When respondents who were undecided were asked to choose a candidate, Wiener’s support rose to 51% and Chan’s to 44%. The survey was conducted by email by the University of Southern California’s Democracy and Fair Elections Lab.

The results may not fully capture Chan’s support among monolingual Chinese-speaking voters, who can be difficult to reach through polling, said Christian Grose, the survey’s lead pollster and a USC political science and public policy professor. Chan, who represents District 1, has previously done better in the vote count than in pre-election polls.

Before the June primary, three surveys showed Chan roughly even with Saikat Chakrabarti at about 20% each. Chan ultimately placed second with 29.7% of the vote, while Chakrabarti received 17.9%. In precincts where more than 30% of registered voters were Chinese American, Chan led Wiener 35% to 31%, according to an analysis cited in the poll coverage.

The latest poll found that 31% of Asian and Pacific Islander respondents favored Chan, compared with 59% for Wiener. The team emailed surveys in languages other than English to some respondents identified through surname matching and included language options in English-language survey emails. But Grose cautioned that reaching voters who speak only a non-English language remains difficult.

“Chinese voters are more likely to support Chan,” Grose said, while adding that the team saw indications other Asian voters may lean more toward Wiener. Those finer-grained findings were not published because the subgroup sample sizes were small.

The survey also found Chan drawing support among both Democratic Socialists of America members and Republicans, while voters who approve of the Democratic Party were more likely to support Wiener. In a race between two Democrats, Republican voters and those who lean Republican could help determine the outcome, Grose said.

Support for the candidates also differed by income: higher-income voters were more likely to favor Wiener, while lower-income and working-class respondents leaned toward Chan. Grose said San Francisco’s growth in higher-income residents could help Wiener.

Pelosi has endorsed Chan. The poll found that among respondents who strongly approve of Pelosi, 54% supported Wiener and 46% Chan. Grose said he expected Wiener might have done even better among Pelosi supporters without the endorsement.

The poll’s open-ended responses reflected competing campaign arguments. Respondents most often associated Wiener with being “corporate/donor-backed” or “effective/accomplished,” and Chan with being “progressive/liberal” or “NIMBY/anti-housing.” Despite widespread unfavorable views of the Israeli government among respondents, Israel-related issues appeared to play a minor role in how they described the candidates.

The poll was conducted by email, a method Grose said has become more useful as people increasingly ignore calls and texts from unknown numbers. Its results offer a snapshot of the race, but the close margin and potential difficulty reaching some language communities underscore the uncertainty in predicting the vote.