[Election-Methods] Bucklin

[Abd ul-Rahman Lomax]

As I've done some research on Bucklin -- not in libraries, 
unfortunately -- I've been coming to the conclusion that Bucklin 
should receive more attention than it has.

It is a very simple method; it can be summed by precincts. It can be 
counted with any election equipment that can handle multiwinner 
Plurality -- i.e., standard equipment -- like regular Approval, It's 
easy to understand and vote (more about that below). And it has 
reasonable behavior, quite similar to Approval; but it is Majority 
Criterion compliant.

The version of Bucklin that I'm familiar with is what was used in 
Duluth, because it is described in Brown v. Smallwood, the complete 
decision is available on the internet, and is worth reading.

In this version, first and second rank votes were limited to one 
candidate or the ballot was spoiled. The third rank -- there were no 
more ranks -- allowed voting for any number of candidates.

I would change that only in one way: I'd allow multiple votes in any 
rank, so a voter could vote exactly as in Approval, *or* could choose 
to exclusively rank.

Then, if the votes for first rank fail to find a majority winner, the 
second rank votes are added in. And if no majority winner is then 
found, the third rank votes are added in. I think that in these 
elections, plurality victory was allowed; it simply becomes less 
likely that there will be majority failure.

Contrary to FairVote propaganda about Bucklin -- most information on 
the net was helpfully supplied by FairVote -- Bucklin appears to have 
been very popular, and no examples are known where it produced a poor 
result. The case in Brown v. Smallwood was one where the result 
overturned by the court was clearly just, and the reversal -- a long 
time after the election -- was very poor public policy. Problem is 
that the voters counted on the system, thus very likely did not first 
rank with standard plurality strategy. It is highly likely that, had 
the election been Plurality, Smallwood would also have won. So, 
essentially, the court cheated the voters. For what?

That the election method was allegedly unconstitutional need not have 
been sufficient reason to reverse the election; the court could 
instead have required a new election. However, they awarded it to the 
first-preference winner. I smell a rat.

There was a request to reconsider, and the Court issued an additional 
ruling on that, simply reaffirming their original decision. But the 
petition is interesting. It was asserted that the majority -- strong 
majority -- of legal opinion at the time did not agree with the Court 
decision. There was also a dissenting opinion in the decision, 
pointing out that there had been no legal opinion that Bucklin -- it 
was actally called, I think, preference voting -- violated the 
one-voter, one-vote principle, and the FairVote claim that this was 
the basis of the decision is, shall we say, oversimplified. A number 
of attorneys in positions of public trust have issued opinions that 
Brown v. Smallwood applies to Instant Runoff Voting, and we will 
probably see, soon, court challenge of the Minnesota implementation of IRV.

My own opinion is that Brown v. Smallwould *does* apply to any system 
of alternative votes indicated on the same ballot, and there is 
plenty of evidence for that in the decision itself. The only legal 
opinion I've seen that confirms the FairVote position on this is from 
lawyers who are advocates for IRV. On the other hand, it is also my 
opinion that Brown v. Smallwood was a corrupt decision, and my 
amazement is that it was allowed to stand. Given that Bucklin was 
quite popular in Duluth, I conclude that, probably, a major party 
that did not want to see Bucklin spread, through it's control of the 
state legislature and the court, was able to shoot down Bucklin, and 
they had too much power at the state level for the local supporters 
of Bucklin to prevail.

Since Brown v. Smallwood was never confirmed anywhere else, as far as 
I've been able to find, the precedent from it applies only to 
Minnesota, and a new consideration in Minnesota might quite possibly 
overturn it. I'm a little worried that IRV advocates may be able to 
influence that decision to focus only on the specific form of 
"runoff" involved in Bucklin, rather than the more general 
alternative vote situation, and the alleged problem that later 
preferences expressed can hurt your first preference. That is the 
tack that I'd expect them to follow; so it might be important for 
Approval advocates to see that a friend of the court brief is filed 
on the more general aspects of the situation. There is no substantial 
legal opinion -- outside of interpretations of Brown v. Smallwooed -- 
that Approval violates one person, one vote, since, in the end, only 
one vote counts, the one cast for a winner, or, alternatively, *no* 
vote on that ballot counts, i.e., affects the result. The necessary 
argumentation was already presented in Brown v. Smallwood by the 
dissenting justice.

