[EM] Why I think IRV isn't a serious alternative

[Abd ul-Rahman Lomax]

At 12:35 PM 11/26/2008, Jonathan Lundell wrote:

>This (TTR vs IRV) is a matter that we can simply disagree on.

Give it some time. Below, you indicate that you are "partial to the 
iterative process." So -- why not iterative process in public 
elections? It can be done, you know, the claim of impracticality is 
false. Top Two Runoff has the problems it has because the first and 
second rounds are using Plurality instead of a better method. TTR is 
a first step toward iterative process. Wouldn't you agree that it 
deserves study?

It's the most widely-implemented election reform, it has been 
considered an improvement over IRV, historically, and it has wide use 
around the world for single-winner elections. I don't like 
single-winner, fixed-term elections at all, preferring deliberative 
methods, i.e., a parliamentary system, where the people, through a 
representative body, elect officers. That reduces the problem to 
creating a representative body which truly and accurately represents 
the people. That's Asset Voting, which can use STV as a base (works 
fine!),  but can also be done much more simply: use Plurality as a 
base, but when there is majority failure, allow the candidates who 
received votes to negotiate on behalf of those who voted for them.

I've certainly seen knee-jerk objections to this, mostly based on 
assumptions that the candidate set would remain the same, candidates 
would be selfish, can't be trusted, would be corrupted, etc. However, 
that's democracy! Not the restrictions, but the *possibility* of 
them. Corruption under a mature Asset system would be extraordinarily 
difficult, only able to make small nudges.

Think about it: with Asset, Plurality suddenly becomes 
*strategy-free*. Vote for your absolute favorite, period. Choose a 
candidate you can trust. Isn't that the most important quality in a 
candidate for office? In modern offices, certainly major ones, the 
ability to make good choices in delegating authority is crucial. And 
that's what a candidate would do when holding votes for redistribution.

It's a spectacular method. Carroll was probably the greatest mind 
considering voting systems in the nineteenth century; his work was 
largely forgotten until recently. He looked at the biggest problem 
with STV and suggested a simple fix. It's a shame that it's never 
been tried, to our knowledge.

But it also works fine with *any* voting system. The system must 
define "majority" sensibly, and then how to assign the proxies -- 
that's what they really are -- based on votes. With Approval, for 
example, because Approval can't necessarily determine a "favorite," 
the votes would be distributed fractionally among the approved 
candidates. With Range, they could be distributed according to the 
vote. (The total of votes would be normalized to 1 vote, then each 
vote multiplied by the necessary fraction. Or a favorite marker could 
be used -- same with Approval.)

At this point, the important thing with Asset is to spread the 
concept, so that it isn't forgotten, and so that it comes up as a 
possibility whenever serious reform is being considered. It's 
essentially be done before, i.e., proposed, there were serious 
initiatives in the first part of the twentieth century to, for 
example, create a city council in which the winners exercise the 
votes of all those who voted for them. That's really proxy voting, 
which, of course, is practically universal in large business 
organizations, and is common-law allowed with property rights unless 
specifically prohibited.

Now, if iterative process is a good thing (and it's the foundation of 
how democracy operates, being what is done in direct democracies, and 
what is done within the representatives in representative democracy), 
why not "simulate" it, i.e., approach it with some form of runoff 
voting. Absolutely, it's not a new idea. It's what Robert's Rules of 
Order Newly Revised *prescribes* unless a bylaw has created something 
else (such as election by plurality, or IRV that neglects the 
exhausted ballots, which amounts to the same thing, only simply a bit 
more likely to find a majority than plurality). RRONR doesn't allow 
candidate elimination. Top Two Runoff sometimes does not either! In 
those cases, "Top Two" refers only to the right to be on the runoff 
ballot, which focuses the electorate on those two. An electorate with 
a majority preference for another candidate can -- and has -- fixed a 
problem with the primary or other conditions that kept that candidate 
off the runoff ballot. They can do it even with a plurality, if 
plurality is adequate to win the runoff.

But with plurality as a method for the runoff, that write-in 
candidacy is dangerous, it can bring up the spoiler effect. Hence, a 
great improvement over standard TTR would be allowing write-ins, as 
is done in some places -- and using an advanced method for the runoff 
that isn't subject to a true spoiler effect. The simplest is 
Approval. Because of the restricted candidate set (perhaps two on the 
ballot plus one serious write-in campaign), IRV should also work for 
the runoff ... but why bother? Approval does it, and if one must 
allow a favorite expression, Bucklin. Much easier to count. More 
likely than IRV to find a majority of votes for the winner, not that 
this matters much if it's a plurality election, i.e., must complete.

But we should always keep in mind that the biggest creator of 
election paradoxes and problems is "must complete." It deprives the 
electorate of a basic democratic right, the right of the majority to 
express that they are not yet ready to decide. That they would prefer 
another ballot to electing one of the existing candidates. With good 
methods used for the primary and runoff, runoffs could be minimized.

But runoffs ameliorate a very serious problem which afflicts all 
preferential systems and, in fact, Range voting as well: voter 
ignorance. That runoffs allow much better public focus on the 
realistic choices is a classic argument for it. It's *deliberative* 
process, not pure single-ballot aggregation.

So: what's the objection? The classic one is low turnout, but when 
voters care, they turn out. Turnout was phenomenal in the Lizard vs. 
Wizard runoff and the Chirac-Le Pen runoff. Further, various means 
*could* change this. A classic one is to reverse the position of the 
primary and runoff, not in sequence, of course, but in connection 
with the general election.

However, due to Range-like effects, I contend that low turnout in 
runoffs is not pathological. It simply indicates relative low 
preference strength, at least to some degree. Holding a primary with 
a general election means that more voters vote who really don't care 
about the outcome. This is essentially noise, if we are trying to 
maximize public satisfaction with an outcome. More motivated voters, 
which includes more informed voters, will turn out to vote in a 
runoff. This probably improves results and overall public satisfaction.

Chirac-Le Pen-Jospin in France, 2002, caused many to question the 
runoff system. But it was not the runoff that was the problem, that 
made, clearly, the best choice given the input. The problem was 
twofold: a limited method of choosing the top two, strongly dependent 
(like IRV) on first rank preferences, which can easily eliminate a 
compromise candidate, broadly acceptable, like Jospin (almost 
certainly the Condorcet winner, and by a large margin), but *because 
he's a centrist (in France, he's a socialist, he was the Prime 
Minister when running for President)* he wasn't the first choice of 
over 80% of the voters. His primary vote was really about the same as 
Le Pen's, an extreme right candidate. So the problem, really, was 
that the first round did not choose the best two candidates, by a 
large measure. This election ought to give IRV supporters pause. One 
of the methods called "IRV" by FairVote is a closer simulation of 
top-two runoff, it's batch elimination of all but the top two 
candidates. While it is conceivable that voting strategy would have 
changed with this, in France, had this been the method, it's actually 
not likely, the position of Le Pen was a big shock. So with "IRV" in 
France, it's quite a reasonably possibility -- I'd give it about 
fifty-fifty, because the votes for Le Pen and Jospin were so close -- 
that batch elimination IRV there would have elected Chirac. Not 
Jospin. And, in fact, I consider the odds about the same if it had 
been fully-ranked, sequential elimination IRV.

As I've mentioned, IRV was replaced, previously, by top two runoff in 
the U.S. A research project for some enterprising student would be to 
study why that happened. Did they know something we don't know -- or, 
more accurately, have collectively forgotten? Or was this simply one 
more corrupt or blindly ignorant action, like Brown v. Smallwood and 
the elimination of STV in New York?

The telling arguments against TTR, FairVote knows very well.

(1) Runoffs are inconvenient. Missing from the FairVote information, 
of course, is the cost, in terms of loss of democratic process and 
election quality, of eliminating runoffs. There are old arguments for 
runoff voting. You don't see FairVote answering these, they behave as 
if those arguments don't exist!

(2) Runoffs have low turnout. Somehow this is assumed to mean "poor," 
as in a "poor election." We read, often, "turnout was poor." But low 
turnout means, very obviously, that voters who don't care haven't 
voted. This actually removes noise, and almost certainly improves 
results. I've never seen a claim that real runoffs made results worse 
that wasn't coming from someone who benefited from massively ignorant 
vote. (Tends to be Democrats in some circumstances, but this also 
comes from Republicans who are able to touch on hot-button issues and 
convince relatively ignorant voters that they are the best.) The 
primary, if held with the general election, is thus much more 
susceptible to advertising and spin in the media. It's much tougher 
to convince the more motivated voters in a runoff to vote against 
their interests.

