Deception & Misdirection
Agents of Influence: The Pitiful Presidential Pollsters
Most people work just hard enough not to get fired and get paid just enough money not to quit.
—George Carlin
Every four years, presidential opinion polling reliably causes regime media to misplace their poop. But after the actual polling places close and report, the stenographer journalists generally don’t expose which—if any—of these influential prognosticators should be publicly grilled for fouling up.
A current exception proves the rule.
On the weekend before Election Day, pollster Ann Selzer unleashed an Iowa survey for the Des Moines Register purporting to show Kamala Harris leading Donald Trump by 3 points in a state Harris supposedly had no business winning. The “late shift toward Harris,” declared the Des Moines Register, was happening because of older women dumping Trump. An editorial page writer from the Arizona Republic warned the “sudden and potentially seismic” numbers “change the nature of the national race.” MSNBC’s Rachel Maddow spent 11 minutes gleefully promoting Selzer and calling her the “gold standard.” The liberal ladies of The View on ABC celebrated. Transportation Secretary Pete Buttigieg X-posted “Iowa, you have shocked the nation.”
On it went, but Ms. GoldStandard™ missed the Iowa call by two touchdowns, plus a pair of two-point conversions. Trump won Iowa by 13 points, ran the table in the battleground states, and silenced Electoral College critics by winning a popular vote majority.
But there will not likely be any banishment to a shameful retirement. Selzer boldly seized back the media megaphone, and they apparently still have her on speed dial.
“I told more than one news outlet that the findings from this last poll could actually energize and activate Republican voters who thought they would likely coast to victory,” she claimed. “Maybe that’s what happened.”
Well then . . . glad that’s been clarified!
Pollsters have far too much power to drive the national narrative. Selzer—wrong as she was in this one instance—is far from the worst offender.
The worst is hiding in a wide field with too many options.
A week before Selzer’s big miss, the Eagleton Institute of Politics at Rutgers University projected a 20-point Harris margin of victory in reliably blue New Jersey. The actual margin was just 5 points, far closer than any Republican has come to winning the state since 1992.
But while that’s also a two touchdown miss, it’s at least in favor of the winner. More consequential and influential damage was done in the swing states. In these places, where the actual margin is expected to be less than a few points per state and the presidency hangs on the outcomes, pollsters that reliably miss by three points or more shouldn’t be working anymore.
Alas . . .
“This . . . is CNN”
Trump won Wisconsin by 0.8 percentage points, a razor-thin margin consistent with the prior two presidential races in the state. But a poll conducted through October 28 by CNN had Harris winning Wisconsin by 6 points. No other poll in the RealClearPolitics average for Wisconsin at any point in October was within even 4 percentage points of CNN’s pitiful projection.
In another miss by nearly a touchdown, Michigan voters made Trump the winner by 1.4 points, while the CNN October 28 survey claimed a 5-point lead for Harris.
Similarly, CNN’s late October survey had Harris seizing a 1-point win in Arizona . . . where Trump ended up winning by 5.5 points. Of the nearly two dozen October polls listed by RealClearPolitics for Arizona, CNN was one of just two to even credit Harris with a lead.
And while Trump took North Carolina by 3.3 points, CNN gave the nod to Harris by 1.
Add them together and CNN, at the last minute, was projecting Harris to convincingly win the majority of battlegrounds and thus the Electoral College. Instead, she got shut out.
Back in the 2020 presidential race, while not outright calling the wrong winner in most of the battlegrounds, CNN’s final, late October surveys frequently missed by more than a touchdown. They were off the mark by 9.2 points in Michigan, 8.8 points in Pennsylvania, and 7.3 in Wisconsin.
In 2020 North Carolina CNN’s last poll put Biden up 6 points in a state Trump ended up winning by 1.3 points.
Missing with Marist
The 2020 NBC/Marist poll was the only other late October survey to match CNN’s misplaced 6-point lead for Biden in North Carolina. They also badly miscalled Florida in 2020, projecting a 4-point win for Biden in a state Trump won by 3.3 points.
Marist’s 2024 polling didn’t improve.
