At 03:48 PM 6/15/2008, Steve Eppley wrote:
Most analyses of spoiling (and voters' strategizing to avoid
spoiling) narrowly focus on the candidates who reached the general
election ballot. The much bigger spoiling problem is during the
nomination process, when candidates choose not to run in the general
election due to concern they will make the outcome worse if they
run. (Example: McCain in 2000, who chose not to run for President as
an independent after failing to be nominated by the Republican
party.) It's a deterrence of competition. It explains why parties
each nominate only one candidate per office and it explains why few
politicians who care about the outcome join minor parties.
Well, you can call it "deterrence of competition," which is true, but
it is also how the system works. We might notice that deliberative
process works through a series of Yes/No questions, with majority
rule. A series of choices made in a process culminating in a
two-candidate election does come fairly close to simulating a more
sophisticated single-ballot process, and may be superior in some
respects: If, in fact, the Democrats appoint the best candidate they
can find from their party (I'll make that the "most electable
candidate") and the Republicans do likewise, the general election --
if not for spoilers -- would be likely to come up with a good overall
choice. However, there are a pile of ifs in there. Nevertheless, this
is approximately how the existing system works. And to reform systems
we should understand how they work, not merely look at how they don't work.
That sort of analysis falsely makes spoiling in major elections
appear rare, regardless of the voting method (even plurality rule).
When analysis properly includes the deterrence of competition, the
spoiling problem is rampant with plurality rule, top two runoff,
instant runoff and approval.
Mmm... it's more correct that the issue hasn't received sufficient
study. To extend the concept of spoiler to pre-election processes is
problematic. Spoiler effects can, of course, happen in pre-election
process. They are, after all, *elections*, but the application of
"spoiler" to Approval is quite problematic.
As was mentioned earlier, there are two kinds of spoilers: what I
call a "first order spoiler," which is a minor candidate with no
chance of winning who vote splits with a major candidate, causing the
other, less preferred (overall) candidate to win. IRV generally fixes
that, so does top-two runoff and Approval. Approval does it at lowest
cost. Then there is Center Squeeze. IRV fails, reasonably frequently,
I'd expect, in the three-viable-candidate scenario. Top-two runoff
appears to, though it does provide for a recovery mechanism, if
preference strength is sufficient. The influence of preference
strength has largely been ignored. Top-two runoff often does not
actually eliminate any candidates, they merely have a hump to
overcome, and the lower turnout generally expected with runoff
elections should help a write-in campaign to succeed: fewer clueless
voters who don't really care and who might mark one of the two on the
ballot without even being aware of the write-in campaign. In most
cases where there is a center squeeze effect with top-two runoff,
though, the preference strength is probably not enough to motivate a
serious write-in effort.
(And, of course, if it fails, and if there is no majority requirement
in the runoff -- BAD IDEA, as my daughter would say -- then it could
elect the least desired candidate. If you are going to shoot the
king, don't miss!)
The deterrence of competition is the worse problem, in my opinion.
The effect tends to be only two parties that have a chance to win,
each party nominating only one candidate per office, using badly
flawed procedures (primary elections, for example) to decide which
one candidate the party will nominate.
I'm not convinced that deterrence of competition is a problem. Yes,
primary procedures are flawed. Simplest fix: use Approval or Bucklin
or Range (possibly Range 2 or Range 3) for primaries. All of these
would tend to find the most-approved candidates, which would probably
be closer to center, making it more difficult for extreme wings of a
party to take it over. (That happening will generally backfire in
future elections, as the party base evaporates -- party loyalty only
goes so far -- but the damage can take years to undo.)
Furthermore, the two parties tend to become polarized.
In one way. They tend to become centrist in another. And there is
overlap, there is a middle which may affiliate with one party or the
other. Worse than polarization, in my view, is issue linkage. The
party system creates strange bedfellows. I have a friend who is
socially progressive in many areas, just about all, in fact, but for
religious and philosophical reasons is opposed to abortion. Where
does she go? I find it rather odd that people are forced to choose
between what are effectively pro-war, pro-life candidates and
anti-war, pro-right-to-choose candidates.
