This is an examination and comment on the document filed in the case of
Minnesota Voters Alliance, et al, vs. The City of Minneapolis
being the
Plaintiffs Memorandum of Law in Support of their
Motion for Summary Judgment and Declaratory Judgment
At the outset it must be stated that Minnesota
precedent is actually quite clear. Brown v.
Smallwood established that any form of
alternative vote, where, in a single election (as
distinct from, say, a primary election followed
by a runoff or general election), voters cast
more than one vote, making a choice of more than
one candidate, where it is one candidate to be
elected, regardless of details as to how the
votes are counted, is contrary to the State
Constitution (as interpreted by the State Supreme
Court). This decision has never been overturned.
While Brown v. Smallwood was, and is,
questionable, it is precedent in Minnesota, and
it is perfectly proper for the plaintiffs to
request summary judgment on this issue. Given the
clarity of Brown v. Smallwood on the topic, I
would expect it to be quite possible that a lower
court would issue summary judgment in favor of
the plaintiffs; which would then, presumably, be
appealed (eventually) to the Minnesota Supreme Court.
The reasoning in Brown v. Smallwood applied
equally to Bucklin voting (then called
preferential voting, a name also in common use
for Instant Runoff Voting) and to IRV. There is
one statement in Brown v. Smallwood where mention
is made of what is now called "Later no harm," an
election method characteristic that IRV satisfies
and Bucklin voting does not. However, other
arguments presented, and the response of the
court to an appeal for reconsideration by the
City of Duluth, make it clear that the court was
concerned with all forms of alternative vote,
anything other than a simple single choice by the voter.
My purpose here is to examine the arguments in the Memorandum.
The plaintiffs begin with a summary of the issues presented.
>The Minnesota and United States Constitutions
>establishes [sic] a person's right to vote and
>to associate one vote represents the intent to
>elect one specific person to office. A city's
>general election methodology allows ranking of
>candidates, vote transfers, and fractionalizing
>votes to obtain a preferential majority to
>declare a candidate elected to office. Do
>transfers and fractionalizing of votes violate a
>person's constitutional right to vote and to
>associate with one candidate for an elected office?
The first sentence introduces this "right to
associate," though that right, as normally
interpreted, has nothing to do with voting. If we
were dealing with the "right to associate," that
right presumably allows us to associate with more
than one person. The court in Brown v. Smallwood
did not mention, as far as I've noticed, a "right
to associate," this is a spurious invention of the plaintiff.
Brown v. Smallwood did not address
fractionalizing of votes, which is a technique
used in Single Transferable Vote to make vote
transfers more equitable, for the purpose of
creating proportional representation. Extending
Brown v. Smallwood to multiwinner elections is a
more complex issue than its simple application to
single-winner elections. Unstated in the
plaintiff's memorandum is the fact that, when
votes are "fractionalized," the result is that no
more than one full vote of the voter has been
used to elect any winner. However, it is correct
that this process results in the voter not
knowing what single candidate was elected with
the assistance their vote, because there may be
more than one. It is possible, though, to remedy
this with a procedural detail and an allowance of
some degree of approximation. I.e., each voter
would know whom, specifically, their vote
elected, because transfers would be assigned and
referred to voting precincts. All this complexity
I will set aside as something quite unclear under
Minnesota law, being possibly imagined as
conflicting with the spirit of Brown v.
Smallwood, but not explicitly considered by that
decision, which dealt only with single-winner
methods. (Though they mention cumulative voting
in the decision, they were making a different point.)
The real issue here is vote transfers, and the
possibility of contingent votes cast by voters.
Brown v. Smallwood was quite explicit on this. In
response to a petition from the City of Duluth
for rehearing, and on review, they held that
>The decision is sound; and we do right in
>upholding the right of the citizen to cast a
>vote for the candidate of his choice unimpaired
>by second or additional choice votes cast by other voters.
FairVote, in its promotion of IRV in Minnesota,
has claimed that Brown v. Smallwood was concerned
about a different issue, based on this comment made in the original decision:
The preferential system directly diminishes the
right of an elector to give an effective vote for
the candidate of his choice. If he votes for him
once, his power to help him is exhausted. If he
votes for other candidates, he may harm his choice, but cannot help him.
