Friday, March 26, 2010

Response to Gierzynski "Assessment" of IRV

Table of Contents:

1. Introduction
2. response from Prof. Richard DeLeon
3. Response from Terry Bouricius
4. Response from Prof. Jack Nagel
5. Appendix A: further comments by Prof. Richard DeLeon
6. Appendix B: Refutation of Gierzynski's paradox arguments by Terry Bouricius


1. Introduction:

Anthony Gierzynski's "assessment" of IRV is not an impartial academic analysis, and Mr. Gierzynski is not an impartial academic. As Professor Richard DeLeon writes, “Anthony Gierzynski’s assessment of IRV is shoddy, biased, and borders on ridiculous.” His report is misleading and seems motivated by his clearly stated bias against a voting method that accommodates independent and third party candidates. This bias against third parties is not surprising considering that Mr. Gierzynski in 2002 was a Democratic candidate for the Vermont legislature who lost to Progressive candidate Bob Kiss. His analysis of the subsequent IRV election of his former opponent, Bob Kiss, incorporates this bias. His stated solution to the "spoiler" problem, which can punish voters and result in the election of the least preferred candidate, is for third parties to merge into major parties - - a solution that is irrelevant to non-partisan municipal elections, impractical for Vermont and unable to address the potential of independent candidacies that also have been an ongoing feature of Vermont elections.

If one wants a genuinely impartial analysis of IRV, one shouldn't choose a partisan political opponent of the two-time IRV winner in the Burlington mayoral race, merely because he also has an academic credential. There are many genuine experts without axes to grind within the field of political science. Indeed, it should be noted that the primary organization of political science professors, the American Political Science Association (APSA), incorporates IRV into its own constitution for electing its national president.

In addition to this rebuttal of Gierzynski's essay, below you will find comments that have been offered by other political scientists with actual expertise in the area of election methods. These include Prof. Richard DeLeon, Prof. Jack Nagel

=====================

2. Response from Prof. Richard DeLeon:

San Francisco State University professor emeritus Richard DeLeon [ http://bss.sfsu.edu/deleon/ ] was founder of the Public Research Institute at SFSU and its long-time director. He was Chair of SFSU's Department of Political Science from 1994-2000, chair of California State University Social Science Research and Instructional Center and author of numerous award-winning articles and books. He was a consultant to SFSU's surveys on ranked choice voting in San Francisco 2005 and 2006. (rdeleon18@comcast.net)

“Anthony Gierzynski’s assessment of IRV is shoddy, biased, and borders on ridiculous. It uses tortured logic, cherry-picked “evidence,” and dubious arguments to slam IRV and lamely defend two-party rule and the status quo ante. Worst, it distorts and misrepresents San Francisco’s experience using IRV, which has been used in six elections now with very good results. As documented in two university exit polls, San Francisco voters of all races and classes (a) understand how IRV works, (b) like that it allows them to choose their preferred candidates and not waste their votes, and (c) overwhelmingly prefer IRV to the traditional run-off system.”
[Additional comments from Prof. DeLeon are contained in Appendix A]

=====================

3. Response from Terry Bouricius

FairVote policy analyst and former Progressive Party Burlington City Councilor and State Representative

Below is a rebuttal of each of the sections of Gierzynski's "assessment" of IRV.

Discrimination by Complexity:
Anthony Gierzynski argues that by having a more complex ballot, IRV discriminates against less educated voters. At some point, added complexity certainly could introduce significant discrimination into the electoral process. However, Mr. Gierzynski does not provide any evidence that IRV approaches that threshold. In fact, most evidence indicates that education levels have no actual impact on effective use of ranked ballots. Prior to the first use of IRV in Burlington elections, to provide one anecdotal example, third grade students at Burlington's Edmunds Elementary School were asked to use IRV to elect a student council representative. These third-graders easily used the ranked choice system, and there were no spoiled ballots. Gierzynski simply dismisses the fact that the Burlington voters in Wards with the lowest income and education levels had no increase in invalid ballots and were just as likely to mark alternate choices as voters in Wards with higher incomes and education. Instead he relies on an exit poll he conducted in 2006. Gierzynski's 2006 report on that exit poll states:

"In general most voters thought that IRV was a better way to express voting preferences—both theirs and voters’ in general—than the usual system (see Figure 3). Seventy-one percent of respondents agreed that IRV was “a better way to express my voting preferences than the usual system.” Sixty-nine percent agreed that with IRV “the election results will better reflect voter preferences than the usual system.” Regarding any difficulty voters might have had with IRV, only 8.6% of voters said they found the ballot confusing."