(3) Runoffs cost money. Yes, they do. But so does IRV, and the jury 
is out on which costs more. FairVote doesn't back up their claims 
with hard data, and you don't see FairVote mentioning the *very* 
delayed results, with the City of San Francisco spending lots of 
money for hand counting IRV, or lots of money for voting systems that 
can handle IRV. There are many problems here, but the election 
security people are pretty much united against IRV. FairVote claims 
that IRV was a "big success" in Cary, NC, but that seems to be based 
on a single poll, that voters liked the system. On the other side, we 
have a lot of anecdotal evidence that many voters were confused, we 
have statements from candidates and election officials that it was a 
mess. The hand counting was a disaster. Cary is not likely to repeat 
the experiment.

So IRV is proposed, which, it is often stated, "simulates" a series 
of runoffs. However, it doesn't simulate them well, at all. In TTR, 
there is a comeback election about one third of the time -- with 
nonpartisan elections, which is the vast majority of U.S. IRV 
implementations so far. With IRV, same jurisdictions, no comebacks. 
IRV is electing the plurality winner, the candidate most preferred by 
the voters in first preference, vote transfers aren't changing the 
first round result. This is almost certainly due to ballot 
exhaustion, and, unlike what Mr. Lundell might think, ballot 
exhaustion is inevitable, for a large percentage of ballots, even if 
full ranking is allowed. Voters don't want to fully rank, they don't 
want to vote for one candidate over another unless they understand 
both candidates and their positions, and most voters simply are not 
-- and are not going to be -- that informed. They know their 
favorite. That was Carroll's realization, how to empower these 
voters, in a fair way, and to use that information, the favorite.

>  I'd set
>three-rank IRV aside as an unfortunate but hopefully temporary
>response to voting equipment limitations.

It horrifies Chris Benham, an Australian, but it's difficult even for 
knowledgeable voters to make intelligent ratings of all the 
candidates. In Australia, full ranking is always allowed in the IRV 
elections, and in most places it is required or the ballot is voided.

(In Australia, voting is mandatory, it's against the law not to vote. 
This, as well, is highly likely to guarantee major-party dominance, 
because major parties have more resources with which to influence 
relatively uninterested voters, the ones who will simply hear an ad, 
and go "Yeah! Those gays are going too far!" Or the like. -- in the 
other direction, "The big corporations and greedy fat cats are the 
cause of all our problems.")

There is little reason to believe that adding more ranks than three 
would improve results, because the more ranks added, the fewer the 
voters who will use them. I'm in favor of giving maximum power to 
voters, generally, but I'd see a much simpler reform that would fix a 
lot of the problems with IRV:

Allow equal ranking. Make it like Bucklin, which required, in Duluth, 
a single vote only in the first two ranks, then as many candidates 
could be approved as desired in the third rank. Personally, I'd allow 
multiple voting in all ranks (and a ballot marked with the same 
candidate in more than one rank would only be counted as a higher 
rank vote, I'd assume. It would not be discarded. Count all the 
votes, simply interpret a vote like that reasonably.)

This would make IRV into Approval for those who wanted to vote it 
that way. And leave it as Later No Harm IRV for those who wanted to 
vote that way.

Which would voters prefer?

I'd simply use Bucklin, of course, the risk of a true Later No Harm 
violation is quite low.

Now, let's look at the French 2002 election again. The runoff had 
Chirac and Le Pen on the ballot. Le Pen got about 19% of the vote in 
the primary, Chirac a little over 20%, I forget the exact number. Was 
the runoff a close election? No. Chirac won with 80% of the vote. 
Essentially, the only people, for the most part, who voted for Le Pen 
in the runoff were those who voted for him in the primary. Turnout 
was higher in the runoff than in the primary.

