A lot has been written on the Later-no-harm Criterion (LNH). FairVote argues that it’s very important, whereas Warren Smith, the Princeton math PhD who created the modern movement for Score Voting, argues that it’s actually harmful. Emily Dempsey, an advocate for STAR Voting, has her own argument against the importance of LNH.
I want to argue against LNH from a generally different viewpoint entirely. The biggest problem with LNH isn’t about finding compromise, it’s about ignoring voter preferences, both in choosing a winner and in reflecting support for non-winning candidates.
Later-no-harm means IRV 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 IRV (“RCV First Choice Preferences”). 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.
Occupy Wall Street
We saw the same thing in a 2012 exit poll conducted by Occupy Wall Street. Plurality voting and IRV both suffocated third parties.
It’s bad enough that later-no-harm prevents transparency of support for non-winning candidates. But even worse, it can ignore voter preferences that should affect the winner. Consider these hypothetical preferences.
C is eliminated first with 31%, and L wins with 51%.
But suppose just 2%, from the CLR faction, swap their LR preference, creating:
C is still eliminated, but now R wins with 51%.
Summary: A tiny change in preferences changes the winner from L to R.
But NOW suppose a HUGE change in preferences occurs going in the opposite direction. The LRC voters find out something terrible about R, causing them to lower R to 3rd place. And the RCL voters find out something super positive about L, causing them to elevate L to 2nd place.
An enormous shift in public opinion takes place, positive for L and harmful for R. But this huge change doesn’t change the winner back from R to L, even though a tiny change of preferences moved the winner from L to R in the first place.
Bait and Switch
One of the most incisive points in Warren Smith’s article mentioned above is this. Imagine a criterion called Honest Second Choices (HSC), which a voting method satisfies if voting for lesser preferred candidate can never give the voter a worse result. This is what LNH advocates should really be after, and presumably what most of them think LNH means in the first place. But it doesn’t, because the harm in “later-no-harm” is to the candidate. But strategic voting is about the voter looking out for her own best interest. And indeed, RCV/IRV can absolutely cause a voter to get a worse outcome by sincerely ranking additional choices on her ballot.
But it’s mathematically proven that no voting method can simultaneously satisfy HSC and make it safe for a voter to support her sincere favorite, as score voting and approval voting do. So you get to pick one. Either it’s safe for the voter to support her true favorite, or it’s safe for her to support her second/third favorites.
The real problem with LNH is that in order for a voting method to pass it, that voting method must necessarily ignore certain preference changes in an arbitrary way.