What’s the Deal with Ranked-Choice Voting in the Bay

The cities of Oakland, San Francisco, Berkeley, and San Leandro all use the new method of Ranked-Choice Voting (RCV) to elect most of their city officials. But in my experience, even if you live in one of those places, you likely have only a rough handle of how that specific system works. Which is a shame, because it’s a distinct improvement over most existing United States methods and could be a key part of an improved future structure. In this blog post I’m going to relate in simple terms how it works, what it looks like in practice, and how to make your vote count the most in these elections – especially the upcoming one for San Francisco Mayor. I’ll wrap up with some thoughts about how it might be improved.

1. How it works

Put in words, RCV in California works like this:
  • Everyone can vote for up to three candidates in their order of preference: first choice, second choice, third choice. (I’m going to abbreviate these votes as #1, #2, and #3.
  • We start by counting up the #1s. If one candidate gets a majority (50% plus one), they win and there’s no need to go any further.
  • If nobody has 50%, you look to the candidate with the fewest #1s and eliminate that candidate. Take all the people who picked that candidate #1 and change their votes to count for their #2s.
  • If one of the candidates now has a majority after reallocating, stop: they’ve won.
  • If not: repeat. Keep eliminating candidates, from lowest vote total to highest, until you have someone with more than 50% of the vote (diving into the #3s if necessary). 

Or to put it in pictures, here’s a simple illustrative video:
Where do we use this system now? In all the four cities I listed, for Mayor and City Council (or SF’s Board of Supervisors equivalent); and, in Oakland only, for School Board.
Other races in California use different systems.
For State Legislature, US Congress, and statewide races like Governor, it’s the top-two, or“jungle” primary; for county races like District Attorney and Supervisor, it’s yet another system; and then there are judgeships and a wide variety of miscellaneous offices.
For those in San Francisco, the special mayoral election on June 5th will use the RCV system; but all the four cities will have many more races under that system this November.
You see in theory how the system would combine everyone’s preferences more reasonably. If candidate A got 40% of the vote, B got 35%, and C got 25%, but all of the C supporters preferred B over A, it would be the right result to elect B over A.

(I personally think this is a major hidden contributor to US dysfunctionality, that in most of the country the top vote-winner, like candidate A, wins outright. It’s a system that elevates the loudest voice in the room. Did Trump really “win” states where he got 48% of their vote? But that’s a rant for another time.)
2. What we see
So that’s the theory. What about the practice? To hear some describe it, RCV has meant disaster, or at least disastrous confusion.
I’ve dived into the data and found that, first of all, in most local elections, it doesn’t make a big difference either way – or rather, can’t. Over seven years, 2010-17, out of 99 races, 64 either were unopposed, had only two candidates running, or had more candidates but one received 50%. In 28 of the remaining 35, the initial #1 vote-winner still won after reallocating all the votes. After all, usually, the most popular pick really is most popular, because the rest of the public isn’t mad at them.
But the exceptions are important.
Many people were surprised at Jean Quan’s squeaker victory for Mayor of Oakland in 2010, which would not have happened in another system. She, of course, had a lackluster one-term mayoralty but that’s pure hindsight. Based on what the electorate knew at the time, she was a reasonable choice, especially compared to Don Perata, the frontrunner and #1 winner, who very much represented politics as usual. Here is a rough animation I made of how she accumulated the votes:

The “margin” tracker at the bottom of this animation shows how Quan started well behind Perata, but the gap shrunk almost every time votes were reallocated, indicating she kept getting more #2 or #3 support. Importantly, the Tuman #1 voters preferred Quan over Perata by a 39% to 22% margin; among Kaplan #1 voters, it was a wider and decisive margin of 62% to 20%. We saw this in how the campaign was run: Quan, Tuman, and Kaplan spoke well of each other, suggesting supporters rank the others after them – implicitly against Perata as a common rival, elevating the similarities in coalition rather than the divisions in their majority.
As I see it, whatever else happened in the election, the voice of the people was heard like never before. Without RCV, the clashing cacophony of voices would have left the status quo the strongest of them all. With RCV, it turned into a new kind of harmony.
To be fair, there have also been RCV elections with less clear results, especially when the number of candidates gets bewildering. In SF District 10, in 2010, there were a whopping 22 candidates, and the top three got literally 12% each. In the end, Malia Cohen, who had originally placed third, won, but a large number of ballots couldn’t count at all because so many people voted for three candidates none of whom were in the top three. (In this case their ballots were “exhausted”.)
There are technical solutions to the problem of exhausting ballots, but they come at the expense of people voting, of their time and energy. Theoretically, if everyone ranked every candidate running, no matter how many there were, no ballot would ever get exhausted. But even if the balloting system allowed this, who would bother filling in so many choices?
But having as many candidates as 2010 D10 is is rare, and even then it doesn’t assure confusion. In 2014, there were 16 running for Oakland Mayor, but frontrunner Libby Schaaf’s victory was much more obvious and well-determined. For the more minor cases, it should go a long way to expand the space from three to six ranks.
I also see, looking at the history of RCV, that being unpopular doesn’t eliminate the advantage of having the most #1 votes. You have to be pretty broadly unpopular to lose despite that initial advantage. Desley Brooks in Oakland D6 was a fairly weak incumbent in her reelection race in 2014, and all three challengers had cross-support, the kind of voting behavior that can make it easier for RCV to turn out incumbents. But Brooks’s initial edge was still too strong to overcome: she started with 42% of the initial vote to her competitor Nosakhare’s 30%, and while she lost margin with every reallocation, she still won in the end, in another squeaker, 52% to 48%. The initial advantage also meant she needed less help to make it over the top.
When I look at the full range of outcomes RCV has had, I think on balance it most often works the way it was designed.
How does RCV compare to the top-two system the rest of the state is trying this decade? To be fair, top-two can let multiple candidates with majority support to defeat another candidate with plurality support. In my example 40-35-25 race between A, B, and C, top-two would send just A and B to a runoff election in which C supporters could switch their support to B, and win 60-40.
But RCV has big advantages over top-two. First, it eliminates the need for that initial primary election that selects two candidates. A single election requires less fundraising, giving more space to insurgent or grassroots candidates. Under top-two, after knocking out C, A might have ground down B in a long, expensive campaign; in RCV, this is harder.
Second, RCV results are more logical in how they combine voter preferences, while top-two more often has perverse outcomes in multi-candidate races. If you have 4 candidates each getting almost exactly 25% of the vote, it can be practically random which two make the top cut; you can get two from the same party even when the parties are relatively evenly matched. This is a risk congressional Democrats are extra-worried about this year. It was the same kind of perversity in France in 2002 where, using the same system for president, only 17% of the first-round vote was enough to send the racist Le Pen to the second round because the left ran too many candidates. That kind of outcome is not rare at all in top-two. Why keep subjecting ourselves to that?
3. How to make your vote count the most
Now, some practical tips.
If you want to have the most possible say over who wins, your best bet is to rank all three choices with different people. That gives you the greatest chance at helping someone you want to win – or helping avoid someone you don’t want.
Don’t “bullet-vote”, or mark the same person three times. It doesn’t give super-support to the candidate. It works out just as if you marked that person one time and no more.
Mark the ballot correctly: the same list of candidates will appear three times, once for each choice you have. You need to mark one choice on each list, not three choices each on the same list.
If you have a wider field with five or more candidates, do some research and try to find which candidates are more prominent and likely to win. Try to find at least one acceptable person who is likely to make it to the final rounds. If you find three longshots – which is not uncommon around here! – voting for all three of them will likely get your ballot nowhere. Two longshots? Sure, if you like, but maybe save your #3 slot for someone with a chance.
4. What can be improved
Although I think RCV is much better than alternatives currently in political conversation, it’s clear there’s room for improvement. Right now it’s normal for only 60-70% of voters to rank two or three distinct choices: some spoil their ballots, and some vote for just one candidate. I have wondered if this might effectively disenfranchise more disadvantaged voters, because they have less time and leisure to figure everything out; I have data analysis in the works on this subject, but so far, the evidence is far from clear-cut.
But even if it doesn’t harm disadvantaged voters disproportionately, the currently low level of effective voting is a problem.
Certainly, voter education is important. Public organs and candidates need to explain better how to make your vote count (amazingly, some candidates reputedly spread the word to rank them and only them); ballot machines could also check for errors and let people confirm that what they’ve put in is how they want to vote. But there’s a limit to how far education can go. How many people have the time or, frankly, the interest to figure out how they feel about every single candidate?
We might consider more active measures to prevent a bewildering number of candidates from running; for example, signature requirements could rise selectively, such as, only if the number of candidates exceeded six. But this kind of countermeasure also shuts newcomers out.
My big crazy reform idea is delegated choices.Imagine if candidate X could certify before the election: “If I’m eliminated, I want my support to go first to Y, then to Z.” Then voters can still rank-order any three candidates like always, but if all they’ve decided is they like X, they can check a box saying “distribute the rest of my votes the way X wants“. Similar to the old straight-party-ticket option, but more granular. Other organizations, like parties or coalitions, could also put their slates as options on the ballot.
In June, San Francisco will have an open mayoral race for the first time since it started RCV. Hopefully, coalitions will form in productive and instructive ways, and we will further build on the potential for RCV to bring people together.

Leave a Reply

Your email address will not be published. Required fields are marked *