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

[Abd ul-Rahman Lomax]

At 02:14 PM 4/23/2007, Juho wrote:
>Let's say that in the U.S. presidential elections roughly 48% of the
>voters vote D=9, R=7, PW=1 and roughly 48% vote R=9, D=7, PW=1.
>Either D or R wins.

The premise is utterly insane and, quite simply, not reasonable. 
Range is difficult to analyze through the simplistic "this block 
voted this way" kind of analysis we are accustomed to using for 
election methods.

What has been presented here is an extremely close election. 
Essentially, going into the election, it is a tossup who will win. 
Note that by the premises there are *no* voters whom I would call 
"Republican" or "Democrat" based on their votes. There are only swing 
voters, voters where their preference for either major party is weak. 
This bears utterly no resemblance to the real world. In the real 
world, there is a set of voters who are dedicated party supporters, 
and then there are other voters, perhaps the majority, who aren't so 
nailed to a party. A minority, perhaps, would vote as described. And, 
in fact, they are much more likely, I'd suggest, to rate a third 
party candidate higher.

>  In the next elections the Democrats notice the
>possibility of strategic voting and advice their supporters to vote
>D=9, R=0, PW=0. In these elections Democrats win.

Again, we have a scenario in which there are no dedicated party 
supporters, as shown by the initial votes. Yet, somehow, these voters 
who really are on the fence (9 and 7 are both fairly strong votes for 
a candidate), will suddenly obey, as a block, party discipline and 
lie about their preferences in the next election. Remember, if they 
rated the candidates honestly, they were happy with either outcome, 
according to their votes. Where does the motivation come from to 
change this position? I'd say that it would happen only if there was 
massive dissatisfaction with a Republican winner.

Further, note that the PW candidate now gets zero from this group. 
That's really not much different from the vote before. But it is 
totally unnecessary. Why would these voters suddenly drop their 
(small) support for the candidate with no chance to win? Remember, 
it's Range. They do not threaten a front-runner by rating a third 
party candidate or write-in higher than zero. If they voted as a 
massive block, and rated him higher than the median, maybe. But that, 
likewise, is quite unreasonable.

Basically, the premise behind these analyses, and we have seen them 
many times regarding Range Voting, is that voters will express a weak 
preference and then be ... shocked, absolutely shocked! ... that 
their weak favorite is not elected. If they cared this much, why did 
they express such a weak preference?

It's a contradiction, and it is blithely assumed as a condition of 
the analysis. Hence the analysis is next to worthless.

It's about time that this be widely recognized. I can understand a 
newbie coming across with this kind of argument, but Juho is not a 
newbie. He should know better! Perhaps he was having a bad day....

>  In the third
>elections Republicans have learned a lesson and now recommend their
>voters to vote R=9, D=0, PW=0. Now the election is in balance again,
>but the method has in practice reduced to Approval (actually
>Plurality in this example).

Sure. But the premises are totally unrealistic. If you are going to 
propose that Range will *reduce* to Approval, you will have to use 
reasonably likely scenarios. It's fine, I'd suggest, to use 
simplified voting patterns rather than the complex distributions that 
would actually occur, but if you oversimplify them, and make it 
appear that there are, for example, only two or three voting patterns 
in a Range election, you then set up conditions to make it appear 
that Range has reduced to Approval.

The fact is that if even the majority of voters bullet-vote, it has 
not reduced to Approval. What this means, in fact, is that, if the 
election is close, swing voters will decide it in a far more 
sophisticated way than happens in plurality, and without the spoiler effect.

And if it *does*, under some difficult-to-anticipate circumstance, 
reduce to Approval, that isn't a bad outcome!

>This strategy doesn't require the voters be rocket scientists.

That is correct. It requires them to be mindless robots, voting as 
advised by party bosses.

>Probably the strategies would not spread as described above.

Duh! If this is "probably true," then why propose the scenario as 
reasonable in the first place?

>  Maybe
>there just would be discussions between voters and in the media and
>all parties would be impacted in roughly same speed.

Or it would have no effect at all. In reality, party fanatics will 
probably bullet-vote from the outset. And, let me suggest to you that 
if a candidate advised voters to bullet vote instead of voting their 
honest opinions (in a context where this is not going to spoil an 
election), I'd consider that a strong disqualification. That 
candidate would have just lost my vote, because he or she does not 
understand democracy. And if the party advised this and the candidate 
did not disavow it, ditto. I take democracy seriously, and so do many others.

