Ranked Choice Voting in Kalamazoo: As Easy as 1–2–3…

Voters in Kalamazoo, Michigan, should vote YES on the Ranked Choice Voting Proposal.

In the City of Kalamazoo, tomorrow’s election includes a ballot measure for whether the City of Kalamazoo should adopt a ranked-choice voting system, if permitted by state law. I urge you to vote YES on this measure. I’ll explain why, what it would do, and the nitty-gritty of how it would work in this post to convince you to vote YES on Ranked Choice Voting.

Before I get into the argument, I want to explain my background on this issue. I’ve spent several years working as a law of democracy lawyer, promoting good governance and election reforms. I’ve also taught seminars on election law as both a visiting assistant professor and as an adjunct professor at law schools. I’ve published academic articles and white papers on related topics. Don’t rely on my advice because I assert that I have expertise in this area — rely on my arguments below — but I am in fact an expert.

In order to understand why ranked choice voting is superior to our current system, we should begin by examining the flaws of the current approach. Currently, our city elections run using what’s called “first past the post” or plurality voting. What that means is that each voter casts up to as many votes as positions to be filled (one for mayor, three for City Commissioner). Those votes are tallied up, and then the highest single vote getter (for mayor) or highest 3 vote getters (for City Commissioner) are elected. This system’s virtues are familiarity and simplicity. Unfortunately, it produces bad results and requires a significant amount of “strategic voting,” where voters have to guess who other voters are going to vote for in order to cast their votes in a way that make them effective. Because it’s easier to follow, I’ll start with a discussion about the problems and how ranked choice voting reduces them with regard to mayoral elections first, although it’s actually even more valuable for multi-member elections like City Commissioner.

In order to determine who to vote for under first past the post systems, voters need to guess how other voters will vote to avoid wasting their votes. Also, this system can often create perverse outcomes where the results don’t match the preferences of the voters. Let’s take a simple mayoral election, with 3 candidates for mayor: Larry Liberal, Mary Moderate, and Chris Conservative. Because city elections are nonpartisan in Kalamazoo, none of them are the nominees of any party, although people who are in the know understand that MM and LL are Democrats and CC is a Republican. Now consider the choices of three voters: Alice is a liberal Democrat and prefers LL to MM to CC. Barry is a moderate Democrat and prefers MM to LL to CC. Connie is a Republican and prefers CC to MM to LL. Alice wants to vote for LL if LL has a chance of winning, but if LL will be a distant third, she wants to vote for MM to make sure CC doesn’t win. Barry’s preferences are reversed — he wants to vote for MM if MM can win, but if he concludes MM can’t win, he needs to vote for LL to prevent CC from being elected. And of course Connie wants to vote for CC, but if the real race is between MM and LL, she should vote for MM instead. In order to choose who to vote for, voters need to guess (and in city elections, without polling beforehand, it really is a guess) how many votes each of the candidates will get in order to vote intelligently.

What’s worse, if they guess wrong, they can cause exactly the outcome they least want. In this example, it’s pretty clear that either LL or MM should be elected (if Alice, Barry, and Connie are representative of the voters in the city as a whole). Let’s assume that 30% of the voters are like Alice, 30% are like Barry, and 40% are like Connie. If Alice and Barry both think that their candidate is most likely to win overall (or has a good shot), and everyone votes for their top pick, CC wins — even though 60% of the voters prefer either of the other two candidates, CC gets a plurality. Conversely, if Alice decides to vote strategically for MM, but it ends up being a close race between LL and MM with CC in a distant third, Alice could make a reasonable strategic decision that has the direct effect of making her favorite candidate lose.

Ranked choice voting mostly solves this. In ranked choice voting for a single office, also known as “instant run-off voting,” candidates need to get an actual majority of the votes. Voters rank their preferences: Alice ranks LL 1, MM 2, and CC 3 (or equivalently stops ranking after 2), and so forth. If any candidate has a majority of the 1st place votes initially (maybe MM is a very popular incumbent), they just win. If nobody has a majority, the lowest vote getter is eliminated. Anyone who listed them as their 1st place vote is now treated as if they listed their number 2 candidate as their 1st place vote. Then the votes are tabulated again.

It’s easy to see how this largely solves the problem of strategic voting and perverse results in my simple example with 30% Alice, 30% Barry, and 40% Connie. Nobody has a majority initially, so the lowest vote-getter is eliminated. Let’s say that is MM, modifying the numbers to 31% Alice, 29% Barry, and 40% Connie. Now, the people like Barry’s votes are treated as if he had voted for their second place candidate as a first place candidate: their votes are treated as votes for LL, their second choice. Now, in the second round, Alice has 60% and Connie has 40%, and Alice wins. This is a reasonable result: 60% of the voters prefer one of the two Democrats, and Alice has more support among those voters. Alice and Barry both voted for their preferred candidates in their actual order of preference.

Ranked choice voting isn’t perfect — the Connie voters can actually get their second-best result (instead of their least preferred outcome) by voting for MM instead of CC in first place. If either CC or LL is eliminated first, Connie gets her second-best preference, so there is still some marginal advantage to strategic voting. In fact, Arrow’s Impossibility Theorem proves, in rough terms, that no possible voting system can eliminate all advantages to strategic voting. Note, of course, that in this example MM is also a reasonable person for the city to elect: 69% of the voters prefer MM to LL. It’s arguable that MM is a better result, in fact, in terms of representing the preferences of the city than electing LL — so there’s still some risk of outcomes that some people would consider poor representations of the voters’ preferences under ranked choice voting. But the results are much better on average and the incentives for strategic voting are greatly reduced with ranked choice voting.

Ranked choice voting works even better when applied to multimember elections, like the City Commission elections. In those cases, strategic voting is even more important, because out of a list of perhaps 10 candidates, three will be elected. Should you vote for your favorite candidate? Or perhaps the candidates that you like the most out of the ones you consider likely to be elected. Maybe you shouldn’t vote for your favorite candidate at all if you’re sure they will be elected, because then you have more votes to make sure the other candidates you like are among the people elected? And even besides the problems with strategic voting, there’s a problem with lack of representation.

Ideally, a body like the City Commission would represent the full diversity of the community it represents, while still allowing the majority of the city’s voters to elect a majority of the Commission. While a sizeable majority of the population of Kalamazoo are Democrats, it’s still useful to have some Republican representation on the City Commission — if for no other reason than to provide constituent services and a sense of being represented to the Republican voters. Also, it’s useful to have a City Commission that represents the wide range of preferences on other issues that don’t break down as partisan — such as for example zoning and traffic planning. Demographic representation also has a value — promoting a diverse range of different perspectives and life experiences on the Commission.

The current system tends to not produce either diversity of viewpoints or diversity of representation on City Commissions. If 50%+1 of the voters prefer a given type of candidate, they can win all 3 seats every election (provided they can overcome the strategic voting and coordination problems).

Ranked choice voting (in this case, “single transferrable vote” or “preference voting”) can solve that problem by using the rankings to more or less fairly allocate the seats across the preferences of all of the voters, with minimal advantage to strategic voting. If two-thirds of the voters vote for candidates A and B (in some order) as their first two choices, and one third vote for candidate C as their top choice, candidates A, B, and C will all be elected. The system thus produces proportional results while being simple from the perspective of the voters, who just have to rank the candidates in a preference order. The system is fairly complicated in the tabulation process, but from the voter’s perspective, it’s straight forward.

For people who want the nitty-gritty details, here’s how it works. First, there’s the concept of the threshold of election. That’s defined as a number of votes equal to the total number of votes divided by one more than the number of seats to be elected. Any candidate who has more than the threshold of election is entitled to a seat, because there can’t be more other candidates than there are seats with more votes. In the case of the commission, any candidate who gets more than a quarter of the votes can have at most two other candidates with equal or more votes, and so is definitively in the top 3 and gets elected. Second, there’s a concept of fraction of votes consumed to elect a candidate. Imagine 75% of the voters vote for Smith as their #1 choice. Obviously, Smith is elected — 75% is way more than 25%. But it wouldn’t be fair for the other two commissioners to be chosen only by the remaining 25%, precisely because 75% is way more than 25% — if the 75% all prefer Smith, then Jones, then Brown, and the 25% prefer Lee, then Black, then Green, it’s intuitively clear that the result should be Smith, Jones, and Lee (or maybe Smith, Jones, and Brown, depending on the exact numbers). So what we say is that since Smith got three times the threshold of election, only 1/3 of the vote of each person who voted for Smith was consumed by electing Smith. Two-thirds of each of their votes can then be transferred to their next preference.

With those two concepts, we can understand how the system works. First, you look at all the first place votes. If anyone has more than the threshold of election, they’re elected, and the portion of the votes for them that were not consumed to elect them are transferred to their second-ranked votes. (These can be to different people. If 1/3 of the votes for Smith were consumed electing her, than each voter who votes for Smith gets to transfer 2/3 of a vote to their second ranked candidate individually. This is all taken care of by the tabulation process, but it’s like each ballot is reduced to a fractional vote when it’s top candidate is elected, and then recast for the next candidate.). We then check again whether anyone is now over the threshold of election. If nobody is, then we move on to the next step: dropping the lowest vote-getting candidate. We reallocate all of their votes to the next person in the ranking order on each of their ballots. (If some of their votes are fractional votes, they remain at the same fraction as they’re transferred on — because this is an eliminated loser, none of the vote was consumed by their elimination.). We then go back to the first step and check if anyone now has more than the threshold of election. The process continues, electing candidates over the threshold and reallocating the remaining fraction of their votes and eliminating the bottom ranked candidate, until every slot is filled. But again, from the voter’s perspective, they just rank the candidates and then automagically the tabulation process produces a fair, proportional representation on the commission.

As with ranked choice voting for single seat offices, ranked choice voting for multiple seats has occasional rare situations where strategic voting would produce a better result for some voters than simply voting in their preference order. However, it’s rare and not a significant problem, even though it’s theoretically imperfect.

One last detail: currently, state law does not permit cities to use ranked choice voting. So if this proposal passes, it won’t have any concrete effect unless and until the state legislature authorizes it. However, it’s still useful to be part of the movement pushing for ranked choice voting, and if the legislature does pass the authorizing legislation — which is more likely as more cities adopt these proposals — it would then improve how our elections operate.

I know this turned into an enormous block of text, so TL;DR: Vote YES on the ranked choice voting proposal. It would make our elections better.