[Lnc-business] commentary on Scottish STV

[Alicia Mattson]

One thing that I found particularly annoying about the way that OpaVote
reports the results is that it does not report the actual number of votes
attributed to each candidate in each round.  Instead it shows us a percent,
though percentages have nothing to do with who is elected or eliminated.
If someone has a whole number of votes, those bars on the chart meet up
with the vertical lines, but otherwise it doesn't tell you that in Round 3,
Jeff Lyons had 1.8333 votes, and that Parr and Shade had 0.33333
votes...instead it says Lyons had 14.1% of the vote and Parr/Shade had
2.6%. What do the percentages matter when the goal is to get to 2 votes?

I was under the impression that with 13 ballots choosing 9 seats from 46
candidates, the "threshold" of 2 meant that only 6 could be elected.  So I
was surprised when the system listed 9 candidates elected.

On this OpaVote page of a "plain English" description of Scottish STV:
https://blog.opavote.com/2016/11/plain-english-explanation-of-scottish.html

It says things like, "All STV methods have a quota or winning threshold."
Once the "winning threshold" is determined to be 2, it sounds like one
needs at least 2 votes to be elected.  There was another quote that I don't
see now (maybe it was on the generic STV page rather than the Scottish
variant) which referred to repeating the process until there are X
candidates who have exceeded the threshold.

Now I'm noticing in the final paragraph of the Scottish STV page the
following, "If there are N positions being filled, then the election is
over when N candidates have reached the quota or when only N+1 candidates
remain (the winners being the N candidates with the highest numbers of
votes)."  So even the low threshold of 2 is not required for election.
Once there are only 10 candidates left, the top 9 are elected.

-Alicia


On Tue, Feb 26, 2019 at 3:52 PM Joe Bishop-Henchman <
joe.bishop-henchman at lp.org> wrote:

> I also just played the results Ms. Harlos sent out and came to the same
> fascinating (to me, at least) realization. Ranked choice voting works
> great if 800 people are choosing 5 or 9 spots; when it's 12 or 13 people
> choosing 9 spots, there are odd results. Especially if each voter is
> limited to nine choices, resulting in premature exhaustion of several
> ballots.
>
> Here are the final results with number of distributed votes:
> V. Sarwark - 2.00
> P. Bilyeu - 2.00
> Merced - 2.00
> Dasbach - 2.00
> Lyons - 1.91
> Kelly - 1.00
> Shade - 0.66
> Recuero - 0.50
> Myers - 0.50
>
> Forget majority - three of our final 9 were elected with less than one
> vote after fully distributing everything. As the report indicated, 37 of
> the 40 rounds broke a tie by random, including on the final round. With
> that many random choices, we could probably re-run the results and get a
> different final three members each time. (One reading of Scottish STV is
> that while it doesn't require a majority it still requires hitting the
> threshold of 2, and only four people did so.)
>
> I compared it to approval voting (the usual method we have used for
> filling vacancies), assuming (probably incorrectly) that everyone would
> have voted for the same people:
>
> Valerie Sarwark 12
> Paul Bilyeu     11
> Alex Merced     11
> Jeff Lyons      10
> Steve Dasbach   9
> Jennifer Moore  9 (not elected)
> Omar Recuero    7
> Fernando Davis  5 (not elected)
> Johnny Walker   5 (not elected)
>
> Tyler Bargenquast       4
> Brandon Bobbit  4
> Bryan Bombardier        4
> PJ Capelli      4
> Jennifer Flower 4
> Cecil Ince      4
> Kenny Kelly     4 (elected)
> Mayna Myers     4 (elected)
> Kevin Warmhold  4
> Marc Lazerow    3
> Kevin Moore     3
> James Olivi     3
> Fransisco Olvera        3
> Sean Parr       3
> Ashely Shade    3 (elected)
> Sharon Smith    3
>
> The starkest differences between the two voting methods are that J.
> Moore appeared on 9 of our 13 ballots, on all but one in the top six,
> but was not elected, while Shade (appearing on just three ballots as
> 4th, 5th, and 7th choices), was elected by random coin flip ahead of
> over a dozen other people who appeared on more ballots.
>
> I would suggest if we do this again in the future:
> * No limit on # of candidates to rank, to prevent premature ballot
> exhaustion.
> * In an extended ballot where we're all sharing results, those voting
> last have an incredible information advantage for tactical voting.
> Results should be secret until voting has closed.
> * If not requiring a majority, establishing a floor for # of votes to be
> elected. In the end, what we did was for all purposes no different from
> randomly picking 9 LNC members to each pick one person for the
> committee, except worse since three of those elected got less than 1
> vote.
>
>
>
> JBH
>
> ------------
> Joe Bishop-Henchman
> LNC Member (At-Large)
> joe.bishop-henchman at lp.org
> www.facebook.com/groups/189510455174837
>
> On 2019-02-22 04:32, Alicia Mattson via Lnc-business wrote:
> > I have a number of comments I wish to make about the Scottish STV
> > voting
> > system, but I don't have time to write it all up this evening.  I will
> > start with these comments and add more details to these thoughts later.
> >
> > It is one thing to read the theory and rules for a voting system on a
> > webpage, but it's another thing to print out the paper ballots and
> > enact
> > the process yourself to see the effects of each step.  Last night, I
> > took
> > the 7 ballots cast up to that point, printed them out on paper, and put
> > them in piles on the floor to manually experience how it works.
> >
> > It didn't take long before my eyes got wide.  Then a little later my
> > jaw
> > dropped as I realized more and more implications of the process.  When
> > I
> > was done, I paced around the living room in a bit of a rant as I put my
> > realizations into words.
> >
> > Clearly, this voting system was envisioned for situations in which the
> > number of ballots being cast VASTLY outnumbers (by orders of magnitude)
> > both the number of seats being filled and also the number of
> > candidates.
> >
> > In our case, however, if everyone had voted it would have been 17
> > ballots
> > cast to choose from more than 40 candidates to fill 9 seats.  I only
> > saw
> > ballots from 12 people on one election and 13 on the other, making the
> > ratios even worse.  This seems to be in the range of
> > worst-case-scenario
> > for this voting system.
> >
> > With these ratios, the process devolves into essentially a casino game
> > of
> > chance necessitating random candidate eliminations early in the
> > process.
> > After the first round of vote distributions, we might as well just tell
> > the
> > candidates to play Russian roulette...or for fewer dead bodies we could
> > just draw names out of a hat.
> >
> > If I correctly understand the process, then it's mathematically
> > impossible
> > for the number of ballots cast in these two elections to elect more
> > than 6
> > candidates to each committee.  Looking at the vote distribution, we'll
> > elect at least 3 but no more than 6, depending on the outcome of some
> > random selections, so we're in for re-balloting anyway.
> >
> > More specifics later...
> >
> > -Alicia
>