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

[Abd ul-Rahman Lomax]

I'm quoting the current FairVote "introduction" 
to a substantial quote from Robert's Rules of 
Order, Newly Revised (RRONR), on preferential voting.

>Robert's Rules of Order (RRO), the well-known 
>guide to fair procedures, makes the point that 
>an election by a mere plurality may produce an 
>unrepresentative result. It recommends voting 
>methods that can determine a majority winner 
>when electing single-seat offices. At 
>conventions of private organizations, etc., 
>where the electors can cast repeated ballots, 
>RRO prefers a system that allows open ended 
>repeat balloting with no runoff eliminations to 
>finally elect a majority winner. Such a system 
>may be time consuming but can allow a compromise 
>candidate to emerge after a number of ballots. 
>However, in elections where open-ended re-voting 
>is not practical, such as in elections by mail 
>(or governmental elections), instant runoff 
>voting (called "preferential voting" in RRO) is 
>the recommended procedure. In the section 
>detailing the procedure for conducting an 
>instant runoff election RRO states that "It 
>makes possible a more representative result than 
>under a rule that a plurality shall elect..... 
>This type of preferential ballot is preferable to an election by plurality."
>
>The full text is below. (Again, note that the 
>term "preferential voting" is another one for 
>instant runoff voting). It is from:
>
>Robert's Rules of Order Newly Revised
>In Chapter XIII §45. 10th edition, 2000, pp. 411-414
>(Used with permission from The Robert's Rules 
>Association, <http://www.robertsrules.com/>www.robertsrules.com)

The text is close enough to the truth that it is 
plausibly deniable against a charge of pure 
deceiption, but it is also highly misleading. 
Basically, it's designed to leave an easy 
impression that Robert's Rules of Order 
"recommends" "IRV," even though the introduction 
doesn't actually say that (any more, it may have 
said it at one time.) And if "IRV" means what 
FairVote has been promoting in the U.S., 
specifically a runoff replacement -- implicit in 
the name that they invented, and which name was 
not used by RRONR -- then it's directly and 
demonstrably false. The difference is subtle 
enough, what might seem like a technical detail 
to some, but it's quite a good example of "the devil is in the details."

IRV, as used in the U.S., is a form of election 
by plurality. "Preferential voting," as described 
by RRONR, is a method which never, following 
their rules, elects by plurality, but it is more 
efficient at discovering a majority than pure 
plurality, so it will avoid some runoffs, hence 
it is "preferable to an election by plurality." 
However, note, "preferential voting" is *not* 
"instant runoff voting." FairVote wants the reader to conclude that. However:

>Preferential voting has many variations. One 
>method is described here by way of illustration.

Then they describe Single Transferable Vote, 
though they do not name it. Later, in reference 
to it, they call it "this form." However, if we 
look at the description, there is what many might 
think an inconsequential difference. I know I 
missed it at first. If you read, say, the rules 
for Ranked Choice Voting for San Francisco or for 
Australian "Optional Preferential Voting," you 
will find that the word "majority" is used to 
refer, in describing the process, to a "majority 
of ballots containing a vote for continuing 
candidates," i.e., those who have not been 
eliminated. The description of RRONR's example of 
preferential voting refers to "a majority of 
ballots." Exhausted ballots, those where a voter 
only voted for or only ranked eliminated 
candidates, continue to be part of the basis for a majority.

The process continues:

>until one pile contains more than half of the 
>ballots, the result being thereby determined.

In Australia, what they call "Preferential 
Voting" refers to "an absolute majority." That is 
not exactly the same as "a majority of ballots," 
for Australia excludes ballots with some invalid 
or invalidating vote, which allows PV to always 
find a majority, because, in Australia, full 
ranking is required or the ballot is invalid.

