Pollster: 'Oversampling' Can Help Capture Asian-American Views

A look at any major opinion poll before or after any election, and it's rare to see Asian Americans listed. It's even more rare to see Asian Americans listed by ethnicity like Chinese, Korean, or Vietnamese. That's because most pollsters don't take steps to ensure Asian Americans are included in random samples, says Mark DiCamillo, of the Field Poll in California.

Because of that, he says, Asian Americans can be inadvertently rendered invisible. One possible counterbalance, DiCamillo says, is deliberate oversampling.

“If you did not augment the sample with Chinese in Mandarin and Cantonese, and Vietnamese and Korean, you would not be including a fairly significant number of people who prefer to be interviewed in their native language,” DiCamillo said.

As demographics change, pollsters like DiCamillo say they're being forced to gradually make their polling more diverse -- an effort that takes effort and money, often in amounts prohibitive for many.

So in September, DiCamillo collaborated with UC-Riverside political scientist Karthick Ramakrishnan to produce a multi-ethnic poll of Asian-American voters. It was the Field Poll’s eighth large scale multi-ethnic poll to date, and at a cost of inclusion close to $50,000, it required foundation support.

But it did provide a random sample of 1,280 registered voters (including 280 Asian Americans). Targeted augmenting for Asian language speakers raised the number to 1,390 respondents. The value of the method, they say, is in the information most polls can’t provide.

For examples, the survey showed that Asian Americans supported the president with a 44.5 percent approval rate, but it also allowed for a drilling down by ethnic group. Chinese had a lower approval rate (41.8 percent), and a higher disapproval rate (36.6 percent). Vietnamese registered voters gave the president an approval rating of 50.9 percent, and a 14 percent disapproval rate. It also showed 35 percent with “No Opinion.”

Instead of a margin of error of 10 percent, the oversampling cut that to a more acceptable 5 percent.

“We’re increasing the precision and that enables us to look at the Chinese, Korean and Vietnamese subsamples within that, ” said DiCamillo. “That’s the advantage of the oversamples.”