FairVote claims that this is a reference to the
fact that, in Bucklin voting, if you add a second
preference vote for a different candidate, that
vote might result in the different candidate
being elected, where, otherwise, it would have
been a tie, or it might, the same, create a tie,
thus harming the favorite candidate. This is
theoretically possible with Bucklin voting, but
Single Transferable Vote avoids this by only
considering lower preference votes once
more-preferred candidates have been eliminated
from consideration. Thus, it might seem, the
lower-preference votes, if any, cannot harm the
more preferred candidate. Had this been the
primary concern of the court, it would have been
a telling argument. But it was not. The court
went on to say, in the same paragraph:
>Another elector may vote for three candidates
>opposed to him. The mathematical possibilities
>of the application of the system are infinite.
This, then, read together with other language in
the decision, and especially the reasoning
expressed in the court's response to the request
for rehearing, shows that it is the multiple
votes that were of concern to the court.
There was a cogent dissent expressed, by J.
Hallam, in Brown v. Smallwood, and it seems clear
from the rehearing response that there was
substantial disagreement with this decision among
legal experts of the time. The court goes so far
as to note that the petitions for rehearing
claimed that the court went wrong on a plain
proposition involving no difficulty, or, to put
it in the language of one of the petitions, "If
one will put the proposition up to good lawyers
who have examined into the question, five out of
six will say that the statute does not violate
the constitution." Clearly, the decision was
unpopular, the court effectively acknowledges (and explicitly disregards) that.
Even though the plaintiff in the present case
argues on the basis of the U.S. Constitution,
preferential voting, per se, has been found
constitutional by nearly all courts. (One variant
on Bucklin, in Oklahoma, was found
unconstitutional in that state on the basis of
its use of fractional votes. However, this was
quite different from the fractional votes in STV,
which actually add to the effective voting power
of the voter whose vote is divided, which the
effect in Oklahoma was to reduce voting power.
(Perhaps this should be explained. In Duluth
Bucklin, first preference votes were tallied, and
if any candidate gained a majority, that
candidate was elected. Otherwise, second
preference votes were counted and added to the
first preference totals. Note that this only
allows one vote per candidate to be cast. Again,
if a majority of ballots (not votes!) were found
to contain a vote for a candidate, the candidate
was elected. And so on for third preference.
(With first and second preference, voters were
only allowed to vote for one candidate, but with
third preference, they could vote for as many as
they wished; i.e., they could effectively vote
against a candidate by voting for everyone else.
The effect of that, though, would be to abstain
from all other pairwise elections.) In Oklahoma,
perhaps assuming that lower preference
represented some lower approval of a candidate,
fractional values were assigned to the votes.
This has the effect, though, of penalizing the
voter for voting first preference for a candidate not likely to win.)
(Sometimes it is asserted that Bucklin voting,
like Approval, allows the voter more than one
vote. However, just as with Instant Runoff
Voting, only one vote, in the end, is effective,
at most. All other votes are contingent, and
moot, they could be struck from the ballots with
no change to the result. The Minnesota court, in
its reasoning does consider the difference
between a single voter and single vote,
represented by a single ballot, and the number of
marks on the ballot, yet, in one of the more
frustrating aspects of the decision, proceeds to
contradict its own reasoning by being concerned
that the number of marks on the Bucklin ballots
exceeded the number of voters. What matters, one
would have thought, is the number of voters
expressing support for a candidate for election;
when this passes a majority, majority rule has
been satisfied, and all other marks, used in the
process of finding that majority, are moot.
Instant Runoff Voting, though, differs from
Bucklin in a very important way: when Bucklin
fails to find a majority, it has considered every
vote cast, not just some of the votes. IRV will
terminate the process and declare a winner even
though a majority has not been found and not all
the votes have been examined, if the votes
remaining from unexhausted ballots have reached a
majority, not of all valid ballots cast, but of
only that reduced set of ballots. Because Bucklin
does not eliminate candidates, until the very
end, it is more likely to find a compromise
winner, one more broadly acceptable, instead of
merely the one with the most first-preference
votes, like Plurality, or what amounts to almost
the same, like IRV. Nevertheless, both methods,
in most election situations, will perform about
the same. IRV starts to have real trouble when
there are three viable candidates; but that's a
rare scenario. Bucklin probably continues to
perform adequately. What is truly relevant here is this:
Bucklin, like IRV, in the end has discarded,
effectively removed from consideration, all votes
not cast for the winner. This is absolutely true
if a true majority is required to win. (A
majority voting for a candidate is a majority
even if all other ballots, not containing a vote
for that candidate, were blank or contained
irrelevant votes, but still were counted as part
of the basis for majority.) The argument that
Bucklin violates one-person, one-vote, is based
on a shallow analysis concerned with the number
of marks, not the number of voters supporting the
winner (vs. those supporting other candidates).
The Minnesota court made the argument, actually,
then missed the implications and failed to apply it.
The dissenting opinion in Brown v. Smallwood
should be carefully read. It covers nearly every important argument.
To summarize my response to the plaintiff's
summary of the issue. Yes, if Brown v. Smallwood
continues to be upheld, vote transfers violate
the Minnesota State Constitution. However, Brown
v. Smallwood was defectively reasoned by that
court, and should be reconsidered, not merely
reinterpreted to allow Instant Runoff Voting,
because of a peculiarity of that method,
considered a defect by some experts, while
continuing to prohibit better methods, such as
the system of voting that was being used in
Duluth, or other methods under consideration by
voting reformers or in use in various contexts.
Returning to the issues raised by the plaintiffs:
>II. Under an election system, a formulation to
>redistribute surplus votes or second-ranked
>votes can divide a single vote into fractions
>among other candidates. Also a ballot is
>exhausted uncounted when an elector does not
>choose the next-ranked candidate although [the
>elector] may have ranked subsequent candidates
>on the same ballot. Is the equal protection
>clause violated when one vote is fractioned, and
>ballots not counted, resulting in a vote's
>dilution because of that election system?
Again, there are two issues here. The first is
the use of fractional votes in surplus transfers.
The second is the issue of a procedural detail,
not relevant under some election rules, the
handling of empty ranks when there are votes
present at a lower rank. The second issue is thus
a detail of less importance. If it's a problem,
it could be fixed. Treat blank votes as if they
were for an eliminated candidate. I've always
wondered why this wasn't being done in San Francisco.
So the issue of substance in this second issue is
the use of fractional votes. It's not clear that
the plaintiff understands this. Such fractional
votes never represent a dilution of the voter's
power, they, rather, represent a restoration of
it. Fractional vote transfers were used to
rectify an obvious inequity in proportional representation systems.
Consider a situation where three candidates are
to be elected to office. One candidate is very
popular, and let's say that two-thirds of voters
have voted for this candidate. However, the quota
for election, called the Droop quota, is not more
than one vote greater than one-fourth of the
number of valid ballots containing votes. (Why
isn't it one-third? Well, it's complicated; and
it is debated, but if we want to elect the
most-preferred candidates, if all elected
candidates are preferred by the Droop quota or
more, then there are insufficient votes remaining
to elect any more candidates. If the quota for
election is one-third, then any exhausted ballots
will result in insufficient votes to elect the
third candidate. Systems that do not result in
exhausted ballots can use the Hare quota,
assuming proper details.) What do we do with all
those "surplus votes"? Do we neglect them, thus
allowing the remaining one-third of the voters to
elect the other two representatives? This would
give two-thirds of votes in a resulting assembly
to candidates preferred by one-third of the
voters! Since we have only been looking at the
first choices of those who voted for the first
candidate elected, the most popular, clearly we
should, in order not to dilute those votes, look at the second preferences.
There are two basic ways that it is done.
Sometimes, in the counting, when a quota has been
found, the ballots used to find that quota are
then set aside and counted no further. Then the
candidate is considered as if eliminated, i.e.,
the next lower preference is counted. In each
case, one ballot is increasing the total for one
candidate by one vote. However, there is a
problem with this: who your vote counts for
depends on what sequence in which your vote is
counted. This is considered a problem (how real
the problem is may depend on unpredictable
details of the exact election environment), so
the division of votes was invented. If, say, the
quota is one-fourth of the votes, and one-half of
the voters have voted for a candidate, then all
of the second preference votes are counted, but
they are "fractioned," in this case multiplied by
one-half. You got your favorite elected, and
one-half of your vote was used for that, leaving
one-half to be assigned to your second
preference, etc. This keeps, within roundoff
error, one full vote active until it has either
been distributed to create winners or it has been exhausted.
