We use ranked choice in Australia and no one ever looks at the first-preference votes. Its a tertiary question. Everyone only looks at the two-party preferred, the final two-person result. Only politics and maths nerds care about first preferences.
But it is possible in some examples that the final round seems decisive, even if some serious contender was eliminated in an early round. What we have in the video is a race that isn't close, even though it seemed close in one of the early rounds. But it is possible for a race to ACTUALLY be close, even though each individual round appears decisive. This IMHO is much worse, since nobody would ever notice unless they have access to the full data, and they actually go through whole the NP-hard calculation. This paper here gives a (made-up) example where it happens: adsarwate.github.io/assets/pdf/papers/SarwateCS13irv.pdf They use 6 candidates... I think it's probably not possible to do something similar with only 3 or 4. (Alaska always limits their RCV ballots to 4, so it won't be a problem there.)
@@ChrisStaecker Also gotta call out: "No TV news person is gonna want to explain that." It is clear you need to be introduced to the undisputed GOAT of Australian TV - Antony Green.
Then please don't you EVER complain about an election result! The preferential system gives everyone-the voters, candidates, the political parties, a more sophisticated view of how the country feels. The Liberal/National collapse leading to the Albanese government victory was not just a case of a few Labor and Green voters grabbing power. Several "safe" liberal seats fell due to the "Teal Liberals" . In other words, a significant portion of those who subscribed to the right of politics got disenchanted with their own party's polices and governance. This is an absolute gift for Australia, because it means that extremism is not tolerated, be it from the left or the right. And yet if you want to vote for a loony left, or "greenie", you can, and the same goes for hunters, shooters, and fishers, and even Pauline Hanson or Clive Palmer, if that is your wish. Our system is far from perfect, but everybody gets a "say". And we know everyone gets a say because voting is compulsory. This is not "nerdy" or "academic" at all, it is democracy in practice.
Hi from Alaska. I live in the interior city of Fairbanks and am graduating with my BS in math in two weeks. I never heard of this election! Thanks for the excellent coverage. I got to vote in the election that established ranked based voting for our state, and it was a very close call!
The issue with voting in places with FPTP, like the UK and America, isn't actually FPTP, though it's made worse by FPTP and any other voting methods is better; the real issue is that you can't get proportionality by having tiny one-candidate districts. No matter what voting system you use, having small voting districts will still be skewed and broken, and even if you used, say, a Condorcet method (round robin where the candidate (if there is such a candidate) who wins all pair-wise competitions with all other candidates is the eventual winner) the result can even be possibly that a party takes all the seats.
Yes- we also have serious problems with Gerrymandering which would totally evaporate if we switched to systems that worked harder to ensure proportionality. This is not solved by RCV.
Okay, it's NP-hard, but I'm pretty sure N is small in this case... now, _explaining_ that on the evening news, that's a different kind of hard-problem.
@@ChrisStaecker Aye. If I had to guess there's an n! or similar hiding in there. Maybe 2^(n!)- which is gonna be great fun if we ever get up to 6 candidates in a race :-p
@@codertao Yes it is O(n!) to even read the voter data, since each voter chooses a (subset-)permutation to vote for, and there are more than n! choices for those. So basically any analysis of anything interesting about an RCV election will automatically be at least factorial complexity.
There's a second-order effect to having ranked-choice voting that's good for democracy, too. If the red party and the blue party consistently get 90% of first-place votes, but the green party gets 10% of first-place votes... say if the first-round totals are close, like, red 46% and blue 44%, then the blue party gets the message that they should maybe adopt some policies of the green party. Ranking the green party first, though there may be no hope that green wins, is a way for the electorate to communicate what they want to see in government, a way to put a little pressure on the two main parties' policy-making.
At first blush, it would seem to me that simplest solution is to always recount. In other words, have two independent processes for counting done in series. The goal would be refine consistent counts: look for discrepancies, fix them and document them for regression testing. In software development, finding bugs is NP hard. Our solution is to make Continuous Improvement a part of the process. We try to create tests to catch issues, but we also add tests after regressions are found. This adds confidence and trust that the system operated as expected; thus, allowing us to believe the system did and will operate in exceptional circumstances.
Great video! To me this is another point in favor of Approval Voting (as opposed to ranked choice). You still avoid the spoiler effect (mostly), and its a lot easier to interpret the results at the end ("Who has the most approvals?")
I dislike approval voting because it forces voters to strategize a lot, but it is an improvement over FPTP. Given that there is no perfect voting system and an improvement is needed badly, we should cooperate to make sure at least one of these solutions is adopted. Here in Czechia, everything is either proportional or has a runoff round. It is decent, but could be improved.
Oh cool, I thought this was going to be that actual thing wrong with ranked choice voting that I'm unable to foresee. The idea survives another day of plausibility! Props for excellent citizen-reporting.
I scrolled through all the comments thinking I'd find at least one or two proponents of Majority Judgement or other highest median voting systems, but apparently there isn't, so I feel compelled to fill the niche. With MJ you get to grade candidates instead of just ranking them. The system avoids Arrow's paradox (a minority third candidate can't cause the most popular candidate to lose to the runner up). And it provides rich statistics on voter preferences. A major drawback is the fact that it uses the median, not the mean: I don't think the media are cognitively ready for medians.
What provision does Alaska have for the case where, in a three-candidate race, there's a tie for second place? Or worse yet a three-way tie? How do they decide whose votes to redistribute to whom? Do they just carry on with round two including all three candidates again?
