[EM] Dave on approval, ranked ballots

[Abd ul-Rahman Lomax]

At 06:44 PM 7/27/2005, Kevin Venzke wrote:

> > It's a question of election philosophy.
>
>I don't agree with your perception. I don't believe an election method can
>find the candidate who is "most widely approved" unless we really do 
>assume that
>voters are sincere, and use the same definition of "approve."

Yes. That's true. However, there is a clear definition of "approve," and it 
is very simple: a candidate is approved if you contribute to the 
candidate's possible election by marking an approval ballot for that 
candidate, or by otherwise ranking the candidate above other candidates.

While there is no guarantee that this will actually pick the "most widely 
approved" candidate, because of possible oddities in how voters will use 
the method, it would seem likely to do so. Absolutely, if voters approve of 
a candidate merely because they are afraid that another, even worse, 
candidate might otherwise lose, there is quite a possibility of a failure; 
this failure would have resulted from the voter's judgement that it was 
better to have the better of two evils, and, indeed, this is not the same 
as what we would colloquially mean by "approval," nor what we would want. 
However, even here, even if many voters vote in this way, it *still* 
remains likely that the result would be a more widely approved winner.

Note that no method is going to guarantee the desirability of results if 
large numbers of voters vote insincerely.

>  Otherwise it seems
>to me that giving or withholding approval is just a strategic tool to get the
>best outcome for oneself.

That is one possibility, but there is quite another. Which is that some 
voters might indeed prefer that a jurisdiction be governed with the widest 
possible consent of the governed. I can say that I'd be among them!

In other words, I prefer that the widest number be satisfied than that I 
personally be *most* satisfied. I consider the latter situation unstable 
and insecure.

> > "Perfect" implies "without flaw." I'd call the failure to elect a
> > compromise candidate a flaw.
>
>When I said "100% perfect" I was talking about the method's ability to count
>individual ballots in the manner intended by the voter.

Sure, with that special definition (which I must presume means that a 
voter's support for a candidate in a pairwise election will always be 
effective in that election and will not produce contrary effects).

>I consider it arbitrary that you find Condorcet less able to pick a compromise
>candidate. If voters in an Approval election do not approve below their 
>favorite
>candidate because they don't want to rank compromises equal to their favorite,
>then Condorcet might well have picked a better compromise, since at least then
>the voters could explicitly rank the compromises below their favorite.

I'll agree. Condorcet should be able to come close to Approval.

>I think it's a real possibility that Approval elections would turn out to look
>much like Plurality elections.

I've said as much many times. However, allowing overvoting is such a tiny 
change to plurality election rules and procedures that it may be the most 
easily implemented of all the alternative complete election methods. 
Fractional Approval Asset Voting a twist on Warren Smith's Asset Voting, is 
likewise just as easily implemented and is superior to every other method 
known to me of comparable simplicity, but it does this by incorporating a 
deliberative phase where votes are redistributed at will by those who 
received them. In other words, the method itself does not necessarily 
determine the winner(s) unless they receive, prior to distribution, 
sufficient votes. (Theoretically, the method could still allow such a 
winner to decline the office, instead choosing another to serve.)

Indeed, I think it *likely* that, initially, AV results would usually look 
like plurality results. This is a feature, not a bug, for it makes 
implementation less of a threat to existing powers. However, over time, it 
is possible that the effects would shift. Note that IRV, and, indeed, the 
actual results of many election systems could initially resemble plurality 
results in most elections, until the political environment adapts to the 
new possibilities. But I think AV would have the mildest initial effect of 
all, while still being, in potential, a vast improvement.

I keep underscoring this, because I don't see it echoed much: Approval 
Voting is not actually a new method; it merely is a discarding of a rule 
that may have been created through a misunderstanding of what overvoting 
would mean. It's easy to make the mistake, at first glance, that overvoting 
allows a voter more than one vote, and that seems unfair. But it isn't. 
Overvoting, if counted, only allows one vote from a voter to appear in the 
winner's total. Or none. Never more than one. The extra votes, if any, are 
moot.