Voter Roll Trivia

The more we know the facts, the more we can bring truth into the bright light where it belongs. This is our effort to engage and we are doing it by sharing fun facts and only facts. Res Ipsa Loquitur!

Q39 - State party affiliated data is mostly not so good and sometimes not accessible. How would our county, or a candidate, create a fresh and accurate local voter data warehouse?

A.    Great question, great idea.

Having your own "special sauce" of data will make a positive difference when helping selected candidates get into office. There are three different ways to do this:

1) Hire a firm that offers a turn-key solution. They do all the data maintenance, send mailers, integrate with mobile canvass apps. This is often the most expensive method especially when targeting a large audience because carpet-bombing lit and phone banking add up quickly.

2) Partner-up with a mobile canvass app and let them insert whatever data is available so door-knockers and phone-dialers can accomplish campaign objectives. This approach is weak: It's not their job to ensure your data is useful. Their mission is to charge active users per month.

3) Develop and maintain your own data, infuse it with other data points and then maintain that by adding feedback on a frequent basis to improve data efficiency. Then, as-needed, this data can be used to create walk/dial lists for the boots on the ground. Also, this method does not preclude the use of 3rd party mobile canvass apps. What makes this the best choice is that you optimize the data before you deploy it to the volunteers.

All of the above are potentially viable however each solution will influence data overhead and quality. Which one is best is based on intended objectives, skill level of those available who will assist and available funding.

Here's a quick sketch of what's involved. To maximize data efficiency The Core data which usually begins as a WEC voter roll, must be joined with other data from different sources.

What's important but not obvious in the image is that as volunteers generate canvass data, it must be fed back into the system so data quality continues to improve.

Here's a PDF of the above image that contains a few more details. If you would like some help to better understand what's involved, call or text 262-421-1222.

Q38 - Is still around? How is different?

A. was a promotional domain name that was used to launch the prototype version back in October, 2021. When we rewrote most of the code and reformatted data for all 72 counties, the time was right for a non-themed domain name.

What we are working on is being able to quickly generate "Street Summaries" which is a concise report for a single address. It shows Voter Status Reasons along with the quantity of those types. Here's an example of 100 Grant Street, an admin building on small campus. That's a non-trivial amount of voters where zero students reside.

Anytime we use the Voter Status Reason of MERGED, we make sure to continually remind others that a MERGED voter record deletes the voter history. This flies in the face of WEC's claim that they NEVER delete data because the voter roll is a historical archive. Scroll down to Q37 for more about the intentional practice of scrubbing vote history. is still active and it explains some of the above.

Q37. WEC continues to state that the old/incorrect data can't be deleted because the voter roll is also a historical record. Deleting a voter from the roll can't be done because it would potentially disenfanchise a citizen. Is voter data ever deleted?

A.    If the voter roll is a precious critically important dataset, then why do they delete voter history on a regular and frequent basis? When a WisVote record is "Merged" ALL vote history is deleted for that voter's name at that address.

For example, if you voted for 20 years on Elm Street, but then changed to Merged, the vote history is "transferred" to a new VRN sometimes at the same address. What follows is an exmple of how voters end up with multilple Active VRNs.

Let's say you moved from Elm to Main St, your Elm voting history is wiped clean and inserted into Main St. Now it appears you have been voting from Main St. for 20 years which is factually incorrect.

As of Feb 2023, here are some WEC voter roll numbers

  • Total Voters= 7,437,585
  • Inactive = 3,828,207
  • Actives = 3,609,378
  • Felons = 34,656
  • Merged = 1,262,590 <-- this many voters had their voter history deleted
Let's use Mr. 09Bainer as an example. This is what we call questionable data. Source:

VRN: 700218350
Name: 17Kyle 1988William 09Bainer (yes, 17Kyle is the name in WisVote)
SUN PRAIRIE WI 53590     (Dane County)
Status/Reason: Inactive/Merged

The Merged action erases all voter history attached to that name/address. So the VRN ending in 8350 has no voter history. (Notice this which means in the last 81 elections, it appears he has never voted -- even though he did, six times)

17Klye gets a new VRN with SAME address but with merged voter history. Why not just correct the typos with numbers in the name? Why a new VRN at the same address? Here's his "corrected" data with a new VRN. They left 17Kyle in the voter roll, just flopped him to Inactive and went about their day creating a new VRN identity for the voter previously knowns as 17Kyle.

VRN: 700815510 Bainer Kyle W
<= Same address as 17Kyle SUN PRAIRIE WI 53590
Status/Reason: Active Registered 3/26/2019

Also active -- but this one with no voter history. As per public record Kyle W Bainer now resides at a different address. Why is 700815510 still Active?

VRN: 702052237 Bainer Kyle W
Status/Reason:Active Registered 1/14/2023

So, here are the three Bainers, two are active and #3 has voted six times.

In conclusion: By design of the WisVote system, WEC clerks delete voter history each time they flip a voter to Merged. As of a recent voter roll, that's at minimum 1+ million deletions. Let's assume Merged voters cast ballots 30% of the time, that's about 5 million data points that have been deleted. Poof gone! Where did that voter history go? You don't know unless you do some non-trivial research.

Q36. VBAN numbers that don't add-up for Sawyer County. Wards 1-6 do not cross-foot for ballots and votes. Is this a bug in the website or is the core data messed up?

