Contact:

Steve Hill: (415) 665-5044,  hill@steven-hill.com 

Rob Richie: (301) 270-4616, rr@fairvote.org  

Understanding the RCV election results in District 10

A FairVote Analysis

The Board of Supervisors race in District 10 was an unprecedented race in San Francisco’s seven-year history of using ranked choice voting (the first RCV elections took place in 2004). It featured 21 candidates, no incumbent and no obvious front runners.  That resulted in an election in which the winning candidate, Malia Cohen, barely edged out the competition in an exceptionally close race. How close was it?

• The top vote-getter in the first round barely topped 12%.

• In the first round, the leading four candidates (Sweet, Kelly, Cohen and Tran) were all within about one third of a percent (0.34%) of each other.  A fifth candidate (Moss) was within 1% of the first-round leader.

• Those five leaders remained the leaders and within 3.5% of each other (as a fraction of the first round continuing votes) for each of the following rounds until they were the last five continuing candidates.

San Francisco previously had an RCV contest in a supervisorial race that had 22 candidates (District 5 in 2004), but there were two clear front runners in that race that together had about 40% of the first round ballots and helped to order the field. That race was won decisively by the lead front runner.  Similarly, other Board of Supervisors races this year requiring RCV tallies had far higher percentages for the two front runners, including 59% in District 6, 78% in District 7 and 82% in District 2.

But in District 10 the two front runners had less than 25% of the first round ballots. It was the most unusual race that San Francisco has ever seen due to the high number of candidates, the lack of any clear front runners, and ultimately in its closeness among five candidates. Given those atypical circumstances, ranked choice voting was decisive in selecting the legitimate winner preferred by the most voters. 

COMPARISON TO OTHER ELECTORAL METHODS.

To understand the effectiveness of RCV, it is helpful to compare how this extraordinary race in District 10 would have unfolded using other electoral systems, specifically a plurality ("highest vote-getter wins") system or San Francisco's previous two round (November-December) runoff cycle.

Plurality elections.  With a plurality system, the highest vote-getter wins regardless of how low their percentage of the vote.  Plurality elections are used to elect the governor and other statewide offices, such as the attorney general in which this year’s winner will fall well short of having a majority of the vote, as have several recent California governors (as of this writing, the attorney general’s race is still undecided because of the closeness of the contest and a slow ballot count resulting from the vast number of absentee and provisional ballots). If the District 10 supervisorial race had been decided using this electoral system, the winner would have had only 12.06% of the popular vote and barely 2000 votes out of over 17,000 cast.  The second-place finisher would have lost by a mere 30 votes, and the third and fourth place finishers by fewer than 60 votes. This would not have resulted in a very satisfactory or democratic outcome, as it would have meant that 88 percent of voters had selected a different candidate than the winner. San Francisco can be glad that it was not using a plurality system in the District 10 race.

Two round (November-December) runoff.  If San Francisco's previous method with a separate runoff election in December among the top two finishers had been used, the candidate with the broadest support and the actual winner, Malia Cohen, would not have made it to the December runoff.  Instead, the top two finishers facing off in the runoff would have been Lynette Sweet and Tony Kelly. Kelly would have made the runoff over Malia Cohen by a mere 27 votes and over Marlene Tran by only 29 votes.

So Kelly and Sweet would have squared off in December in what would undoubtedly have been a low turnout election.  Typically in San Francisco's old runoff elections, voter turnout would plummet from November to December, sometimes by as much as 40% of the vote.  That's because voters turn out in great numbers for the presidential or gubernatorial races in November, but lose enthusiasm over local races in December. Indeed, in the last December runoff for a District 10 race in 2000, voter turnout plummeted 46% from the November election.  The winner was elected with fewer than 5900 votes, less than 30 percent of the voters that originally showed up to vote in this race in November (see Table 1 below).

Table 1. Runoff election, December 2000

Advocates of returning San Francisco to a separate runoff election say one of its advantages is that, in a close race like this one in 2010, voters would be able to assess the benefits of the top two candidates. However if the past is any guide, not only would few voters return for that second election, but voters would have had to withstand a month and a half of more mudslinging and hack attack campaigning in which they heard the worst about their future supervisor. The tone of the campaign would have become increasingly bitter, and rather than some candidates openly endorsing competitors for second choice (as occurred under the RCV system) voters would have seen “everyone for themselves” behavior from the candidates.