However, Gierzynski notes, among the least educated 14.5% agreed the ballot was confusing. Though not insignificant, it still reflects a small minority of such voters, and does not mean there was any corresponding failure to effectively use the ranked ballot. We do not know how many of these voters who found the ballot somewhat confusing, none the less appreciated the additional ranking opportunity, and how many disliked it. For example, a person might say that their cable TV selections are more confusing than having only over-the-air programming, but still want to keep and be able to use cable television. Since virtually every one of these less educated Burlington voters cast valid ballots (99.9% of all IRV ballots were valid), it is unreasonable to suggest, as Gierzynski does, that any class of voters were disenfranchised or harmed.

Much of the rest of Gierzynski's analysis about ballot complexity relates to a hypothetical situation in which IRV might be used for all, or most general election contests, rather than a short municipal ballot. For municipal elections the current "complex" ballot typical in the U.S. does not apply. Typically, towns are only electing two or three offices, and voter turn-out is far lower than in state elections with "complex ballots" with many races. Contrary to Gierzynski's implication, the relative "complexity" of ballots clearly is far less significant than how exciting the races are in terms of affecting turnout. Races with only one or two uncompetitive candidates do not drive high voter turnout. Because IRV accommodates, and as Gierzynski points out, may even encourage more than two candidates while reducing the threat of a "spoiler" scenario, it does appear to somewhat enhance voter turnout. Burlington's two recent IRV mayoral elections had higher turnout than the two preceding mayoral elections (though not higher than some historic races). Aspen (CO) conducted its first ever municipal election using IRV recently and saw the highest turnout for a municipal election ever. However, IRV certainly does not guarantee a high turnout. For example, a recent San Francisco election in which the incumbent mayor won a landslide victory over only token opposition, did not spur high turnout despite the use of IRV.

Paradoxes:

This section of Mr. Gierzynski's assessment is the most disturbing. It reveals either a lack of basic understanding of voting methods, or an intentional effort to mislead the reader. Gierzynski discusses four voting paradoxes that can occur with IRV, but amazingly fails to note the key fact that these identical paradoxes can also occur under normal plurality elections or two-round runoff elections, and are not at all unique to IRV. He also neglects to mention the dramatic paradox of his preferred system of plurality voting that would have elected the candidate least preferred among the top three candidates in Burlington.

Nothing in this section of his assessment brings any useful information to light, as these paradoxes have nothing to do with any difference between IRV and the voting methods already commonly used throughout the U.S. The only possible purpose for presenting these paradoxes (and not mentioning that they also apply to plurality or runoff elections) is to mislead the reader into assuming they are unique "defects" of IRV.

An exhaustive presentation of paradoxes and election method criteria see

http://fc.antioch.edu/~james_green-armytage/vm/survey.htm

[Appendix B contains a more thorough explanation of the four paradoxes Gierzynski mentions and how all of them also apply to plurality or traditional runoff election systems.]

Addressing the "Real" Problem:

Mr. Gierzynsnki virtually defines "the real problem" as voters getting more than just two choices on the ballot – it is “technical fix” to solve a “political problem” that is the very fact of more than two candidates running. He writes “single seat contests (such as mayor, or US Senator, or governor, or president) provide an incentive for those of similar political mind (that is ideology) to coalesce behind a single candidate in order to win a majority of votes and capture the seat – those that work together to build a majority before elections win, those that don’t, lose. This structural incentives is the main reason the US has a two party system." Mr. Gierzynski goes on to make it clear that he is a strong advocate of the two-party system with only two choices on the ballot.
This view is debatable for state and federal elections and has even less influence in a municipal setting with non-partisan elections. Under a plurality voting system that suffers a serious threat of "spoiler" scenarios, worthy potential candidates often refrain from running so as to avoid causing a split in the majority. Some people may agree with Mr. Gierzynski that democracy is enhanced by such limitations on voter choice, and encouraging behind the scenes negotiations and arrangements prior to the election to avoid having more than just two choices on the ballot. However, I believe that most people believe greater voter choice would be a good thing, so long as the spoiler problem could be mitigated with either a traditional runoff or IRV. Certainly it diminishes some of the worst abuses of attack politics and the impact off campaign spending on those attack ads.

Cost:

As to the cost of transitioning to IRV, Gierzynski again miss-represents reality. In jurisdictions that replace two-round voting systems (either primary and general, or first round and runoff) with IRV, cost savings are the norm (see the response of the Wake County election administrators about the savings from using IRV here http://www.fairvote.org/?page=2543). However, there can be a modest net increase in costs if the jurisdiction is replacing single-round plurality elections. The added cost depends largely on the voting equipment in place, and whether it can easily be adapted to ranked ballots. Some jurisdictions may decide a more democratic election process warrants the additional cost and others may not.

