[EM] Fwd: 2008 election fiasco is preventable

[Abd ul-Rahman Lomax]

At 04:35 PM 7/13/2007, Jonathan Lundell wrote:
>On Jul 13, 2007, at 12:56 PM, Abd ul-Rahman Lomax wrote:
>
> > Yes, it does not solve all problems. But Approval Voting is a very
> > good system, one of the best. Higher resolution Range is better. In
> > the other direction, Condorcet methods are arguably better in some
> > ways. Approval is *clearly* better than so-called Instant Runoff
> > Voting, and it is a whole lot simpler to implement.
>
>"Better" only if you don't care that voting for your second (or
>lower) choices hurts the chances of your first choice.

"Better" is with reference to social utility simulations, the best 
information we have about the comparative behavior of election methods.

When I wrote that Approval does not solve all problems, I was 
referring to the obvious lack of discrimination in the method, such 
that there are no ranks within the Approved and Disapproved classes. 
Range addresses this, as to ranked methods.

Of course, I can also be alleged about Range that "voting for", i.e., 
rating highly, your second choice "hurts the chances" of your first 
choice. However, what must be noted here is that the "hurt" is 
relatively small. After all, this *is* your second choice.

In any case, the first and most serious problem we must address is 
the so-called spoiler effect, which creates serious harm within a 
two-party environment, where third party candidates have little 
chance of winning, no matter what the election method. (We note that 
IRV, "Alternative Vote," is not associated with strong multiparty 
systems, it seems to, like Plurality, encourage two party dominance.) 
In this environment, voting for your second choice has no significant 
effect on the "chances of your first choice."

Unless you happen to Approve both major party candidates. One wonders 
why you even bothered to vote!

Indeed, what we would expect to see with Approval is that supporters 
of major party candidates would mostly continue to vote as they have 
been voting, for one candidate. It is the effect on third party 
candidates that is important; it allows them to participate in the 
real election while still indicating support for their true favorite. 
Obviously, that support is not as clearly established in simple 
Approval as it would be in Range, or, in a different way, in IRV or 
other ranked method.

In short, Mr. Lundell's comment is about a matter which has little 
practical effect in the short run, which is why I did not explain it 
in detail in my post, but only alluded to it.

In any case, the real comparison is not between Approval and Range or 
ranked methods like IRV -- or better ones--, but between Approval and 
Plurality or FPTP, the status quo in the U.S. What Mr. Kok is 
proposing is a very simple and fast reform that does not raise the 
cost and other objections that are practical obstacles to deeper 
voting method reform; all that we have to do is to Count All the Votes.

As stated, it doesn't provide for every need. But it does remarkably 
well for such an obvious and simple step. After all, why *shouldn't* 
we count all the votes? Why assume that voters made a mistake if they 
voted for more than one?

(And if you look at justifications for discarding overvotes, that's 
what you will find. Allowing overvotes does *not*, under any 
circumstances, allow a voter more than one effective vote, which can 
easily be proven by looking at any election where overvotes were 
allowed. Discard all the "extra votes" which were cast for losers, 
and there is no effect on the election. Those votes were moot. They 
are, in fact, "conditional votes," just like the "extra votes" in 
IRV. They only affect certain pairwise contests.)

Those of us who are working for immediate adoption of Approval mostly 
do not think that it is the ultimate voting method. However, if we 
*do* get Approval, there will be, we think, an appetite for being 
able to express finer distinctions, and there are simple and measured 
steps to get there from an Approval base. Approval allows equal 
ranking, and the best ranked methods also allow equal ranking. It's 
quite feasible, with a little extra ballot and counting complexity, 
to add ranks to Approval; then the question is how these ranks are interpreted.

Most of us (the Approval supporters) prefer to use Range methods 
rather than ranked methods, but there is, in fact, a compromise which 
I've come to think is ideal.

Quite simply, a Range ballot, with its rankings of candidates, can be 
interpreted as a ranked ballot, equal ranking allowed. It is then 
possible to determine if there is a candidate who beats the Range 
winner pairwise. Because the Range winner is the prima facie social 
utility winner, it would not be correct to simply award the election 
to the pairwise winner. However, neither do we have a clear 
democratic choice when there exists a candidate who, by rank 
analysis, would have beaten the Range winner in a direct contest.

So we would have an actual runoff. My analysis leads me to think that 
the Range winner would *ordinarily* be confirmed by this runoff, 
except when various election phenomena have caused the Range winner 
to not truly be the SU winner (such as strategic voting, perhaps 
based on misinformation or deception).

It should be noted that simulations show that the Range winner is 
*usually* the Condorcet winner, so runoffs would not be the norm. As 
to the contingency that there is more than one candidate who pairwise 
beats the Range winner, I am at this time only noting that this would 
be quite rare, most likely; we could use any of the good Condorcet 
methods to pick a ranked winner; and the matter, in my opinion, 
deserves further study.

Simulations have shown that Range with a top-two runoff has somewhat 
higher SU results than simple Range; because of normalization error, 
Range isn't perfect. Merely very good.

The elephant in the living room of ranked methods is, of course, that 
pure ranked methods take no account of preference strength; and, as a 
result, can clearly choose seriously suboptimal winners. If you 
really care about "hurting your first preference by also approving 
your second preference," then perhaps you shouldn't approve your 
second preference! The rational choice will depend on preference 
strength and the election context, i.e., the estimated probability 
that an election is going to be close, between the relevant pair of candidates.

If a person prefers pepperoni pizza, but also approves of mushroom 
because the preference is not strong, yes, adding the second 
preference vote can "hurt" the chances of getting the first 
preference, but we make compromises like this all the time when we 
act jointly with others. In healthy environments, especially where 
some have weak preferences and others strong ones, we quite properly 
go for what is most broadly acceptable. Approval is the simplest way 
to achieve this goal, and I've seen it used that way; I've seen it 
used to make a choice where there was a clear preference of a 
majority, but not strong enough to cause the majority to disregard 
the strong preference of a minority; there were a series of optional 
actions presented, and members were asked to state which ones were 
acceptable. While one choice, we knew, had been vigorously defended 
by some members -- it was the status quo -- and words like "over my 
dead body" had been used, when the Approval results were in, it was 
clear that there was one option which was acceptable to almost all. 
And the majority stepped aside; that one option was presented for 
adoption and the vote was *unanimous* to accept it.

(There were religious issues involved. The majority apparently 
decided that holding to what had been traditional for many years was 
not worth avoiding offending a minority, perhaps twenty percent of 
members. In other words, the approval vote for the status quo was 
about 80%, and for the new option, 98%.)