A.    Here is the screen-cap that was referenced above. It shows two voting locations that are not typical. Columns for Ballots and nV (number of votes) should be about the same. Why are these not behaving like the others? That's a darn good Q!. We believe we know the answer, but let's first address the question.

What you have uncovered is of possible significance. VBAN shows data but it's up to people like you to uncover patterns that need a second look.

In the PDF you provided about Sawyer absolutely has numbers that do not align with most other polling locations. When we created VBAN, it wasn't designed to find what you have found. Should we do another version, we will implement some of the tech we have built into

If you look at the far right columns in VBAN, you will see that total ballots counted by method (scanned, hand count, touch screen) do add-up. What you provided in the first four columns are big outliers. That is a potential score, but without more intel, conclusions should not be drawn. But you have discovered a pattern that is worthy of a deeper dive.

At most polling locations, those left-most columns should add-up or at least be close. (see yellow arrows above) Those reflect voters who were registered and showed up to vote on election day. From what you provided, there was a flood of people who jumped out of bed on election day and decided: Holy sweet cheeses! Today I will get dressed and register to vote... and VOTE!

Is that possible? 100%! But if you believe that, we have some math peeps here that would question your police work, Lou.

So now you are at a crossroad. You can take this information to a local clerk and ask them to confirm/reject the numbers. If you have connection to a friendly, they may be willing to assist on the low-low. Absent that, you will likely need to submit an ORR. That goes down on your permanent record.

Be advised, you are questioning 2020 data and that puts you in the room with the others who question "settled" voter data science. At best you're curious, at worst you get put on the tinfoil watch list and bunch of dudes in SWAT gear visit you at 5am grab your hardware.

I salute your efforts and we will supply you with whatever data you need so you can continue your march to the magical Town of Truth and Awarenessville

Q35. has a Status Reason of Moved and Movers. What's the difference? Can Movers cast a ballot?

A.    Let's begin with some definitions. Movers are marked as Mvrs(ERIC) in GhostFinder are people who might be moving as detected and reported by ERIC, the DC-based rogue splinter organization to which we pay $75,000/year for membership. Here's the official movers memo from ERIC telling clerks how to update our data.

Moved are those who have been "confirmed" by a clerk that that they have indeed left the state.

Some history to provide a comparison

So to the first Q, Movers are provided by ERIC and the details of that algorithm is unpublished to the public. A Moved reason is the result of an edit by a clerk from data provided by ERIC, WEC or elsewhere.

The other Q: Can Movers cast a ballot? Any active voter is eligible to vote, so on that one: YES

Q34. There are rules to how you build a database and much of it centers around normalization. Does WEC follow the rules?

A.    With regard to WEC and the voter roll, we don't fully have an understanding of the database design. All we can do is infer from the CSV export file WEC provides for $12,500.

Here are some 1st Normal Form Rules:

  1. Using row order to convey info is not permitted
  2. Mixing data types in the same column -- not allowed
  3. No primary key -- not allowed
  4. Repeating groups -- not permitted
How does WEC score with regard this the 1st rule?
  1. Not sure. They might not be guilty of this one, we're checking...
  2. The Voter Registration Number is sometimes a number, but more often a non-number. They are in violation
  3. The Wisconsin voter roll contains no primary key. They are in violation
  4. Voter history uses repeating groups. They are in violation

Bottom line, the WEC voter roll fails the most basic database design test. However, we can't draw a hard conclusion because a CSV export could be an intentionally flawed version. The $64 question: Does the WEC data indicate ineptitude or intentional obfuscation? Our default answer: Both

Q33. People are all pounding their fists on the table to "clean up the voter rolls!" but is that being done? Is WEC or anyone taking action to improve the data?

A.    Let's use duplicate VRNs (voter reg numbers) as an example. Specifically, we will use Kristen who in 2020 had 11 duplicate VRNs. Here's a snip from way back when

Why would a voter database have the same person with 11 identical voter numbers? This is clear evidence that Voter Reg Numbers are not unique. There were 1000+ examples just like this in 2020.

Fast-forward to SEP22 WEC data. Here's how they modified Kristen to not have so many pesky duplicates...

The person/people behind the curtain pulling the data update levers decided the best method to eliminate duplicates is to swap out a leading digit for another. In this case they removed the 7 and replaced it with 2-6 and 8-11. For a reason few will ever know, they swapped out one with 51, I guess because using a 7 would have not solved the duplicate problem.

Not very creative or subtle, but effective as WEC can today state there is only one voter with a duplicate VRN. And as we all know, one occurence of something is referred to a glitch, something that can never be called widespread.

So answer: YES, they are making changes. You decide on the value of said changes.

Q32. When I load WEC data into Excel, there are many duplicate VoterRegNumbers. Is this correct?

A.    A WEC Voter Registration Number (VRN) is not a number. As per the scant WEC documentation...

A Voter Reg Number is a non-number customer number encrypted with a number.

Here's the official PDF from the data masters at WEC.

Just so we are on the same page... an alpha-numeric is not a number. e.g., A12 is an alpha-numeric because it contains one character that is not a number. A number can only have 0-9. Anything other 0-9 means it's an alpha-numeric. Another example: A phone number 2625551212 is a number, but (262)555-1212 is an alpha-numeric.

Now, back to the question...