As noted, Trump won Michigan last week by 1.5 points, but the Marist poll released on October 30 projected Harris up 3. In Pennsylvania, Trump won by 2 points but Marist called it for Harris by 2. Marist’s October 30 poll for Wisconsin put Harris up 2, while Trump won by 0.8 points.
As was the case with CNN, Marist’s late polling for the 2024 battlegrounds also projected the wrong overall winner in the Electoral College.
As the kids like to say: “You had one job!”
It is possible for professional polling firms to at least get the winner of the Electoral College correct, if not all the precise states. Atlas Intel, for example, correctly nailed Trump as the winner in all seven 2024 battlegrounds.
The Grey Lady’s Lousy Polls
The New York Times–Siena College poll wasn’t so fortunate, mis-projecting Harris as the winner in their final polls for Nevada (missed by 6.8 points), North Carolina (off by 5.3 points), Georgia (3.3 points) and Wisconsin (2.8 points.) That’s already enough to get the final electoral count profoundly wrong, but they also punted and predicted ties in Michigan and Pennsylvania.
So, the “All the news that’s fit to print” newspaper failed to call the winner correctly in six of the seven battlegrounds.
It was only slightly less ugly in 2020, when the NYTimes-Siena final survey outright mis-called Biden rather than Trump the winner in Florida (off by 6.3 points) and North Carolina (off by 4.3 points.)
Though they correctly projected Biden the winner in Wisconsin, Arizona, Michigan, and Pennsylvania in 2020, the NYTimes-Siena numbers in each of these battlegrounds meaningfully (and sometimes outrageously) overstated Biden’s margin of victory. In Wisconsin, Biden eked out a 0.7 point win, underperforming the NYTimes-Siena ludicrous prediction of an 11-point blowout. The poll called Arizona for Biden by 6 points, but he won by just 0.3 points, close to the 0.9 point Real Clear Politics average prediction favoring Biden.
Michigan was similar: Biden won by 2.8 points, but NYTimes-Siena projected him up 8 points. And in Pennsylvania, the poll projected Biden winning by 6, but he won narrowly with a 1.2 point margin.
Influentially Biased
Astute readers may notice another similarity. In addition to reliably and badly missing over the past two presidential elections, each of these three polls managed to reliably miss in the same direction.
RealClearPolitics ranked 2020 multistate pollsters by both degree of error and direction of error bias. For example, the least error-prone pollster in 2020 was InsiderAdvantage, with their polls off by an average of just 2.4 points, with 67 percent of the misses overstating the strength of Republicans and 33 percent overstating Democrats. Next best was Susquehanna, with an error average of 2.5 points and a Republican bias of 60 percent.
The NYTimes-Siena average gap between prediction and reality in 2020 was 5.1 percentage points. That’s already a big “oops” that’s well outside the claimed error margin of almost every individual poll. But then there’s the directional bias, showing a whopping 92 percent of the mistakes made by the NYTimes-Seina 2020 polls overstated the strength of the Democratic candidate.
The average Marist polling miss in 2020 was 5.7 percentage points, with 83 percent of the errors favoring Democrats.
And then there’s CNN . . . 7.3 percent average inaccuracy with their wrong calls in 2020 favoring Democrats 100 percent of the time!
RCP’s 2020 pollster scorecard found four other polling firms that missed the truth by an average of more than a touchdown (7 points) per race, and in each case the bias was also 100 percent favoring Democrats.
Not one of the 20 pollsters on the 2020 list had a 100 percent bias that favored Republicans, while 10 of them (half!) scored a perfect 100 percent bias favoring Democrats.
Only three of the 2020 pollsters showed any skew favoring the GOP at all, and those three were ranked as the three most reliable. The two best were already named: InsiderAdvantage and Susquehanna. The third “most accurate” spot was held by Trafalgar Group, with a 2.7 percentage point average miss and 86 percent bias toward Republicans.
“Let me tell you the polls that count, and those are the polls a couple of weeks before the election,” claimed Rush Limbaugh. “That’s when the pollsters worry about holding onto their credibility. Those are the polls that everybody remembers.”
That was probably true when he said it. But today’s regime media lets bad and biased pollsters worry less and move on. In 2028 we’ll be reading about Selzer and all the rest of the gold standards as if nothing happened.