[EM] IRV vs Plurality

[Abd ul-Rahman Lomax]

At 02:38 PM 1/14/2010, Kathy Dopp wrote:
>On Thu, Jan 14, 2010 at 1:44 PM, robert bristow-johnson
><rbj at audioimagination.com> wrote:
> >
> > On Jan 14, 2010, at 1:17 PM, Kathy Dopp wrote:
> >
> >> On Thu, Jan 14, 2010 at 12:34 PM, robert bristow-johnson
> >> <rbj at audioimagination.com> wrote:
> >>>
> >>>
> >>
> >> I'm glad to hear you don't support IRV/STV methods.
> >
> > not over Condorcet.
> >
> > i dunno what else you could have drawn from either my posts here or at the
> > FairVote.org site where i took on Rob Ritchie *several* times 
> (and i've seen
> > you there, too).
>
>My entire life's focus is not in following all election methods debates.

Kathy is a voting security expert, not a voting systems expert. She's 
learning, though. And she knows to ask experts, but people do express 
their opinions in process. Which is good, the fastest way to learn is 
to make a mistake and read responses to it, and it helps if there are 
experts or others who know who will be patient enough to do that.

Would that FairVote and Rob Richie had listened, they'd have learned 
and they would have modified their strategy to focus on deeper and 
more effective goals. The ultimate goal of FairVote, in its 
foundation, was proportional representation, but they got stuck on a 
political choice as to how to implement it, and then on a strategy to 
faciliate the adoption of that method by using a method which is 
crazy for single-winner, but which is better for multiwinner. When 
one is determining representation, the goal is to have one vote from 
each voter go to one candidate, or to be split among candidates, and 
the goal is to maximize this so that as few votes as possible are 
wasted and the maximum number are represented.

So Later-No-Harm makes sense when finding representatives where most 
votes will find their way. But STV, the general method, is still 
defective, and there are better methods, including ones much easier 
to canvass, so trying to get IRV in order to get the basic STV voting 
system in place was putting the cart before the horse. A political 
error that, I'm sure, looked good at the time.

And what happens with political errors when someone bets their career 
on it? Summary: it gets hard to change course, unless the person 
involved is able to see beyond their own limitations. Not a common 
skill among most political activists, who are trained to be bulldogs 
who don't give up no matter what arguments are tossed at them. Debate 
skills. People are taught how to debate to win, not to produce the 
most sensible result after all has been considered. It's too often a 
piece of an adversarial system, with gladiatorial combats, and then 
the crowd gives thumbs-up or thumbs-down. Entertaining, but not 
particularly efficient for producing intelligent and sustainable 
decisions with a distracted crowd. There are much better ways that 
will work with people-as-they-are.

And it only takes a few people to realize this to start building the 
structures. It is *not* necessary to convince the masses. That will 
come later, after they have examples to look at, which is what most 
people need.

So my political recommendations are based on what is already known, 
what has been already tried, with only minor variations beyond that. 
Approval voting, one might note the critics state, can default to 
Plurality if most voters vote for their favorite and leave it at 
that. *That's fine.* Most voters, indeed, will vote that way in most 
elections. So it's *harmless* and can be tried, particularly since it 
is essentially no cost to just Count All the Votes, and we should be 
doing that anyway! Does anyone think that it's an irrelevant and 
worthless fact to know how many ballots in Florida 2000 contained 
votes for each of the candidates, regardless of whether or not some 
were overvotes? Even if this information couldn't be used in that 
election because the rules required disregarding overvoted ballots. 
Instead, what's reported is this: overvotes are reported only as 
spoiled ballots. The votes on them are not counted at all. So the 
damage is concealed from us. If a rule has disproportionate impact, 
it can't be seen.

Count All the Votes. And then, I claim, we should use the votes that 
are counted, and political theory generally says that Approval 
Voting, which is simply a matter of Counting All the Votes, is quite 
a good method, superior to plain Plurality, and simply defaulting to 
Plurality if people just vote for their favorite. The extra votes do 
no harm, generally, and can help if there is a minor candidate who 
might otherwise be a spoiler, if the supporters of that candidate are 
willing to add an additional vote for a frontrunner. By definition, 
this is normally a relatively small number of people. But if we are 
concerned about the spoiler effect, here is a simple fix that costs 
nothing. It's not a complete fix. It gets more complete if we use a 
ranked ballot and phased-in approval votes. I.e., Bucklin.

