There’s lots of exciting analysis here on this crazy Alaskan special election. And you’re probably not going to see this kind of analysis in many other places given how complicated ranked-choice voting is to decipher. There’s also an odd fascination with many media outlets just blindly repeating RCV advocacy claims. Even RCV critics in the media tend to have strange arguments.
This election was truly weird. Conservatives got a worse outcome for ranking their favorite. They even got a worse result for showing up at all. And the Democrat, Peltola, could have lost by getting more 1st choice votes from Palin supporters. RCV even framed the race as being between the wrong two candidates.
But let’s start by making sure everyone is on the same page. Because of Representative Young’s death, there was a special election, and this was ranked-choice voting’s (RCV) first use in Alaska.
In 2020, Alaskan voters passed a ballot initiative by half a percent. This initiative implemented an open primary system with RCV in the general. As news to some, this was RCV advocates’ second attempt in Alaska. RCV failed on another ballot in 2002 when nearly two-thirds of voters opposed it.
So what’s this new approach look like?
The primary has all the candidates, regardless of party, on the ballot, and voters are allowed to choose just one candidate. The top four go onto the general.
The general election had traditional ranked-choice voting. This is a ranking method that simulates sequential runoffs using voter rankings. After voters submit all their ranked ballots, here’s how RCV then works:
- Check if any candidate has >50% of 1st choice rankings. If yes, that’s the winner. If not, continue.
- Eliminate the candidate with the fewest 1st choice rankings. Transfer those ballots to the voters’ next-choice preference (if they marked one).
- Check if any candidate has >50% of the 1st choice rankings of the active remaining ballots. If yes, that’s the winner. If not, continue to eliminate the candidate with the fewest 1st choice rankings until the winning threshold from the remaining ballots is reached.
Hype & Claims
Alaskan voters were told a lot about what to expect with this new voting system as well as what it did. For instance, Reason magazine repeated a claim about voters being freed to vote their conscience:
“But with ranked choice voting, voters are freed up to vote their conscience, rather than whoever they think is most electable, without having to worry about voting for a “spoiler” candidate.” – Joe Lancaster, Reason.
Here’s another one:
The claim here is that voters can always support their favorite candidate and if that candidate doesn’t win, their vote will count towards a compromise.
Real Clear Politics made the popular claim that RCV guaranteed a majority:
“They were also voting to guarantee that the winners of their elections would have to win at least 50% of the votes.” – Jason Grenn, Real Clear Politics.
Just go ahead and hold onto those RCV claims. Here they are for reference:
- You can always support your favorite without worry
- Your second choice will come into play if your first choice can’t win
- The winner will always get a majority
(Spoiler: none of these are true. Read on to see why.)
Note that one candidate, Al Gross, wound up dropping out and wouldn’t be replaced. These were the results of all 48 candidates in the primary:
When voters choose among candidates who are similar and can only choose one of them, you see something called vote splitting. Vote splitting happens when voters’ support divides between candidates. Consequently, similar candidates reflect artificially lower support, and dissimilar candidates (particularly ones that signal that they’re viable) can appear to have relatively higher support.
As a result of this vote splitting, we see lots of candidates with very small amounts of support. Many of those candidates have artificially low support, partly because voters fear they aren’t viable. But even among the frontrunners, there is further vote splitting. Indeed, having a top-4 is necessary as a buffer to make sure that the best candidate in the pack is more likely to make it in the general. Otherwise, we could miss out on the best candidates entirely. Even then, in some cases, the vote splitting may be too much.
As we see, Palin significantly led the pack. We know there’s some obvious distortion here because of the inevitable vote splitting. We saw this same distortion even with RCV in our analysis of the Democratic primary.
1st Choice Results: General
Remember when we talked about vote splitting? While that can happen when voters are forced to choose one candidate, it also happens when voters are forced to rank one candidate as first. And that can eliminate otherwise strong candidates.
Let’s take a look at the 1st round under RCV (red highlight shows the eliminated candidate):
Unlike the primary, the Democrat took the lead here. This is likely due to an absence of liberal candidates to split the vote with as well as the two Republicans splitting their 1st choice votes. Interestingly, support for Republican candidates outnumbers support for the Democratic candidate 2-to-1.
RCV now eliminates Nick Begich first because he had the fewest 1st choice votes. Let’s see what we have for the next tally.
Final RCV Tally
Complete RCV Results Table (yellow highlight indicates the RCV winner):
Peltola was able to hold her lead. Begich is eliminated and 11,290 of those ballots were exhausted due to not ranking anyone else or other ballot issues. The remainder of Begich’s next-choice preferences went to Palin over Peltola at a roughly 2-to-1 rate.
