Tags
2022 election, Bo Diddley, Gregory Isaacs, Morrissey, Queen, ranked-choice voting, RCV, Wall Street Journal
Also two weeks ago Thursday, The Wall Street Journal had an opinion piece about Alaska, its Ranked-Choice Voting (RCV) system, and the implications for the mid-term election.
I don’t particularly like Ranked-Choice Voting, and for a number of different reasons. I might concede that it could help with party primaries, where a large list of contenders needs to be narrowed down in preparation for a traditional general election. Furthermore, in most (but not quite all) respects I find RCV inferior to “approval voting,” a method by which you can vote for as many or as few as you like. Detailed arguments over that, though, will have to wait*.
Instead I’d like to put together, here, an RCV voters’ guide for those who find themselves subject to this voting method unwittingly or even unwillingly, but who want to maximize their effectiveness within the system. I am inspired by the WSJ article in what it doesn’t offer. It suggests that the Republican Party will likely fail to regain the traditionally-Republican seat come November, despite having an understanding of why they lost in August. The WSJ criticism is aimed towards Republicans, which is to say the candidates and the party’s leadership. I’d like to speak for the voters. For a number of reasons, I’ve spent a lot of time trying to figure** this system out and Alaska’s experience clarified a few things for me.
If you wish to come along with me, it may be because you are already in a jurisdiction that uses this voting method. Perhaps you truly want to understand the process better before you go to the polls. I hope that, even as lengthy as my post is, it will help you to do that. It may be, instead, that you are in a State or locality that is considering changing over to this method – and many are “considering,” even if it is a longshot. In that case, these details might help you form an opinion on how RCV works and whether it is needed in your locality. Either way, you can follow through my analysis of a hypothetical ballot and join me as I form my own concrete ideas about how I should approach an RCV election. If you’d rather not wade through my extensive prose but still want to cut to the chase, you can just jump ahead to the end.
Are you still with me? Good. As you imagine standing in the voting booth with me, consider we each hold a ballot where we must vote to fill three available seats and have nine names on the ballot. Such a specific example will allow me to refer to the numbers more simply. In a traditional election, this would be a “vote for no more than three” office. Typically, you’d have six or fewer vying for the three slots, as the two main parties have nominated exactly three for the office. RCV imagines various scenarios – an open primary system or maybe multiple, smaller parties – that would often increase the candidate list beyond this traditional, “binary” choice.
Before we start on the specifics, I’ll make one more stab at a short version. The key to correctly voting an RCV ballot is to consider it as programmed instructions for a sequential series of elections. After you choose your favorite (as you would on any ballot), you then pick a second choice to vote on the N+1nth ballot when your first choice has been eliminated on the Nth. And so on until only M+1 ballots remain (M being the number of winners i.e. 3 in my sample). Sound complicated? Well, we haven’t even started yet.
Looked at it this way, there is no point*** in ranking more candidates than the one you favor plus, maybe, some others you’d accept if your favorite gets knocked out. The problem is that we can’t know what’s going to happen once your ballot is turned in, so you really need to plan for every contingency. You might also need to make some choices – what works for one contingency may end up being counterproductive in a different and unexpected case. The details that follow help organize such a planning exercise.
Step 1: Tell Me, Who Do You Love?
The first step is probably to consider who you like. It is just possible, as unlikely as it may seem, that the ballot offers you more than enough acceptable candidates. If so, then your job is pretty much as this voting method is portrayed in the media. You simply have to decide the order of preference for the candidates that you like.
I’ll emphasize this again; you don’t need to vote for all of the candidates. The magic number for our sample ballot is 6; six candidates need a ranking. In general, you have to vote for the number of candidates (let’s call this Cn) who are to be eliminated. So, for a different example, in the case of a single seat with four candidates, you should be ranking your top three.
This math is important because voting for more than that number means you are guaranteed that your bottom rank (or ranks) will not be used and voting for any less leaves you potentially left out of the final, determining voting round. So always vote for exactly Cn candidates, right?
Well, maybe.
Before I go on, I seem to be developing some sort of a math journal paper with my notation. I’ll just summarize here what we have so far:
- M: Number of open seats (3 in our sample ballot, 1 in a “typical” election).
- Cn: Total number of candidates vying for the open seats. (9 in our sample ballot. This number must be more than 2 for a single-seat contest for RCV to have any meaning.)
- N: Number of rounds of balloting, assuming that a majority is never achieved prior to eliminating all losing candidates. N can always be calculated as M – Cn. It is also the ideal number of candidates which you should rank. (6 for our sample ballot.)
- Cs: The number of sickening candidates. See next section for an explanation.
Step 2: but I’m So Very Sickened
The other item you have to look at is whether there are one or more candidates by which you absolutely cannot abide. Let’s call this number Cs (for sickening candidates). If this number happens to be the same as M, again you are in luck. On our imagined ballot, you would simply have your three candidates which you will not rate.
