Friday, March 31, 2006

Burlington Experience with IRV by Caleb Kleppner

Caleb Kleppner was hired by the city of Burlington, VT to design the instant runoff voting or IRV voter education program, to train pollworkers about IRV, and to train city staff on the software used to tally the IRV election results. He also assisted Director of Elections Jo LaMarche with the design and evaluation of the pre-election logic and accuracy testing of Burlington’s voting equipment.

By way of background, he worked for FairVote – the Center for Voting and Democracy for five years when he was living in San Francisco. In that capacity, he drafted the IRV legislation adopted by the voters in March 2002 and worked with city and state officials in the development, testing and deployment of the optical scan voting equipment used to conduct IRV elections in 2004 and 2005 in San Francisco.

How did Burlington’s first IRV election go?

By all accounts -- my personal observations of the Burlington wards and from the media coverage -- the election went extremely smoothly. Voters didn’t have trouble filling out their ballots, and pollworkers didn’t have problems with the IRV portion of the process. I think you’ll hear testimony from some pollworkers, but the comment that stuck with me was that the people staffing the IRV help desks were bored. The Burlington Free Press headlined their story about the voting, “Burlington Voters Ace Instant Runoff.”

In terms of logistics, polls closed at 7 pm, and we ran the IRV tally and announced the results by 9:05 pm.

Some have speculated that IRV might be too difficult for voters, in particular, too difficult for low income voters. Was it?

Let’s imagine what we would expect to see if the system was confusing for voters and especially low income voters. Then we can see if that happened.

First, we would expect lower voter turnout. In fact, citywide voter turnout was more than 25% higher than any mayoral election since 1999, which is as far back as I was able to go. In the lowest income ward, Ward 3 turnout was 25% higher than the 2003 mayoral election and 72% higher than the 2001 election. Thus, we have no evidence that IRV depressed turnout citywide or in Burlington, nor in low-income neighborhoods.

Second, we would expect more voters to skip over the mayor’s race compared to other races. This year, about 1% of voters skipped over the mayor’s race. In contrast, 2% skipped the mayor’s race in 2001 and 24% skipped it in 2003. In Ward 3, 1% skipped the mayor’s race, the citywide average. In every city council ward, more voters voted in the mayor’s race than in every other race – something that was not true in 2003. In 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. Thus there is no evidence that IRV led to more voters skipping the mayor’s race citywide or in low-income areas.

Third, we would expect the rate of invalid ballots to rise. This year, the invalid ballot rate for mayor was one tenth of one per cent, meaning that 99.9% of voters cast a valid vote. I do not have data for past mayoral elections, but I know from national experience that this is an extremely low invalid ballot rate. For example, in the 2000 presidential election in Florida, the overvote rate was around twenty times higher.

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.

Fourth, we would expect voters to forego the opportunity to rank candidates. There were 3 major candidates in this race. Voters ranked on average 2.9 candidates, and in Ward 3 voters ranked 3.1 candidates (highest of all 7 wards). 82% of Burlington voters ranked more than one candidate, and 83% of Ward 3 voters ranked multiple candidates. Thus, no evidence that IRV discouraged voters from ranking multiple candidates, and in fact, Ward 3 voters ranked more candidates than voters in other wards.

Fifth, we would expect large numbers of exhausted ballots, meaning ballots that did not rank one of the top two candidates. Over 89% of all voters and 93% of Ward 3 voters ranked one of the top two candidates. Thus, the vast majority of voters cast a vote that counted in the decisive round of the instant runoff – and those that didn’t were largely Republican voters who quite rationally may have decided to refrain from expressing a preference between the Democrat and the Progressive.

Even though this was the first time Burlington voters voted in an IRV elections, there is no evidence that IRV discouraged voters from participating and there’s no evidence that the system posed any burden for low income voters.

Note that these conclusions mirror what happened in San Francisco, where some people speculated that IRV might be disadvantageous to Asian Americans. Professor Rich DeLeon of San Francisco State University analyzed 18 hypotheses and concluded: “Based on the evidence …, the score is zero for 18” that IRV disadvantages Asian American voters in San Francisco.


The administration of this election was extremely smooth, and voters demonstrated they are perfectly capable of participating effectively in an instant runoff election.


At 1:40 AM, Anonymous Anonymous said...

The voters were not confused by IRV. The voters of Pierce County do not like the Montana primary. The voters of Pierce County will like IRV better than the Montana primary.


Post a Comment

<< Home