[EM] IRV vs Plurality

[Jameson Quinn]

2010/1/8 Kevin Venzke <stepjak at yahoo.fr>

> Hi Dave,
>
> --- En date de : Ven 8.1.10, Dave Ketchum <davek at clarityconnect.com> a
> écrit :
> > I said "approval", not
> > "Approval".  I read Range ratings of A-1, B-9, and C-2
> > as saying B is much more strongly approved than A or B.
>
> You are looking at the meaning of Range ratings on a ballot, but you
> don't seem to care how Range uses this information. That's what I'm
> talking about.
>
> Kevin Venzke
>


If I understand you correctly, I think that this is an important point, and
should be expressed more clearly. It is, in my opinion, one of the ONLY
valid arguments against Range, and yet it is (again, my opinion only) in
some cases a telling one.

The point is that Range encourages naive voters to vote non-strategically.
If even sophisticated people like Ketchum's can laud range ballots'
expressivity, without noting that strategy could limit this expressivity, we
can imagine voters, often far less sophisticated, will fail to understand
range strategy at all. Some voters will, strategically, vote approval-style,
and other voters will NOT.

A range proponent might argue that this just means that Range will, in
practice, perform somewhere between the (excellent) theoretical Range result
and the (pretty good) theoretical Approval result. Indeed, simulations show
that, for any *randomly distributed* constant proportion of strategic
voters, Range performs well, and in fact outperforms most other methods
(including Condorcet methods) given the same proportion of strategic voters.
However, this argument assumes random distribution. Such an assumption is
not only unjustified in the typical election; in fact, there are grounds for
assuming the contrary. Voting groups are, by definition, ideologically
distinct, and thus there is every possibility for their propensity for
strategic voting to be distinct. If groups do thus vote strategically at
different rates, Range can give seriously deficient results - with
sufficient strategy gap, even the worst candidate can win. (Condorcet
methods, on the other hand, will tend to still elect the Condorcet winner
despite any strategy, unless results are close and/or some voters use a
specifically misguided strategy).

I think it would behoove Range advocates to think about these issues,
because one horror story of this kind could be more obviously unfair than
many of the worst pathologies of other systems.

However, there are forms of Range voting which would avoid this problem.
Most obviously, there is Approval itself, which is technically a form of
Range. That can be extended to any version of Range which disallows or
strongly discourages most "weak" votes. For instance, there could be options
to vote only 0 ("worst"), 1("bad"), 50 ("intriguing"), 99 ("better"), and
100 ("best"). 50 is labeled as "intriguing" because the option is intended
particularly for little-known candidates whom a voter believes, but is not
sure, deserve support, not for well-known candidates who are simply of
average quality. This would give additional expressiveness beyond Approval,
without a significant risk of pathological situations where a strategic
minority imposed their will over an unstrategic majority. For instance: in a
standard "two frontrunner" scenario, where no voter was uninformed about
either of the frontrunners and so nobody rated either as merely
"intriguing", such a system could not give the "wrong" strategic-plurality
winner unless the gap were less than 1% (ie, closer than 49.5%-50.5% of the
two-way vote).

Jameson Quinn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20100109/d728e34c/attachment-0004.htm>