15 | Medium | 2 | 5 |
| Delaware | D by 3.38 | Very low | 2 | 4 |
| Florida | D by 5.02 | High | 4 | 51 |
| Georgia | R by 0.48 | High | 3 | 57 |
| Hawaii | R by 0.80 | Very low | 1 | 2 |
| Idaho | R by 0.16 | Very low | 1 | 3 |
| Illinois | R by 0.95 | Low | 3 | 9 |
| Indiana | D by 5.77 | Medium | 3 | 19 |
| Iowa | D by 4.26 | High | 4 | 31 |
| Kansas | D by 2.85 | Low | 3 | 5 |
| Kentucky | D by 6.13 | Low | 3 | 10 |
| Louisiana | D by 3.11 | Very low | 1 | 8 |
| Maine | D by 5.51 | Low | 2 | 7 |
| Maryland | R by 0.37 | Low | 3 | 5 |
| Massachusetts | R by 0.48 | Low | 3 | 11 |
| Michigan | D by 2.65 | High | 4 | 67 |
| Minnesota | R by 1.09 | Medium | 3 | 11 |
| Mississippi | D by 3.06 | Very low | 1 | 2 |
| Missouri | D by 3.16 | High | 3 | 21 |
| Montana | D by 1.48 | High | 4 | 12 |
| Nebraska | D by 2.38 | Very low | 1 | 2 |
| Nevada | R by 1.39 | High | 4 | 45 |
| New Hampshire | R by 3.79 | High | 4 | 41 |
| New Jersey | D by 4.51 | Medium | 4 | 13 |
| New Mexico | R by 1.14 | Medium | 4 | 9 |
| New York | D by 0.74 | Medium | 4 | 11 |
| North Carolina | D by 1.15 | High | 4 | 65 |
| North Dakota | D by 2.01 | Low | 2 | 4 |
| Ohio | D by 0.86 | High | 4 | 23 |
| Oklahoma | D by 1.83 | Very low | 0 | 0 |
| Oregon | R by 0.68 | Medium | 3 | 9 |
| Pennsylvania | R by 0.15 | High | 4 | 59 |
| Rhode Island | R by 0.92 | Very low | 2 | 5 |
| South Carolina | D by 3.69 | Low | 1 | 2 |
| South Dakota | D by 5.52 | Very low | 4 | 9 |
| Tennessee | D by 4.47 | Low | 2 | 7 |
| Texas | D by 1.22 | High | 4 | 21 |
| Utah | D by 0.30 | Very low | 2 | 4 |
| Vermont | R by 0.80 | Very low | 2 | 2 |
| Virginia | D by 3.20 | Medium | 3 | 12 |
| Washington | R by 6.84 | Medium | 2 | 6 |
| West Virginia | D by 3.21 | Low | 2 | 6 |
| Wisconsin | R by 0.08 | High | 4 | 36 |
| Wyoming | D by 0.17 | Very low | 1 | 2 |
A state can lean strongly Republican and still have a Democratic polling bias, or the other way around. The color shows which party polls may overstate, not which party is expected to win the state.
The estimates use Senate polling and election results from 2014 through 2024. A large estimated miss can come from a small sample. Oklahoma has no usable Democratic-versus-Republican Senate race in this dataset, so its estimate relies on its region and has very low confidence.
Explore the full calculations in the spreadsheet, or download the underlying data. This page presents a published set of estimates; it does not automatically refresh when the spreadsheet changes.