Is this complicated? Yes. Is it fair? Well, up to
the election of the last candidate, yes, it is
clearly fair. With the last candidate, the
election effectively becomes the same as an
instant runoff voting election, with the problems
associated with that. In summary, fractioning the
votes, as is done with multiwinner STV, does not
dilute them in the sense of weakening them;
rather, it maintains their potency. Votes are
only fractioned after having been partly used to elect a winner.
However, if Brown v. Smallwood is allowed to
stand, the court would be consistent if it were
suspicious of any form of Single Transferable
Vote, no matter how fair it is. The court in
Brown v. Smallwood explicitly rejected arguments
based on election method performance or value:
their answer to these was, essentially, "Fine!. Change the constitution!"
The third issue raised by the plaintiffs is with
regard to the right of municipalities to create
"election systems affecting general elections."
There was a similar issue, I think, with Brown v.
Smallwood, but the present issue would have to do
with details regarding the 1983 statute or other
relevant statutes. It's not an issue that I feel
a need to address. My understanding has been,
though, that Minneapolis was within its rights to
implement a voting system if that system did not violate the constitution.
Reading further in the Memorandum, I find plenty
of irrelevant argument. The core, though, relies
entirely on Brown v. Smallwood or similar
reasoning, as it should. (Only Brown v. Smallwood
makes this issue clear enough, on the face, to
justify a motion for summary judgment.)
However, instead of simply quoting Brown v.
Smallwood and showing its application to Instant
Runoff Voting, they dilute their argument by
rearguing Brown v. Smallwood, which might be an
unskillful move. It invites re-examination of
that case, and it's my opinion that a careful
re-examination will result in its reversal. And
the arguments they give, at least some of them,
are novel, and defective. Consider this,
examining fractional vote transfers in STV:
>Furthermore, in multiseat elections, the single
>transferable vote scheme reflects the likelihood
>of a diminished and vanished right to political
>association and right to vote when surplus votes
>and transferred votes result in fractions of a
>vote between two opposing candidates. An
>elector's intent to politically associate with a
>candidate cannot be realized if his one vote is
>divided between two candidates with opposing political beliefs.
>
> From this example, an elector does not know the
> destiny of his vote, and will not realize his
> intent for political association. The fraction
> of ".5" split between the Democrat and the
> Independent cannot reflect the political intent
> of an elector since it cannot be associated
> with two ideologies likely diametrically opposed to each other.
First of all, this concept of "political
association" is entirely invented by the
plaintiffs or the plaintiff's counsel.
Ideologies are constitutionally irrelevant.
And, worst of all, if there are "ideologically"
incompatible votes, it is because voters vote for
candidates, not ideologies. The idea that the
vote, voluntarily cast by the voter, under no
coercion, "cannot reflect the political intent"
of the elector, is preposterous. It is a direct
expression of that political intent. This is
seriously poor reasoning, period, not to mention seriously poor legal argument.
I will examine the vote transfer chart that they
present at the end of this document.
In another legal gaffe, the memorandum proceeds
to rely upon Bush v. Gore as a precedent, when
the Supreme Court, in a flash of candor, stated
that Bush v. Gore was not to be considered a
precedent, but was based on unique circumstances (as I recall).
This argument could be, possibly, quite
inexpedient politically, and, yes, Virginia,
politics has an effect on legal decisions. Bush
v. Gore was highly unpopular with most legal
experts, and raising it as a precedent is waving
a big red flag that says, "Watch out for spurious arguments."
The memorandum then considers the legality of
Minneapolis adopting its own election method,
and, in particular, the STV method. I have paid
less attention to this argument; however, one
part of it stood out to me. The plaintiffs claim
irreconcilable conflict with state election law
regarding election contests over "who received
the largest number of votes legally cast." They
conflate this with "first preference votes,"
assuming that all other votes wouldn't be a part
of that number. This is, thus, the same issue;
the conflict arises only if vote transfers are
not "votes legally cast." Suppose that the voting
method was simple Approval voting. Vote for any
candidate you care to support. I.e., you may vote
for more than one, but never, of course, may you
cast more than one vote for any particular
candidate. With such a system, there is a clear
definition of which candidate received "the
largest number of votes legally cast." This was
the candidate whom the largest number of voters
chose to support. If we consider alternative
votes as being similar, as expressions of
support, but used only if one's first preference
is eliminated, we still end up with the winner
being the one whom the largest number of voters
chose to support, whether through first preference votes or alternative votes.
