Is Oklahoma's voter database suspicious?
A couple of friends have posted links to an article alleging shenanigans in Oklahoma's voter registration records. The article is on a website known for sensationalistic headlines, but that article linked to an analysis on another site more temperate in tone, but missing important context.
The bottom line: The voter ID number patterns which the analyst found suspicious in the short time he spent with the database have explanations rooted in the history of Oklahoma's voter registration system, specifically the transition to our first statewide registration computer system in 1990 and its replacement with a newer system in 2011.
The headline on The Gateway Pundit screams "Jerome Corsi: Oklahoma Added to the List of States with Irregularities in Board of Election Voter Registration Databases Suspected of Fraud." Corsi gets off to a misleading start:
In a highly suspicious August 27 run-off mayoral election in Tulsa, two relatively inexperienced political operatives with pedigree-quality, radically woke Democratic Party credentials beat a conservative Republican CPA, attorney, businessman, and pastor with a long history of community service, in Oklahoma, by the narrowest possible margins.
It wasn't a run-off -- it was the general election, with a run-off to come in November. Karen Keith, a 16-year county commissioner and 30+ year TV news reporter and anchor endorsed by the police and firefighter unions, and Monroe Nichols, an 8-year state representative endorsed by a former mayor, two former governors, and the daily paper, are hardly inexperienced. Brent VanNorman's long history of community involvement happened in other cities and states. Corsi doesn't appear to know that this election is officially non-partisan, with no party labels on the ballot. VanNorman was one of three registered Republicans running for mayor, but the only information about the party affiliations of the candidates was on websites like this one, not on the ballot. (I wonder who fed Corsi the above characterization of the Tulsa election.)
More about the dynamics of the mayoral race below. It was impressive that VanNorman did as well as he did with so little money, no name recognition, and a standing start with only three months to introduce himself to the voters. I'm not shocked that he failed to make the runoff; I'm amazed he came so close to making the runoff and beating the one-time front-runner, and I'm frustrated with the Republicans who withheld support entirely or until it was too late to matter (Tulsa County GOP, Kevin Stitt).
Corsi quoted a New York State-based analyst named Andrew Paquette, who has been acquiring voter databases from election boards and analyzing them, focusing on patterns in voter identification numbers as a potential indicator of fraud. Corsi goes on:
Paquette has charged that State Board of Elections official voter registration databases may contain cryptographic codes of intelligence agency complexity that enable rogue actors to obtain official state voter ID numbers for non-existent fraudulently created voters in an apparently criminal scheme designed to facilitate the certification of fraudulent mail-in votes.
(I have detected irregularities in Corsi's spelling of Paquette, which he spells "Pacquette" about as often as he spells it correctly.)
The article embeds Paquette's report but doesn't link to it, which is rather dodgy. Here is Paquette's report. Paquette's report is provided on his site as non-OCRed images, which is also rather strange. Here is the beginning of his Oklahoma analysis, which is much more circumspect than Corsi's lurid prose:
I have spent literally one day looking at Oklahoma's voter rolls. Much less if you subtract the time it took to import each county's database into a master database for study. In comparison, it took weeks before the first hints of voter roll algorithms were found in Ohio and New Jersey, and even longer in New York. With the caveat that this is not enough time to yield a definitive response either way, here are a few preliminary observations:
In this entry I'm going to focus on Paquette's suspicions about Oklahoma voter ID numbers and registration dates. He created scatterplots of date of registration on the X axis and was surprised to see that the voter ID numbers don't correlate in any obvious way to registration dates before 1990:
What this plot tells us is that ID numbers generally ascend as registration dates become more recent. There is a large break in CID numbers between about 720,000,000 through 800,000,000 that occurs in 2012. This break is found in all other OK counties. The reason for this is unclear, particularly for a state with a population size that is unlikely to ever exceed the available unused numbers. A close-up of numbers on the left of the plot reveals another break in the numbers in the year 1990, after which they ascend normally. Earlier numbers do not follow a normal ascending pattern, but are found in any year from 1950-1990, regardless of number size. That is, a high number from the series is just as likely to be from 1950 as 1990, but later numbers always ascend with the year. This is different from some counties and bears further investigation (Figure 2).
