[EM] Remember Toby

[Dave Ketchum]

Here we get led down a path of rejecting methods for being worse than  
plurality in some way, without considering whether they may be better  
in other important ways.

Agreed that Approval is better - and very little different.

Asset deserves a bit of study - the candidate you vote for can be part  
of deciding who gets elected - sorting this out makes it more complex  
than plurality for me.

IRV sounds great to many for what it offers voters - vote for more  
than one as in Approval, but rank them to indicate which you like best.
      What it tells the vote counters sounds good until you look  
close:  Look only at what each voter ranks highest; if this identifies  
a winner - fine; if not, discard the least liked of what was looked at  
but failed to win and try again.  Usually this will discard losers and  
expose a deserving winner.  But sometimes what I describe next happens  
to one the voters really liked:
   50 A
   51 B>A
   52 C>A
   53 D>A
Counting: 50A; 51B; now 51A; now 53D beats 52C.

Condorcet looks much like IRV to the voters.  Counters, looking at all  
the ballots above say, will see 153A beating 53D.

On Jun 2, 2011, at 3:14 PM, fsimmons at pcc.edu wrote:

> There are only two single winner methods that are uniformly better  
> than Plurality, i.e. that are better in
> some ways and worse in none.  These two methods make use of  
> Plurality style ballots, and those voters
> who want to use Plurality strategy (marking only their preferred of  
> the two frontrunners) can do so without
> incurring a worse result than they would get in a Plurality election.
>
> The two methods are Approval and Asset.  My remarks in the first  
> paragraph explain why neither of
> these methods is in any way worse than Plurality.  To see that they  
> are in some cases better, consider
> the following points:
>
> In the case of Approval, if many voter s also mark the candidates  
> they prefer over their Plurality choice,
> the results will often be improved.
>
> In the Asset voting case, consider that when you trust your Favorite  
> candidate’s ranking of the other
> candidates, you can mark you favorite and not worry about Plurality  
> strategy.   It appears that between
> eighty and ninety percent of the voters would rather have their  
> favorite do the ranking.  Where do we get
> that figure?  We get it from  Australia where the vast majority of  
> voters just copy their candidate cards
> onto the ballot.

The Aussies are required to rank every candidate - a chore few want to  
do for themselves.  If voting for as many as in Approval the American  
voter should see little pain in ranking these few that they approve of.
>
>
> In summary, we have shown that both of these methods are at least as  
> simple and have at least as good
> results as Plurality by treating the ballots as Plurality ballots,  
> and that obviously safe and beneficial
> departures from Plurality strategy yield significant improvements in  
> both cases.  Therefore, these two
> methods are uniformly better than Plurality.
>
> Although there are many other methods that are better than  
> Plurality, there are no others that are
> uniformly better, i.e. no other method Pareto dominates Plurality.
>
> When we propose a method to replace Plurality, if that method is  
> worse than Plurality in any aspect at
> all, you can be sure that the opponents will focus on that aspect.

A reason for caution BUT it is proper to consider magnitudes of both  
gains AND losses.
>
>
> But who can rationally oppose a change to a method that is uniformly  
> better than the status quo, except
> by proposing what they think is a better method?  But that supposed  
> better method can be shot down if
> it is worse than Plurality in any aspect.  Take IRV, for example.   
> It has more complicated ballots than
> Plurality.  And unlike Plurality it fails monotonicity, just to  
> mention two aspects.  No matter that its clone
> independence and later no harm features may completely compensate in  
> the minds of some people; it is
> not uniformly better than Plurality.
>
> Even DYN which is a hybrid that allows Asset voting at one extreme  
> and Approval at the other is not
> uniformly better than Plurality, because the ballot is slightly more  
> complicated.  In every other way it is
> better than Plurality, Asset voting, and Approval.   So far I have  
> seen no method that is uniformly better
> than DYN, but the trouble is that DYN is not itself uniformly better  
> than Plurality because it needs a two-
> bits-per-candidate ballot instead of a one bit per candidate  
> ballot.  Our voting public may not be ready for
> that much change in the ballot.  All of the other proposed methods  
> except various three slot methods
> like MCA use more complicated ballots than DYN.
>
> Is DYN too complicated?  If so, we are stuck with ordinary Approval  
> or ordinary Asset Voting.  They are
> the only choices simpler than DYN that dominate Plurality.