Now, there is a problem with sequential
elimination, in that it doesn't treat all
lower-preference votes equally. Specifically, it
does not count them until the voter's first
preference has been eliminated. This can cause a
candidate to lose who would clearly win in a
direct contest with the IRV winner, because the
votes for that candidate were not "uncovered"
until the candidate was eliminated - and these
votes are never even counted, typically. However,
this is a complex issue, and for this to be the
basis of a constitutional challenge would require
far greater legal skill than is being exhibited.
Now, about the STV vote chart they provide. At
the start, I will explain the Droop quota used.
It's probably easiest to understand that quota,
why it is V/(N+1) + 1, rounded down, V being the
number of valid ballots cast and N being the
number of candidates to be elected, by
considering first the case where N = 1. V/2 + 1,
rounded down, is a simple majority. That is, it
is the lowest number of votes that is more than
half. With two candidates, the quota is the
lowest number of votes that is more than a third.
And so forth. It seems complicated, but it's actually pretty simple.
Then, when a candidate is elected, the candidate
usually has more than the quota of votes. If we
subtract the votes used from the votes needed to
meet the quota, then we have unused votes,
"surplus votes." So that this group of people is
fairly represented according to the proportion of
this group in the electorate, these votes are
reassigned according to the lower preferences on
the ballots. If these voters all voted
consistently, it would not be necessary to use
fractional transfers, for, quite simply, the
excess votes would go to a single candidate. Thus
a group that was a vote more than half the
electorate would get two winners if three are
being elected. (If this seems unfair, then we
should really look at deeper systems of
proportional representation that involve electing
more than three candidates! It's better than the
standard plurality at-large system that would
give all the winners to a majority of voters.) In
fact, however, voters are not robots, voting
consistently with each other. So what is done is
to consider that these voters have had a portion
of their vote used to elect their favorite (if
these were all first preference votes), and then
they individually cast their second preference votes at a reduced value.
The plaintiffs consider a two-seat election, with
four candidates and 10,000 voters. The candidates
are a Republican, a Democrat, and Independent,
and a Green, and the first preference totals are
4000, 3000, 2000, 1000. This is an example, it
seems, that they created. It's not a realistic one.
They definitely did not create and describe the
example in order to make the STV system clear.
Indeed, it may have been part of their desire to
make it seem as complicated as possible.
Be that as it may, or not, the quota with 10000
votes and 3 seats is 3334. The Republican is
first elected, using up this number of votes, out
of 4000 total, leaving 666 votes unallocated.
Each voter is then assigned 666/4000 vote to be
assigned to that voter's next preference. They
don't explain this part, they just do the math,
it seems that they don't realize that it is the
voters who are controlling the vote transfers,
through their ballot choices. To make the
expression maximally unclear, they do the math
backwards, first calculating and expressing, as a
decimal, the fraction (4000 3334)/4000, i.e.,
16.65%, and then multiplying it by 4000 to get
666.
They don't state it very clearly, but apparently
the Republican voters voted second rank as 3000
for the Independent , 1000 for the Green, and
none for Democrat. While that's thoroughly
unrealistic, I can accept it just as an example. So the transfers are
To the Independent, 0.1665 x 3000 = 499.5 votes
To the Green, 0.1665 x 1000 = 166.5 votes.
They seem quite disturbed by the 0.5 vote
fractional parts, as if there is some voter who
has half their vote going to the Independent and
half to the Green. That's not at all what is
happening. What is happening is that each voter
is casting, through their second preference,
0.1665 vote. Which happens to come out to totals
with exact half-votes in them. It could be any
number; the method described in the Minneapolis
procedure calculates the surplus ratio to four decimal places.
The way they state it is, "The transferred vote
in this example fractionalizes one vote into
one-half for two candidates, the Green and the
Independent." Which is quite incorrect. What has
happened is that 3000 votes for the Independent
have been reduced in weight to 499.5 votes, due
to those voters having already elected a
preferred candidate with most of their vote, and
likewise 1000 votes for the Green are fractioned
to 166.5 votes. There is no "vote" which was split in two.
