The problem with approval voting

[Bruce R. Gilson]

I keep noting that approval voting (unlike score voting with a few different levels) leads to a dilemma of how to vote on a candidate who you neither want to bring down (in case his principal opponent is one you dislike) nor to support too strongly (when his principal opponent is one you like better) if you don't know very well how the others might do. If we had approval voting in primaries, in 2008 I would have certainly approved Giuliani and McCain and not approved such as Santorum and Huckabee -- but I would not know how to handle Romney (as opposed to 2012, where Romney was by far better than all the others in that year's primary in my eyes). In plurality voting I voted for McCain -- a slightly insincere vote, as Giuliani was my first preference, but I figured by that time that he had no chance.

In 2016, I will probably vote for whichever of Kasich or Christie still seems to have a chance by the time the Maryland primary comes around. If I were voting on a score ballot, both of them would get a maximum score and Cruz a minimum. But there are a number of in-betweens. I really don't like Trump, but I see him as better than Cruz. I'm not too hot for Rubio (though I suspect he will be the ultimate nominee) but I don't know how I'd vote on him in an approval primary as I would rather see him than Trump or Cruz, but really want Christie or Kasich. So an approval primary would face me with this dilemma again.

[William Waugh]

If you know how you'd vote in a finer-grained score election, you can have the same effect in a coarser-grained one (e. g. Approval) by converting your chosen scores into probabilities using some source of random numbers, and approving with those probabilities.

[Key words: Score Voting, Range Voting]

On Tuesday, January 26, 2016 at 2:54:12 PM UTC-5, Bruce R. Gilson wrote https://groups.google.com/d/msg/electionscience/WolRcDGFMdU/5hvWuMOoEgAJ

[Warren D Smith]

On 1/28/16, William Waugh <2knuw...@snkmail.com> wrote:
> If you know how you'd vote in a finer-grained score election, you can have
> the same effect in a coarser-grained one (e. g. Approval) by converting
> your chosen scores into probabilities using some source of random numbers,
> and approving with those probabilities.
>
> [Key words: Score Voting, Range Voting]

--you *could*, but the fact is, most voters don't do that. Consequently
large differences between approval and score voting are
seen, empirically, in elections.

Also, if you like approval voting, but are stuck with plurality voting, then
we could get the same effect by, say, if you approve N<=5 candidates, then
plurality vote for a random one of them with chance N/5, and skip
voting with chance 1-N/5.
Presto, approval effect got with plurality voting.

Well, no.

[Bruce Gilson]

​Thanks for replying for me. Since you're the expert in this topic here, I am happy to see my ideas confirmed.

[William Waugh]

On Friday, January 29, 2016 at 12:36:53 AM UTC-5, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

On 1/28/16, William Waugh <2knuw...@snkmail.com> wrote:
> If you know how you'd vote in a finer-grained score election, you can have
> the same effect in a coarser-grained one (e. g. Approval) by converting
> your chosen scores into probabilities using some source of random numbers,
> and approving with those probabilities.
>
> [Key words: Score Voting, Range Voting]

--you *could*, but the fact is, most voters don't do that.  Consequently
large differences between approval and score voting are
seen, empirically, in elections.

Have there been any empirical studies that kept a system in place for long enough and over elections that mattered enough to draw the attention of party strategists?

I see every reason for party strategists to notice what strategies would work (as revealed by game theory) and to advise the voters in their party how to vote, and and I see every reason for voters who care about winning for the ideals and values they share with their party to follow that advice.

Also, if you like approval voting, but are stuck with plurality voting, then
we could get the same effect by, say, if you approve N<=5 candidates, then
plurality vote for a random one of them with chance N/5, and skip
voting with chance 1-N/5.
  Presto, approval effect got with plurality voting.

Well, no.

If it's "no" rather than "yes", why did you write it?

[William Waugh]

On Tuesday, January 26, 2016 at 2:54:12 PM UTC-5, Bruce R. Gilson wrote https://groups.google.com/d/msg/electionscience/WolRcDGFMdU/5hvWuMOoEgAJ

I would favor finer-grained Score voting (with 1, .99, .90, .50, .10, .01, 0) over mere Approval if it weren't for concerns about fraudproofing the tally.

[Warren D Smith]

> Also, if you like approval voting, but are stuck with plurality voting,
>> then
>> we could get the same effect by, say, if you approve N<=5 candidates, then
>>
>> plurality vote for a random one of them with chance N/5, and skip
>> voting with chance 1-N/5.
>> Presto, approval effect got with plurality voting.
>>
>> Well, no.
>>
>
> If it's "no" rather than "yes", why did you write it?

--There is such as thing as "reductio ad absurdam."
Those of us who have managed to reach or surpass the ancient Greek
stage of human development, are aware of that.

[Clay Shentrup]

On Wednesday, March 2, 2016 at 9:16:15 AM UTC-8, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

> Also, if you like approval voting, but are stuck with plurality voting,
>> then
>> we could get the same effect by, say, if you approve N<=5 candidates, then
>> plurality vote for a random one of them with chance N/5, and skip
>> voting with chance 1-N/5.
>>   Presto, approval effect got with plurality voting.

No. E.g. suppose Approval Voting would yield Liberal1=60%, Liberal2=60%, Conservative=40%. So liberal voters, who all "approve" both liberal candidates, randomly vote for one of the two liberals with Plurality Voting.

Result: Liberal1=30%, Liberal2=30%, Conservative=40% 

[William Waugh]

Yes, there most certainly is such a thing as _reductio ad absurdam_. But the topic was simulating finer-grained Score voting by strategy when faced with Approval. Are you saying you can reduce that to an absurdity? I haven't managed to reconstruct a step that you didn't mention, to get from simulating the effect of finer Score given Approval to this new idea you introduced of simulating Approval on vote-for-one Plurality.

[Clay Shentrup]

On Wednesday, March 2, 2016 at 6:12:24 PM UTC-8, William Waugh wrote:

On Wednesday, March 2, 2016 at 12:16:15 PM UTC-5, Warren D. Smith (CRV cofounder, http://RangeVoting.org) wrote:

I haven't managed to reconstruct a step that you didn't mention, to get from simulating the effect of finer Score given Approval to this new idea you introduced of simulating Approval on vote-for-one Plurality.

I'm pretty sure I just demonstrated, via reductio ad absurdum, that it can't be done.