In addition, a December runoff would have been expensive for candidates, undermining campaign finance reform. So-called 'independent' expenditures' focused on attack ads soared in the old runoff system, with the San Francisco Ethics Commission finding they quadrupled in December runoffs. The top two candidates in the runoff would have needed to quickly raise a lot of money, giving an advantage to whichever candidate has either more personal wealth or access to those organizations with wealth. And San Francisco taxpayers would have had to foot the bill for administering a second election, taking general fund revenue from other pressing needs at a time when the city is financially strapped.

Conclusion:  A separate, second round runoff in December would have excluded the strongest candidate (Malia Cohen) from the runoff, exacerbated the ever-growing problem of money in elections, and resulted in a winner elected by a smaller, less representative electorate than a November election. Despite these shortcomings, a two round runoff would be far more preferable than the plurality method.

Ranked choice voting (also known as instant runoff voting).

The dispersed vote in the District 10 race challenged RCV more than any previous contest in San Francisco. Cohen's share of the first round ballots more than doubled during the course of the RCV tally but ultimately fell well short of 50% of the first round ballots. Still, the overall number of voters participating in the 2010 RCV race were comparable to the number of voters in the November-December 2000 runoff cycle. The number of voters participating in the December 2000 runoff and the final round of the RCV tally were in a similar range, as were the number and percentage of votes won by the winning candidate in both elections (see Table 2 below and compare to Table 1).

Table 2. RCV election, November 2010

Despite the challenges of this District 10 race, RCV has several features that make it desirable compared to a separate December runoff.  First, it elected the true winner with the most support, Malia Cohen, whereas a separate runoff among the top two finishers would have resulted in Cohen not making it into the runoff at all. Yet, in one indicator of Cohen’s strength as a candidate, election simulations show that she would have defeated any candidate paired against her in the final round. Cohen was the legitimate winner even though she would have lost by less than 0.4% in a plurality voting election and would have failed by 27 votes to make it into the December runoff.

In addition, Cohen, an African American candidate, was able to benefit greatly from the supporters of other African American candidates in the race (third place finisher Lynette Sweet and Dewitt Lacy in particular), resulting in the election of a black supervisor in this district historically won by black candidates (note that if no black candidate had won this race, that would have been the first time in decades that the Board of Supervisors had no black members). Cohen was more successful than any other candidate at building coalitions, which is plainly evident in following the trail of runoff rankings that accrue to her during 19 rounds of the RCV tally. About 60 percent of the voters in this race used all three of their rankings. The ranked ballots allowed coalition-building to be a decisive factor in who was elected ultimately, instead of the winner benefitting from the usual hack attack, big money campaigns.

Conclusion:  This race with 21 candidates and no clear frontrunners challenged the ability of any electoral system to accommodate it. This was a very unusual, even atypical race that San Francisco had never seen before and may not see again for a long time. Given the parameters of this race, RCV functioned smoothly to produce a winner that was preferred by the most voters. It fostered a degree of coalition-building as candidates and voters used the ranked ballots effectively, and unlike other races this race was substantially free of negative, mudslinging attacks as the multi-candidate field focused on seeking the second and third rankings from the supporters of other candidates.

Given the challenges of the District 10 race, reasonable people will disagree if it would be more valuable to have either a second election that helps voters to choose among the top two finishers in what is likely to be a low turnout, mudslinging election, or to finish the election in a single November election where the winner can emerge from the large pack of candidates on the strength of coalition building. But it is important to note that the negatives of a separate runoff election between the top two finishers would be prevalent no matter how many candidates enter the initial race in November, while the challenges that stem from using RCV decline as the number of candidates decreases to the norm that we have seen in other San Francisco RCV contests. The District 10 race was atypical in its complexity.

Postscript: Addressing Three Common Myths about RCV Elections in San Francisco

1. "RCV is responsible for delaying election results."