There is a solution, but few willing to look at it yet. The earliest
hints, in fact, go back to Lewis Carroll, publishing a pamphlet in
1884 that suggested a simple modification of STV: representation of
voters by candidates who received votes. The process still created a
peer parliament (he used the Droop Quota and an STV ballot, but ....
no exhausted ballots. I'd guess -- I haven't gotten the original text
yet -- that he assigned any exhausted ballot to the first choice on
that ballot, to be distributed at the discretion of that candidate.
This system allows voters, if they choose, to vote for one, the
person they most trust to represent them. Whether or not that person
gets a lot of votes isn't important. This is a *choice* of proxy, not
actually an election. The election takes place deliberatively,
through negotiation, except for those seats where a candidate has
received enough votes directly. No wasted votes. It seems that for
most people, even and maybe even especially those familiar with
standard voting systems, it takes quite a while for the idea to sink
in. Ultimately, it could mean the end of elections as we know them,
with elections being replaced with something far more democratic.
What elections take place, then, take place deliberatively. For
starters, deliberative process is Majority Criterion Compliant,
Condorcet Compliant (under reasonable interpretations), and, in fact,
takes account of preference strength informally.
One name that's been given the cause of the polarization is the
"center-squeeze effect." That is, candidates who want to win avoid
moderate positions because candidates at those positions will tend
to lose due to being sandwiched between the candidates to the
"left" and the candidate to the "right."
Center squeeze doesn't cause polarization; rather polarization causes
center squeeze, combined with Plurality methods.
A much older analysis was the model of "two candidates under risk
of 3rd candidate entry" (given plurality rule). That analysis was
more realistic than the conditions in the median voter theorem. The
median voter theorem leads to the prediction that when there are
exactly two candidates competing on one issue dimension, they will
both try to take the position of the median voter, given plurality
rule or any voting method that reduces to majority rule when there
are 2 candidates. The prediction of the more realistic analysis
where more than 2 candidates can compete is that neither of the 2
major candidates will compete at the median position, since if
either does, a 3rd candidate at a position a little further from
the median can enter the race and win.
If vote splitting is allowed to do its work. Absolutely, whether
there is incentive or not, if the two parties run separate primary
processed, the winner will tend to be at the center of the party, not
at the center of the electorate. So a two-party system will tend to
oscillate, as one party wins, starts to exercise its power, and
alienates the middle by going too far to one side. So then the other
party wins and repeats the same. *Usually* the parties don't go very
far to the extreme very quickly. But it could be pretty dangerous.
Some think, I'm sure, that it *has* been dangerous and more than dangerous.
On the matter of polarization, Instant Runoff looks as bad to me as
plurality rule and top two runoff.
Depends. IRV with a majority requirement (which is *actually* what
Robert's Rules describes, contrary to FairVote propaganda) would
likely be better than top two runoff. But who would want to spend the
serious money involved for IRV when there are better, simpler systems
to improve the first stage of top-two runoff. Simple example:
Approval. Both elections are Approval, the primary and the runoff.
Write-ins are allowed in both elections, so, *theoretically*, the
runoff could fail to find a majority winner. My own view is that this
risk in public elections is quite low, and, because majority failure
could indicate that something serious is going on, it's worth the
risk (and expense). But I think that majority failure in a runoff
would be so extraordinarily rare, particularly if the runoff were
Approval, that it's not worth worrying much about it. I could be
wrong, which doesn't mean that a majority requirement is bad. Might
be better to make it easier for people to vote!
Polarization of two big parties may create a false impression that
the voters too are polarized. It may also increase actual
polarization among voters, since voters won't hear as many moderate arguments.
That's right. Voters, in fact, aren't nearly as polarized as public
campaign rhetoric -- which doesn't come from the average voter --
might make us think. It's pretty obvious that real voters aren't so
attached to one party or the other winning, long-term. Many, many,
think of themselves as "independents," whether or not they are
registered as a member of one party or not.
One reason some people are clamoring to replace plurality rule with
Instant Runoff appears to be occasional election outcomes like Bush
in 2000, when Nader was a spoiler. Most Nader voters preferred Gore
over Bush, so Gore would have won if Florida had used Instant Runoff.
Maybe. What about Buchanan, etc.? I agree that Gore likely would have
won, but I don't think it is by any means certain. Nader was arguing
that it didn't make any difference, that a vote for Gore was a vote
for the status quo, and, I think as well, if Bush won it would help
hasten the end of corporate dominance. Many Nader voters wouldn't
have voted for Gore.
