A key reliability-enhancing provision of Minnesota law is that all transactions result in a paper record. Photo by Nicole Neri/Minnesota Reformer.
Architects consider the people who will use a space, the structural elements that will give that space stability, and the surfaces that people will see and interact with, down to the handle on the front door.
A polling place has the same elements. The voters and election judges are the heart and soul. The foundational structure is built of paper: The signed voter certificates and marked ballots that provide a stable record. And the touch points one sees upon entering the space are the election equipment — computerized equipment, in a modern polling place.
Many Minnesota polling places have three kinds of computerized equipment.
- Pollbooks aid in voter check-in and election-day registration.
- A ballot marking device provides an alternative for voters who prefer not to mark their ballots with a pen — perhaps based on their abilities.
- An optical-scan tabulator checks the ballot’s markings before depositing it in a ballot box.
Each of these systems has its own story for why it is important, what precautions are taken to ensure its reliability, and how — if at all — it can be networked with other systems.
Two-thirds of Minnesota’s counties use electronic pollbooks, with the remainder using traditional paper pollbooks. Because all the most populous counties are in the electronic pollbook camp, most voters experience this technology. Although Minnesota law provides each county the flexibility to select its own vendor, all the counties have made the same choice, KNOWiNK’s Poll Pad.
Electronic pollbooks can improve efficiency and reduce errors. One feature that became particularly important in 2020 is that they ease the coordination of polling-place and absentee voting. All accepted absentee ballots need to be noted in the pollbook so that the voter doesn’t vote again at the polling place.
With paper pollbooks, if lots of absentee ballots arrive after the rosters are printed, marking them into the pollbooks manually is a huge job. Also, once polling-place voting is underway, no newly accepted absentee ballot can be opened until the pollbook is checked to see whether the voter voted in person. That too would scale up poorly without the use of electronic pollbooks.
For this reason, the Legislature expressly allowed electronic pollbooks to be networked beyond the polling place for purposes of updating absentee ballot records. Most commonly, this is done using Wi-Fi communication from the pollbook to an access point within the polling place and then cellular data communication from the access point to a county server. If you see the blinking lights of an access point in your polling place, that’s what it’s for. (The pollbook systems also make use of short-range Bluetooth networking between the tablet computers and printers.)
Aside from updating absentee ballot records, other aspects of the pollbooks’ operation are handled locally, unlike in some other states. No off-site communication is needed to check in a voter or process an election day registration. That way, smooth polling place operation is not jeopardized if there are problems with networking or a central server.
The ability of Minnesota’s pollbooks to operate independently is one aspect of a focus on reliability. Given that the world includes attackers, reliability necessarily encompasses cybersecurity. However, it goes beyond it — many systems fail even without coming under attack.
Another key reliability-enhancing provision of Minnesota law is that all transactions result in a paper record. When a voter is checked in, the pollbook prints out a voter certificate, which the voter signs. When a voter is registered, the pollobook prints out a voter registration application, which the voter signs. These paper records are retained for 22 months, immune from any subsequent computer problems.
After making it through the sign-in process and receiving a ballot, the next computer system a voter may encounter is a “ballot marking device” (BMD). All polling places have at least one BMD, rarely more. The particular kind is dependent on the county. As with pollbooks, Minnesota has taken a more cautious approach to this technology than in some other states.
Rather than directing all voters to use a BMD, our polling places offer them as an alternative to pens. Minnesota’s BMDs print markings onto the same paper ballots as are used for voting by hand; they don’t print a specialized format onto blank paper. Therefore, all votes are recorded in a uniform, human-verifiable paper record, mostly without any involvement of computer technology. This largely obviates the security concerns raised in other states.
All BMDs go through the same certification and testing processes as the tabulators, described subsequently. This includes testing prior to each election with the specific ballots for that election.
There is no reason for a BMD to be networked, although the ImageCast Evolution, used in 6 counties, is integrated with a tabulator and may have limited networking for that reason.
