Ranked Choice Voting: Wrong for Seattle
Seattle voters have a chance to adopt one of two alternative voting methods this November 2022.
- Approval voting: Vote for as many candidates as you like (no ranking). That’s it. The two candidates with the most votes advance to the general, just like now. The ballot looks the same. The counting is the same. You just replace the one-vote limit with “vote for one, or more”. Approval voting is used in St Louis and Fargo.
- Bottoms-up voting: This is a novel form of ranked voting, never before used in the USA. Voters can rank up to five choices, and the tabulation is conducted by eliminating the candidate with the fewest votes until only two candidates remain.
Lots of mathematical analysis and historical evidence says that approval voting is better at choosing the winner who most satisfies the will of voters, especially when taking strategic “gaming the system” into account. For instance, see these voter satisfaction efficiency measures from a Harvard stats PhD and former board member with the Center for Election Science.
The red and brown dots for approval voting are almost at 96%, which far exceeds the performance of ranked choice voting (“RCV”) at 80%-85% depending on the amount of strategic voting.
But this isn’t even “ranked choice voting”
Proponents of bottoms-up refer to it as “ranked choice voting”, which has been used in dozens of US cities, including San Francisco and New York City. But while the Seattle version uses a ranked ballot, the method of determining the two winners is virtually unheard of, and never before used in the USA.
Bottoms-up does not attempt to find the two most popular candidates, but instead attempts to find diversity. The ranked voting advocacy organization FairVote describes it here as a “semi-proportional” method.
Diversity is a worthy goal if you’re actually electing multiple candidates to sit on a council, but can actually harm minorities if you’re ultimately going to choose a single winner, as Seattle does via its top-two general election process, mandated by state law.
Here’s a simplified hypothetical election example to demonstrate that.
35% Bruce > Jessyn
33% Lorena > Jessyn
32% Jessyn > Bruce
With these preferences, Bruce and Jessyn are far and away the most popular candidates, each being preferred to Lorena by a crushing 67% landslide majority. But the bottoms-up method eliminates Jessyn with only 32% of the first-place vote, and advances Bruce and Lorena to the general. And not only is Jessyn preferred to Lorena by 67% of voters, she’s preferred to Bruce too, by 65% of voters. Jessyn is by far the most popular candidate overall, preferred to each rival by a two-thirds majority; and yet Jessyn is the first eliminated.
Why does this happen? Because, by design, RCV doesn’t look at the 2nd choice preferences of Bruce voters, and thus it doesn’t accurately measure the overall preference of the electorate. And the announced totals won’t even reveal the hidden support for Jessyn; voters will just see that she’s eliminated with 32%. Voters feel like they’re casting this precise statement of their ranked preferences, but then The Algorithm™ selectively ignores some of that ranked preference detail.
More precisely, the lower a voter ranks Bruce, the more influence that voter has over who challenges him in the general. This means that the challenger will, by definition, be a statistically weaker opponent, and thus more likely to lose the general by a big margin.
For the progressives in this example, it might feel like a victory to see Lorena make the general. But it’s a fleeting joy, as Bruce crushes her with a 62% majority. Whereas Jessyn might seem a less ideal choice, but she would at least give progressives a meaningful chance to tip the balance of power ever so slightly in their direction. And a competitive election would give both candidates an incentive to fight for every last vote.
A real life example
Washington voters just saw this vote splitting dynamic play out in real life, as Washington’s third congressional districts results were as follows.
Marie Perez Democrat 31.0%
Joe Kent (Trumpist) Republican 22.8%
Jaime Herrera Beutler (anti-Trump) Republican 22.3%
Republican incumbent Jaime Herrera Beutler, who voted to impeach Trump, was eliminated due to vote splitting with Trump-backed Republican Joe Kent. Now Kent will likely go on to defeat Democrat Marie Perez. This is exactly the same outcome we would get from the bottoms-up ranked voting system proposed for Seattle.
Approval voting solves this
With approval voting, “2nd choice votes” count the same as 1st choice votes; a vote for your 2nd choice counts the same as a vote for your 1st. You just vote for as many candidates as you want to, no ranking. Many ranked choice voting advocates see this as a flaw, but in these examples we see how great a benefit it can actually be.
In our theoretical election, many Bruce supporters could and would also approve Jessyn as a 2nd choice compromise, and that expression of preference wouldn’t go ignored. Thus Jessyn would be much more likely to advance and give Bruce a run for his money.
And this would have also played out in Washington’s third district. Democrats, fully aware of their odds of losing to the Trumpist, would in large numbers make the pragmatic choice to cast a back-up vote for never-Trump Republican Jaime Herrera Beutler. Thus they could at least work to prevent their worst outcome.
In this race, Beutler was the compromise/consensus candidate, and ranked choice voting would have eliminated her. This is exactly the kind of dynamic we want to avoid. Approval voting is the simple pragmatic system to find consensus and avoid uncompetitive elections.