Why IRV is Ill-Conceived

If you've read the explanation of Condorcet Voting at this website, you know that cyclic ambiguities are resolved by dropping pairwise defeats. Instant Runoff Voting, on the other hand, drops candidates on the basis of the current top-choice votes at each round of the runoff procedure. The IRV method of dropping candidates is too coarse because it often capriciously throws out some of your expressed preferences, and often those are your more important preferences that get discarded. The only preference IRV is sure to count is your first choice.

What would you say about a project manager who makes irrevocable decisions based on only a small fraction of the available information? That's what IRV does, when it eliminates candidates based only on a count of 1st place rank positions.

The result is that the compromise that you need can be immediately eliminated because your IRV traveling vote hasn't reached him or her yet. You vote is on your favorite when your compromise needs your vote to avoid elimination. Then your favorite loses to your last choice. Contrary to what IRVists claim, that isn't a contrived scenario.

Condorcet reliably counts every pairwise preference that you want to vote. Approval doesn't let you vote all of your pairwise preferences, but it reliably counts all the pairwise preferences that you consider important enough to specify. With Approval, you decide which of your pairwise preferences will be counted. With IRV, IRV's capriscious idiosyncracies decide that for you.

I liken Approval to a solid, reliable handtool. A rank method is an automatic labor-saving machine. Condorcet is a deluxe automatic machine. IRV is a shoddy, defective automatic machine. You'd be much better off with the reliable handtool than with the defective automatic machine.

Let me mention a few IRV failure examples:

Example 1, with three candidates:

40: A,B,C
25: B,A,C
35: C,B,A

Candidate B is the candidate who beats each one of the others pairwise. That is, candidate B would beat each one of the others in separate pairwise contests. Condorcet would choose B.

IRV immediately eliminates B.

IRVists sometimes try to justify that by saying that "favoriteness" is important. Then why don't they just advocate Plurality?

But if IRVists say that "favoriteness" is important, then ask them to justify this:

Example 2, with five candidates:

63: A,B,C,D,E
75: B,A,C,D,E
100: C,D,B,E,A
86: D,E,C,B,A
73: E,D,C,B,A

We can simplify this if we delete all of the preferences that IRV simply ignores for no good reason:

63: A,B
75: B
100: C
86: D
73: E,D

It looks a bit sparse after we delete the preferences that IRV never looks at!

Candidates A and E are eliminated and transfer their votes inward, and then C has the fewest votes and is eliminated.

But not only would candidate C beat each one of the others in separate pairwise contests, but C is the favorite of more people than any other candidate is.

That can be written more extreme (with the sparse format):

Example 3, extreme case:

50: A,B
51: B
100: C
53: D
49: E,D

Now, not only would candidate C beat each of the others in a two-candidate race, but now C is favorite to about twice as many voters as anyone else. And yet IRV still eliminates C, just as in the previous example.

These aren't contrived examples. These are typical scenarios. They work with a wide variety of numbers. All this case requires is that "favoriteness" taper away from the voter-median candidate. The two examples show the wide variety of numbers with which the scenario can happen.

Ask the IRVists to justify that.

IRVists talk about electing a majority candidate, but IRV makes a false majority. Condorcet protects majority rule. IRV will often violate majority rule, as it does in the three examples above.

We want to get rid of the lesser-of-two-evils problem that dominates voting in a conventional plurality system. IRV, like Plurality, will sometimes give people a strategic need to bury their favorite by voting someone else over him/her, to protect a lesser-evil from early elimination.

Approval never gives anyone incentive to vote someone else over his or her favorite. Condorcet meets the criteria listed for it on our technical evaluation page.

For the goals of majority rule, and getting rid of the lesser-of-two-evils problem, Condorcet and Approval bring genuine significant improvement. IRV doesn't.

IRVists like to say that sometimes it won't have the problems I've described, that IRV "guarantees" that sometimes you won't regret voting your favorite in 1st place. Guarantees that contain the word "sometimes" or "maybe" don't seem very reassuring.