On Sunday, Seth Keshel released a new analysis of five Pennsylvania counties that he says alone hold “92,000 extra Biden votes… enough to flip the state!” Keshel is a former Army captain who specialized in military intelligence. He uses statistical analysis to review voter registration trends to predict future results and evaluate the likelihood that past results are organic.

The report claims that voting and registration data reveal that ‘phantom voters’ must be on the rolls in Pennsylvania. For instance, the publicly accessible data shows a 134-year-old man who voted in the 2020 election. Born in 1886, it appears this man registered to vote for the first time in 1998. According to Keshel, it appears the man “waited until he was 112 to finally get involved in the democratic process.”

In Erie County, Keshel says the data shows “40.2% of all voters registered for the 2020 election were registered on the 1st day of any given month,”—statistical probability should be around 1 in 30 or 3.33%. Therefore, the odds of the voter rolls being accurate and natural are so near to zero as to be certain they are not. Keshel acknowledges that “There may be a limited window to explain that voters get the first day of a month as a registration date if the info isn’t fully available.” Still, he says the registration data contains various years that go as far back as 1901.

If some procedure exists for associating a “placeholder” date with authentic voters for whom registration date is unknown, it was not applied consistently nor statewide. In Luzerne County, he says ~35.5% of voters were registered on the specific date of 2/5/1998. In Franklin County, ~22.5% of all registered voters had a registration date of January 1st of various years.

Keshel’s August 8 five-county report is based on Pennsylvania voter rolls available to the public from the Pennsylvania Department of State. The counties he reviewed are Erie, Luzerne, Franklin, Montgomery, and York. These counties are not the most populous in PA; together, they contain only five of the top-ranked cities among the top 20 in Pennsylvania by population (Erie 4th, Lower Merion 11th, Millcreek 12th, York 17th, and Wilkes-Barre 19th).

About the five-county report, he told his followers, “This is only useful if you guys do something locally with it, demand action, and pressure the living hell out of the PA RINO coalition to have these explained.”

All in all, Keshel estimates 92,000 excess votes for Biden in these five counties and 504k in all of Pennsylvania. If accurate, that would have given Trump a win by almost 6%.

Soon after the November 2020 General election, Keshel evaluated Pennsylvania and found “substantial deviance from statistical norms” when looking at voting patterns in Pennsylvania. He produced and signed an affidavit attesting to the results of that analysis, dated November 17, 2020, which were based on voting totals taken from the Decision Desk HQ unofficial tracker.

Keshel’s declaration says, “Party registration is the single most accurate predictor of a subsequent election,” and notes a consistent trend of both registration and election results in Pennsylvania changing in favor of Republicans over time since 2008. However, Biden’s victory came despite a continuing registration advantage for Republicans, inexplicably bucking that predictive trend.

Biden won Pennsylvania by only 1.2%. After looking at the numbers, Keshel wrote that what he found was “horrific” and that it was “Even worse than [he] thought.” On his Telegram, Keshel wrote“Joe Biden ‘won’ PA by 80k. That is 2k higher than Trump won WI, MI, PA in 2016. A sitting president doesn’t travel to PA to rage against audits with a lead of 80k. Joe Biden knows he did not win Pennsylvania.”

Keshel further showed in his 2020 declaration that Pennsylvania appears to have moved to the right since Barack Obama won there in 2008 by ~10%. Obama only won the state by ~5% in 2012, and in 2016 Trump just barely won by less than one percent. Keshel says this is part of an overall working-class voter shift away from Democratic nominees and toward Republican ones—a shift that also affected Wisconsin, Ohio, Michigan, and Minnesota.

In 2016, he had predicted Pennsylvania’s flip to Trump, and based on a “21 to 1 ratio… of net new registration in 4 years,” called 2020 a “slam dunk for Trump.” Per trends in voter registration, he said, “60 of 67 counties should have been more GOP than last time,” but it turned out to be only 20. He also noted that 2020 was “the first election in which [bellwether] Luzerne County failed to be won by the winner of the state” since 1936.

Keshel maintains a “one-stop-shop” Telegram channel called “Captain K’s Maps and Stats,” he appeared on the Dark to Light podcast for an interview on July 2, 2021, and he joined David Clements ‘Pennsylvania Election Audit Debrief’ July 26: