Wednesday, September 07, 2022

It's Not Cheating for Republicans To Lose: Ranked-Choice Voting Edition

I know it's not worth it to engage in Republican histrionics about how ranked choice voting is anti-majoritarian after Democrats won an Alaska House seat last week. The actual objection, as Republicans have made manifestly clear in their behavior over the past few years, is to "Democrats winning elections", and there's nothing deeper than that going on under the surface.

But the arguments they're making about how ranked choice systems are anti-democratic because "60% of the voters in Alaska voted for the Republican agenda" are so transparently ridiculous, and are being repeated with such vigor, that they need to be addressed.

Of course, it is a misnomer off the bat to say that a majority of Alaskans voted for "the Republican agenda". Voters don't vote for "agendas", they vote for candidates. And leave aside the notion that Republicans suddenly care about majoritarianism in a electoral system riddled with anti-democratic elements ranging from gerrymandering to the Senate to the Electoral College.

Nonetheless, it is the case that something feels off when more voters choose candidates from party X but, because they're divided, a single candidate from party Y prevails with a plurality. This can afflict Democrats as well as Republicans (witness worries about Democratic "lock outs" in California's top-two primary system). And it's worth noting that this circumstance is actually very common in a multi-candidate field with first-past-the-post rules. Indeed, Mary Peltola won a plurality of first-choice votes -- she would have won the election without a ranked-choice run-off! (Peltola had 41% of the initial vote, with Palin receiving 31% and Begich 28%).

But here's the thing: when we see voting patterns where 40% of the electorate backs a Democrat, 35% back Republican A, and 25% back Republican B, the reason we think it's unfair that the Democrat wins is that we assume if we asked the supporters of Republican B "if you had to choose, would you back Democrat or Republican A", they'd pick the latter. It's a reasonable enough assumption in a party system, to be sure, and in many occasions I suspect it's an assumption that'd be borne out. But all ranked choice voting does is actually ask the question rather than assume its answer. And it turns out that in Alaska, enough supporters of "Republican B" (Begich) did not prefer Republican A (Palin) over Democrat (Peltola). So the Democrat won, for the simple democratic reason that most Alaska voters preferred her over the most popular Republican competitor. That's not cheating, that's an election!

Put simply, if a majority of Alaska voters' preference was to elect a Republican -- any Republican -- over a Democrat, the voting system in Alaska gave them ample opportunity to make that choice. They chose otherwise, because it turns out that their preferences weren't that simple. And ultimately, that's what's driving Republican rage here: they think the voters' preferences were wrong, and so it is cheating for their will to have prevailed. Hard to think of a pithier summary of contemporary GOP attitudes towards democracy.


bookworm914 said...

I agree with your analysis overall, but I think the argument would be stronger if it explicitly addressed the exhaustion of Begich ballots. The final round margin b/w Peltola and Palin was about 5000 ballots, whereas about 11,000 Begich ballots were not counted in round 2 because they did not rank a 2nd candidate.
RCV round 2 reallocations:
Begich - Palin 27,053
Begich - Peltola 15,467
Begich - Exhausted 11,243
Begich - rank 2 Overvotes 47
Perhaps most of those exhaustion voters genuinely didn't care to distinguish the other 2 candidates - Alaskans for Better Elections says their exit poll showed 33% of voters ranked only 1 and 75% of those (ie 24-25% of total voters) said that was because they didn't like any other candidate. However in practice, the exhausted Begich ballots are a noticeably smaller percentage than the exit survey found (11000 / 53000 ~= 20% < 33%) and based on ideology we would probably expect that Peltola voters are more likely to disdain all other candidates. So maybe half or more of those Begich voters would have ranked Palin had they understand the ballot better. 20% exhausted ballots is pretty close to the 15% surveyed who said the RCV ballot was difficult.
If those 11000 broke down in the same proportion as the 43000 who did express a 2nd preference, they'd have been slightly less than 2:1 in Palin's favor, which would make the round 2 outcome noticeably closer but Peltola would still win. (Splitting 7250 / 3750 would net Palin 3500 votes, and she would trail by about 1500.)

No new conclusions - Republicans are hypocritical and full of it, as you already said. And if ballot exhaustion is reporting voter sentiment inaccurately, the solution is more voter education not less RCV. But still, I suggest that addressing the claim that "exhaustion is disenfranchising" is part of the project?

Erl said...

