Lessons from Alaska
Voter complexity, shaping candidate behavior, and implications for 2022 and beyond
(photo: Loren Holmes/Alaska Daily News/TNS)
Yesterday’s results from the special election in Alaska make me think of two things: (1) ranked-choice voting’s ability to capture complex preferences, and (2) implications for November’s national Congressional election.
Ranked-choice voting and the complex preferences of Alaskans
In a race to fill a Congressional seat left vacant by the passing of Alaska monument Don Young (R), the winner was Mary Peltola (D). A Yup’ik, Peltola is the first native Alaskan to be elected to Congress. She bested two Republicans, Nick Begich and Sarah Palin. How did Peltola do that in a state that has voted 53-54% Republican in recent years?
The naïve answer might be, by splitting the vote. But Alaska has implemented a voting reform in which the final round is a ranked-choice election. In theory, people who preferred Begich or Palin could have their vote get reassigned to the other Republican. But not enough of them did that. Why not?
Ranked-choice voting is said to favor the election of the most moderate candidate. If we consider today’s election from the candidates’ point of view, that idea failed. This diagram illustrates the point:
(Please pardon the crude drawing. I’m traveling!)
Begich was the middle candidate and could have been the consensus choice. But he was squeezed out in the first round from both sides, leaving a choice between two candidates farther from the median Alaskan voter.
Now think not just of candidate positions, but also voters’ point of view. Imagine that they were distributed as follows:
In this view, it made sense for some Begich voters to choose Peltola, as long as they were not too bound by partisanship. This being Alaska, that is quite plausible.
Still, this doesn’t account for the fact that 21% of Begich supporters didn’t list a second choice at all. There was considerable antipathy between Begich and Palin supporters.
A ranked-choice voting expert might say that Begich supporters threw away power of their vote - an “exhausted” ballot, in RCV lingo. Normally, exhausted ballots are considered bad, a sign of voter fatigue, or hostility to the new method.
But alternately, we might consider interpreting ballot exhaustion as a choice - and abandoning the idea of a single axis of polarization from left to right.
Consider Sarah Palin as an election denialist or even just as an outsider who’s been out of Alaska politics for too long. In this case, voters might have additional, more complex preferences. The landscape of voters might look something like this:
In this higher-dimensional picture, Palin has features that aren’t on the standard left-right axis - and those features were too far-out for some Begich supporters. In this case, their indifference between the Peltola and Palin might be a rational choice. Under this interpretation, the eventual outcome captured public opinion well.
However, that’s not the end of the story. An even more important test lies ahead.
It is said that a ranked-choice voting rule can shape the behavior of candidates by incentivizing them to build coalitions rather than fight one another. (Other voting reforms such as approval voting might do the same. I am not taking sides here…not without evidence!) For example, in last year’s NYC mayoral primary, Kathryn Garcia and Maya Wiley campaigned for each other's supporters, and it almost worked. Such coalitions can also affect governance. This is a potential benefit of ranked-choice voting, depending on the political landscape.
To improve their electoral chances, Begich and Palin might be well advised to reassess their strategy in time for November‘s regular election, when the seat is filled for a full two years. Will they encourage their supporters to list second and third choices, giving them more voice in the final outcome? We will find out.
Alaska and the November national election
The next question is what this race may mean for national politics. Special elections are prized by election geeks for their ability to take the temperature of actual voters in the run-up to a midterm election, without relying on polling data.
Before today, the immediately preceding five special elections were more favorable to Democrats than the 2020 election. Such calculations are possible using detailed precinct-level voting data for the Biden v. Trump race. (See this spreadsheet from DailyKos Elections). Does the Alaska election add a sixth? Some election analysts and commenters have noted this pattern.
However, I wouldn’t be so quick to conclude that the effect is quite so large. Peltola won the final round because voters were given the option to express complex preferences. National voters are not in a position to do so. In the first round of voting, 60% of Alaskans chose Republicans (Begich or Palin). By this measure, the outcome was actually more favorable to Republicans than the 2020 Presidential result, which was 53% Trump.
Even so, the median of the six special elections (one of which is Alaska’s) is still 5.5 points more favorable to Democrats than the 2020 election. That is fairly eye-popping, considering that Biden won the popular vote by 4 points. A total margin of D+9.5% would be unbelievable. And in fact, I don’t believe it.
A longer view: what do a dozen-plus special elections tell us?
Let us look at still more data. This year, there have been 34 special congressional and legislative elections. 15 of them have taken place since May 2, when the Dobbs decision striking down Roe v. Wade was leaked. In those elections, the median outcome was 4 points more Democratic than the 2020 Biden-Trump margin, for a total of D+8%. That’s still a lot, considering that about D+2.5% is necessary to retain control, thanks to geographic and gerrymandering advantages for Republicans.
A note of caution is in order. Although special elections are real elections, their turnout patterns may be different from the November election. Decision Desk HQ notes that the pro-Democrat advantage in special elections depends on educational attainment. In particular, the smallest advantage occurs in counties where fewer than 40% of voters have a bachelor’s degree. In those counties, Democratic overperformance is only 1 percentage point over 2020. That gets to D+5%, still enough to retain control, but it’s closer than what the overall special-election measure would predict.
Finally, for an even less-favorable measure, see the generic Congressional ballot, whose recent median I calculate as D+2.5%. That’s on the knife edge of going either way. This measure tends to move after Labor Day and is worth watching.
Democratic gains in the House this year would not be unheard-of. Charles Franklin notes that although the President’s party usually loses seats in the midterm election, there are multiple exceptions. In the modern age of political polarization, which I date to 1994, the President’s party has gained seats in 2 out of 7 cases.
So for control of the House of Representatives, there is more suspense than one would’ve expected a few months ago. The Alaska special election doesn’t tell us as much as we might want, but it adds to the picture. For now, it is true that movement toward Democrats in the last few months, as well as a smaller-than-expected gerrymandering advantage, have made this an open question.