Conclusion:

Here Mr. Gierzynski has turned reality on its head with his analogy to climate change scientists. Virtually every election method expert knows that the paradoxes Gierzynski discusses are not unique to IRV, but apply to plurality or traditional runoff elections as well. If anything, in his analogy Gierzynski represents the climate change denier, being outside the norm among electoral method experts (I again refer you to the fact that the APSA itself incorporates IRV into its own constitution). Many leading comparative political scientists support instant runoff voting over plurality voting. Some of them can be found on this list: http://www.fairvote.org/irv/endorsers.htm

=====================

4. Response from Prof. Jack Nagel

Fullbright Scholar Jack Nagel is graduate chairperson of the Department of Political Science at Wharton, University of Pennsylvania. His area of research includes elections and voting theory. He is Co-Editor of the Journal of Policy Analysis and Management, on the Editorial Board of Electoral Studies, and on the Editorial Board of Political Science.
(nageljh@sas.upenn.edu)

Prof. Nagel made this comment after reading Professor Gierzynski’s report:

“IRV always prevents the victory of a candidate that the majority of voters like least. A runoff plan that elects any candidate with a plurality provides no such guarantee. As for repeal advocates’ desire for a two-party system, the two-party system has already broken down in Burlington, so voters will be better served by an election method like IRV that is well adapted to multi-party races. ”



=====================

5. Appendix A

Further comments from Prof. DeLeon:

1. Here's a link to my original report on my re-analysis of the PRI exit poll data for the Nov 2004 election in San Francisco:

http://www.sfusualsuspects.com/resources/docs/Rich%20DeLeon%27s%20Working%20Paper%20on%20IRV%20in%20SF%20--%20Sept%2011%202005.pdf

Most relevant regarding alleged negative effects of ballot complexity etc. are Tables 1 (understanding IRV) & 2 (information burdens of IRV) on p 4, the bottom line critical results in Table 6 (overall preference for IRV vs old system), and Table 7A on p 8, with the interpretive note: "Notice that voters who gathered more information were much more inclined to favor RCV (71%) than were those who gathered less (52%) – a result some readers may find surprising, especially if they view information-gathering as a burden or cost of voting rather than as a benefit."

2. Here's a relevant excerpt from my written testimony to Minnesota State Senate :

QUESTION 1: Did San Francisco voters understand ranked-choice voting?

ANSWER: Yes. More than half the sample voters (52.8%) said they understood ranked-choice voting (RCV) “perfectly well,” and another 35.3% said they understood it “fairly well.” About 12% said they had at least some trouble understanding RCV, including only about 2% who said they did not understand it at all. See Table 1, attached.

QUESTION 2: Did RCV require voters to gather more information about candidates compared to past elections?

ANSWER: Yes, on net, with about 31% of sample voters saying they gathered more information than in the past and only about 7% saying less. The vast majority of voters, however, reported no difference. Whether the need to gather more information is a good or bad thing, however, is an open question. There is evidence, for example, that voters who gathered more information were actually much more pro-IRV than were those who gathered less.
3. The piece I did for the Los Angeles people (see attached) might be of some use. This is based on PRI's exit poll of voters in the Nov 2005 election, with an exclusive focus on the responses of self-identified Latino and Latina voters. Esp. see Tables 3 and 4 and the interpretive notes, particularly this bit following Table 4: "More than four out of five Latino voters in all income classes reported that they understood RCV perfectly or fairly well.

4. In general, there's no evidence I've seen in the 2004 & 2005 PRI exit polls that IRV ballots are so complex or burdensome that they discriminate against lower SES classes of voters by confusing them & discouraging their turnout in elections, etc. For me, the bottom line is that SF voters, including lower SES voters, overwhelmingly prefer IRV to the old system, and for good & plausible reasons (fewer wasted votes, more able to vote for preferred candidates, etc.).

5. One last thought: A much more "complex" version of rank choice voting (in the form of its use in full-blown PR/STV systems) was used by voters in a number of U.S. cities between 1915 & 1960, including some big ones like Cincinnati & Cleveland in Ohio, and the voters had no problem with using it with good results -- and this long before computers & when the average schooling level was much lower. See Kathleen L. Barber's book, Proportional Representation and Election Reform in Ohio (Ohio State Univ Press, 1995). In her conclusions, she writes: "The conventional wisdom came to be that the PR/STV ballot was too complicated for people to understand, causing turnout to decline. In four of the five cities this turned out not to be true...." p 308. Re claim that ranked choice voting leads to greater conflict, she writes: "In none of the Ohio cities using PR elections was there a significant increase in conflict among council members. In fact, it appears that the electoral pressures of ranked voting led to greater consensus as campaign styles changed." p 308. Her overall conclusion is that: "In these five Ohio cities, PR/STV was demonstrated to be an electoral system technically capable of facilitating public decision making in complex communities as well as producing fair representation." p 309.

