By Stephen Fisher, 3rd November 2026
Today the USA is holding its midterm elections. The poll-based forecasters and prediction markets are near certain that the Democrats will win control of the House of Representatives and they favour the Democrats for control of the Senate. Those predictions are directly or indirectly informed by opinion polls and historical experience, including the errors of opinion polls in previous elections. Election results can often end up being judged in light of expectations set by forecasts. That can get particularly complicated if the polls are systematically out, leading some to complain that the polls got it wrong while others claim that a party underperformed.
This blog post sets some of the poll-based forecasts for today’s elections in an historical context, using simple metrics, and thereby considers how the results might best be interpreted if the expectations of Democrat success are met, or not.
The House
If the Democrats win the House then it will be confirmation of what is widely expected and no surprise, given the polls. It will be an achievement but not a remarkable one.
To win half the seats in the House FiftyPlusOne and ElectIt both estimate that the Democrats need about a +4 point lead in the popular vote. That is a measure of how the electoral geography (including the boundaries of districts) advantage the Republicans.
Also, by historical standards a comfortable Democratic win of the House in the midterms with a Republican in office would not be unusual. In the 2018 midterms, during Trump’s first presidency, the Democrats won an 8.6-point lead in the share of the vote and 235 seats (more than the 218 required for control). Similarly in 2006, under George W Bush, the Democrats won with 233 seats and a 5.9-point lead. Under George H W Bush’s midterms, the Democrats won 267 seats and a 7.8 point lead. Admittedly that was 1990 and still part of a long period during which the Democrats consistently won House elections in both general elections and midterms regardless of which party won the presidency.
The Democrats have won the House in all the midterm elections under Republican presidents since 1954 with just one exception: 2002. The Republicans won then most likely because they benefitted from a rally-round-the-flag effect after the September 11 2001 terrorist attacks. Moreover, at that point the US defeat of the Taliban in Afghanistan was considered a major success with relatively little loss of life (especially American lives) and Iraq had not yet been invaded. The circumstances of today’s election are rather different.
The Harvard and Caltech forecasting team, ElectIt, say, “Given today’s estimate of the generic ballot, D+8.8, we expect a national popular vote of about 54.0% [of the two party vote].” That D+8.8 is similar to the average for other major forecasters here. 54.0% of the two-party vote would be the largest share of the vote for the Democrats since the 54.4% they secured in 2018 and larger than the Republican’s have managed since before the Second World War (sources vary slightly but see here and here for the historical record).
The nationwide share of the vote will not be known for several days. If the Democrats are on course for around 54% of the two-party vote, on par with 2018, they should be up by about +6 points on average since 2024. Early results might be indicative of that.
More broadly, given that in 2018 and 2006 the Democrats won a remarkably similar number of House seats (235 and 233 respectively) and given that they are forecast to get a similar number or perhaps more this year (e.g. see John Sides spreadsheet here), the Democrats should be disappointed if they do not achieve that.
But perhaps redistricting since 2024 will limit their success. According to an article from August in the Washington Post, “if Democrats win all of the most competitive districts classified by the Cook Political Report, they would win 235 seats under the 2024 map. But under the current map, they would only win 223 seats.” On that basis redistricting potentially costs the Democrats 12 seats, but the impact will depend on the level and distribution of votes. Given the partisan effect of redistricting, Democrat performance relative to 2006 and 2018 should probably be judged more on their share of the vote than on the number of seats they win.
The Senate
The Democrats need a net gain of four seats to win control of the Senate.
Apart from Ohio, all the most competitive Senate seats were last contested in 2020, the year that Trump contested Joe Biden’s election victory. Then the generic-ballot polling average was around D+7 but the Democrats won the House with only a 3-point lead due to polling error. Nonetheless, that is a relatively strong baseline for the Democrats to defend.
Despite that, the Democrats are expected to make net gains rather than losses, if only because this is now a midterm election with a Republican president rather than a (close) general election. As the discussion of the House election polls and forecasts show, the Democrats are expected to perform at or above the level of 2018, when their share of the House vote was 3 points higher than it was in 2020. That is another reason to think that the Democrats ought to make gains in the Senate rather than suffer any losses.
However, there are some states that the Democrats won last time that are at risk this year, including Michigan, Minnesota and New Hampshire. The polling averages suggest the Democrats will hold those with leads of 6, 6, and 9 points respectively. That implies swings to the Democrats since 2020 of +2, +0.5, and -3.5 points respectively. All of those are much less than the Democrats might have hoped for and well below the level the party ought to achieve if their performance is on par with 2018 and 2006.
The states most likely to flip from Republican to Democrat (with the swing since 2020 required to win) are Alaska (6.4), Iowa (3.3), Kansas (5.7), Maine (4.3), North Carolina (0.9), South Carolina (5.1), Texas (4.8) and Ohio for which there is no 2020 baseline. (In Ohio, Jon Husted was appointed senator after JD Vance became Vice President. Now Hunter is standing in a special election to serve the remainder of the term that started in 2022 when Vance was elected.)
Various factors affect Senate elections and lead to variability over time and between states, not least incumbency and demography. Here I consider two simple indicators of what a party might reasonably be expected to win that are independent of polls: the swing required since the previous election and the track record of winning Senate seats. Actual performances might be reasonable explained with reference to other factors, but required swing to win and past electoral success are reasonable benchmarks to start from.
In 2024 Maine elected an Independent to the Senate who votes with the Democrats, so a Democrat failing to win the Senate race today should be considered disappointing, despite a substantial 4.3-point swing being required.
Similarly, the Democrats won the Senate race in Ohio in the 2018 midterms during the first Trump presidency. Given that the Democrats ought to do at least as well as they did in 2018, losing Ohio today should be considered a disappointment for them.
Based on a required swing of just 0.9 since 2020, North Carolina is the easiest target seat for the Democrats. If they do not win there, they will be notably short of 2006 and 2018 levels of performance.
With a similar logic, the required swing of 3.3 since 2020 to flip Iowa means that a Senate gain there would be on par with, or only slightly surpass, the Democrats success in 2006 and 2018.
On the basis that the Democrats could be expected to win Iowa, Maine, North Carolina and Ohio while retaining the seats they are defending based on the kinds of performances they put in at the 2006 and 2018 midterms under Bush and Trump, there is a path for them to win control of the Senate that they reasonably ought to achieve the equivalent of today.
It may not be precisely that scenario that produces a net +4 gains for the Democrats in the Senate. Simulations by FiftyPlusOne show much more uncertainty in how the national vote share translates into seat tallies for the Senate than the House. They find Republicans managing to win 50 seats (and so keeping control) while the Democrats have anywhere between a +3.6 and +10.5 lead in the national generic congressional vote. We will have to just wait for each Senate result to be declared, and maybe wait a long time for Alaska, before we know whether the Democrats have won control of the Senate.
Alaska, Iowa and North Carolina all elected Democratic senators in 2008, when Obama won a big victory in the aftermath of the financial crisis. If the Democrats do not win Senate seats in those states this year, it should raise questions as to why the Democrats are not rising to the levels they achieved in 2008 when both this year and then the elections are in the context of a beleaguered Republican administration. That said, Alaska not only requires the biggest swing since 2020 but it has been complicated by an electoral administration controversy.
South Carolina’s last Democratic senator, Ernest Hollings, was re-elected as recently as 1998, but first elected in 1966 and was then a segregationist. Texas has not elected a Democrat to the Senate since 1988, and Kansas not since 1932. Failure to win any of those three should not be considered a setback even if Democratic hopes are dashed. Winning any of them would be historic.
