The truth behind these viral Trump ‘fraud’ graphs
Donald Trump and his Republican allies have launched a barrage of legal challenges attempting to flip election results in key swing states, with the President refusing to concede to Joe Biden until all avenues are exhausted.
Mr Biden was named the President-elect over the weekend after edging ahead of Mr Trump in several battleground states including Pennsylvania, Georgia and Michigan, bringing him to the 270 electoral college votes required for victory.
The Trump campaign hopes it can still turn things around through a combination of legal challenges, audits and official recounts, alleging significant voter fraud in a handful of Democrat-run urban centres was sufficient to push Mr Biden over the top.
While so far Republicans have not been able to provide proof of voter fraud, they are collecting evidence in the form of hundreds of sworn affidavits from witnesses to the alleged illegal conduct, a number of which have already been submitted in court filings.
In Michigan, for example, a lawsuit filed against Detroit election officials includes testimony from one poll challenger, Andrew Sitto, who claimed to have witnessed "tens of thousands" of suspicious ballots being brought in via the back door of the counting floor at the TCF Center in the early hours of November 4.
"I specifically noticed that every ballot I observed was cast for Joe Biden," he said in the affidavit. "While counting these new ballots, I heard counters say at least five or six times that all five or six ballots were for Joe Biden. All ballots sampled that I heard and observed were for Joe Biden."
At least one account appears to have already fallen apart, however - Pennsylvania US Postal Service whistleblower Richard Hopkins, who claimed to have witnessed supervisors illegally backdating mail-in ballots, "completely recanted" after being questioned by investigators, Democrats on the House Oversight Committee tweeted late Tuesday.
Mr Hopkins' claims were included in a Trump campaign lawsuit filed on Monday which alleged Pennsylvania officials created "an illegal 'two-tiered' voting system" that unfairly favoured mail-in ballots and encouraged "ballot fraud or tampering".
But in another twist, Mr Hopkins appeared in a video on Tuesday night with Project Veritas, which first aired his claims, claiming he did not recant and that he was pressured by USPS investigators to water down his story.
The video included an alleged secret recording from the interview, in which an investigator tells Mr Hopkins the "storm is getting crazy" and he had been called to "try to reel it back in before it gets really crazy".
George Washington Law School Professor Jonathan Turley, who testified for Mr Trump during his impeachment, argued against immediately dismissing all allegations out of hand. "I have expressed scepticism on past claims, but we now have sworn allegations of fraud. Why not look at the evidence?" he said.
Prof Turley said TV networks were "dismissing any and all allegations within the first 24 hours, as tabulations were continuing", describing it as "like saying that a patient has a low white blood cell level but insisting on stopping testing if you cannot conclusively say that there is cancer".
Rod Blagojevich, the former Democrat Governor of Illinois who was jailed for corruption but released earlier this year after Mr Trump commuted his sentence, on Friday weighed in to support the President's allegations.
"If the question is, are the Democrats stealing votes in Philadelphia, my answer is, 'Is the Pope Catholic?' It's a time-honoured tradition," Blagojevich told Newsmax. "Big Democrat-controlled cities like Chicago, my hometown, Philadelphia … to do precisely what they're doing now, I've never seen such a magnitude because this I think is an indication of just how widespread it is, how deep it is."
He said he didn't think it was confined to Philadelphia. "My instincts, again coming out of Chicago Democratic politics, my instincts tell me it's going on in Atlanta, it's going on in Detroit, it's going on in Milwaukee, it's going on in Las Vegas," he said.
"It's like what Justice Powell said about pornography. 'You can't define it, but you know when you see it.' And coming out of the Democratic Chicago political establishment, I know how they operate. They control polling places, they stop votes when their candidate's behind, and then in the wee hours of the morning, in the dark of night, the stealing starts."
Blagojevich claimed "we've seen that in big numbers, unprecedented numbers in this election in Michigan and Philadelphia". "It's outrageous, and the fact that they're doing it with impunity … is because the media is simply looking the other way," he said.
Unfortunately for Republicans, "you know it when you see it" is likely not the legal standard they will need to meet if they hope to be successful in court. Aside from specific testimonies from alleged witnesses and whistleblowers, Republicans are also attempting to make the "statistical case".
