[EM] IRV vs Condorcet

Abd ul-Rahman Lomax abd at lomaxdesign.com
Wed May 26 21:12:23 PDT 2010


At 10:03 PM 5/26/2010, robert bristow-johnson wrote:

>On May 26, 2010, at 8:19 PM, Dave Ketchum wrote:
>[about IRV]
>>     Backers make a big deal of "majority" - but it is of the final
>>stacks, not of all ballots.
>
>what it is, is *a* majority.  for a particular pair that is left
>standing after the other candidates are eliminated by the IRV STV
>rules (which is the essential problem with IRV).  assuming no ties,
>each pair of candidates drawn from the candidate pool has an intrinsic
>majority.  the question is: which majority is the salient majority?

Once upon a time, there would have been no question. "Majority" has a 
few meanings, but it never meant "majority of all those voting for 
the top two, excluding all other ballots cast in the same election." 
Robert's Rules calls it, in the counting rules, just "majority," and 
that allowed IRV enthusiasts to believe that they meant last-round 
majority, if they didn't read too carefully, and FairVote went on 
promoting this even after it was pointed out that Robert's Rules, in 
the instructions for the clerk, mentions that voters should be told 
that if they don't rank all the candidates, there might be a failure 
to get a majority, and the election would have to be repeated. It is 
totally explicit.

In San Francisco, the voter information pamphlet on the RCV question 
said that the "candidates would still be required to gain a majority 
of the votes." It didn't say "majority of the votes for the top two, 
left after eliminations." It said "majority of the votes," and unless 
someone read the question carefully, they could easily think that 
"majority of the votes" meant majority of *all* the votes. My guess 
is that the people on the ballot information committee thought that 
too. They had simply swallowed FairVote propaganda, which hasn't been 
really explicit about this majority thing, most of the time.

I've pointed out that this concept of "last round majority" could be 
used to claim that there is a very simple change to Plurality that 
allows it to always find a majority. This is far cheaper than IRV and 
produces the same results, almost all the time, in nonpartisan elections.

It's simply. Just use the STV elimination and count it as if it were 
IRV. Eliminate the lowest vote-getter in each round, and ballots 
which only have a vote for that candidate, until a candidate left has 
a majority of the remaining votes. If you want a more sophisticated 
version, allow the ballot to approve multiple candidates. Presto! A 
majority in every election.

Now, if that isn't a majority, why is the IRV majority a majority?

In fact, we can take the process one step further. Wouldn't it be 
desirable to have unanimity in every election? Very simple to do, eh?


>>     Suppose Tom, Dick, and Harry share all the top rank votes, and
>>Joe gets all the 2nd rank.  Then if raced in pairs Joe would get
>>twice the votes of each of them - but Joe is invisible in IRV.
>
>or, we could change Joe's name to "Andy" and Tom and Dick to "Bob" and
>"Kurt", leave Harry out of it, and this hypothetical becomes less
>hypothetical.

Cool. Leave Hairy out of it. Much easier.

David didn't exactly express this well. He means that Joe could be 
the unanimous choice of every voter in second rank, and lose, simply 
because the first rank votes of Joe were less than those of Tom and 
Dick. Those first rank votes could be almost equally divided, so we 
have an IRV winner based on one-third of the vote (suppose the Joe 
voters truncate), whereas Joe would beat that candidate two to one in 
a direct face-off. That's horrible performance. To be sure, that's 
extreme. The situation in Burlington wasn't that bad, just an 
ordinary IRV failure to respect a majority position, in favor of the 
Democrat, who would have beaten all the other candidates in pairwise 
races, and probably would have won under Bucklin, as well. Or 
Approval or Range, my guess. 




More information about the Election-Methods mailing list