Voters can handle Instant Runoff Voting by Caleb Kleppner

Testimony to Pierce County Charter Review Commission
Presented April 13, 2006

1. Introduction

Good evening. The title of my remarks tonight is, “Voters can handle instant runoff voting.”

Voters can handle instant runoff voting, and I’m pleased to hear from the Auditor that if the voters pass instant runoff voting, she can and will implement it better than anywhere else in the country. Running an election is a difficult job, and it’s certainly true that instant runoff voting adds some burden to the job, but as San Francisco and Burlington have demonstrated, it’s entirely manageable.

If there’s one thing I want you to remember about my presentation, it’s that voters can handle instant runoff voting.

That doesn’t mean that IRV is right for Pierce County, and I’m not going to try to convince you it is. Your job as commissioners is to figure out that out, by thinking about what criteria to use to evaluate an election system and then weighting the advantages and disadvantages of the various proposals according to those criteria. I’m confident that that at the end of the charter review process, you’ll put some good proposals before the voters.

So instead of trying to convince you that IRV is the answer, I want to simply demonstrate that when used in actual public elections, IRV doesn’t pose any problems for voters. I’ll try to do this by discussing how it has been implemented, what issues were raised, how those issues were addressed, and what the election results and exit polls show.

No voting method is perfect, which Kenneth Arrow received a Nobel Prize for proving, which means that all voting methods have disadvantages. But in the case of IRV, the idea that IRV is too difficult for voters is not one of those disadvantages. So if you want to oppose IRV, you’ll have to find another argument. I think I’ve heard all the arguments against IRV, and I’d be glad to enumerate them if you’d like, but I’ll feel like I accomplished my job if I can get you to use a valid argument against it rather than an spurious one, such as it’s too difficult for voters.

2. Some quick background on my experience with IRV

I worked for FairVote, a non-profit organization that promotes fair elections, from 1999 to 2004. I helped IRV legislation for San Francisco, Burlington, Vermont, Vancouver, Washington, and many state legislatures. I have run ranked ballot elections for a major multinational professional services firm as well as a listener-sponsored radio network with 100,000 members. I designed the federal and state testing protocol for the voting equipment and software used in San Francisco’s IRV elections, and I designed the voter education program and pollworker training for Burlington’s first IRV election. I also tested the IRV tallying software and trained city staff on its use in Burlington. I am currently interviewing to be the executive director of the Cambridge Election Commission, which uses ranked choice voting for its municipal elections.

3a. IRV in practice: San Francisco 2004

San Francisco voters passed the IRV charter amendment in March 2002 and used it for the first time for 7 of 11 county supervisor races in November 2004 using the same voting equipment that the city had been using since 2000.

What happened on Election Day? It was a hotly contested presidential election, and voters formed long lines before polls opened at 7am. But voters walked into polling places, filled out their ballots, put them in voting machines, walked out and then overwhelmingly told exit pollers, “I understood IRV, IRV was easy, and I prefer it to the old system.” For example, 87% of voters said they understood IRV well or fairly well, and voters preferred IRV to the old system by a 4:1 margin.

What do the results show? Did voters decide not to vote because of IRV? Did they skip the supervisor race? Did they cast an invalid ballot? Did they decide to rank only 1 candidate?

Voter turnout was higher in this election than in many years, mainly due to the presidential election. Drop-off, which is the percentage of people who show up at the polls but don’t vote in a particular race, ranged between 5 and 12% for supervisor, which may seem high until you learn that drop off for US senator was 7% and for congress and state legislature ranged from 8% to 16%. And the drop off for supervisor was no higher than in previous years. 99.3% of voters cast a valid vote, which is pretty good considering that one of those races had 22 candidates. Voters ranked on average 2.5 candidates out of a maximum of 3 possible.

How did the media react? Many members of the local media truly didn’t believe that voters would be able to rank candidates and that the Department of Elections would be able to run an instant runoff election, especially since San Francisco has had a distinguished history of election foul ups. You might have heard about ballots being dried in a microwave during one election or ballot box tops found floating in the bay after another. One hit piece during the IRV campaign, said, “The Department of Elections would screw up a two-car funeral.” The day after the election, the San Francisco Chronicle headlined its story: “New vote -- it's a go: Premiere of ranked-choice voting method mostly gets a thumbs-up -- few glitches,” which for a San Francisco election is a rave review.

And pretty much everyone had to admit that voters handled the new system and it didn’t pose problems for pollworkers, whose jobs really didn’t change much, except they had to post an IRV poster in the polling site and answer a few questions about IRV.

