I’m not persuaded of approval voting. My guess is that voters will
bullet-vote for the candidate they like best among those they know
about; candidates will encourage bullet voting on their own behalf;
pundits will have nothing better to say, and voters will have no motive
to award more than the minimum number of approvals.
   A danger of quadratic and entropic measures is that they don't
impose the constraint that the number of survivors has to be small
enough to make ranked voting effective on the second round.
   CJC

On 25/08/2023 15:05, Kristofer Munsterhjelm wrote:

Colin,

Your concern about approval voting is corroborated, at least anecdotally,
by evidence from Fargo, ND -- the only US jurisdiction to use approval in
general elections. In 2022, 60%
https://citizendata.com/tracking-voter-support-for-electoral-reforms/ of
voters reported voting for just one candidate in the field of seven. At
least two candidates (at least according to quotes in an article
https://democracysos.substack.com/p/what-fargo-reveals-about-approvalfrom
Rob Ritchie) encouraged their supporters to bullet vote for them.

MJG

On Fri, Aug 25, 2023 at 4:52 PM Colin Champion <
colin.champion@routemaster.app> wrote:

On 8/26/23 01:23, Michael Garman wrote:

Given that we're talking about approval as a first stage of a multistage
method, the more relevant comparison would probably be approval plus
runoff, like in St. Louis. I don't know the analogous stats for the
primary there, however.

I don't think approval would be good enough for a single-winner election
(Burr dilemma and incommensurability) as mentioned before. But choosing
five finalists is different matter than choosing a single winner.

-km

Yes, the rule of thumb I quoted should be thought of as a quick and dirty
approximation to 2^entropy or e^entropy, depending on whether entropy is
measured in bits or nats.

Here's another related rule whose relevance and convenience is more obvious:

Let n be floor of the quotient of the total number of ballots and the
cumulative support of the cumulative vote winner. [Cumulative votes are
like approval votes, except they are counted fractionally instead of whole.
]

This choice of n makes possible a simple version of STV for selecting the n
runoff finalists.

Let Q be the number of ballots divided by n. The largest faction will be no
larger than this quota, but after eliminations and vote transfers all n
remaining candidate factions will be within one transferred  vote of Q in
size.

Before continuing he pairwise wins and losses are determined once and for
all ... with all ties resolved.

The pairwise loser of the two smallest faction candidates, has her first
place votes transferred respecting their ballot preferences as much as
possible ...consistent with the upper limit imposed by the quota Q.

Once a candidate reaches Q votes, his stash of votes is frozen.

Keep going in this way until exactly n factions remain.

The candidates that correspond to these remaining factions are the
finalists to participate in the runoff ... a runoff that will have its own
method and new ballots.

I recommend for the initial narrowing stage (that we already detailed) an
option of VPR (Vote for a Published Ranking) for those voters that find it
too tedious to generate their own ranking or partial ranking.

I offer this in the spirit of brainstorming ..
not intending a finished product!

fws

On Fri, Aug 25, 2023, 7:06 AM Kristofer Munsterhjelm km_elmet@t-online.de
wrote:

On 8/25/23 22:51, Colin Champion wrote:

Here's a funny idea that I probably wouldn't propose seriously. Each
voter is assigned a random candidate pair in advance and asked to
determine which is better. This "sampled Condorcet matrix" is then used
to determine the finalists.

It probably wouldn't work because the voters wouldn't want to be told
what contest to focus on, and the error (what pairs are noisy) wouldn't
be uncorrelated as some contests would be considered more important than
others.

But I just got to think about it because, if any method is to be easier
than ranking, it has to ask less of the voter than ranking (or the
product of data required times people asked should be lower). Approval
obviously does ask less, as does your "pre-established candidate order"
idea. As does this, and sortition-based hybrids (fewer voters).

So even though that idea may be "thinking too far outside the box", the
guideline/heuristic of thinking about how to reduce the information that
needs to be gathered may be a good one for finding methods that do work.

-km

If the voters bullet, no problem ... they should not approve anybody they
don't feel like supporting. Approval frees the voters to approve more than
one if they want ... a freedom they don't have under Plurality FPTP.

But it would be amazing if the result of bullet voting was anywhere near a
uniform distribution...the only case that would make the entropy large
enough to impede the narrowing of the candidate fold.

This level of entropy is so unlikely in this voting context that it is not
worth worrying about.

Think about the election resulting in Arnold's first gubernatorial win in
California. It was FPTP. Yet most of the candidates fell far short of the
front runner averages.

On Fri, Aug 25, 2023, 1:52 PM Colin Champion colin.champion@routemaster.app
wrote:

If you want optimal strategy, vote for your Plurality compromise and
everybody you like better.

So Approval improves the result when people apply the "everybody you like
better" clause, and leaves it no worse (nor better) until they get more
familiar with the method.

On Fri, Aug 25, 2023, 4:23 PM Michael Garman michael.garman@rankthevote.us
wrote:

I have entertained the idea of ranking only few first candidates as a means to reduce the amount of information that needs to be gathered. At least in Finland people are used to voting by writing the number of their favourite candidate inside a big circle in the (cardboard) ballot paper (that does not contain much more than that single circle). It would be easy to add few more circles in the ballot paper, and people would quickly understand that it is ok to write the numbers of few favourites in it, rather than only one.

It is important to keep the participation level high in the elections, and therefore it is important to make voting as easy as possible. Allowing only few top candidates to be ranked of course reduces information, and thereby makes the results less accurate, but that may sometimes be a cheap price to pay, depending on the preferences of the voters, and on what kind of voting practices they are used to and feel comfortable with.

In some elections, like in two-party countries where there are two clear frontrunners, already ability to rank two candidates could be enough. That could mean ranking your true favourite and one of the frontrunners. But in general, I was thinking of something like allowing three to five top candidates to be ranked. The ideal number of ranked candidates may depend on the total number of candidates, on how many of them have some chances to win etc.

You could allow enthusiastic voters to rank even more candidates, but maybe it is ok to set a hard limit also for them, just to keep the ballots simple, and in order not to give the impression that you need to do a lot of work to cast a "full vote".

I note that voters may see this kind of voting also as "approval voting" since they are able to point out few candidates that they "approve", and leave others ranked at the bottom. There is no need to change the method though. For example regular Condorcet is fine. Being ranked at the bottom by default works well enough as a "disapproval" of a candidate.

BR, Juho

Juho,

It's good to hear from you and to get your insight from practical
experience in Finland.

So maybe voters could draw a circle and put dots in the circle to represent
the candidates worth the trouble of including in the circle ... with better
candidates near the center.

An optical scanner with the right software could rank the dots according to
their distance from the center, etc...

... or something like that!

fws

On Fri, Aug 25, 2023, 11:58 PM Juho Laatu juho.laatu@gmail.com wrote:

I
The choice of n should be flexible enough that if two candidates both had
more than 70 percent approval, and nobody else got more than 49 percent,
then n should be only two.

Perhaps every finalist should have at least 71 percent (about root .5) of
the approval of the candidate with the most approval opposition to the max
approval candidate.

That 71 percent parameter is open to adjustment .

The idea is that we should admit into the final stage anybody with almost
as much approval as Chris Benham's max approval opposition challenger.

fws

On Fri, Aug 25, 2023, 11:58 PM Juho Laatu juho.laatu@gmail.com wrote: