Retired psychologist and League of Women Voters member
Jeanne W. Eisenstadt with a psychology-based criticism
which appears largely invalid although initially it may sound plausible.
Somewhat related criticism by M.Balinski & R.Laraki
based on "measurement theory." We completely destroy it.
However, there are a few genuine criticisms of range voting.
They all seem to be related to strategic voters.
Range voting actually reacts very pleasantly if all the voters
act strategically; one can prove Range Voting
maximizes pleasant surprise
elects "beats-all" winners whenever they would exist
with honest voters
(both under plausible assumptions about how the strategic voters behave).
And range voters in 3-candidate elections can always be "semi-honest"
(i.e. never say A>B if they believe B>A)
without ever sacrificing any strategic oomph; and range voters can avoid ever
rating their true favorite below somebody else – again
without ever sacrificing any strategic oomph.
We're unaware of any comparable package of theorems showing nice behavior with strategic
voters for any non-range voting system.
Also, strategy in range voting is
compared to in many other systems;
see puzzle 40
for a theorem showing that that the best strategy for group of co-feeling voters
can fail to be "all vote the same way" in Approval, Borda, Schulze-Condorcet,
and IRV voting, whereas with range voting the best strategy for the group is
always simply to pick the best vote, and then everybody casts it.
So what, then, is the criticism?
The problems for Range Voting arise when either
Not all of the Range Voters – only some –
vote strategically; and in a way correlated to their politics.
If for example, voters preferring A over B express intense
opinions, but those preferring B over A express mild opinions, then
A can win the election, thus "penalizing" the B-voters for their "honesty."
To this we reply that it is naive to believe that only one side will employ
strategic exaggeration in votes; if one does, the other will too. Further, even if that did
happen, it would tend to be self-correcting over time. Further, the mild-opinioned
voters have little basis for complaining, since their preferences were only mild
and since they voluntarily chose to express them. (The more mild their opinions, the more
likely such a strategy will work against them – but the less it matters!)
Here is a succinct statement by Professor T.N.Tideman
of the criticism, and responses by Kok
and Lomax, and finally a response based on
ongoing computer simulations, which seem
to prove that (surprisingly) this whole concern is actually not a concern; even if
the worst fears of the critic here come true, then range still is superior
to IRV, Condorcet methods, etc. Also see the
other set of computer simulations.
Or when the "plausible assumptions" we alluded to above
are violated – because the voters try to be strategic but
do not have enough knowledge to do so intelligently enough.
Chris Benham has given a nasty example
illustrating how range voters (by trying to be strategic without enough
understanding of the situation) can backfire badly. Fortunately,
of Bayesian Regret
indicate that Benham-type scenarios arise rarely enough, and/or
are not severe enough often enough, so that Range Voting still seems to be
the best method among all common proposals on average.
Also, in our detailed discussion of the Benham example
we explain why it depended on ill-informed voters, and show that
if those voters had been a little better-informed, they would not have fallen
into Benham's strategic-voting trap. An even-nastier trap is, however,
stated which is not so easily escaped, but it is more complicated than, and presumably
a lot rarer than, Benham's example.
Finally, one can argue (with at least some validity) that range voting
too-honest voters more than it victimizes them.