[Election-Methods] Dopp: 11. "Could deliver unreasonable outcomes."

[Abd ul-Rahman Lomax]

>11. Dopp: "Could deliver unreasonable outcomes
."
>
>Unreasonable outcomes are less likely with IRV 
>than with any other single-seat voting method in use today.

Top-two runoff? Approval Voting (used in a number 
of professional societies)? Borda Count, used in 
various forms in a few places for governmental elections?

Top-two runoff can produce much the same 
unreasonable outcomes as IRV, but there is a 
serious safety mechanism. In most places, 
write-in votes are allowed in the runoff, so if a 
truly bad result happens, the voters can fix it. 
Usually, though -- in every case I'm aware of -- 
the two candidates remaining include one good 
enough that the voters don't bother to work as 
hard as it would take for a write-in to win. With 
IRV, that flexibility does not exist.

>  Every single voting method ever proposed can 
> deliver "unreasonable outcomes" in some 
> scenarios, but real-world experience has shown 
> IRV to be one of the best methods.

Because he says so? Is there any academic 
consensus on that? (FairVote knows that there is 
not, and certainly voting systems experts, 
generally, have a low opinion of IRV, and only 
support it, sometimes, in comparison to 
Plurality. For some reason, voting systems 
experts don't much think about Top-Two Runoff, 
they mostly focus on deterministic methods that 
always produce a result, always complete an 
election from a single ballot. Which is one very 
serious limitation, incompatible with majority rule, actually.

>  The overwhelming number of election method 
> experts agree that IRV is fairer and more 
> democratic than plurality voting even if some 
> might prefer other theoretical voting methods.

Seriously spun. Sure, better than plurality 
voting, better than diving into a swimming pool 
with no water in it. Better than dictatorship. 
Better than lots of things. But ask those same 
experts about Top-Two Runoff, which is the real 
choice, the real alternative in most places. IRV 
is actually being implemented in places that had 
top-two runoff, not plurality. Is it better. 
What, I wonder, does FairVote think Robert's Rules of Order would say?

>  The American Political Science Association 
> (the national association of political science 
> professors) has incorporated IRV into their own 
> constitution for electing their own national 
> president. Robert’s Rules of Order recommends IRV over plurality voting.

FairVote knows the truth about this, and the 
statement has been very carefully worded to 
convey the desired impression without actually 
lying. The impression conveyed is false, the 
words used are true. Genuine, class-A political 
spin. Fortunately, I've been working on these 
issues for about nine months now, with the 
Wikipedia article, which was at one time full of 
FairVote spin, enforced by sock puppets and Rob 
Richie himself, edit warring using an anonymous 
unregistered account to keep all genuine 
criticism out of the article and maintain all his 
talking points, including the two above.

What's the problem? Go look at the APSA 
constitution and, sure enough, you will find a 
provision that if there are three or more 
candidates for the office of President-Elect, the 
"standard method of the alternative vote" is to 
be used. And the method is described, and it is 
STV, or what we call, loosely, IRV. But what does 
this *mean.* Richie wants us to think it means 
that political scientists prefer IRV. Sometimes 
the APSA factoid is even listed as a "recommendation" of IRV.

However, how does APSA *actually* elect its 
Presidents? Surely if we are going to look to the 
political scientists for guidance on election 
methods, we'd want to see what they actually use, 
and STV is in the Constitution only as a contingency.

The President, with the advice and consent of the 
Council, another elected body, appoints a 
Nominating Committee which names a single 
nominee. If there is no other nominee, this 
candidate is elected at the Annual Meeting. 
However, it is possible to nominate other 
candidates by petition. The last time there was a 
petition candidate was almost forty years ago. In 
order for them to use IRV, there would have to be 
a second petition candidate. The chances of that 
can be estimated at once in 1600 years.

Wait, what about the Council? Ah, yes, the 
Council. Plurality-at-large. There are as many 
candidates nominated for the Council as there are 
seats, but there are some petition candidates, 
often. Voters vote for as many seats as are open, 
and the candidates with the most votes win. So 
... political scientists are actually using 
Plurality. Period. They are not actually using 
IRV, and they set it up so that the need for IRV 
would be extraordinarily rare. Should we do this.

