[EM] The worst about each system; Approval Preferential Voting (new name for an MCA-like system)

[Abd ul-Rahman Lomax]

At 02:30 PM 5/26/2010, robert bristow-johnson wrote:

>On May 25, 2010, at 1:20 PM, Jameson Quinn wrote:
>
>>What are the worst aspects of each major voting system?
>>
>>-Plurality: Everything. It routinely requires dishonest strategy
>>from a large minority, or even a majority, of voters. Enough said.
>
>except some unnamed folks here (whose posts i don't see anymore) think
>that it's better than IRV.

I am *so* relieved that Mr. B-J doesn't have to suffer through my 
posts any more. He should have done this long ago.

Is Plurality better than IRV? Under some conditions.

You are a small town. You hold nonpartisan elections, with a small 
number of candidates. Plurality is better than IRV. Why? I'll just 
start with a few:

1. IRV under the general conditions of nonpartisan elections almost 
never changes the result from Plurality. People don't realize this 
because they tend to think of "spoiler effect," which usually depends 
on partisan elections and a small party or independent candidate 
pulling off a few votes that flips the election. IRV can fix that, 
but at huge cost, and, notice, it is turning a victory for one 
almost-winner into that for the other almost-winner. Many imagine 
that IRV would have rescued the nation from George Bush, but it is 
far from obvious. It might have made a difficult canvass into a 
totally insane impossible one. The fact is that in nonpartisan 
elections, that phenomenon seems to almost never influence the 
outcome. IRV doesn't flip results in nonpartisan elections.

2. When the method is plurality, people know that if they vote for 
their favorite, if their favorite is not going to win, they are 
wasting their vote. IRV can create that impression, but it is a false 
one under center squeeze conditions. By voting for their favorite and 
thus concealing their preference for the candidate who would be the 
majority winner, underneath their favorite who runs second in 
first-preference votes, and who maintains that until the last round, 
they have wasted their vote, they might as well have stayed home. 
Just like Plurality. But they know it, so they can make an intelligent choice.

3. Plurality is much easier to canvass. It's also, in a small town, 
easier to vote. Just vote for your favorite, hang the "strategy." 
People do accept plurality results as fair, and in small town 
government, when Plurality is the method, not very many offices have 
three candidates, so it's moot. The problem in small towns is more, 
sometimes, in getting *anyone* to run!

But I personally believe that finding a majority is important, 
because it is more unifying. IRV, quite simply, doesn't do this, the 
majority it manufactures is, too often, faux. Since top two runoff 
has some of the same problems as IRV -- but it functions better in 
terms of results than IRV -- I suggest using a better advanced method 
for the primary, one that is actually designed to seek true 
majorities, unlike IRV, and that certainly does it better than IRV. 
And that's Bucklin, and it is easy to vote and canvass, and there are 
no reports in the historical record otherwise. It was tried in 
approximately ninety towns in the U.S., in roughly 1910-1920 (a far 
wider application than FairVote has managed, without the central 
organization pushing it), and it was used for party primary 
elections, apparently, for much longer. It's alleged "failures" 
disappear if it is used in its best application, as a way of finding 
majorities without a runoff. It does it, often. And when it fails to 
do so, instead of using it to elect by plurality, just hold the 
runoff! Compared to plurality, you have not lost anything, and you 
have gained a great deal.

In particular, Bucklin does very well at allowing sincere first 
preference expression. That is very important to voters! It is very 
flexible for voters in how to add additional approvals, and this 
becomes much less of a worry in a runoff system. Voters can make 
their decision on adding second preferences, or third preferences, 
with a simple question: which do you prefer, to add lower preferences 
or to have this election go into a runoff? If you don't mind a 
runoff, you are completely free to truncate, if that makes sense to 
you. Most people, historically, in major Bucklin elections, did add 
additional preferences. But we don't have a lot of data. There is a 
project for an enterprising student!

