Website Search

Sunday, December 19, 2010

Weak defense of IRV - Instant Runoff Voting in NC Court of Appeals oddities

Some say instant runoff voting did what it was supposed to do in the NC Court of Appeals Contest. For sure the IRV algorithm that sorts, allocates and eliminates votes is capricious in nature. IRV election is like a crap shoot. The top 2 candidates had 20% and 15% of first round votes. Thigpen led McCullough by 100,000 votes. After sorting, allocating, eliminating and reshuffling the IRV votes, McCullough and Thigpen had a near tie. The top two, Thigpen had 27.62% of the vote from all ballots cast, and McCullough had 27.96% of all votes. But with IRV, majority is redefined to 50% of "remaining" votes, and excludes about 800K voters. Thigpen lost his lead and now trails McCullough by around 6,700 votes. Close enough that Thigpen called for a recount. This was conducted by machine, not hand - last week. Results aren't in yet. 

Weak defense of IRV
Greensboro News and Record. Doug Clark. December 16, 2010.
It's disappointing that Duke professor Mike Munger calls our opposition to Instant Runoff Voting "absurd," and counters with arguments that are just plain wrong.
The IRV, which decides a winner in a multicandidate race by adding second- and third-choice votes when no one receives a majority, achieves a result "precisely the same as if we conducted two separate elections," says Munger.
Talk about an absurd statement. That would be true only if voter turnout wasprecisely the same in a second primary and no voter changed his or her mind about the candidates in the interim. But it's worse than that, because IRV effectively excludes a high proportion of voters from this runoff.
IRV was used in the special election for an appeals court seat, which drew 13 candidates. On Nov. 2, Cressie Thigpen led with 395,220 votes, or 20 percent. Doug McCullough was next with 295,619 votes, or 15 percent. As the top two, under the IRV rules their second- and third-place votes were added. Last week, five weeks after the election, the State Board of Elections announced that McCullough was now the winner with 544,023 votes to 537,445 for Thigpen.
This was, of course, not precisely like a separate runoff for another important reason: 1,943,771 people voted in this race on Nov. 2. But the votes for McCullough and Thigpen after the "instant" runoff totaled 1,081,468. What happened to the other 862,303? These were voters who did not mark McCullough or Thigpen as their first-, second- or third-choice candidates. So the "instant" runoff was like holding a second separate election in which 862,303 people who voted in the first election — 44 percent of the total — were excluded from participating. How does Munger think that yields precisely the same result?
Next, responding to our view that second- and third-choice votes should not count as much as first-choice votes, Munger says "the idea that the votes are second-place or third-place votes is balloon juice." I don't know what balloon juice is — hot air, maybe? — but I do know that the ballot instructed voters to mark their second and third choices as well as their first choice. So they indisputably are second- and third-choice votes. For many voters, they may have been an afterthought.
If these second and third votes are considered to count as much as first-choice votes, then they should have been counted for all 13 candidates. If they were, it's possible that a candidate other than McCullough or Thigpen could have received the most votes. We don't know because the State Board of Elections has not released how many 1, 2 and 3 votes each of the 13 candidates polled.
Finally, Munger employs some "balloon juice" of his own: "The editor seems to prefer plurality rule, but sophomore political science students learn that plurality rule rarely chooses the best candidate in simulations."
Simulations? Why don't we look at real elections? In Guilford County's Senate District 28 race, Gladys Robinson won a plurality election over Trudy Wade and Bruce Davis. There was no "instant runoff." On what basis would Munger contend that Robinson, a community leader for many years and a member of the UNC Board of Governors, was not "the best candidate"?
For that matter, how does Munger conclude that Thigpen was not the best candidate in the appeals court race and McCullough was? Thigpen led McCullough by 100,000 first-choice votes — meaning a lot more voters thought Thigpen was the best candidate. After adding 2 and 3 votes, McCullough led Thigpen by 7,000 votes (pending a recount). On that basis Munger says McCullough was the best candidate? That strikes me as more than a little absurd.
On top of its other faults, IRV is arbitrary. Why not just count first- and second-choice votes? If that had been done, Thigpen's initial lead over McCullough probably would have held up. Why add third-choice votes? That's sliding further from the voters' real preference. Or if the idea is just to count more votes, how about giving voters fourth and fifth choices? Or let them rank all candidates in order of preference? That would be worse, in my view, but it's no less logical than the IRV format applied in this race.
My question to IRV supporters is this:
Compared to a plurality election, does IRV produce an outcome that is faster, less expensive or more credible?
The first two answers are no, plain and simple. Those are facts. The third question requires an opinion. My opinion is no — and professor Munger sure didn't change it.

To receive updates by email visit this link