If one imports a CSV file into Excel that contains 123,0123 what will be the result? It will look like this:

Analogy: There are two people in the room. You ask them to write their ages on a slip of paper. One writes 32, the other 032. Which one is older? Answer: Not sure because one person works for WEC.

Back to our CSV file... this is where some make inadvertent mistakes. When one purchases WEC data, it is provided in a CSV. Let's say someone purchases all of a single Assembly district which might come to 66,000 records. It's possible that file will contain 123 and 0123. When you load the CSV into Excel/OpenOffice, those will appear to be duplicates. BUT, as per WEC, they are not the same because a VoterRegNumber is not a number.

Let's look at what we call the BFC (Bumble Flock Coefficient) inflection point, when we see alpha-numerics become numbers. Below is an image from WEC Sep22 Dane county data where the yellow line indicates where data flips from alpha to numeric.

Yes. You are reading that correctly. As of Sep 2022, Daniel has a VRN of B. Usah has N425-8573-0964- and Kendal's VRN is NEW.

Just after Kendal, the data is all numeric, starting with Thomas who has the awesome VRN of 2!

From the above, you will note that most VRNs begin with one or more zeros. Looking at two state-wide lists, the average amount of numbers that are not numbers is about 70%. So about 5.4 million VRNs begin with one or more zeros.

Now for the $64 question. As per WEC, a VRN is encrypted with the registration number. Let's use Daniel Siehr as an example.

How is the letter B an encryption that contains a number?

One more just for fun: Note in the Sep22 summary above where oooooooooo has quantity = 1... that means one person in the voter roll has a VRN of ten zeros ... 0000000000. The All Zero Award goes to Alisha Loria who has an address in Wauwatosa.

So the answer is MAYBE because older data did have duplicates. As of today, there is only one duplicate in the most recent voter roll.

Q31. What is the origin of the fraud icon showing date curves that have been fashioned into the letter F? Can or StatBandit data recreate this graph to duplicate these?

A.    NO. and StatBandit both use WEC data to generate output. To our knowledge, WEC has never published, and nobody have ever asked them to produce, the data points needed to generate those curves. Do they have the data? Of course they do. However sharing those details would point the finger at the source of the "spike protein" anomaly.

To generate that graph for Wisconsin (or any other state), you can go to the NYT and download the raw AP data. Here's the DATA for Wisconsin and here's the now famous fraud curve we generated from said AP/NYT data.

Let's zoom in on that spike protein anomaly so it's clear where we have no answers. Look at the line that appears to be the trajectory of a rocket launch sometime around 330am. Where did those originate? Most assume somewhere from Dane and Milwaukee county.
Do other states have similar "F" curves? Yes, but only a few. For example, if you look at Kansas data from the same data source, you see what appears to be a normal set of data points.   SOURCE
MyVoterTool and StatBandit both use the official garden variety data provided by WEC which are static summary documents. In order to produce graphs with data that might have conclusions, we need time-stamps and source. e.g., What data was uploaded by WEC to wherever at 3am AND where did that data originate? Those two questions, now long after the election, will likely remain unsolved mysteries.

Q30. StatBandit claims 94% voter participation. Others have said voter turn out was under 70% -- which they say is normal and as such, that StatBandit numbers are utterly absurd or ridiculous. Please help reconcile these two views. Which side is correct?

A. Darn good Q! We reached out to the mega data slayers at StatBandit and here's the skinny. The 94% and the under 70% are not the same metric. Let's dig in...

The under 70% crowd are referring to these numbers -- numbers you will never find in StatBandit:

While StatBandit does have population data, it's not used to derive votes/regs percentages.

We are fuzzy as to the exact reference to 94% as has no singular overall conclusion of 94% -- so this caused us to ponder: Where does 94% exist in the WEC voter results?

e.g., A Dane County search show these two polling locations scored 94%.

City of VERONA		Wards 6-9 	(VERONA CITY HALL)
Before we begin, let's agree: There were two types of voters

  • those who were in pollbooks at 8AM on election day
  • those who were not

Those not registered, but they could show up and register and vote. Those are referred to as EDRs (election day registrations) who represent a group of go-getters who woke up on election day and proclaimed,

"By golly, even though I have done nothing to prepare, I will go register and vote today!"

Let's look at Verona and Middleton

5038 ballots were cast at Verona City Hall, 4837 for Middleton. They voted absentee 82% and 78% with total pollbook-ready registrant quantities at 5362 and 5122.

If we toss the EDRs (288 and 144) into the number wonkulator, you're still in the 94% neighborhood with 89% and 92% respectively.

Set aside the 94% ... how about someone bring up the MANY polling locations that reported 100%? Need examples? They're easy to find. Here are a few where registrants = ballots... that's 100%

True they were all tiny polling locations but let us never forget:

The 2020 election was decided by only 6 votes per polling location.

Small numbers matter

PS. Read the next Q and A. That one was also in the 94% neighborhood and the numbers matched from AP/NYT.

Q29. vote/ballot data is not correct. You should correct the data or take down this website. (not really a Q, but rather a comment via email which is just a good for me! Comments you can ignore :)

A. Anyone who thinks vote and ballot counts are messed up, they should send their belly-aches to WEC. Everything in are numbers obtained via WEC.

