A 🧵 on how terrible and massively error-filled Pennsylvania voter roll data is.
This is important for understanding how much voter fraud there is. A system that cannot prevent innocent errors also cannot prevent malicious errors.
First, the highlights:
1/N
-423 PA voters are older than the oldest known person. 17 are too young. One is yet to be born.
-Almost 1m (12% of PA voters) lack a house number, making their address impossible to verify.
-252 PA voters only list a Post Office Box as an address
2/N
-Thousands of PA voters are registered at single addresses corresponding to homeless shelters and mental hospitals
-42% of likely PA college undergrads who registered to vote with an on-campus address are still registered to vote from their dorm at age 24 or older.
3/N
-14 PA voters have no surname at all, or their surname is a number or Unicode character
-17k PA voters have a phone number with the wrong number of digits, or numbers like “1111111111”
4/N
-370K PA voters registered to vote on a federal holiday when government employees are not working. 75% of these are on New Year’s Day specifically, where “Jan 1” probably means the date is missing.
-15% of all voters (1.4m) have some kind of critical error
5/N
First, a decent working definition of fraud is “wrong data entered for malicious reasons”. A decent definition of data errors is “wrong data entered for innocent reasons”. Cleanly identifying the errors is much easier than identifying the intent behind those errors.
6/N
But the rate of errors should be quite informative in your beliefs about fraud. Because it speaks directly to the question of “how easily can wrong data be entered here, regardless of motive?”
7/N
A system with many safeguards in place against erroneous data necessarily makes fraud much harder to commit. If you can stop wrong data being entered in the first place, you stop both innocent errors and malicious errors.
8/N
A system that allows lots of insane data to be entered into the official databases without triggering any kind of flags or alerts has far fewer protections against fraud. All else equal, this makes fraud much more likely.
9/N
Finally, we are not talking about what theoretical protections are in place. We are talking about the official voter roll data of who can vote in Pennsylvania in the upcoming 2024 election, as of October 17th 2024, as supplied by the Pennsylvania Department of State.
10/N
1. Birthdays
The oldest person currently alive, according to Wikipedia, was born in 1908. How about among Pennsylvania voters?
11/N
Pennsylvania has 262 registered voters born in 1800, 45 born between 1801 and 1899, 67 born in 1900, 22 born in 1901, and 27 born between 1902 and 1907. All up, 423 registered voters are impossibly old.
12/N
At the other end of the scale, to be eligible to vote, you must be 18 by election day, making your birthday before November 5, 2006. By contrast, 14 registered voters have birthdays in 2007 or after. One has a birthday in 2091.
13/N
2. Addresses
First off, against explicit Department of State requirements, 252 people list only a Post Office box as their address. 236 list the PO Box as their physical address. 16 list a PO Box as a mailing address, but no physical address.
14/N
This is the tip of the iceberg of bad address data. Of 9,033,760 records, 11% of all addresses (997,724 voters) are missing a house number. 92,099 are missing a zip code (for both mailing and residential address), and 66 are missing a city.
15/N
Some addresses have large numbers of registered voters. While most seem genuine and innocuous (universities, old age homes, large apartment buildings, religious communities), others are weirder.
16/N
134 voters have their address as the Torrance State Hospital, a psychiatric facility, for instance. 2,249 are registered at 315 S. Broad St, a mailing service for the homeless. 886 are registered at 2913 Kensington Ave, Philadelphia, a homeless shelter with 60 beds.
17/N
A lot of addresses with many registered voters are university campuses. I identify a partial list of 16,843 voters with an on-campus address at Penn, Temple, Drexel, Lincoln, Franklin and Marshall, Swarthmore, Dickinson, Gettysburg, Widener, Shippensburg, or Lycoming.
18/N
On its own, this is allowed by PA Department of State guidance if this is your home address. You might quibble whether an undergraduate dorm is actually your “home”, but let’s let this slide for now.
