Dear Readers, Below you will find a long discussion of ranked choice voting in Albany. Be warned, this piece is about 4,400 words. But I think it’s important to look at some examples of ranked choice voting, and how it would work in our city. For a shorter take, please see my ballot statements here and here. If you get bored or run out of time, you can scan the subheadings or skip down to the closing thoughts at the end. All words in purple are hyperlinks.
In Albany, we’re being asked to consider changing to a different set of voting rules known as ranked choice voting (RCV). The proposed RCV system we are being asked to consider will certainly be more expensive than our current system, adding an extra $26,000 cost for each election, but there is no guarantee is will be any better. All voting systems are imperfect, and a system that is suitable for one city may not be the best choice for another city.
Social Choice Theory
The imperfection of all voting systems was established by Nobel Laureate economist Kenneth Arrow with his famous Impossibility Theorem. In his 1951 book, Social Choice and Individual Values, Arrow jump-started the field of social choice theory. This body of thought attempts to provide a systematic framework to explore how well individual votes and other preferences can be aggregated to the societal level. Arrow’s Impossibility Theorem showed that all voting systems have fundamental flaws, although these might not always be serious. Arrow famously stated, “Most systems are not going to work badly all of the time. All I proved is that all can work badly at times.”
Later another Nobel Laureate, the Indian economist Amartya Sen, extended social choice theory to examine problems in developing countries like the failures in institutions that lead to poverty and famine. You can read more about Arrow (here and here) and Sen (here and here).
There are dozens of alternative voting models. RCV is just one. Before we jump down the alternative voting rabbit hole, it’s a good idea to read about Arrow’s Impossibility Theorem to better understand the complexities of social choice theory. Don’t believe me? Read this. But I’ll leave it up to you to browse the literature. You can also start here or here, or search for “ranked choice voting” in Google Scholar.
Ranked choice voting in Albany
For now, let’s focus this conversation on Albany. The Albany City Council and the Charter Review Committee have looked at the issue of RCV several times over the years. Here is a recent example from the March 19, 2018 city council meeting, item 8.1. The staff report is a useful overview.
Albany is a charter city — like many older cities, Albany has its own set of bylaws, or charter. Most newer cities are general law cities that rely on the State of California’s rules for cities. Currently, general law cities are not allowed to use RCV, and most recent governors, including Jerry Brown and Gavin Newsom, have vetoed RCV bills for general law cities because of their concerns that RCV is overly complicated and will lead to voter confusion.
Because Albany is a charter city, we have the option of using RCV. However, this requires study by our charter review commission, consultation with our city attorney, and a modification of the city charter — which requires a vote of the citizens of Albany. Both the city council and the charter review committee have looked at the issue of RCV repeatedly, and in general have not been interested in adopting it.
This year a group of advocates calling themselves Voter Choice Albany began collecting signatures for a citizen’s initiative to put RCV on the Nov. 2020 ballot. A citizen’s initiative requires signatures from ten percent of Albany’s registered voters, plus some extras as a buffer. In Albany, that’s about 1,200 signatures. The group claims it had collected about 600 signatures before it had to stop because of the Covid pandemic.
Soon after their signature-gathering moratorium, the members of Voter Choice Albany approached the city council and asked it to place their initiative on the ballot as a council-initiated proposal. When the council did not express much interested in doing so, Voter Choice Albany threatened to sue the city if it did not place the measure on the ballot.
Given that lawsuits are expensive regardless of the outcome and given that Albany’s citizens will sign just about any petition, the council concluded that Voter Choice Albany would have probably eventually found their 1,200 signatures. Therefore, the city decided this was not a fight worth having and agreed to put the issue to the voters. That is how we find ourselves where we are today.
The human problem that RCV doesn’t solve
Albany’s adults are usually focused on their families and careers. Serving in a volunteer capacity on a city commission and gaining the experience necessary to serve on the city council can be a thankless task. But it is an important one. Getting residents involved in the workings of city governance is part of that long process.
If we don’t do this work, we are often left in the few months before elections seeking people who are willing to run. By the time election day rolls around, this problem is either solved or it’s not. At that point, no fancy voting algorithm is going to solve the problem for us. It’s too late. RCV is not a solution to our fundamental problem of getting citizens involved.
