[EM] average over time proportionality election method

Raphael Ryan RaphFrk at netscape.net
Mon Mar 6 18:00:01 PST 2006


Abd ul-Rahman Lomax <abd at lomaxdesign.com> wrote:

>At 07:18 AM 3/6/2006, Raphael Ryan wrote:
>>
>>The question comes down to how constant that threshold is.  It isn't 
>>perfectly constant so the system isn't perfectly proportional.
>
>It will never be "perfectly proportional." The quantization noise, 
>very likely, would be too great. 

Well, it would be on average proportional over the long term.  The noise is only short term noise.

>And it should be realized that there 
>is a severe risk.
>
>There is quite often a significant minority in favor of actions that 
>would endanger the community. 

I don't think that in a 100+ legislature that they would end up being that significant.  Also, I think if the districts were multi seat, then the randomness would be reduced.

>While there would be little harm in a 
>few representatives in a large parliament being elected through this 
>scheme, there would arise a risk that more than just a few would be elected.
>
>There are other complications: candidates don't run again for various 
>reasons. The transfer of votes mechanism was passed over with what 
>look to me like major unsolved problems.

This is an issue, I think on retirement or death, excess should be transferrable.  Allowing transfers in every election after the votes are counted is equivalent to asset voting (I think ?) except that candidates can decide to keep some of their votes for the next election instead of giving them all away.

>
>*However*, while the method is interesting, delta-sigma communication 
>works when the update frequency is high. Delta-sigma with an update 
>frequency of a year would not track the electorate very well. Indeed, 
>what this system would do is to influence present assemblies by how 
>the electorate felt years earlier.

This could be a feature, it provides some stability.  However, this would likely be offset by the quantisation noise.

>
>Further, since the goal seems to be proportional representation, and 
>there is already an excellent method on the table that *quickly* 
>creates proportional representation without lost votes, and it does 
>so almost immediately (and without party-list), the idea seems a 
>purely abstract exercise in possibilities, something that we should 
>certainly do, but not likely to be very fruitful in terms of ever 
>seeing the light of day.

The proposed system could be combined with a ranked voting method.  You would rank candidates and the last choice would be something like "7:  candidate B's excess pool".  This would give pretty good proportionality in the short term and near perfect proportionality in the long term.  The main point is to eliminate wasted votes.  

It could also be used with fractional voting too, in fact, that might be useful, there is little point in using up your entire vote on a candidate who is at 90% of the threshold before the election.  

I see your point about asset voting.  A negotiation step can allow the interactions to happen between candidates that would require a complex voting procedure to model.  However, you lose the anonymous/secret ballot effect at the negotiation stage.  There are also issues for hold outs and deadlocks and "chicken" effects.  A deterministic voting system avoids all that.

Also, I was reading Warren Smith's paper.  I think that the mechanism for elimination he proposes still suffers from the middle squeeze effect of IRV.  Perhaps a better rule would be to allow a candidate to voluntarily exclude himself.  However, I guess a candidate can do that by just transferring all his votes to another person.


__________________________________________________________________
Switch to Netscape Internet Service.
As low as $9.95 a month -- Sign up today at http://isp.netscape.com/register

Netscape. Just the Net You Need.

New! Netscape Toolbar for Internet Explorer
Search from anywhere on the Web and block those annoying pop-ups.
Download now at http://channels.netscape.com/ns/search/install.jsp



More information about the Election-Methods mailing list