A special election puts Democrats on track to flip the House

In march 2018 a Democrat won a stunning victory in a special election for Pennsylvania’s 18th district in the House of Representatives, flipping a seat that Donald Trump had won by 20 points. The result made clear that Republicans’ House majority would soon be washed away in the midterms. Ahead of the special election held on December 2nd in Tennessee’s seventh district, which Mr Trump had won by a similar amount, Democrats hoped to repeat the feat. This time, however, Republicans held the seat by nine percentage points—a good-not-great showing for Democrats in a conservative region.

With Republicans’ House majority unscathed, attention has turned to the race’s implications for next year’s midterms. Special elections have been accurate bellwethers in the past, and The Economist’s congressional-forecasting model relies on them (alongside polls and past elections) to predict the national political environment. Using data from the political websites Ballotpedia and The Downballot, the model first compares the results of each special election with those of the most recent general election in the same district. For example, Democrats cut the Republican margin of victory in the seat west of Nashville from 21 percentage points in 2024 to nine this week, a gain of 12 points.

Next, the model estimates the national popular vote for the House implied by this shift. Because special elections tend to attract only the most fired-up voters, whichever party has the wind at its back tends to enjoy a big turnout advantage, exaggerating swings. On average, each percentage point of vote-share gains in a special election has been worth around 0.4 points in the next general election. Adding this reduced swing onto the nationwide results from 2024, the Tennessee race suggests that Democrats are on track to win the midterms by about six percentage points.

Finally, our model combines every special election into an average, weighting the most recent contests and those for Congress more than older ones and those for state-legislative seats. Because the Democrats’ performance in Tennessee roughly matched their average results earlier in this cycle, our special-election-based forecast remains unchanged at around a five-point Democratic win. This is slightly ahead of where the party stood in late 2017 (see chart) and almost certainly enough to flip the House, though it would probably leave the Senate in Republican hands.

Compared with Democrats’ big electoral wins last month in New Jersey and Virginia, the Tennessee results look a bit humdrum. Two local factors suggest they might understate Democrats’ nationwide advantage. First, the party’s nominee, Aftyn Behn, is a former progressive activist who regularly films federal immigration officials, making her a poor ideological fit for a red district. If history is any guide, the party’s candidates in most close races next year will probably be more moderate.

Second, turnout was quite high, with more major-party votes cast (180,000) than the 177,000 recorded in the midterms of 2022. Ms Behn won 18% more votes than the Democratic candidate did that year. This suggests that rather than simply benefiting from unmotivated conservatives who stayed home, Democrats also persuaded a hefty slice of voters to switch sides—meaning that the 0.4 ratio used in our model may be too low.

On the other hand, special elections as a whole have lost predictive power of late. As political parties have polarised by education, Democrats have enjoyed a growing advantage in low-turnout contests, which does not necessarily translate to general elections. The 2024 cycle, in which a wave of support from politically disengaged voters won Mr Trump the presidency, was among the worst performances for our special-election forecast on record. It expected Democrats to win the national popular vote for the House by five points, when in fact they lost it by two.

Because Mr Trump is not running next year, our special-election forecast will, with luck, avoid a similar error. The real test is likely to come in 2028, when a new Republican presidential nominee will have to try to piece Mr Trump’s coalition of lower-turnout groups back together. ■