Note as well that the Democrat only actually won with under 50% of the vote when you consider all the original ballots. RCV results inflate the win percentage by only looking at the winner’s support out of the remaining ballots and dismissing exhausted ballots that don’t have enough rankings on them. This sort of majority is contrived by artificially reducing the number of candidates to two.
This is important because there’s a false impression that RCV guarantees a majority winner (as Real Clear Politics repeated). It does not. In fact, no voting method can guarantee a majority winner in any election when there are more than two candidates. A majority winner may exist in an election, but it is not guaranteed. There was no majority winner in this election.
Further simplification below (invalid ballots and write-ins are removed for simplicity; implied expansion showing preference of collective options is used when voters limit their rankings; see Alaskan site for raw files; non-material adjustments made for total matching; written analysis may use more rounded numbers for easier reading):
RCV got a result. But remember. Any voting method can do that. It’s not yet time to celebrate.
Let’s ask whether RCV gave the right result. And let’s further ask whether the electorate would have been better off if more voters had been dishonest or not voted at all.
The candidate who could beat anyone doesn’t win under RCV
A common way to determine if a voting method was correct is whether it elected the candidate who could beat everyone else head-to-head. In this case, the candidate who could beat everyone else head-to-head (inferred using RCV’s ranking data) was Begich.
See all possible combinations below:
- Begich (101,314) > Palin (63,689) [Begich win by 61%]
- Begich’s 101,314: being 53,810 Begich first choice, plus 47,504 Peltola > Begich > Palin)
- Palin’s 63,689: being 58,973 first choice, plus 4,716 Peltola > Palin > Begich)
- Begich (88,018) > Peltola (79,484) [Begich win by 53%]
- Begich’s 88,018: being 53,810 Begich first choice, plus 34,208 Palin> Begich > Peltola)
- Peltola’s 79,484: being 75,799 Peltola first choice, plus 3,685 Palin > Peltola > Begich)
- Peltola (91,266) > Palin (86,026) [Peltola win by 51%]
- Peltola’s 91,266: being 75,799 Peltola first choice, plus 15,467 Begich > Peltola > Palin)
- Palin’s 86,026: being 58,973 Palin first choice, plus 27,053 Begich > Palin > Peltola)
And as we know, Begich was not the RCV winner. Worse, RCV eliminated Begich in the 1st round—a result of vote splitting of 1st choice-rankings.
How could Begich get eliminated in the 1st round? This is due to a phenomenon called the center-squeeze effect. This can occur in choose-one elections, runoffs, and also RCV. The candidate in the middle has their support divided by candidates on either side of them. Here, Begich split the vote with both liberal and more conservative voters.
RCV chose the wrong winner.
Fun fact! A beat-all (called Condorcet in technical speak) winner doesn’t always exist in every election. Of course, it did exist in this election, and it wasn’t the winner that RCV chose.
RCV punished Alaskan voters for ranking their favorite as first
More interesting than RCV choosing the wrong winner is how it chose the wrong winner. We know that Begich could beat either candidate head-to-head. This means that if Begich had made it to the second round, then he would have beaten any candidate he was up against.
We know the following based on the rankings:
If at least 5,200 (Palin > Begich) voters instead ranked their second favorite (Begich) as first, then they would have gotten their second favorite candidate rather than their least favorite. That’s because Palin would have been eliminated first causing Begich to beat Peltola in the next round.
Instead, Palin voters got their worst outcome because they honestly ranked Palin first.
We don’t often associate the spoiler effect with close elections, but this was a spoiler effect. That is, if not for Palin (who wasn’t going to win) entering the election and taking 1st choice votes away from a stronger candidate, Begich (the best candidate by head-to-head matchups) would have won.
As Alaskan voters can now see, RCV still suffers from the spoiler effect—despite claims to the contrary.
Some Alaskan Voters Literally Would Have Been Better Off Not Voting
This RCV election gets even weirder. Palin voters didn’t create a worse outcome for themselves merely by ranking Palin first. They got a worse outcome for themselves just by showing up.
That’s right. If between 5,200 and 8,500 (Palin > Begich) voters hadn’t shown up at all, those voters would have gotten a better outcome. Begich would have gone against Peltola and still won.
Because Begich was the best candidate given the ballot data, RCV actually would have generated an outcome that would have satisfied more voters if fewer people had voted.
It’s a good thing for Peltola that she didn’t attract more Palin voters—she’d have lost
The strangeness continues. Peltola could have actually gotten more 1st choice votes in this election and caused herself to lose. How’s that? Let’s look.