Now, what about this claim that later preferences expressed can hurt 
your first preference? It is, of course, true; that is, if the voter 
chooses to add a candidate to a lower rank, that vote could help that 
lower-ranked candidate to beat one's favorite. However, how realistic 
is this possibility, under what conditions would it actually occur?

First of all, FairVote claims that Bucklin was abandoned because, 
allegedly because of this harm, most voters did not add lower rank 
votes at all. This is like the IEEE situation with Approval. The IEEE 
implemented Approval, I've seen it believably alleged, when there was 
a possible dissident candidate who might win through vote-splitting. 
So the Board implemented the method to protect their desired winners. 
Then, when the danger passed, the continued existence of Approval as 
a method could have allowed dissident candidates to more easily 
garner votes, and maybe win. So it was removed, and the reason given 
was familiar: most voters weren't using the additional vote possibility.

But that is no reason at all to drop Approval. If *every* voter, in 
an election, does not use the additional votes, there has been no 
cost, nothing was wasted. With hand-counting, Approval *may* incur 
some additional counting cost, *but only to the extent that voters 
use it.* No use, no cost.

Approval, however, fixes the spoiler effect. The effect we are so 
unfortunately familiar with takes place when a candidate with no 
chance of winning the election flips the result from one party in our 
two-party system to the other. In such a context, if voting for 
multiple candidates is allowed (in a single-winner election), most 
voters, I would predict, vote only for their favorite. *By 
definition, most voters' favorite is one of the two frontrunners.* 
Now, when only two candidates are in range of winning, votes for 
other candidates are essentially moot for victory purposes. So for 
this alleged "harm" to take place, the voter must be considering 
voting for both frontrunners. Like voting for Bush and Gore in 2000.

This is, more or less, quite similar to staying home! And this 
alleged effect is the number one reason commonly given for why 
Approval is a poor system, and likewise Bucklin.

No, voters in Duluth loved Bucklin, unless those appealing the 
decision were essentially lying. I'd love to find some newspaper 
articles on it. Bucklin was working. No election method can magically 
create a majority  winner, unless voters are coerced into voting for 
a candidate that they might detest. Forcing voters to rank all 
candidates can be exactly this. "You *will* cast a vote for Adolf 
Hitler or Genghis Khan. But, hey, you are the voter, sovereign. You 
can choose." No, thank you.

I prefer what Robert's Rules prefers: if an election does not 
discover a majority who have essentially voted Yes to the result, it 
has failed. RR actually dislikes top-two runoff, it favors repeated 
balloting until a majority winner is found. Yes, that can be quite a 
nuisance, but, silly parliamentarians, they happen to think that 
majority rule is important, and that no decision is valid without the 
assent of a majority. Unless the bylaws -- *contrary to their 
recommendations* -- allow plurality elections.

This should be the interpretation of truncated ballots: "I would 
prefer that the election fail than that any of the remaining 
candidates be elected." Should the *voters* be able to assert that 
outcome? (It is a variation of None of the Above, and that variation 
actually exists in, I think, standard Robert's Rules elections, where 
a blank ballot submitted by a voter is considered, essentially, a No 
vote for all the candidates; it can prevent the plurality winner from 
prevailing.

And this, then, I recently realized, causes the common FairVote claim 
that voting for a second preference cannot harm your first preference 
to not be true. In Vermont, IRV legislation was proposed some time 
back. Vermont requires a majority of ballots cast, I'm not sure of 
the exact definition, it may exclude blank ballots. If no candidate 
gains a majority, the top three candidates are presented to the House 
for election by secret ballot. Not the top two. So this is closer to 
the Robert's Rules procedure, where there is no candidate elimination 
at all. The House stands in as surrogates for the voters. (Don't 
trust a representative, don't elect him or her!)