Now, next factoid: to repeat it: "Robert’s Rules 
of Order recommends IRV over plurality voting."

Well, it doesn't exactly recommend it. It says 
that "preferential voting" gives a fairer result 
than plurality voting if it is considered 
impractical to use repeated balloting, which is 
what they *actually* recommend, indeed *require* 
unless bylaws specifically permit otherwise.

It then states that "there are many forms of 
preferential voting." We've named some above. 
Borda Count, for example, or what Brams calls 
Fallback Voting, more widely known as Bucklin 
Voting. But Robert's Rules then describes the STV 
method "by way of illustration." And then it 
proceeds to tell us the problems of this specific 
described method: it "deprives" voters of the 
opportunity to base later choices on the results 
of earlier rounds (which is what is allowed with 
Top-Two Runoff) and it can fail to find a 
"compromise winner," and it blames this on candidate elimination.

This latter failure is the "Center Squeeze" 
phenomenon of IRV: a candidate might be 
preferred, collectively, over every other 
candidate, but merely because the candidate does 
not get enough first preference votes, it is 
eliminated before all the votes that would show 
this are counted. Those votes, in fact, are never 
counted. We'll come back to that.

Attempting to make a virtue out of a vice, 
FairVote has invented the Core Support Criterion 
which is a reference to this. Suppose that there 
are three candidates in an election, and their 
partisans are about equal in number. Add a fourth 
candidate who is the first choice of nobody, but 
who is everyone's second choice. According to the 
Core Support Criterion, a candidate should not 
win who is not the first choice of at least one 
voter, so this obvious compromise candidate, 
preferred by every voter by a large margin over 
whoever would win the IRV election, can't win. 
Preferred by, in the example given, a vote of three to one.

Let's make a more reasonable example, the Core 
Support Criterion is based on something utterly 
impossible, I'd say. Suppose there are three 
candidates running a close race. Any one of them 
could have the lowest first round total. With 
IRV, that one would be eliminated. Now, suppose 
that one of the three candidates is really very 
broadly trusted, and is the second choice of 
nearly everyone who does not have him or her as a 
first choice. I can think of candidates for whom that might be true. So we have

34: A>B>C
33+:C>B>A
33-:B>A>C

There is nothing particularly unreasonable about 
this profile. B is eliminated in the first round. 
But B would beat, in a pairwise election, A, the 
IRV winner, by a vote of 66:34, a landslide. And 
with just a few more votes, B would win the IRV 
election as well, *by the same margin*, because 
one of the other candidates would be eliminated 
and those votes for B would be revealed. So a 
very few votes in the first round flips the 
result from B, a landlide winner, to one of the 
others. Which one depends on where the votes taken from B went.

This problem is serious enough that Robert's 
Rules mentions it. FairVote will tell you it is very rare.

Of course it is rare. This problem does not 
happen in a two-party system! And IRV, remember, 
helps maintain a two-party system.

Is this problem inevitable? Definitely not! 
Consider Fallback Voting (a.k.a. Bucklin). The 
first round does not find a majority, so the 
second round votes are added in, the election 
becomes an Approval election. Wonder of wonders, 
B is elected with a vote of 100%! That is 
because, when the approval threshold is lowered 
to include second rank choices, every voter has 
placed B in second rank or higher.

Any Condorcet method would also, of course, elect 
B, because B is the pairwise winner. Approval would probably do the same.

Now, FairVote will doubtless point out that 
Bucklin doesn't satisfy Later-No-Harm. 
Later-No-Harm, however, is incompatible with the 
basic principles of majority rule, which requires 
compromise if decisions are to be made. Robert's 
Rules is utterly unconcerned with Later-No-Harm. 
But here is how it would affect the election:

Some voters would think that if they vote for B 
in second place, this might help B defeat their 
favorite. So they wouldn't vote for B, they would 
"truncate." However, real voters don't mostly 
think like that, even though political activists 
wish they would. If a voter has a very strong 
preference for the favorite over all others, then 
they won't rank others, quite properly. But if 
they have a weak preference, they will; in the 
example given, B is widely known and respected. B 
might not get a two-thirds majority, but it only 
takes a relatively small number of the other 
voters ranking B second to bring B the victory.

Continued with:
Dopp: 12.“Not all ballots are treated equally
”