Single-seat ranked choice voting (instant runoff voting)
RCV as proposed for Albany comes in two flavors — single-seat RCV, also known as instant runoff voting, and a more complicated version called at-large RCV that is rarely used in the United States. You can read more about how they work at the RCV advocacy website, fairvote.org. It is a well-organized site that provides lots of information. Here is what I discovered there:
Nationally, 23 jurisdictions presently or imminently use RCV to elect officials. Of those, 17 exclusively use single-seat RCV (instant runoff voting).
To give you some perspective, the nine-county Bay Area alone has about 100 jurisdictions, including cities and counties. There are thousands of public jurisdictions in the United States.
Instant runoff voting makes a lot of sense for cities that have been conducting primary elections followed by runoff elections. Holding two elections can strain the resources of both cities and candidates, and the instant runoff process saves them both time and money.
However, Albany doesn’t use primary and runoff elections for council members, so one of the main justifications for switching to RCV is missing in our city. If Albany were to switch to instant runoff voting from its current at-large system, it is not clear what we would gain. In practice, the vote-transfer process of RCV doesn’t appear to make much difference.
According to fairvote.org, “There have been 15 RCV races in the U.S. which were won by a candidate other than the first-round leader. That’s 4.2 percent of the 353 single-winner RCV races since 2004.”
In the Bay Area, four cities that vote by districts use instant runoff voting — Oakland, Berkeley, San Leandro and San Francisco. San Francisco has been using RCV since 2004. Oakland, San Leandro and Berkeley have been using it since 2010. According to online voter registrar records, there have been 60 RCV elections in San Francisco and 83 in the East Bay, a total of 143.
In 94.4 percent of these elections, the candidate who won the first round of counting either prevailed in the first round (just like in a conventional election) or won after additional rounds of counting rank-choice ballots. In only 5.6 percent of the elections (8 of 143) did a candidate who did not take the lead in the first round come from behind to win. Both nationally and in the Bay Area, the extended ballot counting of RCV only affects the outcome of about one election in twenty.
At-large ranked choice voting
The RCV initiative on the ballot this November in Albany also allows our city to implement the even more complicated at-large version of RCV. Referring again to the fairvote.org website:
Two jurisdictions exclusively use multi-winner RCV (single transferable vote) – Cambridge, MA and Eastpointe, MI.
Two use a combination of single- and multi-winner RCV – Minneapolis, MN and Palm Desert, CA.
Two use a form of multi-winner RCV called preferential block voting – Payson, UT and Vineyard, UT.
Albany’s RCV advocates are asking the voters to make Albany only the fifth city in the country to adopt at-large RCV. It would be an experiment, and one that, in my opinion, will not do much for Albany. Here is an example of how at-large RCV work from the fairvote.org website.
The example above involves six candidates for three seats in a partisan election. This is not how elections work in Albany. First of all, our elections are non-partisan. Secondly, it would be rare to have six candidates for an Albany election. Finally, we are left to assume that one Republican winning thanks to RCV is a better outcome than electing three Democrats. I’m not sure why, in general, that should be the case.
It can be difficult to picture how adding second- or third-choice candidates, or even lower-ranked candidates, from the ballots of eliminated candidates can shape the eventual outcome of the election. There are three conditions that typically apply in the unusual case that counting the subsequent ranked choices beyond the first round makes a difference:
1) If, after the first round, two candidates are nearly tied.
2) If there are enough candidates to create a depth of ranked ballots that are capable of making a difference.
3) If the voting patterns in the subsequent rounds are sufficiently different from the patterns in the first round.
A useful analogy is to the counting of vote-by-mail (VBM) ballots after the polls have closed. In the past, VMBs were mostly absentee ballots, but over time more and more voters have switched to becoming permanent VBM voters. Imaging the following scenario:
In an election, 100,000 voters go to the polls. After the polls close, it is announced that Candidate A leads Candidate B by 51,000 votes to 49,000. The candidates are nearly tied, separated only by 2,000 votes (as in #1 above). The voter registrar’s office announces that there are 3,000 VBMs that need to be counted. There are enough remaining votes to make a difference (as in #2 above). If Candidate B was the choice of all the VBM voters, she would win by 52,000 to 51,000 votes. However, in order squeak out a victory, she would need only 2,501 of the VBM votes to win.
