Pardon me while I trumpet our poll’s accuracy.

According to the secretary of state’s office, the semi-official results of the races for governor and the U.S. Senate give Jerry Brown a 12-point victory over Meg Whitman and Sen. Barbara Boxer a 9-point edge over Carly Fiorina in the Nov. 2 elections.

As has been reported, of the 14 pre-election polls released within 10 days of the vote, the two that came closest to predicting the actual outcomes were the USC College/Los Angeles Times and Field polls. The most accurate poll on the top two races was the USC/L.A. Times October poll, which had Brown 13 points ahead and Boxer up by 8 points.

The previous month’s USC/L.A. Times poll showed Brown leading by 5 points, while two other highly respected public polls had the governor’s race even (Field) and Whitman leading by a point (Public Policy Institute of California). Whitman’s pollsters quickly branded our survey a “Democratic poll.” Other pollsters linked the finding to a likely-voter model that they said was an “unrealistic” representation of the 2010 electorate.

The October poll results attracted the ire of the GOP candidates themselves.

Bragging rights aside, the main reason the USC/L.A. Times poll was almost dead-on in its prediction of the two races was that it reflected real Latino voter numbers.

As do other pollsters, our samples were randomly culled from voter lists. We called cell as well as land-line phones, surveyed respondents with live interviewers and used a range of measures of past electoral behavior and self-reported enthusiasm among respondents to model likely voters.

We did one thing differently: In the Latino voter interviews, our interviewers spoke English and Spanish fluently.

Why use bilingual interviewers?

While the majority of Latino voters in California speak English, a substantial proportion speaks mostly Spanish or feels more comfortable doing an interview in Spanish. These voters are much more likely to lean Democratic. If these Spanish-speaking Latinos are systematically under-represented in a poll’s sample, the effect is to make Latino voters look less Democratic than they actually are.

When they are fully represented, as they were in the USC/L.A. Times sample, the point spreads between the candidates widen dramatically.

Most survey firms use bilingual interviewers but only after an English-speaking interviewer identifies a problem. Yet the probability of completing an interview with a non-English-speaking respondent under these circumstances is much smaller. As a result, the number of interviews completed in Spanish in most election surveys is less than it should be to reflect the composition of the state’s electorate.

An apt analogy to what I’m talking about comes from a hilarious sketch performed by the comedian Dave Chappelle. (Check it out on YouTube.) The skit centers on the difference between grape drink and grape juice. Grape drink, says Chappelle, is a combination of water, sugar and “purple.” It’s cheap and easy to get. Grape juice, by contrast, is made from the fruit.
Once you taste the real thing, it’s hard to enjoy the ersatz version.

Because most state election surveys don’t use bilingual interviewers from the get-go, their findings amount to drinking grape drink rather than grape juice. As a result, they inadequately portray just how Democratic Latino voters in California are.

Yes, this methodology is expensive and time-consuming. But it may be indispensable to knowing what’s going in the state’s increasingly diverse electorate.

While Latino voters now represent 1 in 5 registered voters in California, they are not the only minority group worth watching.

Asian Americans, currently 13% of the state’s population, make up a rising share of its registered voters (7%). More diverse in terms of language and ethnic origins than Latinos, Asian Americans are also more heavily immigrant, and the range of languages, in addition to English spoken by Asian American voters, includes Mandarin, Cantonese, Tagalog, Vietnamese, Korean, Hindi, Bengali, Japanese, and Hmong, among others.

Is it possible to do surveys in multiple Asian languages? Both PPIC and Field have recently done polls with Asian American voters using interviewers who speak two Chinese dialects, Vietnamese and Korean. The USC/L.A. Times 2010 post-election study is using interviewers who speak English, Spanish, Mandarin, Cantonese, Tagalog, Vietnamese and Korean. Stay tuned for the results. They’ll probably surprise you.