Had Paquette had more time with Oklahoma's numbers (why the big hurry?) he might have noticed that a large share of voter IDs from a given county begin with the same two digits and those two digits match the sequence of the county name in alphabetical order. Adair is 01, Woodward is 77. Tulsa and Wagoner are 72 and 73, respectively. Osage is 57 and Rogers is 66. Historically, back in the days of handwritten indexes, people sorted names beginning with Mc before all other names beginning with M, because a Mc name (a Celtic patronymic) is sometimes spelled as Mac, so McClain, McCurtain, and McIntosh are 44, 45, and 46, and Major is 47. Precinct numbers are six digits beginning with the two-digit county code. (FIPS county codes follow the same order, but are all odd numbers separated by two, which I suppose allows for a new county to be added in alphabetical order without renumbering the rest.)
So a large share of Tulsa County voter IDs, 155,841 out of 390,753 in the August 8, 2024, download, begin with 72. These were all issued on or before April 18, 2011. The rest of the voter IDs begin with 80, reflecting a move to a new statewide election computer system in 2011. Existing voters kept their ID numbers beginning with a county code, but new voters were registered with numbers beginning with 80, a value that would not conflict with any existing voter ID numbers, because Oklahoma has only 77 counties.
(Here is a January 2011 story announcing selection of a vendor for the new system, an op-ed by Oklahoma State Election Board Paul Ziriax, announcing the new system from Hart InterCivic, which included new ballot scanning machines to replace those that were nearly twenty years old, a July 2011 story mentioning the model names of the old (OPTECH III-P Eagle) and new scanners (Hart InterCivic eScan A/T Paper Based Digital Ballot Scanner), December 2011 article showing the difference in ballot styles between old and new machines.)
So why do voter ID numbers after 1990 increase monotonically with registration date, but are seemingly random until 1990? Because the Oklahoma State Election Board got its first statewide computer system that year. In Tulsa County there is a break in registrations between June 15, 1990, and June 30, 1990. If you sort the Tulsa County voter file by voter ID number and filter for registrations prior to June 15, 1990, you'll find that the names are mostly in alphabetical order. The exceptions to alphabetical order are mainly women; women are more likely to change their last name after they get married, but they keep their voter ID number. For example, near the end of those pre-1990 records, I found someone I know with the last name Werner whose voter ID number falls in sequence with people named Zumwalt, which was her married name in June 1990.
My wife and I are four numbers apart, even though I registered to vote in 1981 after I turned 18, and she registered to vote in Oklahoma eight years later, after we were married in 1989. Even though our first names are close together in alphabetical order, there were once six Michael Bateses registered to vote in Tulsa County, four with different middle names, and one with the same middle name and a date of birth six months earlier in the same year. There's still one other Michael Bates -- Michael S. Bates, the retired City of Tulsa human resources director, who was also registered to vote before the 1990 computer system went online -- his voter ID is after mine and before my wife's.
What is likely is that Tulsa County Election Board took its existing alphabetized voter database and entered them into the new system in that order, from A'Neal to Zyskowski. In the current database, those numbers range from 720000002 to 720300782, and the registration dates range from June 23, 1942, to June 15, 1990. I found only three exceptions in Tulsa County, three voters with ID numbers in that range who registered in October 1990, February 1991, and October 1996. Tulsa County's population in the 1990 census was 505,289; 59.5% seems a reasonable ratio of registered voters to population. Today there are 390,753 voters, and the 2020 population was 670,653 -- 58.3%.
There are only 51,323 voter records in that range of voter ID numbers today. A lot of people die or move away in 34 years. Whoever had 720000001 must fall in one of those categories. In 2016, 77,007 voters had ID numbers in that range.
(Here is an August 1990 Associated Press story about the state's then-new election administration computer system, running on Digital Equipment Corporation computers with software by Andersen Consulting, a branch of the Arthur Andersen accounting firm, a May 1990 Okmulgee Daily Times story about election board worker training and reporting that the new system will go into use on July 1, and a July 1990 Oklahoma Press Association story on the new system.)
There are 226 records in the Tulsa County database with no recorded registration date. 189 of these have their addresses redacted with asterisks; among this group I recognize the names of district judges, assistant district attorneys, and others who might be at risk from stalkers. I'm not sure how blanking registration dates helps with security for these people, but there is a correlation. The law authorizing address confidentiality is here; the rules adopted by the State Election Board are here. (30 with redacted addresses have valid registration dates in 2023 or 2024, perhaps reflecting a later tweak to the law or election board procedure.)
Of the remaining 37 with no registration date, they were all born in 1962 or earlier, and all but one are in that initial group of voters entered into the computerized system in 1990. The date may not have been entered on the original record, or perhaps was illegible when the records were entered in 1990.
