Good example of plurality failure

I came accross this.

Might be a good place for an Approval campaign

1 Like

Also I messaged LP of Kentucky and one said:

Approval voting is garbage. It gets gamed and turns into bullet voting or disapproval voting. It’s a damn trainwreck at National every time.

Try Score or STAR instead.


By National, do they mean the National Libertarian Party? Also, how about suggesting IRV with equal ranks allowed (to minimize the need for Libertarian voters to Favorite Betray in close 3-way contests)?

I’d ask them for details, or an example. The way that they describe it being “gamed” may just be a reflection of the Libertarian Party. If there is one dominant faction represented by one candidate, then of course that faction is going to bullet vote. It could be that whoever sent you that reply is more concerned about not getting their way.

That said, at a convention, where people have assembled in one place specifically to vote on stuff, you can and should hold repeated ballots. It looks like the LP actually does use repeated ballots, according to their bylaws, they use repeated ballots until a majority is reached with IRV-style elimination between ballots (though if multiple candidates are below 5% all are eliminated.) I don’t see anything about approval voting in their bylaws.

IRV-style elimination would introduce the IRV pathologies that we criticize, however, so I would not recommend it. Instead I would recommend either the procedure recommended in Robert’s Rules of just holding repeat ballots until there is a majority with no elimination, or, if voters really need to be spurred into a consensus (and you can’t confine them indefinitely like Cardinals), something along the lines of:

  • Voters select a single favorite candidate.
  • If one candidate wins a majority of the single selections, they are selected.
  • Otherwise, an Approval ballot is held.
  • Before the Approval ballot is held, a threshold for elimination/continuing is set. For the first ballot, a preset threshold may be used, such as “1 vote”, but it should not exceed the number of selections the single selection leader on the first ballot received. For later ballots, it should be x votes plus the number of votes for the lowest continuing candidate. (Basically, for the last surviving candidate in a round to continue past the next round, they must expand their support by a satisfactory amount in between rounds). If no one passes the threshold, the least approved candidate is eliminated.
  • The Approval leader is considered the “present default option” and a majority vote is conducted to determine whether to select them. If the vote fails, then the process is repeated.

Would it be possible to let voters fill out a ballot, and then indicate how many rounds they want to keep their support on each candidate? It’d be a simulation of several rounds, with maybe a vote at the end to confirm whoever has the most approvals left. It might also be possible to indicate that you want a candidate to only start receiving support in a particular round and then for it to end at a later round.

Ha! Simple score voting would have been better by far for rectitudinous so-called ‘honest’ voters in this case. Instead of pouring over various links, here, in their own words is what the ever-virtuous Libertarian Party actually perpetrated:

=/ “Just take a look at the Kentucky Libertarian Party’s Facebook post after the results came in” /= :

:~~~~~~~~~~~~~~~~~~~~ //
In an ideal world, we elect Libertarian candidates and advance liberty. Failing that, we push mainstream candidates towards liberty to advance the cause.

But if we can’t do those things, we are always happy to split the vote in a way that causes delicious tears. Tonight there are plenty of delicious tears from Bevin supporters.

Had Matt Bevin not ditched his liberty Lt Governor for a Mitch McConnell picked anti liberty, corrupt running mate who has tried to eliminate Kentuckians jury trial rights, had Matt Bevin not presided over a huge sales tax increase, had Matt Bevin supported any of our key issues on criminal justice reform, marijuana legalization, expanded gaming, cutting taxes, or acted with the least bit of civility, we probably would not have run a candidate. Of course, he did the opposite. And here we are.

We split the vote. And we could not be more thrilled. If our friends in the major parties do not want this to happen again, they should think about passing ranked choice voting. And supporting our issues.

In the meantime, thank you to John Hicks, Ann Cormican, Kyle Hugenberg, Josh Gilpin and Kyle Sweeney for running. Your effort was appreciated.
// ~~~~~~~~~~~~~~~~~~~~:

They said =/ We split the vote. /= And therefor they caused the Republican Matt Bevin, who would have been elected governor of Kentucky in the Nov, 2019 election to actually lose to the Democrat Andy Beshear. Because the Republican had offended them in several ways. And so… they split the vote openly and deliberately!