Typical rules for RCV will simultaneously eliminate all candidates who are tied for last place in any single round. But some different implementations use slightly different rules in the case of last-place ties. The official rules for the state of Alaska say that they resolve a last-place tie "by lot", i.e. randomly. (Coin-toss) It's never happened, as far as I know. See: www.elections.alaska.gov/election-information/#RankedChoice
why use ranked choice tho when you could use star voting or approval voting which are both much cheaper/safer and less influenced strategic voting (which is the whole reason to choose a voting system in the first place!)
Fantastic presentation and analysis. I'm wondering if there is a nice visual that can be generated so that we can see 1st round AND 2nd round outcomes at the same time so that "Close" won't appear close (if it isn't close, as in this case). Hmmm. Have to think about that. Edit: As pointed out by another commenter, there is no need to show the first round results. That's the answer. Easy peasy. :-)
It's worse than that- it is possible in some examples that the final round seems decisive, even if some serious contender was eliminated in an early round. What we have in the video is a race that isn't close, even though it seemed close in one of the early rounds. But it is possible for a race to ACTUALLY be close, even though each individual round appears decisive. This IMHO is much worse, since nobody would ever notice unless they have access to the full data, and they actually go through whole the NP-hard calculation. This paper here gives a (made-up) example where it happens: adsarwate.github.io/assets/pdf/papers/SarwateCS13irv.pdf They use 6 candidates... I think it's probably not possible to do something similar with only 3 or 4. (Alaska always limits their RCV ballots to 4, so it won't be a problem there.)
3 Remarks: A: It's not stupid for Cacy to demand a recount, *IF* her goal was to give Holland the win over Giessel. (She would still have had to find enough votes that Giessel gets eliminated in the first round, but with 140-ish votes out of 16k, that's at least close enough to the 20 vote threshold that it seems somewhat plausible to me.) B: Well, if the media had realized that this was a waste of time, they would ordinarily respond by not wasting their time on reporting on it. C: Something being NP-hard doesn't mean that something is near impossible. It just means that it's guaranteed to get harder _at a ridiculous pace when you add more candidates._ It's most definitely quite easy in the case of 3 candidates. ("You can do it by hand on a napkin" kinds of easy.) And even with 20 or so, it might just take a computer program 10 minutes to brute force the solution.
A: It's unreasonable for a recount to swing anywhere near 140-ish votes in a field of this size. More likely swings of single-digits. And a cursory look at the politics shows that there's no way Cacy would've wanted Holland to win the election. Cacy and Holland are opposite ends of the spectrum, with Giessel more moderate. (You can see almost no Cacy voters list Holland as their second choice.) B: To me it seems worth reporting that a useless recount was happening. C: This is absolutely right- the NP-hard thing is mathematically interesting, but not necessarily very important for real political elections. In Alaska they already limit the final RCV ballot to 4 candidates, so their true margin of victory will always be computable with no real problem.
The most important fact is increased choice for voters, without one or more run-offs. And sure, if you look at it from a politically "religious" standpoint, you are not going to appreciate the subtlety of ranked choice. For example, if a Republican of Democrat is an axe-murderer, and you don't care about that, then you will vote R or D anyway. Under preference voting, you can choose from a field of candidates, not just two. Some are going to be closer to your personal position than others. And preferential voting encourages more candidates to run, which is healthy politically. There is no reason why a field of candidates such as a number of Dems, GOP, and independents can't run. There is no downside. If you are worried about election efficiency, then have compulsory voting, no political financing, and make all candidates really work for those votes. "Name recognition" is not the electoral system's problem, it is the voter's responsibility.
In Ireland we have a single transferable vote system in which the vote count in actually non-deterministic. Yep. It's pretty great, actually. And it works because 1 - there is a huge amount of transparency during the (manual) count 2 - it turns the counting of votes into a spectator sport that people like to bet on 3 - it succeeds in bringing a diversity of 3rd party and independents into government 4 - it turns out that, in practice, recounts very rarely matter 5 - we have *lots* of candidates (in my old district I had 20 choices) I think that 1 & 2 above are the important ones. The government tried to bring in automated counting and people protested in the streets. Irish people understand the logic of the system, but they want to see it happen. Trust *and* verify. Always
@@ChrisStaecker Most of the non-determinism comes in because our parliamentary elections use multi-seat districts to allow for proportional representation. In Co. Wicklow, we had five seats. This means that in order to be elected a candidate had to receive 1/5 of the total vote. If in a round of counting there is a candidate who has a surplus of votes over the quota, the extra votes _as determined by the order of the count_ all get transferred to their next preference down on the list, to be counted in the next round. If no candidate is elected, the candidate with the least votes gets eliminated and all of their votes get transferred to the next candidate down on the list. Between the order-dependence of the count and the reshuffling that occurs due to transfers it becomes entirely possible that different counting orders could result in different outcomes. But every vote gets counted in some *positive* sense towards the outcome. I believe that the Australians use a variation of the same concept. For a more detailed (and accurate) description: www.citizensinformation.ie/en/government-in-ireland/elections-and-referenda/voting/proportional-representation/
@@kumoyuki Interesting- and what determines the order of the count? Just the order in which votes are cast? If they need to recount, do they somehow recount them in the same order?
