(Here's a link to the original website main page.)
On April 19 (2023) the Portland City Council approved the draft proposal for city code changes that adopt ranked choice voting. The flaws explained below are in the revised city code. Hopefully these flaws will be remedied in the future, after experiencing Portland's city elections using the adopted version of ranked choice voting.
The version being presented to the Portland City Council on Wednesday April 12 has been slighly improved from the earlier version (referred to in later sections on this page), yet it still needs the following specific refinements:
The document titled CHAPTER 2.08 - ELECTION OF CANDIDATES specifies details about using ranked choice voting in the 2024 Portland City Council and Portland Mayor elections.
Unfortunately the current wording has some serious flaws. Fortunately it's easy to refine the wording in ways that will prevent unfair elections results. As an important bonus, these wording changes will simplify voter education.
These proposed wording refinements primarily refer to the “policy choices” named Result of Voter Overvoting on the Ballot and Number of Candidates a Voter Can Rank on the Ballot.
These recommendations were written by the subject-matter expert for the Ranked Choice Oregon ballot initiative, which did not collect enough signatures to reach the next stage.
The introductory statement that “voter intent cannot be determined where there is an overvote” is false. That statement is useful in Australian elections where voters write a number next to each candidate name. But ballots here in Oregon use darkened ovals in “choice” columns to indicate candidate rankings. So there will be no uncertainty about what the voter intended when they mark the overvote.
An overvote on an Oregon ballot will clearly indicate the voter is ranking the overvoted candidates higher than the candidates who are marked at lower choice levels. The current wording mistakenly specifies that overvotes be completely ignored, as if the voter had not marked any candidates in the overvoted choice column.
The current definition of an overvote (2.08.010 I) is correct. It says: “Overvote” means a voter has ranked more than one candidate at the same ranking in the same contest. For example, a ballot has an overvote if a voter assigns two candidates the number 1 ranking in the same contest.”
However, the definition of “Highest-ranked active candidate” includes the words excluding overvotes, which is not correct. The recommendation below provides a corrected definition for Highest-ranked active candidate.
The “highest-ranked active candidate” on a ballot can be identified in some cases even though an overvote is involved. For example, if only one of the candidates in an overvote is an active candidate (which means the other candidates in the overvote are inactive candidates) the ballot clearly should be counted as support for the single active candidate.
Consider that when an election is close enough to trigger a recount or audit, the marks for clearly defeated candidates can be ignored so they no longer contribute toward an overvote. In particular, if two candidates are contesting the result, only the overvotes that involve both candidates need to be regarded as an overvote.
The current wording (2.08.010 G) says: “Highest-ranked active candidate” means the active candidate a voter has assigned to a higher ranking than any other active candidate in the contest being tallied, excluding overvotes.
The words excluding overvotes must be removed from this definition. This allows an overvote to be counted in support of a single active candidate when the other candidates in that overvote are inactive candidates.
Here's a suggested corrected definition:
(2.08.010 G)
“Highest-ranked active candidate” means the active candidate a voter has assigned to a higher ranking than any other active candidate in the contest being tallied , excluding overvotes.
If more than one candidate would be a highest-ranked candidate because of an overvote, the ballot temporarily becomes inactive until the next round when one of the overvoted candidates might have become an inactive candidate.
The number of candidates a voter can rank on the ballot interacts with how overvotes are counted. To understand why, consider a ballot on which the voter ranks their first 5 choices in columns First choice through Fifth choice, and the voter wants to mark their most-disliked candidate at the lowest ranking, yet there are 12 candidates and only 7 choice columns. If the voter knows that an unmarked candidate will be ranked lower than all other candidates, the voter can leave their most-disliked candidate unmarked and mark the remaining 6 candidates in columns Sixth choice and Seventh choice. This situation would require the use of overvotes, or else guessing which candidates have no chance of winning and therefore can be unmarked along with the voter's most-disliked candidate.
In this case the overvotes clearly convey the voter's intent, which is to rank their most-disliked candidate lower than all other candidates. Yet the current wording (2.08.30 A. 4.) says: If a ballot contains an overvote, the voter's vote is transferred to the next highest-ranked active candidate on the voter's ballot, if any. This counting rule violates the voter's clear intent.