>>-Approval: divisiveness.
>...
>>-Range: Strategy is too powerful.
>
>i couldn't get the guys at ESF to even acknowledge the obvious
>strategic considerations a voter would face with Approval or Range.
>they just say that "it's mathematically proven" to be better than
>anything else.  Clay Shentrup needs to get on this list and start
>defending his position rather than expecting me to do the same on his
>list.
>
>Clay, i'll take you on here on EM, but not on ESF.  it takes too much
>time and is a far less objective context.

The arguments are the same regardless of the list. On the ESF list, 
you'll get more participation from experts in Range, that's all. 
Here, you'll get a bit more from people who don't like Range. But 
your arguments don't get better in one list vs. the other. To B-J, 
though, it's personal, and personal face is involved, hence his preference.

I've seen approval in real use in a real organization, to make 
choices efficiently. It was the opposite of divisive. The proof was 
that a ratification vote, where the approval results had been 
something like 98%/65% for the top two (out of many options, 
plurality and repeated ballot would have taken much longer to find a 
majority, or would have chosen the 65% immediately, it was the status 
quo and was very popular before the discussion, and remained about as 
popular), was unanimous. The 98% became 100%.

This election was a good example of how the majority criterion can be 
defective. Before the polling, first preference would have been the 
65%. Not only a majority, almost a supermajority. But once it was 
seen how much more acceptable the 98% option was, the 65% voters 
changed their minds. They no longer preferred their original first 
preference because, guess what? People actually consider the benefit 
to the organization of unity. If every time you have an election, you 
offend a third of the people, the organization gradually gets weaker. 
If the election had been Range, it would have been even more obvious. 
Those who preferred the status quo were familiar with it and did not 
realize the depth of opposition. The opposition was actually offended 
by the status quo, and it was a religious issue. Classic Range voting 
Condorcet criterion failure.... I.e., classic Range voting success.

Range and Approval are basically the same method, Range just allows 
finer distinctions to be expressed. Strategy in Range is no more 
powerful than in Plurality. Indeed, Range and Approval default to 
Plurality if voted that way. Except that I don't like them as 
plurality methods, I prefer, greatly, that majority approval is 
required. And that requires that sometimes there be at least two 
polls. In pure democratic practice, the number of polls is unlimited, 
and that's the standard for elections under Robert's Rules, which are 
certainly widely used.

Voting systems are not just for political application, they are general.

And, in any case, listing the systems like this is narrow-minded. The 
best systems are probably hybrids of some kind. It is entirely 
possible to have a hybrid Range/Bucklin system that is also 
Condorcet-compliant. It merely takes, probably only in a fraction of 
elections, a runoff.

(Usually, the Range winner and the Approval (Bucklin) winner and the 
Condorcet winner will be the same. Runoffs are only needed, perhaps, 
to deal with majority failure or the rarer multiple majority problem, 
that a multiple majority might not actually reflect wider approval, 
but poor strategic decisions.)

What Approval can do for small organizations that can vote directly 
is to make the repeated ballot procedure (which is normally 
vote-for-one) more efficient. The only problem with Approval in a 
repeated ballot contest is the contingency that a ballot comes up 
with two majorities, and there is a simple fix: ratify the result! In 
public elections, there is already the tradition and law that if two 
conflicting ballot questions pass, the one with the most votes 
prevails. If people don't like that, they could hold a runoff. 
Nowhere have I seen that done with conflicting ballot questions, 
though. It was assumed that the most widely acceptable result was better.

It is claimed that more advanced systems are harder to vote. There is 
no evidence for that. A hybrid Range/Bucklin/Condorcet system would 
be, probably, easier to vote than Plurality, because it is not 
difficult to pick favorites and to rank. This is actually an argument 
that FairVote makes, and it's often true, it is simply not the whole 
truth. In particular, if it's hard to rank two candidates, it should 
be simple to equally rank. A Bucklin voter can simply bullet vote, if 
anything else is difficult for the voter, and voters should know that 
this is perfectly okay, and, in a runoff system, it's also quite 
safe. Ranking candidates according to what is easy, and not ranking 
the rest, with good ballot and system design, then uses the ranking 
as evidence of clear -- i.e., relatively strong -- preference, and 
the not-ranking as lack of preference. Isn't that what it means?