Statewide Polling Dramatically Lowballs Republican Numbers and Enthusiasm – Again

Polling is at best an inexact science. At worst, it’s manipulative guesswork. So mistakes are inaccuracy inaccurate predictions are inevitable. But why do they seem always to favor Democratic candidates?

In an LA Times article on May 28, 2026, reporter Seema Mehta cited a poll conducted by her paper and the UC Berkeley’s Institute of Governmental Studies. She wrote, “Among likely California voters, 25% support Xavier Becerra, a Democrat and former Biden Cabinet secretary… . Republican Steve Hilton, a former Fox News commentator and British political strategist, has the backing of 21%, while 19% backed billionaire hedge fund founder turned environmental activist Tom Steyer, a Democrat.”

With 58% of the vote counted so far, the AP has it at 27.8% for Hilton; 25.4% for Becerra and 19.6% for Steyer.

With 42% of the vote outstanding these percentages can obviously change, but probably not by too much. It’s notable that the poll totally nailed it for Becerra (poll and actual both 25%) and Steyer (both 19%) but seems to have underestimated support for Hilton by more than 30% (poll 21 vs actual almost 28).

I read the methodology and I want to be clear that I’m not trying to bust the balls of Berkeley IGS Poll Director Mark Di Camillo or anyone else. In fact, that’s my whole point – no one’s doing this on purpose. The bias is baked in.

Reached for comment on his poll, Di Camillo told the Globe, “I would prefer to talk to you about the performance of the polls once all the results are in.”

Fair enough.

The LA Times used the Berkeley poll to launch a narrative regarding a late surge by Steyer. Casting Becerra as the likely top vote-getter, their unbylined pre-election “Analysis” article opined, “Until recently, it was assumed that Republican Fox News host Steve Hilton would also advance, especially after he was endorsed by President Trump. But a new poll suggested Hilton was in a tight race for second place with Democratic billionaire Tom Steyer, who is spending heavily from his own fortune.”

That Steyer surge never materialized, exactly as it failed to do in 2020 when he spent $253 million to come in 7^th in Iowa, 6^th in New Hampshire and 3^rd in South Carolina before dropping out with zero pledged delegates. (For those doing the math, it came to about $3373 per vote and my Globe story on Steyer running an epic bad campaign earned a coveted Drudge hit.)

And I do wonder why university and newspaper polls seem consistently to under-predict support for the Republican candidate.

On Saturday Nov 2, 2024, the Des Moines Register shocked the country with a survey by its respected pollster J. Ann Selzer. In a ruby red state so right-leaning that neither candidate had really contested it, Selzer showed Harris up by three points. Three days later, Trump won Iowa by 13 points. That’s not some kind of harmless miss. Showing that Iowa was suddenly in play caused tens of millions of dollars to flow into the Democratic candidate’s coffers to be spent before the Tuesday election. The results not only inspired the candidate who was reported to be 3 points ahead but worse, it demoralizes the voters of the candidate who incorrectly appeared to be struggling.

It wasn’t just wrongly predicting results it was interfering with those results.

Why did Hillary Clinton wake up on Nov 8, 2016 believing she’d win? Well, public polls showed her ahead nationally by 3–4 points and state polls in Michigan, Pennsylvania and Wisconsin had her as such a lock that she famously didn’t even visit Wisconsin in closing weeks. Even the 538 model, considered cruelly pessimistic by Hillary fans, gave her a 70% chance as voters headed to the polls.

Eight years later, with so much at stake and the first billion-dollar campaign behind Kamala Harris, she too believed victory was inevitable. After the Dobbs decision, Democrats were mobilized and their candidates had outperformed expectations in 2022 midterms and special elections. And all those “hidden Trump voters” that pollsters had missed in 2016 were supposedly accounted for. And yet … all seven swing states plus the popular vote went to Trump.

Why is it always thus? Why aren’t Republicans sitting on polls that show commanding leads only to discover on the day after E-Day that they underperformed?

This isn’t simply biased reporting after the fact. The danger here is that these predictions create self-fulfilling prophecies.

As something with experience not just as a journalist in mainstream media, but in partisan politics, I believe some of this can be blamed on the sense many pollsters share with those in the media. They simply don’t know any Neanderthals who would consider supporting these objectionable candidates. That doesn’t mean they ask the questions in a way that steers the results such as, “Who do you support, the sainted Hillary or the fascist Donald Trump?”

But the rejectionist tone seeps into the questions. Usually the order of the choices is randomized, but they’re asked alongside a slew of other questions. In the state of California, with its propensity for ballot initiatives, if a voter has just been asked three questions about “unfair CEO pay” and “tax billionaires” and then is asked which candidate to support, it’s not a surprise that support for the Republican would be meaningfully depressed.

Even Republican firms are prone to underestimating the size and enthusiasm of the Republican voter base, and it’s not only a Trump phenomenon. In 2014, GOP majority leader Eric Cantor’s private pollsters and all public polling simply didn’t see Tea Party candidate Dave Brat coming along.

If Di Camillo gets back to the Globe after the state takes its sweet time tabulating yesterday’s results, we’ll update this article. And question number one will be, Why was your poll dead on for Becerra and Steyer (and even nailed Mahan and Bianco and pretty close on Porter), but somehow missed a huge pocket of support for Steve Hilton?”