[EM] bullet voting and strategy on Approval ballots.

[robert bristow-johnson]

On Aug 28, 2010, at 9:37 PM, Warren Smith wrote:

> --it is hard for me even to follow this stream-of-consciousness
> postmodern literature type discussion, much less reply to it.


okay, i'll translate it:

>> and i don't quite get whether a bunch of people bullet voted in the
> Burlington IRV elections has much to do with the Approval or Score.

some people chose to bullet vote on the ranked-order ballot in the two  
IRV elections we had in Burlington VT.  that bullet voting was not  
necessary (or should not have been necessary) in order to prevent harm  
to their favorite candidate, assuming that was the candidate they  
bulleted.

on the other hand, Approval and Score *does* potentially harm your  
favorite unless you bullet your favorite.  if a large component of  
people bullet vote their favorite in Approval or Score, the outcome of  
the election is not any different than it would be for Plurality.

> all's i think about that is either they really only preferred one
> candidate and all of the rest were equally bad.  or, out of spite,
> they were saying to themselves "I hate IRV.  I'm gonna vote only for
> my guy, goddammit."

so i contend that in Burlington's two IRV elections, the bullet voters  
had one of two motivations for bullet voting.  one is that they didn't  
like any other candidate other than their favorite, so the bullet vote  
was a sincere vote.  the other is that, because they hated the idea of  
a different ballot than their traditional vote-for-one ballot, even if  
they had a 2nd or 3rd preference, out of spite they did not mark such  
lower preferences on their ballot.

>   "...Then I'm gonna bitch about how my vote didn't
> count if my candidate is eliminated before the final round."

this is me politiking.  i am bitching about these "Keep Voting Simple"  
advocates that claimed they were disenfranchised from the election  
because their vote had no effect in the IRV final round.  but they  
*chose* not to mark a preference between the candidates that ended up  
in the final round.

>  it was,
> to me, interesting about how some of the most rabid IRV opponents here
> in Burlington were prog-hating Democrats.  those bullet voters are the
> only ones (among the big 3) who can claim that they were left outa the
> final round.

so, considering the top three (like you do on your Burlington web  
page), it's only the bullet voters that bulleted Montroll (the Dem)  
that had their ballots "exhausted" before the final round and they  
expressed no vote in selecting or rejected the person that was  
eventually elected mayor.  if they bullted Prog or GOP, their vote was  
in the final round.


>  it's a bogus argument, in my opinion.  they could have
> thought about it and made up their minds about the Prog v. GOP by
> election day, rather than demand the convenience of not having to make
> that decision until 3 weeks later.

these anti-IRV people in Burlington had no clue about the anomalies or  
pathologies you've identified in the 2009 election, Warren.  their  
problem is that IRV (and any other ranked-choice ballot with no two- 
round runoff) requires them to consider candidates other than their  
favorite and to come to a conclusion about them by the initial  
Election Day.  but they want the "convenience" of not having to decide  
about their contingency candidates if they don't have to.  they think  
that it is their God-given right to just vote for their favorite and,  
if their favorite did poorly, then decide later about their second  
choice.  they call themselves "One person, one vote" but they are  
really "One person, two votes".

so i hope my translation of my stream-of-consciousness made it more  
clear.

>
> *If you are an IRV proponent: your arguments (if any) that IRV is
> better than approval voting because with the latter voters will
> "bullet vote" ... are bogus.

i am not saying that IRV is necessarily better than Score or  
Approval.  i am saying that the Ranked-Choice ballot (which is used by  
IRV or Condorcet) is better than the ballot for Score or Approval.  it  
collects exactly the right amount of information from the voter.   
Score requires too much information from the voter and Approval  
requires too little (and makes it impossible for the voter to express  
his favorite candidate without bullet voting that favorite).

>  The evidence does not support you.

the only evidence i am offering is the ballots themselves.  for  
Approval, if you mark your favorite "approved", how do you also  
approve of another candidate (who is preferred less than your fav)  
without potentially harming your favorite?

> FairVote did not produce any evidence. In fact if anything the
> evidence I compiled refutes that.

you're not refuting the basic problem.  any "evidence" you collect  
(about voter regret) does not refute the basic problem.

> *If you couldn't care less about IRV, but are for some reason  
> worried about the
> bullet-voting bugaboo for approval/range voting: well...  now you know
> there is no evidence said bugaboo is worse with approval/range voting
> than it is with IRV.

with the ranked-order ballot (which happens to be used for IRV, though  
IRV is a lousy method to count it), at least the voter can express  
that they like Candidate A better than Candidate B, even though he/she  
approves of both.  you cannot do that with Approval, and that leads  
the savvy voter directly into strategic thinking about how to mark the  
ballot.  and i am experiencing this first-hand in the Vermont State  
Senate race in the Chittenden Senate district where we vote for up to  
6 candidates (a binary vote) and the top 6 vote-getters are elected.   
i would likely "approve" of many more than i am voting for (1 or 2),  
but i don't vote for them because i do not want to harm the 1 or 2  
that i *am* voting for.

> *If you are babbling about "spite" and "prog-hating Democrats" and  
> so forth
> in a mysterious manner:   well, why the heck should anybody care?

just trying to give you some context for why IRV got repealed here.   
if you think that the level of sophistication of the argument rose to  
the level of your analysis on your Burlington page, i regret to inform  
you that it did not.

>  I don't.
> Maybe they are bullet voting out of "spite." Or because they've
> "seen god."  Or they are "tired of it all."  Or they're "rabid." Or
> they have a "weird animus." Who knows?  I DON'T CARE WHY THEY ARE
> DOING IT.  I offered no hypotheses on that question.  I do not accuse
> them of rabies or anything else.
> I merely compiled the data,

*what* data, exactly.  you do not have data from real elections under  
two cases: one was IRV (or Condorcet or whatever) and the other as  
Approval.

we can compare IRV to Condorcet exactly (if we have records of each  
ballot, which we do), because the ballots for the two systems are  
identical.  it is not a stretch to assume that, whether it was IRV or  
Condorcet, that the voters would have marked their ballots identically  
in both cases.  so the two systems can be directly compared.  we can,  
with that last IRV election, speculate as to what would happen if it  
was FPTP and TRR (which is the law we reverted to) by assuming that it  
is always the 1st-choice vote in the ranked ballot that would become  
the single vote in the traditional FPTP ballot.

but we cannot map these ranked ballots or the traditional vote-for-one  
ballot to a sincerely filled out Approval ballot.  we do not know,  
without making even bigger assumptions, how these voters would have  
marked their Approval ballots for comparison.

you can run simulations that sorta guess how these voters might mark  
the Approval ballot, but that's not data the proves anything.

> and found out there was less bullet voting
> with approval than with comparable IRV elections.

so where did you make the side-by-side comparison?  in what election  
did they issue two ballots to the same voters (one IRV and one  
Approval) and tell them to fill out the two for some academic study to  
compare?

> I therefore
> conclude, that any claim there is something specially worrying about
> approval voting which makes it vulnerable to bullets... is baseless.

i stated the basis on my own experience, what the motivation was.  i  
have also asked several (Democrat) voters about what they do for the  
"vote for 6" state senate race in our large district.  i have never  
heard from *anyone* that they vote for all 6.  almost everyone votes  
for one or two candidates. that's bullet voting.  and it's a voting  
strategy.  and it's a voting strategy that has a purpose (which is not  
to harm your favorite candidate).

> If you want to dispute me on that, here is my suggestion:
> find some evidence.

i have no idea of where to run a real-life simulation where voters  
fill out two ballots for the same election (but only one is counted).   
how can we get them to do that?  they will just toss the ballot that  
doesn't count.  but unless you have some rhyme or reason for mapping a  
ranked ballot or a simple "vote-for-one" ballot to Approval, i cannot  
see how you can make any comparison.

i *have* made a comparison of Approval to the "vote-for-six" ballot we  
have in our state senate district, because i know of no person who was  
limited by that six vote limit.  if almost no one gets to the limit,  
that vote-for-six ballot is effectively the same as Approval.  and it  
is clear that there is strategy employed (namely bullet voting 1 or  
maybe 2 candidates) with such a ballot.  that's a fact. that is what  
people are doing, Warren.

> I'm working on finding some more evidence myself, and see no reason
> you should not help.


other than the experience that i have had, and the above discussions  
with others in this location (in Burlington VT, we have experienced  
*both* the ranked-choice ballot and virtual approval ballot, i have  
never lived anywhere else where i have seen either).

--

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."