[EM] piling on against IRV

[robert bristow-johnson]

On May 8, 2010, at 9:50 AM, Terry Bouricius wrote:

>
> A primary concern we have come across consistently from average  
> voters when presenting election method reforms, is the concern about  
> Later-No-Harm. People have some vague concept of Borda rules, and  
> ask whether ranking second choices under IRV might hurt their  
> favorite candidate. Our experience has shown that this is just about  
> the ONLY election method criterion that spontaneously occurs to  
> average voters when learning about new methods. It may not be  
> ultimately the most important criterion, but methods that violate  
> Later-No-Harm seem to have no hope of adoption, and FairVote isn't  
> going to devote resources to futile reforms.

so Terry, in 2009 Burlington, did 495 of 1513 voters that marked their  
ballots as W>M>K (perhaps with some other candidates in between)  
experience LNH?  not so much to their first choice, but to their 2nd  
choice?  how is that different from a Nader voter in 2000 who might  
tolerate President Gore who actually caused W's election?  Keri told  
us in that debate that such a thing couldn't happen with IRV, did she  
actually do her research?

> When asked about other governmental election method reform ideas I  
> mention pros and cons of each and express why I believe some other  
> systems are not achievable (often simply due to the fact that there  
> are no examples of any other governments using the methods, making  
> officials highly reluctant to be the first guinea pig), or are not  
> as good, in my opinion, as STV. In fact I often raise Condorcet as a  
> reasonable method (though I also state why I think it is  
> unachievable in America at present). As an example, in this video   
> of Rep. Mark Larson speaking on IRV to a workshop at the 2009  
> Democracy Fest, http://www.cctv.org/watch-tv/programs/instant-runoff-voting-0 
>  at around 14:00 minutes, you will note that I was the one who  
> brought up the Condorcet concept (Vanilla ice cream was the apparent  
> Condorcet winner of the just completed IRV demonstration). I support  
> Condorcet as preferable to plurality, and have even helped draft  
> Condorcet legislation.
>
> But experience has convinced me that Condorcet is not a reform that  
> can be won.

i don't think anywhere close to a comparable effort has been given to  
that reform.  even with your help (and i know you were trying to lay  
low because you were most often cited personally for bringing IRV to  
Burlington), i felt pretty alone.

> FairVote is seeking to improve election rules NOW, and IRV and PR- 
> STV appear to be the only positive voting method reforms that have a  
> real prospect for enactment in the foreseeable future. The key  
> factors leading to this conclusion are:
> 1. These reforms have been used for governmental elections for over  
> 100 years.
> 2. These reforms have been adopted, and are being used in the U.S.
> 3. IRV has an analog (two-round runoff) already used in America,  
> with which voters are familiar and comfortable (that is elimination  
> of bottom candidates is deemed appropriate by most voters).
> As an example, the commission set up in Colorado to examine  
> alternative voting methods had a lengthy presentation from Range  
> Voting supporters, but they rejected it and favored the single  
> transferable vote algorithm.

the Range/Approval guys just don't seem to get the *inherent* pressure  
to tactical voting that either Range or Approval causes.  i dunno why  
they can't see it, it's obvious to me.  anyone who wants to maximize  
their electoral influence (and who that bothers to vote doesn't?) will  
push the score of the candidate they really support to the maximum or  
approve only that candidate (lest they worry about harming that  
candidate by scoring or approving other candidates) and the method  
will devolve into FPTP with a scaling factor.  it obviously will cause  
voters to agonize over exactly what tactics they bring to the polls.   
i respect Warren's analysis of the 2009 Burlington election, but he  
hasn't gotten me past square 1 in his selling of Range or Approval.  i  
just don't buy the premises.  simply because of the nature of the  
ballot, it causes tactical concern for any voter who is thinking with  
any depth in an election with multiple viable candidates.

the existing establishment of STV in many jurisdictions is not much of  
an asset, because it hasn't always worked so well in those 100 years.   
IRV is like selling a not-very-good upgrade to a crappy operating  
system.  people are dealing with all the frustration of MS-DOS for  
years and Billy Gates then comes with his bloated and buggy Windoze as  
the solution to all of our MS-DOS woes, some adopt it and find out  
later that it also is a POS.  the problem with FairVotes promotion of  
IRV is that the failures of IRV will sully, by association, the ranked  
ballot because too many dumb people cannot (because of FairVote's  
promotion) differentiate between the ranked ballot and the STV method  
of tabulating it.  that was the big mistake from the very beginning.

