Toward Technical Standards for Voting Systems that Accommodate Ranked-Ballot Elections
Overview of requirement of standards for ranked ballot elections:In a traditional election, vote-counting equipment can store running totals for each ballot position. In a ranked choice voting election, ballots must allow voters to rank candidates in order of choice and voting equipment must store an electronic record of each voter’s ranking. In a separate step, all rankings must be aggregated and then tabulated according to the rules of the election. Because of this difference in the role of the voting system, technical standards for ranked-ballot elections must address these two steps independently.
Legislation authorizing ranked ballot elections can address certain details in a variety of ways. One example is duplicate rankings. Some legislation discards ballots when a duplicate ranking is reached, whereas other legislation skips the duplicate ranking. Thus, vendors must specify how their systems handle particular details and if there are any options for handling those details. Questions in red text address details that vendors must spell out.
In the text below, we lay out all the issues that must be addressed in the technical standards as well possible options for each issue. It is divided into three parts:
- Voting and capturing rankings,
- Cleaning data and tabulating results, and
- Result reporting, logic and accuracy tests, and post-election counts.
Part One: Voting and capturing rankings
This part concerns the process by which the voters completes her ballot and the voting system scans it for errors, notifies the voter if there’s an error, and then stores the image of the ballot.
1. Ballots and ranking
Ballots shall be simple and easy to understand. Ballots shall allow voters to indicate their first choice, their second choice, their third choice and so on.
For vendors: What is the maximum number of candidates (including write-ins) that the system can accommodate?
2. Error prevention (DRE equipment)
The voting system shall be able to prevent the voter from casting an invalid ballot by ensuring voters do not rank more than one candidate with the same ranking. It also shall be able to prevent the voter from skipping a ranking.
3. Error notification (precinct-based equipment using paper or punch card ballots)
A ballot with one or more candidates ranked sequentially starting with first choice shall be considered an intended ballot. All other ballots shall be considered error ballots (although only invalid once the voter error is reached – for example, a voter who had ranked two candidate as their second choice will still have their ballot count for their first choice). Errors include the following scenarios:
- Duplicate rankings, such as listing more than one candidate for first choice
- Skipped rankings, such as listing a second choice but no first choice
- Same candidate ranked more than once, such as a candidate ranked first and third.
The voting system shall be able to notify voters if they cast a ballot with an error. The system shall print out a descriptive message in plain English that informs the voter of the error. If error notification is used, the voter shall have the options of correcting her ballot and overriding the error by submitting it “as is.” The voting system may be programmed to notify voters if they rank fewer than a fixed number of candidates in a race.
For vendors: Can error notification be turned on or off at the jurisdiction’s request?4. Write-in candidates
The voting system shall allow the voter to vote for one write-in candidate for each seat being elected.5. Storage of ballot images
The voting system shall store true ballot images and shall be able to produce a human-readable report containing all ballot images. The ballot images should store duplicate rankings, skipped rankings and the same candidates listed more than once exactly as the voter marked the ballot. For example, if a voter casts an undervote as a 1st choice and a vote for candidate A as a 2nd choice, the vote should be stored as (undervote, A) rather than in the “cleaned” or advanced form, “(A, undervote).
For vendors: Does the system keep track of the actual candidates that received the same ranking or does it store a code for duplicate rankings?
Part Two: Cleaning data and tabulating results
This part concerns the process by which the stored data is cleaned and interpreted according to the authorizing legislation and then tabulated to determine the winner or winners of the election.
1. Skipped rankings
Does the system allow for the distribution of the ballot to the next valid ranking?2. Duplicate rankings
Does the system allow for the skipping of no more than a fixed number of rankings and exhausting ballots that skip more than that (e.g., skip from 1st to 3rd, but not from 1st to 10th)?
Does the system allow for the exhausting of ballots when they reach skipped rankings?
Does the system allow for the exhausting of duplicate rankings?3. Same candidate ranked more than once
Does the system allow for skipping of duplicate rankings?
Does the system allow for the counting of duplicate rankings if exactly one of the duplicate candidates has not been eliminated?
Does the system allow for the dividing of ballots fractionally among all duplicate ranked candidates?
Does the system allow for skipping to the next valid ranking?4. Ties
Does the system allow for the exhaustion of ballots after it reaches a candidate ranked more than once?
Does the system allow for resolving of ties for last place by drawing lots?5. Simultaneous elimination
Does the system allow for resolving of ties for last place in favor of the candidates with the most votes in the previous round, or by lots if tied in all previous rounds?
Is it possible to set the tabulation to simultaneously eliminate two or more candidates with the fewest votes if the sum of their votes is less than the candidate with the next most votes?
Is it possible to simultaneously eliminate all candidates with fewer than a fixed number of percent of the valid votes?
Part Three: Result reporting, logic and accuracy tests, and post-election counts
- Result reporting: The system shall be able to produce reports of the number of votes credited to each candidate, along with the number of exhausted ballots, in each round of counting. The system shall be able to produce a human readable file of all rankings on all ballots in the election. The system shall also be able to report the total number of complete undervotes (no candidate ranked) and overvotes for ranked choice races.
- Pre-election Logic and Accuracy tests: L&A testing shall consist of running sufficiently large test decks with predetermined patterns through the system, capturing ballot images, and tabulating the ranked choice results. The ballot images and the tabulated results can be compared to the predetermined patterns.
- Post-election Logical and Accuracy tests (e.g., 1% manual tallies): The system shall have clear procedures for performing a manual tally of ballots in one or more precincts and comparing the round by round results of that tally to a machine tabulation of the ballot images from the same precinct or precincts.