A guide to voter fraud, by the numbers
How serious is the issue of voter fraud in the US?
With the election months away, Americans remain unsure of how they should even plan to cast that vote. Stay home, and run the risk that your mail-in ballot won’t be counted in time to make a difference? Or vote in person and risk contracting Covid yourself? Or, worse, spreading it to someone older and sicker than you and getting them killed? How’s that for a Catch-22?
President Trump opposes universal mail in voting because, he claims, it will lead to large-scale fraud that will undermine the American electoral system. Of course, he’s been clear that he opposes universal voting because it will lead to “millions of millions of ballots” being counted that Trump doesn’t want: “Now, they need that money in order to make the post office work, so it can take all of these millions and millions of ballots,” Trump said in an interview with Fox Business Network’s Maria Bartiromo. “Now, if we don’t make a deal, that means they don’t get the money.” (Fortune)
As we know from our experience of 2020, anyone can say something and get lots of people to believe it without much evidence. That makes getting to the bottom of a more complicated, numbers-heavy topic like voting fraud difficult. Yet voter fraud has been studied before and, at times, extensively, by academics, think tank researchers, and political strategists, and there’s a lot of information out there. This article will walk through some of that detail while giving a broad overview of the state of voter fraud in the US and what different groups of experts (and non-experts) have to say about it. This way, as we get closer to November you’ll be armed with more information to guard against whatever misinformation is certain to be lobbed in your general direction. This article reviews:
- Each tribe’s narrative
- What data do we need?
- Confirmed cases of voting fraud
- Double counting
- Non-citizen voting
- What is the PILF?
- PILF’s data on noncitizen voters
- Solutions? Focus local
- Sources and further reading
1. Each tribe’s narrative
Each tribe? But doesn’t ReConsider usually say that there are multiple tribes and that a false biparty representation is actually more harmful than helpful? Yes indeed. However, when it comes to Trump’s statements about the USPS, I unfortunately think there really are only two tribes in this case. Why? Simply because the issue is immensely divisive. Whether you fear that Trump is stealing the election, or that election fraud will undermine the legitimacy of our institutions, you truly feel that the other group is undermining America’s democratic institutions.
That said – we can still look to polls to get a sense of how big each tribe is. A YouGov poll from August 17th shows that a majority of Americans “disapprove of Trump blocking funding for the USPS,” and a majority of Americans also believe he’s trying to “delegitimatize the outcome of the election because he fears he won’t receive a majority of votes.”
YouGov is generally a well-respected polling agency. But do notice how leading the questions in these polls are. We talk about that more in our podcast on this topic, which will be released this week.
Question: Do you approve or disapprove of President Trump’s effort to block funding for the postal service to prevent mail-in voting in the election?
However, those opinions are still split strongly down party lines.
Question: Donald Trump has repeatedly claimed that the 2020 election will be “rigged” against him. Which comes closest to your view?
Tribe A: Universal mail in voting will lead to unacceptable levels of voter fraud
Narrative: Examples of voter fraud are all over the place. All you need to do is look. Non-citizens registering to vote between 2013 and 2015 in Philadelphia. Deceased people’s names appearing on voter rolls in Colorado. People voting twice. We’ve never had universal mail in voting before, and implementing it immediately will create an unacceptable level of risk of fraudulent voting, thereby undermining the legitimacy of the election. We’ve voted in-person during pandemics before, we can do it again.
To be clear, we oppose universal mail in voting. Absentee voting is fine, since absentee voters have already requested their ballot. The risk is sending out a bunch of ballots to everyone and then having extra ballots left over that anyone can just grab “and do whatever you” want with them. With universal mail in voting, every registered voter receives a ballot even if they don’t request one, which is distinct from absentee ballots, which are also sent in by mail, but requested first.
Tribe B: By limiting universal mail-in voting during a pandemic, Trump is undermining America’s democracy
Narrative: Trump has done a lot to hurt our country over the last four years, but this is crossing a line. By announcing that he intends to limit funding the USPS in the middle of a global pandemic, he’s forcing people to choose between participating in democracy or safeguarding their health (as well as the health of others). This is a purposeful and directed attack on our voting system, one of the things that America has historically used to keep would-be kings out of power. Mail in and absentee voting has been used extensively throughout the country, and the incidence of voter fraud remains infinitesimally small, especially to the outsized impact that voting restrictions would have.
