Eleven months after the 2020 American presidential election, the official results remain so incongruous, they merit an empirical exegesis.
The political establishment’s narrative is that Biden won an unexpectedly close race, and the outcome requires no further examination. Yet, Biden’s victory is so statistically suspicious, so riddled with ahistorical outcomes, that a detailed data examination is necessary to reassure Americans the official result was, in fact, the actual result.
Official tallies record 161.3 million votes cast in 2020. Donald Trump got 75 million of those votes, 12.1 million above his 2016 total, and the most votes ever received by an incumbent president. Joe Biden received 81.2 million votes, the most votes for anyone who has sought the presidency.
Biden received 306 Electoral College (EC) votes to Trump’s 232. The individual vote totals behind that victory show an amazingly slim margin of victory for Biden. He won Arizona (11 EC votes), Georgia (16 EC votes), and Wisconsin (10 EC votes) by a combined 43,809 votes, which made the difference between victory and an Electoral College tie.
Now, let’s consider nine categories of suspicious anomalies that led to Biden’s squeaker of a victory.
I. Census Bureau Data
In 2020, the Census Bureau found 5 million fewer voters than the number of ballots counted. This is the largest gap recorded since these post-election surveys began in 1964. These 5 million excess ballots account for most of Biden’s national popular vote lead. To cite one state-level example, the Census Bureau found 4.8 million voters in Georgia, but Georgia reported 5 million counted ballots.
The Census Bureau’s validated voter survey is a very thorough and comprehensive piece of post-election data analysis. Historically, it has been far more accurate than exit polling and other post-election surveys and studies, as Robert Barnes, a leading political analyst and successful political prognosticator, explained in early May on his “What Are the Odds?” podcast.
The nationwide excess of counted ballots over registered voters in 2020 is extremely unusual. Census data usually finds a very small differential between the number of people they identify as having voted in the previous presidential election and the official total number of ballots counted in that election. In 2016, Census voting data matched almost precisely the number of ballots counted.
Historically, when Census data has differed from the official ballot count, it has tended to overestimate, rather than underestimate, the number of voters. The opposite was the case in the 2020 election.
Most revealingly, the Census data shows the turnout surge was almost exclusively among White blue-collar voters, an overwhelmingly pro-Trump cohort. Yet, somehow, the surge favored Biden in the end.
Turnout in 2020 was 6.7 percentage points higher than in 2016. The Census data on overall turnout, and turnout among specific demographic groups, closely aligns with the macro- and micro-turnout predictions made respectively by Barnes and Richard Baris, the preeminent pollster and managing director of Big Data Poll, and polling data at my firm, Democracy Institute, which forecast a Trump win.
II. Predictive Metrics
During any presidential campaign, a number of leading indicators foretell the election outcome. The mainstream media focuses almost exclusively upon approved media and academic polling, which historically has a mixed record. Unlike the independent polling conducted in 2016 and 2020 by Baris, Robert Cahaly, Rasmussen, Susquehanna, and the Democracy Institute, respectively, the mainstream pollsters sponsored by the major television networks and news organizations performed terribly in 2016 by predicting a Clinton win, and even worse in 2020, predicting a Biden landslide.
“The 2020 polls featured polling error of an unusual magnitude,” a report by the American Association for Public Opinion Research concluded. “It was the highest in 40 years for the national popular vote and the highest in at least 20 years for state-level estimates of the vote in presidential, senatorial, and gubernatorial contests.” The average error in the polls just two weeks before the 2020 election was 4.5 percentage points nationally and 5.1 percentage points in state-level polls.
In a normal election, when the big polls err, non-polling metrics do not. These include party registration trends; the number of votes the candidates received during their party’s primary election; voter enthusiasm levels; the number of (especially small) donors; social media followings; broadcast and digital media ratings; online searches; the number of candidate lawn signs; campaign merchandise sales; and the number of individuals betting on each candidate.
Every non-polling metric forecast Trump’s reelection, and these non-polling metrics have historically had a 100 percent record in indicating who will be president—until 2020. Consider: for Trump to have legitimately lost the election, the mainstream polls needed to be correct, which they were not. Furthermore, for Trump to lose, not only did one or more of these non-polling metrics have to be wrong for the first time ever, but every one of them had to be wrong, and all at the very same time.
This is not, strictly speaking, impossible, but it is extremely unlikely.
III. Voter Registration
Historically, the registration-by-party trend is a very reliable predictor of the election outcome. New and crossover registrations are the best proxy for both voting intention and actual voting behavior. In 2020, Republicans achieved massive registration gains vis-à-vis the Democrats across almost all competitive states. Notably, in 2020 voter registrations in the swing states of Michigan, Pennsylvania, and Wisconsin—won by Trump four years earlier—each trended toward more Republicans registering, foreshadowing an even more favorable electoral environment for Trump than in 2016.
The higher Republican registrations were the continuation of what happened in 2016. In Pennsylvania, for example, 2016 registrations had trended toward the Republicans in 60 of 67 counties. Unsurprisingly, when those newly registered voters cast their ballots, 62 of those 67 counties trended toward Trump, explaining his statewide victory over Hillary Clinton.
But strangely, the same trend didn’t produce the same result in 2020. Statewide registration trended 3 percent to the Republicans between the two presidential elections, with Republicans registering 242,000 net new voters, compared to just 12,000 for the Democrats. This gave the Republicans a massive 21-to-1 registration advantage over the Democrats in Pennsylvania between 2016 and 2020.
Hence, last year saw 60 of 67 Pennsylvania counties trend toward the Republicans in registrations, foretelling a larger Trump win statewide than in 2016, as historically Pennsylvania along with the swing states of Florida and North Carolina has always trended towards the party that made overall registration gains.
However, when the ballots were cast, only 20 of those 67 counties trended toward Trump—which is the opposite of what one would expect. The official results reported Biden winning Pennsylvania with 50.01 percent of the recorded votes.
IV. Bellwether Counties
Since 1980, 19 counties have voted for the winner of every presidential election. This 40-year trend of accuracy was unbroken, until 2020.
These bellwether counties are: Washington County, Maine; Essex County, Vt.; Otsego County, N.Y.; Cortland County, N.Y.; Ottawa County, Ohio; Wood County, Ohio; Shiawassee County, Mich.; Van Buren County, Mich.; Vigo County, Ind.; Sawyer County, Wis.; Juneau County, Wis.; Marquette County, Wis.; Richland County, Wis.; Bremer County, Iowa; Warren County, Ill.; Westmoreland County, Va.; Valencia County, N.M.; Hidalgo County, N.M.; and Clallam County, Wash.
In 2020, Trump won 18 of these 19 counties, but lost the election.
Noteworthy, too, is Luzerne County, Pa., which has been won by the statewide winner in every presidential election since 1936. Republican candidates who have won Luzerne County have won the White House six out of seven times and have always won Michigan, too. Also worth noting is Mahoning County, Ohio, which had only been won twice by a Republican candidate since 1936, until Trump carried the county in 2020. Oddly, Trump lost Mahoning County in 2016—the election he won.
In 1956, President Eisenhower won Luzerne and Mahoning counties and won 41 states. In 1972, President Nixon also won both counties and won 49 states. Simply put, Republican presidential candidates have always won national elections when they win both of these counties.
V. Bellwether States
Since 1896, the winner of Ohio has won the presidency on all but two occasions. Trump won Ohio easily in 2020. When Ohio and Iowa have been won in tandem, as Trump did in 2020, the winner has won 24 of the past 26 presidential elections; 1944 and 1960 were the only exceptions, and the 1960 election was stolen from Nixon. When Florida, Iowa, and Ohio vote for the same presidential candidate, that candidate has won the White House 26 times. The two times the candidate won those three states and somehow didn’t win the White House are Nixon in 1960 and Trump in 2020.
Trump won Florida by 3.3 percentage points, and won both Iowa and Ohio by more than 8 percentage points, yet he apparently lost both Arizona and Georgia. Since 1932, when North Carolina and Florida have been won in tandem, as Trump did in 2020, the candidate who won them has won the national election 17 times out of 18 (the exception being George H. W. Bush in 1992). When won together, Iowa, Ohio, and North Carolina have a perfect record identifying the national presidential winner since 1896. Perfect, that is, until 2020.
