[EM] IRV vs Plurality

[Dave Ketchum]

On Jan 11, 2010, at 4:16 PM, robert bristow-johnson wrote:
> On Jan 11, 2010, at 1:19 PM, Dave Ketchum wrote:
>> On Jan 11, 2010, at 11:45 AM, robert bristow-johnson wrote:
>>> On Jan 11, 2010, at 8:54 AM, Kathy Dopp wrote:
>>>> ...
>>>> Plurality is far better than IRV for many many reasons including:
>>>>
>>>> 1. preserves the right to cast a vote that always positively  
>>>> affects
>>>> the chances of winning of the candidate one votes for
>>>
>>> and more often than not, *hurts* the credible candidate that is  
>>> politically more aligned with the candidate one votes for.
>>>
>>> 50,000 voting for Nader elected W in 2000.  that is a matter of  
>>> fact.
>>
>> Plurality does that only when you vote for one who has a  
>> possibility of winning.
>
> what do you mean.  Florida voters that voted for Nader (who had no  
> chance of winning) were far more likely, if they would come to the  
> polls at all, to vote for Gore over Bush if that was the choice  
> presented to them.  *Gore* had a possibility of winning (and many  
> say that he won, according to the law) and needed 530 more votes in  
> Florida and the history of the first decade in the 21st century  
> would have turned out far different.  there were thousands of Nader  
> voters in Florida and the fact that the election was decided by  
> plurality (and that W managed to prevent the rest of the Florida law  
> from taking effect) meant that the secondary choice of the great  
> majority of these Nader voters was harmed by their primary vote (and  
> only vote) for Nader.

Going thru this carefully:

Kathy said "positively affects".

r b-j said something, which I chose to ignore.

I responded to Kathy, pointing out that, since there was no chance of  
there being enough FL votes for Nader for him to win, voting for him  
did not affect his chances of winning - in other words there is no  
expectable positive effect in voting for such candidates as to their  
chance of winning.

Then your talk of voting for Nader affecting other candidates chances  
is outside the topic Kathy started for this.
>
>>  Sometimes doing that prevents voting for the one you prefer but  
>> expect to lose.
>
> yes.  sometimes that is the case and one must accept that he/she is  
> in the minority.
>
>>>> 2. allows all voters the right to participate in the final counting
>>>> round in the case of top two runoff or primary/general elections
>>>
>>> but IRV does that in an instantaneous way UNLESS some voter  
>>> changes their mind about their alternative candidate.  IRV or  
>>> Condorcet (or any ranked ballot) requires the voter to choose  
>>> *and* *commit* to not just their favorite, but their fallback  
>>> candidate on the same Election Day.
>>
>> With Condorcet the voting is all done on one election day, and all  
>> that the voters rank are considered in the counting.
>>
>> While the candidates and voters must do their preparation before  
>> that one act of voting, that single voting round should be all that  
>> is needed for the counting and deciding on winner.
>>
>> Note that primaries may be used, but there is no need for them such  
>> as is true for plurality - multiple candidates for a party can be  
>> voted for in a Condorcet general election, with voters ranking such  
>> candidates if and when they choose.
>
> but, between the lines here, the opponents of IRV (or any ranked- 
> order ballot) want the Progs and Dems to together field a single  
> candidate (like there would have to be a Prog/Dem nomination caucus  
> at Burlington High School and *one* candidate comes out of that).   
> that, in my opinion, is an insult to both the Progs and the Dems.   
> they are different parties, they can field their own different  
> candidates (or, if a single person is most preferred by Progs and  
> Dems and is nominated by both town party caucuses, then that  
> candidate can still just register once at City Hall).  (in New York  
> state, a single candidate may run under multiple party banners and  
> that state accumulates votes for that single person, even though  
> they are cast under different columns.)   i dunno what Kathy Dopp  
> thinks of NYS.
>
> but, it doesn't matter.  even if the Prog and Dem are different  
> candidates, the political alignment between the two parties is  
> undeniable.  i believe that Burlington voters (and we'll see if this  
> is still the case in March) do not want a minority GOP candidate to  
> win solely because the liberals in Burlington split their vote.  the  
> anti-IRVers insist that this is the price we must pay for having two  
> separate liberal parties.

