[EM] Does IRV elect "majority winners?"

[Abd ul-Rahman Lomax]

At 07:44 PM 1/5/2009, Kathy Dopp wrote:

>IRV/STV cannot claim majority winners, not only because ballots are
>exhausted and not considered in the final counting round, but also
>because not all voters' choices are even fairly and equally considered
>during the counting process - thus resulting in winners who are
>disliked by a majority of voters and overlooking candidates who are
>preferred by a majority of voters casting votes in the contest.

IRV cannot claim to *consistently* find a majority winner, though 
this is indeed claimed by advocates.

I disagree with Ms. Dopp, however, on one thing. The elimination of a 
Condorcet winner doesn't, itself, guarantee that IRV won't find a 
majority winner in spite of that, it depends.

"Majority" doesn't mean "best," necessarily. It simply means that, in 
an election, more than half of the voters chose to vote for a candidate.

What does this mean, "vote for a candidate?" IRV obscures it, and, we 
see, in Burlington, that many voters chose to vote for candidates 
whom they clearly didn't support, because they apparently had an idea 
that it was a "good thing" to rank all the candidates. By ranking all 
the candidates that were on the ballot, they were, effectively, 
voting for a candidate whom they most certainly didn't approve and 
would likely prefer to see a runoff than to elect this candidate.

Usually this is moot, because when they bottom-ranked a candidate 
whom they detested, they were only voting for this candidate over a 
write-in; so only with a massive write-in campaign could this become 
an issue, where the top two were a write-in and the candidate these 
voters ranked last.

Basically, IRV, when it fails to find the Condorcet winner, where, 
indeed, the ballots show an eliminated candidate as being preferred 
to the IRV winner, nevertheless can claim that the winner won by an 
absolute majority, if that happens. Usually, though, because of 
truncation, which will be fairly common for those who voted for the 
Condorcet winner (that is, they truncate with that candidate, who is, 
by the conditions of this problem, probably in second rank for many), 
there will be majority failure as well.

Thus if there is a real runoff *and if write-in votes are allowed in 
the runoff*, the voters *could* fix it. Robert's Rules of Order 
describes STV as a means of finding majorities and *not* as a runoff 
replacement, a true majority of ballots cast is still required. What 
they may not have realized is that IRV is particularly bad at this, 
in nonpartisan elections -- i.e., practically all elections in bodies 
advised by RRO. American Preferential Voting -- Bucklin -- does much 
better, because, if necessary to find a majority, it reveals all the 
votes and counts all of them. And it is *far* simpler to canvass. 
Just count all the votes in each rank. Add them as necessary to find 
a majority.

If they are all added and no majority is found, then there are two 
paths to take: terminate with a Plurality, which is honestly shown in 
the results, or hold a runoff. Because Bucklin was sold as finding 
majorities from a single ballot, the latter was never considered. 
This should sound familiar. Those who ignore history are condemned to 
repeat it.

>Abd ul is right that Top two runoff is a lot better system, and TTR is
>most likely less costly, definitely is easier to count and more
>auditable, is fairer, and both elections are monotonic too.

TTR has some obvious problems, which are easy to fix. Quite simply, 
use a better method in the primary, one which will efficiently find 
true majorities, thus reducing the need for runoffs.

Then, allow write-ins in the runoff, and use a better method for that 
as well. Bucklin, two-rank, would be *great* for this, because it 
would allow write-in voters to still participate among the two 
candidates on the ballot. Usually, in the majority of actual office 
elections like this, there would be a majority in the primary; it's 
hard to say what percentage, though, it depends on context and the 
number of candidates.

If they are partisan elections, a majority in the primary gets more likely.

Then, I'd predict, a majority would be found in the runoffs, almost 
always, yet it would no longer be a majority coerced by the method. 
Write-ins would have a chance, and the situation where something went 
wrong in the primary can be fixed *if the voters care enough about it.*

Basically, a proper goal would be to reduce runoffs, but not to 
eliminate them. It is also possible, with some kinds of plurality 
results, to predict with very high confidence that a runoff would 
produce the same result. However, a majority requirement is very 
clear and simple, and, in my view is quite adequate.

Where one could set the bar lower than that isn't clear to me. I'd 
suggest that with better methods in the primary and runoff, with a 
majority requirement in the primary (not in the runoff), problems 
would be minimized, and analysis of results over many elections would 
reveal where it might be safe to set a lower bar than majority in the 
primary, thus avoiding some more runoffs.

First priority: preserve Top Two Runoff. It's the most advanced 
voting system we have in common use in the U.S., particularly in 
nonpartisan elections, which is where it is mostly used.

There are some grounds for preferring IRV in partisan elections, 
where it does, sometimes, improve results, but in the nonpartisan 
elections where FairVote has promoted IRV most heavily, IRV is 
clearly inferior. They only manage to sell it in that context by 
making false promises, such as:

Find majorities. In fact, where doesn't find a majority in the first 
round, it doesn't usually find one at all. This isn't obvious to most 
people until they study the method and how it actually works.

Reduce negative campaigning. There is no evidence for this, actually. 
Think about it: most seriously negative campaigning takes place in 
partisan elections, where there are normally only two truly viable 
candidates. The supposed motive behind the reduction is an attempt to 
gain second-preference votes, but second-preference votes, among 
supporters of the top two, are moot. IRV never counts them! They only 
become relevant in a three-way race. From the San Francisco 
experience -- with nonpartisan elections -- there isn't any evidence 
of a reduction in negative campaigning. Some minor candidates, who 
don't usually waste a lot of effort excoriating each other anyway, 
since it's pretty much all moot, have cooperated, encouraging their 
supporters to vote together for them, holding common campaign events. 
While this may be a beneficial effect, to some degree, it's moot in 
terms of finding the winner, almost always. And it doesn't reduce 
negative campaigning.

Lower costs. As Ms. Dopp notes, this may be directly false. It's a 
complex issue: voting equipment and software can be expensive, and 
standard equipment and software, that jurisdictions may already have, 
often can't handle IRV. Hand auditing can become a nightmare with 
IRV. There isn't any actual analysis from experience showing reduced 
costs, only some projections, rather optimistic ones, that it might.

IRV is being promoted as a cost reducer in L.A., when there are much 
simpler methods, including better timing of elections so that the 
primary is held with the general election, that would reduce costs 
far more. Further, what should really be on the table as an 
alternative, American Preferential Voting, as it was called, or the 
Grand Junction method, or Bucklin, provides most of the benefits of 
IRV without the complicated counting and center squeeze problem of 
IRV. By 1918 or so, 55 cities in the U.S. were using it. I'd still 
like to know what happened to it: there were many articles published 
and available on-line about the *implementations* of APV, but it's 
very difficult to find anything about its disappearance.

And there are other arguments presented which are similarly misleading.