If Candidate B does get 2,501 VBM votes and Candidate A gets 499, then Candidate B wins by 51,501 to 51,499. In this example, Candidate B would have to take more than five-sixths of the VBM ballots (83.33 percent). During the in-person ballot box voting, she only earned 49 percent of the vote (as in #3 above). Therefore, it seems unlikely that Candidate B will be the eventual winner.
Simulated examples from Albany elections
Keeping these concepts in mind, let’s apply them to a hypothetical Albany example. In the 2018 city council election, there were two open seats and three candidates. Here are the three candidates and their vote totals:
Peggy McQuaid 4,716
Rochelle Nason 4,245
Preston Jordan 4,009
Total votes 12,970
Note that this election was not held under at-large RCV rules, but let’s use this as an example of how an at-large RCV election would play out. In an at-large RCV election with three candidates, in addition to making a first-rank vote, voters also would make second- and third-ranked choices. In an at-large RCV election with two open seats, any candidate that gets more than one-third of the vote is automatically elected. In this case, one-third rounded up is 4,324 votes. Peggy McQuaid is elected with 392 over-votes, or votes over the minimum she would have needed. Neither Nason nor Jordan are over the threshold of 4,324 votes. Nason is 79 votes short, and Jordan is 315 votes short.
In the next step, McQuaid’s 392 over-votes must be redistributed to the remaining candidates in proportion to the second-rank choices of the voters who ranked her first. We have no information on those choices, but we can make some reasonable guesses. Nason would need 79 of McQuaid’s 392 over-votes, or 20.15 percent, to have the 4,324 ranked choice votes necessary for election. Jordan would need 315, or 80.36 percent.
Given that the two female candidates were both incumbents, it is reasonable to assume Nason would get more than 20.15 percent and would be declared the second winner. In this example, 1) the leaders were not close to being tied, 2) there were enough ranked ballots to make a difference, and 3) the voting patterns were unlikely to be sufficiently different from the first-round votes. Therefore, the conventional election and an at-large RCV election would have yielded the same results.
Here is another example based on Albany’s 2018 school board elections:
Total votes 17,967
Again, this election was not held under at-large RCV rules, but let’s use this as an example of how an at-large RCV election would play out. In at-large RCV election with five candidates, in addition to making a first-rank vote, voters also would make second- through fifth-ranked choices. In an at-large RCV election with three open seats, any candidate that gets more than one-quarter of the votes is automatically elected. In this case, one-quarter rounded up is 4,492 votes. Hinkley and Duron are elected with 430 and 83 over-votes respectively. Doss is short 1,400 votes, Blanchard is short 1,552 votes, and Stapleton-Gray is short 2,054 votes.
In this election, there was one slate—the Albany Teacher’s Association (ATA) slate of Hinkley, Duron and Doss, and the two white male incumbents. For the sake of simplicity, let’s assume that the over-votes in the ATA slate all stayed within the slate. In other words, all of the ATA slate voters placed Hinkley, Duron and Doss in first- through third-ranked positions and the incumbents in fourth and fifth positions.
If so, when the over-votes are distributed at the end of the first round of voting, Doss gets all of Hinkley’s 430 and all of Duron’s 83 over-votes. That brings Doss’s vote tally to 3,605 votes, still 887 votes short of the threshold level of 4,492.
The candidate with the least number of votes, Stapleton-Gray, is now eliminated, and his votes are redistributed to the second-ranked candidates of the 2,438 voters who voted for him. Again, we have no information on those, but we can make some educated guesses. If Doss gets 887 of the 2,438 transferred votes, (36.38 percent), then he is over the threshold, and he is the third candidate elected.
However, if the voters tended to vote for the incumbents as a slate, and therefore Blanchard gets at least 1,552 of Stapleton-Gray’s transfer votes, (63.66 percent), then that puts Blanchard across the threshold and Blanchard is elected rather than Doss. This example points out two things contrary to misperceptions about RCV. First, slates form and they matter, and RCV doesn’t necessarily favor candidates of color.
Note in this example that 1) two of the candidates were nearly tied, 2) because there were two more candidates than open seats, there were enough ranked ballots after the first round to make a difference, and 3) if the voting in subsequent rounds were sufficiently different than the first round, RCV can matter. If Doss was elected in this scenario, then the conventional and at-large RCV results would have been the same.