The infographic below helps to convey this interaction between overvotes and the number of choice columns.
Here's a wording recommendation about counting overvotes. In addition it helps to resolve the issue of how many choice columns should appear on the ballot.
(TBD) If two or more active candidates become the highest-ranked active candidates then temporarily this ballot is counted as an inactive ballot. If the election result is close enough to trigger a recount or audit then such ballots can be matched with equivalent ballots on which the same active candidates are overvoted and the matched ballots are apportioned among the same-ranked active candidates in equal whole numbers. For example if two ballots rank candidates A and B as the current highest-ranked active candidates, one ballot counts as support for candidate A and the other ballot counts as support for candidate B. Any unmatched ballots are ignored during that elimination round.
In other words, counting overvotes is only needed when counting them might cause a different candidate to win. And that counting would be done during a recount or audit where the counting does not completely rely on primitive software that fails to count overvotes.
If anyone claims that software cannot count overvotes as described here, please point them to this open-source software.
The above changes require minor wording changes in one section that describes the single-winner tabulation process. Here's a suggested change for instant runoff voting tabulation. The specific changes are indicated by the strikethrough and the underline.
(2.08.020 B 2 a)
If no candidate has a majority of votes, the active candidate with
the fewest votes is defeated and the non-defeated active candidates
retain the number of votes counted for them in any prior round.
Each vote on a ballot that was counted for a defeated candidate in
the prior round is then transferred to the next highest-ranked active
candidate on that voter's ballot that is not an overvote, if any.
If a ballot's next highest-ranked active candidate involves an overvote for more than one active candidate in that overvote the ballot is temporarily counted as inactive until the next elimination round.
The handling of overvotes is slightly different when the tabulation method is the single transferable vote. In this case when a ballot has an overvote, and the highest-ranked active candidate is the only active candidate among the overvoted candidates, and that candidate is elected, the ballot's transfer value becomes zero, and it becomes an inactive ballot. This means it cannot contribute toward electing any further candidates.
This rule avoids the complications that would arise otherwise. Here's a suggested wording for this provision. Only the underlined portion is changed.
(2.08.020 C 2 b 2) Each ballot counting for an elected candidate is assigned a new transfer value by multiplying the ballot's current transfer value by the surplus fraction for the elected candidate, with the result truncated after four decimal places. “Surplus fraction” is calculated by subtracting the election threshold (“T”) from an elected candidate's vote total (“V”), then dividing that number by that elected candidate's vote total, and then truncating that number after four decimal places, where the candidate's “vote total” is the total transfer value of all ballots counting for a candidate in a round of tabulation. Any ballots that helped to elect the candidate based on being the only active candidate in an overvote are given a transfer value of zero.
The current rules do not specify how an unmarked candidate should be ranked compared to a candidate who is ranked at the lowest rank. The unmarked candidate can be ranked at the lowest rank printed on the ballot, or ranked below all the ranked candidates. Each of these two interpretations has advantages and disadvantages. Here is a suggested interpretation:
(TBD) A candidate who is not marked at any choice level is regarded as ranked below all the marked candidates.
Another missing rule is how to interpret the ranking of write-in candidates on ballots where that candidate's name is not written. This detail might seem unimportant. Yet candidates and voters will ask about this important issue because it can affect who wins.
There are three possibilities for interpreting this situation. (The non-named candidate can be ranked at the lowest printed choice level, the same as an unmarked candidate, or lower than all the named candidates.) Here's a suggested wording that specifies the compromise alternative.
(TBD) If a voter writes in the name of a write-in candidate, all the other ballots on which that candidate's name is not written are interpreted as if that candidate is an unmarked candidate.
A provision for batch elimination needs to be added. Batch elimination gives permission to Election Officials to eliminate two or more clearly unpopular candidates during the first elimination round.
This provision simplifies counting during election night. Also it provides another way to dismiss write-in candidates who are named on only a few ballots. This provision takes advantage of the fact that the exact elimination sequence among unpopular candidates doesn't affect who wins.