There are 9 records in the Tulsa County database where the registration date is earlier than the birth date. All 9 are part of the records that were input into the system in 1990. It's reasonable to guess that these were data entry errors when transcribing from cards to computer.
So, excluding the 219 redacted records, that's 46 voters out of 390,534 with blank or impossible voter registration dates, only 12 thousandths of one percent -- 0.012%. Ideally there wouldn't be any, but the county election board would have been wrong, as they entered the existing registration rolls into the computer system in 1990, to drop a duly registered voter for lack of a legible voter registration date.
Not all counties tracked voter registration dates prior to the 1990 computer system. The Osage County roll has only 48 voters with registration dates on or before June 15, 1990. 3,168 registration dates are blank out of 29,442. Wagoner County has only 44 voters with registration dates that pre-date the 1990 computer system, and 3,802 blank registration dates out of 51,987. But Rogers County apparently tracked registration dates before the 1990 computer system: It has only 805 blank registration dates out of 64,776 voter records; 23 of that 805 belong to voters with redacted addresses.
In summary, there's nothing weird about Oklahoma voter ID numbers or registration dates that isn't explained by the history of Oklahoma's computerized election management system. There are a small number of oddities that reflect a reasonable number of clerical errors.
Another concern raised by Paquette is that purged records are not retained in the database. The definition of a purged record is that it has been removed from the database. The State Election Board does, however, provide a separate file containing all records deleted statewide in the last 24 months. I am surprised that Paquette overlooked this data source. The file I downloaded on September 30, 2024, contains 244,388 records, of which 90,381 were removed as a county transfer (removed from the database of the voter's former county), 83,701 were deleted for inactivity, 51,647 as deaths (health department, next of kin, nursing/funeral home, written notice), 7,965 were deleted as duplicates, 3,128 for felony convictions, 2,572 are labeled CONF. NOTICE - STATE or CONF. NOTICE - COUNTY (possibly indicating that a confirmation postcard to the address on record was returned with a change of address out of county or out of state), 2,518 as a state transfer, 2,254 license surrender, 222 mental incapacity. The file includes the date of deletion, ranging from October 3, 2022, to September 29, 2024.
Paquette also complains about possible clones -- records with the same date of birth and first and last name. I haven't checked this yet; that would require looking at the entire state at once, and that involves loading all 77 separate county files into one database. I'll take a look at that at a later date.
On the jump page, I've got more detail regarding the dynamics of Tulsa's mayoral race and why Jerome Corsi's description doesn't fit the facts, on the contents of Oklahoma's voter registration database, and on why I don't trust any headline on The Gateway Pundit.
The political dynamics of the 2024 Tulsa mayoral election:
While his opponents had been running for over a year, Brent VanNorman got into the race only three months before election day. VanNorman has an impressive resume but has only lived in Tulsa for three years, and most Republican elected officials withheld an endorsement. Governor Kevin Stitt endorsed him the night before the election, too late to matter. The Tulsa County GOP endorsement came a mere six days before the election. Congressman Kevin Hern and former Attorney General John O'Connor were among the few prominent Republicans to make a timely endorsement.
As of the pre-election contributions and expenditures reports, Keith had raised $729,911.28, Nichols had raised $464,014.06. VanNorman did not submit a pre-election contribution and expenditure report. His 2nd quarter report showed that he had raised only $92,555 to that point. Informally, I've heard that he raised about $200,000. City Councilor Jayme Fowler, who dropped out of the mayoral race after a year, raised $273,720.
Because of her years in front of a camera and positive public image, Karen Keith was considered a prohibitive favorite, and some potential Republican candidates opted not to make the race because of it. Public polls had her above 50% of the vote, and she won 53.4% early voters by mail with 1,638 of 3,066 cast. Her support began to drop in response to the scandal at the Tulsa County Family Center for Juvenile Justice, which she and her fellow commissioners knew about long before it became public. A poor performance at the NonDoc/KJRH debate drove her numbers further down and elevated VanNorman. VanNorman finished second in ballots cast on election day, but not by enough to overcome the absentee vote lead Keith had already locked in before public opinion began to turn against her.
