[EM] Student government - what voting system to recommend?

[Abd ul-Rahman Lomax]

At 11:08 AM 4/25/2007, Tim Hull wrote:
>That's the answer to what would likely happen if Range voting were 
>implemented anywhere of significance - cards like those used in 
>Australia would appear telling voters how to vote.  Granted, it 
>would probably happen under any preferential system, but in range it 
>is almost guaranteed that bullet voting would be encouraged "except" 
>by minor party candidates, which MAY recommend voting for major 
>party candidates as well.

Sure. We can expect the appearance of cards. What we *don't* know is 
how widely the cards would be followed. It may depend on other 
details of the system.

>Though I like the basic idea of range, I will say I have 
>reconsidered somewhat when it comes to contentious elections.  In 
>this case, I do see it degenerating into Approval really fast.

I'd suggest looking at the problem more carefully. First of all, if 
it "degenerates" into Approval, it has become a system which many 
experts think quite a good one! Approval *is* Range, just not as 
finely resolved.

Secondly, the scenarios I've seen that allegedly show reward for 
Approval-style voting actually don't. Indeed, one that was proposed 
yesterday or the day before, here, show severed punishment of some 
voters (40%) for bullet-voting, Had they actually voted Approval 
style, and voted for *both* candidates whom, by the sincere ratings, 
they really almost equally preferred, they would have had a vastly 
superior outcome. Next time, indeed, we might expect, they would 
realize that to drastically derate the centrist candidate, as they 
did in the scenario, they are lying to the system and shouldn't be 
surprised that the system pays attention to their lies and gives them 
what they did not want!

>   While Approval is a decent system (better than FPTP for sure), 
> third parties would still have significant trouble breaking through 
> (major parties will always bullet vote, and a large number of 
> third-party supporters will vote for the major party as well).

Again, assumptions are being made that have no basis in reality. Yes, 
*many* major party supporters will simply bullet-vote. But many 
others would not. I'm a registered Democrat and pretty consistently, 
under the present system, vote for Democrats. But there are certainly 
times I would have voted for third-party candidates as well.

No system is going to come into the present environment and suddenly 
award third parties the victory. IRV certainly isn't going to do it. 
Warren Smith has pointed out that IRV is practically political 
suicide for third parties to support. Sure, it allows them to gain 
votes (as would Approval, much more simply), but if they ever 
approach parity, once again the spoiler effect appears with a 
vengeance, and could bite them so badly that they would be doomed as a party.

Once again, the criticism of Approval does not really stand as a mark 
against Approval. Under Approval, as under all Range methods (and 
some other methods as well), third parties can start to gain votes 
without throwing elections to the dogs. That's what they need now. 
Approval behaves better than, say, IRV, as a third party approaches 
parity. Range does better. And there are hybrid systems that do even 
better, such as Range plus a runoff if there is a Condorcet winner 
who is not the Range winner also.

If you can't win the pairwise contests *and* you can't win the 
sum-of-votes contests, I'd suggest, you shouldn't win!

And the objection I've seen to the runoff proposal is that, 
supposedly, the voters have already expressed how they would vote. 
But this objection assumes that voters will maintain the same 
preferences, even though they now have some important new 
information: how acceptable the other candidate, the Range winner, 
is. Some voters will vote, essentially, to maximize overall 
satisfaction, as long as that result is quite acceptable to them. If 
they don't, well, I *do* believe in majority rule. Which is not the 
same as the Majority Criterion, which applies to preference systems 
and pays not attention to preference strength, as the Pizza example 
shows so clearly. The majority in the Pizza election will generally 
have no problem with accepting its second choice, once it knows how 
the rest of the electorate feels about their first preference. Wouldn't you?

>  We may see "shifts" between one party being major and another 
> being major from election to election as voting results start to 
> demonstrate shifts in support and voters vote accordingly, but it 
> seems like each election will continue to be a basically two-party 
> competition in and of itself.

For how long? And is this a problem? What happens, typically, is that 
the major parties adopt the popular positions of the minor parties. 
Other things take place as well. In Minnesota, I was reading, there 
really is a three-party system, and one of the parties is the 
Democratic Labor Party, which is Minnesota's version of the 
Democratic party. The third major party is the Independence Party, 
which is somewhat of a moderate libertarian party, which does manage 
to elect some officials.

I don't see better election methods as, by themselves, bringing third 
parties up to parity,generally. But I also don't think that 
necessary. They make it more possible, by allowing third parties to 
get a leg up without damaging the election. IRV is widely proposed 
for this purpose, but Approval does it far more simply. Just stop 
tossing ballots that are "overvoted!" Just Count All the Votes!

I have been unable to find any justification anywhere for tossing 
overvoted ballots, it seems to have been assumed from the beginning 
that they must be errors. Robert's Rules, talking about it, says that 
they should be discarded "because they are errors."

There is an argument that is made, sometimes, that discarding 
overvoted ballots allegedly prevents fraud. However, it is quite 
clear that overvoting rules have, instead, been used to perpetrate 
frauds, in various ways. Some have proposed that Approval ballots 
have a specific No vote, which then can allegedly detect ballot 
alteration fraud. But I don't think that will work very well either. 
It is better to protect against fraud in far more direct ways, most 
specifically by the ballot imaging proposals I've made, which are 
already practical, do not increase public costs at all, and which 
would nail down public confidence in vote counting. Since the public 
could verify it, easily.

>In my mind, that leaves IRV and Condorcet as serious voting reform 
>proposals.  While IRV isn't the best thing in the world (it can 
>screw up when a new party becomes "major" in an election), it could 
>help third parties with a solid coalition with a major party ( i.e. 
>nearly all Democrats vote Green #2, and all Greens vote Democrat #2).