So – which one is true? Let’s dig into some of the data and see if we can learn something interesting.
2. What data do we need?
Voter fraud is a fairly broad category, and can include a number of different types of crime. But we can file most examples into three categories:
- Intentional voting fraud
- Double counting
- Non-citizen voting.
Intentional vote fraud is just what it sounds like – someone made a premeditated effort to game the system by voting more than once or getting someone else to vote more than once on their behalf. Double counting happens when an individual appears on multiple voter rolls, either from a premeditated effort to vote more than once, or because they have multiple address and forgot, or registered in a new town and forgot that they were on the previous voter rolls. It’s a broader category that includes both intentional and unintentional cases of voter fraud. Lastly there’s non-citizen voting, which happens when a non-US citizen ends up on the voting rolls, either intentionally or unintentionally. How can someone accidentally end up on a voting roll? Well, as we’ll see, the integrity of voting systems depends heavily upon the local county administering the election, and the operational and technological capabilities of different counties vary meaningfully.
But first, let’s take a look at intentional voting fraud.
3. Confirmed cases of voting fraud
The Heritage Foundation (a conservative leaning think tank according to AllSides.com, MediaFactCheckBias) keeps a fairly extensive record of intentional voting fraud dating back to 1993. Heritage’s database includes cases that were caught and in most cases someone was convicted of a crime.
The shortcoming of Heritage’s database is that it only contains confirmed cases, which means it is not necessarily a representative sample. For example, perhaps certain types of voting fraud are less likely to be caught, so they’re not picked up by the database. It may be that only the most successful fraudulent voters don’t get caught, and they happen to steal the most votes. Or, it may mean that there isn’t that much voter fraud. It’s hard to tell, because the mere fact that they’re convicted adds a selection bias to the sample. You can’t draw inferences from the entire American population based on it. For that you’d need a different type of sample (more on that soon).
That doesn’t mean that Heritage’s database is useless, just that the conclusions you can draw from it are limited (for a more extensive critique of the Heritage Foundation’s coverage of voter fraud, you can take a look at Brookings response to it. But some China experts have also made the case that Brookings has close ties with Chinese governmental agencies, so who knows).
You can’t infer the true incidence of voting fraud from Heritage’s database. For that, you need to do a little work, which fortunately for us a number of academics, statisticians and political scientists have already looked into. Don’t worry though, you won’t actually have to read any of those academic papers yourself (unless you want to, of course! All links and sources to this article can be found below and interspersed throughout). We’ll summarize all of that for you in the next section
4. Double Counting
Double counting is when a single individual appears on voter rolls twice – usually once in one location, and a second time in another. Double counting can occur either intentionally or unintentionally, for example if someone moved to a new state, registered, but forgot to notify their prior state that they’ve moved. That makes it difficult – but not impossible – to get some sense of the scale of double counting all over the country.
I think you can reasonably assume that most cases intentional cases of voter fraud result in some form of double counting. This is worth noting because it means that intentional voter fraud is captured in the broader “double counting” category. This isn’t a bullet proof assumption, so if I’m wrong here leave a comment below and let me know.
Venn diagram is not to scale.
Double counting is difficult to catch for a couple of reasons. Usually, elections are administered at the county or city level, and each municipality has its own way of keeping tabs of voter rolls. That means administrative systems that don’t coordinate well with one another. The multiplicity of systems makes it difficult to pull data from disparate sources to make an accurate accounting of double-counting in different states.
Researchers regularly reference the challenge of finding reliable data to properly investigate duplicate voting. One data-oriented investigation that was presented to the white house’s Presidential Advisory Committee on Election Integrity (in 2017) said exactly this: “The variability in access, quality, cost and data provided impedes the ability to examine voter activity between states.” A paper authored by researchers from Stanford, Harvard, University of Pennsylvania, and Microsoft echoed these sentiments.
Despite the challenges in making this estimate, these researchers still constructed different statistical models to draw the most general conclusions about voting behavior that they can given the data that is accessible. A team of five researchers estimated that the total number of double 21,724 and 33,346 double votes cast in the 2012 presidential election. The other paper I mentioned was prepared by Simpatico Software Systems and presented to the white house. It found something similar for the 2016 presidential election – that around 40,000 duplicate votes were cast in total in the US presidential election.
What’s interesting here is how close the estimates from these two sources that are using different statistical models. They vary by about 10-20,000, but both estimates are the same order of magnitude. 40,000 therefore seems like a reasonable high-end estimate of the number of double counted votes in both 2012 and 2016.