Now, suppose that write-ins had been allowed in the runoff. Given 
that Jospin was probably much more popular than Chirac, and given 
that Le Pen simply wasn't going to win the runoff, even if Chirac and 
Jospin evenly split the vote -- i.e., it would have been 40-40-20, 
the problems with this election, if considered serious, would have 
been fixed, and quite safely, with a write-in campaign for Jospin. 
I'd predict he'd have won. But if not, then, indeed, the voters would 
have spoken. The ballot would have given Chirac some edge. But with 
very motivated voters, really exercised about the election, many of 
them saying that they were going to vote for Chirac with a clothespin 
on their noses, I'd say that the on-the-ballot edge would have been 
fragile. No amount of additional write-in campaigns would have 
allowed Le Pen to win, he was massively rejected and his only support 
was his core support. So much for the Core Support Criterion, 
FairVote's attempt to create an election criterion that can make 
sense to people who don't understand election methods, as an attempt 
to cover up IRV's massive criterion failures. It passes criteria that 
can be shown as actually undesirable! And it fails the ones that 
voting systems experts -- and basic democratic process -- consider 
important, such as electing someone like Chirac or The Lizard. IRV 
passes the Majority Criterion, because that criterion only refers to 
first preferences, but it can easily elect someone, in theory, who 
would lose by a landslide to a different candidate. Far from being 
rare, elections like the French one and the Louisiana Lizard v. 
Wizard, show that it must be reasonably common. You can't tell from 
the Australian IRV data, they don't provide the critical ballot 
information. IRV buries the necessary preferences, these are votes 
cast by voters which are never counted. (Consider what we'd think 
about a Plurality election where they didn't count votes from voters 
because it had become clear that a candidate was going to lose. As 
with a recount where, say, votes for candidates who didn't get 
something near a plurality were ignored. Why bother adding them up? 
-- Because we should Count All the Votes. Every Vote should Count, or 
at least be counted. Face it, IRV is a really poor election method, I 
wouldn't use it even for a poll in a democratic assembly. I'd use 
plurality because it's quick, and if I wanted better data, I'd use 
Approval or Range. As a poll, of course, Range is stellar.

One of my realizations a while back: while Robert's Rules of Order 
prohibits overvotes on written ballots, it is silent about them with 
respect to show-of-hands votes. There is no rule preventing a voter 
from voting for more than one candidate in a voice vote or 
show-of-hands or rising vote. I'm sure it happens. These votes are 
counted. (Voice vote is rare for elections, in modern times. It's how 
the Spartans did it, the winner was the one with the loudest vote. 
It's a kind of Range voting, actually.) Why does RR instruct the 
clerk to not count overvoted ballots? Because the "intent of the 
voter cannot be discerned." This assumes that the voter does not 
intend to overvote, that it was a mistake. It's against the rules 
because it is a mistake and it is a mistake because the rules require 
it not to be counted. Circular. There is no discussion as to why 
overvotes should be prohibited. I've been unable to find legislative 
history that explains the no-overvoting rules; apparently, as with 
Robert's Rules of Order, it's simply been assumed that people will 
vote for their favorite and leave it at that. In deliberative 
assemblies, it works, because a majority is required. The need for 
additional votes only appears when one goes to a single-ballot method 
that resolves without a majority being needed. In an assembly, 
single-vote may actually be better than Approval, except for polling. 
(Robert's Rules is a bit down on polling, considering it redundant to 
voting, but polling can make deliberative process proceed much more 
quickly. Range polling even more quickly.)

>  My problem with TTR is that
>it's almost as bad at encouraging strategic voting as FPTP is. Better,
>yes, but not good. Approval is also a strategy game that I'd rather
>not play; sure, I can imagine elections in which Approval is easy and
>relatively non-strategic, but it's also easy to imagine otherwise.

Stated, but no evidence that this is an actual problem. "Strategic 
voting" here means that voters cast an informed vote. They tend not 
to do that with TTR or with IRV, especially in first preference, 
which is all that you get in TTR, and which might be all that counts 
in IRV. I see little reason to believe that first preference votes 
would shift significantly between TTR and IRV. Both systems encourage 
people to vote for their favorite (exclusively in TTR, and first 
preference in IRV.)

