While I appreciate the topic and it's an interesting thesis, I don't buy it. One of the goals of a voting system should be to reduce the need to vote strategically, so that voters can just express their honest preferences. After this election, a lot of voters who most prefer Palin have to be asking themselves if they should rank Begich first next time around, to give an artificial sense that he has a "strong core of support". (This is analogous to how many voters in a closed primary consider "electability".) If voters get better outcomes by voting strategically as blocs, that's not going to help us move past the two-party system.
That all said, I do hope that this particular questionable outcome and the general shortcomings of instant runoff voting don't sour the public on all types of ranked voting, or more generally on structural reforms to our democracy that are badly needed.
Hi Steven. I agree with your article in the main, but not your assessment of STAR Voting. Without good core support a candidate will lose in the instant runoff. The mix of the two types of support are different from RCV, but I prefer the mix in STAR.
Ranked Choice Voting failed in Alaska, just as we all predicted it would. It failed to elect the candidate preferred by the voters. It failed to prevent the spoiler effect. It failed to make it safe for voters to honestly rank the candidates. It failed to prevent vote-splitting.
If the voters prefer candidate A over candidate B, and your voting system chooses B, your voting system is fundamentally broken.
Broad support seems to be ignored in this case. Please consider giving each candidate some version of a score for core support and broad support. Then combine those scores for each candidate to make some version of a net score. These scores could determine the winner outright or perhaps could determine who is eliminated or who proceeds to the next round. Please see numbers below at end of this comment for three examples of various ways to create a net score.
Using only core support as a means to determine who is eliminated in a particular round appears to me to violate the criterion that both support levels should be considered. Note that Begich got 28% first ranking, which is fairly respectable in a three-candidate comparison, and Palin only did about 3% better than that. I would suggest that based on overall considerations, Palin should have been the one eliminated. To completely ignore the broad support of Begich (71.7%!) effectively exaggerates (over-emphasizes) the core support.
Method 1:
Percentage ballots with 1st or 2nd place votes (in essence, a combined measure of core support and broad support):
Begich Palin Peltola
71.7% 48.0% 50.2%
Method 2:
Begich Palin Peltola
% score for 1st place votes 28.5 31.3 40.2
% broad support: % 2nd place votes* 60.4 24.4 16.9
Combined score 88.9 55.7 57.1
( * What % of remaining ballots besides those with first place votes did a candidate receive 2nd place votes?)
Dec 13, 2022·edited Dec 13, 2022Liked by Steven Hill
One thing that I haven’t seen in any of the comments or analysis yet is: if a different type of election system is implemented, wouldn’t the type/mix of candidates likely change as well?
Assuming that a different system (~especially~ once established for several cycles) would result in similar candidate behavior/ positions, let alone similar lists of candidates, seems … rather imprecise.
- - - -
(Perhaps considering that could make it hard to sustain arguments for/against any particular system.)
People who might ~never~ consider running for office in ‘first-past-the-post’ or RCV environments, for example, might be willing to run under another scheme such as ‘approval voting’. Likewise, divisive candidates who currently do well in FPTP-style elections might moderate themselves or choose a different career path altogether.
Any recommendations for existing writings comparing different electoral systems ~without~ assuming that candidate pool would remain relatively equivalent?
1. "Core support" is a concept that hasn't been properly defined. To define it such that it can be intrinsically important we must define it in terms of cardinal preferences, not just in terms of how many voters ranked a candidate first.
2. Condorcet does value both core support and broad support - it's just more flexible about it than RCV is. Under Condorcet methods, a candidate still benefits from having a lot of core support - but if they don't, they can still make up for it by having a lot of extra broad support. Under RCV, if a candidate lacks enough core support then their broad support is ignored altogether.
3. Condorcet embodies simple and intuitive values: that all voter's preferences should matter equally, and that candidates should be incentivized to care about all voters equally. The values of RCV ("A candidate who has virtually no core support, even though the candidate did have broad support, that candidate should not win. You need BOTH core and broad support to win an election under RCV.") are incompatible with values of Condorcet.
4. When I say that I prefer the values of Condorcet to the values of RCV, you don't need to "make a dash for the rest room". I prefer the values of Condorcet because I believe that incentivizing candidates to care about all voters equally will make a wide range of polarization-induced disasters, such as a civil war or democratic backsliding, less likely. If you convinced me that necessitating core support (rather than merely valuing it) leads to better consequences for society than optimizing for depolarization, that would convince me to endorse RCV values over Condorcet values.
What doesn’t make sense is to use two totally different and arbitrary methods for a post-hoc evaluation of the candidates, and then to likewise arbitrarily decide which of those two methods is a priori superior as an argument against the other one.
You haven't explained why you prefer Hare to bottom-two runoff or why you want to forbid equal-ranking except at the bottom.
Failing to elect the Condorcet winner when there is one is manifestly minority rule in a circumstance wherein majority rule instead would have been possible.
Failure of Frohnmayer balance opens up a crack into which capital can insert its wedge and thereby continue to prevent democracy and enforce the GHG emissions associated with endless war, as a consequence of which, humanity shan't have many more generations to survive.