However, it's always good to check your work. To make a comparison, we downloaded the AP/New York Times election feed and randomly selected one polling location: Town of NEW CHESTER Wards 1-3 (Adams county)

Vote totals were 100% match. We could do more to verify the data, but we already did this a year ago. If anyone finds any bloopers, be specific when reporting them 'cuz we aim to please.

Q28. As per the WEC voter roll, what is the average age of the registrants for Racine County, D63, Village of Rochester?

A. Trick question! Your answer should have been a question. Which age?

The WEC voter roll has a date, but it's when (maybe) the person registered to vote. WEC knows, but does not supply registrant DOBs -- which probably is a good thing, however the last four of the SS would have been very handy when searching for duplicates.

So age of the person based on WEC data? Unknown

However, using NON-WEC data, a sample size of 2,993 registrants, the average human age is 55 years.

Wait, we have more.

The average age of the human at time of registration: 45 years and the average time since they registered was ~10 years ago. And now for the bonus round...

Let's set a benchmark so we are on the same page regarding date math: Q. What's the average of Monday to Friday? A. Wednesday

What's the average date of all the registration dates for Rochester as per our NON-WEC data? July 30th, 2011 which means registration date age, as of this post, would be ~10 years, 8 months.

So your take-away: Avoid using WEC voter roll dates to calculate voter age.

Q27. Is it true that counties with lots of FIDO keys are potentially guilty of or planning election hanky-panky?

A. Maybe. But let's make sure we start with FIDO-101.

A FIDO key is not some magic starport that beams you into the deep dark underbelly of the WEC voter roll.

If you are over the age of 40, you might recall a term known as a dongle. Back in the old days, software was distributed on disks. So if you wanted to purchase a C or Pascal compiler from Borland, you strolled into Borders, plopped down $99 and walked out with a fat-stack of manuals and floppy disks.

What did Borland do to prevent you from giving those disks to someone else? Zip, zero, nadda. That was a spot-on example of a non-dongled product because the price of the product was low. Dongles did not grow on trees.

But if you puchased some point-of-sale application for $30K, they would sell you "seats" via LPT dongles. You want four printers in the kitchen? That will cost you four dongles + the $30K.

For any application to grant access for a device, it needed to be able to confirm a dongle was connected. A FIDO key is the same thing except it's plugged into a USB port. These days, machines with LPT ports are mostly found in computer junk yards, but USB ports, they're everywhere.

Anyone can purchase a FIDO key. You can grab one from Amazon for $20 and maybe have it delivered tomorrow. A FIDO key has no value until it is associated with a user. A FIDO key must be used with a legit user ID and password for access to a FIDO-enabled application like WisVote.

So the quantity of FIDO keys is fully unrelated to unauthorized access. Where hanky-panky would occur... you only need to ask this question:

What individuals (clerks et al) were given a userID and PW to WisVote AND what level of security did they have?

If (darn big if) we knew that, we could also ask WHEN and WHERE were those IDs were used to change data.

So instead of asking how many keys are issued, how about we ask for a list of EVERYONE that has/had access to WisVote since 2020. IMO: That list is just as likely to appear as a bunch of new hi-res Zapruder frames.

Q26. How many registration numbers are in the WEC voter roll?

A. Ha! Trick question! Silly rabbit. WEC voter registration numbers are not numbers, they are a mixture of characters and many appear to have been produced by a room full of over-caffeinated monkeys pounding on keyboards.

The WEC voter roll has about 7 million records, and it keeps growing every darn day. Of those 7 million records, about 1.3 million of those are not numbers. That means about 18% of WEC registration numbers are just plain goofy.

WEC data wrangling tip of the day: If you are performing diff comparison reports, be aware that string matching will yield output that looks like monkeys helped.

Q25. What does 148,216 represent as it pertains to this image?

A. As of 3/18/2022 ... 148,216 is the number of "net" additional people added to the WEC voter roll since the Nov 2020 election. Very few are aware of what net means.

Good news! Until they raise the price, the cost per registrant continues plummet like the Russian Ruble. Current market value: $1.75 per 1000 while the price in 2020 was a mind-blowing $1.79 per 1000.

Q24. Do other states charge $12,500 to purchase voter roll data? Is there any other state more greedy than Wisconsin?

A. No. Wisconsin is the only state that charges that amount. There are 22 states that charge less than $100 and 11 of those are no charge. However, Wisconsin is not the most greedy as Alabama took 1st Place at $37,000 per copy.

If you wanted to acquire data for all 50 states, assuming unrestricted access, the total price tag would be about $120K.

Source: Availability of Voter File Information Oct 2020

Q23. Are there any counties with 1/1/1918 registrants that ALL VOTED in 2020?

A. Yes. We selected five counties and as suspected, Dane and Milwaukee rose to the top of the leaderboard. Every one of the 35,999 registrants who have an application date of 1918 in those two counties voted. Are there more in the state? Likely, but these two counties answered the question.

Is there anything wrong with this? Yes. The data is inaccurate as there are no humans alive that are 122 years of age. WEC has admitted this is just a data glitch, part of the conversion process of moving from the old system to the new. Why don't they fix the glitches? That's a question someone should ask WEC.

Q22. What are the top 3 counties in registrants (2021 WEC voter roll) to population (2020 census)? (these population numbers are for all people, not just those of voting age) Source

A. Oddly enough the top 3 in our list: Vilas, Waukesha and Winnebago. This does cause us to scratch our heads. Why would most have a ratio of > 1?

