A poll worker at Martin Luther King Center in St. Paul on Election Day 2020. Photo by Nicole Neri/Minnesota Reformer.
This is part of an occasional series on election administration. Read part 1, “Who does what?” Part 2, “Who can vote in Minnesota?” Part 3: “How and why polling places are computerized.” Part 4: “How absentee voting works.” Part 5: “Reconciliation.”
Election officials do their best to get results out on election night or the next morning, but always with a note that they are unofficial. Perhaps preliminary would be a better word than unofficial.
These results do come from an official source, unlike news reports of estimates derived from exit polling. But they haven’t been through all the checks and revisions that occur following the election. Those post-election activities are essential to producing final, official, certified results that merit confidence. Nothing done quickly can have the same level of assurance.
Turning the unofficial results of election night into the official results that allow a winner to receive their election certificate is called canvassing. In Minnesota, the canvass process includes administrative checks, formal board action, and a manual audit of randomly selected precincts.
Election administrators in each jurisdiction produce a “canvass report” that tabulates the votes cast, as well as the number of voters, and Election Day and advance registrants. This largely repeats the information from election night. However, it reflects several kinds of cross-checking and the revision of any errors found through that checking.
Most basically, the administrators check that the vote counts were correctly uploaded to the Election Reporting System. They do this by closing the loop back to the definitive source of information: the printouts from the tabulators.
Discrepancies don’t mean an evil-doer corrupted the data upload. They typically reflect such mundane errors as transposed digits in hand-keyed numbers. The great thing about the process, though, is that it will find any discrepancy, regardless what caused it. If the numbers all check out, then whatever might have gone wrong didn’t.
Election administrators also double-check that all the batches of ballots are accounted for in each precinct. That may include ballots voted at the polling place, by mail, in person at the county, and in person at the city. All the batches should have been included on election night — Minnesota doesn’t report absentee votes separately, unlike some states. But in the rare event that one batch was inadvertently left out, that too can be fixed.
During the canvass period, the unofficial results on the Election Reporting System may get updated multiple times as individual precincts’ vote counts are revised. These revisions are almost always small, but the repeated changes in the totals can be unnerving. They reflect the election administrators’ devotion to transparency, providing their current best understanding of the results. The alternative would be a curtain that lifts when the results are final.
Once the canvass report is complete, the administrators submit it to a canvassing board for certification. Strictly speaking, the administrative document is known as an abstract of the votes cast; it only becomes the canvass report when adopted by the canvassing board.
Municipal and school-district elections are handled by local canvassing boards in those units of government. All other elections go initially to a county canvassing board in each county. The county canvassing board certifies the tabular information in the canvass report and declares the winner for each county office and for any state office contained entirely within the county.
For state and federal offices, the county canvassing board serves as the first tier in a hierarchical arrangement, aggregating information and feeding it up to the state canvassing board. Just as the county canvassing board receives results from each precinct, the state canvassing board receives the canvass reports from each county.
The state canvassing board declares the winner of each federal office and of each state office that crosses county lines. For state general elections, the state canvassing board meets three weeks after Election Day.
Ordinarily the state canvassing board declares its results the day it meets, but the statute allows it up to three days to do so. That makes 3.5 weeks the outside limit of the waiting time for official results. They can still be challenged — recounts and court contests are the topic of the next column in this series. But all the necessary steps are complete.
Everything discussed thus far amounts to glorified addition. The votes were already counted on Election Day, or somewhat earlier for absentee ballots. All that was necessary during the canvass was to compile those results into progressively more inclusive sums. But one other process during the canvass period goes beyond the record keeping and addition. It checks the accuracy of the actual vote counting.
Routine checks of vote counting are present in many states and are known generically as tabulation audits. Minnesota’s particular tabulation audit is called the Post-Election Review (PER). Because it involves humans looking at ballots and counting the votes by hand, it is easy to confuse with a recount. But unlike a recount, it isn’t a procedure for resolving a dispute about the election’s outcome — it is a routine process to ensure that the optical-scan tabulators are operating reliably.
Some states conduct their tabulation audits after the canvass is complete so that there is a fixed target to audit: The election administrators have already committed to their version of the vote counts. Other states conduct their tabulation audits before certification of the canvass so that if any discrepancies are found, the results can be revised accordingly. Minnesota gets the best of both worlds by slotting the PER in between the county and state canvassing board meetings.
At the county canvassing board meeting, the canvass report is presented and officially certified, providing the baseline tabulation to be audited. The county canvassing board only then selects by lot the precincts to audit. The election administrators need to be prepared with audit-ready results in all the precincts.
If any substantial discrepancies are found in the randomly selected precincts, more precincts are checked, in the worst case leading to a full recount. But even if only minor discrepancies are found, they get forwarded to the state canvassing board as revisions to the initial county canvass report. That way, the finally certified results are as accurate as they can be.
Some minor adjustments are quite common, resulting from the ability of humans to discern votes expressed in nonstandard ways that the optical scanners can’t read. For example, a voter may have marked two different candidates, but crossed one of them out. The human can see that this is a vote for the other candidate — a Minnesota statute says so. But for the optical scanner, it looks like the voter selected too many candidates, which counts as no vote at all.
Situations such as this — in which the PER discovers voter intent that the optical scanner is not designed to detect — do not count toward the judgment whether the scanner is operating properly. Only an unexplained difference between the hand count and machine count is treated as a red flag. But even the explained differences are fed up to the state canvassing board so as to produce the most accurate result.
The acceptable number of unexplained differences is quite small. From 2006 to 2020, it was 0.5% of the ballots, except in the smallest precincts. That limit was never exceeded in any of the hundreds of precincts checked each election cycle. Starting in 2022, the Legislature has lowered the limit to 0.25%. Indeed, the new limit is even lower, in percentage terms, for large precincts: No more than five differences are ever allowed. This is particularly stringent when one considers that a single misread ballot can manifest as two differences, such as one fewer vote for candidate A and one more vote for candidate B.
In each state general election, the counties audit the races for president or governor, U.S. representative, and U.S. senator, if up for election. The number of precincts randomly selected in each county depends on the number of registered voters and of precincts. Even the smallest counties audit at least two precincts, and the largest (Hennepin) audited 13 in 2020.
Could a problem escape detection by the PER, yet affect the declared winner of an election? Yes, in theory. A problem could affect only one of the unaudited races. Alternatively, discrepancies in an audited race could be below the PER’s threshold, yet enough to tip a really close race. Or larger discrepancies could be concentrated in only some precincts — ones that lucked out of the random selection. In practice, however, the PER provides greater assurance when taken in context of the pre-election testing and the potential for recounts.
In particular, Minnesota’s options for recounts are well suited to catching the problems that evade the PER. One kind of recount handles very close races, where the relevant errors might slip under the PER threshold. Another kind of recount is good at handling large errors concentrated in a few precincts, which could evade the PER’s lottery. These recount options are the topic for the next installment, along with the court review that serves as the final fallback.
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.