19/N
I call you an “undergrad” if you currently list a campus address, and were aged between 17 and 22 when you registered to vote in PA. 98.7% of campus addresses were registered at this age, consistent with campus housing at many universities being mostly for undergrads.
20/N
How old are these former PA undergrads now, while still registered to vote from a campus address? 42% (7,056) will be 24 years old or more at the election. 29% (4,771) will be 26 years old or more. 5% (885) will be 30 years old or more!
21/N
It is possible that some of these are professors or graduate students living as resident assistants. But do you really expect 42% of them to still be living on campus (not just enrolled at the university, or in off-campus housing) two years after their likely graduation?
22/N
These are not just dead addresses either. Of 9,273 undergrads that voted in the past, 6.5% (601) were 23 or older when they last voted from a campus address. 3.8%(356) were 24 or older. 2.2% (213) were 26 or older. 0.7% (65) were 30 or older.
23/N
3. Names
There are lots of weird names, for sure, so a lot of seemingly strange names are probably genuine.
24/N
For instance, 77 PA Registered voters have surnames of 1 character or fewer. The most common is “O”, at 29, which is actually a Korean surname. 18,280 registered voters have first names of one character or less.
25/N
But even here, there are some cases that are odd and hard to understand. There are 9 surnames that are entirely blank, one surname of “6”, two of “7”, one of “9”, and one of “\”.
26/N
4. Phone Numbers
The Department of State says listing a phone number is optional, and only around 42% of voters list one. But there appear to be no safeguards against entering crazy or fake numbers.
27/N
16,704 voters included the wrong number of digits. 787 PA voters entered phone numbers that were either “1234567890”, “9876543210”, “0000000000”, “1111111111” or other versions of the same digit repeated ten times.
28/N
5. Registration Dates
Even the timing of voter registration is puzzling. Since 2011, 368,760 people somehow officially registered to vote on a federal holiday. In 2024, where it’s easy to find the data, these are all also holidays for Pennsylvania State Employees.
29/N
We know that the forms weren’t being processed by government employees on those days. But even if the number represents something else (e.g., when someone filled out the form?), the patterns suggest a lot of them are errors or missing data.
30/N
Specifically, 75% of them are on New Years Day. It seems likely these are actually just missing data being labeled as “January 1st”. While there are reasons why birthdays might be masked, it’s very hard to see why registration dates would also be.
31/N
It’s worth noting – these are the errors you can identify just from voter roll data alone. If you merge with other data sources, you can find many more. “Stop Bogus Ballots” (@bogus6199) has done great work here
They find 2,405 people registered to commercial addresses, 346,505 invalid addresses, 27,672 people who moved and left no forwarding address, and 262,488 who permanently moved out of state. They analyze a number of other states too.
33/N
6. Totals
How important are all these errors in practice? Let’s limit ourselves to critical errors affecting the integrity of the registration: missing address data, only a PO Box, impossible ages, undergrads above age 23, and registration dates of New Year’s Day.
34/N
We get 1,362,907 cases of critical errors, equal to 15.1% of all registered voters in the state of Pennsylvania.
35/N
Consider just how many different types of errors we have here. It seems quite likely that many of these are simple screw-ups. And indeed, the more egregious the violation is, the more you might think this is true.
36/N
For instance, if you want to come up with a fake birthday to sneak in fake voters without raising suspicion, “January 1st, 1800” is about the last date you’d pick. Maybe it means the person was born somewhere without a birth certificate and doesn’t know their birthday.
37/N
Maybe it’s the result of some court order, like witness protection, or the Address Confidentiality Program for victims of domestic violence, stalking or abuse (though as the name suggests, this focuses more on addresses)
38/N
But if the system will allow you to enter a birth year of “2024” or “2091” or “1800”, and nothing happens, and that makes its way into the official voter record… what else does it allow to be entered?
