Originally posted on June 23, 2021. Reposted here without edits.
Yesterday’s Democratic primaries held in New York City employed a novel voting system entitled ranked choice voting. Specifically, instead of opting for a single candidate for a given office, voters were given the opportunity to select up to five choices, ranked in order of preference. Vote counting is then conducted in a series of rounds, each governed by the following loop1:
- If any one candidate has more than 50.0% of the vote, they are declared the winner and the count is stopped.
- Otherwise, the candidate with the lowest current vote count is knocked out. If the associated ballots list another candidate that is still “alive” in the race, they are “reassigned” to that candidate and the vote tally is recomputed.
In principle, this mechanism is supposed to ensure that votes for minority candidates are not wasted, but instead are reassigned to more popular second, third or even fifth choices – depending on how many candidates and rounds of vote tabulation there might be. Proponents of the system suggest that, on the one hand, this helps further democratise the electoral process, while also encouraging candidates to band together into quasi-blocs or “factions”, such that at least one of them would ultimately amass enough votes, even by virtue of being someone’s second or third – or fifth – choice, for the entire “faction” to remain competitive.
At least, that is how it is supposed to work in theory.
For some evidence of how this system works in practice, we turn to a recent Politico story covering the New York City mayoral race2. Roughly a third of the way through, the article includes a quote from a certain Alex Clemens, described as a political strategist and lobbyist, which states that out of the 429 ranked choice elections held thus far in the US, in 414 cases the candidate with the most first-place votes ended up winning. In other words, most of the time there was no difference in outcomes between ranked choice and conventional voting contests. One might present several hypotheses as to why this might be the case:
- Overwhelming advantage. Plenty of elections, for example in constituencies with long-time incumbents and entrenched party machinery, are never even close to being competitive. Adding ranked choice optionality to a contest where the leading candidate is bound to receive 70%-80% of the votes is more than superfluous.
- Too few candidates. Quite obviously, if there are only two candidates competing for a given seat, the addition of ranked choice accomplishes nothing.
- Too few viable candidates. It is conceivable that an election may include a number of marginal candidates, say, polling at under 2% of the total vote. In this case, transferring a fraction of this 2% to another candidate will often have only a negligible impact on the race as a whole. As well, transferring votes between several 2%-type candidates is still unlikely to aggregate enough votes to compete with vote leaders, who might have started out at 20%-30% or higher.
- Factional splits. Simply put, in order for ranked choice voting to work, there must be, in the voters’ minds, some viable alternative to their first choice. A wide enough rift, for example between a political party’s left and right wings, might produce a situation where a number of voters prefer their chosen candidate and only their chosen candidate, with all or most of the alternatives viewed as unpalatable.
There may well exist other explanations as well, and testing them would entail going through the specifics of every single one of those 429 electoral contests. I am perfectly happy to let others take up this particular torch; in the meanwhile, however, I would also share my personal impressions of how the system did, and did not, function specifically in the context of New York City Democratic primaries.
To begin with, let me outline, without identifying specific races or candidates, the decisions that I was faced with when filling out my ballot:
- In each of two separate races, there was only one candidate, and a marginal one at that, whose programme came close to what I would term as palatable. Voting for this individual alone was a waste of a vote, however ranking the other candidates, all of whom I found to be wholly unsuitable, was a complete non-starter. One could not even fall back on the good old “lesser evil” strategy, as they were all different shades of “evil” insofar as my personal politics are concerned. In a case such as this, ranked choice accomplishes very little.
- A couple of other races presented a different problem, namely that most of the candidates were both acceptable to me politically and strikingly similar to one another on matters of policy. Here, ranked choice worked best, I felt – when one was ranking five nearly identical candidates against one another to make sure that one of them, any one of them, ultimately made it into the top ranks.
- In one instance, a local race did not use ranked choice voting at all. One presumes this was some sort of an incidental – or a deliberate – legislative oversight. Nevertheless, to go through the entire two-paged ballot of ranked choice races only to see a simple first-past-the-post contest at the end felt somehow incongruous.
As well, it struck me that, mathematically, ranked choice votes only matter if at least one of the ranked choices is a top tier candidate. Thus, in the Democratic mayoral primary, where pre-election polls showed three of the eight leading contenders to each have under 3% support – and preliminary election results, before giving account to ranked choices, have them at 2.2%-2.8% – ranking those three and only those three names was still the equivalent of a wasted vote. The only way to take advantage of ranked choice would be to include a fourth name from among the two or three leading contenders for the party’s nomination, even if this fourth individual carried no political appeal whatsoever to adherents of any of the first three “marginal” contenders.
What one ends up with, then, is a system designed to funnel supporters of marginal candidates to leading contenders – who seem to win most of the time in any case – so as to inflate their vote tallies. A cynic might go even further and suggest that ranked choice simply gives the voters more to do without changing the underlying political or electoral picture. In any case, genuine utility seems to arise only in those instances when there are multitudes of equally acceptable, and roughly equally polling, individuals in the race, or when there is a clear and even split between two blocs. One suspects these cases to be relatively infrequent, however, at least if practical results referenced by Politico are anything to go by.
To an extent, one even struggles to define the exact problem that ranked choice is supposed to be solving, or at least one that cannot be resolved by simpler and more efficient means. If, at the end of the day, all ranked choice accomplishes is the channelling of votes from marginal candidates to one or two vote leaders, then why not simply have a second round run-off instead? And if ranked choice is meant as a bridge between a first-past-the-post voting system and proportional representation, then why not…simply switch to proportional representation?
In this sense, it is also instructive to look at some of the critiques of ranked choice as employed in New York City that have emerged within the past 24 hours. For example, Politico has published another stor 3 on the city’s mayoral primary where the candidates were blamed for not forming blocs or coalitions and instructing their voters to vote accordingly. But this assumes that a bloc actually exists, and that at least some of the candidates are similar enough to one another in terms of policy and political appeal. Moreover, as noted above, someone in the ranked choice list must necessarily be from among the leaders, which leads to two separate questions – why is a voter’s first choice a marginal candidate and not one of the leaders in the first place, and why should said leader seek the support of the consituents of said marginal candidate. Mathematically speaking, of course, it is perfectly possible to construct just such a race where blocs of sufficiently similar candidates exist, and where pooling “bloc” votes together produces a tangible impact. Moreover, if enough elections are run under ranked choice rules, someone, somewhere, is statistically bound to reproduce in reality just such a race.
And yet, in 414 out of 429 races none of it mattered. Nor did it matter for me personally in the two races where I basically was opposed to all save a very distant “lesser evil” candidate, because of who survived the gauntlet of the local party machine to actually compete.
- A basic WHILE-DO statement, for those familiar with programming languages.[↩]
- Durkin E., Giambusso, D., “‘No one is gonna steal the election from me’: Echoes of 2020 in NYC’s mayor’s race”, Politico, June 22, 2021, retrieved June 22, 2021.[↩]
- Heath, R., “How NYC messed up its mayoral election”, Politico, June 22, 2021, retrieved June 22, 2021.[↩]