Nate Silver Calls Out ‘Boring’ Pundits Missing the Biggest Midterm Clues
The pollster earlier predicted a big win for Democrats in the Nov. 3 elections.
A polling guru has weighed in on election analyses with the midterms only two weeks away.
Nate Silver, the founder of FiveThirtyEight, urged his fellow pollsters to “take a longer view” as he warned against oversimplified assessments of the November elections.
“The thing that’s annoying about election analysis is that you’ll never be criticized for saying ‘thing that happened last time will happen again’ which makes people kind of boring and risk-averse and also leads to a lot of overfitting from n=1 or n=2 samples,” he wrote in a series of posts on X.

Nate Silver on X
Overfitting is an error that occurs when a model sticks too closely to its training data, to a point where it fails to make accurate predictions or conclusions on any other set of information.
ADVERTISEMENT
“It’s kind of vaguely empirical, I guess,” Silver went on. “But in principle, one advantage of the models is that ‘thing that happened 13 out of the last 17 times is a good bet’ is more robust than ‘thing that happened 1 out of the last 1 times is Just How It Works Now and will happen again.’”
Polls are predicting a blue wave in this year’s midterm elections.
Jonathan Drake/Reuters
“Not saying there aren’t interesting [questions] here about when ‘old’ data becomes obsolete. But part of the reason to take a longer view is that even when there’s a regime change, it’s hard to know when the regime will change again,” he added.
Silver acknowledged that there is a “very robust” pattern of the sitting president’s party performing poorly in the midterms—just as many polls have shown the GOP is likely to get trounced by Democrats.
Get the best of The Beast delivered right to your inbox
By clicking “SUBSCRIBE” you agree to our Terms of Use and Privacy Policy.
But he also pointed out that the pattern “doesn’t cleanly answer ‘can Democrats really win in Kansas?!’ but it sets various floors and ceilings.”
Silver earlier upgraded the Democratic Party’s chances of winning the Senate to 75 percent.
Earlier this month, Washington Post polling director Scott Clement asked, “Are midterm polls underestimating Republicans?”
Republican pollster Patrick Ruffini said polls from 2018 to 2024 conducted at this stage overestimated support for Democratic Senate candidates by 6.2 points in Iowa, Michigan, and Ohio. State Navigate’s Chaz Nuttycombe also noted that rural Republicans didn’t answer their survey in an Iowa legislative district, warning it could be a “Canary in the Cornfields for Iowa polling this year.”
Pollsters have also cautioned that inaccurate sample groups, skewed data sampling, and unknowable voter behavior can all affect the outcome of a poll and could be inflating support for the Democrats.
“It’s impossible to know whether polls will overestimate Democrats or Republicans before any election,” Clement asked. “But understanding why polls missed in recent years, how pollsters have adapted, and what the data shows this year can help set expectations for this fall.”
Register below to read this article for free or subscribe
to unlock unlimited access to The Daily Beast.
Monthly
$1
First month then $5.99/month
Annual
$35
First year then $59.99/year
Premium
$79
First year then $119.99/year
*Substack access provided by the next business day, using your subscription email. Choosing the Premium plan constitutes your permission to share your subscription email with Substack and your agreement to Substack’s Privacy Policy.
Already have an account? Sign In
Looks like you already have a subscription!
You're all set!
Thanks for subscribing.


ADVERTISEMENT
Sign in or create an account
Login dialog
Loading comments…
ADVERTISEMENT