We surmised that throwing in the ballot totals might shed some light. No light was shed.

Hmm. Back to head-scratching.

Q21. Which city received the most Zuckerface/CTCL cash to help ensure a COVID-free election in 2020?
BONUS: What was the amount per voter? (true, this technically is not a crappy WEC data question, but variety keeps us sane)

A. Duh. The city of Milwaukee received the lion share of the "keep elections safe" booty. Of the CTCL grant, Tom Luxembourg Barrett and gang received about 34% of the total. The amount per voter? The CTCL budget for Milwaukee was $13.82 per voter. Big shout-out to city of Racine that bitch-slapped the competition with $80 per JB vote.

Here are the top ten cities (population) that received 92% of the total $10 million zuckbucks, which represents 22% of total ballots cast in all 72 counties. Whoa - sucks bigtime for Oshkosh, the city that wasn't invited to the prom.    QUESTIONS/FEEDBACK

Source: Ballot/Vote data provided by WEC via StatBandit

Columns defined
CTCL '16 Basis: The CTCL budget allocation
Zucker Award: Amount that went to that city
Adult Population: Census data of peeps who can legally vote
WH, BL, HI: WHite, BLack and HIspanic
PreRegs: Voters registered before election day
ED Regs: Voters who registered on election day
TotalRegs: You can do the math
2020 Ballots: Total ballots cast as per WEC
TO: Turnout (ballots/totalregs)
$/Vote: Comparison to CTCL budget using total votes
J Votes: Votes for JB
$/JV: Allocation of $ for Joe votes (because we all know, this was their target)

Q20. What is the most unusual duplicate registration "number" in the WEC voter roll? Bonus points: How many?

A. This was one of those trick questions and the "number" was the clue. Congrats and big shout-out to Sally from Iowa county for picking up on that. The answer is, the most unusual "number" in the WEC dataset is NOT A NUMBER because, Western Technical College is found in the Voter Reg Number field. So bottom line: The data element WEC refers to as the voter registration number is not a number.

As per the official WEC database documenation, the field known as Voter Reg Number is not a number. Here's how they describe it: Alpha-numeric customer number encrypted with voter registration number

So when you see a Voter Reg Number like 701472814, do not let your eyes play a trick. That is not a number, it's what's called an alpha-numeric. As for their definition, that the number is encrypted with the number is kinda weird and Wester Technical College is a great example.

The bonus We only looked in one county and found ~600 of them. But the question someone should ask, how is this "alpha-numeric" anomoly part of the official voter roll? Two possibilities: 1) Database ineptitude or 2) Intentional obfuscation

Q19. One might assume that old people, especially dead ones, will have a lower voter registation number. Does the WEC data support this hypothesis?

A. We selected Vilas county to see if a correlation existed between registation numbers and registration dates. Results: ~62% which does indicate some correlation.

The county data, via a scatter chart, was not what we expected. Sample sizes: 1,000, 5,000 and 10K.

Because we converted the registration date to a number, we can compute the average record age of all registants for Vilas county which comes to Tue Mar 22 1977 18:14:10 GMT-0500 (CDT) or 45 years ago -- a blistering example of how "stats" can be tasty stupid fodder. Vilas has almost 11,000 registrants that have a reg date of less than 1920. That's like 30% of everyone registered in the county. But we assume these turn-of-the-century numbers are WEC-plugged values and as such, do not represent the true age of the record.

For comparison, Jefferson county has an average registrant date of Tue May 25 2004 06:13:44 GMT-0500 (CDT) or 18 years ago. Here's a scatter of 5000 registrants. The total county, of ~90K registrants, has a wimpy correlation of 47%.

So, does WEC data show a date-registration number correlation? MOSTLY NO

Q18. Gobble gobble. How many people registered to vote on Thanksgiving? (this one was published just after T-day)

A. Thanksgiving is one of those events that doesn't fall on the same date each year. This year is was on the 25th. Last year it was the 26th. In 2019 it was the 28th and the year before that... it was the 22nd.

The answer: We don't know. We never even tried to find the answer. The only reason for this question was we needed something to post next to the cartoon turkey. Hope you are enjoying the left-overs.

Q17. Milwaukee county is #1 on the leader board with 282,886 registered voters who applied before 1920. Of all pre-1920 registrants, Milwaukee accounts for 49% of them. How many other counties combined equal that same percentage?

A. After Milwaukee, it took the next 36 counties to equal that quantity pre-1920 registrants.

What's interesting maybe is the ratio of really old people to the total registrants for each county. So for Milwaukee, which had 1,378,269 on the voter roll last November, only 2.4% were really old.

Yes, we have heard the tales of the 1918 dates. As per WEC folklore, when they converted the old to the new, many application dates were missing. So instead of leaving that field blank, someone had the idea of using 1/1/1918 as a default date.

Ok, fine. Let's go with that. If that's the case, one would expect all counties to have a somewhat similar ratio of really old people to total registrants. How is it that Door and Vilas counties average 32%? (click the image to zoom)

Milwaukee had 1,378,269 registrants -- the other 36 had 4,570,626. Milwaukee had almost 50% of the WEC-generated pre-1920 dates. If anyone has any theories, we would love to hear 'em.

Q16. How many views does this triva page generate? (in part due to

A. As of today here is the click volume along with the distribution of where they originated:

WI(20%)  WA(9%)  OR(4%)  IL(4%)  US(63%)
Canada(13%)  Ireland(1%)  Germany(3%)  FBook(1%)  UNK(52%)  CCP/RU(10%)
Chrome(30%)  Win10(17%)  Win7(7%)  Android(7%)  iPhone(5%)  Mac(5%)

CCP/RU are clicks from China and Russia. Unknowns are IPs that did not resolve to a location and likely represent those dastardly search engine/spy bots.

Q15. What's an example of a false positive as it pertains to multiple people connected to the same address?

A. There are many examples of FALPOs when one assumes all USPS adddresses within the WEC data are deliverable. They are not.

A registrant is supposed to have a valid mail-to address. By valid, it means the voter can be reliably contacted via that address. If a registrant entry is challenged, they must respond to a postcard asking them to confirm they are still at that address.

Here is an example of a soft false positive

This address was reported as having 10+ voters connected to this one dwelling. This qualifies as a FALPO because someone did not include their unit/lot/trailer number when they registered. You can test this in by entering 834 E Broadway. Here's the output for that search.

This is a soft false positive because it's possible the USPS mail carrier would recognize the name and make the delivery. A better example would be when a registrant address points to a university or college without a res-hall and unit number.

Q14. What are the terrible ten counties that were direct recipients of the Zukerface cash and what % of total ballots did they represent?

A. We had to toss Waukesha in there because the city by that name received $42,000. The total ballots cast by these 10 counties represents 50.22% of all ballots for the state. FYI: RU is a Reporting Unit, also known as a polling location, includes one or more wards.

Question to ponder: Why did Racine get 22 times the amount of Zucker funding than Waukesha which is twice the population size?

DISCLAIMER: These counties are not literally terrible. We just needed a word that began with the letter T. None of our insightful blathering here is ever intended to cast a dark cloud over an entire group just because one of their cities received large sacks of unrestricted cash from a group funded by Facebook.

Q13. Is it possible to use the WEC voter roll to find irregularities for all elder care facilities in the state?

A. YES, but you would first need the addresses of all facilities in the state. You would then join those addresses with the WEC dataset. Now you have all the registrants for all the elder care facilities.

Downside: That will be a large haystack in which to find the needles. Using one of the five types of licensed facilities as a benchmark, the report will contain ~100,000 registrants.

Q12. In November of last year, there were 6,994,642 in the voter roll. In July of 2021, about 206,000 registrants were purged. How many fewer are in the voter roll today?

A. That was a trick question as it implied the answer is less than 6,994,642. It's possible a "purge" of 206K did happen in July, but without before and after datasets, we have no idea which records were flipped to inactive.

Bottom line: Since November of 2020, the voter roll has increased by 120,842 voters. Total number of registrants is now 7,115,485. So EVERY published article where they indicate Wisconsin election officials have removed is 100% false. When they say "removed" it just means the registrant has been flipped to inactive, and can an inactive registrant be made active again? YES

Q11. How many email addr,esses in the voter roll include a com,ma?

A. Way more than there should be. You don't fling around commas willy-nilly and most everyone knows a comma is not a permitted character in an email address. However, if we published the actual quantity, that would open the door for critics to slam this for being focused on something with no value. So sorry, no numbers for you.

Just so we are clear .. email addresses must include only RFC-compliant characters, which everyone knows are these:
  • Numbers 0-9
  • Upper and lowercase letters A-z
  • Plus sign +
  • Hyphen -
  • Underscore _
  • Tilde ~
  • Period .

Yes yes, everyone also knows-- email addresses are optional. But from a business logic perspective, why would WEC allow them? What's interesting is the public registration website DOES NOT ALLOW a comma. So that would lead one to conclude these pesky commas are being inserted after they are part of the voter roll or being entered via some Stargate back-door data entry interface.

Oh, and by the way -- there are commas in phone numbers too.

Q10. FACT: There are many versions of the WEC voter roll in the public. One group purchased the dataset 28 times leading up to the 2020 election. With all this data out there, why are we not sharing this information?

A. On multiple versions: YES that is fact. There are many versions of this data out in the wild.

Why is the data not shared? That's an essay question we will answer soon. The answer is a mind blowing, so if your mind is easily blown, you best stay away.

Restrictions on what you can do with the list? NONE After you pay $12,500 for the crappy data, you can go nuts with the info contained therein.

Here is WEC's official position on this matter...


Q9. Is the Wisconsin voter roll big data? Do I need special equipment to analyze it?

A. For most people YES. It is big data because if you tried to load all 72 counties into Excel, your hardware likely will not complain, but Excel will scream like a horny alley cat at 2am. The entire voter roll is about 7 million records, or in spreadsheet-speak, 7 million rows you will never load in Excel. When you consider the full view, meaning to include the columns of data, that comes to about 840 million points of data.

So why YES for most people? Most have no idea what it means to split a file and likely would have never needed to do so.

On special equipment... YES. The first equipment you will need is a phat brain with data analysis and mathematics know-how. Absent phat brains, it's very unlikely you will know or care about grabbing your own version. Here is one of the tools we use to paw through the data.

We use the full WEC dataset (7 million records) and smash through that in short order. If we are building an extract for a county or a custom query, it takes a minute or two to generate a clean subset. When we need distribution plots or group-by summaries, we can cut through all 7 million records in ~10 seconds.

