“Ranked Choice Voting”? I support it, but…
The best-known proposal for voting reform is flawed. Fortunately, it’s fixable. Can we keep increasing momentum and also fix it?
Politics is broken. In the US; in the UK; and probably even in Canada.
Voting reform can help fix it. These are the main countries still using first past the post voting almost exclusively. (I could also have included India, which uses FPTP for many but not all elections.) And that voting method is definitely at the root of the political problems. It systematically leads to wasted votes and unrepresented groups. Gerrymandering, spoiled elections, polarization, zero-sum politics of spite, narrow bipolar debate, money-driven politics; FPTP makes all of them worse.
So, we need to fix that. And we can! These days, voting theorists know a lot about voting methods that avoid those problems. Proportional representation methods can represent almost everybody, with only a tiny sliver of wasted votes. For single-winner elections, there are methods which promote honesty and compromise and growth rather than strategy and intransigence and stasis-punctuated-by-whiplash.
In fact, in an important sense, the problem isn’t too few solutions, but too many. It doesn’t take much to think up a new voting method. Designing one that’s actually a good idea is a little bit harder, but still possible, even for just a dedicated amateur. If each activist has their own homemade solution, nothing is going to get done.
So we need a way to agree on a common strategy, so that perfectionism doesn’t get in the way of practical progress. Luckily, voting theory is rooted in multiplayer game theory, which has some ideas about how to coordinate. One way would be to have a coordination mechanism imposed from outside—like in 2004, when British Columbia had a citizens’ assembly on voting reform, and over 100 randomly-chosen voters representing the entire province met for almost a year to decide on the best system for the province (they settled on a form of STV, a system discussed below). There are a number of reasons that answer was specific to that time and place, but certainly if this kind of assembly/deliberative poll/focus group were repeated in the US, it could do a lot to get activists working together. I, for one, would commit to following the conclusions of a group like that, if it were run in an unbiased way.
Failing that, the way to coordinate is to have everyone actively seek a “Schelling point” — basically, the simplest and/or best-known idea that would actually work for enough people to form a critical mass.
But what does it mean to “actually work”? In principle, voting reform could help solve all the varied issues I listed in paragraph 2 above, and more. In practice, though all of those issues are important, the ones that could build a critical mass of reformers are the first two: gerrymandering and spoiled elections. These two problems affect different kinds of elections, so the solutions are likely to be different voting methods.
Gerrymandering affects multi-winner elections—for things like Congress, state legislatures, city councils, and school boards. In FPTP, these are run by dividing the voters into single-seat districts and having a separate election in each. When you do that, the dividing lines matter a lot. If politicians control the line-drawing, they often deliberately use that power to advantage themselves and/or their party. But even if the lines are drawn neutrally, a substantial fraction of the votes (sometimes more than half) don’t help elect anyone, so those voters have no accountable representative.
The “spoiler effect” happens in single-winner elections. Because FPTP forces voters to pick just one candidate, two similar candidates in the same race can “split the vote” and lose — even though either one of them would have won easily without the other in the race. In other words, they spoil the election for each other. To avoid this, voters often learn to ignore all but two candidates in each election, leading to a two-party monopoly on power; that tendency is called “Duverger’s Law”. The two-party monopoly, in turn, puts perverse and destructive incentives onto candidates, parties, and the media; and thus encourages polarization, spite, corruption, stale ideas, and policy whiplash.
Since single-winner and multi-winner elections are different situations, they need different solutions. If those solutions were related, that would make each of them a better Schelling point—but only if both of them work.
Two of the best-known voting reform proposals for these two situations are related. I’m talking about the proposals that have recently become known as RCV, ranked choice voting. For single-winner elections, that means the thing we Americans used to call IRV, instant runoff voting (Brits called it the Alternative Vote, while Australians call it Preferential Voting). For multi-winner elections, it means STV, the single transferable vote.
In both of these systems, voters rank their choices in order of preference, and the vote-tallying procedure involves eliminating the weakest candidates and transferring votes until the winner(s) have enough. In IRV, “enough” means over 50% of the remaining votes; in STV, it means a quota, defined to be just large enough that if each winner has one quota of votes the leftover votes are less than a quota.
So do these two proposals solve their respective problems?
In the case of IRV, the single-winner form of RCV, the answer is basically yes. As a voting theorist, I kinda hate to admit that, because IRV’s theoretical defects (primarily nonsummability, nonmonotonicity, and center squeeze) make it just about the weakest reform proposal. But does it solve the spoiler problem? Pretty much. In the most famous spoiled election — the Florida presidential election of 2000—IRV would have allowed Nader’s votes to transfer, largely to Gore, ensuring a true majority rather than a spoiled election.
