Thursday, November 17, 2011

Analysis of Ranked-Choice Voting: SF D.A. 2011

The bulk of the media attention to last week's local election in San Francisco has been directed at the mayoral race — the first competitive contest under both public financing and ranked-choice voting (RCV). However, digging a bit beneath the surface of the election of the District Attorney sheds some light on how RCV played out. I was able to download the entirety of the ballots from a public website. To be clear, I cannot tell how a specific person voted; however, I can tell you exactly how many people voted for Bill Fazio in 1st place, George Gascón in 2nd place, and David Onek in 3rd place: 1,055. I can also tell you that the progressives who supported the ticket of Onek–Bock–Fazio (recommended by the San Francisco Bay Guardian/sfbg.com and others) had the unintended effect of helping George Gascón to an easy win. Although she was the second choice of the "anybody but Gascón" crowd, Sharmin Bock's core supporters clearly were not fully on board with the concept.

An important caveat is that the data I downloaded were preliminary results, and do not include the last 2,603 ballots (1.39%) to be processed.


Before I get into the heart of the discussion, there are some interesting factoids buried in the data. The most popular ballot cast — and indeed the only one more popular than completely blank — was for George Gascón only, with no backup choice. A total of 24,045 people, or 13.92% of voters who expressed any preference at all, voted only for George Gascón, including the 102 who voted for him only as their 2nd choice and the 45 who voted for him as both 1st and 3rd choice but nobody for 2nd choice. Far more relevant is that Gascón was marked somewhere on the ballots of 107,731 voters (62.36%), compared with 106,391 voters (61.58%) who supported Sharmin Bock. David Onek was in a distant third, with only 78,713 voters (45.56%) giving him the nod at any level, barely ahead of exactly 72,000 (41.68%) who supported Bill Fazio. (The percentages add to more than 100% because I am combining 1st, 2nd, and 3rd choice votes in the totals.)


Bock eliminated
Round 2 pickup
from Bock
total votes
Gascón 80,538 +15,786 96,324
Onek 44,606 +12,528 57,134
Gascón's margin of victory: 39,190

Onek eliminated
Round 2 pickup
from Onek
total votes
Gascón 80,538 +6,805 87,343
Bock 42,391 +27,940 70,331
Gascón's margin of victory: 17,012


In the first round of the RCV process, Gascón led with 72,453 votes, with Onek at 40,066 and Bock at 35,920. Because it was mathematically impossible for Bill Fazio (18,114 first-place votes, and only 72,000 total votes) or Vu Vuong Trinh (6,210 first-place, 24,963 total) to overcome Gascón's lead, they were both eliminated in the first round, with their votes transferred to their 2nd or 3rd choice selections. That left Gascón still with a sizable lead but less than 50%, with Onek and Bock neck-and-neck for runner-up, but Onek's first-choice advantage enabled him to squeak past Bock into the final round, even though Bock picked up half again as many votes from Fazio and Trinh as Onek did. Thus, Sharmin Bock was eliminated and her votes redistributed to Gascón and Onek. Those Bock voters, though, supported Gascón by a margin of 15,786 to 12,538 for Onek, allowing Gascón to coast to an easy victory.

If Sharmin Bock had been able to pull ahead of David Onek, his supporters' next-choice votes would have favored her by more than 4-to-1 over George Gascón, 27,940 to 6,805. The final outcome would have been the same — Gascón's lead was simply insurmountable — but it would have cut his margin of victory by more than half. If the overall race had been closer, Gascón's fate could easily have ridden on who came in second.


Total votes by rank
1st 2nd 3rd total
Sharmin Bock 35,920 51,660 18,811 106,391
George Gascón 72,453 22,012 13,266 107,731
David Onek 40,066 20,333 18,314 78,713


A few more factoids: Sharmin Bock received more 2nd-choice votes than George Gascón and David Onek combined, and also led in 3rd-choice votes. She was the favored 2nd-choice not only of the Onek–Bock–Fazio "anybody but Gascón" ticket, but also of Gascón's supporters and Trinh's supporters, but Bill Fazio's supporters backed Gascón. Bock also came in 2nd among voters who chose only one candidate, with 8,678 votes to Gascón's 24,045, Onek's 6,187, and Fazio's 5,317. The Bay Guardian's "clean slate" (O-B-F) received 10,721 votes, making it the third most popular ballot, after "Gascón only" and "no vote." Out of 184,808 total ballots, including 12,045 undervotes, only 90,072 voters selected three different candidates as their 1st, 2nd, and 3rd choices. A total of 45,800 voters chose only one candidate; of those voters, 37,768 marked their candidate only as a 1st choice, 301 only as a 2nd choice, and 157 only as a 3rd choice.

There has also been a great deal of discussion about voter confusion regarding the process of ranked-choice voting. Some of the odd ballot markings do indeed demonstrate that some voters were unclear on the concept. For example, 157 voters chose no one for 1st or 2nd, but selected one of the candidates as their 3rd choice. (Those ballots were promoted to 1st-choice, so they did count.) While it is tempting to picture a voter making a stand for "None of the Above" as their first choice, defiantly digging in on the second, with a reluctant concession to reality in the third-choice slot, it's rather more likely they simply didn't know what they were doing. Even more confused were voters who chose the same candidate for 1st and 3rd, but a different one for 2nd. A total of 1,312 voters made that nonsensical mistake, led by the 240 who voted for Gascón–Fazio–Gascón. There were 9,277 voters who marked the same candidate in two or three consecutive slots; the extra ranking was simply disregarded. Still, with 172,763 marked ballots, the overwhelming majority cast clear and unambiguous ballots, suggesting that the "voter confusion" angle is little more than media hype.