A variant on this is to pull 3 ballots out and take the majority (if none, keep pulling until there is a plurality then stop).

This would make the extreme outlier results less likely, but still maintain most of the properties you said. For example, if there are 40% violets and 60% blues, then chance of a violet winner becomes only 35.2%. If there are 10% violets, their chance of winning is only 2.8%.

Depending on how risky you want to be, you can replace 3 with 5, 11, 31, or 101. For example with 11 ballots, 40% violet in the population, their chance of winning is 24.6%; with 10% violets, their winning chance is only 0.03%.

Reminds me of of the cache eviction strategy 2-random (pick two entries at random and evict whichever is oldest).

Assuming that the 1% extreme minority wants to turn it into a dictatorship, you'd be guaranteeing that sooner or later the democracy would fail.

If used to elect something like a president, yeah, that could be problematic. But if used to elect a legislative body then its not really a problem. Sure 1% of the body would be pro-dictatorship, and occasionally a bit more due to the randomness, but its not a serious risk that a majority of the body would turn out to be pro-dictatorship.

> Assuming that the 1% extreme minority wants to turn it into a dictatorship, you'd be guaranteeing that sooner or later the democracy would fail.

Luckily a single election isn't enough to make a dictatorship, in the US.

There is never a guarantee that a government will stand forever.

Sure, but we could have separate rules for changing how the voting works, compared to the day to day operation of the groups. Also, on the good end, maybe the 1% have some amazingly good ideas that the other side would never implement otherwise.

Or just use approval voting: you're allowed to checkmark as many candidates as you want (instead of just one), and whoever gets the most points wins.

1. No spoiler effect. If you really want Green to win, but you think Blue is the lesser of two evils, voting Green and Blue at the same time is strictly better than voting only for Blue.

2. It's about as easy to understand as FPTP. No funny business where politicians mislead people about how the voting system actually works. Approval voting has a few downsides, but they are obvious downsides, so everyone is on equal footing.

3. You can still use those nice scantrons and hole punches.

4. No weird questions about "should I give the candidate I want most a 1, or should I give the candidate I want most the 10?"

5. It's a majority rules system and not a dictatorship.

I don't really get the love that approval voting gets.

Approval voting requires the voter to decide whether or not to approve the lesser of two evils. It seems to me that this isn't a simple question it depends on the odds of different outcomes, level of preference between two evils, etc. To me that seems like it would make voting harder for most people, and politicians would absolutely be trying to mislead people about the appropriate strategies for who to approve and not approve.

I thought most people vote for who they like, or at least don't know they dislike, or just for those in their party. If that assumption is true, then approval voting wouldn't be nearly as challenging as you're making it the case to be.

Sure, approval voting is easy if you don't care about making effective use of your vote.

If you want to make effective use of your vote, you've got to make calls about supporting you people you dislike, but don't dislike as much as the other guy.

Is it worse in that respect than regular "pick-one, winner-take-all" voting? Ironically, it might be the lesser evil.

The important feature that approval voting has, and plurality lacks, is that approving your favorite candidate is never against your own interests. What you do with your second favorite is a matter of strategic voting, but that’s what you were doing in FPTP anyway.

Honestly, I'm not sure.

Pick-one, winner-take-all is pretty bad. Approval voting might be slightly better in some ways, but I'm not sure its better enough to make up for being just a little bit more strategically complicated.

I think on balance Approval Voting would be an improvement, by removing the spoiler effect and allowing new parties to emerge (and old parties to split), but you're right that burdening voters with difficult strategic questions is not a decision that should be taken lightly.

Of course, FPTP has its own strategic problems, leading to people having choose whether to vote against their actual preference, but this is at least a familiar problem for people that have experienced this type of election before. I suppose that voters who are confused by Approval Voting could continue to treat their ballot as if it were a FPTP ballot, but this does feel like it is giving more voting power to people who are more intelligent (which some people here might undemocratically see as a positive) or more devious (which people might not see themselves as).

For me, though, the biggest issue with Approval Voting is that it makes the counting process harder, which risks entrenching the need for untrustworthy electronic voting machines. Approval Voting does at least satisfy the Summability criterion, but, if counting by hand, the ballots ideally need to be divided into 2^N separate piles (for N candidates), or there needs to be N separate running tallies of votes, as opposed to under FPTP where there only needs to be N piles, and the piles could even be weighed to count them.

I don't believe that it is an unreasonable requirement that an electoral reform proposal allows easy hand counting of ballots, as MMP can be just as easy to count as FPTP. (If a separate ballot paper is used to express party preference in addition to candidate preference, that effectively means 2N piles, but the Zweitmandat system allows the party preference to be inferred without needing the second ballot paper).

This is called random dictator: https://en.wikipedia.org/wiki/Dictatorship_mechanism#Random_...

> Power never falls into a state where the majority party can stay in control.

That usually only happens when power gained by elections is used to change the system, often outside of the formal rules of change. Obviously, no election system can safeguard against this, since it is an effect outside of the electoral system.

> Every vote counts.

Only one vote counts. This is literally the textbook example of a system than fails the non-dictatorship criterion.

> Extreme minority votes still matter

Well, again, since only one vote matters, it's only an extreme minority of votes that matters, but it's extremely unlikely to be one from an extreme minority position.

> Only one vote counts. This is literally the textbook example of a system than fails the non-dictatorship criterion.

This depends on how you state the non-dictatorship criterion. Wikipedia's discussion refers to a "single pre-determined person's preferences." But since the dictator is selected randomly, they aren't pre-determined.

Further, every vote counts in the sense that every additional vote for a candidate increases the odds of them winning the election. The candidate has an incentive to get additional votes no matter how many votes they currently have. You don't have the situation where its futile to vote because the outcome isn't going to be close.

In fact, in that scheme, votes for a minority party are more significant than they are in our current system.

I think an additional neat property is that the representation of a particular district would be proportional in time. I.e. 40% red, 60% blue district would be represented by a red member 40% of the time and a blue member 60% of the time.

> Extreme minority votes still matter. If 99% of the population is blue, and 1% of the population is violet, violet supporters still feel their voice can be heard.

Why do they feel this?

Proportional representation has many of the good qualities listed here but is probably more politically feasible.

Amend this so that the time in power is proportional to their vote.