Did it go perfectly? No, elections almost never go perfectly. Some voters were confused about whether or not they had to rank 3 candidates, and some pollworkers falsely told voters they had to rank 3 candidates. But these situations were quite minor for the first use of a new system, and by monitoring polling sites, the Department of Elections was able to correct them.

3b. IRV in practice: San Francisco 2005

In 2005, San Francisco had 3 citywide races on the ballot: city attorney, treasurer and assessor.

This was the first citywide use of the system, and the first use for about 40% of the city. Again, voters walked in polling places, filled out their ballots, and reported they liked the system.

This time, 99.6% of voters cast a valid ballot. Dr. Chris Jerdonek estimated that IRV increased the turnout by 168% of the number who would have participated in a December runoff if the city had not adopted IRV, with especially large increases in traditionally low turnout areas.

3c. IRV in practice: Burlington 2006

Last month, Burlington used IRV for the first time to elect its mayor. There were 5 candidates on the ballot plus a write-in slot. The city hired me and a partner to design the voter education program and additional pollworker training. To make sure the election went well, a trained city employee staffed a special IRV help desk in every polling place.

What happened? Voters walked in, got their ballots, ranked between one and five candidates, put their ballots in the voting equipment, and walked out. 90% of voters said they knew they’d be asked to rank candidates for mayor, voters preferred IRV to the old “vote for one” method by more than 3 to 1, and 91% disagreed with the statement, “The ballot was confusing.”

So few people used the IRV help desk that the staffers were bored, and one precinct sent their help desk person home for the afternoon lull. The polls closed at 7pm, the city uploaded the data from all 7 polling sites, and announced the final, official tally at 9:05 pm, which included absentee ballots, since they get delivered, verified and fed into the voting machines in the polling places.

What did the media say? The Burlington Free Press headlined their article, “Voters ace instant runoff,” and a noted Vermont political commentator said the system went “smooth as silk.”

How did voters do? The valid ballot rate was 99.9%, voters ranked on average 2.9 candidates, and 90% of voters ranked one of the top two candidates. In Burlington, the Republican Party is the third party, and of the people who gave their 1st choice to the Republican candidate, about 1/3 ranked the Progressive next, 1/3 ranked the Democrat next, and about 1/3 said a pox on both of their houses. That looks to me like rational voter behavior. The drop-off for mayor was about 1%, which was lower than the 2% in 2001 and the 24% in 2003, and was much lower than contested city council races, which hadn’t been the case in previous elections.

These 3 elections are pretty strong evidence that voters are in fact capable of ranking candidates, and if you have a decently designed ballot, do basic voter education, and give pollworkers a little extra training, voters not only can handle it, but they like it.

5. Effect of IRV on low-income voters, racial and ethnic minorities, seniors and youths

When people claim that IRV is too hard for voters, I often ask them if they can rank candidates in order of choice. Most say they can, but they often say they are worried about other people. Who those other people are depend on where you are, but I’ve heard them all: less educated, poor, Asian Americans, Latinos, African Americans, old people, young people, you name it. And I’ve often heard these arguments made by members of these very groups. Strange that I, a well educated middle class white male, have more faith in the ability of these groups than some leaders in those groups, but I’ve seen voters of all stripes use IRV successfully, and the data is clear.

Rather than showing you data about everyone of these groups, which I could do if you had the time and interest, let me talk about two groups who were most prominently represented to be disadvantaged by IRV: Asian American voters in San Francisco, and low-income voters in Burlington.

Professor Rich DeLeon, of San Francisco State University, used the individual rankings combined with census data to test 9 hypotheses about whether IRV disadvantaged Asian American voters. Those hypotheses were:

Supporters of Asian American candidates would have, compared to voters in general,

a higher proportion of voters who didn’t rank one of the two finalists;
a lower average number of rankings;
a lower proportion of second & third choices;
a lower proportion of "effective ballots,"

In precincts, the higher the proportion of Asian Americans,

the higher the proportion of drop-off supervisor races;
the higher the rate of over-voting (ranking more than one candidate as first choice);
the higher the proportion of "exhausted" ballots;
the lower the mean number of choices (ranks) used in voting;
the lower the proportion of second or third choices made in voting;

He tested the 9 hypotheses in two supervisor districts that had high percentages of Asian American voters and several Asian American candidates. What did he find?

“Nine hypotheses with clear predictions were tested in each district, adding up to 18 opportunities for the available empirical evidence to reveal patterns of data at least consistent with, if not proof of, the arguments advanced by some critics that SF's new RCV system systematically disadvantages the city's API voters vis-à-vis voters in other racial/ethnic groups. Based on the evidence presented here, the score is zero for 18.”

