Extradimensional Cephalopod might break an Ethics Alarms record today with three COTDs—I’m not sure yet: stay tuned. This one the earliest of the three, includes his cogent analysis of ranked choice voting systems, which I am on record as hating. As I have learned more about the Democrats donating to the nuttier Trump-endorsed Republicans in state primary contests, my hatred is even more entrenched. The more opportunities a system creates to game it, the less trustworthy it is. In my view, ranked choice voting asks the voter to try to game the system. Count me out.
Here is EC’s Comment of the Day on the Alaska special election item in “Labor Day Weekend Ethics Warm-Up, 9/2/2022: Which Are The Pod People And Which Are The Fascists?”
***
At first I took issue with your characterization of the Alaska election, on the basis that ranked choice voting would have removed the spoiler effect. However, I realized that’s not quite true.
Under first-past-the-post voting, if 60% of voters preferred the Republican party but were evenly split across two Republican candidates, then the Democratic candidate would have won with a 40% plurality.
In the same scenario but with ranked choice voting, one of the Republicans would have been eliminated and their votes would have gone to the other Republican candidate, who would have won. In that situation a Republican wins, just like in the FPTP scenario, but also the Republican voters get to vote on which Republican wins without risking their party losing. That is, if they all have a Republican as their second choice.
The Republicans lost because not only did the Democratic candidate have a plurality of first choice votes, but many of the people whose first choice was the less popular Republican Begich had the Democrat Peltola as their second choice.
If Begich had not run, the results would be the same. If Palin had not run… that’s a good question. Unfortunately they don’t appear to tabulate the second preferences of the votes that are not reallocated, which seems like something they should be required to do. However, it is entirely possible that Palin’s voters would have put Begich as their second choice, and therefore if Palin had not run, maybe Begich would have won. That’s a potential spoiler. But since Palin was apparently the more popular Republican candidate, why wouldn’t she be the one running?
I could see a possible argument that ranked choice voting is bad because it encourages parties to run spoiler candidates under the assumption that there can be no spoilers with ranked choice voting. But who was the spoiler here? Begich didn’t affect the election, and Palin might have affected it but she was actually more popular between the two of them.
This is a problem with literally any voting system, regardless of how it’s tabulated, because it’s not always possible to reconcile people’s preferences. A voting system will always have edge cases where it may not fulfill some desired criteria for how it should work ideally. This principle is known as Arrow’s impossibility theorem. Technically, it applies to first-past-the-post voting as well, because surprise! FPTP voting is just another variation of ranked choice voting. It’s the one with the rule that only first choices are counted. In the vast majority of realistic scenarios, that makes it worse than many other forms of ranked choice voting.
Regarding voter preferences, it’s unlikely though theoretically possible to have some voters who prefer A to B, some who prefer B to C, and some who prefer C to A, in such proportions that different types of ranked choice voting will produce different results. There is no one perfect form of ranked choice voting that will avoid all the problems we might object to.
However, that does not imply that FPTP is objectively better than other forms of ranked choice voting. Even if voters are only allowed one choice, those preferences still exist. No voting system can reconcile them in a perfectly objective way. We have to pick the criteria that matter most to us. Do we want to reward candidates for being the least unpopular? Or do we want to reward them for being the first choice of the most people? The method we choose depends on our answer. I assert that the form of ranked choice voting used in Alaska has better results in the most realistic scenarios than FPTP does. I think the way to discourage strategic voting might be to select the electoral system at random from a curated list after people have cast their votes but before the votes are counted.
One thing is for sure, though: it is naive to assume that fixing the voting system will be sufficient for fixing the country. Voting communicates so few bits of information that it’s not a meaningful way to influence nuanced policy. In order to reap the benefits of an improved voting system, we also need to get the public talking to each other, negotiating, and issuing thoughtful, responsible, and reasonable ultimatums to elected officials.
Right now many people feel that voting hardly matters because one politician is about as bad as another. We want to build a world where people fulfill the responsibility of voting, but don’t care that much about the voting system because the top candidates are all decent enough.
Does that all make sense?
Ethics Alarms 
One of the things I think I LIKE about this form of voting is that it gives voters a chance to make a protest vote without necessarily “throwing one’s vote away” in the process.
If I don’t like either candidate in an otherwise-two-party race, I can vote FIRST for say . . . Santa Claus . . . and then second for the “least worst” of the major candidates.
Now imagine (and this CAN’T be all that hard considering the last few elections) that a large number of people agree that both candidates are awful and vote the way I described. In the first round you’d have a lot of “candidates” eliminated (You know: Santa Claus, Mickey Mouse, Superman, “None of the Above”, etc.) and the election would be decided on the 2nd-choice votes of many ballots.
That would send a big message about whether or not a candidate has a mandate from the people or whether they mostly were just holding their noses in the booth. I’d like that a lot.
–Dwayne
Oh, wow. That would be awesome. As much as I’d hate for that to end up happening, I’d love to see some news anchors and pundits try to spin the fact that neither candidate got more than, say, 20% of the first-choice votes.
Which reminds me of how I can clarify what I distrust and dislike about RCV. Let’s say there are five candidates. A voter ranks them A-1,B-2,C-3, D-4, and E-5. And that’s how he should vote. BUT, you see, he fears that B is the candidate most likely to defeat A, his favored one, so he leaves B off the ballot entirely. That makes his vote A-1, but C-2, D-3 and E-4. But that’s a lie. That’s not the way he really ranks the candidates—that’s the way he decides is most likely to elect A. The system encourages dishonesty, indeed rewards it. Verdict: Unethical
Thanks, Jack! I appreciate your willingness to engage on the topic.
As it happens, before I saw your reply, I was trying to think of ways that voters could game the system. The angles I considered boiled down to trying to rearrange the order in which less popular candidates were eliminated in order to boost the second place leading candidate towards a majority. (I assumed the first place candidate would not need a boost.) It sounds like that’s what you’re considering in your example: getting B eliminated first so that B’s votes pass to A, even though B is preferred over the other non-A candidates.
In my first version of that scenario, I did the algebra with four candidates: two frontrunners, a Republican “R1” and a Democrat “D1”, and two less popular candidates, also a Republican and a Democrat, “R2” and “D2”. (…Um, no pun intended.) Let’s say R1 is leading with a plurality but someone wants to make D1 win. They want D2 eliminated first so D2’s votes pass to D1. For simplicity, I assume R2 voters made R1 their second choice and D2 voters made D1 their second choice.
I concluded that under those conditions it would be mathematically impossible to game the system by influencing which less popular candidate was eliminated first. If D2 and R2 were evenly matched and a mere handful of voters could tip the balance and make D2 go out first, that could not give D1 a majority. Therefore, there would be a second round in which R2 was eliminated and R1 received a majority of votes and won. If D2 had enough votes to give D1 a majority upon transfer, then D2 could not be made to be eliminated in the first round: D2 would be too much more popular than R2. In order to make the second-place frontrunner D1 beat the first place frontrunner R1 with a majority, both less popular candidates would need to be eliminated and some people who voted for R2 would need to pick D1 as their second choice. I call that the system working as intended.
I ran the algebra with five candidates as well: frontrunners R1 and D1 and trailing candidates R2, D2, and D3. With two sources of secondary Democratic votes to feed to D1, that would be enough to cause D1 to surpass R1 and win a majority… and it also would not matter in what order the elimination took place, as R2’s elimination couldn’t give R1 a majority.
Again, this is assuming that people only vote for candidates from one party. If they don’t do that, I suspect the system becomes too unpredictable for any regular voters to game it. The only reason people would attempt to game the system by voting against someone they liked is if they weren’t already informed of the math, or if someone tricked them into thinking it was possible. We can take measures against that. The math says that the best bet is to just mark your ranked ballot honestly, based on what you actually think of the candidates.
Does that all make sense? I can post the steps if you like.
I still say most of the work of democracy must take place before anyone casts any kind of vote.
“I still say most of the work of democracy must take place before anyone casts any kind of vote.”
All makes sense. Whether having my vote go to someone I didn’t vote for, and maybe never would, still seems to be a deal-killer ethically, though. No?
“Whether having my vote go to someone I didn’t vote for, and maybe never would…”
Where did you get this premise? Your vote will never go to someone you don’t indicate on the ballot. It’s the same effect as having a runoff election that you don’t attend (or for which you turn in a blank ballot) because you don’t like any remaining candidates.
In the scenarios for which I did the math, I was assuming for the sake of simplicity (and for the sake of a hopefully more robust argument against my own position, the more honorably to refute it) that every voter would specify predictable secondary choices. You don’t have to do that on your real life ballot.
Does that address your point? I’m not sure I understood it.
I may have been assuming that you saw something in RCV that I haven’t figured out yet. What did: “in your example: getting B eliminated first so that B’s votes pass to A, even though B is preferred over the other non-A candidates” mean, if not that someone who votes for B will have their votes “passed to A”? What DOES that mean?
Oh, from your example, I thought that your hypothetical voter wanted to get B eliminated even though B was their second choice, because they wanted A elected as their first choice and assumed that most of the people who voted for B would have put A as their second choice. In other words, the only reason I could think of that your voter would dishonestly leave B out of their ballot was that they figured if B were eliminated, A would get most of B’s votes. I may have paraphrased that clumsily.
It’s just that the way the math works out, it seems there’s no reason to vote dishonestly, because as far as I can tell the order of elimination won’t make a difference in giving A a majority. Does that make more sense?
The paradigm should be one man, one vote, for one candidate. If the required plurality is not reached by any particular candidate then there should be a run-off with the top two contenders. But then I am an advocate of the KISS (keep it simple stupid) in all things, especially politics, gender and taxation.
I agree with the idea of a runoff election, but it seems unnecessary to make people do all the work of voting again if no majority is reached.
Wouldn’t it be simpler if they just filled out one ballot, and the ballot effectively said, “if your first choice is eliminated, who would you vote for in a runoff election? And if your second choice is eliminated, who is your third choice? Et cetera…”
That way the entire runoff election process could take place while only requiring people to go in to vote once, like some sort of… “instant runoff”, if you will.
Perhaps some people can’t grasp the concept of a hypothetical situation like “pretend your first choice is eliminated; who is your second choice?” If that’s the case, they almost certainly don’t understand policy costs and risks enough to competently judge any candidates, and their vote was meaningless noise even under first-past-the-post voting.
Does that make sense?
It makes sense if we posit a fairly informed electorate. However, I think it’s hard enough to get people out to the polls in the first place these days. I don’t think we’re going to get them to go the extra mile. In Alaska, if you recall, the people who did not make a second choice very likely played a pivotal part in the final outcome.
As well, I think that the political parties will do their best to influence how their voters fill out the ballot. I recall during the recall election of Newsom in California, his campaign encouraged people to leave the second half of the ballot blank (California recalls such as that evidently are bifurcated — first you vote yes or no on the recall, and second you vote for the candidate you want to replace the recalled office holder). If the recall supporters had prevailed, that might well have allowed a Republican to win with much less than a majority of the vote.
I think my preference would be for an actual runoff election. It could be that in a multi-candidate election, such as they had in Alaska or if you had minor party candidates, that there could be a plurality threshold. If you get, say, 40% of the vote (and perhaps you have to have a certain margin over the second place candidate) then you win without a runoff. Contrast that to Georgia (and I think Louisiana) where you must have an absolute majority.
On the gripping hand, this is one of the beauties of the federal system we have. States can do their own thing and we can even have Alaska (and perhaps Georgia) as an example to the others of how not to do things.
Are you talking about ranked choice voting, or democracy in general?
A runoff election is just like ranked choice voting, except that it wastes people’s time because you could have collected all the same information from a ranked choice ballot.
Why do people keep assuming that the Alaska election didn’t reflect the preferences of the voters? Citizens voted, and the votes were counted, and Peltola was elected. Riddle me this: Who should have been elected instead, Palin or Begich, and why?
Alright, I’ve collected the arguments people have brought against ranked choice voting and condensed my counterpoints. What do you think?
1. RCV is more complicated than voters can follow.
Counterpoints: Not if we educate them competently, like we already sometimes do with regular ballots. If they’re not capable of comprehending ranked preferences and a ballot that accepts them, then their vote would be meaningless noise even under a first-past-the-post system. If that describes most voters, we’ve got a bigger problem. However, standardized testing indicates that many children can understand and fill out bubble-sheets correctly, so adults should be alright.
