later-no-harm

clay shentrup
4 min readJan 2, 2020

many people believe and/or claim that ranked choice voting or “RCV” makes it safe to rank additional choices after your first choice, due to the fact that RCV satisfies what’s known as the later-no-harm criterion. but this is simply false. here’s a simple example with three groups of voters with their preferences ordered as follows.

35% mary-nick (i.e. mary in 1st, nick in 2nd)
33% sarah-nick
16% nick-mary
16% nick-sarah

nick is eliminated, and then the winner is mary. observe that the sarah voters—who have nick as their second choice—were harmed by ranking nick in second place. that elected mary, their third choice. a better strategy would be to rank nick in first place, because then he wins, and those sarah voters get their second choice instead of their third. so this claim that RCV makes it safe to rank your second choice is demonstrably not true.

and in case you thought this was contrived, this example is based on the 2022 alaska house special election. the candidates were sarah palin (r), nick begich (r), and mary peltola (d).

strategy

let’s order the potential ballots a palin supporter could cast in this election, from most to least effective.

  1. begich-palin (causes begich to defeat peltola)
  2. palin-begich (helps begich defeat peltola only if palin is eliminated before begich)
  3. palin

shell game

RCV advocates argue that the best strategy is #2, because #2 is better than #3. but of course that doesn’t make sense, since #1 is even better still.

the subtler part of this shell game is that the RCV advocates focus on what’s safe for the voter’s favorite candidate, when strategy is really about what’s safe for the voter themself. for instance, third party supporters usually vote for their lesser evil between the two major parties—showing that strategy is not about protecting your favorite candidate, but about protecting voter—yourself.

the cure is worse than the disease

perhaps even more notable is that this property of RCV—that a vote for your second choice doesn’t count until and unless your first choice is eliminated—is the whole reason the alaska election had an unreasonable anti-majoritarian outcome. begich (r) was preferred to peltola (d) by a fairly substantial 5% margin, and 60% of voters favored a republican. yet peltola, the democrat, won. this happened because RCV completely ignored the second place rankings of palin voters, who favored begich to peltola 10-to-1.

the intention of the design of RCV is to ensure that palin voters can’t harm palin by ranking begich as a second choice. and this can sound great on the surface, but it creates unintended consequences! in order to prevent those begich votes from being used against palin, RCV just ignores them. but this means those begich votes also can’t be used against peltola! and that’s why we got this whole mess in the first place.

approval voting

RCV advocates often use this later-no-harm argument to claim that the competing approval voting method—where you simply pick all you like, rather than rank candidates—would lead to voters simply “bullet voting” for their favorite candidate, and no one else. after all, why approve your second favorite if that might cause your favorite to lose?

after the lesser evil

but when we consider strategy, the issue becomes, “what do you do after you’ve strategically voted for your lesser evil?”

with RCV, palin voters who rationally/strategically vote for begich simply can’t give palin their full support. the best they can do is rank her second, just behind begich.

but with approval voting, a palin supporter who tactically approves begich as the “lesser evil” still has every incentive to also vote palin (and, in principle, anyone else they prefer to their lesser evil strategic pick).

and it’s for this reason that the truth of the matter is the exact opposite of what RCV advocates claim.

transparency

in addition to the matter of electing the right candidate and mitigating the negative effects of strategic voting, later-no-harm also means RCV lacks transparency. look at these results from an online polling experiment using several different voting methods.

elizabeth warren wins an outright first-round majority, so later-no-harm ensures we never even see the 2nd, 3rd, etc. preferences of any of the voters with RCV. the results look almost identical to the status quo in the upper-left. it appears that vote splitting between the four black candidates (harris, rice, demings, and bottoms) has artificially lowered their apparent support. meanwhile, STAR voting and approval voting reveal a much different story. warren still wins, but it’s not a landslide. this is what happens when you solve vote splitting, which RCV fails to do.

and here’s what happened with maine used RCV for their 2020 presidential election. look at those pitiful numbers for the minor partie candidates like jo jorgensen of the libertarian party or howie hawkins of the green party.

occupy wall street

we saw the same thing in a 2012 exit poll conducted by occupy wall street. plurality voting and RCV (aka “IRV”) both suffocated third parties.

approval voting and score voting resulted in a dramatic increase in support for third parties.
IRV aka RCV barely made a difference in showing third party support.

conclusion

properly understoond, later-no-harm is a bug rather than a feature.

--

--

clay shentrup

advocate of score voting and approval voting. software engineer.