The whole key to RCV strangeness is changing who gets eliminated first. That’s because the ballots going to the candidate eliminated first are the only ballots guaranteed to have their second-choice preference looked at. If there’s any way to change who gets eliminated so that a new candidate advances and changes the outcome, then you’ve got a break in the election. There are many ways to break RCV elections.
Imagine if Peltola reached across the aisle and spoke directly to Palin voters. Imagine that she empathized with their position and identified issues they cared about that Palin and even Begich ignored. And let’s say that as a consequence, Peltola got the first-choice votes of between 5,200 and 8,500 voters who would have otherwise ranked only Palin.
What happens as a result? Palin would have gotten eliminated in the first round and Peltola would still not be able to beat Begich. Peltola is in an odd position where she has to ride one of two sweet spots. She has to either attract either very few or a whole lot of Palin voters. Anywhere in between and she loses. In the actual election, her “sweet spot” was attracting few Palin voters, which she did successfully.
Strangely, some RCV advocates claim this kind of oddity is fine so long as the winner didn’t actually lose by getting too many votes. I disagree. It’s always bad for a voting method to put a candidate in a situation when it’s possible for them to lose by getting more votes—wherever those votes may come from.
This is one of the weirdest anomalies that can occur in RCV, and Alaska had a front-row seat. One reason it’s likely not more talked about is that RCV’s own complexity shields the ability of the media to easily see these issues.
In contrast with the claims made by the media and other organizations, we see several discrepancies.
Unlike the claim in the Reason article, we see that voters can’t always vote their conscience without it coming back to hurt them. Palin supporters saw firsthand that by naively ranking her first, they got their least favored outcome.
We also see that your next-choice preference doesn’t come into play if that candidate is already eliminated (sometimes incorrectly). FairVote’s claim that voters have a reliable backup doesn’t shake out because (like Begich in this election), RCV may already eliminate that candidate. And if your next-preferred candidate is already eliminated by RCV’s algorithm, it’s like you never ranked them at all.
And unlike the Real Clear Politics claim, we see another example of how RCV doesn’t guarantee a majority winner (again, no method can). Even through an arbitrary elimination process, Peltola only got 48% of the total vote. And we know she doesn’t win all her head-to-head elections.
Future & Alternatives
This is not RCV’s first catastrophic failure. And given its usage, it won’t be its last. In an election back in 2009, the city of Burlington, Vermont used RCV. And in a similar fashion to Alaskan conservatives, Burlington conservative voters were punished for ranking their favorite as first. Conservatives’ reward for their sincerity was a mayor from the Progressive Party. This Progressive Party candidate was also the wrong winner in the same way as this Alaskan election; a different candidate was able to win in all head-to-head matchups.
Burlington repealed RCV immediately following that election, though they recently voted to bring it back. This is partly because RCV has become in fashion within progressive circles.
It’s easy to become shortsighted with elections, particularly when your favorite party gets elected—even if undeserved. It’s important to remember that the job of a voting method isn’t to get your favorite candidate elected. A voting method’s job is to elect the best candidate for a given community of voters so that everyone is represented as well as possible. Of course, this never happens perfectly.
The issue with ranked-choice voting is beyond its complexity, though that also shouldn’t be overlooked. It’s important to know that there’s a simpler reform that performs better. For example, a method that lets you pick all the candidates you like (approval voting) is now seamlessly being used in US cities due to its simplicity.
Liberal voters who got their way this time must note that RCV is a chaotic voting method. That a candidate can win by getting less support and lose by getting more support is indicative of that chaos. The issue with chaos is that not only are its individual movements a bit random, but its randomness stretches in many directions—including directions opposite your favor. Instead of hoping the chaos shakes in one’s favor, we must search for voting methods that are not only good at picking good winners but are also reliable.
RCV is a bit of a random pendulum. Next time it may hurt liberals. To show how close liberals came to a bad outcome, keep in mind that if only half the Begich voters who didn’t differentiate between Palin and Peltola were convinced to rank Palin next, there would have been a different outcome. Having more Begich voters indicate a 2nd choice ranking isn’t a stretch either. Consider that those who ranked Palin and Peltola first indicated their next-choice preference at much higher rates.
This election should be a warning to others. First, open primaries can also suffer from vote splitting when you force voters to choose just one candidate. If you’re doing open primaries, you should let voters pick all the candidates they like (approval voting) to address the vote splitting. You should also let voters pick all they like in the general election to avoid the oddities seen in this election.
We’ll be doing additional analysis on this election with approval voting. Stay tuned and follow our work. You won’t see this analysis anywhere else.