Are Elderly Votes Being Harvested in VA?
EPEC Team's Senior Data Analyst asks: Why Are So Many 'New' Registrations being created For voters In convalescence?
Senior Analyst Rick Naigle’s recent analysis of Virginia’s voter-registration rolls shows odd spikes in first-time registrations for elderly voters aged 77 and up — and more odd items.
Virginia Registered Voter List (RVL) as of April 2025 shows 6.2 million registrations. Assuming each is a unique voter (individuals must be age 16 or older in order to register to vote), Virginia should know that it has at least 6.2 million living residents between age 16 and age 116.
If age at the time of registration is random, we would expect maybe a hundred or so voters to register after turning 85 years of age.
Yet we have found 9,913 voters in the Registered Voter List who registered to vote only after turning 85 years of age (6,578 females, 3,266 males, 68 undefined, and 1 gender X).
When we adjust to include first-time registrations aged 77 and up, the number grows.
Our analysis also found odd patterns of voter-registration spikes in five-year intervals, in addition to a few hundred voters who appear to have been registered prior to the age of 16. And we counted one hearty soul who was registered for the first time at the age of 115. But here we focus on the spike in first-time registrations of elderly voters.
Why are there so many?
Below are some slides of the analysis, and other observations.
PDF version of EPEC’s full analysis can be downloaded here.
More Analysis:
The chart below shows four voters registered to vote in 1902 (number not visible on the graph). Anyone who registered to vote in 1902 would have had to be 21 years of age in 1902, meaning they would have had to have been born no later than 1881.
If that were the case, we would have four voters age roughly 143 or 144 years of age on our Registered Voter List (2025 - 1881 = 144).
We presume we are dealing with data quality issues, but we cannot rule out manipulation of elderly voters.
Analysis Methodology
We computed the age of voters at the time of registration by subtracting Year of Birth from Registration Year.
Computed voter ages at the time of registration ranged from -82 years of age to 115 years of age. The median age (50% below, 50% above) was 31 years of age.
Average age was computed to be 33.5 years of age. Age standard deviation was computed to be 14.4 years. From this we can compute an “approximate” Upper Control Limit = Average + 3x Standard Deviation = 76.7 years of age.
So, we have Minimum (-82), Median (31), Average (or Mean) (33.5), StdDev (14.4), and UCL (76.7).
What is the UCL (Upper Control Limit) and why is it of interest?
Statistically speaking, we are interested in variation which is more than 3 standard deviations from the mean. Any registrations where the voter was older than 76.7 years of age might be considered “special cause” variations. We computed the UCL as an approximation.
We did not verify the actual distribution of the Age When Registered computations, but this is close enough for us to identify registrations that may end up being challenged, if not audited for accuracy.
We added some “slack” to the UCL. We used Age 85 as the Upper Control Limit for special cause auditing — it should be “close enough.”
The chart below shows the registration dates for voters who registered to vote when they were 85 years of age or older. We observe 9,913 voter registrations where the voter was 85 years of age or older. We can see a pattern in the data.
The maximum of registrants 85 years of age or older occurred on 15 October 2024, when 126 senior citizens were registered to vote. That was probably the last day for in-person registration for the 2024 General Election.
It looks like there is a spike associated with the last day of registration for most General Elections since 2012.
“Special cause” might be the last day to register. It is worth noting that there was also a registration spike of elderly voters on 5 November 2024 (Election Day).
This is new, and probably occurred because Same Day Registration was allowed during the 2024 General Election.
We must still ask the following questions:
Why are there so many elderly voters age 85 or older registering to vote for the first time?
Are these voters in convalescent centers?
Are we seeing elderly voters having their ballots being harvested, possibly without knowledge or consent?
Are elderly voters submitting their ballots absentee via third party harvesters?
Are these ballots being completed by someone other than the voter?
We do not know the answer to these questions. Yet the questions remain after observing many voter registrations more than three standard deviations from the mean.
We would expect perhaps a hundred, but not nearly 10,000.
These registrations should be audited by General Registrars or the Department of Elections.
The following Localities had 150 or more voters on the RVL who registered to vote when they were Age 85 or older:
Out of 133 Localities in Virginia, there were 132 Localities with voters registering at age 85 or older. The list of all Localities, in alphabetic order, follows:
We would expect to see a few voters registering to vote on or after their 85th birthday, but the number should be no more than a few hundred.
