OPINION:
With the second impeachment trial of Donald Trump about to start, we’ll be hearing a lot about fraud in the 2020 election. Republicans will claim that fraud was widespread and incontrovertible, while Democrats will say that these claims are false and intended to cast doubt on a free and fair election that they won.
How can the American public rationally evaluate these claims? As nonpartisan election researchers and professors at Stanford University, where we run a research lab studying these issues, we ask two key questions about any claim we hear.
First, what is the evidence offered for the claim? Many claims about voter and election fraud seem intuitive, inviting quick acceptance but falling apart under scrutiny.
Consider popular claims that many thousands of dead people voted in the 2020 election. The evidence for these claims comes from comparing public records on who voted in the 2020 election to death records and noticing that some individuals listed as voting this past November share the same name as people who died before the election.
But many people share the same name. In a country of 330 million, you are sure to find many people with the same name, some of whom voted in 2020, and others of whom passed away in 2020 without voting.
Following previous research in political science, and mirroring work that election administrators do on a regular basis, our research lab explored this last year in Washington state’s all-mail elections — the type former President Trump claimed were especially prone to fraud. While we found many thousands of instances between 2011 and 2018 where a deceased person shared a name with someone who voted, nearly all of these cases involved distinct people: One person voted, another person with the same name but a different birthdate passed away.
Journalists who have examined specific claims about dead people voting in 2020 have found the same types of patterns. Dead people are not voting in large numbers; the evidence offered for this claim is junk, based on accidental or intentional misuse of voting and death record data.
Another popular claim is that heavily Democratic counties used the surge in absentee voting as cover to fraudulently increase President Biden’s vote share.
In late December, economist John Lott claimed to provide rigorous statistical evidence that this fraud occurred in Georgia and Florida. Mr. Lott’s procedure to identify the fraud boiled down to a comparison of two differences. First, Mr. Lott compared precincts in heavily Democratic counties to precincts just over the border in more Republican counties. Second, he compared precincts in two heavily Republican counties. His final estimate compared the difference involving Democratic counties to the difference involving just Republican counties.
But, research from our lab, in collaboration with Prof. Andy Eggers at the University of Chicago, discovered that this research design has a fatal flaw: The conclusion is based on the arbitrary order the Republican counties are recorded in the data set. If we reverse the order they appear in the data set, Mr. Lott’s procedure produces “evidence” consistent with pro-Trump voter fraud. Once we employ standard statistical techniques to avoid this arbitrary ordering, we find no evidence that heavily Democratic counties in Georgia or Pennsylvania padded Mr. Biden’s lead. Again, this is a claim based on faulty data.
In other cases, though, the claims are based on true facts but make incorrect conclusions.
For claims based on true facts, we ask a second question: If the claim is factually correct, does it actually imply the election was fraudulent?
Consider the concern over bellwether counties in the 2020 election. Trump allies pointed to Mr. Biden winning only one of 19 bellwether counties — counties that have predicted the correct presidential winner since 1980. This is a true fact, but does it imply the election was fraudulent?
No. Research from our lab shows that the bellwether counties’ past performance is no indication of their future success. In fact, we find that counties that have predicted several prior elections successfully usually perform worse than other similar counties when predicting the next election. Mr. Biden won few bellwether counties because most of the 19 remaining bellwether counties tended to vote for Mr. Trump in 2016 and had only small shifts in their vote total in 2020.
In a similar vein, some Trump allies have claimed that Mr. Biden’s win was statistically improbable — if not impossible. In a lawsuit Texas filed against Pennsylvania, an expert report from Charles Cicchetti claimed that there was a less than “one-in-a-quadrillion” chance of Mr. Biden winning the election, a claim widely parroted across social media and television news. To compute this probability, Mr. Cicchetti compared Democratic support in swing states in the 2016 and 2020 election and asked how likely the change would be if the underlying support for Democratic candidates remained constant. Mr. Cicchetti found it was very unlikely.
It is true that the election results from 2016 to 2020 changed. But does this indicate nefarious behavior?
No, it doesn’t. Mr. Cicchetti’s test vastly understates the probability of change. We applied his test to all state-level historical presidential election results and found that it flags nearly all results as suspicious for both Democrats and Republicans. Even more dubiously, 6% of elections exhibit a change Mr. Cicchetti claims has a “1 in almost-infinite” chance of occurring.
To sum up: Some of the major facts that are purportedly so surprising as to throw the 2020 election result into question are either not facts, or not surprising.
As nonpartisan election researchers, we are alarmed by the false, misleading, and mistaken claims that the former president and his allies have pushed in their attempt to subvert the democratic process.
But we do not think the solution is for experts to simply say, trust us, there’s no such thing as fraud. Fraud is not impossible, and the public should always remain vigilant about all threats to democracy. As the impeachment trial brings the election fraud debate back to the forefront, we hope that people will look not only at the claims being made, but at the evidence and logic underneath them.
• Justin Grimmer is professor of political science at Stanford University. He is a senior fellow at the Hoover Institution and the co-director of the Democracy & Polarization Lab. Andrew B. Hall is professor of political Science at Stanford University, and, by courtesy, a professor of political economy in the Stanford Graduate School of Business. He is a senior fellow at the Stanford Institute for Economic Policy Research and the co-director of the Democracy & Polarization Lab.
Please read our comment policy before commenting.