• Welcome to the Internet Infidels Discussion Board.

What makes a voting system fair?

Blahface

Member
Joined
Apr 1, 2008
Messages
269
Location
Illinois
Basic Beliefs
Atheist
What do you think the important factors are in determining what is a fair voting system? Below is a list and description of some voting system criteria. I also list whether or not approval voting, score voting, instant runoff voting, and ranked pairs violate each criterion. Please rate on a scale of 1 to 10 the importance you place on each.
1 - meaning that it doesn't matter at all if it passes the criterion.
10 - meaning that it should always pass this criterion.

Cancellation Criterion
Any one voter should have the ability to cancel out the effects of any other single voter.

Passes: Approval, Score, Ranked Pairs
Fails: IRV

Condorcet Winner Criterion:
If there is a candidate that beats all other candidates in a pairwise contest, that candidate will win.

Passes: Ranked Pairs
Fails: IRV, Score, Approval


Condorcet Loser Criterion:
If there is a candidate that is defeated in all pairwise contests, that candidate cannot win.

Passes: Ranked Pairs, IRV
Fails: Score and Approval

No Favorite Betrayal Criterion:
Voters should have no incentive to not maximally support their favorite candidate.

Passes: Approval, Score.
Fails: IRV and ranked pairs if there is no Condorcet winner.

Later No Harm Criterion:
Rating up or giving a positive rating to a lesser preferred candidates cannot cause the more preferred candidate to lose.

Passes: IRV.
Failed: Approval, RP fails if there is no Condorcet winner.

Participation Criterion:
When adding ballots to an existing tally in in which the ballots show preference of candidate A to candidate B, the addition of the ballots should not change the winner from candidate A to candidate B.

Passes: Approval, Score
Fails: IRV and RP if there is no Condorcet winner.

Independence of Clones Criterion:
If you have two candidate that are so close that they are effectively the same, the removal of one from the race must not increase or decrease the other's chance of winning.

Passes: IRV, Ranked Pairs, Score, Approval.
Fails: none
 
Last edited:
these are my ratings

Cancellation Criterion: 6
Condorcet Winner Criterion: 7
Condorcet Loser Criterion: 9
No Favorite Betrayal Criterion: 9
Later No Harm Criterion: 7
Participation Criterion: 8
Independence of Clones Criterion: 9
 
Independence of Clones Criterion:
If you have two candidate that are so close that they are effectively the same, the removal of one from the race must not increase or decrease the other's chance of winning.

Passes: IRV, Ranked Pairs, Score, Approval.
Fails: none
That makes no sense. Suppose there are three candidates, A, A' and B, and nearly everyone prefers A and A' to B, and is indifferent between A and A'. Removing A' will increase the chance A wins from 50% to 100% under all those systems, as well as every other sane system. I think you've misstated the criterion.
 
Independence of Clones Criterion:
If you have two candidate that are so close that they are effectively the same, the removal of one from the race must not increase or decrease the other's chance of winning.

Passes: IRV, Ranked Pairs, Score, Approval.
Fails: none
That makes no sense. Suppose there are three candidates, A, A' and B, and nearly everyone prefers A and A' to B, and is indifferent between A and A'. Removing A' will increase the chance A wins from 50% to 100% under all those systems, as well as every other sane system. I think you've misstated the criterion.
Indeed. The independence of clones criterion is that removal of a clone must not increase or decrease the other, non-clone candidates' chances of winning.
 
That makes no sense. Suppose there are three candidates, A, A' and B, and nearly everyone prefers A and A' to B, and is indifferent between A and A'. Removing A' will increase the chance A wins from 50% to 100% under all those systems, as well as every other sane system. I think you've misstated the criterion.
Indeed. The independence of clones criterion is that removal of a clone must not increase or decrease the other, non-clone candidates' chances of winning.

Yes, that is the correct criterion. My bad.

In my defense though, I posted that late at night right before going to bed as I am also doing right now.
 
Guys, the only fair voting system is the one where the results are the ones I prefer.
 
The criteria seem to be geared to the crazy US/UK voting system of one return per constituency. This is fairly exotic to a Swede, so I won't Google, but a few links to explain the terms
approval voting, score voting, instant runoff voting, and ranked pairs
might have made me look into the alternatives.
 
I prefer ranking every candidate by every voter from one to n. Those who fail to do so must re-vote until they include all candidates ranked. The winner is determined by highest ranking totals.

This way every candidate is assured of being given attention by every voter.
 
I prefer ranking every candidate by every voter from one to n. Those who fail to do so must re-vote until they include all candidates ranked. The winner is determined by highest ranking totals.

This way every candidate is assured of being given attention by every voter.

Sounds like the system we have in Australia.
 
I prefer ranking every candidate by every voter from one to n. Those who fail to do so must re-vote until they include all candidates ranked. The winner is determined by highest ranking totals.

This way every candidate is assured of being given attention by every voter.
Yes. Excellent.