It should be realized that Approval simply makes it safe to vote for 
your favorite. Otherwise Approval "strategy" is the same as with 
Plurality. Want to make your vote count? Make sure you vote for one 
of the frontrunners. With Approval, as with IRV, you can do this and 
still vote for your favorite. Bucklin allows this, as well, Bucklin 
strategy is quite simply in the vast majority of cases.

But IRV, remember, is vulnerable to Center Squeeze, just like TTR. 
Consider the Chirac-Jospin election. There were a lot of voters who 
voted, first round, for other than Jospin, who preferred Jospin over 
any realistic winner. If they have voted strategically instead of 
sincerely, first round, they'd have elected Jospin in the runoff. 
That is, a group of minor party voters, realizing the danger and the 
situation, could have changed the result for the better, not only for 
themselves but for the majority, by voting insincerely, reversing 
expressed preference.

If you are so concerned about strategic voting, why not be concerned 
about a method which makes strategic voting necessary for the method 
to come up with a good winner? With Range methods, normalized sincere 
ratings (that needs definition) will help come up with the best 
winner. Strategic voting, such as it is in range, may help you 
personally, but can damage overall results. And if you vote 
sincerely, the loss in personal gain, if any, is small. That's the 
whole point, something that FairVote arguments completely miss, with 
their example of 99% voting "sincerely" at 100, 99, and then comes 
along one voter who votes "strategically," at 0, 100, and outweighs 
all those "sincere voters." They will allege that this 99% of voters 
will be upset because they didn't get their favorite. However, the 
difference in rating, the loss to these "sincere voters" is tiny. I 
don't think I'd really have any difference in reaction to the news of 
the election of my 99% satisfactory winner vs the 1% one. Unless 
those ratings were really distorted, i.e., I didn't really, among the 
election set, have such a tiny preference strength. In which case the 
100,99 votes weren't really "sincere." They were "stupid"! I.e, not 
votes that would allow the method to determine the best winner, what 
are really considered "sincere" range votes under any reasonable definition.

Basically, most of the criticism of Range I have seen has been 
shallow and ill-considered, depending on knee-jerk reactions (like 
"vulnerable to strategic voting" -- generally undefined with any 
precision -- is a Bad Thing, worse than, say, can Elect a Candidate 
who would Lose in a direct face-off by a landslide to the Method 
Winner. Hint: IRV. The "vulnerability" of Range -- to the kind of 
"strategy" that Range can provide a small benefit for -- is minor, 
and does very little harm. The vulnerability of IRV to election 
pathologies: huge. Most elections, we should realize, would resolve 
with the same winner no matter what the method (among methods 
commonly proposed). So even a large series of "successful" IRV 
elections -- where the Plurality winner in the first round was 
elected, and there is no data to show much more than that, would not 
reveal the weaknesses, which take place in a minority of elections. 
Only simulations can uncover this with reasonable efficiency and 
accuracy, otherwise statistical variations swamp the possible results.

>I confess that I'm partial to the iterative process, at least under
>the right circumstances.

Yup. Nice confession, you are confessing a good thing. What 
circumstances? What makes it break down?

Obviously, iterative process takes more time and requires more work 
on the part of voters. Yet that is *all* we see in deliberative 
bodies, which also can be short of time. They have means of making it 
more efficient, such as committee systems which predigest decisions. 
In a community that values consensus, I've seen this work 
spectacularly. We are accustomed to committee systems that are 
politically controlled; that could be avoided. That is, the process 
is both hierarchical and iterative. It's all under the control of the 
majority, when the system is healthy, because the majority can 
overrule any officer's decision.

