How the 2026 Polls Stack Up Against Recent Elections

After pollsters underestimated Republicans in 2016, 2020, and 2024, the 2026 midterms tested whether the industry would again underestimate Republicans, as many Republicans predicted, or would swing the other way, as pollsters overcorrect for their previous errors. The answer turned out to be the latter, as the polls significantly underpredicted Democratic performance in a few key races, though the overall error was still slightly below historic average for Senate contests.

The biggest polling misses this cycle were in Texas, where Democrat James Talarico was underestimated in the polls by 9.5 points, as he won by 12.1 points, but the final RCP Average had him up by 2.6 points. Polls underestimated gubernatorial candidate Gina Hinojosa by a similar margin, as she beat incumbent Gov. Greg Abbott by 7.1 points, while polls had Abbott up 2.3, a 9.4-point swing.

Among the toss-up Senate races, the next highest polling error was in South Carolina, where polls significantly underestimated Democrat Annie Andrews. In the final polls, Republican Darline Graham Nordone led by 2.6 points, while Andrews won by 4.5 points, a 7.1-point miss.

The cycle’s other significant misses came in states that were swing states in 2024 but that RCP already rated as “Likely Dem” this year. In Georgia and North Carolina, polls heading into Election Day had Democrat Jon Ossoff up 8.6 and Democrat Roy Cooper 11 points, even though President Donald Trump won both states in 2024. Yet on Election Day, they won by even greater margins: 18 points for Ossoff and 19.6 for Cooper, a 9.4- and 8.6-point swing in their direction. Georgia Democratic gubernatorial candidate Keisha Lance Bottoms also significantly outperformed the polls, which had her up 0.3 points, while she won by 8.7, an 8.4-point swing.

Even the largest of these errors, like the 9.5-point miss in Texas, aren’t unprecedented in recent history. In 2022, polls missed by 11.5 points in Washington’s Senate race, and in Maine’s 2020 Senate race, they missed by between 9.6 and 15.6 points, depending on the poll.

This year’s misses didn’t follow a consistent pattern, as pollsters were fairly accurate in other races in both reliably red states and recent swing states. In Kansas, polls only underestimated Democrat Adam Hamilton’s lead by 1 point, as they had him up 4.1 and he won by 5.1. In Michigan, Democrat Abdul El-Sayed won by 7.1 points, similar to his final RCP Average lead of 5.3 points.

In 2026, not every race’s polls underestimated Democrats. In Nebraska, where Republican Sen. Pete Ricketts faced independent Dan Osborn, the final RCP Average had Ricketts up 3.7 points, while he won by 6.2. In Alaska, where the count is still ongoing, polls also slightly overstated Democrat Mary Peltola’s position, as she led by 1 point in the final average but currently leads incumbent Sen. Dan Sullivan by just 0.3.

Overall, the polling error of the 2026 midterm election wasn’t much greater or worse than in previous elections. The RCP Average missed by 3.5 points on average in toss-up Senate races, slightly worse than in 2024, when it missed by 2 points in Senate races and 1.6 in swing state presidential contests. However, it’s slightly better than the polls have been in Senate elections between 2008 and 2024, whose average error is 3.8 points.

In 2026, the polls missed in Democrats’ favor, underestimating the Democratic candidate in seven of the nine toss-up Senate races, the opposite of recent presidential elections, when both Senate and presidential polls overestimated Democrats. On average, the RCP Average overstated Republicans’ margin by 2.7 points in this year’s toss-up Senate races, well below the 6-point overstatement of Republicans in 2022.

Over the longer run, these misses have tended to even out. In Senate toss-ups between 2008 and 2024, the RCP Average overstated Republicans by an average of just 0.4 points, as cycles that favored one party were offset by cycles that favored the other. In swing state presidential polling over the same period, the average miss was 0.4 points in the other direction, overstating Democrats.

The comparison to recent historical data indicates that the error this cycle was largely in line with past elections. The RCP Average’s 3.5-point miss in toss-up Senate races was slightly better than the 3.8-point average from 2008 to 2024. The large misses in Texas, South Carolina, and Georgia will likely draw the most attention from this cycle, but taken as a whole, the 2026 polling looked much like the polling of recent midterms, with a few misses in individual races rather than a broad failure across the map.