[Election-Methods] Strategy/polling simulation for simple methods

[Abd ul-Rahman Lomax]

At 11:40 PM 1/21/2008, Kevin Venzke wrote:
>For each round of polls I can also note the rate of agreement between the
>poll winner and the sincere CW (if there is one), sincere MF (if there is
>one), and social utility maximizer given the voters who show up.

Other writers have long noted this, and I've, in particular, pointed 
out that participation bias favors the SU maximizer (which I 
generally call the "range" winner). Thus top-two runoff is a much 
better method than often thought. The only problem is it's 
vulnerability to center squeeze in the first round. Approval with 
top-two runoff should do much better, likewise Range with pairwise 
analysis and a runoff if there is conflict between the range winner 
and a pairwise winner (beating the range winner0 and when there are 
cycles, consideration of all of the cycle members, though that's not 
a possibility I've explored much, it would be extraordinarily rare, I 
suspect. Probably Range winner vs. any candidate beating the Range 
winner pairwise would be quite enough.

Besides, there is a much better method: parliamentary elections 
through deliberative process, with an Asset elected parliament, with 
direct voting allowed by all electors (all those who hold votes from 
the original secret ballot.) Deliberative process is, of course, 
Condorcet compliant, and it allows the repeated balloting (v. 
Robert's Rules of Order) which has the settling effect Venzke 
describes. Fully democratic (direct participation or participation by 
chosen "proxy," which is what it amounts to), creates peer assembly 
as part of the same electoral process, efficient, fair, and 
intelligent. Down side?

Let's see -- there will be logrolling (horrors!). Candidates will say 
one thing and do another (double horrors! -- rule one: don't vote for 
someone you don't trust, period. Under present systems you have to do 
that, or you vote for someone you think you trust from media 
impressions, but very few can vote for someone they would actually 
hand their baby to for care without further exploration of who they 
are in person. Only with Asset (or delegable proxy) can you actually 
vote for the person you most trust, without any consideration 
necessary of whether or not that person is "electable," because, by 
voting for him or her, you *are* electing.

Turns out Lewis Carroll got the Asset part right, in the 1880's. What 
more lurks behind the looking glass?

>The strategy used varies by the method. Approval is the easiest:
>Expectation can be perfectly calculated from the previous round of polls.
>Approve above expectation.

Right. This work has been done before (by Lanphier?). Though maybe 
not with this kind of simulation. It is easily predictable on pure 
theoretical grounds.

>FPP, antiplurality, and combinations involve estimating the value and
>likelihood of all ties that could be broken from a given vote.

I did analysis of Range Voting vs Approval strategy (expected outcome 
for various votes), by looking only at the outcomes for votes which 
actually affect the outcome (create or break tie). All other votes 
affect the voters' overall expected utility for voting, but not the 
expected utility between vote options. This allows an almost-exact 
study of expected outcomes; this is published on the Range Voting 
site. There was further work to be done, but, hey, I've got 
Attention-deficit Hyperactivity Disorder, and one thing that means is 
that I can do some serious work, with interesting results, at the 
beginning, but finishing it, nailing it shut, and sending it off to 
the publisher, almost never happens. Story of my life. Anyway, I've 
got lots of nice nearly-woven scarves lying about. Trying to figure 
out how to finish them. It will take help, that's clear. If I could 
do it on my own, it would mean that my ADHD diagnosis wasn't 
accurate, and that the doctors who prescribe me controlled substances 
to treat it are misguided.


>Hopefully this is a somewhat interesting read. Any thoughts?

It *is* interesting, but I didn't read beyond what I commented on. 
Someday I should. Kevin, would you like to work on a paper formally 
publishing all this? We *can* get it published, and if it's 
published, there will actually be some serious attention paid to it, 
and, hey, if it's published, you can put it on Wikipedia. This list 
is a form of peer-review, better, in fact, than the peer review 
boards of most peer-reviewed publications, but, because there is no 
decision made, it's all informal, it's wasted and without serious 
consequence (except that all of us who are paying attention now know 
something or know it better, but ... as far as informing a broader 
audience, useless and mere navel-gazing among election method 
aficionados. To move beyond this, as one suggestion, join and 
*participate* (don't just wait for me to do something and then react) 
in the Election Methods Interest Group, 
electionmethods at yahoogroups.com). It's an FA/DP organization, which 
has lots of consequences, among them that you can join, pick a proxy, 
then go on no-mail or special notices status, and wait for your proxy 
to ping you if your proxy thinks it needed. If you are interested in 
election methods, there is *no* good reason not to join. You can even 
be a supporter of -- horrors! -- IRV and join, and a fair number have.

Thanks for your work, Kevin.