Two criteria though:

1) Number n should be equal or greater than the number of candidates, to prevent voters being forced to grade any two candidates the same;

2) Number n should be equal or superior to 10 times the number of candidates (or more) to allow voters to mark appropriately their possibly very strong likes and dislikes as they see fit.
EB
 
I prefer ranking every candidate by every voter from one to n. Those who fail to do so must re-vote until they include all candidates ranked. The winner is determined by highest ranking totals.

This way every candidate is assured of being given attention by every voter.

If you force people to rank all candidates, a lot of people might just not want to mess with it and not show up to the polls. If Australia didn't force people to vote, I can't imagine that their turnout would be that great – that's even considering that they give them the option to let the party decide the rankings. Another problem with that in combination with the Borda count is that people will just bury the candidate they know they don't like and that could lead to somebody accidentally getting elected who is worst candidate for that constituency.
 
I prefer ranking every candidate by every voter from one to n. Those who fail to do so must re-vote until they include all candidates ranked. The winner is determined by highest ranking totals.

This way every candidate is assured of being given attention by every voter.

If you force people to rank all candidates, a lot of people might just not want to mess with it and not show up to the polls. If Australia didn't force people to vote, I can't imagine that their turnout would be that great – that's even considering that they give them the option to let the party decide the rankings. Another problem with that in combination with the Borda count is that people will just bury the candidate they know they don't like and that could lead to somebody accidentally getting elected who is worst candidate for that constituency.
You are in effect arguing that the only truly democratic voting system would be too dangerous to implement, people being apparently too stupid or too lazy to be given the responsibility to elect representatives.

Also the possibility you are describing, that the worst candidate be elected, is very, very unlikely to occur. This would require that no or very few voters actually support somebody else than the worst candidate AND that in any event more voters ACTUALLY support the worst candidate. Just ranking last the ones you don't like won't produce this result. In particular, this is very unlikely to occur in larger elections, not least because you can count on party members and party supporters to vote for their candidates. And if only the worst candidate happens to have supporters then the place really deserve to have him as leader.

So there is another principle I want to add to what is a good voting system, which is that mechanisms to ensure that the worst candidate couldn't be elected shouldn't deny voters the right to have the only democratic voting system that exists.
EB
 
Since nobody so far has bothered to favour totally different systems than US/UK, I'll have a go.

Fairness in local or national voting is, to me, when every party will get the number of seats in the respective parliaments that corresponds to their percentage of the total votes. Who will get elected is decided by the candidates' position on the ballot*).

Roughly: In a certain community, party A gets 40% of the votes, B gets 30%, C collects 20% and the five other parties share the rest. Suppose there are 50 seats. A: 20 (the first 20 people on their ballot of perhaps 50 candidates), B: 15, C: 10 and the final 5 are distributed similarly. There might be a percentage threshold that the 5 smaller parties manage, and if they're sufficiently close, each will get one representative.

The proportionality will, for example, cancel out any efforts at gerrymandering. It's the total, not the results from individual constituencies, that matters. For nation-wide voting, a party that is strong in a limited geographical area but presents no ballots anywhere else, will return MPs if the local total gets them a sufficient percentage of the national total.

*) There used to be possible to cross out candidates you don't like, to disturb the party official order. Possible use: cross out all males to increase female representation. Now replaced/amended? by a possibility to explicitly vote for one (1) candidate. A candidate that the official party decision has been placed in just below the expected cut-off, could (and have managed to) by popular vote oust even pretty senior contestants higher on the ballot.

Please comment if this makes (no) sense to the system(s) you're acquainted with.
 
Forcing people to rank candidates has the downside, that in lack of a strong opinion, people might just default to whatever order the candidates are listed on the ballot. I think in ranking there should always be an option to lump all irrelevant candidates in one bin, or as in approval voting just say which ones are acceptable.
 
Forcing people to rank candidates has the downside, that in lack of a strong opinion, people might just default to whatever order the candidates are listed on the ballot. I think in ranking there should always be an option to lump all irrelevant candidates in one bin, or as in approval voting just say which ones are acceptable.

Easy to fix. Use electronic gaming process to randomize names on ballots as voters vote. Oh, yes, require elections to run until all voters have voted. Begin to fine voters after 7 days for not having voted yet.
 
Forcing people to rank candidates has the downside, that in lack of a strong opinion, people might just default to whatever order the candidates are listed on the ballot. I think in ranking there should always be an option to lump all irrelevant candidates in one bin, or as in approval voting just say which ones are acceptable.

Easy to fix. Use electronic gaming process to randomize names on ballots as voters vote. Oh, yes, require elections to run until all voters have voted. Begin to fine voters after 7 days for not having voted yet.
So candidates that would have, if there were no compulsory ranking, be counted as equally good or bad, would get roughly equal number of votes from people who just revert to the default ranking. A complicated solution to a problem that was never a problem in the first place: all this would do is introduce a new failure mode and risk of tampering with the process without any gain whatsoever.
 
Back
Top Bottom