SF's use of IRV defended in BeyondChron article.

In a recent piece at BeyondChron.org, San Francisco resident and Instant Runoff Voting proponent Chris Jerdonek makes the case against repealing IRV. While he devotes most of his time to impaling IRV's opponents, there are a few substantive claims related to the actual merits of IRV. Before looking at them though, I want to preface this by stating that I believe Instant Runoff Voting is probably a little better than the delayed Top-Two Runoff that San Francisco used to have (although it is much worse than Score Voting and Approval Voting). My "angle" is merely objective factual reporting. Throughout the years, the vast majority of IRV proponents have made egregiously misleading and even false statements in support of the system.
[IRV is] recommended by Robert's Rules of Order
This claim is false.
[IRV] was used during the first-ever election of Chinese Americans as mayors of both Oakland and San Francisco.
This is technically true, but the implication seems to be that IRV deserves the credit for this increase in diversity. But San Francisco mayor Ed Lee came in first, and progressive icon John Avalos came in second, in first-place votes. Under the old system, they would have gone to a second round runoff. So is Jerdonek implying that Lee would have lost to Avalos with the traditional runoff?
Since the first use of RCV in 2004, minority representation on the San Francisco Board of Supervisors has doubled from four supervisors to eight (out of 11 seats, a whopping 73%); Asian American representation has quadrupled from one supervisor to four. San Francisco’s Board of Supervisors is the most diverse and representative of any major city in the entire nation.
This is a shining example of the post hoc ergo propter hoc fallacy. The implication is that IRV is responsible for this increase in diversity, when it could have been a mere coincidence. Consider that in 29 of the 31 races to which IRV has so far applied, the winner was the candidate who would have won with Plurality Voting without even holding a runoff. And of the two exceptions, one of them was the infamous 2010 District 10 supervisor race, in which Malia Cohen (who was third in first-place votes) won in the 20th round with 4321 votes from 18308 voters who expressed at least one ranking in that race.

In a recent debate, San Francisco supervisor Sean Elsbernd (who is fighting to repeal IRV) challenged IRV architect Steve Hill to name a concrete example in which IRV was responsible for the increased diversity. And even if Hill could produce one, what would that mean? It would mean that one of those winners, who came in first in first-place votes, would have lost in a head-to-head matchup with the person who came in second. In that case, could we really even say IRV had produced a better outcome?
In addition, adoption of RCV has led to supervisorial winners securing on average 29% more votes than supervisors had with the old December runoff system. 
There have also been important elections in which turnout went up in the runoff. In the 2003 mayoral runoff between Gonzalez and Newsom, there were 253,872 ballots cast (54.46% of registered voters), compared to 208,028 in the general election. Whereas in the recent highly contentious 2011 mayoral race, using IRV, there were 194,215 votes. A rough analysis by Warren Smith, the Princeton math Ph.D. who created ScoreVoting.net, shows that the change in turnout has been statistically insignificant.

Lastly, this is a pretty unimportant issue, since the number of voters required to reach statistical significance is vastly lower than any of these figures. If the IRV proponents have a valid complaint with turnout in runoff elections, it's that there's an alleged demographic skewing (older, whiter, richer voters turning out) that happens, due to e.g. students being out of town in December. Needless to say, this has nothing to do with runoff elections per se. It is merely about when the elections are held. Sean Elsbernd has discussed a compromise in which we go back to a September/November election/runoff scheme, which would hopefully address this skewing problem.
And San Franciscans have saved approximately $7 million by not holding unnecessary December runoff elections for the past eight years in which 40 races have been decided using RCV.
The cost issue is very complicated. IRV elections are more costly to conduct than a regular Plurality election (although cheaper than Plurality+Runoff), and San Francisco has spent hundreds of thousands of dollars on voter education — e.g. a billboard in my neighborhood that explained how to properly rank a ballot. I suspect this $7 million figure is a little high, but let's assume it's in the ballpark. SF's Annual Budget totaled $6.8 billion across all funds for the 2011-2012 year. So this is a savings of 0.103% of our yearly budget. Not that $7 million is insignificant, but just to put that into perspective.

The more important point is that this savings has nothing to do with the adoption of IRV per se. It comes from eliminating runoff elections, which you could do without adopting IRV at all. In that case, people might complain that you'd no longer get a "majority winner". But as I pointed out above, the result would have plausibly been the same in about 94% of elections. And IRV doesn't exactly guarantee a "majority winner". For instance, in the 2009 IRV mayoral race in Burlington, Vermont, the Progressive won, even though a majority of voters preferred the Democrat to the Progressive. In fact, IRV could have potentially done that even if the Democrat had more first-place rankings than the Progressive.
though ranking candidates is completely voluntary, and numerous studies have shown that voters often select a single candidate out of choice, not confusion
This is highly misleading. Yes, people may bullet vote "intentionally", but that can often happen because they don't understand how IRV works. IRV proponents often contend that a benefit of IRV is that it satisfies the Later-no-harm Criterion, meaning that a vote for your second choice cannot hurt your favorite (although it can hurt you). Given that, most people who bullet vote would be better off ranking additional candidates, in case their favorite is eliminated. But they may naively assume that they are helping their favorite candidate more by bullet voting. Indeed, I have found that even my fellow software engineers who vote don't understand how IRV works. Good luck to the non-techie folks.
the American Political Science Association uses it to elect its president (and that group knows a thing or two about elections).
It is extremely doubtful that a political science organization has expertise in voting methods, which have a lot more to do with game theory and economic principles (heavy math) than with the traditional "liberal art" of political science. I'm willing to bet that if I had a conversation with the APSA members who instituted IRV, they would not have the faintest idea what Bayesian Regret is, or what the Favorite Betrayal Criterion is.

Now here are some organizations with a focus on math, who use Approval Voting, the simplest form of Score Voting — both of which are vastly superior to IRV in basically every respect.
  • Mathematical Association of America (MAA), with about 32,000 members
  • American Mathematical Society (AMS), with about 30,000 members
  • Institute for Operations Research and Management Sciences (INFORMS), with about 12,000 members
  • American Statistical Association (ASA), with about 15,000 members
  • Society for Judgment and Decision Making
  • Social Choice and Welfare Society
  • International Joint Conference on Artificial Intelligence
  • Public Choice Society
  • European Association for Logic, Language and Information
  • Econometric Society
  • National Academy of Sciences (for the selection of members)
  • The Voting Power and Procedures branch of the London School of Economics selected Approval Voting from a pool of 18 potential voting systems.

There are several good arguments to be made in favor of IRV over the traditional Top-Two Runoff. Unfortunately most IRV proponents have avoided them in favor of specious and misleading (and sometimes even blatantly false) arguments.

No comments: