As TYT previously reported, the results of several recent high-profile Kansas primaries last month were delayed for many hours due to a computer “glitch” in Johnson County, the state’s most populous county.
Before the glitch, sitting Governor Jeff Colyer reportedly led Secretary of State Kris Kobach by 44 votes in the Republican gubernatorial primary, and Brent Welder led Sharice Davids by seven points in the Democratic primary for Kansas’s third House district seat. After the glitch, Kobach and Davids took the lead in their respective races. Colyer and Welder both conceded without requesting recounts.
The debacle caught the media’s attention for many reasons, including that Johnson County was using a new voting system in the election: ExpressVote touchscreen ballot markers from Election Systems & Software, LLC. The software was certified by the federal Election Assistance Commission (EAC) on July 2, 2018. The Kansas City Star called it, “the first time in the country that voting machines with this kind of particular configuration were used.”
The EAC’s certification letter, which TYT obtained from the county, was signed by Brian Newby, a former Johnson County election commissioner who reportedly got his current position on the EAC in 2015 after Kobach recommended him for the job. (Before leaving Johnson County, Newby made headlines for significant personal and financial transgressions.)
It was current Johnson County Election Commissioner Ronnie Metsker, a Kobach appointee, who recommended that the county buy this particular system.
On August 27, a county press release said that last month’s reporting delay was caused by a problem with the “backend reporting code,” which ES&S has now “rewritten” and submitted for federal certification. The release assured voters that “ES&S is testing an alternate reporting process” in case the re-written code is not certified in time for the November election. The press release further advised that the problematic “reporting code” was “separate and distinct from the vote tabulation software” and “new ExpressVote machine[s].” According to the release, the actual “votes were tabulated accurately, and results tapes were printed at the poll sites…”
In a joint phone interview, TYT spoke with Metsker and two ES&S representatives regarding the process around last month’s elections. Asked about the county’s basis for saying that “votes were tabulated accurately,” neither ES&S nor Metsker could offer an explanation of how they knew that to be true.
Cohn: [W]hat was done to make that determination that the votes themselves were tabulated accurately?
ES&S Vice President of Software Development Gary Weber: Yes, the casting of votes and the tabulation of votes actually takes place on the voting device, in this case the ExpressVote voting device. The non-performing software was on the back-end selection results, aggregation and results reporting system, a very distinct part of the system.
Cohn: …[B]ut how did you determine that the votes—before they ever got to that problematic software—that the votes were tabulated accurately by the ExpressVote?
ES&S Senior Vice President of Government Affairs Kathy Rogers: I guess I don’t understand your question, Jennifer. It’s a paper-based system, the system certified by the Election Assistance Commission. The tabulation of the results were never in question, it was merely a slow reporting issue, of the election results that night; so there hasn’t been, and still isn’t, any concern over the actual accuracy or the sanctity of the votes. Those are secure.
TYT asked Metsker whether the county had compared the results tapes referenced in the press release to the results that were publicly reported to ensure that the results weren’t altered after they left the precincts (a reconciliation conducted in some states). Metsker said he was not sure he understood the question.
Cohn: Okay, so the results tapes…am I correct that they provide the vote totals from each individual ExpressVote machine?
Metsker: That’s correct…
Cohn: Okay, and it is printed election night at the polling places?
Metsker: That’s correct…
Cohn: …[T]he reason why I’m asking is…I have concerns…that the reported total after the uploading delay actually matched what came out of the machine, and so I’m wondering, did anybody make that comparison between the results tapes and the reported total, whether unofficial or official?
Metsker: I don’t even know for sure what you’re asking.
In an email to TYT, Jody Hanson, Johnson County’s director of public affairs and communication, said the county also did not use the results tapes during the post-election canvass.
“The results tapes were not brought to the canvass because the official vote totals come from data derived from the master USB sticks from each polling location,” Hanson wrote. “Results tapes are typically checked in and accounted for internally by Election Office staff. Some states may derive their official totals from the results tapes. That’s not the case in Johnson County, Kansas.”
In response to a public records request by TYT, the county advised that it would cost $3,487.00 plus mailing costs to obtain a copy of these results tapes. TYT requested a PDF of the reported election-day totals shown for each precinct, known as the “Statement of Votes Cast.” Although a comparison might detect irregularities that occurred after the totals left the precincts, it would not establish whether the votes themselves were tabulated correctly in the first place. The poll books, copies of which the county says can be exported in Excel, also would not establish this.
