Commentary: Accuracy of Hispanic Polling
By Edward T. Rincon, KERA Commentator
Dallas, TX –
By the year 2042, the Census Bureau tells us that over half of this nation's population will consist of African-Americans, Hispanics and Asians. As these ethnic segments become a larger part of our nation's population, knowing what they believe and what they value will be increasingly important in guiding national policies and programs. The close competition in the current presidential race, for example, may very well be shaped by the political sentiments of potential swing voters like Hispanics.
Despite the changing demographics, however, many pollsters continue to use outdated methods to measure the public opinion of our nation's diverse audiences, and sometimes provide misleading information to decision-makers who rely heavily on such polls. The traditional margin of error associated with a poll tells us how close the poll results are to the true population value, but it provides no assurance that the poll is not misleading or biased in other important ways.
Polling audiences like Hispanics, for example, poses special challenges that can compromise the statistical integrity of a poll or survey. From my 30 years of experience in polling Hispanic and other multicultural audiences, I have observed five common practices in the polling industry that can produce misleading results. For example:
- When contacting Hispanic households, it is not a good idea to rely exclusively on the terms "Hispanic" or "Latino" to identify the desired respondents. The terms Hispanic or Latino are strange concepts to many Mexicans and persons born in other Central or South American countries where the terms are not used. Asking a respondent's country of origin is helpful in capturing more of these individuals.
- Not all Hispanics want to hear a greeting in Spanish when called by pollsters, especially native-born Hispanics, who comprise 60 percent of the U.S. population and are more proficient in the English language. Unless bilingual interviewers are used to greet respondents in either language, a study can end up with a higher proportion of foreign-born Hispanics who are more comfortable with Spanish.
- Many pollsters and sponsors are conducting English-only surveys, despite the fact that two-thirds of Hispanics, and nine in ten Asians, prefer an interview in their native language. The exclusion of native-language speakers from surveys tends to distort important statistical indicators, such as satisfaction ratings, media audience size, health outcomes and others.
- The sample size of Hispanic respondents in a poll is often too low for the geographic coverage at hand, making it very difficult to find important relationships even when they exist, and obscuring the variation in Hispanic attributes. Using small samples is analogous to examining a lab specimen with a low-powered microscope - the detail is not observable unless the power is sufficient for the job.
- And lastly, the practice of using Spanish surnames exclusively to select a sample of Hispanic households will likely miss about 30 percent of Hispanics that do not have common U.S. Spanish surnames due to intermarriage or birth in a foreign country. Consequently, other sampling techniques are needed to capture this segment.
As a final word of advice, you should avoid the tendency to quickly embrace the results of polls involving Hispanics without first understanding how the poll was conducted, especially when the stakes are high. Polling organizations should be encouraged to adopt some of these practices and disclose such information to the public where it can be readily seen, not buried in obscure sections of the report. Regardless of their motivations, outdated polling practices have no place in a nation that is undergoing such dramatic demographic shifts.
Dr. Edward Rincon is President of the research firm, Rincon & Associates and an instructor at SMU.
If you have opinions or rebuttals about this commentary, call (214) 740-9338 or email us.