BY GEORGE GAO
One question we’re often asked is, Why aren’t Asian Americans shown as a separate group when differences among whites, blacks and Hispanics are discussed?It’s worth noting that Asians are indeed included in our U.S. surveys. While we often do not break out their standalone views, Asians’ responses are still incorporated into the general population figures that we report.Nonetheless, it’s a good question, and one we hear frequently, so we put together a summary of some of the methodological and other issues on accurately polling U.S. Asians.
The demographic challenge of polling Asian Americans
The most obvious hurdle is the relatively small number of Asians in the U.S. compared with other races and ethnicities. According to 2014 Census Bureau estimates, 5.4% of U.S. adults are Asian, compared with 11.7% who are black, 15.2% who are Hispanic and 65.1% who are white.
A typical survey of the U.S. population consists of 1,000 adult respondents. This threshold reflects the way in which many researchers attempt to balance survey cost with survey quality: A national sample of 1,000 interviews yields reasonably good precision for major subgroups defined by gender, age, race and ethnicity; much larger sample sizes will reduce the margin of error minimally but also make polls much more expensive. (Pew Research Center samples are typically 1,500 respondents or more.)
A common rule of thumb used at Pew Research Center and elsewhere is to only report subgroup estimates if they are based on at least 100 respondents. In a perfectly representative survey with 1,000 adults, we would expect about 152 Hispanics, 117 blacks and just 54 Asians – with the latter subgroup falling well under the 100-respondent limit. The size of this Asian subgroup would simply be too small to make reliable estimates of the views and experiences of Asian American adults as a whole.
For example, if we were to find that “50% of U.S. Asians approve of …,” the margin of error for the Asian subgroup of 54 respondents would be 13.3 percentage points – meaning that somewhere between 37% and 63% of Asians in the country would “approve of” the hypothetical topic. That range, or “confidence interval,” would simply be too large to report.
While increasing the sample size would help, it’s not enough to address another issue ….read full article here.