Related Articles
Election forecasters think this year’s presidential race is too close to call. Nevada? A complete tossup. Pennsylvania? Anybody’s guess.
The race looks clearer to Seth Keshel, a Trump supporter and former Army intelligence captain who made his own forecasts for November. That’s in part because, unlike other election modelers, he factors in widespread election fraud — something that officials say has not happened in modern elections — into his predictions, skewing them in favor of former President Donald J. Trump.
The unusual work has made him a celebrity among election deniers, even earning him a 40-state speaking tour in front of conservative crowds and recognition from Mr. Trump, who has lauded him as “highly respected.”
Pennsylvania? It’s going to Mr. Trump, Mr. Keshel predicts, partly because Democrats failed to pass election laws that he says could “rig” the race in favor of Vice President Kamala Harris. Michigan? Mr. Trump could lose there, he said, partly because the Democrats control all three levels of government, giving them the ability to enable widespread fraud.
“I would say it’s clear that Trump is performing better than he did in 2020,” Mr. Keshel said in an interview.
Election deniers like Mr. Keshel have spent four years since Mr. Trump’s electoral defeat in 2020 blaming the loss on a conspiracy theory involving voter fraud conducted through mail-in voting, ballot harvesting and automatic voter registration. They continue to push the idea even though courts and election officials have repeatedly refuted and dismissed the claims. In truth, voting fraud in the United States is extremely rare.
Mr. Keshel’s influence illustrates how deeply electoral conspiracy theories have seeped into Republican politics, so much so that even election forecasting — normally a field grounded in data — is influenced by the elaborate fiction of widespread fraud.
Forecasting experts who reviewed Mr. Keshel’s work said that, like many election deniers, he appeared to select data to support his deep belief that Mr. Trump should win and ignored opposing data.
Forecasters use reams of information to predict what will happen on Election Day, injecting their own ideas for how each data point should be weighed. Most of them, including Nate Silver, the modeler behind the Silver Bulletin, are giving a slight edge to Ms. Harris after feeding their complex models copious data on polling, economic indicators, historical election results and other data. An average of various polls calculated by the election data team at The New York Times also shows that most of the swing states are too close to predict.
Mr. Keshel instead predicted that Mr. Trump would win at least 281 Electoral College votes to Ms. Harris’s 226, with Georgia and Michigan remaining impossible to predict.
Keshel vs. the Polls
Mr. Keshel bases his calculations on widespread voter fraud, a widely debunked claim. He tends to see a race that favors Mr. Trump even more than an average of various polls calculated by the election data team at The New York Times.
#g-table-box {
max-width:600px;
}
#g-table-box ,
#g-table-box .g-artboard {
margin:0 auto;
}
#g-table-box p {
margin:0;
}
#g-table-box .g-aiAbs {
position:absolute;
}
#g-table-box .g-aiImg {
position:absolute;
top:0;
display:block;
width:100% !important;
}
#g-table-box .g-aiSymbol {
position: absolute;
box-sizing: border-box;
}
#g-table-box .g-aiPointText p { white-space: nowrap; }
#g-table-Artboard_1 {
position:relative;
overflow:hidden;
}
#g-table-Artboard_1 p {
font-family:nyt-franklin,arial,helvetica,sans-serif;
font-weight:regular;
line-height:19px;
height:auto;
opacity:1;
letter-spacing:0em;
font-size:16px;
text-align:left;
color:rgb(55,116,177);
top:1.3px;
position:static;
text-transform:none;
padding-bottom:0;
padding-top:0;
mix-blend-mode:normal;
font-style:normal;
}
#g-table-Artboard_1 .g-pstyle0 {
font-weight:700;
height:19px;
color:rgb(0,0,0);
position:relative;
}
#g-table-Artboard_1 .g-pstyle1 {
font-weight:500;
line-height:15px;
height:15px;
font-size:13px;
text-align:center;
color:rgb(119,118,119);
top:1px;
position:relative;
}
#g-table-Artboard_1 .g-pstyle2 {
font-weight:600;
height:19px;
color:rgb(0,0,0);
position:relative;
}
#g-table-Artboard_1 .g-pstyle3 {
font-weight:700;
height:19px;
