[EM] Why is wikipedia so biased pro-IRV?

Jameson Quinn jameson.quinn at gmail.com
Thu Feb 24 23:46:08 PST 2011


> There is a bit of research supporting that view which is quoted in the
> Wikipedia piece "John J. Bartholdi III, James B. Orlin (1991) "Single
> transferable vote resists strategic voting,"<http://www.isye.gatech.edu/~jjb/papers/stv.pdf>"
> I can't say my understanding of it is all that great, but it seems to be
> based on computer modelling of elections to find out how often tactical
> voting can make a difference. My questions are, I suppose, how good is this
> research? Is there other research with computer models that contradicts
> this result? How well can we quantify the differences?
>

 I just read the abstract of that paper, and I believe that I understand
what it is saying. Here's my understanding.

It's not that they actually did any computer modelling of any elections.
What they did was mathematically prove that, roughly, even if you knew
everyone else's ballot, there are some possible situations where even the
most powerful supercomputer could not tell you your best strategy, and could
not tell you whether the election was nonmonotonic.

That's a mathematically interesting, but practically meaningless, result.
For one thing, for such an "unreadable" election to happen, there must be a
significant number of serious candidates. With, say, "only" 7 serious
candidates, a computer could check the result of 5040 possible ballot
orderings in a small fraction of a second. With 3 serious candidates, you
can essentially do it in your head, or, with a little practice, intuitively.
For another, it doesn't matter if strategy is sometimes hard
("NP-complete"); it matters if it's sometimes easy. If I can strategize
today, I don't care if there might someday be a situation where it's too
confusing for me to figure out a strategy. The two Burlington examples prove
that a simple strategy could repeatably work in real life.

Strategy is not the only problem with IRV, and perhaps not even the most
serious one. Even with all honest voters, IRV can give bad results, and
systematically disadvantage centrist candidates. But strategy is a problem.
In a world of stable party identification and 2 or 3 major parties along a
clear one-dimensional ideological spectrum, it would be easy to
overcompensate for IRVs honest flaws and entrench two parties, as
unassailably and unfairly as with plurality.

JQ
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20110225/2705f1b5/attachment-0004.htm>


More information about the Election-Methods mailing list