We should compare what would presumably have happened if they had used single selection (‘choose one’), approval, and simple score voting. Let’ see what resulted with single selection. It’s almost for-sure that by obtaining a chunk of votes for their ‘dummy’ Libertarian candidate, they caused the Republican to lose:

(I make lots of mistakes with simple arithmetic, but this is probably mostly right.)

Andy Beshear ( D ) got 711955 votes.
Matt Bevin ( R ) got 707297 votes.
John Hicks ( L, dummy ) got 28475 votes.

And the Libertarians presume, almost certainly rightly, that if Hicks had not run, Bevin would have gotten just about all of Hicks’ votes, a total of 28475, would have gone to Bevin. 707297 + 28475 = 735772. So Bevin would have won. But no.

Now let’s try approval voting. The most reasonable expectation is that Bevin would have won with about the same numbers he would have gotten in the case just above – if, however, Hicks had not run. The results would then be:

Andy Beshear ( D ) got 711955 votes.
Matt Bevin ( R ) got 707297 + 28475 = 735772 votes.
John Hicks ( L, dummy ) got 28475 votes. (Here, Hicks did run.)

So in this case, Bevin wins.

Except for one little thing. The Hicks voters would have been… a bit less than ‘honest’. If we assume they had wanted Hicks more than Bevin, and had ‘voted their conscience’, so to speak, they would not have granted Bevin any votes, and the result would have been just the same as it was in the single selection case – and Beshear would win instead.

Now let’s see what simple score would produce. They can give between 1 and 10 votes, or explicit abstention. If they had voted shrewdly using the strategic hedge, plus, they really wanted to win, the results would probably be something like:

Andy Beshear ( D ) got 10*711955 = 7119550 votes.
Matt Bevin ( R ) got 10*707297 + 7*28475 = 7272295 votes.
John Hicks ( L, dummy ) got 10*28475 + 7*707297 = 5235829 votes.

But if they have simple score and everyone wanted to be rectitudinous Dudley Do-Rights and vote ‘honestly’ they would probably get something like:

Andy Beshear ( D ) got 10*711955 = 7119550 votes.
Matt Bevin ( R ) got 10*707297 + 5*28475 = 7215345 votes.
John Hicks ( L, dummy ) got 10*28475 + 5*707297 = 3821235 votes.

So with simple score, everybody can vote ‘honestly’ and the result would probably be the same as if the Libertarian ‘spoiler dummy’ had not tried to be a spoiler.

Do observe in the statement they gave above, that even the Libertarians have been indoctrinated into supporting the stealth spoilered ranked choice voting. It seems we have a P.T. Barnum “step right up, everybody’s a winner” scenario going on here.

Maybe something like this can be done with Condorcet i.e. take the Copeland winner in case of a Condorcet cycle, and then check if anyone can beat the Copeland winner and subsequently receive majority confirmation in their own right. Condorcet could also be used in a consensus-seeking process by maybe starting to tolerate a few pairwise losses, perhaps particularly the weakest ones, after a few rounds pass.

Interestingly, IRV does allow the major party supported by more voters to win. Which means that if Libertarians push for IRV, they will be removing their own leverage (in being able to field spoilers). That might be a good argument to sell to them.

The Beshear-Bevin margin was 4,658. With Approval, if even one fifth of the Hicks voters had also approved Bevin, then Bevin would have won. But why did none of the Bevin voters in your example approve Hicks?

(Again, we parse the word “honest” not as “rectitudinous” but more “truthful” or “reflective”, i.e. your vote accurately reflects your true preferences as much as the system allows.)

Libertarians can threaten to bullet vote in IRV, or even give their preferences to the other major party (if they are willing to spoil the election, they could extort the party they prefer more by threatening to elect the other party with their 2nd choice vote.)