> Of course, IRV and STV can also be repealed (and have been in some  
> jurisdictions).

as we have found out.

> However, it is important to understand why. Some people have  
> suggested that IRV's failure to elect the Condorcet candidate caused  
> its repeal in Burlington.

i suggest that it was used as an excuse, a ruse, to discredit the  
adoption of IRV over the FPTP/TRR that existed before.  we both know  
that it was completely disingenuous for the IRV opponents to point to  
the 2009 election and accuse it of "thwarted majority" because the  
method they were dragging us back to would not elect the unambiguous  
majority winner either.  if fact the method they took us back to  
totally obscures that information because the "Vote for One" ballot  
will not even collect sufficient information for us to even know who  
had the majority support over any other candidate.

> That is not correct. The repeal had everything to do with the  
> support for the Republican plurality winner, a heavy Republican- 
> leaning turnout in March,

there were *many* Dems that were against IRV in Burlington (mostly  
also from Wards 4 & 7).  some i worked side-by-side with in Howard  
Dean's prez campaign.  if it were not for those Dems, IRV would have  
survived this year (since it was so close).  and it was also because  
of such poor turnout from the other more liberal wards.

>  combined with the anti-Mayor Bob Kiss sentiment resulting from a  
> local financial scandal. If Bob Kiss had been the Condorcet winner,  
> there STILL would have been the same "Keep Voting Simple" repeal.

not likely.  in 2006 he was the CW (and also the plurality winner).   
there wasn't as much of a fuss then (but there was some).

*but* if Condorcet was the law and the "3rd place" candidate (who was  
the CW), i believe also there would have been a big fuss.  but they  
couldn't have disingenuously used the "Thwarted Majority" critique as  
they did this year.  and far fewer Democrats would have jumped on the  
"Keep Voting Simple" bandwagon.  i don't think the repeal effort would  
have succeeded since it was already so close.

> Any method that does not affirm the traditional plurality winner as  
> the "rightful" winner (all of the methods advocated on the Election  
> Methods list) would be subject to the same repeal risk.

that is true.  but when IRV fails to elect the CW, we know there are  
other pathologies that pop up because of that.  it's *because* in 2009  
the CW was kept out of the final round that the IIA effect ("Spoiler- 
lite"), LNH-lite effect (which punishes non-tactical voting), non- 
monotonicity effect all occurred in that election.  if it was  
Condorcet in Burlington in 2009, they couldn't have made any of those  
arguments nor the precinct non-summable criticism.  *all* they could  
argue was that their plurality (of 1st choice votes) winner was not  
elected.

so if we're gonna improve the operating system, let's really improve  
it.  because if we bring a so-called improvement and it causes the  
computer to crash, people will want to go back to their tried-and-true  
command-line DOS and be very wary of later proposals to improve it,  
even if the later proposals really *do* improve it.

> I would hope supporters of other reforms would try to enact their  
> preferred reform in non-governmental organizations and municipal  
> government elections anywhere in the U.S. that is not already  
> engaged in IRV or STV reform, so we can get some real world  
> experience with them.

i do too.

> FairVote will absolutely NOT seek to undercut such reform efforts.

but, by promoting IRV as synonymous with Preferential Voting  
(essentially as the only manner a government can tabulate the ranked- 
order ballot) and denying the carnage when it crashes, FairVote ends  
up undercutting all reform efforts.