Do you believe that it is the "most democratic" practice to take the candidate that would win the most votes in an election against Peltola AND win the most votes in an election against Palin and NOT elect that candidate? That sounds contradictory to democracy. In each of those pairwise elections among the entire voting population, Begich would have been chosen by more people.
Primaries let partisan diehards narrow the field before the rest of us get to weigh in. So does RCV/IRV. In each round of an IRV election, ONLY the 1st place votes are taken into consideration. Giving special significance to that "core" of support means that divisive candidates (including ones that would actively harm large subsets of the population) could win based on the fervor of their base. If everyone's vote is equal, why are the people in that "core" given more weight than the rest of us?
RCV is no antidote to partisanship.
Dems are happy with the Alaska result, but this can cut both ways.
In proportional representation IRV/RCV works better. More than one voice represents everyone. But if we are going to have a single politician represent the views of everyone in a geographical district, they should reflect consensus and broad support, not a "core" of supporters.
If we really want diverse, representative views present in policy making, we should use Citizens Assemblies (sortition) wherever feasible.
LOL, in the Final Word category: here is a tweet storm about results from a new study from Nicholas Stephanopoulos, Kirkland & Ellis Professor of Law at Harvard Law School, putting a cherry on the cake. Real world data bests public choice theory bunkum!
"I just posted this paper, written for a symposium at Washington & Lee, on the real-world record of instant-runoff voting. A key question about any voting method is how often it elects the "Condorcet winner," favored by voters over any other candidate. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4763372
In American IRV elections, the Condorcet winner was elected in 181 of 183 races over the last couple decades. Examining foreign IRV elections in Australia and Scotland, I find that the Condorcet winner was elected in 191 of 193 races.
These are very high rates: about 99%. The gains from switching to a Condorcet-consistent method that's guaranteed to elect the Condorcet winner are therefore small. From a Condorcet efficiency perspective, IRV is already excellent.
The paper also tackles the related question of how often there's a Condorcet winner in the first place. The answer is almost always: in 183 of 185 American IRV elections and 193 of 193 foreign IRV elections.
Contrary to the fears of public choice theorists, then, voters' candidate preferences are almost never cyclical. There's almost always a single candidate whom most voters prefer to any other candidate (in a head-to-head matchup). From this standpoint, voters look pretty rational."
While I appreciate the topic and it's an interesting thesis, I don't buy it. One of the goals of a voting system should be to reduce the need to vote strategically, so that voters can just express their honest preferences. After this election, a lot of voters who most prefer Palin have to be asking themselves if they should rank Begich first next time around, to give an artificial sense that he has a "strong core of support". (This is analogous to how many voters in a closed primary consider "electability".) If voters get better outcomes by voting strategically as blocs, that's not going to help us move past the two-party system.
That all said, I do hope that this particular questionable outcome and the general shortcomings of instant runoff voting don't sour the public on all types of ranked voting, or more generally on structural reforms to our democracy that are badly needed.
Hi Steven. I agree with your article in the main, but not your assessment of STAR Voting. Without good core support a candidate will lose in the instant runoff. The mix of the two types of support are different from RCV, but I prefer the mix in STAR.
Ranked Choice Voting failed in Alaska, just as we all predicted it would. It failed to elect the candidate preferred by the voters. It failed to prevent the spoiler effect. It failed to make it safe for voters to honestly rank the candidates. It failed to prevent vote-splitting.
If the voters prefer candidate A over candidate B, and your voting system chooses B, your voting system is fundamentally broken.
Broad support seems to be ignored in this case. Please consider giving each candidate some version of a score for core support and broad support. Then combine those scores for each candidate to make some version of a net score. These scores could determine the winner outright or perhaps could determine who is eliminated or who proceeds to the next round. Please see numbers below at end of this comment for three examples of various ways to create a net score.
Using only core support as a means to determine who is eliminated in a particular round appears to me to violate the criterion that both support levels should be considered. Note that Begich got 28% first ranking, which is fairly respectable in a three-candidate comparison, and Palin only did about 3% better than that. I would suggest that based on overall considerations, Palin should have been the one eliminated. To completely ignore the broad support of Begich (71.7%!) effectively exaggerates (over-emphasizes) the core support.
Method 1:
Percentage ballots with 1st or 2nd place votes (in essence, a combined measure of core support and broad support):
Begich Palin Peltola
71.7% 48.0% 50.2%
Method 2:
Begich Palin Peltola
% score for 1st place votes 28.5 31.3 40.2
% broad support: % 2nd place votes* 60.4 24.4 16.9
Combined score 88.9 55.7 57.1
( * What % of remaining ballots besides those with first place votes did a candidate receive 2nd place votes?)
Method 3:
Borda count scores:
Begich Palin Peltola
189332 149715 170750
One thing that I haven’t seen in any of the comments or analysis yet is: if a different type of election system is implemented, wouldn’t the type/mix of candidates likely change as well?