39/N
If voter rolls are presumed to be so error-filled that crazy numbers can be entered, and defenders of the system say “oh, it’s probably a typo or missing data or a court related confidentiality order”, then how is anybody meant to use any of this data to identify fraud?
40/N
If there is no data that can be used to detect fraud, then why should you have any confidence that fraud is being prevented in the first place?
41/N
How sure are you that bad actors aren’t able to also register much less insane fake information? If you can’t rely on checking the data yourself, you’re left with the state’s assurances that the process is airtight and only valid entries can be recorded.
42/N
Like, you know, officially registered eligible voters who are born in 2091, or with a surname of “\”, or who list a Post Office Box as their only address.
43/N
A lot of the interpretation of these errors comes down to which side is presumed to have the burden of proof. People who claim that there’s very little evidence of voter fraud rarely actually provide any concrete evidence that the errors *are* innocent mistakes.
44/N
They might be! But they also might not. At a minimum, the argument that you should have confidence in the integrity of the voting process because of all the ironclad safeguards in place to prevent fake voters being registered seems extremely dubious.
45/N
If the system of voter registration is so broken as to allow this many different types of errors into official data, this close to an election, I argue that the burden of proof should go the other way.
46/N
The State of Pennsylvania should explain why anybody ought to have faith in the security and integrity of the election process in the state, given the almost total failure of basic error checks. I have no idea how they would do this.
47/N
Pennsylvanians deserve much better than this. Unfortunately, this close to the election, I do not think they will get it.
48/N
If you liked this thread, please re-tweet the first post!
I am increasingly convinced that one of the worst societal choices the west made was deciding that housing should be a vehicle for generating investment wealth, rather than something that stays as cheap as possible.
The related problem was trying to square the circle of "house prices should go up" and "housing should be 'affordable'" by subsidising loans for housing, which just makes the cost problem worse.
"Affordable" comes to mean, in practice, "I can get my name on the title deed, notwithstanding that it takes me longer and longer to pay it off". This is very different from "cheap".
There is a certain kind of opportunistic genius I associate most with the Greeks, in this case an old friend of mine.
When reflecting on Germany/Costa Rica game, once Spain lost to Japan, Germany couldn't go through. But they still had 5 minutes to play. What could they do?
They had enough time to turn around and score three own goals, to make sure that Costa Rica won, and then Spain wouldn't go through either.
This would have been a wonderfully Greek move - out-of-the-box thinking and willingness to endure embarrassment in order to punish hated European rivals.
A friend asked me a few days ago whether I was planning to look into voter fraud this election. I replied that I wasn't. I came to the conclusion last time that it wouldn't matter what we documented, it wouldn't change the result.
Unless State GOP parties did something to stop this kind of stuff in advance, it was hopeless. And sure enough, the states that were useless and dubious last time have chicanery and surprising results again this time. Pennsylvania. Arizona. Wisconsin. Michigan.
I spent days last time trying to get the @PAGOP account to tweet something, anything, about the fraud taking place in Montgomery County. Nothing. Not a peep.
"As Wallesteimer described the atmosphere in the mid to late 2020s, 'From here on out, both parties' leaders began to suspect that if they lost power, they were liable to lose their freedom, if not their lives. ...
After reaching this conclusion, they began to justify their own escalations as being a necessary precaution against the presumed intentions of their opponents. This in turn justified those opponents in their own beliefs, and their own escalations. ...
If Trump wants to do one last good economic policy, he should put @JohnHCochrane on the Fed board with the sole mandate of implementing Narrow Banking, come hell or high water.
Actually, giving him a broad mandate would be better, but this would be a great place to start.
First, by using the Supreme Court original jurisdiction, it at least forces them to take a position on it. This can't just be slapped down by some no-name judge in Hawaii.
And if there's anybody with the social authority to overrule these states, it's the Supreme Court.
Second, this forces the media to do something they've studiously ignored - focus on actual serious allegations going on, not just ignoring the issue or cherrypicking the stupidest claims.