If you took every confirmed case of voter fraud in the US since 1950 and shoved them all into 2020, they would not be enough to flip the election in most -if any- states. The notion that there is widespread voter fraud that MUST BE STOPPED is a false one, circulated entirely for the purpose of fooling the gullible into supporting thinly-veiled modern equivalents of poll taxes and literacy tests.
The GOP is full of bullshit, I'll give you that.
I will, however, add the bane of all voter fraud detection system statistics.
- Detection rate is easy to acquire.
- Precision is easy as well, albeit fuzzy and imprecise at times.
- False omission rate, however, is extremely difficult to acquire if not straight out illegal.
In layman's terms:
- We know quite well how often the system is dinging.
- We know somewhat well how likely it is that the system is correctly dinging.
- We hardly know anything about how likely it is that the system is incorrectly not dinging.
What we can say is that the false omission rate is
probably small due to the effort required but probably not
inconsequential for
tight races.
Worse, trying to measure the probability requires either intentionally committing voter fraud or convincing a large swath of the population that a hugely important (fake) election is being held.
We can calculate it using the detection rate, the precision, and the sensitivity. However, that is just kicking the can down the road since sensitivity suffers the same issue as the false omission rate: we hardly know anything about it, and measuring it is difficult if not illegal.
False Omission Rate = Precision * ((1 - Sensitivity)/Sensitivity) * (Detection Rate/(1 - Detection Rate))
If everything worked flawlessly, no voter ID laws, and the fraud rate is low and not too concentrated in one area, then sensitivity can probably be approximated by the turnout. Based on
NYC DOI data from the 2013 NYC mayoral election with a 13% turnout, when the system wasn't working correctly, the sensitivity dropped to 3%. (The methods used by the DOI were legally questionable, by the way)
As an example, I estimate the UK 2017 parliamentary elections to have had a false omission rate of 0.1 ppm +- 90% (rounded to 0 to 8 fraudulent votes that went undetected), based on
police reports. The MP seat of North East Fife was decided by two votes.
The garbage error bars are due to the fuzziness of the police data (inconclusive reports and reports that lack evidence).