(Remember, folks, the flap in the Texas Assembly when the Speaker 
refused to recognize a motion to vacate his office, refused to 
recognize an appeal from the floor on that refusal, and declared the 
meeting adjourned without vote? It was fascinating to watch him, 
essentially, get away with it? He flagrantly violated the rules and 
basic democratic principles. The Parliamentarian resigned and he 
appointed two cronies to replace her. How did he get away with it? 
Everybody wanted to go home, they didn't have the political will to 
challenge him. He was a very skilled politician. The public was mad 
at the legislators because it was led to think that it was all their 
fault for being "partisan." Though the effort to remove him from 
office was actually bipartisan. Etc. That was an Assembly not 
accustomed to standing up for its rights. They could, quite simply, 
have refused to adjourn. When he left the room, it was as if he'd 
left his office and become absent. Standard parliamentary procedure 
can deal with a situation like this, so unless the Texas assembly 
(House?) had given up its rights in the standing rules -- apparently 
they had not -- it would have only taken one member with the 
knowledge and skill to fix the situation. Instead there was chaos. I 
find it remarkable that such a body didn't know how to deal with the 
situation. It's not like a chair never tried to take over and run 
things undemocratically before. -- Speakers do often sidestep 
democratic process, but they do so with the continuing consent of the 
majority, which avoids confronting it for political reasons. Since it 
is the majority which has the ultimate power, the situation 
continues, often until it gets really, really bad. Politics. It can 
create a lot of inertia, not in a good way.)


>  US political parties have used it in the
>past, and it's suitable to an open-ended convention setting.

That's right. But it was considered an abuse of power, allowing the 
"fat cats" in "smoke filled rooms" to make the big decisions. This is 
a populist argument based on the premise that we can't trust 
"politicians." It's part of what keeps the U.S. public from 
exercising its real power. Not allowing experts chosen by the public 
-- which could easily be done, and should be how parties are run -- 
to make decision means that big decision are referred to the public, 
through primaries. Supposedly better, in theory. But, in fact, highly 
vulnerable to manipulation by, what? Fat cats. Really Fat Cats. Using 
the media, which they buy.

We can overcome this, when we are sufficiently exercised, which we 
become when it gets really bad. Unfortunately, it has to get really 
bad to wake us up. There is a better way, and it is *not* getting 
everyone to be continuously active in politics. Better voting systems 
are a tiny, tiny part of that, Plurality is actually good enough, if 
the public is well organized outside of government, outside of big 
party structures, but independently and efficiently. And freely. It's 
about communication, not control. It's about trustworthy advice. How 
can we manage to create that, create a system that will give us the 
best advice, without creating centers of power that can be corrupted.

We know how to do it, but few even realize the need. That's what we 
need right now, simply more people who can see the problem and 
possible solutions.


>  The US
>Greens used a kind of "live IRV" in their 2004 convention, with
>multiple vote-for-one rounds with elimination and/or withdrawal
>between rounds. But there are lots of reasons that we don't want an
>open-ended process for public elections, or even a process (like
>GPUS-2004) that guarantees eventual termination, but with an uncertain
>number of rounds. I prefer the IRV compromise to TTR or Approval.

You have not stated why, and real election results show a serious 
problem. They don't show it in a way that enables me to say, it 
happened in *this* election, but statistically, it's pretty clear. 
We'll know even better when we have more elections to analyze, we 
only have data on a relatively small number of IRV elections. But 
this is the summary: in nonpartisan elections -- the vast bulk of IRV 
elections in present U.S. implementations -- there are no "comeback 
elections." None, so far. In a series of primary elections in Texas 
-- that were TTR -- there was a comeback election in 29% of them. The 
runner-up in the primary ended up beating the primary leader in the 
runoff. Now, like IRV, TTR encourages voters to vote for their 
favorite in the primary. My sense of real voters is this is what they 
do, and it is what they've been encouraged to do. In political 
primaries, as well, most voters do just that, particularly if it is 
runoff voting. They trust, generally, that one of the top two 
candidates in first preference will be a reasonable choice, and they 
can then make their choice in the runoff if there is majority failure 
in the primary. I think there is very little first-round strategic voting.

In these IRV elections, the first-round leader always goes on to win 
the runoff. In theory, IRV does better than Plurality. In practice, 
all IRV does is collect votes from a larger sample and, in 
nonpartisan elections, it turns out that vote transfers don't change 
the preference order, generally, for the remaining candidates. That 
was a surprising result, but, on reflection, it makes sense. We see 
something different with partisan elections, under certain 
conditions, particularly those where there is only one minor party, 
on one side, with significant vote strength. That party creates 
vote-splitting in the first round, hence vote transfers *may* reverse 
the first-round result. Have two minor parties, one on each side of 
the spectrum, this effect disappears, and that seems to be a more 
stable situation. It's known in Australia, where they have a fairly 
stable system, two big parties (on of them technically a coalition, 
but functioning non-competitively, they don't both run candidates in 
any individual race, I think), that the primary leader almost always 
goes on to win the election. Exceptions are likely to be elections 
where the primary results were close, close enough that statistical 
variations could shove the result in the other direction.

