[Election-Methods] Clone-related problems (was Re: Clone related problems in Range/Approval)

[Steve Eppley]

Hi,

I agree with Mr. Lomax that parties' main purpose is to coordinate 
campaigns, if he means coordinating the *votes* by assembling a 
coalition large enough either to win or to elect a "lesser evil" 
compromise that defeats a "greater evil."  Given traditional plurality 
rule or top two runoff or Instant Runoff or many other methods, parties 
fail to coordinate a large enough coalition if they nominate more than 
one candidate per office. (Borda has the reverse problem.  It would 
cause nomination of a huge number of inferior clones.)

However, I don't agree that parties would continue to nominate no more 
than one candidate if the voting method reduces spoiling so much that 
when spoiling occurs the winner is similar to the candidate who 
otherwise would have won.  This is what I believe the occasional 
spoiling would be like with a voting method that tends to elect centrist 
compromises, since it would create a strong incentive for candidates who 
want to win to take similar (centrist) positions on the issues.  
(Candidates would be concerned that if they take a minority position on 
an issue, some candidate X could take a majority-preferred position on 
that issue and identical positions on the other issues, so that a 
majority would prefer X.)

Mr. Lomax' concern that a huge increase in the cost of campaigning would 
result if a party nominates more than one candidate is legitimate but I 
think it's very speculative.  Here are some counterarguments.  First, by 
canceling or shortening the primary elections there would be an 
immediate savings of money. (Consider how much money Clinton, Obama and 
Edwards spent--and continue to spend--after the Democrat contenders were 
winnowed down to three.) 

Second, I believe there will be resonance in the voters' minds when 
multiple candidates advocate similar policies and make similar 
criticisms of other parties' candidates, even if the co-partisan 
candidates are limited by having to share a fixed amount of campaign 
money.  Imagine a debate pitting three against one, even if the one is 
granted time to repeat himself.

Another counterargument is that nominating multiple candidates makes it 
more expensive to attack them all (assuming they don't attack each other 
fratricidally).  Even if the cost of campaigning does increase 
significantly, it will be offset by the cost of attacking them all.

Parties could adopt rules to minimize fratricidal attacks.  The parties 
control a significant amount of campaign money (donations to the party 
rather than to individual candidates).  They could spend it on 
coordinated ads.  Or they could award it to attacked candidates and deny 
it to attacking candidates. (Perhaps the Democratic super-delegates 
could do something similar now to induce Clinton and Obama to be nicer 
to each other and campaign against McCain, by indicating they will vote 
for the one who campaigns most against McCain between now and the 
convention.)

I don't believe Mr. Lomax' example about how much money Clinton and 
Obama are spending against each other is strong.  Clinton and Obama are 
not running against McCain yet, they are running primarily against each 
other. (Sorry about the pun.)  Their short term goal is to be nominated. 

Mr. Lomax's point about reduction of negative campaigning against 
similar candidates appears to undermine his argument about Clinton & 
Obama, who are similar to each other.

If the voting method will tend to elect centrist compromise candidates, 
it won't matter if non-centrist parties nominate only one candidate.  
Politicians who want to win will position themselves as centrist 
compromises.  Furthermore, if parties nominate only one candidate, other 
candidates can run as independents and would choose to do so if they 
believed they would win or not worsen the outcome.  One example I've 
written about is the 2000 U.S. Presidential election, when McCain could 
have won running as an independent against Gore and Bush if the voting 
method had been better.  Most Gore supporters preferred McCain over 
Bush, most Bush supporters preferred McCain over Gore, and I believe 
enough voters (about 15%) preferred McCain over both Gore and Bush to 
make him the sincere Condorcet winner.  Rod Kiwiet of Caltech told me a 
few years ago he agreed McCain was the sincere Condorcet winner.

{Going off on a tangent here}  When Rod agreed about McCain, it was 
right after he'd conducted a pairwise poll that showed Arnold 
Schwarzenegger was sincere Condorcet winner in the election that 
recalled California Governor Gray Davis.  A goal of the poll was to 
check whether the public's political pairwise preferences really are 
transitive (IF A>B AND B>C THEN A>C) which Rod said had been assumed by 
political scientists but never checked.  {End of tangent}

There's an old adage about not putting all one's eggs in one basket.  
Much less information about the public's election day preferences is 
available during nomination season than closer to election day.  By 
nominating only one candidate, the party can fail to nominate the one 
who can win.

Finally, returning to the issue of the cost of campaigning and whether 
it could sharply limit the number of nominees, let me remind readers 
about some properties of the VPR family of voting methods: 

    Voting for a Published Ranking
    Prior to election day, each candidate publishes a ranking
    of the candidates.  On election day, each voter selects one
    candidate. (What could be simpler?)  Each vote is replaced
    by the ranking published by the voter's selected candidate.

Assume for this discussion that the algorithm VPR uses to tally the 
rankings doesn't suffer from Borda's awful "inferior clones" problem and 
that one of the following conditions is true:

    1. The algorithm elects within the top cycle.

    2. The votes are published, then each candidate may choose to
    withdraw from all the rankings before the rankings are tallied.
    (Candidates can withdraw to elect a compromise and defeat
    a greater evil.)

As I've written previously, I believe much less campaign money would be 
needed for good compromise candidates to win given VPR.  Their main goal 
would be persuading other candidates to rank them higher than those 
considered worse by these other candidates (or by these other 
candidates' "significant others": big donors & influential supporters).  
It's theoretically possible a good compromise candidate could win even 
if no one voted for her (just as it's theoretically possible a candidate 
can win given many other voting methods even if no one ranks or rates 
her topmost).  Surely it will be much less expensive to get no votes (or 
a few) than to get enough voters to vote a candidate high enough to win 
given a non-VPR method.

VPR would also mitigate Mr. Lomax' concern that too many candidates on 
the ballot would confuse too many voters.  He expects voters could 
competently select favorites when many candidates are on the ballot.  
Given VPR, selecting favorites would be more than sufficient.

Regards,
Steve
---------------------------------------
Abd ul-Rahman Lomax wrote:
> At 06:23 PM 4/13/2008, Juho wrote:
>   
>> Let's assume that the set of candidates consists of groups of clones.
>> For example there can be multiple parties and each of these parties
>> has multiple candidates. We further assume that typical voter
>> preferences are such that they prefer all their own party candidates
>> clearly over the other candidates (A1>A2>A3>>X>Y>...).
>>
>> The claim that I don't recall having seen before is that in Range and
>> Approval it makes sense to the parties not to nominate multiple
>> candidates.
>>     
>
> It also makes sense for them not to nominate octopi or beer steins.
>
> Parties exist, as a major reason, to coordinate campaigns. If a party 
> has three candidates running against one from another major party, it 
> may have to spend, for positive campaigning, three times as much. 
> There will also be additional inefficiency in negative campaigning, 
> because, to some degree, they may attack each other. Look what is 
> happening now with Clinton and Obama.
>
> The claim that advanced election methods, starting with IRV, will 
> reduce negative campaigning is pure fluff, there is no evidence at 
> all that this is true, and some contrary evidence from San Francisco. 
> It may reduce negative campaigning between similar candidates, to be sure.
>
> Having a huge number of candidates on the ballot, with a 
> single-winner system, is really quite confusing to voters. Sure, as 
> happens at present, voters may be able to pick a favorite and a 
> most-disliked, but what's in between may be more or less junk data. 
> This is why some Range advocates think that blanks should be 
> discarded, but I consider that totally impractical at this time, and 
> dangerous to boot. Blanks should probably be minimum rating, though 
> there is some argument for them being midrating (exact) or possibly 
> some other middle rating, perhaps calculated from all the other votes.
>
> I don't see Range being used to replace primaries. Rather, I see 
> Range primaries and then a Range general election.
>
> And I prefer, as well, Range systems that allow favorite designation, 
> with a real runoff if pairwise comparison (including the favorite 
> designation) shows that there is a candidate who beats the Range 
> winner pairwise. And this would not have so much of a clone problem. 
> People could give, for example, all the Republican candidates a 100% 
> if they really want to make sure that the party wins, as best they 
> can, then pick one favorite. So if there is a Republican win, they 
> have not just abstained from determining which one won. There are 
> similar systems that would do condorcet analysis on the Range votes, 
> and if a voter rated one Republican at 100 and another at 99, the 
> difference would be trivial in terms of electoral power; such a vote 
> would mean that a Republican win -- either one -- was greatly valued, 
> yet there remained a minor preference....
>