The election was no different than previous elections, with minor fraud incidents that did not change the outcome.
(91% probability)Question
Was there widespread fraud in the 2020 US election?
30 Nov, 2020Answer
Background
The 2020 US presidential election officially concluded with president Joe Biden declared the winner over his opponent, former President Donald Trump. The results have been disputed by many, including Trump, claiming the Biden victory was obtained illegally, by various means of election fraud. Due to the complexity of the US election system and the number of fraud claims raised, the issue is not easily settled, and a probabilistic analysis is needed.
Hypotheses Considered
Calculated Results
Calculated Results
1
91%
The election was no different than previous elections, with minor fraud incidents that did not change the outcome.
91%
2
4.4%
The election outcome was manipulated through a centralized mass computer fraud, involving a significant portion of US electronic voting equipment.
4.4%
3
4.3%
The election outcome was manipulated through the centrally coordinated effort of multiple people.
4.3%
4
0.6%
The election outcome was manipulated through many local and uncoordinated frauds, whether human or machine-based.
0.6%
Starting Point
Initial Probabilities
Name
Initial Likelihoods
No Effect
85%
Centralized fraud
7%
Computer fraud
2%
Decentralized fraud
6%
According to research, almost all presidential elections are accompanied by low-level fraud that does not affect the outcome. There have been a few known cases of centralized and decentralized frauds in small elections, while there is no known prior for computer fraud.
To account for yet undiscovered frauds and new technologies, we generously increase the initial likelihood of fraud, and especially computer fraud.
Name
Initial Likelihoods
No Effect
85%
Centralized fraud
7%
Computer fraud
2%
Decentralized fraud
6%
Evidence
Effect
Fraud Claims
Name
Effect
Updated Likelihoods
Nearly all of the fraud claims in the 2020 elections were either debunked or remain unproven. No claim or combination of claims has been found to be both true and sufficient for changing the outcome of the election. This has been the conclusion of virtually all government and election officials, recounts, courts, experts, and the governors and secretaries of states, in the disputed states. We have independently validated a sample of the key conclusions.
Yet some claims do exemplify windows of opportunity for fraud, and some actions by officials in those situations (especially in Georgia) are somewhat abnormal.
Name
Effect
Updated Likelihoods
Results vs. Expected results
Name
Effect
Updated Likelihoods
When comparing the results of the election to polls predictions, the results in the Georgia runoff election, and the voting history in the disputed states, there is no evidence of fraud. These factors greatly reduce the plausibility of claims that allege hundreds of thousands of fraudulent votes in a single state. Smaller frauds - which are the majority of the claims - would probably not be detected by these tools and comparisons.
Name
Effect
Updated Likelihoods
Voter Turnout
Name
Effect
Updated Likelihoods
While turnout increased significantly in the 2020 elections, the increase was nationwide and bipartisan (though it has leaned more Democratic). Statistical analysis also showed no indication of fraud. As such, we slightly increase the likelihoods of the No effect hypothesis.
Name
Effect
Updated Likelihoods
Margins in disputed states
Name
Effect
Updated Likelihoods
The small margins in the disputed states reduce the likelihood of an effective decentralized fraud. With no coordination between participants, there is no way of knowing when to stop altering the vote in order to avoid raising red flags.
The margins in the disputed states are small compared to the results nationwide. But this seems to be the standard, as challenging these results (in courts, media, etc.) has the greatest chance of success.
Nonetheless, Democrats winning 6 out of the 7 states with the smallest margins is a little less likely and so we generously increase the likelihoods of the effective frauds.
Name
Effect
Updated Likelihoods
Missing evidence
Name
Effect
Updated Likelihoods
Election frauds are sometimes accompanied by whistleblowers or anomalies detected by fraud detection software. The lack of these in the 2020 elections slightly reduces the likelihood of effective frauds.
Name
Effect
Updated Likelihoods
Election systems vulnerabilities and defenses
Name
Effect
Updated Likelihoods
Contrary to exaggerated claims by regulators and the media, the 2020 election was not beyond the possibility of an effective fraud, but it is still unlikely.
Though the systems are somewhat secured and, more importantly, decentralized and mostly disconnected from the internet, there are still significant and proven methods of hacking individual systems and a few proven, yet complicated, methods of hacking significant portions of the electronic election system.
Almost all voting machines in the US, and all of the machines in the six disputed states, print paper ballots that are checked by the voter, which should create a discrepancy with hand recounts in case of digital fraud. No such discrepancy was found but there are a few known methods of bypassing even the paper audits.
As such, we generously increase the likelihoods of effective frauds.
Name
Effect
Updated Likelihoods
Discussion