That is not correct. The fact that the statewide attorney general's race remains undecided due to a half million still-uncounted absentee ballots should put this myth to rest. Delays in RCV results are entirely due to two reasons: 1) With so many absentee and provisional ballots, San Francisco takes a long time to get all the ballots scanned in both non-RCV and RCV contests; and 2) the Department of Elections has made a policy decision to only release first rankings on election night, and to delay running the RCV tally until the first Friday after the election, when in fact it easily could run preliminary tallies on election night and each day thereafter. As the Registrar of Voters in Alameda County (which uses the exact same RCV voting equipment as San Francisco) recently stated in a PBS news broadcast, running the RCV tally literally takes seconds. It’s as quick and easy as walking to the computer that is used as the central tabulator, hitting a few keystrokes on the keyboard, and seconds later you have the results. Each time you run the RCV tally, you are doing it with the dataset of ballots that have been scanned to that point, and that provides preliminary results. It is simply untrue that you need to have all ballots scanned before the RCV tally can be run, since all results in all races, whether non-RCV or RCV, are preliminary until the election is officially certified 28 days after the election. Other jurisdictions that have used RCV elections have run the tally on election night using the exact same voting equipment that San Francisco has. Unfortunately by waiting until Friday to run the tally, the Department of Elections creates a false impression of “wild swings” in election results for some races.  In Districts 2 and 10, the candidate ahead on election night in first rankings in fact was losing to another candidate, and that would have been immediately obvious if the RCV tallies had been run for both races on election night. 

2. "RCV was responsible for Ed Jew (who was later convicted of bribery) winning in District 4 in 2006."  

Ed Jew would have won a plurality “highest vote-getter wins” election and most likely would have won in a December runoff as well. In 2002, Jew, who is Chinese-American, had a strong showing and finished third in this highly Asian American district. He was well-known in the district, and in the 2006 race he was the most popular Asian American candidate, finishing first in first rankings and then picking up significant second and third rankings from the supporters of other Asian American candidates.  The second-place finisher, Ron Dudum, was not Asian American and it is unlikely that Dudum would have been able to win a majority in this highly Asian American district in a December runoff. Dudum also failed to win a majority in the 2002 runoff when he ran against another Asian-American candidate, and a FairVote analysis of the 2006 ballots shows that any of the top three Asian American candidates in 2006 would have defeated Dudum if paired against him one-on-one. So Jew’s win with RCV is not surprising given the majority voting pattern of his district.  Some have speculated that the troubles that caused Ed Jew to eventually resign from office (he lived out of his district and was soliciting bribes, for which he was later convicted) would have been uncovered if there had been a December runoff, but this contention is incorrect.  The solicitation of bribes by Supervisor Jew did not occur until AFTER he was elected; and he had been living out of his district for a number of years prior to his election, but it was only uncovered during the investigation regarding the bribes. Indeed, when Mr. Jew ran for the same seat in 2002, it was not discovered during the campaign that he was living out of his district.  

3. "With RCV, you can win by being everyone's second choice."

RCV now has been used to elect San Francisco leaders in every November election since 2004. Many of those elections required RCV tallies to determine a winner, but this is the first year in which candidates who led in the first round were defeated by candidates who came from behind to win. That happened in two races, Districts 2 and 10, and both of those races were extremely close in which the ultimate winners initially trailed their top opponents by less than 1% in the first round.  With RCV, you can't ignore coalition-building but you also need to be a lot of voters' first choice as well. RCV represents an appropriate balance: you have to earn enough  first choice support from some people who really want you, but you can't be so polarizing that backers of all other candidates overwhelmingly reject you. Winners, ultimately, are those that have a strong core of support as indicated by having a substantial number of first rankings but also have a broad base of support indicated by having your share of second and third rankings from the supporters of other candidates.

Background info:

* Watch a seven-minute segment on the PBS New Hour with Jim Lehrer on Jean Quan's victory in the RCV mayoral election in Oakland: http://www.pbs.org/newshour/bb/politics/july-dec10/oakland_11-19.html

* Read Steven Hill's commentary in the San Francisco Chronicle on "The New Politics of Ranked-Choice Elections": 
 http://www.sfgate.com/cgi-bin/article.cgi?f=/c/a/2010/11/14/EDUH1GBFVT.DTL