[EM] Approval reducing to Plurality

[Juho]

On Aug 30, 2010, at 3:42 AM, Kevin Venzke wrote:

> Hi Juho,
>
> --- En date de : Dim 29.8.10, Juho <juho4880 at yahoo.co.uk> a écrit :
>> De: Juho <juho4880 at yahoo.co.uk>
>> Objet: Re: [EM] Approval reducing to Plurality
>> À: "Election Methods" <election-methods at electorama.com>
>> Date: Dimanche 29 août 2010, 16h59
>> Yes, polls that are arranged before
>> the actual election is an interesting approach for Approval
>> since Approval may fail badly if there are more than two
>> potential winners in the actual election. Particularly this
>> could offer some help in the Burr dilemma set-up where e.g.
>> the right wing has two equally strong candidates.
>> (Approval may perform better when there are three candidates
>> tied in a triangle.)
>
> Yes I don't think the Burr dilemma scenario is one that Approval is  
> likely
> to choke on in practice. I think polls or parties will work it out to
> eliminate the dilemma, in the same way and for the same reasons that
> intraparty contests don't tend to survive up to election day in  
> Plurality
> elections. There is so little possible gain, and everything to lose.

If R1 and R2 are under the control of one party then the party can  
control the game, but if R1 and R2 come from two separate and  
competing parties, then it may be very difficult for one party to give  
up and donate the victory to their competitors.

>
> If a triangle type of situation makes it to election day (which  
> seems more
> plausible to me, if there's a cycle or something) I do agree that
> Approval will probably handle it better than a Burr dilemma, but  
> only by
> a little bit... Resolving the election somewhat arbitrarily as opposed
> to having good win odds for the worst candidate.

Yes.

Near ties are quite common in various elections nowadays. There is a  
tendency to get e.g. 51% - 49% results. Also in Approval situations  
where cycles are not clear but where the winner is not yet known are  
common. The set-up may always evolve all the way until the election  
day, voters will change their opinion, different polls may give  
different results, and candidates may make last minute moves (e.g. to  
react to what polls have shown, to attack the leader). In this  
situation Burr dilemma style asymmetric set-ups and more symmetric  
triangle style set-ups with near ties may be quite common. Also linear  
opinion space offers some interesting possibilities (strong extreme  
right wing candidate + weaker moderate right wing candidate).

>
>> The problems are however not quite over yet. One potential
>> problem is strategic answers in the polls. It would make
>> strategic sense to R1 supporters to start the polls by not
>> showing any support to R2. That could make R1 look better
>> than R2. And that could lead to a game with the polls, both
>> groups trying to present themselves as the stronger group.
>
> Yes but by the end of the day I just don't think that will be present.

It may be that playing with the polls is not widespread since people  
don't want to think that far. Bullet opinions might be more common in  
polls than at the actual election day when one has to make sure that  
some sensible candidate will win.

> Unless nobody actually cares about winning the race for the team, at
> some point before the election there needs to be a decision. If this  
> is
> really a problem I can imagine having primaries etc.

One thing that works against finding a consensus before the election  
is the optimistic attitude of the voters. It is typical that voters  
believe and hope that their favourite will win and will get the  
remaining needed supporters before the election day. They may also pay  
more attention to polls saying that their grouping/party is likely to  
win. It is also possible that the other right wing candidate is seen  
as their strongest competitor and therefore voters of both groups  
stick to the idea that it is not their group that should give up.

It may often happen that in some poll one of the right wing candidates  
is ahead of the other, and as a result the whole world starts thinking  
that that candidate is a "winner" and the other one is a "loser". And  
as a result the "winner" will win with a clear margin (although the  
game could have gone also the other way quite as well). But also a  
tied situation all the way until the election day is a common  
phenomenon.

>
> Really nothing new should be revealed on election day. If the outcome
> would be "Burr dilemma kills one side," that's exactly what the polls
> should be saying before the race. If they want to avoid that they need
> to try harder to consolidate. (I don't see rank ballot methods as much
> better here, incidentally. Even with marginal improvements in  
> ability to
> handle the situation, there is still very little to gain compared to  
> what
> can be lost.)

Let's take Finland as an example. The last three presidents (in five  
elections) have come from the left wing although right wing has  
majority. There are two leading right wing parties and one leading  
left wing party. There are of course also many other factors  
influencing the end result but one can not avoid thinking that also  
the competitive position between the right wing parties has an effect.  
The last three elections used TTR. Before that there was an electoral  
college with three rounds (and a mixed direct+electoral college system  
in the last elections of that format). Based on this experiment it may  
be that even in the second round of TTR the right wing has not found a  
consensus to vote for the only remaining right wing candidate. (The  
gap between right and left wing may also not be so strong that it  
would stop people voting across that dividing line.)

These examples do not prove anything but they show that it is possible  
that even in TTR having multiple about equally strong candidates of  
one wing might be harmful to that wing. Approval does not have the  
"clarifying" second round. Also typical rank ballot methods do not  
have second round. It may be that it would be easy to rank R1>R2>L  
since when voting that way one anyway clearly states that R1 is better  
than R2. Voting R1>L>R2 (before knowing that R2 will beat R1) might  
require more "competitive attitude" than voting for the left wing  
candidate at the second round of TTR. But this is too complex for me  
to tell. Real life experiments might show something more. Clone like  
candidates may be a problem in many ways. Rank ballots could be a good  
way to try to overcome this problem.

>
> There is the thought that Approval forces the voter to strategize...
> But in practice all the strategy (if you can recognize it as such)
> should be in what happens with the polls and whatever affects the  
> polls.
> On election day the strategic voter probably only has one meaningful
> decision to make, and it's a binary one.

This works fine if there are only two leading candidates left at the  
election day and voters have been educated well enough to understand  
that they should approve exactly one of them. But it seems that at  
least in multiparty countries situations where there are more than two  
potential winners are not uncommon (this seems to be the rule (or at  
least very common) in Finland).

>
> The downside for Approval is what happens when assumptions about  
> quality
> of polling information and the ability of polls to reach equilibrium
> don't hold. I.e., when voters are unable to play the game effectively.

In some sense the poor quality of polls may also improve things. I  
mean that one of the right wing candidates seems to lead clearly in  
some polls (although this is a mistake) that could help people make up  
their mind and jointly with media "declare" one of the right wing  
candidates to be the leading right wing candidate.

(There is also the other side. Often people seem to give support to  
the non-leading candidates and find problems in the leading ones  
(maybe seeing them as arrogant and too strong) and thereby make the  
set-up more even at the election day.)

My summary is simply that near ties seem to be slightly more common  
that mathematical models would suggest. Partly this may be due to  
inaccuracy in measuring things, variation in opinions and polls,  
uncertainty of how things will evolve, and the very uncontrolled  
thoughts of individual voters. I note that also in most two-party  
elections the outcome is not clear before the election day. The same  
rule may also apply to three and more credible parties / candidates (=  
typically at least two potential winners, often also three, at least  
in Finland).

Juho


P.S. In Finland the (first round first place) support of the three  
strongest candidates in the last five presidential elections were  
43-19-17, 49-21-18 (second term), 26-22-20, 40-34-13, and 46-24-19  
(second term). In the third election one of the three strongest  
candidates did not come from the three biggest parties. In all  
elections the candidate with most first preference first round votes  
was the left wing candidate (that won also the final round). In the  
left wing smaller parties are quite likely to vote for the large left  
wing party candidate (that more centrist than the other left wing  
parties, that sometimes don't even run their own candidate to make  
sure that the left wing candidate will make it to the final round). In  
the right wing there are few additional smaller parties. (And then  
there are the greens.)


>
> Kevin Venzke
>
>
>
>
> ----
> Election-Methods mailing list - see http://electorama.com/em for  
> list info