[EM] Two simple alternative voting methods that are fairer than IRV/STV and lack most IRV/STV flaws

[Abd ul-Rahman Lomax]

At 09:29 AM 1/15/2010, Chris Benham wrote:
><snip> With "repeated balloting" there are no eliminations?  As I 
>undersatnd it, in "Top Two Runoff" all but the top two first-round 
>vote getters are eliminated if no candidate gets more than half the 
>votes in the first round.

Yes. The standard voting method under Robert's Rules of Order (and I 
think under other standard parliamentary rules) is vote-for-one, with 
repeated election until a single candidate gains a majority. As many 
rounds are used as necessary, and there are no eliminations; in fact, 
"the election is repeated" which means the whole election process 
including nominations. But someone might move to mass-nominate 
existing candidates, though that's probably a bad idea. I'd vote 
against it! After all, that candidate set produced majority failure, 
and some of them might simply withdraw, and that's what really 
happens, it only gets tricky when three candidates insist on staying 
in, but these people may directly negotiate outside the election 
process itself. If the voters don't compromise, they don't get to go home.

(Seriously, they can go home, all it takes is a majority on a vote to 
adjourn. They decide, by majority vote, like nearly all decisions in 
parliamentary procedure, which is more important, hammering away and 
getting a result, or getting some rest and trying again the next 
session. They can also decide, by majority vote, on some caretaker 
office-holder pending final resolution. There is no way that, where 
it's practical, I'd trade deliberative procedure for a mere voting 
system that is satisfied with less than a true, voluntary majority to resolve.)

Yes, that's top two runoff. The major down side is, of course, center 
squeeze, but it turns out that in nonpartisan elections, center 
squeeze may be vanishingly rare. The famous examples were partisan elections.

I think this is a major discovery: in nonpartisan public elections, 
voter rankings tend to average out such that voters with a particular 
first rank preference have little overall bias, compared to the rest 
of the voters. The implication of this is that Plurality, as long as 
the number of candidates does not climb too high such that noise has 
a big effect, reproduces what IRV does, because the vote transfers 
don't affect the rank order, they merely raise the vote totals proportionally.

This doesn't hold with partisan elections, where a candidate's party 
affiliation is disclosed on the ballot, thus providing a handy guide 
to voters which allows them to fall into partisan grooves, which, 
then, cause the supporters of a particular candidate to be, indeed, 
biased with respect to the rest of the population. In the U.S., on 
average, a Green voter will prefer a Democrat over a Republican.

This is of great practical importance. In the U.S., top two runoff is 
*mostly* used with nonpartisan elections, or with party primaries, 
which amounts to the same thing. (Voters in party primaries get no 
guidance from party affiliation!) Because of the expense and 
inconvenience of runoff elections, FairVote was able to sell IRV as a 
method to "gain majority results without runoff elections!" But the 
results of IRV and Plurality, we are seeing, are the same.

(We can argue that with IRV, voters are more likely to vote their 
sincere first preference, and I'd agree, but in terms of results, 
this is highly unlikely to affect them, because if the first 
preference is at all perceived as possibly winning, in nonpartisan 
elections, voters will sincerely vote for them if they prefer them, 
and the only additional votes will be from other voters compromising, 
and in nonpartisan elections, these votes won't shift the results, so 
the determining factor is the initial first preference position of 
the electorate, unless it's close.)

With real runoffs, though, the electorate gets to scrutinize the top 
two. In San Francisco, there was a winner of an IRV supervisorial 
election, with almost 40% of the valid ballots after transfers, in a 
field of, as I recall, about 23 candidates. It turned out that he did 
not live in the district and was ineligible to serve. In a runoff, 
it's practically certain that this information would have come out. 
The same is a general truth: in a runoff, voters now have a reduced 
set to consider, and will make more intelligent choices, overall.

So FairVote has been damaging the most advanced voting system in 
general use in the United States, replacing it with a method that is 
allegedly cheaper, but with high implementation costs, and, in 
reality, the reduction in expense could have been accomplished by 
abolishing the majority requirement. Which the IRV implementation 
did, the measure explicitly took it out of the election code.

But my point is that there is a much better way, which is to use a 
better system for primary elections *without* abolishing the majority 
requirement. Because there are no eliminations and, if necessary, all 
the votes are counted, Bucklin is more efficient at finding 
majorities. It will still fail sometimes, but less often than IRV< 
which never counts some of the votes.

Remember, in nonpartisan elections, the supporters of a candidate 
will show the same distribution of votes as the rest of the 
electorate. So the supporters of the runner-up are quite likely to 
vote in similar proportions for the frontrunner as the rest of the 
electorate. Possibly, because of perceived frontrunner status of the 
favorite, they will be less likely to add a second preference vote 
for another frontrunner, but many will. And these may take the 
plurality leader over the majority line. Analysis of voting patterns 
in San Francisco showed that it would, but it's difficult to predict 
accurately how many voters will avoid adding such a lower preference 
vote, due to Later-No-Harm concerns.

The question will hinge, likely, on preference strength. Seriously 
partisans of the runner-up won't do it, but many or most voters 
aren't so partisan, they really just want a decent result. My guess 
is that, compared to top-two runoff, Bucklin for the primary will 
eliminate about half of the runoffs or more. Bucklin for a runoff 
(could be two-rank Bucklin, probably sufficient) will then make the 
runoff safe for write-in candidates, if allowed, and, as well, allow 
voters to, at least, express dissatisfaction with the top two by 
voting for a write-in. (The ballot should not be considered spoiled 
if they do that, the vote would simply not be counted for either 
ballot candidate in the first round, but only in the second, where 
the voter, if the voter cares enough, will add a vote for one of 
them. Or if not, the vote could cause majority failure in the runoff, 
which is valuable information even if not actually used to determine 
the winner.)