A Better Democratic Primary
How better voting methods could lead to a better nominee
In early 2016, many thought Trump couldn’t possibly win the Republican primary. Even as he came a close second in Iowa and then won in New Hampshire and South Carolina, polls showed that a majority of Republican voters disapproved of him and had several other candidates they preferred. How could he possibly win if most voters opposed him; surely, the race would soon narrow to two frontrunners, and Trump would fall behind.
Of course, as we now know, having numerous opponents turned out to be not a weakness, but a strength. He kept his opposition divided, and as he continued to rack up victories, he became the presumptive nominee. Many of the Republicans who had previously disapproved of him fell into line, and he ended up with a handy majority in the last few contested states (though still just 45% of the popular vote in the primary as a whole).
This divide-and-conquer dynamic was not inevitable, but neither was it just good fortune. It was the consequence of a choose-one voting system, also known as plurality voting. By restricting voters to support only one of the options, this system is almost designed to make the most controversial candidate into the winner. If you can say something that makes 40% of the voters love you and 60% of them recoil, that actually helps you win, as long as the 60% remain divided among several competing options. That is, of course, the environment Trump instinctively thrives in.
For 2020, we can already see that there will be many candidates for the Democratic nomination. If we don’t want to nominate a Democratic version of Trump — if not in 2020, then maybe in 2024 or 2028—we need to reform the primary voting system, moving to something that encourages more consensus, and thus expands, not contracts, the Democratic “Big Tent”.
I’m going to analyze 5 possible voting method reforms for the Democratic primaries, explaining how they’d work and their relative advantages and disadvantages. These 5 possible reforms are: single-winner Ranked Choice Voting (aka Instant Runoff Voting, or the Alternative Vote); Approval Voting; STAR voting; and 3–2–1 voting. It’s important to note that every one of these would be a huge step up from the current choose-one system. Even their worst “disadvantages” relative to each other are tiny relative to the crushing problems with choose-one.
It’s important to note that every one of these would be a huge step up from the current choose-one system.
Throughout this article, I’m going to assume that the overall structure of the primary stays the same: individual state-by-state primaries spread over weeks or months. Reforming that overall structure is a promising idea, but it’s beyond the scope of this article.
Note: I won’t explore much of the voting theory of proportional delegate allocation or transfer of delegates from candidates who withdraw; that would be a separate article.
Ranked Choice Voting (RCV)
Though not as simple as approval voting (explained next), Ranked Choice Voting is probably the best-known of these reforms. To vote, you rank choices in order of preference. To count those votes, you first tally up top choices. If no candidate has a majority, you eliminate the candidate with the lowest tally, and move their ballots to tally with each voters’ next-highest preference (eliminating or “exhausting” any ballots that have no more preferences). You continue such eliminations until some candidate has a majority of the remaining ballots, which will always happen at or before the time that just two candidates remain.
When using RCV for a multi-state primary system, you could in theory give all the state’s delegates to the winner. It would probably be fairer, however, to give them a portion of the delegates proportional to their final winning percentage, and divide the other delegates according to the remaining ballots’ first choices.
RCV would allow voters to be far more honest than under plurality. Votes for minor candidates aren’t wasted, they’re simply transferred to later preferences; so voters could afford to show support for non-frontrunners, and the relentless pressure to drop out of the race would be reduced.
Nevertheless, like our current choose-one method, this system would be biased towards celebrity candidates, and away from centrists. Candidates with high name recognition would rarely be eliminated early, and thus they’d pick up votes from second- and third-place preferences in these early elimination rounds. Then, when the race got down to 3 candidates, often the center of the three (which, in a Democratic primary, would mean the center of the party, not the center of the electorate as a whole) would be “squeezed”, losing votes on both sides, and thus be eliminated. Because of this “center squeeze” problem (which has occurred in several historical RCV elections such as in Burlington, VT; British Columbia; and Queensland, Australia) RCV can thus elect more extremists than the other options below. Remember, though: the current choose-one system has an even bigger problem with extremists.
In the 2020 Democratic primary, center squeeze would likely take the form of a Biden-versus-Sanders race, with the other candidates prematurely eliminated from serious contention. Personally, I believe that this would be bad for the party. Even if Biden or Sanders is the ultimate winner, I think it would be good to have a system where supporters of other candidates feel that their enthusiasm has gotten a fair chance, so that as many people as possible are brought into the race.
Approval Voting
This is by far the simplest voting reform. Voters approve as few or as many candidates as they want; the candidate with the most approvals wins. That’s it — for a single-winner race.