RRONR talks about the possible effect of incomplete ranking:

>Sometimes, for instance, voters decline to 
>indicate a second or other choice, mistakenly 
>believing that such a course increases the 
>chances of their first choice. In fact, it may 
>prevent any candidate from receiving a majority 
>and require the voting to be repeated. The 
>persons selected as tellers must perform their work with particular care.

This is a description of the Later No Harm 
Criterion. RRONR actually errs here, because that 
additional ranking may cause the election to find 
a majority and thus terminate, whereas 
withholding it could, as they describe, cause the 
election to fail, and then their favorite might 
win in subsequent process. There are some 
conditions, not uncommon, where this is a reasonable possibility.

The main point here, though, is that "this form 
of preferential voting" isn't "Instant Runoff 
Voting." It reduces, but does not eliminate, the 
possibility of needed some further process, such 
as a runoff, depending on what the bylaws 
specify. If the bylaws do not specify what 
happens, the default in Roberts Rules is that the 
election -- the whole process, including 
nominations -- is repeated. They do allow 
existing nominees to be included by default, 
unless they withdraw. No eliminations.

In real public runoff elections, the rules 
generally specify that the top two remain on the 
ballot, none of the others are placed there. Some 
rules, such as the default in California, allow 
write-ins in the runoff, and these write-ins 
sometimes win, they are not irrelevant.

Then, RRONR goes on to criticize "this form of 
preferential voting," i.e., the STV form.

>The system of preferential voting just described 
>should not be used in cases where it is possible 
>to follow the normal procedure of repeated 
>balloting until one candidate or proposition 
>attains a majority. Although this type of 
>preferential ballot is preferable to an election 
>by plurality, it affords less freedom of choice 
>than repeated balloting, because it denies 
>voters the opportunity of basing their second or 
>lesser choices on the results of earlier 
>ballots, and because the candidate or 
>proposition in last place is automatically 
>eliminated and may thus be prevented from becoming a compromise choice.

Note that the common form of Top Two Runoff is a 
compromise with what Robert's Rules of Order is 
prescribing. They would have a vote-for-one 
election process repeat indefinitely until one 
candidate has a majority. It's possible to have a 
public election system that is very close to 
this, but it would require a shift in our concept 
of what an election is. If the U.S. electoral 
college were elected with proportional 
representation, and if it were functioning as 
originally conceived, it could continue the 
election through an indefinite number of ballots. 
But, aside from this, it's been considered 
adequate to have two stages, when a majority is 
not found in the first stage, and the probability 
that the election will complete with a majority 
is greatly enhanced or even ensured, the latter 
being done by rigorous elimination of all candidates but the top two.

RRONR is not a book about voting systems, these 
systems are actually a detail for them; the rules 
*require* that all decisions, large or small, be 
made by majority vote on Yes/No questions. There 
are easy tweaks to the two-stage election process 
that make it far closer to the pure democracy 
that RRONR requires by default. But RRONR is a 
manual about existing practice, and only the IRV 
form of preferential voting is in current use for public elections.

However, the problem they describe, that it 
"denies voters the opportunity of basing their 
second or lower choices on the results of earlier 
ballots," is very real, and is a common argument 
for Top Two Runoff voting, which really does 
shift results from the first ballot based on the 
new campaign and new consideration by the voters. 
Further, the elimination of a "compromise 
candidate" is what is called Center Squeeze, and 
with nonpartisan elections, it is reasonably 
common; if it were not reasonably common, RRONR 
would not mention it. Other forms of preferential 
voting do not suffer from this problem.

So, does Robert's Rules of Order recommend IRV?

No.

What RRONR suggests for use under some conditions 
is different from "Instant Runoff Voting." If we 
used STV, as an example, for primary elections 
(the first stage in a Top Two Runoff practice), 
and we used it as a two-winner process -- which 
functions better than single-winner IRV -- and we 
considered the election done if any candidate 
gained an "absolute majority," and if, then, a 
runoff is needed, we allowed or continued to 
allow write-in votes, we would have a system far 
closer to what Robert's Rules suggests than what we have as IRV in the U.S.