So, if I add a second preference to my IRV ballot, and my first 
preference is eliminated, my second preference vote could put that 
candidate over the top to gain a majority. Whereas if I had not, and 
no candidate gains a majority, my second preference -- who, with IRV, 
must be in third place for this to happen -- would still have a 
chance of being elected by the House. Likewise with Robert's Rules 
and further election process.

In the Vermont IRV legislation, a ballot instruction was included, 
and that instruction repeated the claim that a lower preference vote 
could not harm your first preference. That ballot instruction was not 
true. I don't think the IRV legislation came close to being passed; 
if it had, I wonder if the problem would have been caught.

In approximately 1996, FairVote invented the name "Instant Runoff 
Voting," and managed to make this the common term in the United 
States. They use the term to cover many different forms of 
Preferential Voting, with different rules and conditions. Thus, even 
if an implementation, such as in Hendersonville, North Carolina, 
isn't what would be called "IRV," they can, and do, claim the "victory."

But, of course, Bucklin is also worthy of the name "Instant Runoff 
Voting." It proceeds in rounds, just like IRV. If there is no 
majority winner in the first round, then there is a virtual runoff, 
an Approval election. The claim that Approval hasn't been used in the 
U.S. is bosh. Bucklin "runoffs" were Approval elections.

On the one hand, FairVote wants to make us think that Approval is an 
untried method, never mind many centuries of use (like 500 years?) in 
Venice, never mind elections of the Pope for centuries, and, of 
course, never mind Bucklin.

On the other hand, when it suits some of them, it is claimed that 
early U.S. presidential elections were a form of Approval. If so, 
it's nothing like any form of Approval I've seen. The essence of 
Approval is that one may vote for or against all candidates; in the 
simplest implementation, a blank is considered a No vote. In the 
early elections, in the electoral college, electors had two votes to 
cast, and there would be two winners. That's not Approval, that is 
multiwinner Plurality (actually, majority, a majority was required), 
with two votes for two winners. Routine, actually. FairVote has 
promoted the article written by Nagel on the so-called Burr Dilemma, 
which is an invented and imaginary dilemma that supposedly would face 
electors in that situation. In fact, both candidates won, with no 
sign of defection. The problem is that the electors voted as 
promised, in the party system that had developed, so both candidates 
were tied, both with a majority, throwing the election into the 
House, where the rules were different, in order to choose which one 
would be President. Not Approval at all, but Nagel simply states over 
and over that it is. Makes one wonder about "peer-reviewed" 
publications.... by political scientists. (The scientists aren't at 
fault, really, it's the *process*, a familiar theme of mine.)

(Nagel argues that the dilemma does exist in Approval elections that 
are pure approval, and Warren Smith seems to have bought this 
argument. But if one looks, as I did above, at the scenarios where 
such a supposed conflict would exist, it doesn't make sense. Smith 
also seems to think that Approval is vulnerable to the Burr dilemma, 
but Range is not. He seems to forget that strategic voters in Range 
will typically vote approval-style; Range gets its performance 
improvement over Approval from, actually, very few voters voting 
intermediate votes. In fact, from my own study, it appears that *one 
voter* may bring substantial improvement from a certain point of 
view. Thus if the Burr Dilemma actually applied in any significant 
way to Approval Voting, it would also apply to Range. But it does not.)

I find Bucklin interesting because it solves the spoiler effect 
through an Approval-type mechanism, but it begins, in the first 
round, as a standard majority-required election, as it was originally 
implemented, and, even if multiple votes are *allowed* in the first 
rank, voters don't have to use that option, allowing flexibility if 
it *does* happen that there are three front-runners.

Bucklin, because it satisfies the Majority Criterion, may be more 
palatable than pure Approval. It might even be enough for it to have 
only two ranks, instead of three. Bucklin ballots could also be 
analyzed pairwise, for other interesting possibilities in hybrid 
Approval/Ranked methods. But right now, Approval is probably the 
method to advocate for immediate implementation -- except that 
Bucklin, with its history, *might* be easier to get. And some would 
argue it is a better method.