So although an Asian American group campaigned against IRV, calling it the latest “Chinese Exclusion Act,” and then declined our offer to work together on voter education aimed at Asian American voters, it turned out that Asian American voters didn’t have trouble with it.

Burlington doesn’t have a lot of Asian American voters, but there is a sizable population of low income less educated people who are most concentrated in Ward 3. How did voters in Ward 3 do compared to the rest of the city? Did their turnout go down? Did they skip the mayor’s race? Cast invalid votes?

Ward 3 turnout was 25% higher than the 2003 mayoral election and 72% higher than the 2001 election. 1% of Ward 3 voters skipped the mayor’s race, the citywide average. Ward 3 voters were five times more likely to skip the contested city council race, which did not use IRV, than the mayor’s race. So IRV didn’t cause low-income voters to skip the mayor’s race; many more skipped the city council race.

In Ward 3, there were exactly two invalid ballots out of nearly 1,200 voters. It’s hard to have a lower rate than that. Again, no evidence that IRV led to increased invalid ballots, nor that voters in Ward 3 were more likely to cast invalid ballots.

Did they rank fewer candidates than voters in other wards? No, in fact Ward 3 had the highest average number of rankings of all 7 wards. They also had the highest percent (93%) of voters who ranked one of the top two candidates, and the lowest percent who ranked only 1 candidate.

I could go on about African American voters, Latino voters, Australian aboriginal voters in the outback, voters in Papau New Guinea in the 1960s or ethnic minorities in London, but I think you get the point: voters of all stripes can handle instant runoff voting.

There just isn’t any evidence that IRV compared to plurality elections is too difficult or confusing.

On the other hand, there is strong evidence of differential participation in primary and runoff elections. For example, in San Francisco, which used to hold a December runoff election after November local elections, voter turnout generally went down all across the city. But voter turnout was lower in low income and minority neighborhoods, and it decreased even more in those neighborhoods. So you might conclude that a December runoff is “too confusing” for those voters, but a better conclusion to reach is probably that December is a terrible time to hold an election, and it’s a lot easier to get people to vote in November. The same dynamic may well occur in Pierce County in an August primary election compared to a November general election

6. Poll workers and election officials

Running elections is tough work. I have a great deal of sympathy for the many hardworking auditors, city clerks and other election officials. They have a lot of work to do, and with new Help American Vote Act requirements, their jobs have gotten even more difficult. But the implementation of IRV is manageable, as San Francisco and Burlington have demonstrated with different vendors and different software, and by eliminating portions of an August election, you make that election slightly easier and cheaper.

Polling place procedures change very little, but pollworkers do need additional training to respond to questions about IRV.

The voting equipment, instead of storing vote totals, has to store individual rankings, which are then aggregated after the close of polls. The IRV tally is then applied to the aggregate set of rankings. To see what that looks like, check out San Francisco and Burlington’s websites, where all the data is posted.

It’s not a trivial change, but it’s not a difficult one. In a state like Washington, where people are increasingly turning to vote by mail, you might consider going to an all mail election, and eliminating polling places. That makes election administration easier, and since the central scanners used in absentee voting already capture ballot images, it makes the conversion to IRV even easier.

7. Political implications: who wins, impact on two-party system

The candidate with the most 1st choices win 95% of IRV elections in Australia. There have been 11 recent IRV races in the United States, and the initial frontrunner has won them all.
IRV lessens, but does not eliminate, negative campaigning, especially between candidates who might be logical allies, such as Libertarians and Republicans, Democrats and Greens, and independent candidates and whichever other candidates they form alliances with.
It broadens the range of candidates,
It broadens the range of issues discussed and covered by media

What IRV doesn’t do:

End the two-party system or elect lots of extremist or third party candidates. In Australia, for example, which has used IRV to elect its House of Representatives for 80 years, I believe that no third party candidate has won a general IRV election. I believe one third party candidate won a special election to fill a vacancy, but that’s a pretty small number, especially considering that there are several third parties that win election to the Federal Senate using a different system. IRV elects the candidate preferred by a majority, and unless you’ve got a pretty strange majority somewhere, that winning candidate isn’t going to be some kind of fringe candidate.

Again, whether or not these things are good is a value call for the people of Pierce County, but this is what happens when IRV has actually been used in American public elections.

8. Conclusion

Now for the short part: What did I want you to remember? Right, voters can handle instant runoff voting. Thank you very much for the opportunity to speak, and I’d be pleased to respond to any questions or comments you have.
Burlington Free Press