The case for instant runoff voting -- two more examples
One test of a good voting system is whether it effectively prevents a candidate to enter the race as a "spoiler." There is probably a technical term for this, but there ought to be a stability of results. If A would beat B in a two-candidate race, the addition of C to the list of candidates shouldn't result in a victory for B. By extension, adding a candidate to an n-candidate race shouldn't hand the election to someone who would have lost the n-candidate race.
Because many jurisdictions don't have any sort of runoff at all, and only a handful use instant runoff voting, we often see elections where the winner is someone who might not have won with fewer candidates in the race. Sometimes the winner is someone who might have lost head-to-head with several of the other candidates, but wins the multi-candidate race because the other candidates split a common core constituency.
Here are two more recent examples.
Yesterday there was a special primary election in Ohio's 2nd congressional district, a seat previously held by Rob Portman, who is now U. S. Trade Representative in President Bush's cabinet.
Here's the final result (PDF) in the Republican primary:
JEAN SCHMIDT 14232 31.35%
BOB MCEWEN 11565 25.48%
TOM BRINKMAN, JR. 9211 20.29%
PAT DEWINE 5455 12.02%
ERIC MINAMYER 2111 4.65%
PETER A. FOSSETT 1026 2.26%
TOM BEMMES 695 1.53%
JEFF MORGAN 400 0.88%
DAVID R. SMITH 374 0.82%
STEVE AUSTIN 221 0.49%
DOUGLAS E. MINK 100 0.22%
Ohio has no primary runoff, so Schmidt wins the primary despite the 69% of the vote against her.
Had the 4th through 11th place candidates not been in the race, the distribution of their votes to the top three could have put any of the top three in first place. Even an Oklahoma-style primary runoff wouldn't fix the problem, as the top three were close enough that we can't know which of them would have finished 1-2 if the other eight candidates had not been in the race. Instant runoff voting would have produced a winner with the support of a majority of the voters. The beauty of instant runoff voting is that you eliminate the spoiler effect of an additional candidate.
We had another example in last week's Republican primary for New Jersey governor.
Doug Forrester 108,090 35.94%
Bret Schundler 93,926 31.23%
John Murphy 33,662 11.19%
Steven Lonegan 24,346 8.10%
Robert Schroeder 16,691 5.55%
Paul DiGaetano 16,551 5.50%
Todd Caliguire 7,472 2.48%
It's been argued that Steven Lonegan acted as a spoiler for Bret Schundler, peeling off enough conservative support to keep him from once again winning the GOP nomination. The counter-argument is that if the other minor candidates had not been in the race, most of their votes would have gone to Forrester. A runoff is the only way to know for sure.
A simple two-candidate runoff probably would have been sufficient to provide a clear outcome, but theoretically Murphy could have finished second in a three-way race -- the bottom four candidates had enough combined votes that if they had been out of the race and all their votes had gone to Murphy, Murphy would have finished second.
Subscribe to this blog's feed:
And if all the votes had been combined, blah blah blah.
And if a frog had wings, he wouldn't whomp his butt when he jumped.
I'm not a fan of having election after election. Primaries are not head-to-head elections; they never were. They are a winnowing-down process for the big dance. If votes get split in the primary, that's the breaks. A well-organized campaigner who seeks to unite the vote instead of splitting it would overcome these hurdles.
Even in a close race between two candidates on the ballot, you may still have a winner without 50 percent because of a few votes to a last-minute write-in candidate.
And another thing, the instant runoff voting sounds OK, but I think it's fraught with confusion and fraud. The election system has enough problems as it is. There's no sense in making a bad system worse by throwing a complicated new thing into it.
Well, W, Mike is absolutely right, and instant runoff is the solution. Majority-vote requirements are designed to keep fringe candidates with limited support from prevailing over candidates that are more generally accepted by the electorate.
Now, runoff (2d round voting) is the most common practice, but runoff typically results in about 30% voter rolloff (not voting) from the first to second round, and only about 30% of the time does the initial leader lose in the second round. When a leader has over 40% and leads by at least 5 points, they win 95% of the time. Brad Henry os one of that rare 5% that overcomes the 40+5 rule.
Instant runoff would still not have fixed the problem in Mike's second example. Under IR, voters cast an initial ballot and a second preference. Votes cast for candidates 3-through-X are typically redistributed on their second preference across the top two finishers. Now, pure preference voting, which reallocates votes up by eliminating the bottom finisher until a majority winner is revealed makes more sense and is more efficient, but also is more complex.
As for IR being "fraught with confusion and fraud," that is simply not true. Any electoral system can be fraught with fraud, but it is not a function of the voting rule, but the culture and legal environment under which voting takes place.
Any runoff is preferable to any one-shot election. And IR is prefereable to a second ballot because it mitigates against losing preferences. pure preference voting makes the most efficient translation of preferences into votes and is most likely to lead to the election of a Condorcet candidate (the candidate who can beat all other comers head to head), which is Mike's real concern.
Good post Mike.
For future reading, folks, check out Sam Merrill's MAKING KULTICANDIDATE ELECTIONS MORE DEMOCRATIC and Chuck BUllock and Loch Johnson's RUNOFF ELECTIONS IN THE UNITED STATES.