FairVote has claimed that Approval has never been used in the U.S. 
for a public election. It's deceptive, there were something like a 
hundred towns that implemented Bucklin in 1910 or so to 1920 or so. 
Bucklin phases into approval voting if a majority isn't found in a 
round, adding additional approvals that the voters may have added in 
lower ranks. If all ranks are counted, it becomes pure approval 
voting. So that's one exception. The other is that approval voting is 
used, in effect, when there are multiple ballot questions. If any of 
them pass, the one that wins is the one with the most Yes votes. This 
is quite equivalent to approval voting with a majority requirement. 
Is that an election? Well, it's a choice between multiple options 
based on votes on a single ballot. I'd call that an election. To call 
it otherwise is to miss the close similarity.

>Reread. Unlike you stating "You don't read" as if you have all-seeing
>ability to know that I never read, I **never** said "You are an IRV
>proponent". Please try to stop confusing your own imagination with
>reality and notice what I actually wrote and what you actually have
>ability to know, when responding to emails.

What I've seen from Robert is arguments that are the arguments used 
by IRV proponents. But maybe I'm wrong. And he might support some of 
those arguments but still be against IRV for other reasons. But my 
concern is the deceptive arguments that have been advanced by 
FairVote, including their arguments against other voting systems, and 
it's very important to expose these.

Many of them fooled many of us for a long time, including me. Tell 
me, does Robert's Rules of Order "recommend IRV"? Remember, this is 
being said in a context where a community is considering implementing 
IRV as a deterministic election method. Would that be supported by 
the parliamentarians who edit RRO, for use in a voluntary 
organization? And the answer is no, and I missed it, even though I 
was very involved with discovering the problems with IRV and even 
though I was familiar with parliamentary procedure. It's easy to 
miss, because it's easy, once one has the public IRV counting 
procedure in one's head, to read the procedure described there as 
being the same as the IRV procedure. It's not. There is a critical difference.

Take a look at the RRO copy that FairVote hosts, found at 
http://archive.fairvote.org/?page=1797, and see if you can spot it. 
Now, when I found this, FairVote activists argued tenaciously that I 
was blowing smoke, that my interpretation was ridiculous. But then I 
noticed something else, in the same section, that I'd overlooked 
because it isn't in the section describing the procedure. But it 
makes the meaning of the procedure crystal clear. It's not what the 
FairVote activists were proclaiming, and any parliamentarian would 
confirm that. By the way, I'm not formally trained, but I have served 
as a parliamentarian.

Would a parliamentarian miss the difference? Possibly, but for a 
different reason. They might not understand the public counting 
rules, because they would assume, unless they noticed the difference, 
that the terms being used had their normal meanings. They don't. They 
have a special meaning, easily overlooked, as it was overlooked by 
the supposedly neutral commission that wrote the voter information 
pamphlet for San Francisco in 2002. If some members of that 
commission knew what they were saying, they were deliberately lying, 
and the effect, regardless, was that the voters of San Francisco were 
deceived about how IRV works. And if you are not an expert, and don't 
put a lot of time into the topic, most people would be deceived.

And I'm deliberately not revealing this straight out, this time, 
because I want to encourage people to look for themselves. Think of 
it as a game, if you don't know what I'm talking about.

Since I found this and pointed it out, FairVote activists, some of 
them, have become a little more careful about making the claim, they 
phrase it so that the claim becomes sort-of true, but the 
*implication* of it remains deceptive. From an insider report, I do 
know that at one time FairVote held meetings to educate people how to 
advocate IRV, with various arguments to make and emphasize. One of 
these activists came after me on Wikipedia, but, unfortunately, the 
guy was really smart and actually read my response. And said, "I've 
been working on the wrong side," and he started trying to help what I 
was doing. But most activists dedicate themselves and inoculate 
themselves against any arguments from "the other side." It's part of 
the ethic of loyalty to the cause, and that ethic is part of what's 
wrong with our system, overall. Voting is a small part of it.

