Polls as Psyops: What CNN’s Own Numbers Say When Nobody Edits Them

Eighteen days before the November 3 midterms, the most useful piece of election analysis on American television has been turned into a weapon by one side and ignored by the other.

Earlier this fall, CNN chief data analyst Harry Enten asked the question every campaign professional asks in private: what if the polls are wrong? He ran the 2026 map through a polling miss of two to four points in Republicans’ favor. The result: Republicans hold the Senate 53–47, flip Michigan, and keep the House. “If the polls undershot the Republicans by greater than three points,” Enten said, “that’s all you need,” putting a Republican House at 218 seats or more.

This week the clip resurfaced across conservative social media under the banner that “CNN has officially confirmed” a Republican hold. Meanwhile, the outlets celebrating double-digit Democratic leads carried on as if the segment had never aired.

Both reactions are psyops. Neither is analysis.

What Enten actually said

The full segment was two-sided. Since 1998, Enten noted, polls have underestimated Republicans about a third of the time, Democrats about a fifth of the time, and neither party roughly half the time. He then ran the mirror image: if the polls are underestimating Democrats by the same two to four points, Democrats take the Senate 53–47 by flipping Maine, North Carolina, Iowa, Texas, Ohio and Alaska. Take today’s polls at face value, and Democrats win the chamber 52–48.

That is the real story. The entire range of plausible Senate outcomes, from a 53-seat Republican majority to a 53-seat Democratic majority, fits inside an ordinary polling error. A three-point miss is not a scandal. It is a routine Tuesday for the polling industry.

Yet the version on social media strips out the mirror image, and the version most legacy outlets present strips out the warning. One side sells a comeback; the other sells inevitability. Both are selling turnout, not truth.

The warning is not hypothetical

Enten has made the case more than once. On September 30 he showed that Republicans beat their September 2024 polls by nine points in Iowa, six in Ohio and five in Michigan. “Fool me once, shame on you,” he said. “Fool me four times”, and you have to ask whether it happens again. In Ohio, Republicans outran their early-October Senate averages by seven points in 2016, eight in 2018 and seven in 2022. On October 12, he added that early-October generic-ballot polling has underestimated Republicans in 14 of 16 elections since 1994, by an average of about four points.

Now set that against what is being sold as consensus. As of October 8, Nate Silver’s model gave Democrats a 93 percent chance of winning the House and a 75 percent chance of taking the Senate. Kalshi traders put the House at 90 percent Democratic and the Senate at 62 percent. Pew finds that 55 percent of Americans expect Democrats to win the House. Those probabilities are only as good as the polls feeding them, and the races that decide the Senate are concentrated in Michigan, Ohio and Iowa, the states where Republicans have repeatedly outrun their numbers.

The counterargument deserves a fair hearing. In the 2025 off-year elections, polls underestimated Democrats by about six points. The same CNN analyst noted that when Democrats lead the generic ballot by nine points this late in a cycle, they have won the House 15 of 15 times since 1942. Midterms are not presidential elections. Donald Trump is not on the ballot, and the low-propensity voters who made presidential polls miss may stay home.

An honest analyst holds both possibilities at once. A partisan narrative holds only one.

Why I lean toward a Republican hold

Three factors tilt me toward the Republican side of that error band.

The first is the Democratic brand. Bryan Bennett, who founded the Democratic-aligned Navigator Research, calculates the Democratic Party’s net favorability at -13 this quarter against -17 for Republicans: a relative brand advantage of just four points, compared with 22 in 2006 and 17 in 2018. Yet Democrats lead the generic ballot by seven. In every midterm since 2006, the Democratic generic-ballot margin ran at or below the brand advantage; this year it runs above it. His worry, and mine, is that likely-voter models are too generous to people who dislike Trump but do not much like Democrats either. That is a turnout problem no weighting scheme can fix.

The second is the party’s nominees. The insurgent wing won Michigan’s Senate nomination with Abdul El-Sayed running on Medicare for All. Michigan is the first state to flip in Enten’s Republican scenario. Expectations of a Democratic Senate rest partly on a candidate whose electability case is untested, in a state that has repeatedly overstated Democrats.

The third is the information ecosystem itself. This summer a phantom outfit called Median Strategies released fabricated polls in Los Angeles and Wisconsin. The Los Angeles “poll” was picked up by the California Post and promoted by Mayor Karen Bass’s own campaign before it was exposed as fake. Roughly 30 new pollsters have appeared this year, and the person behind Median was reportedly a 21-year-old recent graduate. When a fake topline can move a mayoral campaign, the numbers have become ammunition. I made that case in these pages in August; the evidence since then has only strengthened it.

California is where the error lands

This matters in Sacramento, not just Washington. The Proposition 50 map was drawn to put five of California’s nine Republican-held seats within Democrats’ reach, yet the Cook Political Report rates only three California races as truly competitive: CA-22, CA-13 and CA-48. In the 22nd, David Valadao is defending a district where Democrats hold a 16-point registration edge. In the 13th, Adam Gray won by fewer than 200 votes in 2024, and the two Republicans combined outpolled him in June’s primary. In the 48th, Jim Desmond and Marni von Wilpert are neck and neck. In my own backyard, the only public polls of the CA-6 race between Kevin Kiley and Richard Pan listed by the New York Times come from a firm affiliated with Kiley.

These seats will be decided by a few thousand votes, where a three-point error is the whole ballgame. If national polls are overstating Democrats by even Enten’s modest margin, the seats Proposition 50 was built to deliver become coin flips, and the House majority Democrats are counting on could be settled in the Central Valley. Ballots are already in mailboxes; the registration deadline is October 19, and early voting centers begin opening October 24.

Insist on the whole forecast

The answer to psyops is not counter-psyops. It is insisting on the full range of outcomes. A forecast that gives one party a 93 percent chance while the decisive states sit inside a normal polling error is not a measurement. It is an expectation, and expectations are being used to move money, volunteers and voters. The viral “CNN confirmed” clip is no better. Enten confirmed only that a modest miss would change the result, in either direction.

My read: the House leans Democratic, the Senate is a genuine toss-up that tilts Republican if history repeats, and the shape of Washington next year is far less settled than the forecasts imply. Neither party should believe its own press. Voters should trust neither party’s polls. And everyone should vote as if the margin will be one.