In 2012, 126 million Americans voted in the residential election, and in 2016 that number increased to 138 million. This would imply that double votes accounted for between 0.17% and 0.29% in 2016 and 2018. This is clearly more than we want, since we want zero. However, as the Stanford, Upenn, Harvard and Microsoft research paper point out, the steps necessary to remove one of the two these double counted registrants could “impede approximately 300 legitimate votes for each double vote prevented.” That would represent between 10-12 million Americans whose registration could be cancelled in the pursuit of eliminating 40,000 double votes.
That would represent between 10-12 million Americans whose registration could be cancelled in the pursuit of eliminating 40,000 double votes.
5. Non-citizen voting
Non-citizen voting makes up a very small percentage of votes cast. The Brennan Center, a left of center think tank according to Media Fact Check Bias, conducted an analysis of 23.5 million voters and found 30 examples of non-citizen voting violations, or about 0.0001%. The Washington Times, a right-leaning media company, makes the case that voter fraud is a problem by presenting individual cases of it. The downside to this approach is that citing specific cases anecdotally doesn’t give you a good sense of the scale of the problem. It tells you that it is happening, but not how much or how often it’s happening. It doesn’t provide insight into whether voter fraud has a meaningful impact on election outcomes.
This isn’t a pass for voting fraud, by the way. We should use the full force of the law to prosecute people who intentionally break the law. In an ideal world, there’d be no voter fraud. But, as 2020 has clearly shown us, we don’t live in a perfect world. We’re stuck analyzing policies based on the reality we’re confronted with rather than the one we wish for.
There is one larger-scale study (note: I’m not saying I did a comprehensive literature review, just that this is one I happened to come across) about non-citizen voters, and newspaper after newspaper after newspaper after newspaper cites it. In it, an analysis of 8 counties in Virginia revealed 1,046 non-citizen voters were found on Virginia state’s voter rolls. Its contention was that finding this many non-citizen voters in such a small number of counties implies that the size of the problem could potentially be much larger.
This larger-scale study was conducted by an organization called the Public Interest Legal Foundation (amusingly referred to as “PILF”). What the heck is PILF?
6. What is the PILF?
In PILF’s two whitepapers, titled “Alien Invasion In Virginia” and “Alien Invasion II: The sequel to the discovery and cover-up of non-citizen registration and voting in Virginia“, PILF makes the case that non-citizen registrations have almost exclusivity been caught by accident. Therefore, the fact that 1,046 registered non-citizens were found in 8 of Virginia’s counties in 2016 is a sign that this is just the tip of the iceberg and the problem must actually be bigger.
However, PILF doesn’t go beyond the observational study to attempt to estimate the true size of the problem. That lets them make claims like “large numbers of ineligible aliens are registering to vote and casting ballots, sometimes for decades, before they are removed from the registration rolls” (Public Interest: Alien Invasion II) without specifying what they mean by “large”.
In addition to its data being insufficient to generalize, PILF screwed the pooch when they published data on the 2018 elections. Essentially, they requested voting records from the state of Virginia to attempt to quantify how many non-citizen voters there were. When they received that information, rather than reporting the correct number, they mistakenly reported twice as many non-citizen voters as actually existed. This resulted in an erroneous finding that 1 million ballots across the US were “returned as undeliverable” in 2018. They argued that this was a sign of growing, unaccounted-for ballots.
However, it turns out that there were actually fewer undeliverable ballots in 2018 as in 2016, and PILF just reported the wrong figures. They subsequently corrected their data, but by that time mass-media outlets had already picked up and reported on those incorrect figures.
PILF also claimed that 28 million ballots wer sent out in the 2012, 2014, 2016 and 2018 elections and unaccounted for – that is, election officials don’t know what happened to them. PILF believes that these unaccounted for ballots pose the risk of vast election fraud: “These represent 28 million opportunities for someone to cheat.”
That’s not exactly right, though. Something like 30% of mail-in ballots are unreturned for “any given election,” which makes sense because we know that something like 40-50% of registered voters don’t vote anyways, whether in-person or by mail. As this Snopes review of the PILF’s claim explains, implying that these 28 million ballots were “missing” is akin to arguing that the millions of registered voters who choose not to vote in any given election are “missing” voters.
Now that we have a little background on the PILF, what do their numbers actually say?
