[EM] publicly acceptability of election methods

[Russ Paielli]

Folks,

You may recall that I posted a message a couple of weeks ago regarding 
the public acceptability of election methods. I would like to return to 
that topic.

As I wrote then, I don't believe that complicated methods will ever be 
considered acceptable for major public elections, at least not in the 
foreseeable future. You may not want to hear that, but please don't 
shoot me -- I'm just the messenger.

What is too complicated? Nobody knows the exact answer to that question, 
of course, but let me tell you what I think.

I think you can forget about any method that cannot be explained in two 
or three sentences understandable by persons of average intelligence. 
Maybe that can be stretched to four sentences, but that's really pushing it.

I think you can forget about any non-deterministic methods, cardinal 
ratings methods other than approval, Condorcet methods involving defeat 
dropping, and anything involving bubble sorting, candidate withdrawal, 
or automatic vote changing and recounting (e.g., AERLO).

As you all know, IRV is gaining some traction, but mainly in 
ultra-liberal strongholds such as SF and Berkeley as far as I know. As I 
pointed out previously, the IRV counting procedure at each round is 
identical to conventional plurality. That similarity, combined with the 
illusion that it gives minor parties a fair chance, makes it 
superficially acceptable.

What "good" methods, then, are publicly acceptable? I think Approval is 
simple enough, but it faces two major problems. First, it may frighten 
die-hard Republicans and Democrats too much. Second, and more relevant 
to this discussion, is that voters don't like to be forced to rank all 
their approved candidates equally. Naive voters prefer IRV because it 
allows them to express their preferences by ranking the candidates -- 
never mind that many of those preferences may not be counted.

What, then, is the answer? If you will be kind enough to indulge me once 
again, I still think that what I called "Ranked Approval Voting" (RAV) 
has a lot going for it.

It is based on Approval Voting, which many of the participants here seem 
to hold in high regard. Yet it removes the common objection that 
Approval forces the voter to "rank" all the approved candidates equally. 
It finds the Condorcet winner if one exists, or it eliminates candidates 
based on a reasonable criterion until a CW is found.

It is also very easy to explain:

Rank your approved candidates. The CW wins if one exists, otherwise drop 
the least approved candidate until a CW is found.

A nice feature of this method is that, unless I am mistaken, it 
automatically finds a winner in the Smith set without having to define 
or even mention the Smith set. That greatly simplifies the public 
explanation.

I have come to the realization that equal ranking should probably be 
allowed, but I still don't think the unapproved candidates should be 
ranked. Asking the voter to provide an approval cutoff separate from the 
rankings is unwise as far as I am concerned. It is an invitation to 
insincere strategy, and it complicates both the voting rules and the 
voter interface. And please remember that Approval allows ranking of 
neither the unapproved *nor* the approved candidates -- yet most of you 
still appreciate it.

Please remember also that many voters are uneducated and downright dumb. 
Think about trying to explain your method to octogenarians who never 
graduated from high school and have never used a computer. And try to 
remember that well over 99% of voters would never spend their time on 
this email forum. As far as most of them are concerned, we're hopeless 
nerds. I hope I haven't rambled too long.

--Russ