Voter Registration Analysis: Cuyahoga County Ohio

Part 1 of the "Nothing To See Here" Series

Purpose

The purpose of the “Nothing To See Here” series is to ask the question:

Can I 100% trust the integrity of the voter registration roll?

Overview

Let’s start with basic facts about this analysis for Cuyahoga County.

• Obtained from publicly available data. Don’t blindly believe me, verify it yourself.

• Registration roll obtained in Oct 2023.

• 876,206 registrants listed.

• 71% No Party Specified (Unaffiliated), 20% Democrat, 9% Republican, <1% Other.

• 641,754 have status of ACTIVE. The remaining 234,452 are CONFIRMATION.

The table above indicates that 36.7% of voters with no specified party have the status CONFIRMATION. Also, of the 234,453 voters for all parties who have the status CONFIRMATION, 99% of them have no party designation. This will be delved into more later.

According to the OH SoS the meaning of CONFIRMATION is:

It means that your county board of elections has mailed you a confirmation notice, but you have not yet responded to that notice to confirm or update your information.

Voters who have received a confirmation notice and fail to respond to that notice or participate in any voter activity (such as voting or updating their voter registration information) after a period of four years are subject to cancellation [my emphasis].

Note the specific term “are subject to cancellation”. It does not say “will be cancelled” after 4 years. I know of someone who has been deceased for more than 4 years in Cuyahoga County and they STILL have a status of CONFIRMATION.

Time Series Data - Peak Analysis

Let’s look at registration patterns over time starting in 2021.

During this period of time, OH was a member of ERIC up until June 16th, 2023. Do the huge spikes in registrations look like normal human behavior? In other words, do people decide on the same day en-masse to register to vote SPONTANEOUSLY? Obviously not, so what explains that spike patterns?

Look at the percentage distributions on the spike days. 88-92% are have no party affiliation. So am I to believe that people who do not want to specify their political party all choose to register on primarily three “spike days” in OH? This seems unlikely to me. You don’t need to be a data expert to understand that.

On Monday June 12th, 2023, the ERIC contract with the state of OH was 5 days from being cancelled. That date corresponds to Spike #3. Coincidence? A parting “gift” to OH voters from ERIC of 8,049 registrations in one day?

Let’s now focus on 2023 only.

Here you can see spike 3 again and additional spikes 4-6. Not only are the absolute number or registrations seemingly outrageous compared to the rest of the days in 2023, but the percent distributions are glaringly tilted to unaffiliated voters (not specified or null on the roll).

Finally, let’s look at those days (and only those days) in 2023 when the number of registrations exceeded 600.

Can you believe that the absolute number of registrations and the distribution between parties represents random human behavior? There have been 37 days so far in 2023 when more that 600 people decide to register to vote. Or is it registration harvesting? Motor voter data dumps? Or is it outright unrealistic data without any rationale explanation?

Registration Counts Per Unique Address

Here are the top 25 address in terms of the counts of registrants at that address.

Here is a compilation of the top 10 locations

Social Services - 1,121 registrants

Apartment Building - 532 (unit number is not included in the registration address)

Apartment Building - 465 (unit number is not included in the registration address)

Homeless Shelter - 446

Apartment Building - 417 (unit number is not included in the registration address)

Apartment Building - 359 (unit number is not included in the registration address)

Apartment Building - 357 (unit number is not included in the registration address)

Apartment Building - 346 (unit number is not included in the registration address)

Apartment Building - 336 (unit number is not included in the registration address)

Apartment Building - 333 (unit number is not included in the registration address)

The top 10 locations would represent 4,712 ballots that if mailed, have to be delivered without a specific unit number within the street address to the correct person.

Am I to believe there is 0% fraud occurring in such cases? 0% fraud is the ONLY acceptable standard.

And that all these ballots make their way into the rightful hands? None of them are picked up by others in the mail room and voted illegally? The chain of custody is 100% immutable? Am I to take all of that seriously? Should you?

Let’s be “generous” and assume that 5 people can legitimately use the exact same address. In Cuyahoga County there are 9,078 address that have 6 or MORE registered voters at the EXACT same address. This is 155,139 voters at these address with at least 6 registrations. Am I to believe that all of these people who are mailed ballots are legitimate voters? This group is ~ 18% of the entire voting public in Cuyahoga County!