FYI for out-of-staters who may read this: Oklahoma absentee ballots must arrive to the Tulsa County Election Board by the time the polls close on election day. A timely postmark is not enough. Like Florida, Oklahoma doesn't allow boxes of ballots to "show up" after election day. Mail-in absentee ballots must be notarized, and the law limits the number of ballots any notary can witness. Mail-in absentee ballots and early voting ballots must be counted and reported before the county election board can begin processing and reporting results from the machines from the precincts; meanwhile the result printouts from those machines have already been posted on the door of each precinct. It prevents corrupt election officials from knowing how many absentee ballots they need to "find" to change a result, and it ensures that we have a full unofficial total on election night, typically within three hours of the polls closing.
The hand recount, conducted under the oversight of the Tulsa County Election Board secretary (a Republican appointee who had previously been a conservative local talk radio host), and election board members appointed by the Republican and Democrat county chairmen, revealed that 38 ballots accepted by the ballot scanners had little stray marks in one or more other boxes of mayoral candidates; by state law, these had to be interpreted as overvotes and not counted for any candidate. Monroe Nichols lost 11 votes in the hand count, Karen Keith lost 13, and Brent VanNorman lost 14. A trusted observer who was present told me that he got as close to the ballots as he could without touching them and saw that all were of the same size and thickness and all had the expected perforation marks on the edge where they are torn out of the bound ballot pad by precinct workers. In other words, there was no indication that someone had slipped in fraudulent ballots.
State election board records
Oklahoma voter registration and voter history records are public records. Candidates and political parties use them to figure out which doors they should knock on and which mailboxes to target with their glossy oversized postcards.
State election board records don't contain phone numbers, but there are companies that will take the state election board data and find the best match in the phone company's database, based on name and address. Political parties and campaigns will also use companies that match your name and address to marketing data about you, all in an effort to find the people most likely to vote for the candidate, to find the best way to persuade you to their side, and the best way to motivate you to turn out on election day.
But the state election board has only your name, residential address (so they can determine which precinct you live in and the districts in which you are eligible to vote), mailing address if it differs (so they can send you a voter ID card and communicate with you), date of birth, date of registration, party of registration (REP, DEM, LIB, or IND), and a list of elections in which you have cast a ballot.
In the olden days, around 1998 into the early 2000s, getting the latest statewide database involved sending a cashier's check for $150 and then getting a box the size of a small microwave oven containing six open-reel computer tapes. You then had to have a way to convert those tapes to files on a computer. There was a company called Metro Data that would put them on a CD for a fee. Eventually, the state election board figured out how to put the data on a CD themselves. Sometime in the mid-2010s, they set up a process involving downloads.
Why I don't trust The Gateway Pundit
The Gateway Pundit is renowned for sensationalistic click-bait headlines that aren't backed up by the news stories linked in the blog entry. Andrew Paquette was quite measured in his analysis, stating where his knowledge is incomplete, but none of that circumspection made it into Jerome Corsi's Gateway Pundit summary or the headline.
I saw that firsthand in 2011, at Americans for Prosperity's RightOnline conference in Minneapolis. It was an exciting event as we heard from Republican candidates who were preparing to launch their campaigns to win the right to challenge Barack Obama's re-election bid. One of those candidates, Herman Cain, held a special press conference for a group of bloggers, and I was there. Cain raised his voice slightly when a leftist blogger raised a question that Cain had already addressed at the outset. NiceDeb and Stacy McCain have contemporaneous accounts of the press conference and confrontation that line up with my recollection, and NiceDeb also has video.
Jim Hoft, The Gateway Pundit's proprietor and, at that time, sole blogger, was not at RightOnline, but he nevertheless falsely characterized Herman Cain at the blogger press conference as "screaming." (Note that the commenters at the time didn't buy Hoft's characterization.) I haven't trusted Gateway Pundit ever since, and I have been even more put off by Hoft's ingratitude toward Dana Loesch, who with her husband Chris helped to nurse him back to health after a major health scare in 2013; Hoft trashed her over their differences of opinion in the 2016 Republican primary race. Hoft has a Filipino "husband" who is 30 years younger.
I generally resist the urge to click when someone shares a Gateway Pundit headline on Facebook or Twitter, but when I do, it becomes quickly apparent that the headline was written as click-bait, not as an accurate summary of the original news story that The Gateway Pundit article quotes.
0 TrackBacks
Listed below are links to blogs that reference this entry: Is Oklahoma's voter database suspicious?.
TrackBack URL for this entry: https://www.batesline.com/cgi-bin/mt/mt-tb.cgi/9247
Subscribe to this blog's feed: