Consider this outcome:
| votes | 1st place | 2nd place | 3rd place |
| 9 | Biden +18 | Warren +9 | Sanders |
| 8 | Sanders +16 | Biden +8 | Warren |
| 7 | Warren +14 | Sanders +7 | Biden |
By IRV, Warren is eliminated, and Sanders wins, 15:9 over Biden. By Ranked voting, the final count is Biden 26, Sanders 23, Warren 23, and Biden wins.
This certainly seems to represent the votes better than IRV. Or consider this:
| votes | 1st place | 2nd place | 3rd place |
| 9 | Biden +18 | Warren +9 | Sanders |
| 8 | Sanders +16 | Warren +8 | Biden |
| 1 | Warren +2 | Sanders +1 | Biden |
The count is Biden 18, Sanders 17, Warren 19. Warren wins. Even though few wanted her for first place, everybody had her for first or second. Wouldn't this be a better outcome than the other options, where half the people would get their last choice?
For this reason, ranked voting favors candidates with broad appeal. Candidates that alienate 1/3 of the electorate will struggle against those who do well across the board, especially in races with more candidates.
Ranked voting does not suffer from the Love-Hate Paradox. Looking at the first example above, scoring 2 points for last place, the count would be Bernie 25, Warren 25, Biden 22: The candidate Ranked voting chooses first for "best" is chosen last for "worst." In the second example, the count is Bernie 19, Biden 18, Warren 17: again showing that the "best" candidate places last when choosing the "worst." Neither does the Participation Paradox function: there is no way that not voting will help your candidate.
Cloning
Now we get to Ranked voting weirdness. Consider this election for favorite color:| votes | 1st place (x3) | 2nd place (x2) | 3rd place (x1) | 4th place (x0) |
| 4 | Yellow +12 | Red +8 | Green +4 | Blue |
| 3 | Red +9 | Green +6 | Yellow +3 | Blue |
| 3 | Green +9 | Blue +6 | Yellow +3 | Red |
| 2 | Blue +6 | Red +4 | Yellow +2 | Green |
Yellow: 20
Green: 19
Blue: 12
Fair enough. But what happens when olive green and pale green are added to the election, like this:
| votes | 1st (x5) | 2nd (x4) | 3rd (x3) | 4th(x2) | 5th (x1) | 6th (x0) |
| 4 | Yellow +20 | Red +16 | Green +12 | Olive +8 | Pale +4 | Blue |
| 3 | Red +15 | Green +12 | Olive +9 | Pale +6 | Yellow +3 | Blue |
| 3 | Green +15 | Olive +12 | Pale +9 | Blue +6 | Yellow +3 | Red |
| 2 | Blue +10 | Red +8 | Yellow 6 | Green +4 | Olive +2 | Pale |
Red: 39
Yellow:32
Olive: 31
Pale Green: 19
Blue:16
Nobody changed their opinion about whether they liked Red or Green better. But the addition of other green options to the election caused green to rise from 3rd place to 1st!
Weird, yes. But worse than "weird," because this means you can game the election. The winner becomes not so much who is the favored choice of the people, but who can best create this slate of candidates that will be consistently voted in an ordered block. An election of Conservative vs. Moderate vs Liberal 1, Liberal 2, and Liberal 3 will almost always go to Liberal 1. This is because when voters vote liberal, the best the Conservative or Moderate can do is 4th place, but where they don't vote liberal, the worst Liberal 1 can do is 3rd place.
Spoilers
Ranked voting is also subject to Spoiler effects. Suppose a candidate with no hope of winning draws some votes from Red. Pink perhaps:| votes | 1st place (x4) | 2nd place (x3) | 3rd place (x2) | 4th place (x1) | 5th place (x0) |
| 4 | Yellow +16 | Pink +12 | Red +8 | Green +4 | Blue |
| 3 | Red +12 | Green +9 | Yellow +6 | Blue +3 | Pink |
| 3 | Green +12 | Blue +9 | Yellow +6 | Red +3 | Pink |
| 2 | Blue +8 | Red +6 | Yellow +4 | Green +2 | Pink |
Red: 29
Green: 27
Blue: 20
Pink: 12
Again, no one changed their opinion of Red, Green, or Yellow. But the 4 Yellow voters, by liking pink better than red, caused Yellow to win.
Dark Horse voting
This is a "deliberate spoiler" trick, although it comes with risks. Yellow can use the hopeless Blue to win the election:| votes | 1st place (x3) | 2nd place (x2) | 3rd place (x1) | 4th place (x0) |
| 3 | Yellow +9 | Blue +6 | Red +3 | Green |
| 1 | Yellow +3 | Red +2 | Green +1 | Blue |
| 3 | Red +9 | Green +6 | Yellow +3 | Blue |
| 3 | Green +9 | Blue +6 | Yellow +3 | Red |
| 2 | Blue +6 | Red +4 | Yellow +2 | Green |
Red: 18
Green: 16
Blue:18
Again, people's opinions of, and votes for Red vs. Yellow haven't changed. But by voting 2nd place for Blue, Yellow make Red lose. There is a natural limit to this game, obviously. If a couple Red voters had tried to do the same thing, Blue would have won, which was the least desirable outcome for the Red and Yellow voters. So people aren't likely to do it to extremes. Yet it is easy to imagine that on rare occasions, people will vote up an obvious loser who isn't as much a loser as they thought, and elect the candidate they most abhor.
Conclusion
Ranked voting has a number of advantages over IRV. It maintains the appearance of the winner being the actual favorite, and it doesn't produce contradictory results. However, it opens the door to lots of electoral gamesmanship. The last thing a worried electorate needs is an electoral system with candidates accusing each other of gaming the system.Fortunately, there is yet another alternative. Next up: Range voting.
No comments:
Post a Comment