ranking versus approval: was: Measurement Error
Topics: Democracy
03 Dec 2000
From: Ervan Darnell
From Vince:
> From: "Ervan Darnell"
>
> > At 01:14 AM 11/13/2000 , Vincent Kargatis wrote:
> > >On improved voting systems: instant-runoff/ranking voting seems to me
> superior to any of the others, better than approval, and way better than 1
> person/1 vote.
> >
> > More on that later....
>
> I am interested in hearing the "more".
Okay, I'm going to punt on the complete analysis because it's quite complex. I'll settle for a couple of points and a reference. The ground rules that I assume here are that each voter has some specific preference value for each candidate (summing to unity) and the purpose of the election system is to maximize utility across all voters. The obvious problem with 1 vote per person is that the preferred candidate never gets a vote because the only useful strategy (for a given election) is to choose between one of the two front-runners. From a libertarian perspective, I think this is a broken model because the outcome of an election is not equally important to all voters and that assumption is built in to the model. But playing this game as it stands:
Ranking is better than one vote per person but still problematic for at least two reasons:
1) It still has the one-vote paradox, I want to cast my highest ranking for one of the two leading contenders, even if they are my two lowest ranked preferences.
2) I'm forced to arbitrarily elevate or reduce my opinion of a given candidate. If I think one has it all and the others are all scum, I rank the good guy "1" and am then forced to rank the least obnoxious "2", i.e. almost as good as the guy I really like.
Approval voting has neither of these failures. It avoids the first problem by the simple reason that I can vote for as many candidates as I like. It avoids the second problem because I'm evaluating every candidate independently. The obvious downside is that one is no longer left with a preference expression between the approved candidates (nor a preference between the unapproved ones if that happens to matter). However, the effective error in the estimation of personal utility is on average less than in ranking systems. Here are the best references I could find easily:
http://www.kellogg.nwu.edu/faculty/weber/papers/approval.htm
http://bcn.boulder.co.us/government/approvalvote/center.html particularly:
http://bcn.boulder.co.us/government/approvalvote/altvote.html
None of these are perfect. Paradoxical situations can be created to defeat almost any system. One can get ever closer by increasing the complexity. For instance, ranking and approval can be combined in some sense by giving everyone 100 votes to cast, scattered any way they wish between the slate of candidates. In theory, that's better for many situations (but not all I believe). As a practical matter, it's unlikely that voters would think through how to use such a complex system to good advantage.
On a personal note, I founded a microcomputer hacking club at UMR years ago (when owning your own hobby computer was still an elitist thing). I wrote approval voting for officers into the club charter. It was difficult convincing the administration to accept it, because they regarded approval voting as something scary and untried. Even more depressing, I found that an above average cut of college students had difficulty understanding it or how to use it. We so rarely had more than 2 candidates for a position, people did not grow accustomed to it. Perhaps if it happened every time, the situation would be different.
-------------------------- eGroups Sponsor -------------------------~-~>
eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/9699/1/_/220122/_/975888938/
---------------------------------------------------------------------_->
Home