Assuming that a different system (~especially~ once established for several cycles) would result in similar candidate behavior/ positions, let alone similar lists of candidates, seems … rather imprecise.
- - - -
(Perhaps considering that could make it hard to sustain arguments for/against any particular system.)
People who might ~never~ consider running for office in ‘first-past-the-post’ or RCV environments, for example, might be willing to run under another scheme such as ‘approval voting’. Likewise, divisive candidates who currently do well in FPTP-style elections might moderate themselves or choose a different career path altogether.
Any recommendations for existing writings comparing different electoral systems ~without~ assuming that candidate pool would remain relatively equivalent?
I've written an in-depth response on my blog: https://voting-in-the-abstract.medium.com/rcv-and-core-support-e0d1780a9184
Some of my main points:
1. "Core support" is a concept that hasn't been properly defined. To define it such that it can be intrinsically important we must define it in terms of cardinal preferences, not just in terms of how many voters ranked a candidate first.
2. Condorcet does value both core support and broad support - it's just more flexible about it than RCV is. Under Condorcet methods, a candidate still benefits from having a lot of core support - but if they don't, they can still make up for it by having a lot of extra broad support. Under RCV, if a candidate lacks enough core support then their broad support is ignored altogether.
3. Condorcet embodies simple and intuitive values: that all voter's preferences should matter equally, and that candidates should be incentivized to care about all voters equally. The values of RCV ("A candidate who has virtually no core support, even though the candidate did have broad support, that candidate should not win. You need BOTH core and broad support to win an election under RCV.") are incompatible with values of Condorcet.
4. When I say that I prefer the values of Condorcet to the values of RCV, you don't need to "make a dash for the rest room". I prefer the values of Condorcet because I believe that incentivizing candidates to care about all voters equally will make a wide range of polarization-induced disasters, such as a civil war or democratic backsliding, less likely. If you convinced me that necessitating core support (rather than merely valuing it) leads to better consequences for society than optimizing for depolarization, that would convince me to endorse RCV values over Condorcet values.
What doesn’t make sense is to use two totally different and arbitrary methods for a post-hoc evaluation of the candidates, and then to likewise arbitrarily decide which of those two methods is a priori superior as an argument against the other one.
You haven't explained why you prefer Hare to bottom-two runoff or why you want to forbid equal-ranking except at the bottom.
Failing to elect the Condorcet winner when there is one is manifestly minority rule in a circumstance wherein majority rule instead would have been possible.
Failure of Frohnmayer balance opens up a crack into which capital can insert its wedge and thereby continue to prevent democracy and enforce the GHG emissions associated with endless war, as a consequence of which, humanity shan't have many more generations to survive.
Steven,
Do you believe that it is the "most democratic" practice to take the candidate that would win the most votes in an election against Peltola AND win the most votes in an election against Palin and NOT elect that candidate? That sounds contradictory to democracy. In each of those pairwise elections among the entire voting population, Begich would have been chosen by more people.
Primaries let partisan diehards narrow the field before the rest of us get to weigh in. So does RCV/IRV. In each round of an IRV election, ONLY the 1st place votes are taken into consideration. Giving special significance to that "core" of support means that divisive candidates (including ones that would actively harm large subsets of the population) could win based on the fervor of their base. If everyone's vote is equal, why are the people in that "core" given more weight than the rest of us?
RCV is no antidote to partisanship.
Dems are happy with the Alaska result, but this can cut both ways.
In proportional representation IRV/RCV works better. More than one voice represents everyone. But if we are going to have a single politician represent the views of everyone in a geographical district, they should reflect consensus and broad support, not a "core" of supporters.
If we really want diverse, representative views present in policy making, we should use Citizens Assemblies (sortition) wherever feasible.
LOL, in the Final Word category: here is a tweet storm about results from a new study from Nicholas Stephanopoulos, Kirkland & Ellis Professor of Law at Harvard Law School, putting a cherry on the cake. Real world data bests public choice theory bunkum!
https://twitter.com/ProfNickStephan/status/1769752230012272691
@ProfNickStephan
Mar 18
"I just posted this paper, written for a symposium at Washington & Lee, on the real-world record of instant-runoff voting. A key question about any voting method is how often it elects the "Condorcet winner," favored by voters over any other candidate. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4763372
In American IRV elections, the Condorcet winner was elected in 181 of 183 races over the last couple decades. Examining foreign IRV elections in Australia and Scotland, I find that the Condorcet winner was elected in 191 of 193 races.
These are very high rates: about 99%. The gains from switching to a Condorcet-consistent method that's guaranteed to elect the Condorcet winner are therefore small. From a Condorcet efficiency perspective, IRV is already excellent.
The paper also tackles the related question of how often there's a Condorcet winner in the first place. The answer is almost always: in 183 of 185 American IRV elections and 193 of 193 foreign IRV elections.
Contrary to the fears of public choice theorists, then, voters' candidate preferences are almost never cyclical. There's almost always a single candidate whom most voters prefer to any other candidate (in a head-to-head matchup). From this standpoint, voters look pretty rational."