[EM] Is there a PR voting method obeying "participation" criterion?

[Abd ul-Rahman Lomax]

At 11:21 AM 11/18/2009, Terry Bouricius wrote:
>Abd wrote
><snip>
>"Under Robert's Rules, if a voter writes something on a ballot, the voter
>has voted, and the vote is counted in the basis for majority, "
><snip>
>
>This is not necessarily correct.

Bouricius raises a doubt, but does not impeach the general assertion. 
First of all, the context is single-winner elections. Special rules 
exist for multiwinner elections, where a voter has more than one 
vote. But even there, Bouricius is incorrect. We'll see.

He is arguing because he knows that the conclusion one might make 
from is are that IRV, as being proposed in the U.S., is a poor 
method, that the "preferential voting" that Robert's Rules of Order 
describes is different in this critical aspect from IRV, whereas the 
top-two runoff method which is aggressively being replaced by IRV 
actually does satisfy the intention of Robert's Rules more closely. 
And FairVote has extensively used the supposed recommendation of 
RRONR to promote IRV, and lots of election officials have fallen for it.

>  Abd is probably relying on the statements
>on page 402-3 of RRONR 10th edition, that even illegal votes cast by legal
>voters are included in the basis and that a ballot that registers "any
>evidence of having some opinion" should be included.

Indeed. Pretty clear, eh?

>However, a voter who casts a ballot may "partially abstain" by marking
>fewer candidates than allowed (see "Right of Abstention" page 394).

Yes. So? Bouricius, as has been common, wants us to jump to a 
conclusion. Amazing, actually. Does he really not understand what he's doing?

>Abstaining (as with a blank ballot) removes the ballot from the basis of a
>majority calculation (see "Majority Vote -  the Basic Requirement" page
>387).

Full abstention has that effect. Partial abstention in the case 
mentioned does not. How is it that Bouricius covers this up, implying 
otherwise? His reference is to the "right of abstention." It says 
nothing about the basis for a majority. The section on voting, which 
does cover this, Bouricius doesn't mention.

P. 427: "In an election of members of a board or committee in which 
votes are cast in one segment of the ballot for multiple positions on 
the board or committee, every ballot with a vote for one or more 
candidates is counted as one vote cast, and a candidate must receive 
a majority of the total of such votes to be elected."

This is no different than single-winner. Every elected member must 
receive a vote on a majority of non-blank ballots cast. (One can read 
this section differently, but "a vote for one or more candidates" is 
clearly intended to be the same as the previous idea that a ballot 
contains a possible vote, even if unintelligible.") In any case, if a 
voter casts one vote, and not the multiple number to which the voter 
is entitled, the voter still is part of the basis for a majority for 
*all* the elections. The voter has not, in this sense, abstained from 
those elections. An abstention would take place in the second form 
for multimember elections given, where seats are elected one at a 
time. In that case, the "majority" is a majority of all those who 
vote on that election, and the "right of abstention" refers to this.

Balloting is repeated if necessary to elect remaining members, and 
this means entirely new elections, not restricted-set elections. So 
if four were elected, our of six total, then voters would have two 
votes allowable on the next ballot.

The principle behind Robert's Rules is clear: no action is taken 
without the explicit approval of a majority. I'm not aware of any 
exceptions, and the preferential voting form RR describes is not an 
exception, that is where it differs from "IRV." RRONR does indeed 
mention election by plurality as a possibility, but also clearly 
considers this an action to be taken only when alternatives are 
impossible. It requires an explicit bylaw allowing election by 
plurality. What the editors of RRONR probably didn't understand was 
that preferential voting, of the sequential elimination kind (i.e, 
the IRV counting method) almost never finds, in nonpartisan 
elections, a majority where it isn't found in the first round, unless 
members actually do rate all candidates, or come sufficiently close 
to this. I don't think they had extensive experience with 
preferential voting. They describe preferential voting as a means of 
finding a majority, a true majority, without repeated balloting, 
which would happen if voters basically vote for all but one candidate 
(in sequence of preference). This may work in small societies where 
all candidates are well-known, but not in situations where voters may 
only be familiar with one or two candidates.

This continued insistence that IRV is compatible with RRONR is an 
example of political polemic that doesn't care if it deceives. That's 
what's wrong with American politics, much more than the voting system!

>  Thus in an IRV election it is arguable either way as to whether a
>ballot that abstains as to any preference between two finalists (registers
>no opinion on this particular question) should be included in the basis or
>not. The actual practice of organizations using IRV (preferential voting)
>on which RRONR is based, indicates rather convincingly that exhausted
>ballots are not used in the basis for calculating a winning threshold.

Bouricius may indeed argue that this or that practice is best, but 
what he's trying to do here is make it seem like RRONR would approve 
of this. It's quite clear. If a vote isn't found on a majority of 
ballots for a candidate, that candidate cannot be elected. Period. 
They are explicit.

Yes, that's the practice in some organizations. And many 
organizations use plurality, and this is merely a more sophisticated 
form of plurality election. Both require bylaw changes, they can't be 
used under the default rules. My guess is, as well, that some of 
these organizations quite simply missed the point, and implemented 
"last round majority" without realizing the implications. What is 
quite likely is that, long-term, simply allowing election by 
plurality would, in most elections, produce the same result with much 
less fuss.

But there is a much better option: a different form of preferential 
voting, Bucklin. Bucklin is more likely to find a majority, and if it 
does, that majority does satisfy the intention of Robert's Rules. The 
reality is that often even Bucklin won't find a majority, that was 
the case in many classical Bucklin elections. But when it does find a 
majority, which is more likely than with IRV, because IRV conceals 
many of the votes underneath non-eliminated candidates, the majority 
is real, a majority of voters explicitly decided to support the winner.

IRV has been implemented recently in a handful of jurisdictions. 
American Preferential Voting, as it was called, was implemented in 
over 90 jurisdictions in the U.S., at its peak. What happened to it? 
*I don't know." There is a great research project for some 
enterprising student with access to the necessary libraries. There is 
a huge amount of material on Bucklin leading up to the peak, 
available on-line. And I've been unable to find much about what 
happened next. It's as if it just disappeared with nobody noticing 
it. (IRV was also used a bit, and similarly disappeared, and possibly 
for the same reasons, one of which would be that both methods were 
oversold as finding majorities, when no voting method that does not 
coerce voters is guaranteed to do it. Bucklin simply does a bit 
better job, finding perhaps as much as half of the majorities that 
IRV misses. Both methods, in the end, saw most voters voting for one, 
not adding additional preferences.

In nonpartisan elections. It's in partisan elections that we see the 
possible utility of IRV, but only when there are two strong parties 
and a weak party that might spoil the election. Bucklin continues to 
function even beyond the point of three strong candidates, IRV breaks 
down, badly.

>Abd and I have been around and around on this in the past, and I have no
>desire to revisit the topic, but I just wanted to indicate that this is
>not an open and shut case as Abd suggests.

It's open and shut, under Robert's Rules, elections require a 
majority of those considered as "voting" to vote for the winner, and 
you have voted if you have cast one vote, even an unclear one; but 
*certainly* if you voted for one of the candidates on the ballot, 
it's irrelevant if that candidate is "eliminated." The vote is not 
eliminated, it stands, and Bouricius showed nothing different. Clever 
try, though!