According to University of Michigan Computer Science Professor Alex Halderman, who studies election technology, the only way to know if “votes were tabulated correctly” is to hand count the paper ballots, or other voter-verifiable paper trail, in an audit or “recount,” and compare the result against the electronic tally. Forensic audits are typically blocked by courts that side with counties and vendors arguing that their voting software is proprietary.
Professor of Statistics Philip B. Stark of U.C. Berkeley, who invented post-election Risk Limiting Audits (which are endorsed by the League of Women Voters, Common Cause, Verified Voting, and the American Statistical Association), says that, “hand-marked paper is more usable by voters without disabilities, not only for recording intent but also for verifying that the paper ballot correctly portrays that intent. We cannot force voters to check that the paper ballot accurately reﬂects their preferences.”
Johnson County’s ExpressVote machines do not use hand-marked paper ballots. Instead, they generate computer-marked “summary cards” containing visually unverifiable bar codes. These bar codes, which can only be read by computer, are the only portion of the summary card counted as your vote.
Metsker stressed that the summary cards, which he calls “paper ballots,” also include human-readable text, which voters are supposed to be told to review to verify that the text accurately reflects their intended selections:
Metsker said, “What I call it is the paper ballot that is marked by the machine, that the voter is able to hold in his hand and review to make sure what is marked on that ballot reflects the choices that he made as he reviews the summary screen before he presses the button to say print ballot.”
Voters, however, reported confusion and conflicting instructions at the polls as to whether they were supposed to verify the summary cards at all. The county’s election worker and supervising judge training manuals include no such instructions.
Metsker told TYT that, as far as he knows, the ExpressVote machines themselves do not instruct voters to verify the summary cards.
Cohn: … [D]oes the screen itself tell voters that they should verify the paper summary cards?
Metsker: I don’t believe so…
Metsker acknowledged that the ExpressVote machines purchased by the county give voters the option to deposit the summary cards directly into the bin without verifying them at all.
“The version that we have has an option for both ways,” Metsker said. “We instruct the voters to print their ballots so that they can review their paper ballots, but they’re not required to do so. If they want to press the button ‘cast ballot,’ it will cast the ballot, but if they do so they are doing so with full knowledge that they will not see their ballot card, it will instead be cast, scanned, tabulated and dropped in the secure ballot container at the backside of the machine.”
Stark said the county’s decision to buy ExpressVote machines with this direct-deposit option means that a malicious hack could alter some votes without the possibility of detection in a manual audit or recount.
“Because the ExpressVote as configured in Johnson County prints the summary card only after the voter chooses between the direct-deposit button (bypassing the opportunity to verify the summary card) and the return-card-for-verification button, a bug or malicious hack could alter votes only when the direct-deposit feature has been selected, that is, when the voter cannot check whether the paper accurately reflects his or her preferences,” Stark said.
“A manual audit or recount would not be able to detect that, because the votes on the paper ballots would match the electronic tally,” he said. “Printing the summary card after the voter chooses whether to review it is a serious design flaw from a security perspective. This is essentially what happened in the Volkswagen ‘dieselgate’ fraud: If a system can tell that it is not being checked, it can be programmed to misbehave only when it is not being checked.”
Metsker told TYT that the county will voluntarily conduct some sort of audit, but said he did not know whether the county would allow public observers.
Metsker: We do an audit at the conclusion of every election, that’s been going on for years and years and years here in this office going back as far as anyone in our office can recall…
Cohn: Okay, are the paper audit trails that you described…part of your audit that you conducted or that you will conduct?
Metsker: That’s correct
Cohn: [Y]ou actually will look at the paper audit trail?
Metsker: That’s correct…It’s a practice we do voluntarily as an internal control and just to make sure our effort is integral…
Cohn: When does that take place, that audit?
Metsker: We’re in the process now. We have a whole bunch of processes that we execute following an election, and that’s coming up pretty soon in our sequence.
Cohn: And will that be publicly observable?
Metsker: … I don’t know that we’ve ever had anyone ask, so I’d have to take a look into that to see what we’ve done in the past. No one’s ever asked me that question.
Cohn: Would you do that for me?
Metsker: I said I’d have to take a look into that.
Cohn: … [Y]ou’re willing to take a look into that?
Metsker: I’ll take a look into…where our procedure will be and what we would be willing to do.
