[EM] Comments on a posting from Benjamin Grant

[Jameson Quinn]

2014/1/4 Michael Ossipoff <email9648742 at gmail.com>

> [quote]
> Let's say that election after election people
> see that more and more people are voting for Nader,although he is not
> winning.  Thinking optimistically (as some people like to) that this might
> be the year that Nader could take it all, they put all their money on Nader
> - they vote Nader, but *not* Gore. The result? Gore's numbers drop, Nader's
> numbers rise a little, but Bush still get's the most!
> [/quote]
>
> That's the old standard criicism against Approval. The answer to it hasn't
> changed, over all the thousands of times that objection has been made.
> Approval isn't a pairwise-count methods. Approval doesn't aspire to, or
> claim to achieve, the pairwise-cound ideal, whereby we can help Compromise
> against Worst, while also helping Favorite against Compromise.
>
> For that, we need rank-balloting.
>

I believe that this is a mistake. Yes, it is true, only in a ranked method
can the same ballot simultaneously count full-strength for Green over Dem
and for Dem over Repub (or of course similarly for Libertarian>Repub>Dem).
But that's not really needed. All you need is a method that will count your
ballot full-strength for whichever you prefer out of the two actual voted
frontrunners (which is critically different, of course, from the two
presumed frontrunners going into the election).

And I believe that in practice, median systems (Majority Approval
Voting/MAV; Majority Judgment; Graduated Majority Judgment; Majority Choice
Approval; ER-Bucklin; etc.) will do that. Assume Green voters give the
Green candidate an A, the Dem a D, and the Repub an F; and Dem voters all
give the Repub an F; then as long as there are over 50% Green+Dem, Repub
will safely lose; and the minute there are over 50% Green, the green will
win. (Again, in a right-leaning area, the same argument applies for a
generally rightist third party such as the Libertarians).

Yes, you can construct artificial scenarios where a median system does not
elect the CW. However, the "blurrier" you make the electorate (with
different voters within a faction casting similar but not identical
ballots), the harder it becomes to make such scenarios come out. In
practice, I am certain that the electorate would be "blurry" enough that
such scenarios would be vanishingly rare.

Why is MAV better than Ossipoff's proposals (IRV, Benham, and Woodall)? For
one thing, none of the 3 are summable. In practice, that is not an
insurmountable barrier, but it does mean that software is many times harder
to certify, and that fraud-prevention checks are similarly trickier and/or
less reliable.

For another, MAV is easier to understand. Instead of an iterative and
comparative process, you can give each candidate a single number or grade,
and the highest number wins. Or, if people want to understand the entire
process, a Bucklin-based explanation gives a good understanding of what's
going on, and is relatively easy to carry out in practice in a hand-counted
election (or in a show-of-hands election, though this leaks information and
allows a certain amount of additional strategy if voters aren't scrupulous
about ignoring the leaked info).

For a third, I am confident that MAV results would give significantly
better social utility than IRV, and slightly better than Benham and
Woodall. That is to say, MAV clearly does better with the same amount of
voter honesty, and I see no reason to believe that voters would be any more
honest under the latter systems, and certainly not by enough to offset
MAV's advantage.

...

Like Ossipoff, I am convinced that Approval is the first step to election
reform; simple to understand, a huge step up from plurality, and a step
towards any good system. But it's worth having a response to scenarios like
this one, where approval is still not good enough. For me, that response is
median/bucklin systems, such as MAV.

Jameson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20140104/bedc8abe/attachment-0004.htm>