These could be valid registrations arising from downsizing and family relocation decisions, but the magnitude of late-life registrations warrants an audit/explanation.
We also question the process involved with merging duplicate registrations, and whether some of these files could be duplicates that are being merged without maintaining the voters’ original date of registration.
That may also explain why so many elderly first-time voters are showing up in the voter lists.
EPEC Team is standing by to provide more information necessary (such as RVL IDs) to General Registrars and/or the Department of Elections if they feel more audits are warranted. ##
Rick Naigle is Senior Data Analyst for EPEC Team and a member of its Board of Directors.
More about this article.
The article implies all of these elderly voters are in assisted living facilities. This is not the case. We observed "clusters" of first time registrations for the elderly at locations correlating to elder facilities and to elder living communities. This might be explained by attention being given to making sure new registrants are registered to vote, BUT ... BUT ... if the voter is already registered elsewhere in Virginia, the voter registration address should be updated - a new registration record should not be created. We cannot tell, with certainty, whether the Registration Date has been updated on an original registration from prior years. This is a possibility, but best practices would record the update date in a separate field, not in the "created date" field, which essentially is what the Registration Date field SHOULD be capturing.
We received some comments and questions from an election official addressing Henrico County concerns.
This is in response to questions received from an election official.
We looked at Registered Voter List data - the "voice" of the process - and observed something "statistically significant," meaning unexpected results attributable to special cause, not random variation. Some action, or actions, is causing far more elderly "first time registrations" than is statistically expected based upon population data (Commonwealth wide).
We did not see a "spike" per se in first time registrations, we saw a pattern which, based upon the underlying probability distribution of the data, suggests "special cause" - something is happening to create more elderly "first time" registrations than is statistically expected, and the number far exceeds random probability.
We observed a "statistically significant" anomaly. This means the variation is not random - the source of the variation is Special Cause. It MIGHT be elderly moving from out of state to be in the vicinity of top-notch medical facilities. It MIGHT be voter registration drives which result in new registrations being created when existing registrations exist. Perhaps some localities are checking for existing voter registrations while other routinely create new registrations. If they are new registrations being created without checking for existing registrations, we'll end up creating a group of voters with multiple voter IDs - all "active," and these "errors" create a ballot harvesting or duplicate vote risk.
Locality is reported merely for identifying the GR responsible for checking / auditing each statistical anomaly. Defects are measured in EACHES, not averages. Each of the identified "over age 76" IDs is its own outlier, statewide. There are pockets of outliers at specific addresses. We identified these "groupings" so root cause analysis can be performed by respective General Registrars or local party officials. These are non-partisan errors. Aside from creating a potential ballot harvesting risk, the next greatest risk is the loss of voter history ... when a new ID is assigned, the old voter history may end up getting deleted if the old ID is removed.
The probability distribution for Registration Age is the Weibull distribution, shifted to the right (from 0) with the start being Age 16 - the earliest age someone can register to vote. The MEAN for the distribution is in the mid-thirties, and three standard deviations establishes and upper control limit (a 3 Sigma limit) around age 76. We ASSUME the Registration Date records the date the Voter ID # is assigned to the individual. This MAY be a bad assumption ... some localities MIGHT be UPDATING the Registration Date when an existing voter record is updated. We THINK these are all NEW registrations, not existing registrations which have had the address updated to reflect the voter's current location.
Our upper control limit (sigma limit) is based upon the statewide population of registered votes. It is POSSIBLE that certain localities may have a greater number / higher concentration of elderly, in part due to the availability of top notch medical facilities. Again, it is the the CURRENT voter AGE, that caught our eye, it is the REGISTRATION AGE. Should these be NEW registrations, or should they be CHANGE OF ADDRESS registrations where the voter's ID does not change.
ELECTION OFFICIAL OBSERVATIONS / QUESTIONS follow:
ELECTION OFFICIAL
"Two observations and a couple of questions after quickly reviewing this.
Observations:
When we see registered voters after 85, Henrico lists 423. It appears somewhat consistent with the population (related to other counties). In other words, I don't think Henrico is out of alignment with Chesterfield or Goochland (51). Henrico is listed at 423 out of ~330,000 and Goochland with 51 out of ~36,000 (total population) and is relatively consistent around .1%. I realize that does not prove whether a problem exists or not; just that Henrico appears to not be an outlier."