=====================

6. Appendix B: Refutation of Gierzynski's paradox arguments
by Terry Bouricius


All of the paradoxes Anthony Gierzynski describes apply to common plurality or two-round runoff election systems as well as IRV, and are thus irrelevant to deciding between these voting methods. He also neglects to discuss the far more serious paradoxes of plurality elections (such as the fact that the candidate who the large majority of voters agree is the single worst candidate can win under plurality rules).

Gierzynski describes a paradox known as the "Condorcet cycle" paradox (the rock/paper/scissors cycle of winners). This is simply a restatement of a well know fact among political scientists that with three or more choices, it is mathematically possible that no single candidate is favored by a majority of voters over all other candidates. Such a cycle could exist, even if unrecognized in any election using plurality rules. The declared "winner" of a plurality election with less than 50% could even be the single candidate that the majority of voters believe is the worst choice (which can't happen with IRV). Likewise if there is an actual tie between two candidates, there will be no "majority winner." These non-majority realities are not "caused" by IRV or any other voting system. They are simply mathematical truisms and have no relevance to choosing a voting method. No voting system can ultimately guarantee to find a majority winner since there may not be one to be found. However, the standard method to try and find a majority winner used in the U.S. is to reduce the field of candidates to two finalists and hope that in the runoff there isn't a tie. This is the same logic used by IRV.

Another paradox Gierzynski mentions deals with what is called the "Condorcet winner." In the Burlington example he uses, the candidate in third place (Andy Montroll) was the preferred compromise second choice of the supporters of both of the stronger candidates (Bob Kiss and Kurt Wright). If Montroll had somehow made it into a runoff against either Wright or Kiss he might have won. Using the logic of a traditional runoff, having less core support than Kiss or Wright, Montroll with 23% support did not make it into the final IRV runoff tally. However, it would be possible to write election rules that made a series of pair-wise comparisons of all five candidates in that race to see if one candidate would theoretically beat all comers. There are advocates of such voting rules within academia, though there is no track record for such a system for governmental elections. One reason such a system would have a hard time passing muster, is that it would be possible for the candidate who came in dead last in terms of "first choices" to be the Condorcet winner simply by being inoffensive. However, it seems bit disingenuous for Gierzynski to fault IRV for failing to elect this weak compromise candidate who also would have lost under plurality elections (finishing in third place) or a traditional runoff election (being eliminated before the final runoff.). IRV found the "majority winner" in the same sense that a traditional runoff election does…by narrowing the field to the two strongest finalists and seeing which is preferred by more voters. Just as in a traditional runoff election, some voters may sit out the final runoff, by staying home in the case of a separate runoff, or by declining to indicate which of the two finalists they prefer over the other, in the case of IRV. This right to abstain should be respected, but should not prevent the conclusion of the election process with the declaration of a winner.

He mentions the "non-monotonic" paradox, in which a candidate who receives more votes from voters who otherwise supported an opponent, thereby goes from being a winner to a loser. This ironic possibility exists in any runoff election or any election based on two rounds of voting in which some candidates are eliminated in the first round. It is not actually the additional votes going to this candidate that cause the switch to a defeat, but rather the change in relative positions of the other candidates that results from the swapped votes, such that this otherwise winner ends up facing a stronger opponent in the final runoff. For example, in any runoff election (whether instant or two round) there is a possibility that a voter might help her favorite candidate by instead of voting for that favorite, by voting for a weak opponent in the first round, so that her favorite candidate will have a better chance of winning in the second round. Of course this strategy can easily back-fire if too many voters attempt it and is counter-intuitive in IRV elections. Since under IRV voters cannot make this strategic switch in the first round only, it is a far less likely scenario than under two-round runoff elections, where voters can switch to their true favorite in the second round. Thus this particular paradox is significantly more likely to be a factor in a traditional runoff election than in an IRV election. A court affidavit by Prof. Jack Nagel in a Minnesota court case involving this issue is available here: http://www.fairvote.org/docs/MinneapolisAffidavitofJackNagel11-3final.pdf

Note that the Minnesota court and on appeal, the Minnesota Supreme Court, concurred and upheld IRV.

A fourth paradox he mentions is the "no-show" paradox, which again is common to both IRV and all runoff elections throughout the U.S. It is possible in a traditional runoff election, that by staying home during the first round of voting, voters may keep their favorite candidate from advancing to the final round. But in this final match-up, if their favorite were a candidate, he or she would lose to a candidate the voter thinks is the worst candidate. However, by sitting out the first round of voting, an acceptable compromise candidate might make it into the final match-up against the hated candidate, and this compromise candidate has broader appeal and can win.