If Cs < M, then you will not rate them, but you also have a decision to make about those who remain. That decision may depend on other factors.
In the original article about Alaska, the author pointed out that there was no information on the rankings assigned for each ballot. The State only released the result of each “instant runoff.” As complaints about this lack of transparency emerged, the State released JSON files which contain some or all of the ballot information in the format that the State system processes the data. Presumably the release of ballot data will eventually lead to a full analysis of that data, regardless of format. At this point, however, it’s not clear how detailed any given (and, at this point, hypothetical) election release might be.
Why dwell on this? The answer is that your vote might have two distinct meanings. The obvious meaning of your voting comes from how your ballot is used in any given round during the election’s determination of a winner. However, if the full ballot rankings for all votes are eventually made public, then your rankings are a permanent record of your intent as a voter which, to you, may be just as important (if not more important) that then actual consequence.
At this point you may be thinking, “what the hell is this guy on about?” Let’s go back to that Cs < M.
Let us also use our sample case. 9 candidates for 3 spots. We also assume that there are two entirely unacceptable options. Of the remaining seven, some are good and some are kinda bad, but none are abhorrent. So what do you do? Do you rank the worst of the remaining 7 as “7”, even though that vote can never possibly be counted? Maybe you do, if you want it on record that “7” is better than “8” and “9.” Or maybe you withhold a rank from “7” because you don’t want it in the record that you “voted for” this candidate whom you would never actually vote for.
All of this, by the way, assumes that the full record of votes is going to be published. Because how you allocate “7” will never show up in the actual results.
It gets worse though.
What if your analysis of the candidates determines that Cs > M. That is to say, what if there are at least four really, really bad candidates? Again, you have to make a choice.
You can rank all N candidates knowing that, with your bottom choice, you’re actually voting for someone whom you don’t want to be elected. But you might want to do just that so as to defeat an even worse choice. You might also do so to signal to the post-election number crunchers that there are, indeed, degrees of horribleness. The problem of course is that a vote for a candidate (whether in theory or in practice) is a vote for a candidate. Do you really want to vote FOR someone whom you can’t stand to see elected?
Step 3: Now the game has just begun
Step 3 would be to fill the ballot. One special feature of the RCV voting method is that, at each ballot elimination and assuming your ballot is still in play, your vote is worth more than it was in the previous step. This is because, inevitably, some other voters’ ballots have already been thrown out.
To put it another way, while one of the advantages touted about RCV is that it assures that the winners always have a majority of voters, this isn’t quite true. The caveat is this; the winner is whoever has a majority of voters on the final ballot. In San Francisco’s municipal elections, for one example, the winners have been chosen with something like only a third of the original ballots in play. In many ways, then, as many as 2/3rds of the voters in an election become disenfranchised, even though they leave the ballot booth thinking they’ve voted!
Therefore, one good way to think of your ballot rankings, having picked your best candidate, is to imagine that said candidate is eliminated from the ballot. Then pick the best candidate from those who remain. Continue this exercise six (N) times. Why do it this way? Because this is the way the ballot will actually be used.
Or would be, in your own worst case scenario.
Step 4: So easy when you know the rules
However, if you’re going to play the game, you ought to play your best game. While nobody can know a priori what is going to happen after the first ballot, we frequently can make a good guess.
This optimization step is, of course, optional. If your number one choice is certain to make it to the final ballot (either as a winner or as the final contest), then there really is no need to rank anyone else. You pick your candidate and then see who is the winner. Yes, RCV would mean resubmitting the same ballot multiple times, but if that’s your vote, that’s your vote.
For me, however, this seems like a suboptimal use of the tools that I’ve been given. If the vote goes six elimination rounds, and my guy comes in number one, I’ve essentially “sat out” all the intervening rounds (whether because my guy already had 50% + 1 or because my guy wasn’t in danger of being eliminated until the very last ballot). I don’t want to be sidelined if I can figure out how to participate.
Let’s imagine that I have three candidates that I actually like. My favorite guy is going to be a tough sell. He’s a bit outside the mainstream and I have no idea whether he can win or not. He has a chance, but it’s a stretch. My number 2 is an odds-on favorite. No guarantees, but he’s almost certainly to be among the three winners. My third choice is a radical. I’m not sure he’d exactly be the best person for the job but he espouses certain positions that nobody else will say (at least out loud). On that good-on-some-not-so-good-on-others basis, he rises to number 3. So should I rank them 1-2-3? No, I should not.
What I will do (and maybe have done – see**) is rank my #3 as my first choice. That way, I’m actively pulling for one of my guys during the round that he is eliminated and, let’s face it, his elimination is going to be inevitable. Once he’s gone, then I can fall back to my #1 to try to help him do the improbable. If he fails, then my vote goes to #2.