So why are there comeback elections in nonpartisan runoffs? (A 
political primary is "nonpartisan" in this sense, the candidates are 
not representatives of opposing parties) And not with IRV? It's 
pretty obvious: Two big effects, *both of which probably improve the results*.

What we don't know is which effect is more important.

(1) Voter turnout. Voting in a runoff is *usually* less convenient 
than voting in a primary. Even when the convenience is the same, 
because the candidate set has been reduced to the two most popular 
candidates, many voters may thing both of them adequate, they don't 
have a big preference between the two, so they accept the result. If 
they can't write in a vote (the most common situation, I think), and 
they dislike both candidates, likewise, no incentive to vote. Lower 
turnout does not indicate a problem. (But if we really think there is 
a problem, we'd set quorum rules. But that is another story.) Lower 
turnout means, pretty clearly, that the voters are more highly 
motivated voters, voters who have a significant preference between 
the two candidates on the ballot. Thus lower turnout shifts results 
toward Range results, by a factor that is sincere, it can't be faked. 
That's typical of a whole class of voting systems that use auctions 
or bidding on lotteries; these systems are design to extract sincere 
utilities from voters. Inconvenient elections do it for free, so to 
speak. In a manner that probably has a roughly equal impact on rich 
or poor. Certainly the unemployed may find it easier to vote!

(2) Increased voter knowledge regarding the remaining candidates. 
This is one of the classic arguments for runoff voting, and it is 
almost certainly true. So voters *change their minds*. They can't do 
that with IRV, nor with any deterministic, single-ballot system.

My conclusion is that TTR, even with its flaws, could be a better 
system than even Range as a deterministic method. But, of course, we 
can fix the flaws -- it's rather easy! -- and one of the fixes might 
be to use Range in the primary, ensuring that there is some way of 
extracting "approval" information from the primary -- such as 
defining any rating more than 50% as being acceptance of the result 
being rated, making that explicit in the ballot instructions. 
Further, this should be studied, the ballot might allow the 
expression of a "favorite" that is independent of the ratings, i.e., 
the voter might top rate two candidates, voting approval style, but 
then indicating a preference. This, indeed, I'd advise if the range 
resolution is low, like lower than 10. (Range 2 is actually a rather 
nice system.... except that it would be missing this favorite 
indication, it would be too *expensive* to lower the next-preferred 
candidate to midrange -- which could be defined as acceptance, it's a 
small difference from "more than half" to exactly half.

Immediately, the conversion of top-two runoff to IRV should stop. 
It's a step backwards. It's certainly possible to argue that IRV is 
better than raw Plurality, that's a different matter, particularly in 
partisan elections. But against top two runoff? No. Bad Idea. 
Reversing old reforms, actually. Existing IRV implementations should 
be studied, carefully. We now have some IRV experimentation, that's 
good, actually, even though the way it came about was typically 
deceptive. In Minneapolis, we may be seeing a new STV-PR usage, that, 
likewise, is good, probably a significant improvement. So what we 
need now are some Approval implementations, most urgently. Given that 
Approval is essentially free, is theoretically better than IRV in 
terms of overall voter satisfaction, it should be a no-brainer, 
highly unlikely to do any harm. Because of the very low expense and 
improvement in terms of fewer spoiled ballots, this really should 
have been the first reform, and, then, maybe we'd have gone to 
Bucklin, much cheaper than IRV, performs better. And has substantial 
history behind it in the U.S.

Or Bucklin might come first. It's more complicated than Approval, but 
still easy to vote. All the votes are counted (i.e., if any second 
preference votes are counted, they all are, not just the ones for the 
leaders.) Isn't vulnerable to Center Squeeze, yet does allow voters 
to go, first, for their favorite.

I've even gotten some IRV supporters to say, yes, Bucklin looks 
pretty interesting. It basically answers the objection that many have 
to Approval, an inability to express who they like most.