In NY, or any other place doing Plurality, groups such as liberals  
have a difficult challenge in preparing for election:
      Get together and give a single candidate a better chance of  
winning.
      Divide up the group's votes among multiple candidates - likely  
causing all such to lose.

With IRV the multiple candidates are a bit less of a disaster since a  
voter can vote for more than one.  Still trouble since the counters  
only look at one of such candidates on any ballot at any instant.

With Condorcet as many of the multiple candidates as desired can be  
voted for by any voter, with the same or different ranks, and all be  
seen by the counters at once.
>
>>>> 3. preserves voters' right to understandably verify the election
>>>> outcomes because the counting is simple enough for them to do,
>>>> precinct summable
>>>
>>> so does Condorcet.
>>
>> And Condorcet gives a more accurate view since the ballots more  
>> completely state voters desires and all that they say gets counted.
>
> the ranked-order ballot gets just the right information from the  
> anonymous voter.  Approval and traditional FPTP do not ask enough  
> questions and Range demands too much information from the voter.
>
> the ranked ballot in IRV is not the problem with IRV.  it's the way  
> IRV interprets the ballot info, counts the votes, and declares the  
> winner that is the problem.

That IRV does not permit candidates to share a rank is a minor  
annoyance.
>
>>>> 4. preserves the right for local precinct control of the counts  
>>>> or in
>>>> the case of election contests that cross county lines, local county
>>>> control of the counting process
>>>
>>> so does Condorcet.    i like precinct summable too, but it isn't  
>>> the end-all requirement for an honest election.
>>>
>>>> 5. is far less costly than the IRV counting process
>>>
>>> not in Burlington.  once the infrastructure was set up (the ballot  
>>> scanning machines didn't have to be changed at all, the difference  
>>> is that the precinct results (that had a record for how each  
>>> ballot looked) were transferred to city hall and a computer did  
>>> the rest.  because of FoI laws, this record is available for  
>>> public scrutiny and has been scrutinized.
>>
>> Topic seems to be that a second look at a ballot is required in IRV  
>> after it is determined that the top rank lost.  In Condorcet all  
>> the looking is done in one pass.
>
> well, no.  in Burlington the ballots are physically scanned once.   
> the data for each individual ballot is parsed and re-examined for  
> each IRV round.  for Condorcet, they would be scanned once and you  
> *could* write the code to parse each ballot once and rack up  
> pairwise totals for all candidate pairs, one ballot at a time.  but  
> you also could parse the whole ballot data (all of the ballots)  
> again and again for each candidate pair.  it's just a matter of how  
> one likes to write their code, that wouldn't matter in any case.   
> the computer might not report results until all of the parsing is  
> done (and it has the result), or it might do it as the parsing is  
> done so some people can have some election-night excitement watching  
> the totals increase (like watching a race).  but, for a town the  
> size of Burlington VT (the largest town in the state, but VT ranks  
> dead last in the size of the largest town in the state) it should  
> take a second or two for a modern PC to do the whole thing no matter  
> if it's one pass for the whole thing, or one pass for each candidate  
> pair.

What matters is effect, not how smart programmers are as to details.   
The first look at an IRV ballot is concerned only with a voter's top  
rank.  Later looks skip over losers and respond with remaining top  
rank.  This looking costs, though perhaps not much as described above.

The possible excitement tangles with the secrecy laws - reporting in a  
manner that identifies how ANY ONE voter voted needs preventing  
(needed protection of voters).

Dave Ketchum
>
> --
>
> r b-j                  rbj at audioimagination.com
>
> "Imagination is more important than knowledge."