The story of San Francisco’s June 2018 mayoral election is one of time and space: when San Franciscans vote and where they did it.
By now you probably already know that London Breed prevailed in the race over former State Senator Mark Leno by about 1% (or 2,000 votes) after all the ranked-choice votes had been tabulated.
The outcome of this close race was in question for about a week after Election Day. The initial returns showed Breed winning, but that changed quickly. Leno took the lead by the end of election night and continued to narrowly lead through each daily update for the next several days.
Here’s how it eventually shook out for the top three candidates:
Time: late deciders broke for Breed and Kim, and just broke Leno
The reason Leno led for such a long time initially is because the Department of Election processed different populations of ballots at different times. The first returns processed (about 79k) were vote-by-mail (VBM) votes received well in advance of Election Day and were extremely favorable to Breed. The next returns processed were those cast in person on Election Day (another 75K) and showed a marked improvement for Jane Kim.
Jane Kim’s election day boost was enough to change the result, after calculating the ranked-choice vote (RCV), from London Breed winning by 4% to Mark Leno winning by <1%, because the vast majority of Kim’s voters put Leno as their next choice.
The remaining 100,000-ish votes that had yet to be tabulated were mostly VBM voters who waited until the last minute to post their ballots, or turn them in at a polling location. These voters were like early VBM results in that they strongly preferred Breed; they were also like election day in that they preferred Kim over Leno.
The reason these chronologically disparate populations matter is because it’s a window into how San Franciscans made their decisions.
Voters who made their decision (and sent in their ballot) early preferred Leno over Kim by a wide margin, but preferred Breed most of all. Voters who waited until the very end preferred Kim over Leno.
This dynamic mirrors the Google search activity in the last 30 days of the election.
Voters who were still making their decision, researched Jane Kim and London Breed at about twice the rate they searched for Mark Leno. This suggests that Mark Leno relied on the support of voters who already knew him from his decades in public service, and less on new voters who weren’t already familiar with his legacy.
Such a hypothesis is supported by Leno’s consultant, who believed that these early voters skewed much older than the subsequent populations. This also suggests that Leno’s single-minded focus on campaign finance failed to move voters that didn’t already have an affinity for him.
Kim beat Leno by 6,000 votes among election day and late VBM voters. Leno only prevailed over Kim because of strong support from early VBM voters (where he beat her by 6,000) and transfers from other candidates (he received a lot of ranked-choice votes from Alioto and Greenburg).
Space: Leno wins only at home, Kim wins everywhere, but Breed just wins
San Francisco’s voter turnout was very high, but it wasn’t consistent across the city’s 11 equally populated supervisorial districts. The voters of District 8 (Castro, Noe Valley, Glen Park) voted at more than twice the rate of their peers in District 11 (Excelsior, Crocker-Amazon, Outer Mission).
This lopsided turnout was good news for Mark Leno. No candidate received as many first choice votes as Mark Leno did in District 8.
However, the impact of District 8’s massive turnout was muted by the fact that the vote in there split relatively evenly (Leno only beat Breed by ~600 votes out of the 36,000 cast) among the top three candidates. Leno, Breed, and Kim all received more first choice votes from District 8 than they did in any other district. Kim’s third place finish in District 8 still yielded her more votes than her first place finish in District 9 (Mission, Bernal Heights, Portola).
Lower-turnout districts had a disproportionate impact on the results if those voters overwhelmingly coalesced around one candidate. Voters in Districts 2 (Marina, Cow Hollow, Pacific Heights) and 10 (Bayview, Dogpatch, Potrero Hill) heavily favored Breed. Despite overall lower turnout, they were Breed’s third and fifth largest sources of votes, respectively. They were the only places where the vote for Breed exceeded the combined votes for Leno and Kim. She netted more than 2,000 votes over Leno and Kim combined in those neighborhoods – which was almost exactly Breed’s margin of victory.
Kim and Leno received almost exactly the same number of first choice votes, but the patterns of support could not be more opposite. Leno’s support (blue indicates areas where he did better than average) was extremely concentrated. He was very popular in a few areas, and quite unpopular outside of those areas:
The narrowness and depth of his support among District 8 voters is further evidence that he was relying on people who voted for him in the past.
Kim’s support (blue now indicates areas where she did better than average), on the other hand, was less dense, but much broader:
She beat Leno in 7 out of 11 districts. He beat her only in four, but handily enough that he came in second overall.
While Leno and Breed relied on their home districts to substantially boost their first choice vote totals, Kim’s support in the district she represents (District 6: SOMA, Tenderloin) was quite low. She came in a distant second to Breed, and only beat Leno by ~50 votes.
Her lack of support in District 6, which was an issue during her State Senate race, may have doomed her chances. Had she done as well in her home district as Leno and Breed did in theirs, she would have come in second place overall by at least 2,000 first place votes.
Breed’s support (blue indicates areas where he did better than average) was deeper than Kim’s and broader than Leno’s, which is how she beat them by more than 10% among first-choice votes.
Breed’s winning coalition shares some similarities with Scott Wiener’s victory in the November 2016 State Senate race. Like Wiener, Breed performed very strongly with voters in Districts 2 and 7 (West of Twin Peaks). However, Wiener relied on similarly high support from District 8, which Breed split almost evenly with Leno. Breed’s coalition instead included supercharged support from District 10 and much more support from District 5 (Haight, Hayes Valley, and Western Addition) than Wiener had received.
You, Me, and RCV
Given that none of the candidates received (or expected to receive) a simple majority of the first place vote, the election results hinged on ranked-choice voting system.
Mark Leno was the single biggest recipient of transfer votes. He received over 52,000 votes over the course of 9 ranked-choice elimination rounds, more than twice as many as London Breed. Almost half (46%) of Leno’s final round votes came through transfers from other candidates – only 21% of Breed’s votes came through transfers.
Most of Leno’s votes came from Jane Kim supporters. Almost 70% of her votes transferred to him, which suggests that their strategy of encouraging their supporters to vote for the other as their second choice was very effective. Some authors have suggested that Kim cannibalized enough votes that would have otherwise gone to Leno to cost him the election. It seems far more likely to me that her presence boosted overall turnout in a way that was ultimately to his benefit.
Given the dearth of enthusiasm for Leno outside of District 8, there is reason to doubt that the lion’s share of Kim voters would have turned out if she not been on the ballot. This was among the closest, highest-turnout open mayoral elections in living memory, and it certainly didn’t get that way because there were too many candidates on the ballot.
Breed’s supporters voted early, and they voted everywhere; Leno’s base in the Castro voted often, but not enough elsewhere; and Kim’s voters mostly turned out in droves on election day. Any slight variation in these factors could have given us a different mayor.
Aren’t you excited to do this all over again next year?
Andy Mullan is the Beacon’s very own data guy. He analyzes data and their impact on local politics in our regular feature, SFByTheNumbers. Find more in-depth, obscure data analysis and adventures in new adulthood at his blog, Lord of the Fails and his twitter @askmullan.
Sign up for the Bay City Beacon weekly newsletter! It's a free way of getting the best of political gossip and cutting-edge culture in your inbox every Friday.