What's In a Name? Sampling Latinos by Surname

Latinos, the largest growing minority in the United States, present special challenges to survey researchers. In addition to posing the same difficulties inherent to polling any minority, Latinos also comprise subgroups representing different countries of origin. Some Latinos have lived in the United States for several generations while others have only just arrived, and the ability to speak English and Spanish varies widely. Along with other cultural and demographic attributes unique to this group, such considerations require researchers to be particularly meticulous when attaining representative samples of the Latino population.

Regardless of their increasing numbers, Latinos still currently comprise only 13.8 percent of the population, according to the U.S. Census Bureau's March 2003 Current Population Survey supplement. This means that initial screening questions will encompass the lion's share of cost in most Latino telephone surveys. In fact, a survey that costs $50,000 to interview any adult age 18 and older could cost three times as much when screening only for Latinos, overall sample sizes being equal. For most researchers, this constitutes a significant cost concern that often spurs them to look for alternative sampling options.

One such method of sampling Latinos is the use of listed samples consisting of people with distinctive Latino surnames. However, a comparison of Latinos by surname status finds significant concerns with the validity and reliability of this type of sampling procedure.

Surname sampling is an attractive option because it can cut the cost of a survey in half compared to a disproportionate stratified sampling design, the procedure generally used for hard-to-reach populations. However, there is a price to pay for the use of surname sampling in terms of coverage. Currently, approximately 65 percent of U.S. households have listed telephone numbers. Thus, the remaining 35 percent has no chance to be selected in this design. This problem is something survey researchers have had to deal with for years in all designs using listed samples. Most have found moderate differences between listed and unlisted samples.

What is particular to the surname design is the question of whether respondents with distinctive surnames are significantly different from individuals without them. It is critical to understand the extent to which Latinos with distinctive surnames are representative of Latinos in general. Some concerns become obvious simply by considering the potential makeup of Latino families. For one, Latino women who have married non-Latino men are almost certainly excluded from Latino surname lists. Aside from concerns about basic demographic measures, this leads to the question of whether there is bias in measures of acculturation and assimilation in Latinos when using listed surname samples. Such concerns clearly underscore the importance of understanding whether Latino surname sampling is biased, and if so, the extent of this bias.