2. RCV encourages voters to game the system and vote dishonestly.
Counterpoints: As far as I can tell, that’s mathematically untrue. I don’t have a comprehensive proof, but I did some algebra with four and five candidates and a few assumptions about party affiliation and it indicates that trying to swing a few votes to influence who is eliminated first doesn’t make a difference in which frontrunner gets a majority. On the contrary, under RCV the voter is incentivized to rank their preferences honestly. It’s actually first-past-the-post voting that encourages dishonest voting, also known as “holding one’s nose and voting for So-and-so” instead of a candidate you actually approve of, just to make sure the other major party doesn’t win.
3. RCV encourages bland candidates that try to appeal to everyone.
Counterpoints: On the one hand, having candidates that appeal to everyone would be a really nice problem to have for a change. On the other hand, we also don’t want candidates who do so by being maximally ambiguous and minimally decisive. Luckily, RCV actually encourages candidates who take risks with ideas for overhauling the status quo to replace zero-sum tradeoffs with win-win opportunities.
Republican and Democratic politicians do not lead or accomplish much because they cannot afford to lose their designated voter base. They are limited in what ideas they can discuss safely. They’re more likely to demonize their opponents than push innovative ideas.
4. “One person, one vote.” If your first choice is eliminated, you shouldn’t get to reassign your vote.
Counterpoints: If every person gets the same number of votes, isn’t that the important thing? Isn’t that the real point of the phrase? Also, we already have runoff elections where people get a chance to cast another vote in the same race for the same office. Those don’t seem to be objectionable. RCV works exactly the same way except it streamlines the process by collecting the initial votes and the runoff votes at the same time. That’s why this version of RCV is sometimes called “instant runoff.” If the United States had started out with RCV, I can virtually guarantee that nobody would be demanding a switch to first-past-the-post.
I don’t suppose anybody finds my reasoning sufficiently compelling? Does anyone have any points against RCV I didn’t list here?
Addendum:
Further research indicates that in eliminating the spoiler effect from first-past-the-post voting, ranked choice voting introduces a lack of monotonicity, meaning that in a close race, adding more votes ranking a candidate as the first choice can actually result in that candidate’s loss. This effect might only appear when there are at least three candidates who each have roughly equal support, which is why I didn’t pick it up when I did the algebra for four and five candidates with two frontrunners.
That said, in my opinion, RCV’s elimination of the spoiler effect more than makes up for this flaw in a political context. My reasoning is as follows:
The cause of the lack of monotonicity is circular preferences, where roughly a third of voters prefer A to B, another third prefer B to C, and the remaining third prefer C to A. There isn’t an obvious winner among the group, and depending on which voting system is used and the exact vote totals, the election could arbitrarily settle in favor of any of the three options. They’re all valid equilibrium states.
I’m not sure how common this situation is with political candidates, but one would reasonably hope that if three candidates receive approximately the same number of votes and there is no clear group preference of any candidate over any other, we might hope that they’re all decent candidates and it doesn’t matter so much who wins.
Furthermore, even if we used a first-past-the-post voting system, those circular voter preferences would still be there. The voters just wouldn’t get a third option to begin with, because the parties and the voters would be afraid of the spoiler effect. Someone other than the voters would end up choosing who doesn’t get to run, instead of the voters finding out how all three candidates stack up against one another. Or the spoiler would run, and the election would be spoiled.
FPTP voting cannot fix the problem of resolving circular preferences objectively, because it’s also just a version of ranked choice voting: one where only the first choices are counted. The only thing FPTP does is prevent the issue from coming up, at the cost of incentivizing politicians to demonize their opponents and avoid innovation, prioritizing the appearance of effectiveness over actual effectiveness.
Admittedly, it might be possible to use dishonest, tactical voting with RCV in order to determine who is eliminated first and thereby who gets the majority. However, it would be much more difficult to do that with RCV than with FPTP. If we really want to prevent tactical voting, though, we might randomly select the exact RCV algorithm from a curated list after all the ballots are in.
For the sake of honesty, I do think that all ballot rankings should be recorded, instead of only recording the vote reallocations when one candidate is eliminated. That way people can judge whether the system seems to be working in practice as well as in theory.
And, as always, most of the work of democracy needs to be done before citizens or legislators cast any votes. RCV just helps incentivize that work.
Thoughts?