When we need to push data into GhostFinder, we often use a spreadsheet to inspect/clean the data before it's imported into the ghost database.

But getting back to the original question. Is a single WEC set big data to us? NO.

7 million rows is no big deal. However, if we had 10 versions of the same, that would be incredibly useful and would cross over into the true definition of big because one dimensional has just become two.

Who is this we? We are just two expert data guys, one from Mequon, the other from Chicago. We are contributors to, but not a member of that group or any other. We are free agents looking to do the right thing.
UPDATE: We no longer contribute anything to because they are no longer connected to any election integrity projects.

Q8. Why are you so critical and negative? Is it because you are angry because Trump didn't win? All of you who are publishing these random numbers here are meaningless. You lost. Why can't you get over it?

A. Good Qs! We are very critical here, but we refer to it as critical thinking. If that was the meaning, then we are guilty as charged.

Negative? We would need a specific example of that. Perhaps you are referring to our disparaging remarks about the WEC data. That's not being negative, that's knowing what crappy data looks like. We are data masters, we know stinky data when we see it.

On Trump-related anger? None here. Zero. While we believe there was some extreme funny-business with the election data, we're not angry about it. That emotion is contrary to good investigative work. We are in fact, happy data scientists who love the invigorating sound numbers crunching.

On You Lost, Get over it... If a person applies for a position and claims to have received really good grades in school, what's wrong with asking to see an official transcript? One would think the applicant would be eager to have someone discover said goodness via independent verification.

Our mission is to seek the truth and to understand how process failures have created the big ugly dataset known as the Wisconsin voter roll.

Q7. What county has the most entries in the voter roll with no registration date?

A. Contrats to Trempealeau (from La montagne qui trempe a l'eau which means the mountain that is steeped in the water) county who takes first place with a devilishly impressive showing with Dane and Outagamie way back in the pack.

Trempealeau County    666
Dane County            93
Outagamie County       42

Q6. Did 1000 voters register in St. Croix county with a birth year of 1918?

A. NO    But before we tee-off on this one, St. Croix had 5,059 dead people on the voter roll last year. Likely inactive, but that does not mean none of them voted. We can't conclude that with only one dataset -- nobody can.

Now, on to the 1918 thing... there were 5,737 were on the voter roll who would have been about 121 years old for the last election. Same as above, likely all inactive, but that doesn't mean...

In conclusion: A county clerk could respond: No issues here. We found nobody born in 1918 on our voter roll. Case closed. ... but there's always some fine print

The registration date (also referred to as Application Date) within the voter roll is when (maybe) they applied to vote, not when they were born.

Let's assume they were 18 years of age when they applied, that means their registration date would have been 1900'sh.

There are, in fact, no regsistrants who applied to vote in 1900 in St. Croix. So the answer of No issues, Case closed is technically correct which would allow them to skirt the issue of the 1918 registration date. (of which there we 5,737 of them)

One last thing: The voter roll for the state DOES have people with an application date of 1900. Are you ever up at 2am watching drug commercials while wondering about the distribution of the state-wide voter roll who are really old?   Wonder No More

Q5. Have dead people ever voted?

A. 100% YES, but to what extent, we have no idea. When data costs $12,500 per set, it very much hampers the ability to draw useful conclusions. Gosh, it's almost like they do this intentionally to avoid that pesky transparency thing.

Another example of monster bad data quality:

Here is the distribution of voter status for registrants who registerd before 1920. When you register, you must be at least 18 years of age. Therefore, a 1918 registrant would have been born sometime around the turn of the century. That would make them about 121 years old today.

These numbers represent the 37 counties that make up ~80% of the vote. This means, as per WEC data, that there were 100,000+ active voters who, if they actually registered on that date, are older than the oldest known living person in the US, who we believe to be Thelma Sutcliffe. Go Thelma!

It's our hope you are outraged and send hate mail, but don't shoot the messager. We always welcome your thoughts and encourge you to share even if that includes inane bitching and moaning about our outrageous WEC-based numbers presented here.

Q4. Does Racine county have 23,000+ duplicate phone numbers? Is this a problem?

A. YES there were 23,262 occurences of a phone number ending in 9050, most connected to Racine county. However it's important to note that 11 other counties are assocated with this number. But the big news is that this number has been around before 2020 and only a small portion of the registrants tied to this number are active.

Is this a problem? YES. Because if this is promoted, it's potentially an unintentional false flag, or worse, an intentional one. It's never useful when people run with numbers they don't understand. How this originated months ago is the work of a potential goofball or worse, because this voter roll tidbit is borderline meaningless.

One last item: This however is an example of crappy data logic. It's assumed that a family might all use the same phone number. But for WEC to allow this many duplicates is another example of data they willingly choose not to validate.

However, phone numbers and email addresses are optional. Therefore, duplicate entries can be dismissed as nobody cares.

Q3. How many registrants, state-wide, have a zip code of 99999? Bonus points: Which county has the most?

A. Last year 2,405 registrants had a 99999 zip code in their mailing address. The good news, that's a very tiny portion of the 7 million on the total roll. More good news, only 15 of those have a voter status=Active. So total inactives with 99999 was 2,390.