An example of when at-large rank choice voting works
Imagine a small agricultural town in the Central Valley. Although the Latino population is in the majority, not all the adults are documented, so they have only 40 percent of the 10,000 voters in town, while the Anglos have 60 percent, or 6,000. Voting is at-large, just like Albany’s current system. There are five seats on the city council to be filled, so each voter can vote for up to five candidates. Both the Anglos and the Latinos run five-candidate slates.
Let’s further assume that Anglos only vote for Anglo candidates, while all Latinos only vote for Latino candidates. All voters use all five of their votes. Within the two groups, all candidates are equally popular. Under these assumptions, it is easy to predict the outcome of the election. All five Anglos are elected with 6,000 votes each, while none of the Latinos are elected because they only got 4,000 votes each. This is known as block voting.
In towns like this, Anglo-only city councils can be sustained for decades. The State of California stepped in to remedy this type of voter suppression with the California Voter Rights Act of 2001 (CVRA), which typically requires such cities to move to district elections. In this case, our little Central Valley city could be split into five voting districts with roughly equal populations. Two might be Latino majority, two Anglo majority, and one mixed. With district-based elections, Latinos can easily win two or more seats on the city council.
But what if Anglo and Latino families all live in mixed neighborhoods? The town could consist of single-family homes occupied by Anglos, with one apartment building on each block occupied by Latino farm workers. In that case, moving to voting districts will not remedy the voter suppression. Since the town is homogeneous, all potential districts would have the same ethnic balance.
In this situation, at-large RCV make sense. Under this model, all voters in town are given ballots with 10 lines to rank-order their choice for the five Latino and five Anglo candidates. The 6,000 Anglo voters rank order the Anglo candidates first through fifth, and the Latino candidates sixth though tenth. The 4,000 Latino voters do the opposite. At the end of the first round, each Anglo candidate has 6,000/5 or 1,200 votes. Each Latino candidate has 4,000/5 or 800 votes. The election threshold under at-large RCV rules is 10,000/6, or 1,667 votes. No candidates win in the first round.
As the ranked votes from last-place candidates are transferred, Latinos win two seats, which require 3,334 of their 4,000 votes. Anglos win three seats with 5,001 transferred votes out of their total of 6,000 votes. Note that each group is represented proportionately. Latinos have 40 percent of the voters and 40 percent of the five seats, while Anglos have 60 percent of the voters and 60 percent of the five seats.
This proportional outcome is the story RCV supporters like to tell. However, in real life, there are not just two interest groups. They can be a mix of Anglo and Latino, male and female, low-income and high-income, gay and straight, meat-eating and vegetarian, and so forth. When voter identities are more complicated, it is not clear how well RCV does to proportionately represent all the possible groupings, especially if, as in Albany, there are typically only two or three seats open in each election.
The at-large ranked choice voting record of Cambridge, MA
Cambridge is the city across the Charles River from Boston. It is the home of Harvard University and the Massachusetts Institute of Technology. Cambridge is one of four cities in the United States that uses at-large RCV voting. The city has used at-large RCV continuously since 1941. The more recent election results are online. I found the City of Cambridge’s RCV voting records for their School Committee and their City Council for the five elections that occurred in the odd years from 2011-2019.
Let me discuss the School Committee first. There were six members elected in every odd year in the five elections from 2011-2019. The number of candidates ranged from a minimum of 9 and a maximum of 12. I listed the top six vote-getters in Round 1 of the voting and compared that list to the list of the six elected winners at the end of the at-large ranked-choice voting process. Here is what I found:
In four of the elections, those occurring in 2011, 2013, 2015 and 2019, all six of the Round 1 winners were eventually elected. The at-large RCV process didn’t change the outcome. In 2015, the sixth- and seventh-placed candidates at the end of Round 1 were separated by only 12 votes, and in the at-large RCV process, the seventh-placed candidate acquired a few more votes than the sixth-place candidate and was elected instead.
Now for the Cambridge City Council elections: Using at-large RCV, Cambridge elects nine members to its city council in odd years. I found records for the same time period as the school committee elections, 2011-2019. The total number of candidates in each election varied from 18 to 26. In three of the elections, 2011, 2013 and 2017, all nine of the top vote earners in Round 1 were elected to the council. The at-large RCV process made no difference.
In 2015, the 11th-placed Round 1 candidate replaced the ninth-placed candidate during the at large-RCV process. In the 2019 election, the 10th-placed Round 1 candidate prevailed over the seventh-placed Round 1 candidate to earn a seat on the city council. The at-large RCV process made a minor difference in two elections.