In a follow-up email to Hanson, TYT requested a copy of the manual or guidelines for conducting the audit, and asked whether the public would be allowed to observe. Hanson responded that there “isn’t a manual or public involvement.” Hanson said:
As we shared on the phone other day we have many post-election steps to closing out an election. We collect, compare, and package voting machine tapes after every election as we complete post-election close-out. For the past 15 years, we have voluntarily conducted this post-election procedure in our office to ensure the election was conducted fairly and accurately. As with many processes or procedures that take place in offices every day, there isn’t a manual nor public involvement. It’s just the way we choose to conduct business with the goal of continuous improvement.
Under Kansas law, state-funded recounts occur only if the victory margin is less than ½ percent and are conducted electronically, not manually. Otherwise, candidates must fund manual recounts themselves and are reimbursed by the state only if the manual recount reverses the election outcome.
This self-funding requirement can discourage recount requests by losing candidates in primary elections, who would already face pressure to concede for the sake of party unity. As noted, both Welder (who had a seven-point lead against the Republican District 3 candidate in a February poll) and Colyer conceded defeat despite the computer snafu in Johnson County.
Kansas recently passed a new audit law going into effect next year, but the law will not require manual audits for most races. Rather, it will require manual audits of 1% of precincts in the following limited instances:
- In presidential election years: one federal race, one state legislative race, and one county race;
- In even-numbered, non-presidential election years: one federal race, one statewide race, one state legislative race, and one county race; and
- In odd-numbered election years: two local races, selected randomly after the election.
Moreover, unlike hand-marked paper ballots, use of touchscreens like the ExpressVote can alter election outcomes simply by failing to work properly on Election Day (no vote flipping required), a scenario that even the most robust manual audit would be unable to rectify. Metsker insists that emergency paper ballots are available for hand marking and that poll workers are trained to distribute them if the ExpressVote machines run into trouble.
Cohn: … [D]id you have paper backups at the polls in case the ExpressVote machine wouldn’t work at all, in case there was some kind of failure?
Metsker: Yes we did, we have redundancy and we have backup plans.
Cohn: So you actually did have paper ballots at all the polling places?
Metsker: Yes, ma’am, we did.
Cohn: Can you tell me about how many? Is it enough for…a certain number of hours?
Metsker:… [W]e have enough at every location that if we have a problem we can prepare more and bicycle them over there as needed. We don’t have enough for the entire day, but we have enough to get started in case there’s a problem like that.
Cohn: So for about how much time…during peak hours would you say?
Metsker: An hour and a half or so.
Cohn: …And are poll workers instructed to use those if the machines don’t start?
Metsker: Absolutely, that’s part of their six hours of training.
Cohn: Okay, because I think there was some confusion about that, I read that anyway in someone else’s article.
Metsker: There were a lot of things said that weren’t true, so don’t believe it if it’s in print.
Although the county’s election worker and supervising judge training manuals instruct workers to provide paper ballots for hand marking upon request by “gold voters” (defined as voters who “do not want to vote on a machine”), it does not instruct workers to provide them if the voting machines fail. And voters in Johnson County reported that election workers did not appear to have a backup plan when machines failed in certain locations on Election Day, causing some voters to “give up and go to work.”
Some polling places also experienced long lines, a problem that Susan Greenhalgh, policy director with the National Election Defense Coalition, attributes to the county’s use of ExpressVote touchscreen machines. As Greenhalgh testified last year to the state legislature in Georgia (which almost passed a bill authorizing this type of system until election-integrity activists intervened), touchscreen systems tend to cause long lines because they limit the “number of voters that can vote at any given time…to the number of [touchscreen] devices [at the polling place].” By contrast, with hand-marked paper ballots and scanners, “voters need only a pen and a place to mark their ballot, making it easy to scale-up during busy voting periods and decreasing lines and wait times.”
In fact, the likelihood of touchscreen-induced long lines was predicted and explained by election-integrity advocates in an article published by the Shawnee Mission Post shortly before Johnson County’s controversial decision to buy the ExpressVote system in May 2018. At the time, Johnson County Election Office Administrator Nathan Carter responded that, “We do not anticipate any delays due to using touchscreen voting machines because our county’s voters are used to voting on touchscreens” and that the “only change for voters will be reviewing their paper ballot showing their votes before inserting the paper ballot back into the machine and casting the ballot.”
Despite the long lines and other problems that arose during the ensuing primaries last month, the county has decided to move forward with its contract to acquire 1,000 more ExpressVote touchscreens for use in November. On Aug. 30, Kobach re-appointed Metsker to another four-year term.
Jennifer Cohn is an election-integrity advocate. TYT has asked her to investigate the transparency and verifiability of last month’s voting in Kansas and to chronicle her efforts. You can follow her on Twitter here.