Michigan race ends with acrimony and funky polls — again

Also in Washington, Senate Minority Leader Chuck Schumer (D-NY) said, “This is a despicable betrayal from Rep. Stevens.” And Neera Tanden, a party activist who was a senior adviser in the Biden White House, said, “Standing with Mike Rogers is a betrayal of everything Haley Stevens has fought for her entire career. It’s also a betrayal of all of her supporters that believed her.”

Stevens also took a beating in the press. In an appearance with CNN’s Anderson Cooper, she found her motives and reasoning challenged so aggressively that she said, “It sounds like you’re upset with me, too, Anderson.” Indeed it did.

Stevens surely knew how her party would react, but perhaps she thought she might receive better treatment in the media. After all, look at the praise lavished on former Wyoming Republican Rep. Liz Cheney when she broke with her party in 2024 and appeared with the Democratic presidential candidate, Vice President Kamala Harris. But that was then — remember that Cheney was turning on Donald Trump — and this is now. If Haley Stevens thought she would receive the media’s “Liz Cheney Country Over Party” award, she was sadly mistaken.

On Saturday, before Stevens’ announcement, el Sayed said he did not want Stevens to attend a big party unity event on Sunday. “I don’t want any distractions,” el Sayed said. Let’s just say that el Sayed did not get his wish.

The poll immediately seemed like an outlier. Seven recent polls listed in the RealClearPolitics average showed el Sayed with a lead from 1 to 7 points. The New York Times had it at 11, which felt like it was in line with the unreliable polling that has bedeviled this race and others.

In the Democratic primary, el Sayed went into election day with a 10.2-percentage-point lead over Stevens, according to the RealClearPolitics average. In the actual election results, el Sayed won by exactly 1 point.

Even before primary election day, when el Sayed was contending not only with Stevens but with Michigan state legislator Mallory McMorrow, unreliable polling played a part in the race. In late June, a poll appeared seemingly out of nowhere from an outfit called “Wedgewood Polls” showing el Sayed at 46% and Stevens at 39%, while McMorrow’s support, which had been in the mid-teens and 20s, fell to 9%. Calls arose for McMorrow to quit the race, which she did, leaving it to el Sayed and Stevens.

Later, Wedgewood Polls published a statement on X revealing that the organization was fake and the poll was fake. It was all an “experiment.” In a post on its now-deleted account, the group wrote, “We wanted to test a simple hypothesis: How easy is it for an entirely fabricated polling firm to gain traction, to be taken seriously, and expose how little scrutinization exists in media and public commentary? The short answer: Far too easy.”

Who was behind Wedgewood Polls? It’s still not clear; it might have been some Democratic dirty trick designed to boost el Sayed or Stevens. In any event, the immediate victim was McMorrow. “Here’s the impact of this ‘experiment,'” she wrote on X. “Donations dried up. Volunteers were devastated and pleading for guidance on what to do. Reporters hounded me with calls to drop out.”

So McMorrow left the race to el Sayed and Stevens. Now, there’s another poll, this time from a real polling outfit — Siena Research, working with the New York Times — showing el Sayed with another big lead. New York Times journalists seemed clearly nervous about it. Almost immediately after releasing the Michigan poll, along with polls from other states that showed Democrats with smaller leads, the paper published an article by its poll expert, Nate Cohn, with the headline: “The Final Polls Show a Blue Surge. Can We Trust Them?”

It was an extraordinary thing for a news organization to say about a big poll it had just released. In this case, the New York Times was worried about something known in the polling business as “partisan nonresponse bias,” a phenomenon in which supporters of one political party are less likely to respond to pollsters than those from the other party. For the last several years, the nonresponse has come from the Republican side — GOP supporters are less likely, sometimes far less likely, to cooperate with pollsters than Democratic supporters.

“This cycle, Democrats are once again responding to Times/Siena polls at a much higher rate than Republicans,” Cohn wrote. “Across these final battleground polls, registered white Democrats were about 30 percent likelier to take the poll than registered white Republicans. That’s actually a bit higher than in the final polls in 2020 … The persistence of this kind of nonresponse bias introduced the risk of systematic polling error like those in 2016 or 2020.”

To correct the nonresponse problem, pollsters use a technique called “weighting,” in which they estimate what each candidate’s support would be if voters from both parties responded in numbers that accurately represented their percentage of the electorate. They’ve been doing that for a while in an effort to correct the mistakes from 2016 and 2020 to which Cohn referred. And now, they’re weighting more than ever.

Cohn even revealed what the result of the new Michigan poll would have been if the weighting formula from earlier elections had been applied to the latest results. With the old formula, el Sayed’s lead would have been 16 points. Using the new formula, the New York Times pegs el Sayed’s lead at 11 points. They’re hoping that is right this time. But they don’t really know, which is why they publish all sorts of cautionary notes.

BYRON YORK: HAS NO TAX ON TIPS/OVERTIME/SOCIAL SECURITY HELPED THE GOP?

There are plenty of nonresponse problems in other key Senate states as well — Alaska, Kansas, and others. But Republican voter reticence might be particularly pronounced in Michigan because el Sayed has so often accused his critics of Islamophobia. He makes it seem that an inordinate number of the people who criticize him are doing so out of bigotry. So say you’re a Republican voter, you’re going to vote against el Sayed, and some polling company calls you up to ask your opinion on the race. Why talk to some stranger about it?

Of course, we’ll know the result soon. In the primary, el Sayed’s lead turned out to be one-tenth the size the polls suggested it was, but he still won. After Tuesday’s election, he’ll either be the next senator from Michigan or a new case study in partisan nonresponse bias.