The trouble with the argument is that in a close spoiled election,
the number of voters who prefer the spoiled outcome is a large
minority; there won't be a consensus that the spoiled outcome is
worse. When the winner of a spoiled plurality rule election has a
large minority of the votes, given only the votes it doesn't matter
to the well-being of society which of the top two wins. It only
appears to matter, since the combination of the spoiling and the
polarization leads to a big difference in policy choices. The
closeness of the vote totals means one cannot tell, given only the
votes, which set of policy choices is better for society, so in my
opinion the ever-present problem of nomination deterrence is much
worse than occasional spoiling in general elections.
Hard to tell, don't you think? Certainly there are common opinions
that have never been proven to be true. Such as, for example, the
opinion that low turnout in runoff elections is a bad thing. It may
be a good thing, shifting results toward Range results (by
preferentially devaluing low-preference-strength votes.)
P.S. - I presume some people will say the 1992 Clinton-Bush-Perot
election was spoiled, since some polls indicated Bush would have won
if Perot hadn't run. In other words, that Perot took more votes from
Bush than from Clinton. That may be. However, there was apparently a
majority cycle, according to a poll by Gary Jacobson of UC San
Diego. If Perot had dropped out, Bush would have won. If Clinton had
dropped out, Perot would have won. If Bush had dropped out, Clinton
would have won. So which one was the spoiler?
Most would assign this title to the minor candidate, without a chance
to win. Otherwise it would be a Center Squeeze effect.
Furthermore, the smallest of the three majorities was the majority
who preferred Bush over Clinton. That means Clinton would have won
anyway given the Condorcetian methods most popular among the members
of this email list, if Jacobsen's numbers were right.
P.P.S. - Kathy Dopp's request for arguments about Instant Runoff was
expressed in terms of support and opposition. Those terms are
absolutist shorthands that omit the key information about what the
item is being compared against. By the same reasoning that a voter
could oppose Clinton (when compared against Obama) and support
Clinton (when compared against McCain), someone could both support
and oppose Instant Runoff. How should someone who believes Instant
Runoff is slightly better than plurality rule and Condorcet is much
better than Instant Runoff respond to her request? Preference orders
are more informative than statements about support or opposition,
since they reveal the comparisons being made.
Preference order is certainly more informative than core support or
first-preference-only, i.e., Plurality. However, missing from it is
preference strength. Approval incorporates a kind of preference
strength measurement through averaging of votes; the independence of
approvals makes this work. Range does even better, of course, by
allowing the direct expression of preference strength. It's an error
to consider that ratings are "sincere," they are votes, just like
Approvals. Just as some theorists dismissed approval because
"approval" doesn't have any fixed meaning, the same has happened with
Range, the ratings don't have a fixed meaning. But Range, regardless,
allows voters to feed preference strength information into the
hopper, and "strategic exaggeration" is quite possibly an oxymoron.
There is no motive, no reward, for actually exaggerating one's true
preference strength. The idea that this is comes from assuming some
"sincere preference strength" which is then modified by
"exaggaration" in order to "improve the outcome." Wait: under Range,
if it improves the outcome, that's a sincere voter preference. What
Range does is to restrict voter preference strength choices so that
the sum of them is one full vote. So one assigns the vote strengths
where they matter; they are von Neumann - Morganstern utilities,
adjusted according to probabilities so that the voter puts the votes
where they count.
My own proposal has been Range with pairwise analysis. If the Range
winner is unbeaten by any other candidate, and the Range winner has
majority approval (approval cutoff must be specified, probably by
setting it as 50% Range), then the Range winner is elected. If there
is a candidate who beats the Range winner pairwise, there is a runoff
between that candidate and the Range winner. It should be noticed
that runoff elections generally test preference strength. If, indeed,
the Range winner was by significant preference strength, the Range
winner has a serious leg up on winning the runoff, even though the
pairwise winner would, as most election theorists look at it,
"automatically" win, an argument I've seen many times when this was
proposed. No, I don't think so. Differential turnout and voters who
reconsider could turn it the other way, and properly so.
----
Election-Methods mailing list - see http://electorama.com/em for list info