When your goal becomes winning instead of honesty and maximization of 
consensus, based on accurate information and fair consideration of 
the various points of view, you have become an actual enemy of 
society, overall, and if you help instead of harming, it's almost accidental.

Focus on pure winning makes sense in the heat of a gladiatorial 
contest, but, note, the gladiators served a very unhealthy system, at 
the expense of themselves, they were pawns, sacrificed for 
entertainment, fighting each other to the death, which, rather 
obviously, wasn't good for gladiators. Sooner or later someone else 
is faster or stronger or one slips.

> > about this, Kathy, i don't believe your veracity at all.  since March of
> > 2009 (when Burlington IRV failed to elect the Condorcet winner 
> and all sorts

Kathy may make mistakes, but I'd be astonished to find her lying.

>In my own imagination, I **do** support the Condorcet method, although
>I don't know how to solve the Condorcet cycles or how often, if ever,
>they might occur.

There are Condorcet-compliant methods, and the first-order intuition 
of most of us who start studying voting systems is that a Condorcet 
winner should always win the election. Turns out, no. Not 
necessarily. But the exceptions are probably relatively rare, and, in 
order to understand it, you need to have a deeper understanding of 
the science of public choice than is possible with only consideration 
of pure ranking.

>However, I see that once again you are certain that you know more
>about myself, what I do, and my own positions, than I do via your own
>imagination, which it seems, you feel no need to verify with any
>outside facts.
>
> > but my position here has *never* been as a proponent of IRV, but 
> a proponent
> > of Condorcet and the ranked ballot.

Kathy hasn't been big on ranked ballots, because they have generally 
been associated with the seriously complicated method of IRV, or the 
less complicated, but still complicated, Condorcet methods. Again, as 
I mentioned, the Condorcet Criterion looks good, it's "intuitively 
satisfying." Unfortunately, it depends on pure rank order, neglecting 
preference strength.

Still, if all you have is a ranked ballot, and equal ranking is not 
allowed, the Condorcet Criterion is probably the best that can be 
done. That's because a simple ranked ballot does conceal preference 
strength information. Warren at one point discovered a paper where an 
analyst noticed an anomaly with a deeply ranked Condorcet ballot 
where the Condorcet winner was rather obviously the wrong choice. But 
he made an assumption of equal preference strengths, averaged, over 
the rankings, or a fair and reasonable distribution of the candidates 
in issue space. I forget the exact argument.

But we can get at it in this way. Consider a Borda ballot. Borda is a 
ranked method which assumes equal preference strength in each 
preference expressed. As normally used, equal ranking is not allowed, 
and all ranks must be assigned (or the voter's ballot is discounted 
in some way). So a Borda ballot with (N+1) candidates translates to 
Range N, with the restriction that a range vote can be assigned to 
only one candidate, and all possible Range votes are advisedly used 
for a full strength vote.

For those not familiar with Borda, here is more detail: the voter 
ranks the candidates, say it is favorite to least favorite. There are 
various ways of stating the canvassing method, but one way would be 
that each candidate is assigned a value from the rank, with the 
highest rank being the number of candidates minus one. The lowest 
rank is then zero in value. The value is the number of the candidate, 
starting from zero, proceeding up to the highest ranked candidate. 
The winner is the candidate with the highest total value summed from 
all the ballots.

This is quite equal to Range N, with no overvoting allowed at any 
rank and no empty ranks allowed. If you allow overvoting at any rank 
and therefore empty ranks, it's Range. Borda assumes that if you rank 
A>B and these are in sequence, a vote strength should be assigned 
according to how many intermediate candidates are ranked in between. 
This is obviously an approximation, even if the voters are Borda's 
famous "honest men." He missed the point! If the approximation is way 
off, as it can be when there are clones, or large missing segments of 
the issue space not represented by a candidate, the method can be wacky.

Donald Saari somehow seems to have overlooked that Range, which he 
criticizes heavily, is the same as Borda, which he actively promotes, 
and the only difference is this imposed restriction that is obviously 
artificial and which doesn't correspond to reality. Basically, like 
many voting systems activists and experts, Saari doesn't trust the 
voters. That's where I differ. They may make mistakes, but if they 
express a strong preference or a weak preference, when they have the 
unconstrained choice, I assume that an expressed strong preference 
*is* a strong preference, because there really is no *significant 
reward* for lying about it. Saari and others miss this completely, 
imagining that voters will necessarily cast strong votes because they 
"want to win," but neglecting the fact that voters have other goals 
than simply maximizing their favorite's chances and doing everything 
to hurt the opponent of the favorite.

