Benford's “law” does not detect voting fraud
cambridge.org
cambridge.org
Benford's law requires that the dataset cross multiple orders of magnitude to kick in. Precincts are generally of the same order of magnitude of size (1000s of people), which is insufficient for Benford's law to apply.
Suppose that's true. So what? We analyze the results. We apply science. It can't change the result if it wasn't supposed to.
You give the devil his due. Because then given the result doesn't change, you can't be accused of covering up a fraud when it wasn't actually there. Which is what happens in the alternative.
Casting doubts and entertaining the witch hunt is damaging the democracy by eroding the trust of people in the election system. Opening the way for a coup in the future.
During the debates Biden pledged to wait for the vote to be certified before declaring victory.
[1] In favor of the Republicans, I mean. Some Republicans, primarily Trump, did claim fraud in favor of the Democrats, but were unable to find any evidence for their claims.
The mantra for Trump's opponents and the media in the first year or so of his presidency was that "Russia hacked the election".
No, he didn't.
Waiting for the election to be "independently certified", which is what Biden was asked about, is a very different thing than waiting for it to be "certified" without qualification.
Can you provide a reference for that assertion? I tried to find it myself, the best thing I found is this blog post, describing the wording "independently certified" as curious:
https://excessofdemocracy.com/blog/2020/10/what-does-it-mean...
> Suppose that's true. So what? We analyze the results. We apply science. It can't change the result if it wasn't supposed to.
You are implying that it is perfectly fine for Trump to spread wild conspiracy theories with no evidence because if those conspiracy theories are false then they will be proven false and no harm will be done. Except irreparable harm WILL be done to the democratic process in America, because there are tens of millions of Americans who worship Trump as an infallible god and will forever believe his claim that the election was stolen from him, even if all the allegations are thoroughly debunked and no evidence turns up demonstrating any kind of fraud.
There are plenty of people suggesting that Trump should concede immediately without going through any of those processes.
> You are implying that it is perfectly fine for Trump to spread wild conspiracy theories with no evidence because if those conspiracy theories are false then they will be proven false and no harm will be done.
The problem with Trump is that just about everything he says is a half-truth. It won't be entirely true but it also won't be entirely false.
Some people think he does this on purpose. It causes his opponents to cover him by pointing out the part that's false and his supporters to defend him by talking about the part that's true, but in either case they're both talking about Trump and they're both talking about the issue Trump wants people to talk about. I assume that anyone who can figure out how to get Trump to stop doing this would be awarded a Nobel Prize.
The trouble is that it means you can't just dismiss everything he says as a lie, because it's only half lie. I expect that when this is all over he'll have uncovered some modicum of voting irregularities but probably not enough to flip as many states as he needs to win, if any. But he's entitled to try.
Now if only somebody would win that Nobel Prize.
https://en.m.wikipedia.org/wiki/Risk-limiting_audit
I think PA and Georgia already do this!
...demanding sufficient evidence that there is reason to expend resources looking for a problem before taking them seriously, especially if they are the apparently-defeated candidate.
Isn't that what's happening? They make their claims, others dispute them, it's a public debate. The original post was an objection to "talking about fraud" -- but talking about it is part of the process of evaluating whether the evidence is sufficient. The answer could be no, but you don't know that before you hear them out.
> especially if they are the apparently-defeated candidate.
It would be interesting to know of a case in which the apparently-victorious candidate was making allegations of election fraud sufficient to affect the outcome.
If you're a heterosexual man, imagine asking your spouse for a paternity test on your children. And when that goes over like a lead balloon (which it would), imagine replying "So what? We analyze the results. We apply science. It can't change the result if it wasn't supposed to."
> you can't be accused of covering up a fraud that wasn't actually there. Which is what happens in the alternative.
So you think that Trump has no problem complaining about fraud when he has no evidence, but won't complain about fraud after an investigation shows there was none? Not a chance—he'll just complain that the investigation was a fraud as well.
One is of whether fraud happened. So far, the stuff before the courts have mostly been thrown out due to lack of evidence. Which supports the idea of due diligence and that the system works.
The second is whether the system itself is compromised. The problem here is that one can keep irrationally escalating accusations saying that the watchmen of the watchmen of the watchmen are all in cahoots. And we know from stuff like flat earth society that there are people willing to believe anything. The danger here is that if you have a large portion of the population and their political representatives believing that the system is rigged, that they might as well do everything in their power to rig it in their favor, which erodes the pillars of democracy.
Stand-Up Maths on Youtube already beat you to it.
Netflix's "Connected" had an episode which was probably prescient on this topic.
There was an election analyst at the end of the episode, who said that Benford's law is broken when people vote strategically against a candidate or change their vote when their preferred candidate drops out.
That has nothing to do with why Benford's law doesn't work. Benford's law doesn't apply to precinct vote counts because precincts are roughly equally sized and Benford's law only applies to data that approximately follows a power law distribution and spans at least a few orders of magnitude.
Can’t say I would have thought that to be possible 10 years ago. Hah.
The tweet you was about an item of political controversy excessively well-covered in the mainstream media (making it generally off-topic even as an independent story for HN), and was not at all germane to the post about Benford's law that it was offered as a comment on (making it off-topic as a comment on that post.)
There are venues in which linking that tweet might be appropriate, one can even imagine a hypothetical HN comment thread on which it might be on-topic. However, it wasn't on-topic where it was offered.
> Don't understand why simply reporting something as relevant as what Trump tweeted is reason enough to get flagged. when its directly relevant to this article
Trump's tweets -- well, this Trump tweet, I suppose its theoretically possibly he might sometime tweet something where this would not be true -- are completely irrelevant to the utility of Benford's law in detecting voting fraud.
> and what your president tweeted is news worthy
Conventional newsworthiness (because we have a conventional newsmedia) is generally tangential and often a negative to HN-relevance.
> When your president retweets that 2.7 million votes where deleted during this election by dominion software.
Were this a thread about Dominion software, or about the general incidence of electoral fraud, or about this particular election at all, there might be an argument that that was relevant. Its not, and its not.
> I made no claims of its validity
Even if the tweet were otherwise relevant, its validity, whether or not you make an explicit representation about it, is not irrelevant to the question of whether it should be upvoted, downvoted, flagged, or simply left unmoderated when presented here.
This standard doesn't really get held up in other discussions on here, people mention tangentially relevant things on here all the time.
Two, Walter R. Mebane, an expert on election fraud, had a response to that paper: https://www.cambridge.org/core/journals/political-analysis/a...
And three, Mebane also released a short paper reviewing the 2020 results: http://www-personal.umich.edu/~wmebane/inapB.pdf
tl;dr: There are very minor anomalies in some numbers but either way we have to wait till we get the final numbers in before we can do real analysis.