It stands out like a fluorescent blinking neon sign from
summary data table
in the world's only (then)
study of range and Approval voting vis-a-vis USA third-party presidential candidates
range voting gives lots more votes to every third-party
candidate (but in the 2004 election, still not enough to
prevent them from losing)
than does either approval voting or plurality voting.
[Similar results also arose in other studies, e.g.
Beaumont Texas 2006,
Occupy Wall Street 2013,
Why is that? I believe I know the underlying reason for it, and it has profound consequences.
For reasons that will be explained later, we call this the "nursery effect."
The reason is that a lot of range voters think thusly:
"I don't believe most of these third-party candidates have any chance to win.
So while maybe I'm going to exaggerate somewhat by giving 0s to the
candidates I think have the most chance of winning but whom I dislike
– just to try to make maximally sure they don't win – on these no-hope candidates
I can afford to express my undistorted honest opinion. So I'm not
going dishonestly to give them 0 or 99. I'm going to honestly give them the
score they deserve."
If the rules were changed and those voters were
forced to vote approval-style (equivalent to range voting but
where only 99s and 0s are permitted – intermediate values are illegal with AV)
then those same voters would, in the vast majority of cases, just give
those third party candidates zero.
That is what experimentally happens. Even with 0-100 range voting where it was certainly
not forced by the rules, by far the most popular
score, was flat zero. (In fact there averaged about three zeros per ballot.)
The experimental fact is range leads to huge vote-count gains
over approval voting, for all third-party candidates.
(And with plurality, third-parties can
really forget it – nobody wants to waste their precious single vote on a no-hoper.)
Now let us talk about the profound consequences of this observed voter behavior.
Because it generates far more votes for third-parties, that gives those third parties
huge reason to unify
behind range voting. With that unification accomplished,
enough power is generated
to actually give voting reform a chance of happening! Note that this unification
cannot happen behind approval voting, so by mistakenly pushing it, voting
reform advocates would be going down a no-chance dead-end road.
Once third parties start to become large, powerful, organized, experienced, moneyed, etc
and hence are perceived as
actually having good chances to win, e.g. once they reach over-20%-of-the-vote popularity,
then range voting is going to stop giving them
a big vote-count advantage versus approval voting. Why? Two reasons. First,
voters at that point will consider them important enough to be worth
strategic exaggeration (99 or 0 score) in their vote. This will cause RV and AV to
act similarly with respect to that candidate. Second, it is simply mathematically
impossible for, say, a highly popular candidate like
Nader (who, our polls said, would have got 21%
in the 2004 election had it been held with approval-style voting) to
get amplified by a factor of 5 by switching to range. (In fact, range only pushed
him up only to 25%.) In contrast, it is
easily mathematically possible for Badnarik (the Libertarian candidate,
with only a fraction of 1% support under Approval Voting),
to get amplified by a factor of 5 by switching to range. And in fact, our study
indicated range amplified him
by a factor of fifteen.
That is why we call it the "nursery effect." Small third parties are like infants.
They are small, weak, and inexperienced.
They cannot survive in the jungle. They need shelter and
care. Range voting gives them that nurturing so they can grow up
instead of being stomped on by an elephant. They get the benefit of all
the voters who like them honestly expressing their opinion. That can easily
bring them up to 10-20% support level with range voting. Once they have that
support level, then they can attract donors, and supporters, and candidates to
throw into the election gauntlet. So they can grow. But under approval voting,
they would be getting 2% or less
voter support. They would have a much harder time
attracting donors, supporters, and candidates. So they would just die in infancy.
The effect of range voting on third parties is like being sheltered in a nursery.
But the effect of approval voting on third parties is: too-small parties
get stomped on by elephants.
Note that once they grow up and get big and tough, RV throws them out of the nursery.
It is even possible that this nursery effect is so powerful that it would
make approval voting
lead to permanent2-party domination,
just like plurality and IRV.
We do not know that. What we do know, is range voting has a
heck of a lot better chance of not doing that.
Also, removing the
range voting hurts third parties
because fewer voters are knowledgeable
about third parties, hence more voters X those scores ("no opinion")
and if those Xs were no longer treated as Xs, but instead as the lowest
possible score 0, then that would hurt the third party candidates. Meanwhile
almost all voters have opinions about the major-party candidates – so that for
those parties this X policy choice has almost no effect.
Note that with range voting (especially with blanks),
voters have the opportunity to express their
honest opinion. In many cases the strategic incentives are very small for them
to dishonestly distort that opinion. This is good. The more honesty and the less
dishonest strategic distortion there is in voting, the better results happen for society.
Simulations show this quality-improvement massively clearly
(see paper #56 here
or our Bayesian Regret discussion or our brief
In a nutshell: range voting==>gives voters opportunity to express
honest opinion with often-small incentive to be dishonest==>more
honesty among voters==>better
election results for society==>improved world.
Approval voting, in contrast, is telling voters, "throw your honest opinion
that Joe Candidate should score 37
in the garbage. We aren't interested and refuse to listen." As a result,
you get significantly worse election results for society and a worse world.
The amount by which the world becomes worse can be and has been
This quality advantage of range versus approval is maximized in elections
in which there are many candidates (say 10) and a comparatively large amount
of voter honesty. The Iowa 2008 caucuses
meet these two criteria, as did the French 2002 election.
That is one reason we need to push for range voting in Iowa 2008.
And there also are even more important reasons for that.