Thursday, October 22, 2020

The Rate of Voter Fraud is So Low It's Almost Statistically Impossible to Swing the Election

Recently, voter fraud has been a huge topic of focus in the upcoming election - brought up repeatedly by President Trump, including at the first presidential debate, as well as over the years as justification for stricter voter ID laws.

But just how prevalent is voter fraud, either with respect to mail-in ballots, in-person impersonation, or double voting? How often does it actually happen?

I compiled a meta-analysis by combining the findings from over a dozen studies/papers/reports on the subject, using sources ranging from President Trump's Presidential Advisory Commission on Election Integrity, to MIT, a fairly reputable university, to Reuters, as neutral/balanced a source as there is. The elections analyzed ranged from 1998 to 2018, with the research originating from universities, the Washington Post, Republican Administrations (both at the state and federal level), the Heritage Foundation (conservative think tank), and the US Court of Appeals for the Fifth Circuit. When a range of outcomes was provided, I always took the highest end of the estimate. All voting data was gathered from the United States House of Representatives.

Voter turnout has been consistently increasing, so I found it most important to identify the percent of voter fraud, not the nominal number of cases, and then apply those percentages against 2016.

Overall, my meta-analysis estimate of voter fraud is 0.00089304% - or 1 in 111,977 ballots cast. This is equivalent to roughly 1,222 votes in the 2016 election, which was 0.0409% of the popular vote margin, and 1.57% of the electoral margin (if all applied to the 3 states that swung the electoral college: Pennsylvania, Wisconsin, and Michigan).

Rate1 in...Equivalent in 2016% of Pop Vote Margin% of Electoral MarginSource #Source
0.02%5,00027,3570.92%35.19%1Working Paper
0.007313%13,67510,0030.335%12.87%2Iowa Secretary of State
0.00614%16,2848,4000.281%10.80%3White House (Trump)
0.002548%39,2473,4850.117%4.48%4Washington Post
0.0025%40,0003,4200.115%4.40%5Brennan Center at NYU
0.0001713%583,7222340.008%0.30%6Arizona State
0.0001%1,000,0001370.005%0.18%7Brennan Center at NYU
0.00009032%1,107,1431240.004%0.16%8Reuters
0.00008741%1,144,0931200.004%0.15%9Heritage Foundation
0.00006%1,666,667820.003%0.11%10MIT
0.00001%10,000,000140.0005%0.02%11US Court of Appeals for the Fifth Circuit
0.0000031%32,258,06540.0001%0.005%12Loyola Law School
0.000002924%34,196,79740.0001%0.005%13Washington Post
0.00000017%588,235,2940.230.00001%0.0003%14Kansas Secretary of State
0.00000013%769,230,7690.180.00001%0.00023%15US Department of Justice (W Bush)
0.00089304%111,9771,2220.0409%1.57%Estimate (Meta-Analysis)

Each source is detailed and linked below:

Source #SourceInstitution TypeElection(s)# StatesDate
1Working PaperUnaffiliated201250Oct 2017
2Iowa Secretary of StateRepublican Administration20121May 2014
3White House (Trump)Republican Administration201620Aug 2018
4Washington PostNewspaper2016, 20183June 2020
5Brennan Center at NYUUniversity2000 to 200650Nov 2007
6Arizona StateUniversity2000 to 201050Aug 2012
7Brennan Center at NYUUniversity201642May 2017
8ReutersInternational News1998 to 20181July 2020
9Heritage FoundationConservative Think Tank2000 to 201850Oct 2020
10MITUniversity2000 to 201850April 2020
11US Court of Appeals for the Fifth CircuitFederal Court2000 to 20101July 2016
12Loyola Law SchoolUniversity2000 to 201250Aug 2014
13Washington PostNewspaper201650Dec 2016
14Kansas Secretary of StateRepublican Administration2010 to 20141Jun 2015
15US Department of Justice (W Bush)Republican Administration2000 to 200650Apr 2007
Estimate (Meta-Analysis)UnaffiliatedAllAllOct 2020

What would these rates look like in every presidential election in the last 50 years? To be as conservative as possible, I'll apply both my estimate (above) and the maximum incident rate I found - 0.02% (even though the paper acknowledged "An audit of poll books, however, suggests that such measurement error could explain many of these apparent double votes.")

ElectionPopular Vote MOV% of Margin (Max)% of Margin (Estimate)
20162,984,7570.92%0.04%
20125,081,9000.50%0.02%
20089,550,1760.27%0.01%
20044,078,2110.59%0.03%
2000530,8933.82%0.17%
19968,203,7160.21%0.01%
19925,805,3390.29%0.01%
19887,077,5230.26%0.01%
198416,878,1200.11%0.00%
19808,423,1150.19%0.01%
19761,683,2470.95%0.04%
197217,995,4880.08%0.00%

Never would either estimate of voter fraud have swung the popular vote. In fact, only once would it have surpassed 1% of the popular vote margin (2000; more on that later).

But the United States isn't decided by popular vote - it's decided by the Electoral College. In three elections in the last 50 years, a handful of swing states have decided the election:
  • 2016
    • Swing states: Pennsylvania, Wisconsin, Michigan
    • Margin in swing states (3): 77,744 votes
    • Fraud estimate (Max): 27,357 votes
    • Fraud estimate (Meta-Estimate): 1,222 votes
  • 1976
    • Swing states: Wisconsin, Ohio
    • Margin in swing states (1): 44,578 votes
    • Fraud estimate (Max): 16,321 votes
    • Fraud estimate (Meta-Estimate): 729 votes
  • 2000
    • Swing states: Florida
    • Margin in swing state (1): 537 votes
    • Fraud estimate (Max): 21,119 votes
    • Fraud estimate (Meta-Estimate) 943 votes
Only one presidential election ever could have possibly been swung by fraudulent votes - the 2000 election of Bush over Gore, decided by Florida - and that's if the estimated fraudulent votes were all cast in the same state for the same party. No other election, even the contests of 1976 or 2016, come close.

Whether fraudulent or not, at margins that close, a mistake can swing an election just as easily as voter fraud, like the infamous butterfly ballot in Palm Beach County (in which thousands of people mistakenly voted for a third-party candidate):


The overwhelming conclusion is that the election has to be extremely close for any of this to matter - something that FiveThirtyEight currently gives a 4% chance of happening.


When you discount that 4% chance down further, by the fact that the fraudulent votes would have to ALL be in the same "one or more decisive states" AND all be in the same direction for the same party, it's almost statistically impossible for voter fraud to change the outcome of the election.

It's far more likely, as of this post, while living in the United States, that you would have tested positive for COVID (2.58% of US population) or died from it (0.068%).

No comments:

Post a Comment