[EM] Asset Voting vs STV

[Abd ul-Rahman Lomax]

At 08:33 PM 4/24/2007, Jonathan Lundell wrote:
>On Apr 20, 2007, at 7:41 AM, Abd ul-Rahman Lomax wrote:
>
>>I'm just taking the opportunity to note the similarity between
>>multiwinner STV and Asset Voting. With multiwinner STV the vote
>>transfers are guided by user rankings, and in Asset Voting, by,
>>essentially, a proxy chosen by the voter. They are really the same
>>method, differing only in who provides the tranfer information.
>
>If I understand Asset Voting (and I may not), there's more to it than
>that. An important aspect of STV (without which is loses later-no- 
>help/harm, a critically important property, to my mind) is that, in
>any given round, only the current top choice on each ballot is
>considered.

Well, that's a *little* more. I was noting similarity, not identity! 
With Asset Voting, the transfer information is provided by a proxy. 
(We often assume that this person is a candidate, but not 
necessarily. Indeed, with single-winner Asset there are some reasons 
why we might prohibit candidates from serving as proxies. But that's 
another story. I've assumed total freedom both in registration as 
candidates and in choosing what candidate(s) to name.)

With STV, the vote transfers are determined by a ranked list, so that 
the top name is used unless that candidate is eliminated, then the 
next name, etc. The voter provides the transfer information, which is 
used specifically and rigidly. It's the rigidity that is the problem. 
As I've noted, the problem is not generally serious with multiwinner 
STV, it seems, but it *is* a limitation and affects the ability of 
STV to maximize voter freedom in choosing representatives.

>One could, of course, allow a voter to fill out a truncated STV
>ballot by indicating somehow that the remainder of the ballot should
>follow an ordering specified by some third party (it could be my
>favorite candidate, or my party, or any other appropriately
>registered preference ordering). Such a variation would preserve
>later-no-help/harm, but allow the voter to defer to the judgement of
>a third party.

This kind of hybridization is often possible with methods. For 
example, it is generally considered that Range Voting does not 
satisfy the Majority Criterion. However, if we add a top-two runoff, 
where the Range winner and the Condorcet winner, if there is any that 
differs from the Range winner, face each other in a choose-one 
election, the combination satisfies the Majority Criterion. That is, 
if the majority has a preference, and maintains that preference in 
the face of a differing Range winner, the majority cannot fail to 
prevail. This, actually, I personally prefer to Range alone. But if 
we must have a single step under all conditions -- a serious 
limitation in my view, and a foolish one -- then I refer Range. It 
will almost always choose a Condorcet winner.

(Asset is kind of a trick, it is not an "election method" as it is 
sometimes defined: deterministic, using a single poll as the only 
input. It is, rather, a hybrid itself, between a simple 
meet-the-quota-and-you-are-elected method and a deliberative, 
negotiated process.)