Allegations of Fraud in 2020 US Election

[Ambrosia]

Hey if you really want to bake your noodle and are having trouble following Benfords law google the Zipf distribution.

Vsauce did a video on it a few years back.



(and no that won't help with detecting election fraud either, but learning new stuff is always fun)

 

[shuttlt]

It appears that many on this thread are using Wikipedia's "Benford's Law" article. There is a comment highlighted in the footnotes is as follows..." Raimi makes the brief comment: "...many writers ... have said vaguely that Benford's law holds better when the distribution ... covers several orders of magnitude."

Yet, the above quote has morphed into... "Distributions that do not span several orders of magnitude will not follow Benford's law."

Ralph Rami wrote a review article in 1976 debunking the need for multiple magnitudes. The quote that people are quoting which says there is a need for multiple magnitudes... is exactly opposite of what Rami proved in his hallmark book and publication.

I can't believe I didn't check. The wikipedia article is being edited to claim that benford's law today. Quotes from editors:

Seems like a coincidence, doesn't it? A seemingly unchanged, relatively ignored article, about a law that indicated voter fraud time and time again, now discredited?

and

It actually discredited Benford's law being used for election fraud months before the 2020 election, as you can see from this older edit. However, in recent days an anonymous user repeatedly removed that text, causing it to have to be re-added. In fact, the most recent edit added an extra sentence to cast doubt on the study that supposedly "discredited" the application of Benford's law to elections. So by all metrics, Wikipedia got changed in the direction of saying that Benford's law IS applicable to voter fraud.

The section everybody is quoting was added only on the 9th. Below is the version before people started to edit the article to win the argument post election:
https://en.wikipedia.org/w/index.ph...pected_and_not_expected_to_obey_Benford's_law

I'll try and go back and find what it said before all this fun began.

 

[pgwenthold]

Seriously, "skepticism" means we are supposed accept the claims of voter fraud?

 

[johnny karate]

This has been a skeptic board in name only for a very long time now.

Do you believe that mass voter fraud has occurred?

 

[Meadmaker]

I don't see that that would mean that the least significant digit wouldn't follow benford's law. Surely benfords law applies to each precinct. For any given candidates vote in any given precinct there is ~30% probability that the least significant digit is 1, an 18% probability that it is 2... start adding precincts and a curve should emerge...? No?

It feels like I'm missing something here. Could you explain?

To understand, you have to understand where that 30% comes from. What's the deal? What magic is this?

The answer is that log10(2)~0.30.
The 18% comes from the fact that log10(3)~.48.

In order to follow Benford's law, the kind of data that you are tracking has to be the sort of thing that has some sort of logarithmic element to it. Lots of natural processes have that sort of property. Another thing that has that sort of property is the product of a group of uniformly distributed random integers. In other words, if I roll a die 10 times, and multiply the results together, and take the first digit, the results will follow Benford's law.

On the other hand, sums of data won't do that. If I roll those same dice and add the results, there won't be any digit in there that follows Benford's law. Not the first. Not the last.

A lot of natural processes do follow Benford's law, because they have an exponential characteristic. That's why engineers use log paper a lot.

"How many people voted for Joe Biden in precinct X" is not the sort of thing you can plot on log paper.

So, how do those people who analyzed the Iranian election come up with something that follows Bedford's law? I have no idea. Products of two randomly selected precincts? I think that would work. I think the result might follow Bedford's law, if the results were legitimate. I really don't know how that all works. I don't know what numbers anyone selected to make that work with election data.

 

[pgwenthold]

The section everybody is quoting was added only on the 9th. .

The paper supplied by xjx was published in 2011.

It concluded that there is no correlation between the results of Bedford's law and election outcomes.

 

[xjx388]

Yes, that is one paper. The same one quoted by Wayerin. Equally there are many, many papers, and newspaper articles using Benford's law to analyse elections. You don't prove/debunk something by finding one paper whose abstract agrees with you. Chiropractic, homeopathy, and clairvoyance are 100% legit by that criteria.

It's one paper, true. There are others, here's one.

Therefore, this article will apply the test to the 2009 German Federal Parliamentary Election against which no serious allegation of fraud has been raised. Surprisingly, the test results indicate that there should be electoral fraud in a number of constituencies. These counter intuitive results might be due to the naive application of the 2BL-test which is based on the conventional χ2 distribution. If we use an alternative distribution based on simulated election data, the 2BL-test indicates no significant deviation. Using the simulated election data, we also identified under which circumstances the naive application of the 2BL-test is inappropriate. Accordingly, constituencies with homogeneous precincts and a specific range of vote counts tend to have a higher value for the 2BL statistic.

OTOH, proponents of the "Benford's Law applies" viewpoint have yet to present any citations whatsoever. The fact that it has been used to try and detect election fraud says nothing about the suitability of the Law for elections. Do you have anything?