@@ChrisStaecker I've never gone to personally witness my local count, but a surprising number of people do. That might be because there are so many candidates - in Ireland the chances are good that you know somebody on the ballot :) I believe the ordering is determined on the first count, which is when they figure out how many ballots were actually cast. I never really thought much about repeatability in the count. I guess that establishing a total ordering across all the ballots makes the count repeatable, even if it might produce a different result under a different ordering. I was invited to a standards committee shortly after I moved to Ireland and one of our first tasks was figuring out how to vote on proposals. This led me into a deep rabbit hole on the mathematics of voting systems: it's a really fascinating field. I'm fairly proud of the Irish system because it has nearly all of the theoretical properties required to really establish a fair vote that is resistant to tactical voting strategies. It's just non-deterministic ;)
How would you know for sure that holland's voters didn't have a higher proportion of 2nd choices being casy? It seems like an unproven assumption that the voters for the other 2 candidates would've had the same split in their ranked choices?
The votes are public information, so we do in fact know exactly how many of Holland’s voters will break for Giessel vs Cacy. None of what I say around 2:40 involves any hypotheticals or conjecture- the numbers come straight from the actual votes.
I think there's a practical solution to the problem that calculating the margin of victory is difficult: require the candidate requesting the recount to prove the election could have been swung by twenty votes or less (or whatever threshold one uses) if they want the state to pay for the recount. This doesn't even require the candidate to find the actual margin of victory, just prove that it's less than or equal to twenty votes
Yes- this is a good idea. The recount trigger rules should be phrased like "margin of victory of 20 or less", rather than "difference in votes of 20 or less". And I was being a bit overdramatic on the NP-hard point. The calculation becomes intractible when the number of candidates is high, but for 3 or 4 candidates it is perfectly doable. Even so, like you suggest, there are other efficient (not NP-hard) algorithms which can compute upper bounds for the Margin Of Victory, which is often good enough. In many cases it's not necessary to compute the true MOV.
*specifically* in a politically environment with 2 parties (or two major parties) I'd argue that a simple color coded pie chart will represent the electorates preferences very well. A third is blue, a third is red, a third is a very slightly different red. Who win? One of the reds, let's find out which in the next round
Yes "NP-hard" is a bit of a red herring in cases like this. It only means that the problem becomes intractable as n increases. For n=3, in this case, no problem.
The issue with ranked choice voting is not it’s structure. It’s getting your normal joe to understand and trust how it works. Seeing how much people can’t even trust FPTP elections today, I can see why RCV is running into issues like this.
If you were to create a vote-counting-methods-appreciation-and-discussion club... how would you rank the different vote counting methods for selecting the club's president?
I have a hard time having strong feelings about it. To me, plurality seems fairly bad. I like approval voting for using in real-world low-stakes decisions. For repeated low-stakes decisions I like the "random dictatorship". For more than 2 candidates, the various ranked methods are a little too complex for me to feel strongly about. I feel like RCV (IRV) generally works well, but I can't strongly prefer it over Coombs' method. And I can invent examples where RCV makes "bad" choices.
@@ChrisStaeckerIn most important cases I wouldn’t recommend this option, due to the complexity, but, if all the candidates and voters are enthusiasts about voting systems, I think Ranked Pairs would be nice? Given each voter’s rankings, for each pair of candidates, compute a margin of victory between that pair, and then sort the pairs from greatest to least margin of victory. Then add edges to the digraph one by one according to this list, skipping any that would create a cycle. (Or, like, if it would make a cycle, put it in the opposite direction I guess)
Love ranked choice voting as an Australian. Arguably even more important is voting being mandatory. The US has a serious voter apathy problem which leads to incredibly skewed election results, since only a third or so of the country even votes at all. Plus, since its a requirement that everyone votes, access to polling places (and early voting options) is assured. If a community loses access to voting for whatever reason (racism, an attempt to skew the results by district, etc.) then it's a federal issue that will be resolved.
@@tuppyglossop222 Do they really? That's even crazier honestly. I always thought election days should be public holidays anyway, to make it even easier for people to attend.
NP-hard probably doesn't matter because you don't need an efficient algorithm unless there are tons of candidates. Digging through n! scenarios is no big deal as long as the number of candidates is smaller than maybe eight to twelve or so.
Don't judge too hard.Preference voting is great, unfortunately too lees used. Awesome Alaska used it, hurray! Now you explained it kind of backwards, blaming them for making a bad decision. Let's look at it with a lessons-learned view: • Recounts are an important tool, having a reasonable low barrier is important. • The recount law was made for non-preference voting. Let's extend it to cover more voting systems properly. • Most people, excluding mathematicians and voting systems fans where confused. Journalists are usually good with words, and not with the actual topic they are reporting about (look at a car magazine for examples). Don't blame them, help them. •• First round sums are not the result, so do not look at them, look at the firsts, seconds and so on. • Politics is a game where you have to know the small details and play all the rules. Don't blame the guy who found and used the obscure rules. • Casy may not have understood that there where no chances of winning with a recount. Anyways, it's democracy and rules....
Was the recount actually bad? Sure, Cacy had nothing to gain, but voters have an interest in having their preferences reflected in the outcome. To do this means that it is important to determine if Giessel lost in the first round because then Holland would have been the winner. Of course this is very unlikely because Giessel was the frontrunner, ~100 votes above Holland not to mention Cacy.
The recount didn't hurt anybody other than wasting some time and money, so it wasn't actually really bad. But it also had no chance of producing any useful information, in my opinion. The rounds themselves are just mathematical abstractions, with no individual meaning. So saying "we really want to be sure who won and lost the first round, even if it doesn't change the end result" isn't really getting at anything important anyway.