On the bonus points, here's the leaderboard: MILWAUEE 551, RACINE 431, KENOSHA 224, MARATHON 193, ROCK 118, DANE 90

One must ponder at least two things:

  1. Why would any important data system permit the insertion of invalid data AND choose to not delete it?
  2. What makes those 15 registrants special by allowing them to be Active with an invalid USPS address?

FYI: There is no 99999 zip code in the US and is considered a dummy code. The most frequent use for this dummy is in fiction/movies, when the author or playwright wants to use a number that cannot be confused with an authentic code. This zip code has long been used when sending letters to Santa.

The Social Security Number 123-45-6789 has exactly the same function. Bottom line: 99999 is real but is considered undeliverable.

Q2. Did anyone register at a polling place AFTER the 2020 election AND show as having cast a ballot? Bonus Points: What was the average date and time they registered at the polling place?

Update 8:35 AM 11/9/2021 - As per the WEC database person, times are GMT.
We have been informed that timestamps (date/time) in the WEC voter roll uses Greenwich Mean Time, a.k.a., GMT. We had to confirm this because WEC does not share much publicly that explains the data they sell for $12,500 other than this milk-toasty Q&A page.

Here is the data definition WEC provides. You will note the application date type is wrong. So like the data itself, this document contains a boo-boo. It says that Char(10) is the field space for ApplicationDate, meaning the date/time field can only hold 10 characters. How many chars are in 10/21/2020 05:00:00? More than 10.

GMT Defined: This is the mean solar time at the Royal Observatory in Greenwich, London, counted from midnight. At different times in the past, it has been calculated in different ways, including being calculated from noon; as a consequence, it cannot be used to specify a particular time unless a context is given.

From a database perspective, it's assumed with a timestamp, absent a zone reference, it reflects the local time. If you ask Goog what time it is, it will provide the answer with the zone reference. e.g. Saturday, November 6, 2021 (CDT) or a few days later it would be (CST).

When we looked at timestamps in WEC data, we assumed it was local time relative to the transaction that only happens in Wisconsin. Reason for this assumption: There is only one timezone for Wisconsin (well, technically two but only one at a time)

This could be explained that these "late" registrants were in-line when the clock struck 8pm. As per law, anyone who is in line when the polls close, still gets to cast a ballot.

Wait. We're not done yet. Let's toss aside the GMT stuff for now. Here's another question the data begs someone to ask. Here's a snip (click image to zoom) of some Racine registrants who walked into a polling place in August -- to register. What polling places are open months before an election? When voters walk-in to register months early, this is what the data should look like. (source: Langlade)

However, we can't prove we are correct but nobody can prove us wrong. The WEC dataset does not provide information that could have easily documented this.

If you have info or a comment on this give us jingle or text at 262-421-1222

A. This was a good one. The answer is YES. This was and is a brain twister. How could someone register on the 4th but have voted the day before? If anyone knows the answer, please share. Our guess, it would be explained as a one-time glitch.

Click the image to the left to zoom. Let us point out some irregularities that should cause you to squint and scratch yo' head.

The election process: Register, then vote. In this case, it was vote, then register.

Notice the registration timestamp: They arrived in the wee hours of November 4th. Yet all 21 samples were able to vote the day before. How does that work?

Next, notice these registrations happened AT the polling place. This is called EDR or election day registration. Wisconsin is one of six states that are exempt from NVRA a.k.a., National Voter Registration Act.

Riddle me this batman: What polling locations open at 3am?

Bonus points: First some date string math: What is the average of Monday + Friday? If we convert days to numbers, where Monday=1 and Friday=5, the answer is 6/2 or 3 = Wednesday. To determine the average time these 21 registered to vote, we convert the date to something we can average e.g., So 11/04/2020 12:11:58 becomes 1635943631 which is the number of seconds that have elapsed since 1/1/1970. Converting a date/time to a timestamp is easy. TRY IT
The bonus question answer: 1:37AM (In the entire state, ~4000 people were registered on the 4th at a polling location, all between MIDNIGHT and 5AM)

To learn more take a peek under the hood to see when these odd registrations occurred.

UPDATE: The voter roll no longer contains a time. It shows only the registration date.

Q1. What is the first name of the oldest person in the voter roll who is not yet born and is marked as deceased? This is not a trick question.

A. This was a trick question. Just because we said it was not a trick question dosen't mean it wasn't. That's how trick questions work. It's not possible to calculate a person's age from the voter roll because that info is not provided. The big clue was not yet born.

The correct answer is Ruth, who as per the WEC voter roll , will register to vote in about 1800 years, but who is currently in the voter roll as deceased.

Ruth has the distinction of being the only person in the official voter roll who registered in the year 3806. If anyone knows her, please let us know so we can send her an award, posthumously.

FINE PRINT: Our numbers are derived from the November 2020 WEC dataset. We acknowledge they are allowed 45 days to post final numbers. For this reason, we do not use this dataset to draw conclusions about the 2020 general election.

The intent of everything you see here is our attempt to shake the trees and rattle the bushes to create thought exercises. It is to open minds so we can ask good questions to understand and challenge the horrible mess of our voter roll. Are these numbers correct? YES! But do they mean anything? You decide. UPDATE: This trivia page was started in December 2020 and at that time, we had only one voter roll so much of the data here was from the dataset. Since that time, we have acquired many new voter rolls so newer posts here will use the most recent dataset.