text-align:center;
color:rgb(185,63,61);
position:relative;
}
#g-table-Artboard_1 .g-pstyle4 {
font-weight:700;
height:19px;
text-align:center;
position:relative;
}
#g-table-Artboard_1 .g-pstyle5 {
font-weight:700;
height:19px;
text-align:center;
color:rgb(230,153,49);
position:relative;
}
#g-table-Artboard_2 {
position:relative;
overflow:hidden;
}
#g-table-Artboard_2 p {
font-family:nyt-franklin,arial,helvetica,sans-serif;
font-weight:regular;
line-height:19px;
height:auto;
opacity:1;
letter-spacing:0em;
font-size:16px;
text-align:left;
color:rgb(55,116,177);
top:1.3px;
position:static;
text-transform:none;
padding-bottom:0;
padding-top:0;
mix-blend-mode:normal;
font-style:normal;
}
#g-table-Artboard_2 .g-pstyle0 {
font-weight:700;
height:19px;
color:rgb(0,0,0);
position:relative;
}
#g-table-Artboard_2 .g-pstyle1 {
font-weight:500;
line-height:15px;
height:15px;
font-size:13px;
text-align:center;
color:rgb(119,118,119);
top:1px;
position:relative;
}
#g-table-Artboard_2 .g-pstyle2 {
font-weight:600;
height:19px;
color:rgb(0,0,0);
position:relative;
}
#g-table-Artboard_2 .g-pstyle3 {
font-weight:700;
height:19px;
text-align:center;
color:rgb(185,63,61);
position:relative;
}
#g-table-Artboard_2 .g-pstyle4 {
font-weight:700;
height:19px;
text-align:center;
position:relative;
}
#g-table-Artboard_2 .g-pstyle5 {
font-weight:700;
height:19px;
text-align:center;
color:rgb(230,153,49);
position:relative;
}
Swing states
KESHEL’S RATING
NYT POLLING AVG.
N.C.
Leans Trump
Trump <1
Pa.
Leans Trump
Harris <1
Nev.
Leans Trump
Harris <1
Ga.
Tossup
Trump <1
Mich.
Tossup
Harris <1
Wis.
Leans Trump
Harris <1
Ariz.
Leans Trump
Trump +2
Other states
KESHEL’S RATING
NYT POLLING AVG.
N.M.
Leans Harris
Harris +8
Va.
Leans Harris
Harris +8
N.H.
Leans Harris
Harris +7
Neb. C.D. 2
Leans Harris
Harris +8
Swing
states
KESHEL’S
RATING
NYT POLLING
AVERAGE
N.C.
Leans Trump
Trump <1
Pa.
Leans Trump
Harris <1
Nev.
Leans Trump
Harris <1
Ga.
Tossup
Trump <1
Mich.
Tossup
Harris <1
Wis.
Leans Trump
Harris <1
Ariz.
Leans Trump
Trump +2
Other
states
KESHEL’S
RATING
NYT POLLING
AVERAGE
N.M.
Leans Harris
Harris +8
Va.
Leans Harris
Harris +8
N.H.
Leans Harris
Harris +7
Neb.
C.D. 2
Leans Harris
Harris +8
Mr. Keshel’s analysis starts by examining how each county performed in previous elections dating back to 2020. He tends to see Mr. Trump’s success in rural counties as genuine, while President Biden’s gains in urban counties were often the result of fraud, like ballot-stuffing, that officials have repeatedly said did not occur. His report from 2020 speculated that Mr. Biden received more than eight million “excess votes” that could be attributed to fraud, despite election officials’ findings that such fraud did not happen.
In some states, Mr. Keshel predicted that Mr. Trump will find even greater support from deeply red counties, while siphoning Black and Hispanic voters away from Ms. Harris.
But in other states, like Michigan, he predicted that Mr. Trump’s apparent surge of support in rural counties will be overwhelmed by fraudulent votes from cities. He tweaks the expected vote totals in each county accordingly, giving Ms. Harris an edge in some states.
He bases much of his other analysis on a dubious data point: voter registrations. In some states, like Pennsylvania, Democratic voter registrations are declining. But other election modelers tend to view the data as far less reliable than polling.
“Sure, sometimes voter registrations do point to a real change in the state,” said Lakshya Jain, an election modeler behind the analytics website Split Ticket. “But also, if that were the case, it should be reflected in public opinion polling.”
After predicting a Trump victory this November, Mr. Keshel has publicly joked about receiving a role in Mr. Trump’s next administration, perhaps as the “director of election reform” so he could push to outlaw many of the processes that make elections easier, like automatic registration and early voting.
He also suggested, though, that the past four years have been grueling, and said that, should Mr. Trump lose in November, his appetite for questioning the results was waning.
“If the 2024 election suffers the same fate as the 2020 and 2022 elections,” he added, “I will not be barnstorming the country presenting on how the election was stolen.”
<
