[EM] FairVote on Robert's Rules of Order and IRV

[Abd ul-Rahman Lomax]

At 12:00 AM 12/20/2008, Dave Ketchum wrote:
>Authors of RR have their own primary goals and properly avoid the election
>methods wars that take place in EM, etc, - simply recommending that group's
>rules authors should be careful as to what methods they choose to define
>for their groups.

Robert's Rules are pretty clear: avoid making decisions, including 
elections, without a majority vote, and they don't fall into the trap 
of thinking that one gets a majority by excluding ballots without a 
vote for the top two.

When elections were by mail and when that was expensive, relatively, 
there were arguments for allowing plurality decisions, but Roberts' 
Rules never approved this, they only note that it requires a bylaw to 
allow it. The default rules would allow elections to take place at a 
special meeting, by majority vote of those attending, but not 
election by plurality.

The argument could be made that allowing election by mail can 
facilitate broader participation, but this then raises the plurality 
problem. Election at a meeting can take many ballots.

Hence the suggestion that using preferential voting would be better 
than election by plurality. But what they describe as "preferential 
voting," while the rules are single-transferable vote, do *not* elect 
by plurality, they merely make it easier to find a majority, and they 
suggest that voters be made aware that if they do not rank enough 
candidates, the election might fail to find a majority "and must be repeated."

FairVote has radically misrepresented this section of RRONR, and that 
misrepresentation has been taken up and repeated by election 
officials in places which have implemented IRV or RCV. The method 
described in RRONR is indeed "better than election by plurality," but 
what is being implemented is, in some of the applications, no better 
than plurality: it *is* plurality, almost always. That's with 
nonpartisan elections. There are subtle but crucial differences 
between what RRONR describes and what is being implemented: the most 
important is that election by plurality is allowed, and the dirty 
little secret is that IRV usually, with nonpartisan elections, where 
full ranking is not obligatory, does not find a majority if one did 
not exist in the first round; further, it only rarely -- no examples 
so far in the U.S. with nonpartisan elections! -- finds any winner 
other than the first round leader.

In other words, with all the jurisdictions that have implemented IRV, 
with nonpartisan elections, no results have been shifted from what 
Plurality would have obtained. But results almost certainly *have* 
shifted: most of these jurisdictions were ones that required a runoff 
election if a majority wasn't found, and runoff elections, depending 
on rules, do find a real majority, at least in some senses, and even 
when the method is open to write-in votes, majorities are normal.

IRV is replacing top-two runoff, not Plurality, usually, so the 
comparison with Plurality is a false one. And top-two runoff, while 
certainly not perfect, is different from IRV in a number of important 
ways. Regardless of theory, it seems that about one out of three TTR 
elections results in a "comeback" where the first round leader loses 
to the runner-up. Since IRV is not presenting us with these, in 
nonpartisan elections, we can be fairly sure that IRV is changing 
results from TTR (better) to Plurality (worse).

FairVote, in describing or giving examples of how IRV works, focuses 
on *partisan* elections, where vote transfers follow some relatively 
predictable pattern. Not as strong a pattern as they or voting 
systems theorists often predict, but still strong enough to shift 
results. So the Green candidate is eliminated and *some* of the votes 
go to the Democrat. Not all. Usually, it turns out, there are enough 
exhausted ballots that a majority still is not found. IRV is a form 
of election by plurality, merely a slightly more sophisticated one 
that can *sometimes* fix the spoiler effect.

And who benefits from that? Mostly the major parties, which is why 
IRV, where it is significantly used, is associated with strong 
two-party systems. What voting system is associated with multiparty systems?

Does FairVote want us to know? It's Top Two Runoff. Definitely, TTR 
can be improved; and the most obvious improvement that fits the most 
neatly into U.S. traditions is Bucklin voting, it was used here in a 
number of states, it was popular (with the public and with political 
scientists and with, apparently, the legal profession, see Brown v. 
Smallwood, where the Minnesota Supreme Court, outlawing Bucklin -- 
and, really, all forms of alternative voting -- was quite aware that 
it was disregarding public opinion and prevailing precedent and legal 
opinion. But they were the Supremes, you know, a majority can pretty 
much do what it wants; what's amazing is that decision still stands. 
Maybe this year or next....

But TTR would be improved by Bucklin voting if the majority 
requirement still stands. If there is no majority, then, presumably, 
the top two would face each other in a runoff, and this *might* also 
be done if there are two candidates gaining a majority, which is 
actually quite unlikely, and precedent would indicate that the 
candidate with the most votes would prevail. Another possible runoff 
would be between the Bucklin plurality winner and any candidate who 
beats that winner by pairwise analysis of the Bucklin ballots. This 
little trick would make Bucklin, overall, Condorcet compliant.

But underneath this there is one teeny problem that makes a great 
deal of voting systems theory moot; Lewis Carroll wrote about it over 
120 years ago, and we've mostly ignored it. That will be another post....