Sure, they will do that if they *actually have strong preference.* 
But if their preference is, say, maximally weak, and they know that, 
will they act in that way? Especially realizing that, by hurting this 
clone, they might be shooting themselves in the foot, if the clone is 
the only hope of defeating a much worse candidate!

I've done the study, and my conclusion has been that the optimal 
voting strategy in Range is a kind of sincere vote. And the voters 
will do best if they can choose that vote themselves, if they can 
express preference strength themselves without being forced into some 
specific model that constrains their votes without necessity.

A voting system should be simple to vote. Range voting does require 
some more complex though, but the simplest Range method is quite 
simple, and the strategy is usually quite simple as well, and it gets 
even simpler if it's a primary with a runoff in case there is 
majority failure. The same question can be asked of each candidate: 
would you prefer this candidate to be elected, or would you prefer a 
runoff? If you'd prefer to elect the candidate, vote for the 
candidate. If you'd prefer a runoff (which brings other risks and 
costs), don't. That's quite simple, and it motivates the voter to 
vote sincerely, and it is an effective strategy, that is, it will 
monotonically accomplish what the voter desires, if possible.

Bucklin makes that choice much easier because it allows the voter to 
have the preference expression cake *and* the possibility of avoiding 
majority failure. Same question is asked, but now it's about lower 
rankings. I advocate allowing equal rankings with Bucklin, because 
they are harmless at worst, and they give the voter additional 
flexibility, and they avoid spoiled ballots. It makes each Bucklin 
round of counting into an Approval election, not just the later 
rounds. But a voter would only, presumably, add multiple candidates 
to a rank if the voter really doesn't have a major preference 
betweeen them. It's an honest vote, and strategically optimal. This 
aspect of Bucklin appears to have been neglected by students of 
voting systems, along with the implications of majority requirements 
and runoffs.

What does this have to do with Condorcet failure? Well, Borda and 
Range fail Condorcet, by any interpretation. But it is quite arguable 
that, especially with Range (Borda without the artificial 
constraints), a situation where it fails is one where the Condorcet 
winner is not optimal. Very simple to show, and the simplest 
Condorcet winner is a majority preference winner, and, for terminal 
simplicity we can just consider two candidates from a much larger 
candidate set. Suppose that there are two candidates, A and X, and a 
majority prefer A over X. If one of the set of A and X are elected, 
must it be A? (This is the Majority Criterion, a stronger criterion 
than the Condorcet criterion, any method satisfying the Majority 
criterion must satisfy the Condorcet Criterion, but the reverse is not true.)

Okay, remember that early paper I mentioned. They noticed that if the 
A>X preference was really A>B>...>W>X instead of A>X>B>...W, it was 
obvious that the voters really don't have -- other things being equal 
-- much preference in the second case between A and X. Why should 
this majority preference fail over a strong sincere preference (we 
are assuming sincere votes!) expressed by a minority of voters?

When we began studying voting systems using simulated utilities, with 
Warren Smith being a major pioneer, and when Range voting was 
proposed as a method -- which simply allows voters to express 
utilities for each candidate, on a scale from 0 to 1 full vote 
(though often expressed as numbers from  0 to N for what I call Range 
N) -- it became much easier to express this as votes.

51: A 100, X,  99 ...
49: A   0, X, 100 ...

In real voting, voters are likely to normalize, so if there were only 
two candidates, A and X, the vote shown would be unlikely. But not 
noticed by most voting systems students is that, if these are sincere 
utilities in some way, on some appropriate scale, the A voters aren't 
highly motivated to vote, they wouldn't even turn out, so ... what we 
must assume is that there is some election process underway that 
attracts them for other reasons. We can also assume that there is a 
fair range of candidates, so that the votes shown are reasonable, the 
A voters actually did intend to indicate that they hardly cared at 
all between A and X, whereas the X voters really did intend (and were 
sincere about it) that they cared very much. Perhaps they'd start a 
revolution if A is elected. (And that's not actually far-fetched in 
some situations, you neglect the strong preferences of a minority at 
your peril, don't do it unless you, yourself have strong preferences. 
And realize that then you might have a war on your hands. Are your 
preferences *really* that strong? Maybe you better start talking with 
the other side!)

