Blog/Polling

Following last night’s referendum in Virginia to approve new congressional maps, Focaldata’s first public MRP of the US House of Representatives finds the Democrats on course to win the House, with the party ahead in 228 seats to the Republicans’ 207.

The new Virginia boundaries see a three-seat swing to the Democrats when compared with our modelling on the existing boundaries, with the Democrats now on course to win 9 of the state’s 10 congressional seats. Virginia joins California, Missouri, North Carolina, Ohio, Texas, and Utah in changing its boundaries, all of which are reflected in the MRP model.

The result is driven by a 4-point national lead for the Democrats (49.3% to 45.6%), a tighter result than some public polls suggest. MRP is often able to produce a more rigorous national vote-share figure by correcting for the under-representation of some types of voters in traditional polling methods, and modelling voting intention within small demographic groups in each area before re-aggregating these against the true population.

Fieldwork was conducted online in waves between 23 January and 20 April 2026, with an overall national sample size of 8,058 respondents.

‍

Please note: the geographical location shown for some congressional districts below may not reflect the latest redistricting, but all vote share estimates have been produced on the updated district lines.

‍

Who is moving to the Democrats?

Our results suggest a small shift in voter preferences, but large turnout differentials. Only 6% of Trump 2024 voters are intending to vote Democrat in November, but we see 67% of Harris 2024 voters intending to vote in the midterms, compared to 60% of Trump 2024 voters. Many Republicans may just stay at home in November rather than switching to the other side.

In the two-party vote preference, three key demographics are shifting towards the Democrats. Hispanic voters have seen a 10-point shift left compared to the 2024 presidential race. The Democrats currently lead 61-39 in two-party vote share, compared to a 56-44 Harris lead over Trump.

The Republican lead with voters with high school or lower levels of education has fallen from a 22-point lead for Donald Trump to just under 15-points today, an 8-point move to the left, and similarly, voters earning under $50,000 are now narrowly intending to vote Democrat (51-49 two-party share), also an 8-point shift from 2024.

Across the board, voters who have been most heavily hit by an increase in the cost of living and shifting furthest away from Republicans. There is a statistically-significant relationship between a state’s increase in gas prices over the last 12 months and the swing to the Democrats. In the 15 states with the highest gas price increases, the swing to the Democrats is three times as large as the swing in the 15 states with the lowest gas price increases.

‍

Methodological notes

Demographics used

We have built a post-stratification frame on all 435 congressional districts, including age, gender, ethnicity, education level, marital status, income, and 2024 vote as variables. In addition, we have modelled interactions like race by past vote. We also use a number of state-level fixed effects like median income, incumbent performance in the previous House election, and the percentage of Spanish speaking respondents in each district.

‍

Turnout and likely voter modelling

The estimates produced are of likely voters. Each cell has a modelled likelihood of voting based on a combination of self-reported likelihood to vote and demographic turnout modelling. The overall implied turnout is 49%.

‍

‘Other’ vote share

We ask voting intention in a multi-step process. Those who say they would not vote or don’t know who they would vote for are asked again for their voting intention, with respondents who say they would not vote then removed from the voting sample. In the latest version of our tracking survey, we have taken extra steps to reduce the amount of respondents who select ‘Another party’. In reality, we suspect some of these respondents will be people who are either undecided or would not vote. We will start to see this reflected more in future updates of the MRP.

‍

Candidates

The MRP model assumes that the Democrats and Republicans will both have candidates in each congressional district. In states with a top-two primary system, projections will be updated to match the final candidates following the primary elections. In ranked-choice states (Alaska and Maine), future iterations will also report two-party vote choice in cases where neither party gets over 50% of the vote in our initial projection.

‍

‍

A full breakdown of results by seat and demographic group can be found here.