For a series of primaries, though, the goal is not just to choose a single winner, but to assign delegates. Thus, the candidate X with the most approvals should get delegates proportional to their percentage approval. Then, among the remaining votes that didn’t approve them, the next candidate Y with the most approvals should get delegates proportional to the percent of ballots that approve Y but not X. And so on, until all the delegates are assigned.
One principal advantage of approval voting is that there’s never a good reason not to approve your true favorite candidate. Although the complications of the multi-state primary would erode this guarantee in theory, in practice there would almost never be any clear reason not to vote honestly in this sense. This means that “center squeeze”, as explained above for RCV, would not be a factor; almost certainly, enough voters from both sides would extend approval to the center, so that the ultimate winner would probably be someone with as broad as possible of a mandate.
The main downside of approval is not in terms of outcomes, but of feelings. As a voter, you would be able to approve your true favorites and oppose your true enemies, but for those compromise candidates in the middle, things are a bit tougher. Should you approve the compromise candidates, thus putting them on the same level with your favorites? Or should you reject them, thus putting them on the same level with the very worst candidates? In this “compromise dilemma”, either choice is unsatisfying, and while harmony and good feelings is one possible outcome, infighting is too.
Remember: this is still strictly better than our current choose-one system, where that infighting and division is virtually guaranteed. But the systems below attempt to resolve this and reduce the problem.
In the 2020 primary, the “compromise dilemma” would be toughest for the supporters of 3rd- or 4th-place candidates. If they support only their true favorite, they risk being irrelevant to the final choice; but if they also compromise on approving one of the top 2 frontrunners, they give up on helping their favorite to pass that frontrunner. As I write this today, polls suggest that Harris and Warren are in 3rd and 4th place; and I wouldn’t want their supporters to sour on the party.
STAR voting
The name STAR voting refers to two different things. First, it’s an acronym for “Score Then Automatic Runoff”. But it’s also a reference to the ballot format, which lets voters give each candidate a 0–5 star rating. Once all the ratings are tallied, the two candidates with the highest average ratings are finalists. The finalists are then compared: whichever is rated higher on more ballots, wins. So if you gave 5 stars to candidate X, and 4 stars to candidate Y, your ballot could help both of them to be finalists; but in the runoff, it would count as 1 full vote for X over Y, no different than if you’d given 0 stars to Y.
In the context of a primary, you’d have to assign delegates, not just choose a winner. So, as with the systems above, you’d give the winner delegates proportional to their percentage in the final runoff, then give the remaining candidates delegates proportional to first choices among the ballots that voted against the winner in that runoff.
Because your vote always counts at full strength in the runoff, you as a STAR voter have the luxury of showing your true preferences. In most cases, this method avoids the “center squeeze” problems of RCV, while freeing voters from the dissatisfying compromise dilemma of approval.
However, though “center squeeze” is much less probable than under RCV, it’s still a possibility. Voters who feared the opposite extreme might win would have to be careful to give enough stars to other candidates to ensure that a strong compromise would be among the two finalists. If voters were too uncompromising, the finalists could be two extremists, even in cases where there was some candidate in the middle who could easily have beaten either side.
As I said above, the “center squeeze” scenario in 2020 would probably mean a race that narrowed down to Biden vs. Sanders. Of course, if one of these two is ultimately the favorite of a majority, then good for them; but if they only win by squeezing out another candidate who could have beaten them one-on-one, then the primary isn’t doing its job. I think that this kind of failure is unlikely in STAR, but it’s not impossible.
3–2–1 voting
In this system, voters rate each candidate into one of three categories: “Good”, “OK”, or “Bad”. First, three semifinalists are chosen: the candidates with the most “Good” ratings. Then, two finalists: the semifinalists with the fewest “Bad” ratings. Finally, as in STAR, the winner is the finalist who is rated higher on more ballots.
As with the other methods above, delegates would be assigned to the winner according to their percentage in the runoff, then to the other candidates according to their support on ballots that voted against the winner in the runoff.
In most cases, this method is very similar to STAR, and thus shares the same advantages: allowing voters to honestly express their key preferences. The main difference is that it’s even more resistant to “center squeeze”. Since there’s 3 semifinalists, that will usually include at least one centrist option; and as long as it does, that candidate will almost certainly make it to be a finalist, winning if they get a majority in the last step. The main disadvantage of 3–2–1 is that it’s relatively unknown and untried, having never been used in any high-profile election.
In the 2020 Dem primary, I think that 3–2–1 would have a healthy variety of options in the top 3, and the most legitimately-popular of those three would win. Of course, this is purely hypothetical, because the odds against this reform being implemented in time for 2020 are beyond astronomical.
Final words
I’ve laid out several possible voting method reforms for the Democratic primary. I think any of them would lead to a less-divisive process and a stronger nominee.