@@001HK0 OK I agree if Giessel actually lost the first round then that really is meaningful. But the margin there (Holland vs Giessel) was not small enough to justify a recount. Certainly not a state-paid auto-recount.
I feel like the maths (while NP hard, and that's clearly a high bar for hardness) is the easy bit. What's completely unknowable here is what would have happened if there were an equal number of democrat and republican candidates, or 2 democrats and 1 republican. I don't know (I'm from the UK) how candidates are selected, but I do (think I) know that over here the number, and spread, of candidates standing often skews the situation. Personally I suspect most voters vote along party lines, so this case would be a legitimate republican win. Wouldn't be my my choice, but democracy is (at least some sort of) democracy. I would love for unbiased statisticians to be in charge of the whole process, somehow aggregating votes over policy issues, "values" or somesuch, but I'm not sure there are any unbiased statisticians - and even if there were, the voting population probably wouldn't understand or accept them.
i don't know how you can calculate the number of votes Cacey would get if Giesel lost the first round, or if Holland lost the first round. Personally, I'm neither R or D, but independent. It may be that if Geisel lost, then Cacey would have won if the women who voted for Geisel voted for Cacey, because she is a woman, party affiliation aside. When Mr T was running and Bernie was on the ballot, here in the northeast, there were a lot of young people that said they'd vote for Burnie, but if he didn 't make the cut, they'd vote for Mr. T. (Of course in NH, Bernie DID win the primary, but the superdelegates overrode the the popular will and voted for Hillary in the electoral college.) But I'll freely admit that I am a dumbass : - ) As an aside, in a local election here we had an alderman lose an election by just 5-10 votes; a recount ensued and he won by a couple votes, another recount and he lost (again.) Finally, after a third recount, he won....... wtfork is goin' on?
We can indeed calculate that number of votes because all of the vote information is known- not just the "round totals", but we know exactly how many people voted Giessel 1st with Holland 2nd, vs Giessel 1st with Cacy 2nd, vs etc etc. So there is no ambiguity about how various votes would've shaken out in different scenarios- we can figure it out exactly. What I'm saying at 2:40 with the specific redistributing of votes is not my own guess- we know that this is exactly how it would've happened, because the votes say so.
It's a terrible and unnecessary system anyways, there's not a lot of benefit in analyzing which implementation of it is better than the other and what the general mass thinks about one or the other.
All universal-suffrage systems are terrible. The opinion of the general mass (which includes me by the way) on who's in power has been proven time and again to only be useful as a destabilizing factor and to promote populism. The only voting system for a country's leader that has benefit in it is the one allowed by the shari'a, where the ruler is elected by "decision makers" i.e. those who's opinion has weight in the land. And this system is natural to human societies, when the people who rule the land are the ones which the "decision makers" of that land don't favor - the entire society suffers. When the people who rule the land have support from the "decision makers" things tend to go smoothly. With that in mind, it's obvious that the more people are allowed to vote the less likely it is for the better outcome to occur. And if you will look at the history of the gradual expansions of suffrage you will see that those who had pushed for its expansion did it for that exact purpose - to sacrifice the wellbeing of the society for the sake of the populists.
@@ChrisStaecker Snark doesn't make one smart. The smart move would be to try and prove me wrong, which you won't be able to because what I have presented to you is a rule set by our creator for the benefit of mankind. "Should not He Who has created know? And He is the Most Kind and Courteous (to His slaves), All-Aware (of everything)." 67:14 Calculus had been around and known quite well since before the failed experiment called "democracy" was brought about in the American and French revolutions. And I expect someone like you to know calculus more than a great deal of the population knows multiplication. Doesn't it bother you that over two centuries later no one can even figure out how to deal with the results of a vote? Gather up the best mathematicians, social sci, pol-sci geniuses of all times - bring them in the same room together and even if they'd manage to agree on a single best system it's still gonna fail miserably even after a thousand iterations and you know it.
@@pubcollize I wouldn't presume to try and prove you wrong. I assumed your objections were for mathematical reasons- but now I see that you are coming from a very different point of view than I am. I am a person of faith too, and here in the USA my tradition has produced a crop of folks who want our government to be based on their version of "rules set by our creator". I am 100% opposed to this. As for "Doesn't it bother you?" Yes it does! For some time it has been demonstrated that there can be no mathematically perfect ranked voting system. This is unsettling, but absolutely does not make me want to abandon universal suffrage.
I find the mind boggling part of this entire story to be that the Bars in your graph 📊 … ARE BLUE … 🤔 are they just as sad about this? Was it a graphic beyond your control? Or! was it the graphic of some liberal media outlet? 😅
We use ranked choice in Australia and no one ever looks at the first-preference votes. Its a tertiary question. Everyone only looks at the two-party preferred, the final two-person result. Only politics and maths nerds care about first preferences.
Clearly this is the way to present the results - i.e., the final two person result. Nice.
But it is possible in some examples that the final round seems decisive, even if some serious contender was eliminated in an early round.
What we have in the video is a race that isn't close, even though it seemed close in one of the early rounds.
But it is possible for a race to ACTUALLY be close, even though each individual round appears decisive. This IMHO is much worse, since nobody would ever notice unless they have access to the full data, and they actually go through whole the NP-hard calculation.
This paper here gives a (made-up) example where it happens:
adsarwate.github.io/assets/pdf/papers/SarwateCS13irv.pdf
They use 6 candidates... I think it's probably not possible to do something similar with only 3 or 4. (Alaska always limits their RCV ballots to 4, so it won't be a problem there.)