Since 1932, Michigan and Pennsylvania have voted with Florida, whether Florida has drifted Republican or Democratic. In 2020, Trump won Florida by a comfortable margin, and by a considerably larger margin than in 2016, yet he did not go on to win either Michigan or Pennsylvania.
In 2020, Trump won Florida, Iowa, North Carolina, and Ohio. Before 2020, on the 13 occasions when the same candidate won these four states, that candidate won the presidential election.
VI. Incumbent Performance
Trump’s 2020 vote total was 12.1 million higher than four years earlier. Since 1892, no incumbent president has gained votes and lost reelection, until 2020.
In fact, both President Obama and President Franklin Roosevelt lost votes but were comfortably reelected. Obama lost almost 4 million votes between his 2008 and 2012 elections. The six presidents who lost their reelection bids all had fewer votes in their second election than in their first election.
In 2020, Trump gained 5 percent among female voters, enabling him to perform very well in swing states and counties. Trump experienced record vote totals and increases in Arizona, Georgia, Michigan, Minnesota, Nevada, Pennsylvania, and Wisconsin—yet he somehow lost each state.
For example, Phoenix-centered Maricopa County, Arizona, which holds two-thirds of the state’s voters, had not voted Democrat at the presidential level since 1948. After winning the county by 3 percent in 2016, Trump gained 248,000 net new votes in 2020, which is a record Republican gain. Yet, Trump lost Maricopa County.
The Pew Research Center’s analysis of validated 2020 voters is very revealing. Pew helpfully divides voters by generation. Compared with 2016, Trump either held his own in 2020 or improved with every generational cohort. His most notable gains were among younger voters, members of the millennial and Zoomer generations.
Yet, late on election night, when the ballot counting restarted with huge dumps of mail-in ballots in key suburban and exurban precincts in Michigan, Pennsylvania, and Wisconsin, these are the young voters who, according to exit poll data, allegedly lifted Biden above Trump.
Puzzling data is also found within Trump’s impressive performance among minority voters. In 2020, Trump garnered a Republican candidate’s largest share (more than one-quarter) of the minority vote since Nixon in 1960. Trump improved nationally, including in urban Midwestern areas such as Wayne County, Michigan, Florida’s Miami-Dade County, and throughout southern Texas, where he flipped traditionally Democratic counties.
Trump won 12 percent of the black vote, nearly doubling the 6 percent he won in 2016. Nearly one in five black men voted for Trump. Biden badly underperformed with black voters, securing 87 percent of their votes, a reliable harbinger of a poor electoral outcome for a Democrat.
Pew’s data confirmed Trump’s remarkable gains among ethnic minority voters, especially Hispanics. The survey found 38 percent of Hispanics voted for Trump; in 2016, 28 percent voted for him. Trump lost blue-collar Hispanics by only 14 points, a record low shortfall for a Republican. Trump’s significant Hispanic growth was keenly demonstrated among Tejanos in southeast Texas and Venezuelan- and Dominican-Americans in south Florida. Such Hispanic support for a Republican has historically made it borderline impossible for a Democrat to win Arizona, Nevada, and New Mexico—and yet, somehow, Biden won those states in 2020.
Trump also gained new Asian voters, especially in areas such as Orange County, California. Overall, Trump’s minority vote performance was so stellar, it was arithmetically impossible for Biden to win the election given Biden’s—and any national Democratic candidate’s—dependence upon support from at least 90 percent of blacks and two-thirds of Hispanics.
Despite the loss of these minority voters, somehow Biden outperformed Barack Obama as a presidential candidate. Biden won a record-low 17 percent of counties nationwide. He only won 524 counties, as opposed to the 873 counties Obama won in 2008. While simultaneously bleeding minority voters and losing every bellwether county bar one, Biden secured 12 million more votes than Obama at his 2008 peak.
Did a wave of white “Biden Republicans” save the Democratic candidate? Hardly. Ninety-five percent of Republicans voted for Trump, helping him win the white suburban vote nationally and increasing his vote in Republican-voting, white suburban counties. Trump simultaneously increased his vote in white blue-collar counties, especially among rural male blue-collar whites, particularly in Michigan, Minnesota, Ohio, and Pennsylvania. Trump even flipped some traditionally Democratic counties in these states in 2020, and yet somehow lost these states overall.