Registrations on 1/1/1900

Why are there 23,876 people listed on the roll with a registration date of 1/1/1900 (not a typo)? 21,842 are ACTIVE voters (91%). So there are over 21,000 people in Cuyahoga County who are voting with a registration date of 1/1/1900. Does that inspire confidence?

Here are the top 20 most popular birthdays of people with a registration date of 1/1/1900. These folks for the most part were born in the 1950’s. How long does it take to update the database with a realistic registration dates?

What about the distribution by party of this group? Does it roughly ratio out the same way the overall county does? Nope.

The number Democrats in this group is roughly 2x the overall county ratio. Why do Democrats have twice the representation in this subset compared to the overall county distribution of party?

Here is the age distribution of this group all registered on 1/1/1900. Interesting that it does not include anyone under 46 years of age. The median age is ~ 70. Why?

Also, compare the above to the overall county distribution by age (below).

Clearly, this subset of people registered with 1/1/1900 don’t have the same age distribution pattern as the large county. Is that just random human behavior?

Precinct Voting

To those people who say voting in the precinct on election day, on paper and hand counting the votes on the ballots is not manageable, I would offer this chart. These are the registrations per precinct. Take these totals and let’s assume a 70% turnout. Looking at the graph, there would be ~ 300-500 voters on average per precinct. This seems manageable for hand counting in one of the more populated counties in the country, #35 to be exact.

EDA Hashing Function

In the EDA Excel workbook, there is a built in hashing function. This permits the users to specify which fields on the registration table to “HASH” and then produces a count of the HASH instances.

For example, I can take [LAST NAME]+[REG DATE]+[BIRTH DATE] and concatenate those together, then count the instances of that HASH. Would you expect that there are some number of people that share the same last name, are registered on the same day and have the same birth day in a county?

In the results of this HASH for Cuyahoga County stated above, there are 493 instances where the same last name, registration date and date of birth all coincide more than 1 time. In fact, 4 of those instances include 3 matches. The remainder (493-4=489) are examples of 2 instances where registrants share the same last name, birth date and registration date.

What does this mean, practically? What could be an explanation?

For example, triplets born on the same day, never marry and retain their name from birth, they all go together on the same day and register to vote on the same day.

Does that seem believable? It seems absurd to me!

Maybe there are other rationale explanations worth looking into.

Hash Example 1 - Digging Deeper

Using this HASH as a starting place, I looked more deeply into one of the instances of 3 Hash counts. I am not going to dox the individuals involved by name, but I will say that there were 6 instances of the same unique address being used by people with the same last name for this HASH result. 5 of them share the same registration date. 3 of those people had the same birth day and registration date and address. Two of them are Democrats, the other 4 are “Null” for party affiliation. Imagine that, all 6 LIVE here together, all 6 with the same last name, in this lovely bungalow (below) on the Cleveland’s East Side.

Hashing is a powerful analysis technique and a good starting point for a deep dive.

More importantly, how does the voter roll manage to become memorialized in this flawed state? Sloppy bookkeeping? Deliberate fraud? Both?

Hash Example 2

In Cuyahoga County, there are 4 instances where the HASH of [Last]+[BirthDate]+[Reg Date] yields 3 counts. Wow, what are the chances of that? Let’s delve into this as a group. I exported this HASH group based on last name and further grouped them.

Here are the registration date counts.

What do you notice? The magic ERIC date of 6/12/23 appears at the top. So the conclusion is that on the last database update from ERIC 5 people with the same last name and the same address were registered and 3 of those had the same birthday too. Thanks ERIC!

Delving further into the this subset, there are a total of 11 people who fit the criteria of having the same last name, reg date and birthdate. Within that group remarkably:

• A group of 3 have the same MIDDLE INITIAL in addition to the same last name, reg date and birth date and address.

Hash Example 3

Let’s use a different HASH of [ADDRESS]+[REG DATE]+[BIRTH DATE]. There are a total of 432 instances where the count is 2 or 3. There are exactly 4 instances of 3 counts. What this means is that three people are living together with the same last name, they registered to vote on the same day and they have the same birthday!

Looking closer at the 4 instances of a 3 count group.

