[EM] ER-IRV(whole)

[Chris Benham]

James G-A,
You wrote (Thur.May 26):

>Chris, you write:
>  
>
>>>my example regarding ER-IRV(whole).
>>>45:Right=Left>CentreRight
>>>35:CentreRight>Right>Left
>>>20:Left>CentreRight>Right
>>>First-preference tallies
>>>Right:45       CentreRight:35      Left:65
>>>CentreRight has the lowest tally, and so is eliminated then Right wins. 
>>>This time no coordination was needed. As long as the Right suporters knew
>>>that Right had more first-prefernces than CentreRight, and a
>>>pairwise win against Left, then each individual Right supporter got an
>>>increased expectation by insincerely upranking Left from last to
>>>equal-first  with no risk.
>>>This would also work if the numbers 45/35/20 were replaced with 49/48/3. 
>>>I suggest the right numbers in your "paradoxical" row  should be IRV1,
>>>ER-IRV(fractional)2,
>>>ER-IRV(whole) 5!
>>>
>>    
>>
>	You may be right about this, but let's discuss a bit more. ER-IRV(whole)
>is a good topic that doesn't get explored enough.
>	It seems that the L voters can set things right by voting L=CR>R. Or, if
>some of the CR voters get mad at the R voters and truncate, L wins. It
>seems that the CR voters have little reason to rank R second, since they
>should know that R doesn't need their help to beat L. If there is even a
>hint of R voters betraying CR, I would expect the CR voters to bullet
>vote. 
>
In some real-life elections, for some voters there are more important 
issues than the strategising (or non-strategising) of  the supporters of 
rival candidates.
In rankings election the desirable norm should be that voters give a 
full sincere ranking. If  voters aren't interested in strategy, and they 
have a preference order  they wish to express,
then why shouldn't they give it?
Even if we grant the assumption that the voters are alert to strategy 
threats and are primed to "retaliate", there is no reason why there 
should be "even a hint of R voters betraying CR".
The strategy is low-risk and doesn't require any coordination (unlike 
Pushover in normal IRV), so individual R voters can do it on their own 
initiative.
In Australian parliamentary elections, most  (or at least very many) 
voters decide which party they are going to vote for (usually their 
sincere favourite) and then just take that party's
"how-to-vote card"  from one of the volunteers at the polling station 
and copy it.
So in the above scenario, the R's  "betrayal" of  CR could just take the 
form of springing these "surprising" how-to-vote cards on polling day. 
("Gee, there must have been a mix-up
at the printers!")
Of  course in Australian elections, voting is compulsory and in most IRV 
elections so is full ranking. But sometimes there is another reason why 
voters might want to rank fully.
A few years ago, there quickly arose a new far-right party that was 
widely regarded as racist; and they had a big profile in the media. I 
refer to the rise in Australia of the "One Nation"
party.  There was a campaign  for voters to "put One Nation last!" to 
make a statement rejecting racism.

>I meant a vulnerability score of 5 (now 6) to indicate a really critical
>vulnerability, like the compromising incentive in plurality or the burying
>vulnerability in margins. Do you really think that the paradoxical
>vulnerability in ER-IRV(whole) is at that level?
>
>
>
>  
>
I assumed that the scores were for comparing the vulnerability of 
methods along the rows, not comparing how big a problem the different 
vulnerabilities are relative to each other,
down the columns (within each method).
So I suggested the score of 5/5 because I  can't even envisage another 
plausible method being more vulnerable to this "paradoxical" strategy.  
No, I don't think the problem is as
bad as the Compromise incentive in FPP; but I do think its bad enough to 
disqualify  ER-IRV(Whole) as a serious proposal.

>More importantly (since I'm not even sure that I will cover ER-IRV(whole)
>in my little chart), do you think that the disadvantages of ER-IRV(whole)
>due to this issue outweigh its advantages over ER-IRV(fractional) in terms
>of reduced compromising-reversal incentive?
>
Yes! I regard ER-IRV(whole)  as much worse than plain IRV and that 
ER-IRV(fractional) is probably slightly worse than plain IRV (and 
certainy not worth the extra complexity).
Reducing incentive to compromise-reverse just by swapping it for 
incentive to compromise-compress  only seems to have any point if we 
assume that the voters notice these incentives
and are interested in strategising.
As I put it in a message to Forest, I think the "official assumption" 
should be that they aren't. (And of  course the compromise-reversal 
incentives in plain IRV are usually  much less powerful than they are in 
FPP).

Chris  Benham