The Washington PostDemocracy Dies in Darkness

The election-fraud crew loves polls now

Kari Lake holds her phone with former president Donald Trump on the line during a rally on Jan. 29. (David Blakeman for The Washington Post)
6 min

Republican Kari Lake came up short in her bid for governor of Arizona by a narrow margin last November, landing about 17,000 votes behind out of 2.6 million cast. Lake sued to block the swearing-in of her opponent, Democrat Katie Hobbs, but that lawsuit was thrown out. Her claims that her voters were unable to cast votes were rejected by the voters themselves.

Meanwhile, other statewide races, including the edge-of-a-razor contest for attorney general, were subject to automatic recounts that didn’t affect the results. Lake’s loss wasn’t close enough for a recount but, given that Republicans didn’t gain enough to win in narrower losses, it’s unlikely she’d have made up the required gap.

In other words, Lake lost. She doesn’t like to admit it and, in fact, hasn’t. Instead, she has continued to rail about purported fraud and misconduct in her race as feverishly as she did for years about the 2020 election — a loyalty to former president Donald Trump’s worldview that helped her earn his enthusiastic support in her own race. But, all of that said, she still lost.

On Friday, though, Lake and other allies of Trump seized on a new poll to argue that she hadn’t lost; that, instead, she’d won. This poll, they claimed, offered strong evidence that she’d not only won but won easily.

It does not. And since that is insufficiently blunt, allow me to expand: It not only doesn’t show that, it is a flawed poll from a deeply partisan company that very obviously conflicts with reality.

Sign up for How To Read This Chart, a weekly data newsletter from Philip Bump

To its credit, this is the second time in the past month that the firm, Rasmussen Reports, has been in the news. The first time was when cartoonist Scott Adams seized on one of its partisan, leading surveys to announce to the world that henceforth he planned to avoid Black people. Now we have this new poll, in which the company purports to show (but in no way actually shows) that Lake was the real election winner.

The survey was sponsored by a group called College Republicans United, a sponsorship that Rasmussen’s Mark Mitchell explained to Stephen K. Bannon — yes, him — included reaching out to Rasmussen to conduct this specific poll. The impetus, Mitchell explained, was a different poll that showed broad confidence in the results of Arizona’s elections. College Republicans United got their money’s worth, with Rasmussen reaching a different conclusion.

Predictably. The firm has always leaned right but, in recent years has adopted an aggressively partisan position that’s apparent in the question it asks. Its focus on “likely voters,” even well before actual elections, seemed to lead it to weight its results in a way that advantaged Republicans. But now the firm trumpets its partisanship explicitly, perhaps hoping that other right-leaning groups will bring it sponsored polls to conduct.

A tweet from Rasmussen posted during the Jan. 6 riot at the Capitol.

Speaking to Bannon, Rasmussen’s Mitchell explained how they came to the conclusion that, actually, Kari Lake did win her race.

“A lot of pollsters, ones with maybe less courage, would really use those results to weight the poll to match the outcome of the elections,” he said, “and then bury the questions and pretend that they never even asked them. And, you know, to be honest, these results were just so unbelievable that we had to report out.” Instead of weighting the responses to the results, he said, they weighted it to exit polls.

“What we found,” he added later, “is that voters told us that just four months ago they elected Kari Lake as the governor of Arizona by 8 points.”

Let’s set aside for a moment the broader questions about Rasmussen’s objectivity. Let’s just consider this issue of weighting.

Since the people pollsters speak to when they conduct a poll are never a perfect representation of the population, they weight the responses they get. If you have too few Democrats relative to what you expect, for example, you might give their answers more weight. If you wanted to evaluate the views of the people who voted in an election, it makes sense to weight your responses so that the results mirror the actual vote result. But Mitchell scoffs at that, instead weighting the results to exit polls, meaning that the composition of the responses is structured to look not like the results but like the turnout.

Beyond questions about the accuracy of exit polls — which are themselves weighted to election results, mind you — consider what those exit polls found. There were more Republicans who turned out than Democrats, which Mitchell says is roughly the weighting they used. But a plurality of voters were independents, and they preferred Hobbs by 7 points. Among Republicans, meanwhile, the exit polls from Edison Research show that about 1 in 11 also preferred Hobbs, twice the rate at which Democrats voted for Lake.

In Rasmussen’s results, independents voted for Lake by 14 points while 1 in 6 Democrats and Republicans crossed party lines to support the other candidate. This is simply not credible, even setting aside the fact that we know what actually happened.

Asked to comment, Rasmussen had no immediate response.

Bannon wasn’t going to push back on this, certainly, so Mitchell pressed on, arguing for the accuracy of his firm’s polls.

“If you average all of our September and October generic ballot polling, we were off 1.2% from the official reported National House popular vote,” he said, which is an odd metric to use. In fairness, Rasmussen’s final generic-ballot poll was much better last year (putting the GOP up 5 when they ended up 2.8 points in House voting) than in 2018 when they had the Republicans up by 1 point in an election where Democrats won the national House vote by more than 8 points. Also, Mitchell said, “we nailed the Arizona Republican primary at the end of the summer” — perhaps not surprising given their tendency to overrepresent Republican voters. (Rasmussen didn’t poll in the gubernatorial general election last year.)

Mitchell didn’t need to get into the numbers much to convince people like Lake (who shared the results of the poll on Twitter multiple times) or Donald Trump spokeswoman Liz Harrington. Harrington announced that poll shows Lake likely won in “a landslide.”

The irony, of course, is that Trumpworld has long disparaged polls for their inaccuracies, their failures to capture the true sentiment of Republican voters. Here, though, this dubiously weighted poll from a partisan pollster is hailed as strong evidence that the election was somehow stolen. (How, of course, is never really explained.)

There’s an old saying in politics: The only poll that matters is the one on Election Day. Here, we have a new maxim: The only poll that matters is one conducted by a partisan firm using dubious methodology four months after voters actually weigh in.