@@ChrisStaecker Also gotta call out: "No TV news person is gonna want to explain that." It is clear you need to be introduced to the undisputed GOAT of Australian TV - Antony Green.
Then please don't you EVER complain about an election result! The preferential system gives everyone-the voters, candidates, the political parties, a more sophisticated view of how the country feels.
The Liberal/National collapse leading to the Albanese government victory was not just a case of a few Labor and Green voters grabbing power. Several "safe" liberal seats fell due to the "Teal Liberals" . In other words, a significant portion of those who subscribed to the right of politics got disenchanted with their own party's polices and governance. This is an absolute gift for Australia, because it means that extremism is not tolerated, be it from the left or the right. And yet if you want to vote for a loony left, or "greenie", you can, and the same goes for hunters, shooters, and fishers, and even Pauline Hanson or Clive Palmer, if that is your wish.
Our system is far from perfect, but everybody gets a "say". And we know everyone gets a say because voting is compulsory. This is not "nerdy" or "academic" at all, it is democracy in practice.
We've been collectively battered by first past the post and we still wince when we see its shape elsewhere.
Hi from Alaska. I live in the interior city of Fairbanks and am graduating with my BS in math in two weeks. I never heard of this election! Thanks for the excellent coverage.
I got to vote in the election that established ranked based voting for our state, and it was a very close call!
The issue with voting in places with FPTP, like the UK and America, isn't actually FPTP, though it's made worse by FPTP and any other voting methods is better; the real issue is that you can't get proportionality by having tiny one-candidate districts. No matter what voting system you use, having small voting districts will still be skewed and broken, and even if you used, say, a Condorcet method (round robin where the candidate (if there is such a candidate) who wins all pair-wise competitions with all other candidates is the eventual winner) the result can even be possibly that a party takes all the seats.
Yes- we also have serious problems with Gerrymandering which would totally evaporate if we switched to systems that worked harder to ensure proportionality. This is not solved by RCV.
Okay, it's NP-hard, but I'm pretty sure N is small in this case... now, _explaining_ that on the evening news, that's a different kind of hard-problem.
Yes- for 3 or 4 candidates, it's really no problem. I don't know the work well enough to say at what point it becomes unfeasible.
@@ChrisStaecker Aye. If I had to guess there's an n! or similar hiding in there. Maybe 2^(n!)- which is gonna be great fun if we ever get up to 6 candidates in a race :-p
@@codertao Yes it is O(n!) to even read the voter data, since each voter chooses a (subset-)permutation to vote for, and there are more than n! choices for those. So basically any analysis of anything interesting about an RCV election will automatically be at least factorial complexity.
The general public is able to understand ncaa brackets, but rounds of voting are too complicated? :'(
There's a second-order effect to having ranked-choice voting that's good for democracy, too. If the red party and the blue party consistently get 90% of first-place votes, but the green party gets 10% of first-place votes... say if the first-round totals are close, like, red 46% and blue 44%, then the blue party gets the message that they should maybe adopt some policies of the green party. Ranking the green party first, though there may be no hope that green wins, is a way for the electorate to communicate what they want to see in government, a way to put a little pressure on the two main parties' policy-making.
At first blush, it would seem to me that simplest solution is to always recount. In other words, have two independent processes for counting done in series. The goal would be refine consistent counts: look for discrepancies, fix them and document them for regression testing. In software development, finding bugs is NP hard. Our solution is to make Continuous Improvement a part of the process. We try to create tests to catch issues, but we also add tests after regressions are found. This adds confidence and trust that the system operated as expected; thus, allowing us to believe the system did and will operate in exceptional circumstances.
Recounts aren't free to do. Not as expensive as a full election but recounting everything would be ridiculous and not required.
What do you mean by finding bugs being NP hard? What algorithm is there for finding bugs? Or did you just mean difficult in general?
There's that voice and wit! My UA-cam finally out you back in rotation thank God
Great video! To me this is another point in favor of Approval Voting (as opposed to ranked choice). You still avoid the spoiler effect (mostly), and its a lot easier to interpret the results at the end ("Who has the most approvals?")
I like approval voting a lot!
I dislike approval voting because it forces voters to strategize a lot, but it is an improvement over FPTP. Given that there is no perfect voting system and an improvement is needed badly, we should cooperate to make sure at least one of these solutions is adopted.
Here in Czechia, everything is either proportional or has a runoff round. It is decent, but could be improved.
Oh cool, I thought this was going to be that actual thing wrong with ranked choice voting that I'm unable to foresee. The idea survives another day of plausibility! Props for excellent citizen-reporting.
I scrolled through all the comments thinking I'd find at least one or two proponents of Majority Judgement or other highest median voting systems, but apparently there isn't, so I feel compelled to fill the niche.
With MJ you get to grade candidates instead of just ranking them. The system avoids Arrow's paradox (a minority third candidate can't cause the most popular candidate to lose to the runner up). And it provides rich statistics on voter preferences.
A major drawback is the fact that it uses the median, not the mean: I don't think the media are cognitively ready for medians.
Great video! We should be doing ranked choice voting for federal elections.
What provision does Alaska have for the case where, in a three-candidate race, there's a tie for second place? Or worse yet a three-way tie? How do they decide whose votes to redistribute to whom? Do they just carry on with round two including all three candidates again?