You say: =/ But why did none of the Bevin voters in your example approve Hicks? /=

Well that’s quite reasonable. Then we might get (with strategic hedging):

Andy Beshear ( D ) got 10*711955 = 7119550 votes.
Matt Bevin ( R ) got 10*707297 + 7*28475 = 7272295 votes.
John Hicks ( L, dummy ) got 10*28475 + 7*707297 + 7*707297 = 10186908 votes.

So in this hypothetical scenario, Hicks would actually win. Obviously, these ‘scenarios’ are crude at best and cartoonish at least. But anyone can come up with their own figures, of course. Some people might try computer simulations. Warren D. Smith did wonders with such simulations. However I am convinced that simulations are quite limited in their ability to cope with the many factors involved with election systems, some of which are extremely exotic.

(Again, we parse the word ‘honest’ not as ‘truthful’ or ‘reflective’, i.e. your vote accurately reflects your true preferences as much as the system allows. No; in election methods analysis, we parse ‘honest voting’ as ‘incompetence’, or ‘irresponsibleness’.)

Addendum (edit):


I think you mistakenly led me into strange territory. Perhaps you meant:

=/ But why did none of the Andy Beshear ( D ) voters in your example approve Hicks? /=* (Not an actual quote)

Then we might get, say:

Andy Beshear ( D ) got 10*711955 = 7119550 votes.
Matt Bevin ( R ) got 10*707297 + 7*28475 = 7272295 votes.
John Hicks ( L, dummy ) got 10*28475 + 7*707297 + 5*711955 = 8795604 votes.

But John Hicks ( L, dummy ) would still win.

That requires coordinated voter behavior which is much more difficult to pull off than simply running / not running a candidate. Besides, the media will specifically be saying how IRV defended the winning major candidate from the spoiler. It still runs counter to the Libertarians’ message.

Many present voters vote based on these things called “issues”, and Libertarians tend to be closer to Republicans than Democrats on those “issues”, and so the Ls tend to pull votes away from Rs more than Ds.

That being said, there probably would also be some (albeit fewer) Dems who think the Libertarian is better than the Republican and they might approve D+L (or, in score, give L a low score like 3).

If the Libertarians are more likely to be Condorcet winners than Republicans, then it’s actually the Republicans who would want to Favorite Betray and put Libertarians 1st. Libertarians are often considered a more centrist party, with support from both sides, so they could potentially play that role if enough voters are willing to rank them 2nd rather than bullet vote.

It depends on the type of libertarian it is a big ideological space. Some are still socially conservative and some are very extreme wanting a near anarchist state. Also, in the United States the progressives started calling themselves liberals after the New Deal. This means that Classical Liberals often call themselves libertarians.

In general, I’d say voting reform’s biggest impact in a state like Kentucky is to allow more moderate voters to replace their party’s primary-nominated candidates, who are likely to be nudged extreme by partisan primary voters, not to help elect 3rd parties, seeing as Kentucky is one of the most Republican states out there. If you’re right that Libertarians may or may not appeal much to both sides, then the odds that they’d need to Favorite Betray in IRV are probably very low anyways.

Edit: The gubernatorial races have actually been closely contested in Kentucky over the past decade, which might just be because the Beshear family are very charismatic Democratic politicians, and partially because the Republican incumbents were unpopular. It’s also potentially because of low turnout; the 2015 gubernatorial election had 30% turnout.

The reason I suspect 3rd parties won’t get a big boost in safe states is because they can compete, without harming election results, in any state with a 70/30 or 80/20 balance in favor of one party, yet haven’t managed to grow much. The same applies to states with maybe a 60/40 balance but that use runoffs.

Edit 2: I have apparently been fooled by the fact that Kentucky has 2 Republican senators; a majority of the state is registered as Democrats, and it seems like a very close state overall, excepting that it leans heavily Republican in the presidential race and has Republican supermajorities in the state legislature and the House. So it might be a very interesting state to push a voting reform in, since it probably has the most voters of any state willing to vote cross-partisanly.