• As a group of twelve registrants, the ages range from 19-31, so no older folks

• And most remarkably, at each address, they all have the SAME last name. Practically this means I could have made this a four part HASH of [LAST]+ADDRESS]+[REG DATE]+[BIRTH DATE].

• One instance of this HASH group was registered on that magical ERIC day of 6/12/23

• They are all ACTIVE.

• They are all Democrat or Unaffiliated.

Registrants with the Status CONFIRMATION

Let’s return to the topic of those registrants who have the status of CONFIRMATION. To hone in on these people, run an extract from the original EDA Excel file and put that subset of data in a separate EDA file.

Let’s review again the definition of what CONFIRMATION means.

It means that your county board of elections has mailed you a confirmation notice, but you have not yet responded to that notice to confirm or update your information.

Voters who have received a confirmation notice and fail to respond to that notice or participate in any voter activity (such as voting or updating their voter registration information) after a period of four years are subject to cancellation [my emphasis].

• There are 230 people with a registration date of 10/10/2023 that have the status CONFIRMATION. Does this make sense? People registered themselves (presumably) less than a month ago but are immediately labelled with this status.

• There are 3,519 people who registered to vote in 2023 (presumably) that now have the status CONFIRMATION. Why? They have no voting history yet.

• Remember the top address above for registrants at the Social Services location which had 1,121? Turns out, 816 of those have the status CONFIRMATION or 73%. It DOES makes sense that these centers create a lot of records in the database who apparently need confirmation. It also leads to an inflated registration roll. Is this done on purpose?

• The top 20 address in terms of registrants counts have 3,260 people with a status of CONFIRMATION.

• There are 2,205 unique address that contain 36,627 registrants with the status of CONFIRMATION that have…….wait for it…..6 or more registrants with that status.

• The top 3 zip codes that contain 27,846 (~10% with this status) registrants are in the following zip codes: 44102, 44109 and 44107. All of them are Cleveland west side counties.

• There are 280 people with this status who are greater than 100 years old. Perhaps some of these folks, if they exist, have the mental acuity to vote or perhaps not. Why are they on the roll and more importantly, how long have they been on the roll needed CONFIRMATION?

  

The main point of concern is keeping the rolls “clean”. As I said above, I know someone who died in 2014 and is still on the roll listed with the status CONFIRMATION. Clearly there is some clean up to do here.

June 12th, 2023 - A Special Day in Cuyahoga County

On June 12th, 2023, over 8,000 registrants were added during the same week that the ERIC contract was dropped as explained above. Let’s look at these 8,000+ registrants in more detail using the same technique as explained above. Extract that subset of data and run it through its own instance of EDA.

The overall county has a 75% ACTIVE ratio and is 71% No Party and 20% Democrat. This group above is effectively all No Party & Democrat and 99.6% ACTIVE. This seems to tilt the scale away from Republicans for this last great “dump” by ERIC.

Fore some reason, the younger population decided to register to vote on 6/12/23. Compare this to the normal county distribution above. What explains this behavior of young people?

Let’s not forget about folks over one hundred who also were registered by ERIC on 6/12/23.

Of these 9 centurions, I looked at the address of each of them. Remarkably, (to me at least) 6 address from this group are single family homes (not group homes or large apartments) and 4 of them have one and only one registrant. Perhaps these 4 stellar centurions don’t live alone (who knows) but they got up the gumption to register to vote, maybe when they renewed their drivers license? Watch out!

One remarkable home has a 90 and a 100 year old living together in a single family who registered on the same day, 6/12/23.

Now let’s delve into the largest age group that decided to register on 6/12/23…30 year olds.

• There is one address where 2-30 year olds registered on 6/12/23 who share the same last name, one is Deonte, the other is Derron and the best part is, not only are they both 30, they were born 9 days apart. If they are twins, that is one looooong labor cycle.

• There is one address where 2-30 year olds registered on 6/12/23 who share the same last name and whose first name differs by exactly ONE letter (Abneal versus Arneal) and they have the same birthday, exactly the same date. Wow!

Conclusion

Using the EDA spreadsheet which is distributed free of charge on Cause of America, you can relatively easily run these types of analysis on small data sets. You do not need to be a database administrator.

I am a fan of the micro trends analysis approach, especially when focused on Spikes. I call this informally Spike Analysis.

Give it a try for your county.