Typical rules for RCV will simultaneously eliminate all candidates who are tied for last place in any single round. But some different implementations use slightly different rules in the case of last-place ties. The official rules for the state of Alaska say that they resolve a last-place tie "by lot", i.e. randomly. (Coin-toss) It's never happened, as far as I know. See:
www.elections.alaska.gov/election-information/#RankedChoice
why use ranked choice tho when you could use star voting or approval voting which are both much cheaper/safer and less influenced strategic voting (which is the whole reason to choose a voting system in the first place!)
Approval voting is severely influenced by strategic voting!
Fantastic presentation and analysis. I'm wondering if there is a nice visual that can be generated so that we can see 1st round AND 2nd round outcomes at the same time so that "Close" won't appear close (if it isn't close, as in this case). Hmmm. Have to think about that.
Edit: As pointed out by another commenter, there is no need to show the first round results. That's the answer. Easy peasy. :-)
It's worse than that- it is possible in some examples that the final round seems decisive, even if some serious contender was eliminated in an early round.
What we have in the video is a race that isn't close, even though it seemed close in one of the early rounds.
But it is possible for a race to ACTUALLY be close, even though each individual round appears decisive. This IMHO is much worse, since nobody would ever notice unless they have access to the full data, and they actually go through whole the NP-hard calculation.
This paper here gives a (made-up) example where it happens:
adsarwate.github.io/assets/pdf/papers/SarwateCS13irv.pdf
They use 6 candidates... I think it's probably not possible to do something similar with only 3 or 4. (Alaska always limits their RCV ballots to 4, so it won't be a problem there.)
You tell em!
3 Remarks:
A: It's not stupid for Cacy to demand a recount, *IF* her goal was to give Holland the win over Giessel. (She would still have had to find enough votes that Giessel gets eliminated in the first round, but with 140-ish votes out of 16k, that's at least close enough to the 20 vote threshold that it seems somewhat plausible to me.)
B: Well, if the media had realized that this was a waste of time, they would ordinarily respond by not wasting their time on reporting on it.
C: Something being NP-hard doesn't mean that something is near impossible. It just means that it's guaranteed to get harder _at a ridiculous pace when you add more candidates._ It's most definitely quite easy in the case of 3 candidates. ("You can do it by hand on a napkin" kinds of easy.) And even with 20 or so, it might just take a computer program 10 minutes to brute force the solution.
A: It's unreasonable for a recount to swing anywhere near 140-ish votes in a field of this size. More likely swings of single-digits. And a cursory look at the politics shows that there's no way Cacy would've wanted Holland to win the election. Cacy and Holland are opposite ends of the spectrum, with Giessel more moderate. (You can see almost no Cacy voters list Holland as their second choice.)
B: To me it seems worth reporting that a useless recount was happening.
C: This is absolutely right- the NP-hard thing is mathematically interesting, but not necessarily very important for real political elections. In Alaska they already limit the final RCV ballot to 4 candidates, so their true margin of victory will always be computable with no real problem.
The most important fact is increased choice for voters, without one or more run-offs. And sure, if you look at it from a politically "religious" standpoint, you are not going to appreciate the subtlety of ranked choice. For example, if a Republican of Democrat is an axe-murderer, and you don't care about that, then you will vote R or D anyway.
Under preference voting, you can choose from a field of candidates, not just two. Some are going to be closer to your personal position than others. And preferential voting encourages more candidates to run, which is healthy politically. There is no reason why a field of candidates such as a number of Dems, GOP, and independents can't run. There is no downside. If you are worried about election efficiency, then have compulsory voting, no political financing, and make all candidates really work for those votes. "Name recognition" is not the electoral system's problem, it is the voter's responsibility.
In Ireland we have a single transferable vote system in which the vote count in actually non-deterministic. Yep. It's pretty great, actually. And it works because
1 - there is a huge amount of transparency during the (manual) count
2 - it turns the counting of votes into a spectator sport that people like to bet on
3 - it succeeds in bringing a diversity of 3rd party and independents into government
4 - it turns out that, in practice, recounts very rarely matter
5 - we have *lots* of candidates (in my old district I had 20 choices)
I think that 1 & 2 above are the important ones. The government tried to bring in automated counting and people protested in the streets. Irish people understand the logic of the system, but they want to see it happen. Trust *and* verify. Always
How is it nondeterministic?
@@ChrisStaecker Most of the non-determinism comes in because our parliamentary elections use multi-seat districts to allow for proportional representation. In Co. Wicklow, we had five seats. This means that in order to be elected a candidate had to receive 1/5 of the total vote.
If in a round of counting there is a candidate who has a surplus of votes over the quota, the extra votes _as determined by the order of the count_ all get transferred to their next preference down on the list, to be counted in the next round. If no candidate is elected, the candidate with the least votes gets eliminated and all of their votes get transferred to the next candidate down on the list.
Between the order-dependence of the count and the reshuffling that occurs due to transfers it becomes entirely possible that different counting orders could result in different outcomes. But every vote gets counted in some *positive* sense towards the outcome. I believe that the Australians use a variation of the same concept.
For a more detailed (and accurate) description:
www.citizensinformation.ie/en/government-in-ireland/elections-and-referenda/voting/proportional-representation/
@@kumoyuki Interesting- and what determines the order of the count? Just the order in which votes are cast?
If they need to recount, do they somehow recount them in the same order?
@@ChrisStaecker I've never gone to personally witness my local count, but a surprising number of people do. That might be because there are so many candidates - in Ireland the chances are good that you know somebody on the ballot :)
I believe the ordering is determined on the first count, which is when they figure out how many ballots were actually cast. I never really thought much about repeatability in the count. I guess that establishing a total ordering across all the ballots makes the count repeatable, even if it might produce a different result under a different ordering.
I was invited to a standards committee shortly after I moved to Ireland and one of our first tasks was figuring out how to vote on proposals. This led me into a deep rabbit hole on the mathematics of voting systems: it's a really fascinating field. I'm fairly proud of the Irish system because it has nearly all of the theoretical properties required to really establish a fair vote that is resistant to tactical voting strategies. It's just non-deterministic ;)
How would you know for sure that holland's voters didn't have a higher proportion of 2nd choices being casy? It seems like an unproven assumption that the voters for the other 2 candidates would've had the same split in their ranked choices?
The votes are public information, so we do in fact know exactly how many of Holland’s voters will break for Giessel vs Cacy. None of what I say around 2:40 involves any hypotheticals or conjecture- the numbers come straight from the actual votes.
@@ChrisStaecker oh they were known even before the recount? Makes sense then.👍
That's crazy that it came so close to a third each in the first round. I don't know what it means, but it's interesting
I think there's a practical solution to the problem that calculating the margin of victory is difficult: require the candidate requesting the recount to prove the election could have been swung by twenty votes or less (or whatever threshold one uses) if they want the state to pay for the recount. This doesn't even require the candidate to find the actual margin of victory, just prove that it's less than or equal to twenty votes
Yes- this is a good idea. The recount trigger rules should be phrased like "margin of victory of 20 or less", rather than "difference in votes of 20 or less". And I was being a bit overdramatic on the NP-hard point. The calculation becomes intractible when the number of candidates is high, but for 3 or 4 candidates it is perfectly doable.
Even so, like you suggest, there are other efficient (not NP-hard) algorithms which can compute upper bounds for the Margin Of Victory, which is often good enough. In many cases it's not necessary to compute the true MOV.
*specifically* in a politically environment with 2 parties (or two major parties) I'd argue that a simple color coded pie chart will represent the electorates preferences very well. A third is blue, a third is red, a third is a very slightly different red. Who win? One of the reds, let's find out which in the next round
The ranked choice problem has to be NP hard when there are a bazillion candidates, not just 3. It was a silly recount.
Yes "NP-hard" is a bit of a red herring in cases like this. It only means that the problem becomes intractable as n increases. For n=3, in this case, no problem.
The issue with ranked choice voting is not it’s structure. It’s getting your normal joe to understand and trust how it works.
Seeing how much people can’t even trust FPTP elections today, I can see why RCV is running into issues like this.
If you were to create a vote-counting-methods-appreciation-and-discussion club... how would you rank the different vote counting methods for selecting the club's president?
I have a hard time having strong feelings about it. To me, plurality seems fairly bad. I like approval voting for using in real-world low-stakes decisions. For repeated low-stakes decisions I like the "random dictatorship".
For more than 2 candidates, the various ranked methods are a little too complex for me to feel strongly about. I feel like RCV (IRV) generally works well, but I can't strongly prefer it over Coombs' method. And I can invent examples where RCV makes "bad" choices.
@@ChrisStaeckerIn most important cases I wouldn’t recommend this option, due to the complexity, but, if all the candidates and voters are enthusiasts about voting systems, I think Ranked Pairs would be nice?
Given each voter’s rankings, for each pair of candidates, compute a margin of victory between that pair, and then sort the pairs from greatest to least margin of victory. Then add edges to the digraph one by one according to this list, skipping any that would create a cycle. (Or, like, if it would make a cycle, put it in the opposite direction I guess)
Love ranked choice voting as an Australian. Arguably even more important is voting being mandatory. The US has a serious voter apathy problem which leads to incredibly skewed election results, since only a third or so of the country even votes at all.
Plus, since its a requirement that everyone votes, access to polling places (and early voting options) is assured. If a community loses access to voting for whatever reason (racism, an attempt to skew the results by district, etc.) then it's a federal issue that will be resolved.
Australia votes on a Saturday too. Voting on a work day, as they do in the US and UK is just crazy.
@@tuppyglossop222 Do they really? That's even crazier honestly. I always thought election days should be public holidays anyway, to make it even easier for people to attend.
NP-hard probably doesn't matter because you don't need an efficient algorithm unless there are tons of candidates. Digging through n! scenarios is no big deal as long as the number of candidates is smaller than maybe eight to twelve or so.
Don't judge too hard.Preference voting is great, unfortunately too lees used. Awesome Alaska used it, hurray!
Now you explained it kind of backwards, blaming them for making a bad decision. Let's look at it with a lessons-learned view:
• Recounts are an important tool, having a reasonable low barrier is important.
• The recount law was made for non-preference voting. Let's extend it to cover more voting systems properly.
• Most people, excluding mathematicians and voting systems fans where confused. Journalists are usually good with words, and not with the actual topic they are reporting about (look at a car magazine for examples). Don't blame them, help them.
•• First round sums are not the result, so do not look at them, look at the firsts, seconds and so on.
• Politics is a game where you have to know the small details and play all the rules. Don't blame the guy who found and used the obscure rules.
• Casy may not have understood that there where no chances of winning with a recount. Anyways, it's democracy and rules....
Was the recount actually bad? Sure, Cacy had nothing to gain, but voters have an interest in having their preferences reflected in the outcome. To do this means that it is important to determine if Giessel lost in the first round because then Holland would have been the winner. Of course this is very unlikely because Giessel was the frontrunner, ~100 votes above Holland not to mention Cacy.