Finally, the voter is likely to feed their ballot into the ballot box via an optical-scan tabulator. Just a smattering of precincts in Pine and St. Louis Counties skip this technology and tabulate their ballots by hand; the rest of precinct polling places use tabulators from three manufacturers. The tabulators provide efficient and accurate vote counts.
But why tabulate the ballots at the time they are deposited into the box? Why not just store them all up and feed them through a central tabulator at the end, the way absentee ballots are tabulated? The key point is that precinct optical-scan tabulators aren’t just a mechanism for producing vote counts. They are equally importantly a mechanism for pointing out any problems with the ballot markings while the voter still has a chance to try again.
There are a variety of ways a voter might mismark their ballot. For example, they might mark two candidates in a race where only one is allowed. In primaries, voters commonly vote in more than one party, which would invalidate the entire partisan portion of the ballot. Regardless of the specific problem, the tabulator gives the voter the opportunity to eject the ballot rather than depositing it in the ballot box. An election judge can then set the ballot aside as spoiled and issue a fresh ballot to the voter to try again. Without the technology, this wouldn’t be possible; there’s no other privacy-preserving way to check ballots before they are cast.
The paper ballots are the stable foundational record of the votes cast. They can always be recounted by hand if there’s any question about the tabulation’s accuracy.
And even if no question is raised, a random sample of precincts is routinely checked by hand after every state general election. (Later columns in this series will explain these recount and audit processes.) But if the initially reported results are way off and substantially different results need to be reported after hand checking, faith in the elections process would be substantially eroded. So there are safeguards in place to ensure the tabulators work correctly in the polling places.
First, the design of each voting system is evaluated through a certification process. The voting system is the complete end-to-end combination including the computers the counties use to design ballot layouts as well as the polling-place systems for marking and tabulating the ballots. To be certified by the state, the system needs to have been tested by a federally accredited independent testing authority.
Second, the specific equipment being deployed to the polling places is tested before each election, configured for that election’s ballot designs and operating on corresponding “test decks” of ballots. That way, any specific configuration problems are caught.
For example, imagine if the tabulator were configured with the candidates listed in a different order than is printed on the ballots. Because the tabulator doesn’t read the textual candidate names and instead just goes by physical location on the ballot, this would result in swapped vote counts, if not caught through the pre-election testing.
Perhaps the most controversial question about tabulators is whether they are networked. The answer in Minnesota depends on the county. In some, it is “no” — the tabulators operate entirely off-line, with vote counts delivered only through printouts and removable storage devices.
In other counties, the answer is “briefly” — the tabulators operate offline as in the other counties, but then after the results have been printed and stored to removable media, the tabulator connects in order to more quickly report unofficial results, generally via cellular data service.
The way the tabulators are designed, there’s no risk that election judges will accidentally connect them to the cellular network prior to the finalization of the results.
For example, the model used in Hennepin County and most other counties requires the following steps: 1. open a locked compartment; 2. hold down a physical button for five seconds; 3. press an on-screen button to close polling and print the results; 4. then press another on-screen button to connect to the cellular network and send the results.
The cellular data communication is encrypted by the voting equipment and, in the more modern deployments, is then subject to a second layer of encryption provided by the cellular provider — a so-called Virtual Private Network. This further reduces the chance that public confidence will be shaken when the early unofficial results are confirmed with the printout and removable storage devices delivered from the polling places.
In short, Minnesota polling places can use up to three kinds of computers: pollbooks, ballot marking devices, and tabulators. They can use up to three kinds of networking: Wi-Fi, cellular, and Bluetooth.
They receive three layers of testing: system certification, pre-election testing, and post-election audits and recounts.
And they provide three benefits: accessibility to voters with disabilities; the chance for voters to correct their mistakes; and efficient publication of unofficial results, even in the face of absentee ballots arriving on Election Day.
Our stories may be republished online or in print under Creative Commons license CC BY-NC-ND 4.0. We ask that you edit only for style or to shorten, provide proper attribution and link to our web site. Please see our republishing guidelines for use of photos and graphics.