[EM] British Colombia considering change to STV

[Abd ul-Rahman Lomax]

At 10:18 PM 4/29/2009, Kathy Dopp wrote:
>STV has *all* the same flaws as IRV but is even worse. It is 
>unimaginable how anyone could support any method for counting votes 
>that is so fundamentally unfair in its treatment of ballots and 
>produces such undesirable results.

I don't think Ms. Dopp, who has put quite a bit of effort into 
exposing the problems with IRV, has really studied the situation with 
STV. Obviously, the problems of IRV are the same as single-winner 
STV, they are the same method.

However, multiwinner STV, particularly forms that use fractional 
votes to deal with reassignments, is quite good *with early choices,* 
and only breaks down when there are too many candidates, such that 
eliminations start before there is fair assessment.

I'm going to describe a variation of STV that isn't necessarily 
what's being used, and, in fact, I don't think it's being used 
anywhere in details. But the Droop quota variation may be in use, more or less.

That is, STV chooses winners according to a quota, and were it the 
Hare quota, and if it were required that a candidate receive that 
quota to be elected, it's clear that every STV winner is reasonably 
fair, if certain conditions obtain. The conditions: lower rank votes 
aren't coerced (as they are in Australia), and voters never choose a 
lower ranked candidate when they won't be reasonably satisfied with 
that candidate as a winner. Under these conditions, every winner is 
reasonable as a representative. Every winner chosen before 
eliminations begin is also *ideal* as a winner.

In the first round of counting, every candidate elected is the first 
choice of a quota of voters. Only those candidates are eliminated 
from the subsequent rounds (before eliminations begin) who are 
already elected. Let's assume that a candidate got double the quota. 
The Hare quota is 1/N * (total valid votes) with any choice shown. If 
there are ten seats in an assembly to be elected, and a candidate 
gets, as a first choice, one-tenth of the vote, that candidate is 
obviously a good representative for one-tenth of the electorate.

But what if the candidate gets two-tenths of the vote? If the ballots 
electing that candidate are eliminated, the *faction* "led" by this 
candidate is under-represented. So the idea is to eliminate half the 
votes and to then reassign the rest to lower ranked choices. How is 
this done? Some places use methods like random elimination, and I 
don't know the details. But the fairest method to me seems to be to 
revalue the votes, all of them, and look at second-rank votes. In 
this case, each ballot would then be worth one-half of a vote. So, 
second round, if all these voters voted the same second rank, they 
would then elect their second choice.

This then would proceed iteratively until all ballots have been 
assigned to a winner, or eliminations would need to start. So far, 
every choice has been clearly a good representative.

Problems begin with eliminations, where votes cast are set aside and 
replaced with lower rank votes, not having been used to create a 
winner. Further, the Hare quota isn't used, rather the Droop quota is 
used. It's assumed that some ballots will not contain a countable 
vote, and that some ballots will not end up choosing a 
representative. Maybe the first vote is for some unpopular candidate, 
and there were no second choice votes. That one ballot would prevent 
the election of a full assembly, because a Hare quota couldn't be 
found for the last seat. So the Droop quota is used, the fraction, 
instead of being 1/N, is 1/(N-1). This, then, allows a supposedly 
realistic compromise to be found. In addition, (I'm not sure about 
actual practice), the quota may decline, being revised according to 
the number of remaining seats and the number of "unexhausted" ballots.

Lewis Carroll (Charles Dodgson) realized the problem and invented 
what we now call Asset Voting to deal with it. Asset Voting doesn't 
allow ballots to be exhausted, publishing a pamphlet in 1884. It's 
not at all clear that Dodgson realized all the implications, he was 
just thinking about politics being more or less the same as it was, 
and that voters would vote in more or less the same way. He was 
concerned that many or most voters don't really know much more about 
the various candidates except to know who their favorite was.

So his idea was to assign an otherwise exhausted ballot to the 
favorite, and this vote becomes, as it were, the "property" of the 
favorite, to be reassigned at the discretion of that candidate. With 
this fix, which converts a raw election method into an input stage 
for a deliberative process, and with a Hare quota (to a slightly less 
extent with the Droop Quota), totally fair method of choosing 
winners, not subject to most of the normal election pathologies. But, 
to my knowledge, it's never been tried.

The big secret: if this were done, all votes would count. No votes 
would be wasted. Voters would still be able to use a preference list, 
if they prefer to control vote transfers before their ballot is 
eliminated from direct use. Many people, reading about this for the 
first time, seem to think that this gives too much power to 
politicians, but they, I'd say, haven't thought it all the way 
through. We are electing representatives who will exercise power; 
this allows someone trusted with a vote to have a say in who gets to 
actually exercise the power, counting and using votes that would 
otherwise be *eliminated.* Don't trust a candidate to recast the vote 
in a good way? Why, then, would you trust that candidate to directly 
exercise the power? It makes no sense to me, having thought about 
this for many years.