Sometimes when I bring this up, it's assumed that the X voters are 
fanatics, and it's somehow "bad" to reward "fanatics." Maybe, but 
fanatics are people and have rights as well, and that's a really 
stupid objection, because strong voting preferences are just as 
likely to come from knowledge as from ignorant fanaticism. People 
with knowledge tend to have strong opinions! In some ways. Sometimes 
in other ways knowledge makes us more thoughtful, but that doesn't 
really refer much to knowledge of candidate positions and character.

In the situation described, Range voting will come up with a result 
that reflects optimized overall voter satisfaction: X wins, and 
*every* voter is very satisfied, if we can trust the way they voted. 
And if the A voters really didn't like that outcome, their votes were 
singularly stupid! Yet practically this exact scenario but more 
exaggerated, is asserted as an argument against Range Voting. Suppose 
100 voters with a maximally weak preference (1/100 vote per voter) 
with 99 voters vs. *one voter* who votes for the other candidate, who 
then wins by 1/100 vote! Wow! The preference of 99 voters is 
overpowered by the preference of one single voter. Is that obviously 
bad or what? No, it's not bad. It's actually ideal, though in the 
case mentioned, it really makes almost no difference in overall voter 
satisfaction if either candidate is chosen, it's been created as an 
example on the brink. Personally, I'd be happy to live in a society 
where the trivial preferences of many are set aside to allow 
satisfaction of the strong preferences of a few.

And, I fact, I live in a society that does this all the time on a 
small scale. People will routinely slow down a little, sacrificing a 
little time, but only a little, to let in someone waiting to make a 
turn, a major benefit for that person, comparatively. Friends will 
make a choice, making sure that a strong preference of any member of 
the group isn't violated, even if that means a majority of them give 
up their favorite choice. I've called this the Pizza election, for 
those who might have read this stuff for a while.

Why should we not have a voting system that does this on a large 
scale. The point isn't that we should drop everything and implement 
range voting, but that the Majority Criterion is defective, requiring 
a result that in some situations is very clearly suboptimal. Which 
would you rather satisfy, the trivial preference of a majority by a 
small margin, or the strong preference of a minority that is almost a 
majority. If you do the latter, the overall satisfaction metric for 
the result almost doubles. It's not close, but, of course, the 
example was designed for that.

The point is that when some methods fail the Condorcet Criterion, 
it's a disaster, a true loss of overall satisfaction, but when Range 
fails the Criterion, it's a case where the Criterion is quite likely defective.

What to check it out? Hold a runoff election whenever a Range winner 
differs from a Condorcet winner. It will hardly ever happen, the 
simulations show. What this is equivalent to is asking the majority 
if it's willing to give up its small preference. Was it truly a small 
preference, or was this some product of quirks of voting strategy and 
possibly incorrect knowledge of the situation?

And, from this, can you start to see the power of repeated voting?

Look, this isn't new. Democratic organizations have long used 
repeated voting to elect officers. They have absolutely no incentive 
to use advanced systems (though I now know enough to suggest some 
improvements for efficiency, that will do no harm), because repeated 
voting with a majority required is actually extremely powerful, 
particularly when one realizes that there are no eliminations. IRV 
simulates a series of repeated elections with a single candidate 
eliminated each round, the candidate with the least number of votes. 
That method isn't used in standard deliberative process, the 
"repeated voting" is really repeated elections, and each election is 
entirely new. New nominations are possible! Practically anything is 
possible with deliberative process, it can't be simply predicted from 
static preferences. Maybe that's why it hasn't been studied much. 
It's hard to study!

But it's well known, and it works, far better than the primitive 
systems used typically in public elections. It's possible to move 
beyond those primitive systems in one giant step. Asset Voting. But 
that's a story for another day.
> >
> > i've said multiple times that IRV transferred the burden of having to vote
> > strategically from the majority (in Burlington, that would be the liberal
> > that would have to split their votes between Prog and Dem) to the minority
> > (in Burlington that would be the GOP Prog-haters that discovered that their
> > primary vote for their favorite candidate was instrumental in electing the
> > candidate they least preferred).
>
>GREAT. Well we agree on many things, even if I think you are slightly
>delusional for thinking you know so much more about me than I do re.
>my "not reading" and my positions on election methods.