Make Bucklin require a majority to finish, hold a runoff among the 
top two if there is no majority, it would be  *clear* improvement 
over present top-two runoff systems, and would avoid, probably, my 
estimate, about one-third of runoffs. A little better than IRV, the 
only reason that IRV avoids runoffs is that the majority requirement 
was dumped with the implementations.

The tragedy in San Francisco is that the voters were told that, yes, 
a majority would still be required. The Dept. of Elections video 
continued to claim this. The distinction between "majority of votes," 
which, if you've followed Brown v. Smallwood, means "majority of 
valid ballots," i.e., "majority of voters," and "majority of votes 
for candidates not eliminated" has been lost. When I started editing 
the Wikipedia article to reflect the difference, there were screams 
of "Too much detail, you'll confuse readers!" Right. Some people like 
to have the public imagine it understands something when it does not. 
That what spin-doctors are very good at.

With logic like that, we might as well say, "The winner must gain all 
the votes." Simply continue the elimination process one step further. 
Any vote not containing a vote for the winner is set aside.

What the IRV process does is, as implemented, is look for a majority 
of votes for candidates not eliminated. If you only voted for 
candidates which were eliminated, as all but two will be at the end, 
Sure, this "simulates" a series of runoffs, but with declining 
participation, and voters who went to the polls, and cast their 
votes, for legitimate candidates, sincerely, are considered not to 
exist. They aren't part of the basis for a majority. No matter that 
they actually voted. Technically this is, indeed, a majority of 
*remaining* ballots, not exhausted, but using the word without the 
qualifying language is highly deceptive. A majority favored McCain in 
2008, but Obama won. Raw deal! Oh, forgot to say: A majority of 
Republicans favored McCain.

A majority of voters voted for So-and-So for San Francisco 
Supervisor, so he was elected. Wait a minute, over 60% of voters did 
*not* vote for him. Yes, but they don't count, they did not vote for 
one of the frontrunners, so their ballots were exhausted, as will 
almost certainly happen. You *must* vote for a frontrunner or you do 
not count? Didn't you know that?

Now, please, after reading this, try to say again, "IRV encourages 
sincere voting."

Sure it does. A little, but then, if you are fully sincere, it whops 
you upside the head. You don't count, unless, of course, you are one 
of the people who count, the people who support the winner and 
*maybe* the people who support the runner up, though all these votes 
could be eliminated and the result would not change. That problem, 
the problem that we have elections for people to represent the 
people, and there are *losers*, people unrepresented -- over 60% of 
those who voted in that district in San Francisco -- is the problem 
with single-winner.

STV-PR has the same problems as IRV, but the problems don't really 
appear in the first elections, until eliminations start. There are 
better methods than STV, as usually implemented, we are not 
restricted to a method invented, what, 150 years ago? But certainly 
STV is better than single-district, single-winner, and it gets better 
the more candidates are election from a district. I.e., large districts.

Asset can go much further, allowing effective districts to be created 
on-the-fly by the candidates, creating an assembly that is 
practically total in its representation, with every voter having a 
representative whom they chose, or who was chosen or approved by 
someone they chose.

If the quota is the Hare quota, not the Droop quota, and if electors 
-- those receiving votes in the election -- may vote on matters 
before the assembly, directly (but not introduce motions or speak by 
right to the assembly), then we have *total* representation unless 
the chosen representative defaults and drops out without assigning 
his or her vote to someone else. I'd say that is about as close to 
total representation as we might imagine. In practice, I'm fairly 
sure, it would be rare for direct votes from electors -- which 
devalue votes of those who were elected to seats by them -- to shift 
a result. The dregs, i.e., those who have no seat, because they can't 
come together with enough other electors to choose a common 
representative, wouldn't be large. And since seats are freely chosen, 
presumably by those who trust them, differences in vote would be the 
exception rather than the rule. It's the principle that everyone has 
a vote or is represented, though, that's important. That's part of my 
goal: for us to build a government that we think of as "us," because 
it very clearly is. We would be *connected*.



>BTW, most of the list will probably be aware that the second round of
>the Georgia (US) TTR senatorial election will be held next Tuesday.
>
>http://en.wikipedia.org/wiki/United_States_Senate_election_in_Georgia,_2008