One possible explanation for Biden’s record haul of votes is provided by Seth Keshel, a former U.S. Army military intelligence officer, who conducted exhaustive trend analyses of contemporary population growth, voter behavior, and party registration data in every state. His data demonstrates that 2020’s results defied every historical and contemporaneous trend. According to Keshel’s empirical analysis, Biden’s national popular vote total was at least 8.1 million above what was arithmetically possible for him to attain.
VII. Primary Vote
Another highly reliable presidential election outcome forecast metric is the candidates’ respective shares of primary votes. Since presidential primaries began in 1912, incumbent presidents who have lost reelection each received between 36 percent (Herbert Hoover) and 72.8 percent (George H. W. Bush) of their respective party’s primary support. Presidential landslides were won by incumbent Republicans with primary support from 85.9 percent (Dwight Eisenhower) to 98.9 percent (Ronald Reagan).
In 2020, Trump received 94 percent primary support. In striking contrast, Biden suffered heavy, early primary defeats in Iowa, New Hampshire, and Nevada, and was rescued by winning South Carolina.
Professor Helmut Norpoth’s primary vote model correctly picked the presidential winner every time from 1912 to 2016, except in 1960 (a stolen election) and 2000, when popular vote winner Al Gore lost at the Supreme Court. As Norpoth explains, the primary vote model “is a statistical model that relies on presidential primaries and an election cycle as predictors of the vote in the general election…. Winning the early primaries is a major key for electoral victory in November.”
Norpoth stated: “What favors Trump in 2020 as well is the cycle of presidential elections operating for nearly 200 years…. After one term in the White House, the incumbent party is favored to win re-election, unlike the situation when it has held office for two or more terms.” Norpoth’s primary model gave Trump a 91 percent chance of beating Biden.
VIII. Presidential Coattails
Presidential elections influence the partisan composition of the House of Representatives in favor of the winning candidate’s party. Winning candidates, especially challengers, carry down-ballot congressional candidates over the finish line. Similarly, down-ballot voting is a reliable indicator of success for the top of the ticket.
Some examples: In 1980, Reagan’s victory saw Republicans gain a net 34 House seats; in 1984, Reagan’s reelection saw a net Republican gain of 16 House seats; in 2008, Obama’s win saw his Democrats gain a net of nine House seats; and, in 2016, Trump’s narrow victory saw his party gain House seats while winning all the toss-up races.
In 2020, Biden apparently won the national popular vote by several million votes, yet congressional Republicans flipped 13 Democratic seats, did not lose a single Republican-held seat, and won all 27 toss-up contests. Furthermore, Trump’s party did not lose a single state legislature and made gains at the state level.
It is extremely rare for this to happen. In 82 percent of the elections since 1932, the party of the candidate who wins the national popular votes gains seats in the House.
IX. Enthusiasm Gap
Another reliable predictor of presidential election outcomes is the enthusiasm gap between the candidates’ supporters. The more enthusiastic a supporter, the more likely he or she is to vote, volunteer, or donate. Trump benefitted from this advantage in 2016, as Obama had in 2008 and 2012.
above left: Senator Bernie Sanders (I-VT) campaigns for Joe Biden at a VoteNow rally in Lebanon, New Hampshire on Oct. 3, 2020 (Twitter/Laura O’Neill)
above right: President Donald Trump speaks at a rally in Duluth, Minn. on Sept. 30, 2020 (Dan Scavino/Twitter/Wikimedia Commons)
Our final Democracy Institute/Sunday Express poll found Trump with an enormous 34-point enthusiasm gap over Biden. To confirm this advantage, one could measure: the Trump campaign’s gargantuan grassroots volunteer army; the candidates’ respective social media, digital, and broadcast audiences; the crowd sizes at their campaign events; the Trump yard signs that carpeted much of the country; the merchandise sales skewed heavily toward Trump; and the unparalleled pro-Trump boat and yacht flotillas and miles-long pro-Trump car parades.
A synthesis of the empirical evidence, innumerable anomalies, and predictive metrics leads to an inescapable conclusion about the 2020 presidential election. Although it is statistically possible that Biden won, clearly it is statistically implausible that he actually did. As Biden, himself, stated in a summer speech in Philadelphia, “It’s no longer just about who gets to vote…. It’s about who gets to count the vote.”