Even more confusing if we swap my preference for #1 and #2. If my preference is really for the guy who’s going to win anyway, I want him at the bottom of my “best” choices – still third. Why? Because I want to save that vote for the final round, when it will actually count.
I used the specific example but, if you’re willing to swallow all the complications, this can be generalized. What I want to do is, during any elimination round, always have either a vote for the candidate being eliminated or a vote for the candidate doing the eliminating. Not only am I then able to, with a single ballot, influence every action during the ballot counting, but my votes are part of each tally in the news story which follows.
Too Clever by Half?
There is a downside to this, of course, because everything has its pluses and minuses.
If I could know, exactly, how many votes each candidate was going to get in each round before I voted and submit my rankings accordingly, I could maximize my ballot’s impact. The problem is that I can’t know if my predictions are accurate or not. In fact, if I could precisely know every outcome before it happened, there wouldn’t be any point in voting at all. Since I don’t know the results in advance, might my gamesmanship be based on projections that wind up with me voting the wrong way?
The obvious mistake would be if I overestimate the chances of my favorite candidate and therefore wind up not voting for him during the round when he gets eliminated. If there are other voters (perhaps my readers among them) also trying to game the system, this gets more and more likely.
Another mistake is to forget that both the numerator and the denominator are changing.
Hold on, I’ll explain that one.
If a candidate is among the top 3 but has yet to get a “majority,” withholding your vote for that candidate in the early rounds will not impact the final decision. That is, a vote for any candidate without a majority cannot cause that candidate to win. This simple analysis ignores the denominator. As candidates are eliminated, some ballots are eliminated as well. In this way, each round sees the “favorites” moving closer to that 50% mark, even if they haven’t received any additional (lower-ranked) votes shifted over from the eliminated candidates. In this way, it may be possible that the elimination of an “opponent,” from whom no “second choices” are awarded, could actually change the math enough to secure victory for one of your own favorites. In other words, the point in the process where your ballot determines victory might occur in an earlier round than when you’d otherwise anticipate.
Yet another, albeit a rather unlikely failed outcome, is if I actually choose a lesser candidate over my preferred candidate because I’m sure that lesser candidate has no chance. What if, then, it is my vote that ends up knocking out my preferred candidate in favor of the lesser one? That would hurt.
In Summary
One of the reasons I go through all this exercise is that the math, the probabilities, and the politics all fascinate me. It’s sometimes enjoyable, when I’m waiting to fall asleep, to game out these odd election scenarios in my head. Given that, I probably wouldn’t mind finding myself having to vote in an RCV election once again (because**…).
However, the bigger reason I went through this exercise is that I’m pretty sure that RCV is a bad idea. It is a bad idea for a bunch of different reasons but the largest of these reasons is that so many people won’t understand its complexity. Worse yet, knowing they don’t fully grasp the complexity, they’ll feel that they’re somehow being cheated out of the full impact of their vote. This will either make them unhappier about the results of an election in which they voted or it might keep them from voting in the first place.
This is bad. It is bad in a way that outweighs any possible benefits from the more-advanced balloting method.
Therefore, the reason I have thought about this so extensively is that I want to be prepared to explain why I think this voting method is flawed. It isn’t enough (and believe me, I’ve tried) to explain to someone that the method COULD be gamed. It is not enough to prove, mathematically, that it COULD produce the wrong result. To an advocate, these edge cases seem so unlikely as to be automatically discounted.
Maybe if I explain how I WOULD vote – in fact, how anybody SHOULD compute whom to vote for – maybe then the flaw in this system will be as apparent to others as it seems to me.
*In one case I will not wait because I will almost assuredly forget a recent criticism of RCV. A letter-to-the-editor response to the cited Journal article points out that Abraham Lincoln would almost certainly have lost the 1860 election. He was the radical among the presidential contenders and thus a system designed to find the “middle ground” would have meant an acceptance of the institution of slavery even if perhaps only a partial acceptance. History tells us most assuredly that would have been the wrong choice.
**Among other things, I voted in at least one Cambridge City Council election using a ranked-voting system. Cambridge has a government consisting of nine at-large members meaning a voter chooses from a list of 20-30 names to fill nine seats. It is unreasonable to think the average voter can accurately rank a list of 30 candidates for office or, really, even nine. That’s before that voter starts thinking about how to game the system a bit.
***Everything you do can have an upside or a downside. Ranking more candidates than those that will be eliminated means that you’ll have some rankings that will never be used. If, as the Journal complains about the Alaska election, the raw data are never published, your 7th, 8th, and 9th vote can never see the light of day under any circumstances. If the data are published, someone might use the fact that you voted for your candidate #9, the one you really hated, to argue that you actually found him acceptable.