What I fail to understand is why Approval was rejected above, when it 
allows quite the same thing, much more simply! In the Approval 
example, it is true, we would see Greens vote as described, mostly, 
and the Dems would much more bullet-vote -- but many would also 
approve the Green, it is cost-free as long as the Greens aren't in 
position to win -- but the problem with Approval is that it isn't 
refined enough to allow voters to show preference among those 
approved (or disapproved.) Range does this. So does a method called 
A+, which is just Approval plus a Favorite designation. A+, by 
itself, does not use that information, but it allows measurement of 
relative popularity, which is all that is really being proposed 
above. That's the only information you get out of IRV about third 
parties which are not going to win.

But Range does it better. And the best voting strategy in Range, 
overall, is to estimate one's own Approval level..... The allegation 
that Range will devolve into Approval is based on incomplete analysis 
of scenarios, neglecting the fact that for some voters, bullet-voting 
can drastically harm some voters, and can only slightly benefit some 
voters. We actually have no evidence that Range will degenerate into 
Approval, and, if it has, well, so what? It simply becomes a 
not-quite-as-good method, it does not become a bad method! Which IRV 
can easily do:

>  It does suffer from the "center squeeze", though.  Condorcet, on 
> the other hand, does not suffer from the center squeeze.

Right. But Condorcet, as a ranked method which does not consider 
preference strength, can make spectacularly bad decisions, and in 
real situations. Among such pure ranked methods, Condorcet-compliant 
methods are obviously supreme.

In real-world elections, the differences between Condorcet and Range 
methods can be small. *Usually* the results coincide. Approval also 
*usually* picks the Condorcet winner.

>   However, it suffers from the opposite problem - the so-called 
> "Pro Wrestler" or "Loony" syndrome in an election with a couple 
> polarized candidates and a weak centrist or joke candidate.  In my 
> student government elections, I picture this being a candidate 
> walking around campus in a clown suit and winning based on becoming 
> everybody's #2.  Also, Condorcet's later-no-harm failure may mean 
> people give a less sincere ranking than in IRV, though this failure 
> is far less so than in range.

Again, poor analysis of Range. Later-No-Harm failure in Range is 
dependent upon some assumptions about how voters will vote. And, 
typically, what is not considered is how the "failure" actually 
affects the outcome from the point of view of the voter, it is simply 
assumed that the voter will be dissatisfied if they get a 
slightly-less-than-most-preferred outcome. Not so! As a voter, I can 
tell you that I'd be pretty happy to get a candidate I rate at 80% 
instead of one rated at 100%, in a field of candidates all rated much 
lower than that, with my most-hated candidate right up there with my 
favorite in pre-election polling! Yet, analysts just look at the 
technicalities of the criterion and say, it failed.

It's the same, really, with the Majority Criterion, which easily gets 
confused with majority rule. Majority rule is measured by votes that 
are Yes/No. It is well-known that no election method is perfect from 
every point of view, but this is only with elections where there are 
more than two choices! The simplest election is Yes/No. And that is, 
in fact, a default election method under Robert's Rules, used when 
something more complex has not been chosed, typically for efficiency. 
It happens to be, under RR, Condorcet-compliant, but because motions 
are debated, it *also* will incorporate Range-type information.

You cannot apply Majority Rule, as a principle, to elections with 
more than two candidates, and it really isn't even correct to do it 
with two, for there is always a third option: None of the Above, 
unless the majority has already specifically determined that one of 
the two *will* be elected.

Instead, some majority at some time in the past determined that.... 
Majority Rule means that the majority *today* can decide! Which it 
can, under Approval and Range, as long as it votes its preference 
strictly. If it allows a minority candidate some of its votes ... 
it's no longer a pure majority..... so, big surprise, those allowed 
votes can cause the minority candidate to win.

Another way of looking at Range is that every voter gets N votes to 
cast *for each candidate.* It is as if there are N times as many 
voters, voting Approval. And if the majority of *these* voters cast 
their votes preferentially for a candidate, that candidate cannot fail to win!

>However, PR still seems like the primary thing to shoot for - single 
>winner elections really aren't any good in achieving better representation.

Lousy, in fact. They aren't representation at all, unless you assume 
that California is represented if a Californian is chosen by someone 
or some process.... Single-winner represents *districts*, not people. 
It is really a system that made sense under the crown.

>   For that reason, I can see the logic of those pushing IRV with 
> the intent of moving in a PR direction.

Yes. It is a strategic decision that they apparently made. However, 
they don't reveal that this is why they chose IRV. Instead, they 
pretend that single-winner IRV is a good method. It isn't.

>   I will say that, given honest voters and an absence of "Loony" 
> type candidates, Condorcet produces better results and seems better.

>   However, it is more complex - and is yet another system to 
> discuss. I really think that STV should be the real goal, with IRV 
> used in single-winner elections (when necessary) for consistency - 
> party lists are rotten by comparison, and no other system has been 
> tested and proven for multiwinner to the extent of STV.

STV is quite reasonable multiwinner, though Asset is far more 
accurate, in theory. There is also RRV, which should be better for 
the same reason that Range is better than Condorcet. It is just that 
with multiwinner, the problems of non-range methods don't bit as deeply.

Asset, of course, finesses the entire problem, and is the one method 
which can then convert, easily, to something so close to direct 
democracy that the difference is academic. So if you *really* want to 
look long term, I'd really suggest considering Asset quite seriously. 
It is also quite similar to Delegable Proxy, which is, some of us 
thing, the choice of the future. (Indeed, the essential aspect of 
Asset is only a secret-ballot layer, above that Asset can function 
quite like delegable proxy).

>Thus, I plan on moving in an IRV+STV direction as far as my reforms 
>(with multiwinner STV used for 70% of the seats on my student 
>government i.e. all of the multi-seat districts).

Whatever.