Actually, the strategic voting required for the majority, in 
Plurality, would be for the smaller of the two liberal parties to 
decide to vote for the other, and this is best done by negotiation 
before the election. In New York, with Fusion Voting, two parties can 
nominate the same candidate, thus keeping ballot rights for the 
party, but not splitting the vote. The smaller of the parties can 
negotiate policy concessions from the larger. They can agree on power 
sharing arrangements, etc. But it's not true that "the majority must 
vote strategically." If it's a majority, they can vote sincerely and 
simply and will win. What's been confused here is that the two 
factions together are a majority. But half that faction, roughly, 
doesn't need to change it's vote. Suppose it did. Disaster if the 
other half does the same!

No, to make sense, they need to find consensus on a single candidate. 
Hold some combined primary, at their own expense, and use it to 
determine party endorsement. Get it together, folks, because if you 
don't, the other side may eat your lunch!

"GOP prog-haters" is excessively polemic. Those who supported Wright 
strongly over Kiss, which is apparently the large majority of the 
Wright voters, will indeed discover that problem if they look. This 
doesn't necessarily have anything to do with their preferred 
political party. But party affiliation *does* have an effect in 
Burlington. Probably not a good one!

Some people who start looking at these things assume that most voters 
are affiliated with the party whose candidate they vote for. Maybe. 
Sometimes. But in a town like Burlington, many people vote based on 
personal perceptions of the candidates, probably even a majority. I'm 
politically progressive, generally, though with a fairly common 
libertarian streak, but I'd vote for a Republican for a local 
election, and probably have. But usually because most local elections 
are nonpartisan, I don't know their party and don't care, because 
local politics are not so strongly affected by party affiliation, and 
a Republican who works for his or her town may in the end do quite 
the same as a Progressive, or, more accurately even, some Republicans 
might take actions that are more progressive, actually, than some Progressives.

It's an anomaly that Burlington even has partisan elections for local 
office, I think. We can tell that party affiliation is important to 
some voters because in nonpartisan elections in even very liberal 
towns like San Francisco, vote transfers don't alter preference 
order. That happened in Burlington this last year, and it was a 
strong effect. It's due to party affiliations giving voters a 
how-to-vote card, in effect! If you are a Progressive, you know 
already a certain kind of vote to cast: Progressive>Democrat. 
Likewise if you are a Republican, you have Democrat>Progressive. As 
stereotypes.

But if you are a Democrat, you are in the middle, and you can and 
will split your second rank vote, if you cast one, between the 
Progressive and the Republican, depending on where you sit. And the 
whole thing is complicated by the presence of supporters of 
independent or other-party candidates and unaffiliated voters.

It's obvious to me what Burlington should do, and it's quite simple, 
and they are already prepared to do it, if they so choose. They are a 
three-party town, and so they need an advanced voting system, and one 
is handy, has been widely tested and tried, and is far easier to 
canvass, and allows the voters to vote with full sincerity and 
flexibility. It defaults to Plurality, if voters don't add 
additionally ranked votes, and if they want some vote margin or there 
is a runoff, it works even better and, used in a runoff, it can even 
allow write-in votes without harm. No spoiler effect.

Gee, what would that be? Hint, besides other obvious ones: I didn't 
invent it, it was invented roughly a century ago (more I think, but I 
don't remember the exact date). It's been tried in roughly 100 cities 
and towns in the U.S., but, remember something: advanced voting 
systems aren't necessarily needed in a 2-party system, and when there 
is a 2-party system, the 2 parties don't want competition from minor 
parties and the larger of the two major parties has a strong 
incentive to lock the minor parties out. So, I'm pretty sure, that's 
what they did. That this method was rescinded in all those places 
means practically nothing about its usefulness in a three-party town. 
IRV breaks down horribly in that situation, whereas simple Approval, 
just allow all the votes to be counted, would be much better with no 
cost at all to speak of.