In the past few days, we saw a few polls for statewide races that looked so good for Democrats that you immediately start to question their accuracy. In this piece, I’ll analyze some of these polls and explain how undecided voters might be distorting the headlines.
Contents:
Democrats’ Problem With Undecided Voters in Kansas
The Golden 50
Iowa: A Blowout for Democrats?
Texas: Talarico pushes closer to 50%
Democrats’ Problem with Undecided Voters in Kansas
First, let’s look at the “undecided problem”. Polls aim to have as few undecided voters as possible. A 53-48% polling lead with 1% undecided voters is much more meaningful than 44% to 39% lead with 17% undecideds. Many undecided voters end up voting, and, in the past, they often skewed heavily Republican.
In Indiana’s 2018 Senate polling average, incumbent Democrat Donnelly led his Republican challenger Braun 44.3% to 43.0%. Donnelly ended up with 45%, close to his predicted vote share, but he lost to Braun, who received 51% of the vote, 7 points more than predicted by the polls. This is one of many cases where a large share of undecideds broke for Republicans on Election Day, creating a polling error, even though Democrats didn’t underperform their own polling share. Some of the data points show the same dynamic in 2026, but there is good news for Democrats, too.
A good example of this is Kansas. Democrats don’t need to win Kansas. If they do, it could be seat number 54 for them. However, a recent Emerson poll showed Adam Hamilton (D) leading incumbent Sen. Roger Marshall (R) 45% to 43%. The result is strikingly similar to Indiana’s 2018 polling average, but with fewer undecided voters, because a well-known Libertarian candidate is getting 4% of the vote, leaving 8% undecided voters. But undecided voters in recent polls lean heavily Republican.
In the Emerson poll, Trump 2024 voters made up 3/4 of the 8% undecided share, while Harris voters only made up 15%. In the graph above, you can see that if the undecideds were to break 3 to 1 in Republicans’ favor, Roger Marshall (R) would win by 2 points.
Winning over these undecided voters is difficult for Democrats. These are 2024 Trump voters who are saying they are undecided but who are still somewhat likely to vote. There is a good chance these are shy voters, who end up voting Republican but refuse to admit it in a polling interview, similar to what might have happened in Indiana. Some may be persuaded, but Democrats should also hope that some of these Republican undecideds just stay home. That would change the math. If only half of the 8% undecided voters show up, the race would be exactly tied at 47.9% to 47.9%.
Data bite:
Today’s NYT/Siena polls show Democrat Troy Jackson trailing Susan Collins by three points in Maine, and Democrats leading by 5 points in New Hampshire & Michigan.
The Golden 50
The undecided voter split is not just working in Republicans’ favor in Kansas. We can observe a similarly tilted balance of the group in other states like Texas and Ohio. The good news for Democrats: Candidates in these key states are polling much higher than Hamilton is in Kansas, and they are getting closer to the golden mark of 50%, where undecideds won’t tilt the race anymore. Third-party candidates can also be of help. As you’ve seen above, a candidate can win with 48-49% of the vote if there’s a third-party candidate who clutches around 2-4%. (We’ll keep the crucial battleground of Alaska out of this analysis because the polls in Alaska account for ranked-choice voting. In these polls, Mary Peltola (D) is above 50%.)
For all free Subscribers of Frontline Democracy, I will take a closer look at Texas and Iowa in the next paragraphs. In both states, Democrats recently got fabulous results from an A-rated pollster.