The recount didn't hurt anybody other than wasting some time and money, so it wasn't actually really bad. But it also had no chance of producing any useful information, in my opinion. The rounds themselves are just mathematical abstractions, with no individual meaning. So saying "we really want to be sure who won and lost the first round, even if it doesn't change the end result" isn't really getting at anything important anyway.
@@ChrisStaecker But if Giessel lost in the first round, it does change the end result.
@@001HK0 OK I agree if Giessel actually lost the first round then that really is meaningful. But the margin there (Holland vs Giessel) was not small enough to justify a recount. Certainly not a state-paid auto-recount.
I feel like the maths (while NP hard, and that's clearly a high bar for hardness) is the easy bit. What's completely unknowable here is what would have happened if there were an equal number of democrat and republican candidates, or 2 democrats and 1 republican.
I don't know (I'm from the UK) how candidates are selected, but I do (think I) know that over here the number, and spread, of candidates standing often skews the situation.
Personally I suspect most voters vote along party lines, so this case would be a legitimate republican win. Wouldn't be my my choice, but democracy is (at least some sort of) democracy.
I would love for unbiased statisticians to be in charge of the whole process, somehow aggregating votes over policy issues, "values" or somesuch, but I'm not sure there are any unbiased statisticians - and even if there were, the voting population probably wouldn't understand or accept them.
i don't know how you can calculate the number of votes Cacey would get if Giesel lost the first round, or if Holland lost the first round. Personally, I'm neither R or D, but independent. It may be that if Geisel lost, then Cacey would have won if the women who voted for Geisel voted for Cacey, because she is a woman, party affiliation aside. When Mr T was running and Bernie was on the ballot, here in the northeast, there were a lot of young people that said they'd vote for Burnie, but if he didn 't make the cut, they'd vote for Mr. T. (Of course in NH, Bernie DID win the primary, but the superdelegates overrode the the popular will and voted for Hillary in the electoral college.) But I'll freely admit that I am a dumbass : - ) As an aside, in a local election here we had an alderman lose an election by just 5-10 votes; a recount ensued and he won by a couple votes, another recount and he lost (again.) Finally, after a third recount, he won....... wtfork is goin' on?
We can indeed calculate that number of votes because all of the vote information is known- not just the "round totals", but we know exactly how many people voted Giessel 1st with Holland 2nd, vs Giessel 1st with Cacy 2nd, vs etc etc. So there is no ambiguity about how various votes would've shaken out in different scenarios- we can figure it out exactly.
What I'm saying at 2:40 with the specific redistributing of votes is not my own guess- we know that this is exactly how it would've happened, because the votes say so.
@@ChrisStaecker I assumed as much but to be fair to the original commenter this wasn't explicit in the video
Second place is still a looser, she would have needed 35 votes to win.
Lots more than that.
It's a terrible and unnecessary system anyways, there's not a lot of benefit in analyzing which implementation of it is better than the other and what the general mass thinks about one or the other.
What do you prefer? Or are you an "all voting systems are terrible" kinda guy? (I can respect that)
All universal-suffrage systems are terrible. The opinion of the general mass (which includes me by the way) on who's in power has been proven time and again to only be useful as a destabilizing factor and to promote populism.
The only voting system for a country's leader that has benefit in it is the one allowed by the shari'a, where the ruler is elected by "decision makers" i.e. those who's opinion has weight in the land.
And this system is natural to human societies, when the people who rule the land are the ones which the "decision makers" of that land don't favor - the entire society suffers. When the people who rule the land have support from the "decision makers" things tend to go smoothly.
With that in mind, it's obvious that the more people are allowed to vote the less likely it is for the better outcome to occur. And if you will look at the history of the gradual expansions of suffrage you will see that those who had pushed for its expansion did it for that exact purpose - to sacrifice the wellbeing of the society for the sake of the populists.
@@pubcollize Well that took a turn...
@@ChrisStaecker Snark doesn't make one smart. The smart move would be to try and prove me wrong, which you won't be able to because what I have presented to you is a rule set by our creator for the benefit of mankind.
"Should not He Who has created know? And He is the Most Kind and Courteous (to His slaves), All-Aware (of everything)."
67:14
Calculus had been around and known quite well since before the failed experiment called "democracy" was brought about in the American and French revolutions. And I expect someone like you to know calculus more than a great deal of the population knows multiplication. Doesn't it bother you that over two centuries later no one can even figure out how to deal with the results of a vote?
Gather up the best mathematicians, social sci, pol-sci geniuses of all times - bring them in the same room together and even if they'd manage to agree on a single best system it's still gonna fail miserably even after a thousand iterations and you know it.
@@pubcollize I wouldn't presume to try and prove you wrong. I assumed your objections were for mathematical reasons- but now I see that you are coming from a very different point of view than I am. I am a person of faith too, and here in the USA my tradition has produced a crop of folks who want our government to be based on their version of "rules set by our creator". I am 100% opposed to this.
As for "Doesn't it bother you?" Yes it does! For some time it has been demonstrated that there can be no mathematically perfect ranked voting system. This is unsettling, but absolutely does not make me want to abandon universal suffrage.
With enough candidates there will always be at least one "close" round. Easy content.
I find the mind boggling part of this entire story to be that the Bars in your graph 📊 … ARE BLUE … 🤔 are they just as sad about this? Was it a graphic beyond your control? Or! was it the graphic of some liberal media outlet? 😅
I had to make all the graphics myself because THE MAINSTREAM MEDIA DOESN'T WANT YOU TO SEE THIS INFORMATION!