Senior Trump campaign adviser Steve Cortes, writing for the right-wing National Pulse website, outlined four elements that "firmly point to the intense improbability of the accuracy of the present Biden lead", which he said provide "more than enough reasonable suspicion to require hand recounts and immediate investigation into fraudulent activities".
Here is a quick rundown of those four claims, and how they stack up.
We'll also look at two other popular theories flying around on conservative social media, which is awash with viral graphs, tables and bar charts, all supposedly demonstrating fraud.
In Wisconsin, where Mr Biden leads Mr Trump by around 20,000 votes, total turnout was more than 90 per cent of registered voters.
"Even in a state with same-day registration, such a number seems implausible," Mr Cortes wrote. "Clearly, high turnout was expected in an intensely political year with vastly expanded access to mail-in voting. But the kinds of numbers reported simply defy reasonable expectations."
Mr Cortes said Wisconsin's biggest city, Milwaukee, which was the decisive locale for Mr Biden securing him a 145,916 vote lead, saw turnout of 84 per cent - dwarfing the "more believable" 51 per cent turnout in the similar midwestern city of Cleveland, Ohio.
Fact checkers have refuted this claim, however, arguing that voter turnout in a same-day registration state is not calculated based on registered voters, but eligible voters.
"So when you divide the number of votes cast in Wisconsin - 3,278,963 as of November 5 - by the voting-age population in Wisconsin (4,536,293 as of 2019, according to the elections commission), you get a turnout rate of 72.3 per cent," USA Today wrote.
"That's the highest rate ever behind the 2004 election, but solidly in the range of past presidential contests here."
Another common argument is that Mr Biden's record-breaking vote numbers relative to former President Barack Obama "simply do not seem credible".
"Could a candidate as doddering and lazy as Biden really have massively outpaced the vote totals of a politician who boasted rock star appeal?" Mr Cortes wrote.
Republicans cite various locations where Mr Biden dwarfed Mr Obama's election performance - such as Montgomery County, Pennsylvania, where he won by 131,000 votes, more than twice Mr Obama's 59,000 vote win in 2012.
"Such eye-popping outperformance versus Obama, in just the right places, naturally raises a lot of suspicion," Mr Cortes said.
Former political consultant Scott Hounsell, writing for Redstate, made a similar argument looking back at Wisconsin, citing five suspicious counties where Mr Biden beat Mr Obama's turnout by double digits.
"If I were the Trump campaign, I would be looking at each of these counties with a magnifying glass," Hounsell wrote. "In totality, in those five counties, Biden beat Obama by 83,291 votes, in a state where Biden beat Trump by just 20,000 votes."
Sidney Powell, one of the lawyers now working with the Trump campaign, told Fox News earlier this week "we've identified at least 450,000 ballots that miraculously only have a mark for Joe Biden on them and no other candidate".
That means the person who filled out the ballot only voted for Mr Biden at the top of the ticket, and not any of the down-ballot candidates in closely contested Senate and House of Representatives races.
"This phenomenon appears far more prominently in battleground states, raising the alarm for manipulation," Mr Cortes said.
In Georgia, Mr Trump's vote total "almost exactly tracked the vote totals for the Republican Senate candidates, separated by merely 818 votes out of 2.43 million votes", Mr Cortes said, but Mr Biden "saw an astounding surplus of 95,801 votes over the Democratic Senate candidates".
The Washington Post, in its fact check, said these numbers were "unremarkable". "Americans have a long and demonstrated history of voting for president and skipping all or much of the rest of the ballot. It's a practice known as undervoting," the paper wrote.
"The 450,000 ballots identified as voting only for Biden are in five states. The total number of votes recorded in those states right now is about 24 million. That means less than 2 per cent of people would have cast ballots for Biden and skipped the rest of the ballot."
The paper pointed out that in Georgia in 2016, the presidential race included more than 200,000 more votes than the state's Senate race.
Finally, Mr Cortes cites lower rejection rates for mail-in ballots compared with previous years as evidence that "tainted" votes are flooding the system.
"Democratic governors clamoured for massive amounts of mail-in voting, knowing full well that most states would become overwhelmed and wholly unable to establish the validity and legality of almost all the votes that poured in via mail," he wrote.
In Pennsylvania, just 0.03 per cent of mail-in ballots are being rejected, representing a refusal rate "1/30th the level of 2016".