Completely agreed that any whining re: ranked choice voting is hypocritical; and the Emmer quote in the tweet engages in the Trumpian bad behavior of fudging the margins as well (40.2% is not "less than 40%", it is in fact "more"—although I should note that our host's post gets this wrong, too!)

But I think there is something a bit more interesting at play here. Begich was probably the Condorcet winner of those three candidates—that is, if Begich had advanced to the final round against either candidate it seems quite likely he would have won. (This fact is probably literally knowable—the State of Alaska has all the ballots!—but does not seem to have been published; contrast the robust data coming out of the NY ranked mayoral.)

From a partisan perspective, Palin was (likely) a general-election spoiler in exactly the way that radical primary candidates are: she crowded out the Republicans' best nominee and then couldn't seal the deal with the whole electorate.

Now, this isn't Instant Runoff's fault. As you note, if the winner had been chosen by First Past the Post, Peltola would have won by 7.9 points, not by 3.0. And the IRV accurately captured an interesting and truthful fact about the electorate: Begich voters didn't really care for Palin. (Imagine being Republican A and picking up just over 50% of Republican B votes, while the Democrat picks up nearly 29%! Fuckin embarrassing.)

And in general, in a three-way race, IRV will do strictly better than FPTP in picking Condorcet winners. You could imagine a four-way race where that's not true—where the Condorcet winner happens to lead in the first round, then gets eliminated—but even that would be a coincidence, not really a superior performance by FPTP.

(When I say you could imagine, I mean, "I built a model scenario where that happens". Not that the model has any meaning but it kind of looks like a four way race between three ideologically divided candidates and one minority community candidate. The compromise majority candidate has a small lead in round one. But the minority candidate only has the support of their small community, and is eliminated. The minority community is sharply ideologically divided and goes to the ideological wing candidates in round two, pushing out the centrist candidate; then the final round is between the wings, with whichever has a slight edge winning—but the centrist candidate would have dominated any head-to-head race. Not that I've ever heard of an election like this, but one can imagine it happening. But even in this scenario, the fact that the Condorcet winner leads round one just happens to be true, it's not informative.)

And of course, back in the real world, we can't generalize from the (presumed) choice of this electorate to the behavior of an electorate in a two-way race. Would Begich have been able to turn out Palin-first Republicans in a head to head with Peltola? Would he have been able to turn out Dems in an interparty head-to-head with Palin? We can't say.

But I think this is an interesting reminder that while IRV mitigates some of the obvious failings of FPTP, it doesn't abolish strategic voting considerations altogether.

Erl said...

@bookworm I missed this the first time, but you write: "based on ideology we would probably expect that Peltola voters are more likely to disdain all other candidates"

I would be shocked if this were true.

I cannot imagine being an Alaska Democrat with no preference between Sarah Palin (shame of my state, Trump before Trump, etc.) and Nick Begich III, grandson of Democratic congressman Nick Begich, nephew of Democratic Senator Mark Begich. Palin leaned into 2020 election conspiracy; Begich punted, which is weak but still way better. Etc. I'd doubt that Peltolta could win in my substantially-Republican state, but I'd still be pretty excited heading into the ranked-choice special election: two different opportunities to Stop Palin!

On the other hand, I can easily imagine being a Palin-first voter with no preference between Begich and Peltola. RINOs are the same as Democrats, don't trust the establishment, etc. I can easily imagine a Palin/Nothing voter.

I'm reminded a little of the time I looked into third-party voting in modern US presidential elections. A lot of discourse assumes it's a self-indulgent left-wing phenomenon—but in every US presidential election since 1968 except for two, right-wing 3rd party candidates have beaten left-wing 3rd party candidates. (The exception is of course the two candidacies of Ralph Nader.) There is no particular tendency for the Republicans to stick together any better than Democrats.

So I modeled the following assumptions:

Peltola-None voters the same percentage of Peltola voters as Begich-None voters of Begich voters, ~20%
Palin-Begich voters the same percentage of Palin voters as Begich-Palin voters of Begich voters, ~50%

This gives a single-vote total of 30%, pretty close to the 33% you cite. And it does lead to the counterfactual second rounds I posited above—Begich beats Palin in a rout; Begich beats Peltola in a squeaker. (Although if you crank up the Palin-None votes, Peltola could squeak ahead of Begich too—with greater difficulty.)

Of course at some level it's silly to speculate so intensely on matters of fact, and probably matters of public fact; but I have a very different model of what the electorate was like than your comment suggests.