Note that even though there was a considerable depth of candidates to rank order, it seldom mattered. Also note that with nine open city council seats, any candidate who accumulates more than 10 percent of the votes is elected. Unlike with single-seat instant runoff voting, where one candidate eventually achieves a majority, with nine open seats, no candidate is required to get anywhere near a majority to be elected.
Also notice how different the reality of at-large RCV in Cambridge is when compared to our hypothetical model of a Central Valley farm town. In our farm town example, the ranked-vote transfer process radically altered the eventual winners. In real life in Cambridge, the differences were minor to non-existent.
A brief note on Eastpointe, Michigan
Eastpointe, Michigan, is a city of 32,500 in the Detroit area. According the latest census data, it is now 49 percent African-American. According to the fairvote.org website, Eastpointe is the only jurisdiction other than Cambridge that exclusively uses at-large RCV. This is not quite accurate. As part of a settlement with the federal Dept. of Justice, Eastpointe agreed to to switch to at-large RCV for electing its four city council members. Unlike Albany, Eastpointe has a directly elected mayor, one who is not elected using RCV.
The first election under the new rules took place in November 2019. There were four candidates for two seats on the city council, elected by at-large RCV. The two winners were a white female city council incumbent and a white male. The two candidates who did not win were both African-American, one male, and one female. The conventionally elected mayor was also a city council member who became the first African-American female mayor of the city (more here and here).
We have to keep in mind that this is the first at-large RCV election in Eastpointe. However, it is worth noting that this result is far from what the advocates have advertised about what to expect from an at-large RCV election. It was the conventional election that elected an African-American mayor, and the at-large RCV election that chose two white candidates over two African-Americans.
Closing thoughts about RCV
The advocates of RCV make the assumption that ranked choices provide more information in an election, and that more information is good. But there is another possibility. Perhaps beyond their second- or third-ranked choice, the voters do not research their choices and are confused about the attributes of their lower-ranked candidates. If so, it’s likely they just guess, or fill in their lower-ranked choices randomly.
If so, the RCV process incorporates some information along with a lot of noise. That would explain why RCV elections, either the instant runoff or at-large versions, seldom overturn the first-round winners. For all its technical sophistication, RCV cannot overcome the simple human problem of voters who lack either the resources or the enthusiasm to carefully study all the candidates.
In addition, the notion that every vote counts in RCV is generally not true. In single-seat RCV (instant runoff voting), if one candidate gets more than 50 percent of the vote, the election is over, just like in a conventional election. In at-large RCV with two open seats, if two candidates get more than one-third of the votes each, the election is over, just like in a conventional election. In an at-large RCV election with three empty seats, if three candidates get more than one-quarter of the votes each, the election is over, just like a conventional election. If I didn’t vote for one of the winners in the examples above, none of my votes counted.
At-large RCV creates incentives for candidates to spend more money. In the 2016 council election I joked that my plan was to spend as little money as possible and still come in third. That is exactly what I achieved. If that election was held under at-large RCV rules, I would have had much more incentive to spend more to try to clear the safe-harbor hurdle of getting one-quarter of the votes, which would have protected me from any vote-redistribution surprises. The amount of money candidates need to raise to compete for a volunteer elected position in Albany is a major reason more residents don’t run. RCV won’t help this problem, and may make it worse.
As in the past, this November we are having barely competitive elections for both the city council and school board. For the city council, there are four candidates running for three seats. For the school board, there are three candidates running for two seats. In elections with only one more candidate than open seats, at-large RCV elections are unlikely to yield a different result than conventional elections. That’s because only over-votes of the leading candidates can make a difference, and this is unlikely to happen because there are usually relatively few over-votes, and they tend to be distributed like the first-round votes.
What do we make of at-large RCV? At least for Albany, I think it is much ado about very little. It does little or nothing to encourage citizen participation in local government. And it does little or nothing to change the outcome of elections. It is a Rube Goldberg machine. Perhaps that explains with so few public jurisdictions in the United States use it.
The advocates for RCV in Albany are asking the citizens to become human guinea pigs in an experiment that isn’t that useful to start with. In addition, Albany residents will have to pay $26,000 per election for the privilege of participating in the experiment. I just don’t see any good reasons to do this, so my advice to Albany voters is to vote no.