"First-time mail-in voters typically see a rejection rate of about 3 per cent historically, or 100 times the rejection rate of Pennsylvania in 2020," Mr Cortes said. "This total lack of filtering or controls raises enormous suspicion regarding a seriously-tainted ballot pool in the Keystone State."
In September, Pennsylvania elections officials told counties they were not allowed to reject ballots if the signature did not match the one in the voter's file - which could at least partly explain the much lower rejection rate.
"We recognise the fact that, as people age, their signature changes, I know mine has," Lawrence County election director L. Edward Allison Jr. told the Associated Press. "Different medical conditions, strokes, all that kind of stuff enters into it."
Another popular claim doing the rounds online is that Mr Biden's vote totals violate Benford's Law, a statistical tool which predicts the distribution of digits in naturally occurring collections of numbers.
Benford's Law is used in financial and accounting investigations, and has also been applied to election forensics in countries including Iran - however other researchers have disputed its ability to detect voter fraud.
11 - In the end, Biden's vote data from that page is far more anomalous than Trump's. Here is what it looks like visually: pic.twitter.com/7qPivR9zQX— Statsguyphd (@statsguyphd) November 5, 2020
Last week, an anonymous Twitter account claiming to be a statistics professor posted a viral thread claiming that Mr Biden's vote totals were "extremely anomalous in comparison to Trump's".
The posts spread like wildfire, and "Benford's Law" quickly became a catchcry among conservatives online - prompting a wave of debunkings from experts.
University of Maryland computer scientist Jen Golbeck, writing on Medium, explained that "first-digit analysis" does not work for election results for several reasons, including the size of electoral precincts.
Benford's Law has been applied to election forensics using "second-digit analysis", but even that has been described by some researchers as "problematical at best and wholly misleading at worst".
University of Michigan statistician Walter Mebane, who was the first to apply the second-digit to election forensics, has already written a paper hitting back at "inappropriate applications" of the law to the 2020 election.
"To date I've not heard of any substantial irregularities having occurred anywhere, and the particular datasets examined in this paper give essentially no evidence that election frauds occurred," he wrote.
Finally we come to another viral Twitter thread, again by an "anonymous data scientist", featuring a series of graphs with dots representing batches of ballots over time, and their Democratic to Republican ratio.
This theory is a bit more complicated to explain, but in essence the author claims the graphs show inexplicable increases in Democratic vote ratios for later batches in certain states, and that "this pattern proves fraud".
In Michigan, we see a combination of Wisconsin strangeness, together with the GA/PA weirdness. We see both signs of contaminated ballot dumping, and ballot ratios drifting toward dems when they should not be. pic.twitter.com/kYdAuDqKyZ— CulturalHusbandry (@APhilosophae) November 9, 2020
Even though mail-in ballots were expected to favour Mr Biden, the author bases their conclusion on the theory that batches of mail-in ballots should have a consistent "D-R" ratio because they "get randomly shuffled in the mail".
A Republican-affiliated Twitter account then amplified the theory in another viral thread, adding its own analysis to the same graphs to explain why they were "so consequential".
I've cut the chart into pre-D and post-D so you can see how dramatic and immediate the change from chaotic patterns to total order was.— The Virginia Project (@ProjectVirginia) November 11, 2020
The odds of structure D occurring in nature are effectively zero. That is artificially induced if anything ever was. pic.twitter.com/OlHc7LtMb9
Several people have now had a stab at debunking the original thread, including journalist Isaac Saul, who described it as "embarrassing" and noted the author misunderstood how Wisconsin moved ballots around.
A Twitter user claiming to be an actuary went more in depth, saying the graphs were completely wrong and that the thread was "both deliberate misinformation and the most idiotic data analysis I've ever seen".
"Basically, the thread author takes incredibly high-level New York Times election data and claims that spikes like this are 'evidence of ballot backdating, manufacturing of software tampering', which is laughable because it's based entirely on number ROUNDING in the source data," he wrote.
As for whether the anomalies are explained by this error, it is impossible to know without a complete picture of other factors (e.g. how mail moves). I am left wondering why the anomalies seem systematic (directional). *All else being equal* I'd expect noise to be the result.— Bret Weinstein (@BretWeinstein) November 11, 2020
Originally published as Truth behind viral Trump 'fraud' graphs