(TBD) Batch elimination: If preliminary tabulation information indicates that one or more candidates clearly cannot win according to criteria previously approved by Election Officials, and if the results will not change by eliminating all these unpopular candidates together, then these unpopular candidates can be eliminated together in the first elimination round.
A provision for interpreting multiple marks in the same row is needed. If the entire row of marks is ignored the candidate would be ranked lower than all the other marked candidates, which clearly is not what the voter intended.
This ambiguity is easy to clarify by adding this sentence:
(TBD) If the ballot contains marks for the same candidate in more than one choice column then the highest-ranked mark is used and the other marks in that candidate's row are ignored.
This clarification combined with the clarifications about how to handle overvotes means that any combination of ballot marks can be counted. And it means voters don't need to be strongly warned to avoid ballot marks that otherwise would cause their ballot to be rejected as uncountable. As an economic bonus, these refinements will dramatically reduce the cost of voter education, which taxpayers will appreciate.
When there is a tie between two candidates during any elimination round, and if the tie can affect who wins, the tie should be resolved using information on the ballots.
When the tie can affect who wins, all the ballots, including inactive ballots, should be counted to determine how many voters rank one of the candidates over the other candidate and how many voters have the opposite preference. This “pairwise comparison” reveals which candidate should be eliminated as less popular. Only if neither candidate wins and neither candidate loses this pairwise comparison is it appropriate to determine the outcome “by state law.”
(TBD) “Pairwise comparison” means a one-on-one comparison between any two candidates that counts how many ballots indicate a preference for one of the two candidates over the other candidate and how many ballots have the opposite preference. The candidate with the larger pairwise count is the winner in this pair and the candidate with the smaller pairwise count is the loser in this pair. If both pairwise counts are the same then neither candidate wins and neither candidate loses this pairwise comparison.
(TBD) If there is a tie in an elimination round then a pairwise comparison is used to identify which of the two candidates has fewer supporting votes and therefore is chosen for elimination. If there is still a tie then the outcome is determined by state law.
If there is any concern that the certified election software chosen for these elections can fail to correctly interpret these vote-counting rules, a new section with the following words can be added:
(TBD) If the software used for the election does not interpret ballot marks as described in sections 2.08.10 and 2.08.30 and the software-identified election winner could change as a result of following these rules then an audit is required and the winner or winners must be determined based on following the rules described in sections 2.08.10 and 2.08.30.
The current wording empowers county election officers to claim that their county's voting equipment has a limit on the “number of rankings on the ballot” without any lower limit being specified. This absence of any minimum could empower an election officer to undermine the success of ranked choice voting by claiming a limit of only two or three ranking levels, which would allow fully (with no overvotes) ranking only two or three candidates. To prevent this opportunity for possible sabatoge, a miniumum number of ranking levels must be specified. A miniumum of five ranking levels can be specified as follows.
(2.0830 A 1 a) Election Administered by Single County. If a city candidate election is administered by a single county elections officer, and that officer determines the voting equipment cannot accommodate eight ranking levels on the ballot, the number of ranking levels on the ballot for that election will be at least five ranking levels and is otherwise limited by the maximum number that the officer determines can be accommodated by the voting equipment.
(2.0830 A 1 b) Election Administered by Multiple Counties. If a city candidate election is administered by county elections officers in more than one county, and one or more officers determines the voting equipment cannot accommodate eight ranking levels on the ballot, the number of ranking levels on the ballot for that election will be at least five ranking levels and is otherwise limited by the maximum number that can be accommodated by the voting equipment that can accommodate the lowest maximum number of rankings.
Soon Oregon will be using the kind of ranked choice ballot on which an overvote will convey a clear intent by the voter to rank the overvoted candidates higher than the candidates marked at lower rankings. Therefore it is not acceptable to regard an overvote as an excuse to ignore the ballot as inactive. These recommended wording changes would protect the voter's intent. And these changes allow flexibility regarding software limitations, election-night vote counting, and jurisdiction differences.