Now, actual STV? It's probably reasonably fair up until the choice of 
the last seat, where the IRV problems kick in fully. Before that, 
it's quite possible that eliminations result in less than optimal 
choices, but still reasonable ones, unless full ranking is coerced.

My biggest problem with premature election reform is that instead of, 
say, forming a commission to study voting systems, and making sure 
that this commission hears and considers evidence about all possible 
systems, and then presents a full report, in detail, of all the 
implications, there usually isn't any kind of comparative inquiry at 
all, there is only the raw choice of the present method or a single 
proposed reform; and even when there is a study group, as in 
Colorado, the decisions are made without adequate back-and-forth with 
experts, and, further, recommendations may be made based on political 
expediency (which method might actually be accepted, due to all the 
complex considerations?) instead of on an accurate comparison of 
methods (which would then leave the matter of practical expedience to 
the experts on that, the elected politicians who actually make the 
final decision).

STV should be on the table, but there are other proportional 
representation systems, including variations of STV in use, 
reweighted range voting, and, indeed, Asset Voting. Systems have been 
proposed in the past which also use a quota to elect, but which don't 
reassign votes; instead, they reassign voting power in the assembly; 
this was proposed for, I think, one or two cities in the U.S. in the 
first part of the last century. This is fairer, in fact, than STV, 
and far less complicated, but it bucks the idea that we should elect 
peer assemblies, where every member has the same voting power as 
every other one. Asset Voting solves this problem by creating a peer 
assembly, as we are accustomed to seeing, but, still, with every 
ballot (within certain limitations) having served to elect a seat.

It's even possible for it to be known what each individual ballot 
did, if when votes are reassigned, it's done by precinct; at least 
every voter who voted for only one would know, quite well, what their 
ballot did. (Because the precinct counts wouldn't be exact, but 
merely close, it would be more accurate to say, if a precinct's votes 
were one-tenth of the quota for a candidate, that 90% of the voter's 
vote elected Candidate X. The voter knows what precinct he or she 
voted in. However, the candidate doesn't know exactly which voters 
elected him or her.)

What's truly interesting to me about Asset Voting is that it can 
shade, one small step at a time, into a hybrid direct/representative 
democracy that retains the best features of both. Deliberative 
assembly with identified representatives with special rights: the 
right to introduce motions and to address, by right, the assembly -- 
always the bugaboo of direct democracy because scale makes large 
assemblies bog down; it's not *voting* that's the problem, it's 
deliberation. Extended penumbra around the assembly consisting of 
those who received votes, they are known and identified, but who 
reassigned those votes to create winners. These can be considered 
"electors," and could have the right to vote directly. Because they 
are known, and the votes they recast are known, it becomes possible 
for their votes, should they choose to vote, to be counted, and the 
votes of those they elected then would be adjusted down accordingly, 
fractionally.

My opinion is that these votes would only rarely affect an outcome. 
However, that they are there in theory would act as a restraint on 
the assembly, and, in addition, rapid and efficient recall could 
become possible, but even more than this, the penumbra acts as an 
advisory body for the assembly, directly representing, by free 
choice, the electorate, no compromises.

You want to become part of this advisory body? You can. Just declare 
a candidacy and vote for yourself. (Now, there might be some 
restrictions, and you might have to declare a second choice, for 
various reasons that I won't go into, there might be some minimum 
threshold below which votes would not be reported, it has to do with 
voting security and the possibility of small-scale coercion. I 
personally think that the coercion problem is vastly overblown under 
most conditions; it's only a problem on a large scale, which doesn't 
apply to the problem that a candidate, say a wife, who demands that 
her husband vote for her, and she doesn't get any votes except her 
own.... So it's possible that singletons would not be recognized, or 
there are other solutions.... I can think of some, but ... way ahead 
of necessity!)

The assembly size can be chosen to be the most efficient at providing 
the combination of best process (too many cooks spoil the soup) and 
full representation. The direct voting option means that making a 
compromise on assigning a vote to create a seat is less important, 
for any elector can still vote directly, if matters that much.

The basic problem with all election methods that resolve with a 
single ballot is that *it's impossible for a single ballot to make 
sound choices under all circumstances.* Further, fixed terms create a 
gap between a representative assembly and the people it represents. 
Asset voting, where every voter can vote for the representative most 
trusted, without compromise, and where all the electors thereby 
created, should minimize or even eliminate the gap. Again, because 
votes can be amalgamated on a very small scale with Asset voting, I'd 
expect campaigning for office, as such, to practically disappear. 
Rather, someone just starts by registering as a candidate, and could 
take as many election cycles as necessary to build voting power, with 
no waste. When a candidate has reached the point where their voting 
power is a significant fraction of that necessary to hold a seat, 
they become, in effect, a lobbying representative to the holder of 
the seat, with real and tangible voting power behind it. They not 
only can deliver votes through influence, they directly deliver 
votes, it's all out in the open.

The real importance, though, isn't in voting and votes, it is in 
creating a broad network that *deliberates* and *advises*.