The Washington PostDemocracy Dies in Darkness

No, there was not a secret scheme to swing Ohio to Biden (as evidenced by the fact that it didn’t)

Trumpworld’s favorite scientist continues to spread absolute nonsense on voter fraud

Former president Donald Trump arrives to speak at a rally at the Lorain County fairgrounds on June 26 in Wellington, Ohio. (Jabin Botsford/The Washington Post)
Placeholder while article actions load

Bear with us as we walk through a few numbers.

There are 88 counties in the state of Ohio. On average, about 82 percent of adults are registered to vote in those counties, as determined if we compare current registration figures from the political data firm L2 with the 2019 Census Bureau estimates. There are a few outliers such as Delaware County, just north of Columbus. It has a higher density of registered voters, which I’ll attribute for no good reason to the fact that it’s home to Ohio Wesleyan University, where my parents met.

In 2020, turnout among those registered voters varied by age, as it usually does. In 2015, we graphed turnout in California in the prior year’s midterms, finding that turnout was high among younger voters turning out for the first time, then quickly dropping off. Turnout increases for decades until it starts to decline again as a function, grimly enough, of death rates.

That’s what turnout in Ohio looked like, too. The black line below shows the values across the state, using L2 data (and current age figures) with the distribution of county turnout rates shown as different sized dots.

There are some outliers, but for the most part turnout in each county tracks with the state figure, just as it tracks with the graph from California seven years ago.

There’s a way to represent this numerically. Statisticians measure how closely related two sets of data might be using a figure called r, the correlation coefficient. An r of 1 means a perfect correlation; an r of 0 means there’s no correlation at all. As you might expect, the turnout rates by age in every county in Ohio track closely with the state rate for that age.

More-populous counties (which, as in most parts of the country, also tend to vote more heavily for Democrats) correlate more closely to the state rate for the simple reason that they contribute more to it. If 75 percent of your population lives in one county (not the case in Ohio, but just to make the point), the state figures are going to look a lot like that county.

In summary, then, about 8 in 10 adults in Ohio are registered to vote, and older Ohioans were more likely to vote last year. No particular county stands out as unusual in terms of its turnout pattern by age. And, of course, Donald Trump won the state by about 8 percentage points. All in all, literally nothing here to suggest any effort to throw the state for Joe Biden.

Unless you are Douglas Frank, the physicist who has become the go-to expert for those eager to argue that the 2020 election was stolen from Trump.

Over the weekend, Trump made his triumphant return to electioneering by giving a speech in a windy field in Lorain, Ohio, a bit west of Cleveland. He focused heavily on his false and/or debunked and/or noncredible claims of rampant voter fraud, as he’s been doing for, what? nearly eight months now. But this rally included a little something extra for the various geeks in the audience: a presentation from Frank in which he laid out his evidence that the election was stolen for Biden, including, somehow, in Ohio.

We’ve addressed Frank’s theories before. They are ludicrous. In the past, he’s applied his amusingly goofy system to states such as Michigan, where he is at least in the realm of “this alleged fraud swung the election.” (A review of Michigan’s election conducted by a Republican-led committee recently described Frank’s analysis as “not sound for several reasons.”) On Saturday, however, he was arguing that similar suspicious activity happened in Ohio.

Touting himself as “the guy that [My Pillow CEO Mike] Lindell discovered and has been featuring in his movies” about voter fraud, Frank claimed that counties first inflate their voter rolls to have extra voters to inject into the results. Then they use a “key” to create the results they want. And then they take the extra voters out of the registration. He admits that counties administer their own elections, so he appears to be claiming that each of Ohio’s 88 counties does this as part of some broad conspiracy, but he keeps that part vague.

“So do you think in Ohio we had a clean election?” Frank said at one point, earning a chorus of “No!” from the audience. Again, despite Trump winning by the exact same margin as he did in 2016.

The “key” is the important thing here. He does a lot of hand-waving about how he discovered the key for each state and how it’s a “sixth-order polynomial,” which sounds very fancy and impressive. But Frank’s “key” is actually just the functional equivalent of the state-level turnout rates. What he does is he takes a number of counties and averages the rates by age across all of them. Then he compares that average to other counties and — ohmigosh! — there’s a strong correlation. Well, sure, because there are common patterns to the rates at which different age groups turn out to vote. This is not new.

But: That’s it. That’s the claim he makes. As for the “sixth-order polynomial,” you can make one yourself. Just plug the data into Microsoft Excel and add a polynomial trendline, cranking the “order” value up to “6.” And, voilà.

See that complicated formula? That describes the dotted curve line. And if you think that I’m oversimplifying Frank’s work, I’m not.

“Look, it’s a smooth curve, okay?” Frank said Saturday. “Algebra students, are you ready? That’s called a sixth-order polynomial. That’s what it’s called. And it’s a simple equation and it’s the highest that Excel goes. So guess what? America was stolen by an Excel spreadsheet.”

Yeah, no.

It’s critically important to point out that Frank’s correlation is so high because he’s using this polynomial trendline. He’s choosing to depict the trend in a way that fits it as closely as possible to the data, and then using that close fit as evidence that something sketchy happened. When, again, he’s simply marveling at the fact that turnout patterns among age groups are consistent.

At no point does Frank wrestle with the fact that his “key” applies equally well to counties that preferred both Trump and Biden, much less to the results in the state. One would think that an effort to steal the vote for Biden would mean that there is more evidence of fraud in counties that backed Biden, but by his own admission that’s not what Frank found. It’s bizarre.

He makes other dubious claims, such as that “everybody” in a county is registered to vote, which isn’t true. It’s likely that he’s using old census data, as he has in the past. In 2010, for example, there were 124,000 people in Delaware County, below the 145,000 currently registered there. Using old data for growing places will give a false impression.

But Frank is heavily invested in this idea of “phantom voters.” In an interview with the Truth & Liberty Coalition, Frank claimed that he’d been working with a congressional candidate in Pennsylvania who was suspicious about having lost in 2020. The campaign asked him to estimate how many imaginary voters there had been, with Frank coming up with an estimate of 30 percent. The campaign decided to test that, it said, with a “list of 1,600 doors that we’re going to knock on to look for some of those phantom voters.”

The result?

“We found about 30-32 percent of the doors we knocked on had at least one phantom voter,” Frank claimed.

He provided no evidence for this claim, but let me be very clear about what is being alleged. Frank is saying that a congressional campaign picked out 1,600 doors (how isn’t clear) and then knocked on them. If you have tried to knock on doors to talk to specific people, you will understand how time- and resource-intensive this is. I was in Scranton, Pa., just before the election last year and tried to contact about 30 registered voters. I spoke with maybe seven at their homes, with most not being there and at least one having recently moved. Knocking on 1,600 doors is a massive chore — much less actually contacting people who live in those houses, much less getting people to talk to you about how they voted. The evidence of this happening that Frank presented to Truth & Liberty was two hypothetical encounters.

That Pennsylvania candidate, by the way, was Kathy Barnette, running in the state’s 4th District. (Barnette actually outperformed Trump.) Barnette and Frank met before the election. She interviewed him last summer as part of her campaign.

That conversation was about the coronavirus pandemic, with Frank again presenting complicated graphs and then picking various political points from their midst. Last June, the Associated Press debunked a viral Facebook post in which Frank had claimed that the virus was fading, which it wasn’t.

Put it this way: Like voters in Ohio, Frank has a predictable pattern of behavior.

Lenny Bronner contributed to this report.