Does this DIY statistical analysis explain away recent election results and prove Donald Trump won the 2020 election? No, that's not true: In at least one state addressed by the model, experts trained in the specifics of election data analysis say the Seth Keshel analysis is flawed in the same way prior attempts by non-experts have failed to prove Trump "should have" won Arizona. The methodologies applied by amateur political analysts have not been through the rigors of review by professionals and political scientists, and they make claims that ignore facts on the ground, says a veteran Arizona election data expert. For instance, voters are free to vote counter to their registration and often do in Arizona. Also, expert analysts provide their data and methods to allow independent testing of findings and this one provides no evidence, only a graphic declaring Trump won 8.1 million votes.
The recycled claims about election rigging appeared in an August 2, 2021, Telegram post (archived here) on the @RealSKeshel account, operated by a former Army captain under the header: "You asked - Merry Christmas." It continued:
Here is a tale of the tape for Excess Biden Votes based on trend analysis in the modern political era, considering population growth/decline, recent voter history, and registration information, including registration by party.
My estimates are always lenient, and do not account for cyber flipping of votes.
Users on social media only saw this title, description and thumbnail:
Paul Bentz, a veteran Arizona political pollster whose firm serves both Republicans and Democrats, said Keshel's numbers ignore the realities of politics in Arizona. In an August 4, 2021, email to Lead Stories, Bentz said one way to see Trump's 2020 problem was in the number of rallies:
It seems incredibly unlikely that Trump could somehow double his lead of four years prior in a much higher turnout election where several factors were all breaking against him. If the Trump campaign seriously believed they were winning Arizona by 200,000 votes, they would have not visited the state so frequently in the waning weeks of the campaign.
Bentz, who has two decades' experience being paid by corporate and campaign clients to measure public opinion in Arizona, said it was clear GOP registration efforts would not overcome the 2020 swing against Trump among independents:
Biden won Arizona by a little over 10,000 votes. According to the Keshel numbers, he allegedly received 210,000 "excess" votes which means Trump was supposed to win the state by nearly 200,000 votes. However, Trump only won Arizona in 2016 by a 91,000 vote margin - in a much friendly election (2016 was more Republican, older, and more rural) ...
Democrats and metropolitan voters overperformed in 2020 - our polling showed that both of which would eat into a Trump advantage, not help it. In addition, Trump was bolstered by Independent support in 2016, but all of our polling showed that the independent and unaffiliated voters were breaking against him (+3 Biden) as well ...
It's also important to note that Trump also suppressed his own turnout by casting doubts about early voting - a tool that the GOP has used effectively in Arizona for nearly three decades.
Keshel is described by Trump partisans as an election expert, although he provides no credential to indicate training in statistical or political analysis. Most of his recent tweets concern COVID-19 denialist claims and pro-Trump slogans.
As senior vice president of research and strategy at HighGround, Inc., Bentz is held accountable for tracking political trends for clients. Another consequential change for Trump, he said, was a shift to urban voters, among whom he loses his advantage. Here's how Bentz explained it (formatted for ease of reading):
In 2016:
- Turnout was 74% and the participation in that race was 38.8% GOP/32.1% Dem/29.1% Independent & Unaffiliated.
- A total of 2.66 million people voted of which 21.1% came from "Rural" Arizona.
- In 2016, Republicans held a 6.7-point advantage in all votes cast - and Trump won by 3.5 points.
In 2020:
- Turnout was 79.9% and the participation was 37.3% GOP/34.4% Dem/28.3% Independent & Unaffiliated
- A total 3.4 million people voted (three quarters of a million more people) of which only 18.6% came from "Rural" Arizona.
- The participation advantage for the GOP nearly cut in half.
Gary King, who is director of the Institute for Quantitative Social Science at Harvard University, said the problem with the Keshel claims is they are being posted with none of the material that could be used to check Keshel's work.
"In an area like this, you don't trust," said King, who is president of the Society for Political Methodology and vice president of the American Political Science Association. "You need the evidence. He's making predictions with some model without telling how he's done it."
King said even experts in quantitative analysis get it wrong, because voters aren't entirely predictable, but that experts test out their methods and models by sharing their materials for review by other experts. "The fact is that what counts is what's on the record, not who he is," King said. "I do get analyses from amateurs that are pretty good. This one is not ... There's no data with any provenance at all."
Keshel took to Twitter to mock those who ask him to document his claim.
"You'd better be prepared to show those numbers in court," shriek the leftist freaks who will stop at nothing to prevent audits of their fraudulent elections.
-- Seth Keshel (@SKeshel) August 3, 2021
In December 2020, Lead Stories reviewed a similar DIY political analysis, finding that spreadsheet analysts miss what Arizona political observers know: Voters don't stick to their registered party.
University of Arizona Prof. Chris Weber is a political science Ph.D. with more than 10 years' experience doing statistical analysis of political and social science data. He correctly predicted 2020 defections from the Republican voter base in an October 2020 New York Times op-ed sharing his findings that statewide, approximately 10% of Republicans indicated they would vote for Biden.
Looking at the earlier version of the kind of analysis Keshel appears to have assembled, Bentz said in December of 2020 it's a mistake to assume Arizona Republicans are party-line voters: "There's a reason why Trump was going to Bullhead City and to Goodyear and the places that he was going in the final weeks of the campaign here in Arizona. And that was to drive up margins in the dark red, heavily Republican areas because they knew they were behind in the metro areas. They knew they were trailing among Republican women...and that in some of the swing demographics, including independents, they weren't doing as well."
Bentz likes to remind those who claim election fraud to notice an important down-ballot race: the 2020 election was run by the Maricopa County recorder, a Democrat like Biden, who lost his own bid for re-election in 2020:
You'd think if there was someone gigging the machine that he might try to save his own job.
Weber said the other Arizona-specific event that likely affected loyalty was former astronaut Mark Kelly, running as a Democrat. Kelly easily won, with even more Republican support than Biden, to fill the seat long held by the late Sen. John McCain, the war hero Republican Trump quarreled with and disrespected. McCain was shot down on a bombing run over Hanoi, captured and tortured and held in a prison camp for five and a half years.
