IRV is not really a good voting system. It's better than plurality, but it still doesn't really allow viable third parties. If a third party ever catches on, they could steal first votes away from a major party, and cost both of them the election. Demonstration here: https://www.youtube.com/watch?v=7Q7rzqJ0YS8
I would recommend a condorcet method, or approval voting. Condorcet methods in particular have a lot of nice mathematical properties and are close to optimal with honest voters. I think it's the most likely to allow third parties to actually get elected. You can see some results of simulated elections with different voting systems here: http://rangevoting.org/BR52002bw.png http://rangevoting.org/BayRegsFig.html
They claim that just approval voting is the same level of improvement from plurality voting, as plurality voting is from monarchy. Based on the massive voter satisfaction index improvement. It really is crazy that we still use plurality voting.
I, too, came away with approval voting being my favored approach.
One question to ask yourself, though, is this: in an election where candidate A is loved by 55% of the population and hated by 40%, and candidate B is loved by no one, but tolerated by 80% of the population, who "should" win? It's more of a philosophical question, but depending how you answer it influences how you should choose a voting system.
If you think the 80% tolerated candidate should win, then approval voting is for you. If you think being loved by 55% of the population should be the winner, then IRV or another system may be better.
I personally prefer the boring, centrist 80% tolerated outcome, so I like approval.
Another thing I realized, is that a lot the research doesn't seem to take into account polls ahead of time and iterated preference making. Approval voting might be terrible if it's once-and-done, but if you have polls leading up to it, you can calibrate your "approval level" a little better.
E.g., if Hitler is running, you might say "I approve of everyone else". But if you see polls suggesting that no one is approving of Hitler, then you might raise your standards a bit and not approve of some more candidates.
Note that in practice, people start to vote tactically to prevent radical candidates, so there really isn't any voting system that works great for that purpose anyway.
I think the centrist option is probably the best for a number of reasons. I think a particularly bad president can do much more harm, than an unusually good president can do good. E.g. start WW3, or destroy the economy. Not to mention dividing the country and polarizing our political system.
Second there is a phenomenon called wisdom of crowds. That if you take the middle of the estimates given by a large number of uninformed people, it usually comes out very close to the true answer. The same should apply to voting. The middle preference of a large number of uninformed voters, should be close to optimal.
Third there are studies that show that centrists and moderates tend to have much more accurate beliefs than others. As in, their predictions of future events were the most accurate. Ideological people were extremely biased and inaccurate. That's probably the closest we can get to scientifically measuring how good their policies would be.
Lastly, for values questions, where there is no "right answer", the middle should be chosen. It's the fairest choice.
>Lastly, for values questions, where there is no "right answer", the middle should be chosen. It's the fairest choice.
As a radical, my problem with this is that the "middle" is defined by the power struggle between the "extremes": it always tracks the formerly "extremist" positions of the winning side. For instance, until last week, nobody would ever have said that privatizing Medicare was anything but extremist far-right lunacy. Coming this year, we're going to hear that it's a mainstream, centrist position. That's not even being said as a statement of fact, since it's transparently false -- people love Social Security and Medicare! It's being said as an imposition of hegemony on the part of the electoral winners, as how "we create our own reality now".
I'd expect Approval Voting to have the same property.
Playing around with single winner election methods may be useful for some purposes, but it's not how you really get additional viable parties.
And most positions are required to be single winner, like mayors, judges, sheriffs, governors, or the president. So it makes sense to focus on single winner systems first.
Comparative empirical studies of modern democracies that I've seen (e.g., as detailed in Lijphart's Patterns of Democracy) suggest that this is simply not correct.
> And most positions are required to be single winner
What elected positions need to be single winner is a matter of choice, but, yes there are some, and for those better single-winner methods help. But they won't make additional larties viable the way that proportionality in the legislature would.
What do you mean? How many countries use approval voting or condorcet methods to select parliaments? Because that's what I'm referring to. These systems tend to select for "best compromise" or centrist candidates, which should represent the population pretty well.
Proportionality of representation in the legislature strongly predicts both number of sustainable parties and public satisfaction with government in modern democracies.
> These systems tend to select for "best compromise" or centrist candidates, which should represent the population pretty well.
Not really; it's not like views on political issues are clearly unimodally distributed such that a single compromise candidate does anything but make everyone in the electorate feel like they have no effective representation of their views in government. That's what proportionality avoids.
Even if they did, representing the majority opinion in each district in the legislature and taking the majority for those majorities as the policy is not the same as directly representing the diversity of views in the electorate in the legislature and taking the majority of that diversity as the policy.
An ideal single-winner system aims to approximate the former, an ideal proportional representation seeks to approximate the latter. Even an deal single-winner system doesn't approximate what PR is doing.
Proportionality, particularly, mitigates the difference in representation between geographically diffuse and geographically concentrated viewpoints that occurs in any, even ideal, single-member district system.
I prefer more direct democracy thru elected executives vs say prime ministers or presidents of the council. Which is the best use case for Approval Voting.
But I've never experienced parliamentary government, so take this with the usual caveats.