Re: delayed runoff (done instantly) as research tool / question
- --- In RangeVoting@yahoogroups.com, "warren_d_smith31" <wds@...> wrote:
> In1) I think we have, to a large degree, answered the riddle of why PRV
> delayed runoff, it REALLY is strategic to vote differently
> than plurality strategy in round 1 - BUT how??
> Your advice solicited...
(plurality-runoff voting) breaks two-party domination, while IRV does not.
2) The rabbit hole goes deeper, as you are pointing out. With PRV,
there is never an incentive to vote strategically for anyone except
your first-place choice. But even then, there's only an incentive to
vote strategically if there are 3 or more
perceived-to-be-front-runners ALL OF WHOM are not your sincere favorite.
If there are just two front-runners, like the two from the
major-parties, and one is your favorite, then you'll be HONEST.
If there are just two front-runners, and neither of them is your
favorite, you'll be HONEST.
If there are three front-runners, and one of them is your favorite,
you'll be HONEST.
If there are three front-runners, and none of them is your favorite,
or if there are MORE THAN THREE front-runners, you'll strategically
top-rank your favorite among the front-runners, and then vote honestly
So, plurality-runoff seems to support tri-opoly, rather than duopoly.
- Ugh...I must apologize folks. I overlooked the whole scenario.
Here's a counter-example to my thought process:
% of voters - their order of preference
34% Bush > Gore > Dean
17% Gore > Bush > Dean
15% Gore > Dean > Bush
34% Dean > Gore > Bush
But maybe we can still learn something from this.
Here we see that, even with just three candidates, "IPR" would just
eliminate Gore, and give the win to Bush, just like IRV. Thus the
Dean voters would have a strategic incentive to switch to "Gore > Dean
> Bush", to get Gore instead of Bush. :(If anyone can offer any other useful insights though, please do.
--- In RangeVoting@yahoogroups.com, "brokenladdercalendar"
> So, plurality-runoff seems to support tri-opoly, rather than duopoly.
- --- In RangeVoting@yahoogroups.com, "warren_d_smith31" <wds@...>
>The relevant part of the message Warren refers to:
> --- In RangeVoting@yahoogroups.com, "brokenladdercalendar"
> <thebrokenladder@> wrote:
> > --- In RangeVoting@yahoogroups.com, "warren_d_smith31" <wds@>
> > wrote:
> > >
> > > The property that the (b) rankings will always be honest, makes
> > > this kind of voting system a possibly excellent research tool.
> > > We can use it to get honest opinions from pollees (!!).
> > > Then we can assess stuff like how badly some other
> > > voting system works.
> > > I can hardly believe this technique has not been used.
> > > It's very simple.
> > Back on Oct 8th, Benham said to me:
> > An example of a very bad method that in terms of desirable
> > properties is completely dominated by IRV is what I call "Top-Two
> > Runoff", that you in your piece somewhat oxymoronically refer to
> > as "Instant Delayed Runoff"
> --well, my point is NOT that "Instant Delayed Runoff"
> is necessarily a great voting method.
> My point is, voters are strategically motivated to be 100% HONEST
> in the "part (b)" rank-orderings used in the 2nd round
> to decide between the top 2 finishers from the 1st round.
> Benham said he did not see why these voters were going
> to be more honest than in any other voting method.
>--- In RangeVoting@yahoogroups.com, "warren_d_smith31" <wds@...>wrote:
>To me a "pollee" is someone who participates in an opinion poll, not
> The property that the (b) rankings will always be honest, makes
> this kind of voting system a possibly excellent research tool.
> We can use it to get honest opinions from pollees (!!).
> I don't understand why pollees would be any more or less honest with
> this system than with any other.
a voter in an election.
> To that I respond: "duh!!" Think about it.Yes, all this is obvious for voters in an election.
> Voters cannot predict ahead of time who the 2 top finishers
> in round 1 will be. If they provide in part b anything other than
> a 100%-honest rank-ordering then there is some chance it will hurt
> them. If they provide an honest ordering, it'll always help them.
> And there are no gradations of help (more vs less) - there is
> either help or hurt. Period. Well... all I can say is, if you
> do not find it obvious the only strategic vote is the honest
> ordering in part (b), then think again until it is obvious, and
> then kick yourself about 100 times for not seeing it
>What motivation do you think they have to lie in an opinion poll with
> Now. Why is this good? It is not necessarily a great voting method
> since the part (a) of the votes (used in the 1st round) still may be
> The reason it is good is this is a TOOL for us pollsters to use,
> FOR THE FIRST TIME EVER, to actually acquire honest rankings from
> people. Where they provably,and obviously, have no motivation to
some other method?
- Cycles are so expectable with Condorcet that you do not have a complete
method if it cannot handle such.
So, if you do not accept "Condorcet" as a method name, what title should
we use for an actual, usable method?
On Thu, 01 Feb 2007 06:36:50 -0000 brokenladdercalendar wrote:
> --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
>>Two red flag words - 'IRV' and 'runoff' together catch my eye, so I
>>out my IRV killer:
>> 35 A
>> 33 B>C
>> 32 C
>>If I understand what goes on:
>> Condorcet, by itself, declares C winner.
> Nope. 33 people say B > C, and only 32 say C > B. So B beats C. But
> 35 people say A beats B. But 65 people say C beats A. So it's a
> cycle. Condorcet methods are distinguished by how they resolve
> cycles, so "Condorcet" doesn't pick a winner here. Minimax, Schulze
> method, Tideman ranked pairs, etc. selects a winner.
Dave Ketchum 108 Halstead Ave, Owego, NY 13827-1708 607-687-5026
Do to no one what you would not want done to you.
If you want peace, work for justice.
- --- In RangeVoting@yahoogroups.com, Dave Ketchum <davek@...> wrote:
>Exactly. Condorcet isn't a method. It's a class of methods - the
> Cycles are so expectable with Condorcet that you do not have a complete
> method if it cannot handle such.
sufficient condition is that they obey the Condorcet criterion.
> So, if you do not accept "Condorcet" as a method name, what title shouldThere are zillions of Condorcet methods. I named a few.
> we use for an actual, usable method?