OUR RESPONSE: The Locality is not the outlier. Defects are measured in EACHES, not averages. The records belong to Localities for records management, so we grouped the outliers by identify the General Registrar who manages the registration record.
Each of the identified "over age 76" IDs is its own outlier, statewide. There are pockets of outliers at specific addresses. We identified these "groupings" so root cause analysis can be performed by respective General Registrars or local party officials. These are non-partisan errors. Aside from creating a potential ballot harvesting risk, the next greatest risk is the loss of voter history ... when a new ID is assigned, the old voter history may end up getting deleted if the old ID is removed.
ELECTION OFFICIAL:
"Secondly, the bold statement below
> If age at the time of registration is random, we would expect maybe a hundred or so voters to register after turning 85 years of age.>
just seems to be off. I have no empirical data to prove otherwise, but I personally know three people that moved family members to the area (I believe they were all over 85). For the entire state to only have a "100 or so" seems to be a leap of faith. Again, I don't know how to dispute this; maybe we can ask one of the assisted living facilities to provide some basic data, but especially around the bordering states (Southwest VA, Northern VA/DC) I would think there are more than "100 or so" people moving into the state over 85."
OUR RESPONSE:
Ideally there should be no outliers beyond six Sigma limits, as this represents 99.99966% of the data. Any data points outside this range are considered unusual and may indicate special causes of variation that need investigation. Random variation is still possible outside of the sigma limits, so we would expect some ... a relatively few number ... of values to exceed the upper sigma limit. If it is roughly 3 defects per million. we'll say 10, and multiply that by 6.5M, so we'd expect no more than about 85 ... this is BALLPARK, not exact ... to exceed 76 ... but we have more than 40K outliers. Again, we won't split hairs on the EXACT upper control limit ... but we really should not have more than 100 above age 85. This statement is believed to be a statistically correct statement. Even if the number is 1K, we still have more than 40K which constitute unexplained special cause variation.
ELECTION OFFICIAL:
"Questions:
My question is this: When this analysis is done below, are we talking about newly registered voters (never voted in Virginia before) or are we also counting individuals that moved into a new county (Chesterfield to Henrico for example), moved precincts (Tuckahoe to Lakewood) or changed address (River Road to Parham road but remained in the same precinct)? In other words, does this study pick up when someone moves within the county to a place like Cederfield or Lakewood, for example?"
OUR RESPONSE:
These are excellent questions. We believe the Registration Date means a new registration was created. It is POSSIBLE that some Localities are updating the registration date when a change of address is made. We observed "voice of the process." We do not know the WHY, but we know a check for special cause is warranted.
ELECTION OFFICIAL:
"Can you confirm if these are new voters to the state, new voters to the county, new voters to a precinct and/or just changes in address within an existing precinct?"
OUR RESPONSE:
Additional analysis is required for us to determine whether these are moves, or new registration. We BELIEVE many of these SHOULD be change of address actions, where the Registration Date should remain unchanged. Virginia law prevents us from performing special cause analysis by looking at the source of the registration. GRs should look at the registration records. Do these new registrations correlate to specific registration methods - i.e., are they voter originated via the web, originated by a voter registration drive, or originating from DMV due to drivers license changes? We don't know. ELECT and GRs might want to perform their own follow-up, or ask EPEC to perform additional analysis using ELECT or GR-provided additional data, which currently is not available to us for root cause analysis.
The concerns expressed by the Election Official are valid, and we thank the elected official for requesting additional information and for providing feedback.
We used Grok to verify the upper control limit (mean + 3 standard deviations). Voters who were 76 years of age, or older, exceeded the upper Sigma limit, therefore there are 51,672 elderly voters whose registrations are questionable. This does NOT mean fraud or anything nefarious has happened. It means the number is much larger than expected, and there is "special cause." It might be local emphasis on ensuring elderly residents are registered to vote, in combination with residence addresses NOT being updated if the voter has an existing registration.
There is a risk that many of these voters will be cognitively impaired, and the ballots of cognitively impaired voters may end up being "voted" by a "third party" fraudulently. No proof, just a risk. Some data sources estimate cognitive impairment to be as high as 30% for those over age 75 in Virginia.