At this point, no other candidate has reached the
quota, so vote transfer due to elimination
begins. The Green is eliminated. The Green votes,
1000 votes, will be transferred to the second
preference expressed on the Green ballots, and
the third preferences of the Republicans who
voted for the Green as second preference will
likewise be transferred. They lump these together, increasing the confusion.
They don't provide the ballot data, just totals,
which they do not explain. Of the 1166.5 votes
held by the Green before elimination, they assign
166.5 to the Democrat and 1000 to the
Independent. This is exceedingly odd. 166.5 is
the number of votes transferred from the
Republican voters to the Green as second choice.
It is 1000 ballots with Green marked as second
preference, and deweighted. So are they imagining
that the Republicans who ranked the Green second
would all rank the Democrat third? Maybe. But
they have, then, every voter who ranked the Green
first preferring the Independent as second
choice, which is astonishingly unlikely if these
labels mean anything. Even if they are purely
arbitrary, i.e., "Republican," "Democrat," etc.,
might as well be "Candidate A," "Candidate B,"
etc., vote transfers like this wouldn't make sense.
The vote transfers from the Greens put the
Independent over the quota, so the Independent,
while trailing the Democrat in first preference
votes, by a large margin (2000 to 3000), wins the
second seat. This kind of election behavior,
quite simply, doesn't happen in real IRV
elections. The norm is that leaders in first
preference win the elections. However, strongly
partisan elections could possibly show different
patterns. IRV in Ann Arbor, Michigan, resulted in
the election of a Democrat when previous
elections had been spoiled by a Human Rights
Party candidate gaining some significant
percentage of the vote (I forget the number, it
was high for third parties, 10%?). But a loss to
a candidate initially leading by a ratio of 3 to
2? In order to show that, they had to use
drastically slanted vote transfers that seem to
assume ideological voting based on clear party
differences , yet which also don't make much sense even then.
Further, they neglected exhausted ballots
entirely. They show none. Again, highly unlikely.
Two candidates are being elected, but they show
no exhausted ballots even though some are
reaching down to third preference (the 1000
Republican first-preference voters who ranked the Green second preference.)
But they still try to make hay from this. When
they return to examine this sample election, they
state that "The City cannot declare that creating
fractional votes, splitting the political
associational intent of one voter between two or
more political ideologies is a small infraction
of constitutional rights to rationalize the need
for single transfer voting." Sic. They really should find a good proofreader.
In fact, that "split" vote refers to the 0.5
vote fractional part of the vote totals
transferred, and does not, at all, represent a
single voter's vote being split between
candidates that were not chosen by the voter.
Rather, 3000 Republican voters cast a lower
preference vote for the Independent, and 1000 for
the Green, at 0.1665 vote each, having
experienced the win of their first preference.
They will not be complaining. The Green votes
ended up with the independent and the Democrat;
in no case did this happen without a voter
explicitly assigning their remaining fraction of
a vote to that candidate. By not reporting the
actual ballot breakdowns, they make it hard to
see this, but I can hope that the court will not
be taken in. Voters do not necessarily vote
"ideologically." They vote for candidates, and
"ideology," manifested through party affiliation,
is not a designed part of our system, some might
even consider it a parasitic growth. I highly
doubt it is mentioned in the Minnesota
Constitution, so this memorandum is not legal
reasoning as to the constitutionality of the
voting method, it is rationalization.
IRV is, compared to other, simpler alternatives,
an expensive election method that makes only a
modest improvement in performance over Plurality
and which actually degrades performance compared
to Top Two Runoff.. But it is properly
constitutional, and preferential voting was only
found unconstitutional in Minnesota (and nowhere
else) through some very poor legal reasoning that
just might, indeed, be analogous to Bush v. Gore,
but it's hard to tell after this lapse of time.
Bucklin voting, like other voting reforms, indeed
like any movement toward purer democracy, could
be feared to allow third parties to rise in
prominence, and thus it had to be stopped. And
for that kind of purpose, court majorities have
been known to manufacture entirely spurious arguments.
----
Election-Methods mailing list - see http://electorama.com/em for list info