>
> An absolutist definition of "majority" is never used by anybody,  
> when discussing elections in the U.S. So when we use the term  
> "majority" we do not mean a majority of people in a jurisdiction,  
> nor a majority of the voting age population, nor a majority of  
> registered voters, nor a majority of voters who went to the polls,  
> nor a majority of voters that includes those who skipped a contest  
> in the first round, nor a majority of voters that includes those who  
> skipped the final contest, by staying home in the case of a separate  
> runoff, or ranking neither finalist in the case of IRV. Just as in a  
> traditional runoff, we mean a majority of those who expressed an  
> opinion in the contest between the two finalists.

remember, the problem (with IRV) is *who* gets to be those two  
finalists.  there is a problem with saying that Bob Kiss is *the*  
majority winner when, in fact, a majority of voters agreed that Andy  
Montroll was the preferred candidate over Kiss.  but, because there  
was no cycle, the same cannot be said about Montroll.  in every case,  
in every pairing, Montroll is the majority preferred candidate.  there  
is no sense where Montroll is *not* the majority.  this lack of  
ambiguity cannot, of course, be claimed if there is a cycle.  that,  
and only that, is the weakness of Condorcet.  but if there is no  
cycle, there is *no* sense that anyone other than the CW can make an  
unambiguous claim to the majority mandate.

> Runoffs always exclude from the denominator children, non- 
> registered, non-participating voters, and voters who abstain from  
> the question of which of the two finalists is better (or less  
> intolerable).  Condorcet enthusiast Robert Bristow-Johnson  
> (unintentionally) paraphrased this concept of two-choice majority in  
> his recent email, when he wrote: "between any two candidates, there  
> is always a majority, unless they tie." The only difference is  
> whether one accepts that under a traditional runoff system and IRV  
> the final round is, in fact, a contest between "two candidates."

the problem is if it's the *correct* two candidates.  Condorcet gives  
*every* pair of candidates an equal standing whereas IRV will ignore  
some defeats in a similar (but less flagrant) manner that FPTP ignores  
some salient defeats.  *every* pairwise defeat should be considered,  
not just some, not just one.

> Both traditional runoffs and IRV rely on this concept of narrowing  
> the field to two finalists,

and that concept, while "traditional" has always been flawed, for TRR  
or for IRV.  the *whole* problem with runoffs, whether they're  
"traditional" or instant, is the problem of determining who belongs in  
the runoff.  saying it's the top two 1st-choice vote getters is a  
contrived criterion.  more information of voter preference needs to be  
collected.

> for the purpose of finding a majority winner. This is not the only  
> way to achieve this, but a standard way,

"standard" like Microsoft software or OS is the de-facto "standard".   
at some point, we have to stop justifying the use of the SUBstandard,  
just because "everybody else is using it".

> though I would not dispute that Condorcet can ALSO find a majority  
> winner in most situations.

not "ALSO".  that is the falsehood, Terry.  if there is a CW and IRV  
(or any other method) misses in electing that CW, then Condorcet  
*exclusively* finds the unambiguous majority winner.  Bob Kiss was not  
the majority winner even though a majority of voters preferred him  
over Kurt Wright.  A majority of voters also preferred Andy Montroll  
over Kurt Wright, so Bob and Andy stand on equal ground in that manner  
(actually they don't since Andy's preference over Wright was more than  
3 times greater than Bob's).  So then the question becomes of course  
who, between Montroll and Kiss, did the majority of voters prefer?   
Montroll exclusively had the *unambiguous* majority standing with  
voter preference.

> In Burlington, there was a separate runoff election in 2009 in a  
> city council race (IRV only applied to mayoral races), in which the  
> "majority winner" of the runoff got fewer votes in the runoff than  
> the "loser" got in the first round, due to a substantial drop-off in  
> participation (abstention).

that was my ward, too.  the first-round majority winner was only 6  
votes ahead of the "loser" who eventually became the winner in the  
runoff.

i'm getting tired, Terry.  i'll bow out at this point.

--

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."