>  In competitive
>elections it is quite possible that majority of voters would not stay
>"sincere" but would vote in Approval style.

Perhaps. But, by definition, this is a close election. And thus would 
be decided, quite likely, by the voters who *do* vote more refined 
range. If not, so what?

>  Once strategic voting
>becomes wide enough to be meaningful to the end result, voting
>sincerely could be commonly seen as "donating the victory to the
>strategists". A key property of this evolution process is that those
>parties and individuals that are strategic will have more voting
>power than others (this breaks the possible balance of having same
>percentage of strategic voters in each party).

What is being done here is to extrapolate from an unlikely scenario 
to a vision of the future, as if that future were likely. It is 
stated as if it were fact.

We do not in fact know how voters will vote in public elections using 
Range. Warren is about the only person who has done research on this, 
and his results, albeit constrained by the obvious limitations of his 
study, indicate that voters will, at the outset, vote sincerely. 
There is about to be, we think, more information coming from the 
French elections, where it seems there is Range exit polling taking 
place. But that is still polling.

Most of us think that Approval is an excellent first reform. Simple, 
cheap, and fixes the first-order spoiler effect. We don't know 
precisely how the electorate will use it, but the most likely initial 
effect would be that voting would continue about the same as today 
for most major-party supporters. However, those who are sympathetic 
to a third party may start to approve third-party candidates as well. 
Those whose preference is a third party will, in greater numbers, 
vote for the candidate of that party plus their favorite among the 
frontrunners.

Thus, perhaps, only a small number of votes will shift. But, quite 
often, a crucial number of votes. We can expect an improvement in 
outcomes if we simply stop tossing overvotes. Which was a bad idea 
before its time, rooted in a very old error.

So, if somewhere Range is tried, and it turns out to reduce to 
Approval, big whoop! Why this is advanced as an argument against 
Range is beyond me. Particularly if the Range method is 
low-resolution, such as Range 3. Range 2, of course, is Approval.

>I think the size of the election doesn't influence much on if voters
>become strategic. I think it is more like a balance of media / yellow
>press interest, strength of rumours, overall requirement of "good
>moral" in the society, and (maybe most importantly) the level of
>competitiveness in the elections in question.

I agree with most of this, however, I think that the competitiveness 
of the election may have far less effect than Juho predicts. Most 
voters are *not* party partisans! They tend to prefer one party or 
the other, but a huge number, if you ask them, will describe 
themselves as independent. If a party starts trying to tell these 
voters what to do, it seriously risks losing their support!

No, voters will bullet vote if they have a serious preference, and 
they aren't interested in any third party candidate at all. So?

(It can easily be argued that the optimum Range vote, if the voter 
wants to amplify his effect to the maximum -- not all voters really 
care about this -- is to rate your favorite among the front-runners 
at max rating, the other front-runner at zero, and then other 
candidates as they fall, it is mostly moot. If it is high-resolution 
Range, like Range 100, or Range -10 to +10, which is quite detailed, 
really, then you might slip the favorite frontrunner a notch if you 
like, the effect, in Range 21, would be one-twentieth of a vote lost 
to the favorite frontrunner. I recommend, instead, that Range methods 
allow the specification of a Favorite, which is not used to determine 
the winner. Except perhaps in a tie.... Favorite, instead, would be 
used for ballot position, campaign funding, and for just plain 
understanding how the voters stand. There are other ways that Range 
could be structured, as well.)

I have also suggested that if the analysis of Range ballots shows 
divergence between the Range winner and a Condorcet winner, a runoff 
be held between the two. Some, seeing this, imagine that the outcome 
of the runoff would be that the Condorcet winner would prevail. If 
true, that's fine with me. However, it is much more likely to occur 
that the voting public would take into account how everyone else 
voted, and *might* vote to, instead, elect the Range winner. After 
all, that is the winner who, the poll indicated, would maximize voter 
satisfaction. How important is that to *you*?

Would you prefer to "win" if the winner made (almost) half the 
population serious unhappy, and there was another winner who would 
make *everyone* happy, and yourself *almost* as happy? To me, and to 
many voters, I submit, the answer is obvious. Even if I bullet-voted 
for the Condorcet winner in the full Range election.

To repeat it, what the kind of analysis of Range that Juho presented 
does is to make two contradictory assumptions: voters don't really 
care that much which of two candidates wins, and they will be 
seriously unhappy if one of them wins. If they cared that much, why 
in the world didn't they express that in their vote?