 

[shuttlt]

In March there wasn't any talk that I can see about it being no good under these curcumstances:
https://en.wikipedia.org/w/index.ph...pected_and_not_expected_to_obey_Benford's_Law

and a pre-Trump one in June 2015

https://en.wikipedia.org/w/index.php?title=Benford's_law&oldid=666424201#Election_data

what you have is the following:

Benford's Law has been invoked as evidence of fraud in the 2009 Iranian elections,[19] and also used to analyze other election results. However, other experts consider Benford's Law essentially useless as a statistical indicator of election fraud in general.[20][21]

So, some experts say it's good, others say it isn't. Hence not debunked but perhaps open to dispute?

 

[Meadmaker]

I can't believe I didn't check. The wikipedia article is being edited to claim that benford's law today. Quotes from editors:


and



The section everybody is quoting was added only on the 9th. Below is the version before people started to edit the article to win the argument post election:
https://en.wikipedia.org/w/index.ph...pected_and_not_expected_to_obey_Benford's_law

I'll try and go back and find what it said before all this fun began.

That's hilarious.

(Not you. The fact that a Wikipedia article about an obscure mathematical property is being edited a few days after an election.)

 

[shuttlt]

The paper supplied by xjx was published in 2011.

It concluded that there is no correlation between the results of Bedford's law and election outcomes.

Sure, I'm aware of that. I was talking about the Wikipedia article.

 

[No Other]

Benfords Law isn't black and white. It gets more accurate the larger the range of the sample being considered.

i.e. if you have a set of values that span one order of magnitude it might apply or it might not.

if your dataset spans two orders of magnitude, it might apply it might not, though it's more likely to apply than a dataset of one order of magnitude

if your dataset spans 10 orders of magnitude it is very likely to apply.

The greater the sample size, the greater the refinement. That is pretty much with anything... But that is not what Benford's law is about, it works perfectly fine with one magnitude.

Somewhere along the line of one oom to ten oom there' a point where you can say this dataset is not a naturally occuring set of numbers to a very high degree of confidence.

accuracy is more appropriate.

In order to say a dataset has occurred naturally or has been manipulated you need a high degree of confidence - which doesn't exist when your dataset spans a low number of orders of magnitude. (or doesn't follow a power law)

Provide an example of what you mean.

The short version is WE DO NOT KNOW WHY BENFORDS LAW WORKS - what we can say is that it's a very useful tool in the times and places that we know it does work and, spoiler alert, detecting election fraud ain't one of them.

Exactly how is this tool inappropriate for election results?

 

[slyjoe]

This guy's in on it too.

 

[pgwenthold]

Sure, I'm aware of that. I was talking about the Wikipedia article.

Well that's really silly. Hand-wringing over whether Wikipedia is sufficiently complete?

Talk about pointless.

 

[shuttlt]

Do you believe that mass voter fraud has occurred?

I know we quibble here about voter fraud and election fraud here. There is also the question of what you mean by "mass". In any case, I don't know. Result impacting voter/election fraud/"human error" involving 10s of thousands of votes has certainly occurred in the past. I see no reason in principle why it shouldn't have occurred now.

 

[shuttlt]

Well that's really silly. Hand-wringing over whether Wikipedia is sufficiently complete?

Talk about pointless.

The sections of the article that were being quoted as "debunking" had only been added after the election. I had naively not checked that, and was bringing it to people's attention who might have been similarly naïve.

 

[No Other]

OTOH, proponents of the "Benford's Law applies" viewpoint have yet to present any citations whatsoever. The fact that it has been used to try and detect election fraud says nothing about the suitability of the Law for elections. Do you have anything?

Benford's Law is Empirical what more do you require?

 

[Skeptic Tank]

I'm not trying to do the impossible. He's desperately throwing crap that is served up to him on Qanon and other nutjob news sources on the wall hoping it sticks.

I find that the facts stick better.

Can you point to anyone else in this thread who flipped from Al Franken loving Obama voter to what I am now? Looks like I have possibly the best record for being convinced out of stuff of anyone in here.

As for QAnon? Those people are a joke and I’ve never paid the slightest molecule of attention to their stupidity.

 

[shuttlt]

That's hilarious.

(Not you. The fact that a Wikipedia article about an obscure mathematical property is being edited a few days after an election.)

This happens all the time, see Miriam-Webster doing a "scheduled update" to their definition of 'sexual preference' hours after it was claimed to be derogatory in the Coney Barrett hearing. The past and language is updated continually to better serve the purposes of the present.

 

[shuttlt]

The short version is WE DO NOT KNOW WHY BENFORDS LAW WORKS - what we can say is that it's a very useful tool in the times and places that we know it does work and, spoiler alert, detecting election fraud ain't one of them.

What? Of course we know how it works. It's a mathematical law that describes processes that fit particular criteria. In what sense don't we know how it works?

 

[shuttlt]

Seriously, "skepticism" means we are supposed accept the claims of voter fraud?

Skepticism means you don't call them debunked on the basis of one study that agrees with you.