[EM] Does IRV elect "majority winners?"

Abd ul-Rahman Lomax abd at lomaxdesign.com
Fri Jan 2 12:23:49 PST 2009

At 06:34 AM 1/2/2009, James Gilmour wrote:
>Dave Ketchum  > Sent: Friday, January 02, 2009 6:07 AM
> > Terry and Abd look set to duel forever.
> >
> > Conduct of elections is a serious topic, but both of them
> > offer too many words without usefully covering the topic.
>So let's try a small number of numbers.
>At a meeting we need to elect one office-bearer (single-office, 
>single-winner).  There are four candidates and we decide to use the
>exhaustive ballot (bottom elimination, one at a time) with the 
>requirement that to win, a candidate must obtain a majority of the
>First round votes:  A 40;   B  25;  C 20;  D 15.
>No candidate has a majority, so we eliminate D.
>Second round votes: A 47;  B 25;  C 20.
>It seems that some of those present who voted for D in the first 
>round did not want to vote in the second round  -  but that is
>their privilege.
>QUESTION: did candidate A win at the second round with 'a majority 
>of the votes'?
>James Gilmour

How many people voted in the election? Looks to me like 100. Could be 
more, actually; Robert's Rules considers all non-blank ballots that 
might possibly intend a vote, including overvotes. But let's stick with 100.

How many people voted for A? We don't know, actually! IRV doesn't 
count all the votes. However, what the method has found is 47. We 
know that 47 voters voted for A.

Are the ballots with a single vote for D on them "votes"? Surely 
those voters think they voted. Their ballots were recognized as legal.

The FairVote propaganda sometimes talks about "majority" without any 
qualification at all as to what it refers to; they are depending on 
voters imagining they know what it means, they know that this 
imagination will lead them to support IRV. Sometimes, however, we 
see, "majority of the votes." Or, in what is even more of a stretch, 
"the winner will still be required to win a majority of the votes." A 
requirement implies a standard that can fail. The IRV method can't 
fail to find a "last round majority," it's simple math -- if we except ties.

But in Santa Clara, the arguments went further. "Majority of the 
ballots" was used. Once again, one could weasel out of the claim of 
deception. "Why, we just meant, of course, "majority of the ballots 
containing a vote for a continuing candidate."

But any reasonable person, not knowing the details of IRV, would 
interpret the words to be a general majority, a majority of all the votes cast.

What was found was a majority of unexhausted ballots found to contain 
a vote for the IRV winner. Not a "majority of ballots," which implies 
the general usage.

Further, these arguments are being made in a context where "majority" 
has a very clear meaning, IRV is replacing, usually, top two runoff. 
The primary *requires* a majority, a true majority, in order to 
complete. When you tell these people that they can obtain a majority 
without needing a runoff, they will very naturally assume that you 
are talking about the *same thing.* The voters go to the polls and 
cast their votes. Setting aside informal ballots, if more than half 
of these voters support the winner, a majority has been found. The 
details of the voting system are actually moot. Did a majority of the 
voters who voted support the winner -- regardless of preference order?

A true majority is considered very desirable. IRV and Bucklin were 
apparently replaced by top two runoff, at least in some places, and 
the probable reason is that a majority was desired, and it was 
realized that these methods don't accomplish that, unless you coerce 
voters, as was tried in Oklahoma. (As is done in Australia.)

However, a more sensible approach would have been to use preferential 
vote in the primary, thus avoiding *some* runoffs! I would argue that 
Bucklin is better, because it doesn't suffer so badly from Center 
Squeeze, and it probably provides sufficient LNH protection that 
voters won't be significantly more reluctant to add additional 
preferences than they are known to be with IRV. We saw very 
significant usage of additional rankings in the municipal elections 
where I've been able to find results.

As I've noted, those who support a frontrunner don't have much 
incentive, with either method, to add ranks. With IRV, we don't know 
from the standard reports, how much truncation is present among those 
who vote for the top two.

Further, it seems to make a huge difference if the elections are 
partisan or nonpartisan. In nonpartisan elections -- which is most of 
the IRV implementations so far in the U.S. -- IRV functions almost 
exactly like plurality, the first round winner goes on -- every 
example so far at least before Nov 2008, which I haven't examined -- 
to win the election. In most elections, a majority is found in the 
first round. Same as Plurality!

So, looking just at the runoffs, roughly nine of them, in no case was 
there a "comeback election." In *real* runoffs, it seems to happen 
about a third of the time, that the runner-up goes on to win the election.

Almost certainly, the runoffs were a truer indication of the 
electorate's preferences, expecially if we consider preference 
strength, than the IRV winner would be. I have some ideas as to why 
IRV does simulate Plurality, it has to do with what motivates people 
to vote as they do. But that vote transfers strongly tend to simply 
maintain the lead of the first-round leader is known in Australia, 
and those are partisan elections, where the comeback effect would be stronger.

More information about the Election-Methods mailing list