So is IRV a Schelling point for reformers? Well, a few reformers see its defects as a deal-breaker; but, as the victory of Question 5 in Maine 2016 attests, the lack of those reformers isn’t enough to slow it’s momentum. I strongly believe that approval voting is both simpler and better than IRV, and I’m actively working to help approval voting reach critical mass starting in Fargo, North Dakota this year. And if I got to pick my ideal voting method, it would be 3–2–1, which is no more complex than IRV. But meanwhile, much as I’m reluctant to admit it, IRV has enough momentum to make it the main Schelling point.
That means that, unlike some of my friends, I think that anyone who supports US voting reform should support IRV, even if they prefer something else.
Given that, it would be nice if STV — the multi-winner form of RCV—were also a Schelling point. And in some senses, it seems that it should be. It’s been on the agenda in US politics for over 100 years; it’s used in 3 US cities (Cambridge and Amherst in MA, and the Minneapolis parks commission) and widely in Australia, Ireland, and Malta; and the “RCV” branding helps it benefit from IRV’s successes. But… it’s not going to actually solve the problem of gerrymandering.
Not that it wouldn’t work if it were implemented. As a proportional representation (#PropRep) voting method, it ensures that only a tiny fraction of votes are wasted; so, no matter where the lines are drawn, all other voters would be meaningfully represented.
But to solve gerrymandering, a voting method has to have some chance of passing at the national level. Individual state legislatures have no incentive to fix the very gerrymandering by which they took power, and in many of the most-gerrymandered states voter initiatives are not a viable solution. But at the congressional level, gerrymandering benefits one of the two parties and hurts the other; if the party that’s hurt by it gets power, it is very much in their interest to solve gerrymandering, especially because they could legitimately impose that solution only on states where it’s the biggest problem.
So to be viable, a solution to gerrymandering should be something that incumbent congresspeople from one of the major parties could embrace en mass. But STV is too much of a disruptive change for that. By putting incumbents into newly-created multimember districts, and creating a complex new ballot format, it could very well mean that a lot of incumbents lose their seats, even if their party as a whole gained seats. Furthermore, it would create incentives to fragment large big-tent parties into smaller single-issue ones, which would especially be a problem for a demographically heterogeneous party. That means that if it ever got anywhere close to passing Congress, insiders from both major parties would be sure to kneecap it.
So STV isn’t a viable solution to gerrymandering, and thus not a Schelling point. I’m free to point out its other flaws:
- Complex ballots. I live in Cambridge, where we actually use STV, and in the last election I had a 26x26 array of bubbles to fill in, one bubble per column, in order to mark my preferences. Carefully weighing the advantages and disadvantages of 26 candidates is not easy, and lazy voters who quit after 1 or 2 or 3 run a serious risk of their vote not counting. (Mine didn’t count until my 5th choice. My 2nd and 4th choices were elected, and my 3rd was eliminated, while my vote remained on my 1st choice; so when my 1st was eliminated, my vote went straight to #5. Simple, right?)
- Impractical for more than about 5 seats at a time. This ends up narrowing voter choice.
- Requires centralized ballot-counting, which reduces security and auditability.
- The chain of accountability is opaque. In the Cambridge election I mention above, if I weren’t a voting geek, I’d think that the representative accountable to me was my 2nd choice, because they were the winner I ranked highest. But because of the order of elimination and elections, the representative who is actually most accountable to me is my 5th choice. Understanding that requires reading impenetrable tables of numbers.
- Compromise candidates can be prematurely eliminated, and coalitions can break down into single-issue individuals.
Luckily, there is a form of #PropRep that solves all of these problems: PLACE voting (FAQ here). For nonpartisan elections, that would be PAD voting. Both of these are systems I designed with the above points in mind, so the movements behind them are still far from critical mass. But they actually have the potential to solve the problem, so I think that small as they are, they’re the Schelling point. They are better than RCV from many points of view: political viability, implementation complexity, voter choice, bargaining power of minorities, sizable parties with fully-fleshed-out platforms, local representation, and chain of accountability. And I’m putting in the work to build critical mass, starting with cities with voting problems such as Lowell, MA or Portland, OR.
So here’s what I think every voting reformer should be doing (not necessarily in this order):
- Support IRV (single-winner RCV). For now, it’s the single-winner Schelling point.
- Support other single-winner reforms, including possible future Schelling points like Approval as well as possible ideal points like 3–2–1 or STAR voting.
- Support proportional representation as an overall idea. That includes multi-winner RCV, mixed-member proportional, and other systems; but avoid discussing details of these systems if possible.
- Support PLACE (for partisan elections) and PAD (for local nonpartisan elections), especially in places which urgently need reform such as Lowell, MA.
Note: I realize that the above arguments haven’t given enough evidence to prove that PLACE and PAD are the best solutions for multi-winner voting reform. I could argue that further, but doing so here would make this too long.