U.S. IRV almost always elects the first-round 
leader. (The same is true for Optional 
Preferential Voting in Australia: FairVote points 
to Preferential Voting in Australia as an example 
of what IRV is, but, in fact, the U.S. is using a 
damaged form of Optional Preferential Voting. 
Australia always allows full ranking, the U.S. 
implementations have mostly been limited to three 
ranks, even in races with over twenty 
candidates.) In nonpartisan elections, so far, 
there is not one example of the runner-up going 
on to beat the first round leader after the vote 
transfers. Top Two Runoff changes the first round 
result in roughly one-third of runoff elections. 
Partisan elections are a little different, 
because it's much more likely, when there is only 
one minor party with a percentage of the vote 
that is greater than the difference between the 
top two, that the vote transfers go, for the most 
part, to only one of the top two candidates.

If we want to use preferential voting, there are 
better and simpler and cheaper forms to use, 
including Bucklin voting, which was known as 
"preferential voting," when it saw fairly wide 
use in the United States almost a century ago. 
Bucklin is not only cheaper, it is substantially 
more likely to find a majority, so if it were 
used as a primary, it would be less likely to 
need a runoff. I would estimate that roughly one 
out of three runoffs would be avoided.

There is an extremely cheap method, surprisingly 
good, considered by some voting system theorists 
to be one of the best methods, and by nearly all 
to be better than Plurality or even than IRV, and 
that is what I call Open Voting, known commonly 
as Approval Voting. I don't like the name 
Approval, even though the votes have that effect, 
because it implies that one should vote for all 
candidates that one "approves." Maybe, maybe not. 
Open Voting is simply Plurality except that all 
the votes are counted. If a voter votes for more 
than one candidate, say for two, both votes are 
counted. The winner is the candidate with the 
most votes. Because some people quickly and 
easily think that this violates one-person, 
one-vote, it does not. If, after the election is 
decided, one looks back and eliminates the moot 
extra vote, it would not change the election 
result. Open Voting can be seen as a kind of alternative vote.

Bucklin Voting is Open Voting, except that the 
voters rank the candidates, and counting proceeds 
in rounds like IRV, except that each new round 
adds in the old votes instead of substituting 
them. Thus there are no candidate eliminations, 
but the voter can still specify which candidate 
is the favorite. In the last round, Duluth 
Bucklin allowed the voter to mark as many 
candidates as desired, which allows the system to 
handle, intelligently, many more candidates than 
three-rank IRV. But, in fact, with all forms of 
preferential voting, many voters only vote for 
their favorite, and usually this is relatively 
harmless, and it is especially harmless if the 
methods are used for a first round. Bucklin does 
not suffer from Center Squeeze in real elections. 
When a majority has not been found, it has 
counted all the votes, it has neglected none. IRV 
elections do not report many of the votes which 
have been cast, either because the candidates 
were eliminated while the votes were still 
covered up by a candidate with higher rank, or 
because there were two candidates left, when 
counting stops, and some of these ballots 
remaining may contain a vote for the opposite 
candidate. This happens much more in nonpartisan 
elections than in elections where only a major party candidate might win.

I would recommend that anyone in a position to 
influence what voting systems a community decides 
to use, and who has realized that there is 
deceptive propaganda out there on voting systems, 
subscribe to the Election Methods mailing list, 
for a while, and ask any questions they may have, 
or propose their preliminary conclusions. They 
will get responses from many points of view, 
including from supporters of IRV, but also 
including serious experts about voting systems. 
The EM list includes discussion of many arcane 
details of some complex voting systems, and much 
of this may not be of interest to the general 
public, and sometimes debates rage on this list 
indefinitely, sometimes because the questions are 
difficult, and sometimes because there are 
partisans who will argue till the cows come home, 
but the truth will be there, and someone who 
asks, carefully reads the responses, and then 
asks again, until it's clear, will find that truth.