7. PILF’s data on non-citizen voters
Alien Invasion I was published in September 2016. It reviews 8 counties in Virginia, and finds 1,046 “aliens who registered to vote illegally.” This doesn’t mean that 1,046 illegal votes were cast in elections, just that 1,046 individuals who were non-citizens ended up on voter rolls in these 8 counties. PILF requested more information from the state, and claims that the lack of transparency by the state government to provide this data represented “…obstructionist efforts by Commissioner Cortes,” although PILF’s own records indicate that the state government providing this info to PILF would have violated privacy laws.
Alien Invasion II expands on the first part by looking at non-citizen voting at the entire state of Virginia, rather than in just 8 counties. It finds that 5,500 non-citizens were illegally registered to vote. “Of these illegal registrants, 1,852 cast nearly 7,500 ballots in elections dating back to 1988”. From 1988-2018, nearly 67 million votes were cast in the state of Virginia, which means that those 7,500 illegal votes accounted for 0.01% of the votes over those 30 years. However, it’s worth noting that PILF’s figure of 7,500 illegal votes may still be high, as it seems like the way PILF is accounting for non-citizens in their analysis may not be accurate, according to Virginia state officials.
PILF, who’s main message is that voting fraud is major problem, reports extremely low figures of actual voting fraud.
Separately, it’s worth noting that the general counsel for PILF, Christian Adams, was appointed by President Trump on August 11th of this year to sit on the U.S. Commission on Civil Rights.
8. Solutions? Focus local
If you happen to think that voter fraud is a problem, then that problem seems to exist primarily at a county level. Counties are responsible for handling votes and often have different and incompatible systems, making comparisons between different regions difficult since it’s hard to know for sure if there’s a duplicate voter or not. Companies that are hired to do this sort of analysis, like Simpatico Systems, are forced to rely on matching algorithms to take a best guess.
The thing is – similar complaints as those described above have been made for the last 20 years. If anyone’s to blame, then, it’s local politicians (or local government agencies) who have failed to establish the data management systems necessary to carry out a democratic vote in uncertain times.
Now, could we have predicted the Covid pandemic five years ago? Well, yes. Health experts have been warning us about pandemic risk for years, describing it as inevitable. If we’re not prepared to vote during a pandemic, it’s not due to lack of fair warning, but lack of action at a local level.
If you’re frustrated because voting fraud happens, you have a right to be! Ideally, we’d live in a world where it doesn’t happen at all. Luckily for you – an election is coming up, and you have the opportunity to voice your frustration towards county officials that failed to set up voting systems that could cope with the sort of social disruption that 2020 has thrown our way.
That said, we’re not going to completely eradicate double counting or non-citizen voting in the next three months, and far, far more people would be disenfranchised by short-term measures aimed at limiting double counting than fraudulent votes would be stopped. It’s not worth the trade off.
9. Sources and further reading
Project Vote – Politics of Voter Fraud
Washington Times: No, voter fraud isn’t a myth: 10 cases where it’s all too real
CBS4 Investigation Finds Dead Voters Casting Ballots in Colorado
Potentially thousands of illegal voters in Pennsylvania
Brennan Center – Debunking Voter Fraud Myths
Brennan Center – The Truth About Voter Fraud
The Heritage Foundation’s database on Election Fraud Cases
NBC News – Trump pushes false claims about mail-in vote fraud. Here are the facts.
Fox News: California man charged with mail-in voting fraud in three elections.
Trump disbands Commission on Voter Fraud (Jan 2018)
Project Vote – Politics of Voter Fraud (2007)
“One Person, One Vote: Estimating the Prevalence of double voting in U.S. presidential elections.” Sharad Goel, Stanford University; Marc Meredith, University of Pennsylvania; Michael Morse, Harvard University; David Rothschild, Microsoft Research; Houshmand Shirani-Mehr, Stanford University.
Simpatico Software System’s presentation to the Whitehouse on data mining for voter fraud
The Election Assistance Commission’s bienniel Election Administration and Voting Survey
Want to have better political conversations?
If you want a free guide on how to have better political conversations, click here to get your copy of the ReConsider Principles and Discussion Strategies. Come join us on the ReConsider Facebook Group, where other open-minded folks like yourself engaging in challenging, respectful political conversations.
And don’t forget to subscribe to ReConsider on your favorite pod catcher or Youtube:
2 Comments