How Polls Undercounted Democratic Support on Tuesday

Democrats notched sweeping wins Tuesday across the key statewide contests in New Jersey and Virginia, outperforming expectations from both parties. Pre-election polls indicated that both the Virginia attorney general and the New Jersey gubernatorial races would be tight contests, with margins of only a few points. Instead, the final tallies revealed a stronger-than-expected Democratic showing, setting up questions about what went wrong in the polling.

The difference between the polling and Tuesday’s actual results was fairly sizable. In the Virginia governor’s race, the RCP Average had Democrat Abigail Spanberger up 10.2 points, and in the latest counts, she is up 15.2. In the New Jersey governor race, the RCP Average had Democrat Mikie Sherrill up by 3.3 points, and she went on to win by 13 points in the latest counts.

This is a stark discrepancy between the polls and the actual results. In the swing states in 2024, for example, the largest discrepancy between the RCP Average and the actual result was in Arizona, where polls had Trump up by 2.8 points, and he won by 5.5 points. On average, the RCP Average was off by only 1.8 points in 2024 in the swing states.

Even more irregular is that in the Virginia attorney general election, where Jay Jones is up 6.6 points, as well as in the New Jersey governor race, not a single October poll predicted the race would go in the Democrats’ direction to the degree it did. Where Sherrill is set to win by about 13 points, polls had shown her ahead by no more than 7. In Jones’ race, surveys showed him leading by at most 2 points, with most either tied or giving Republican Jason Miyares the edge.

This leaves the question of why polling was so significantly Republican-favored.

The most likely explanation for the big changes compared to recent polling is the recent effects of the government shutdown. Before Oct. 28, Trump’s approval rating had actually decreased since the start of the shutdown, dropping from -7.7 on Oct. 1 to -6.7 on Oct. 28. However, since Oct. 28, his approval has dropped precipitously to -11, as government programs such as SNAP have started to be cut off, and government workers have not received paychecks. This last-minute felt impact of the government shutdown on voters likely was not seen in early- or mid-October polling, but did affect election results.

The other big factor that Republicans and pollsters will have to deal with going into 2026 and beyond is whether Trump voters can be converted into Republican voters. In 2024, Trump was able to turn out more young people, Hispanics, and working-class voters than Republicans had been able to in previous elections.

Comparing Trump’s popular vote win to the New Jersey election, Trump was able to garner 43% of 18-29-year-olds, compared to Republican Jack Ciattarelli’s mere 31%. Among Latinos, Ciattarelli was also able to garner only 31%, compared to Trump’s 46%.

Both parties tend to struggle in midterm or off-year elections when their own party is in power. The party that won the presidency has not kept the House in the midterms since 2002. However, with Trump, this problem seems particularly troublesome given the latest elections in Virginia and New Jersey and the spring Wisconsin election, where liberal-leaning Supreme Court Justice Susan Crawford beat conservative Brad Schimel by 10.1 points – in a state Trump won by 0.9 points last fall.

Whether Republicans can retain the voters Trump amassed in 2024 remains an open question, and the latest results suggest they are struggling to do so.

For pollsters, this means that in future elections without Trump, polling the same samples that led to some accuracy in 2024 will not work for the midterms and perhaps not for future presidential elections. Instead, to accurately assess the Republican voter base, they will need to design methods that target voters who are likely to vote when Trump is not on the ballot.