CONSIDERATIONS OF SURVEY ERROR IN SURVEYS OF HISPANICS

As the largest and fastest-growing minority population in the United States, Hispanics have become an increasing focus of survey research. The vast body of Hispanic research evidences myriad options regarding sampling, data collection, and weighting, each of which can affect the resulting data about this population. Typical survey designs feature simple random samples (sometimes obtained as part of larger omnibus or general-population surveys), stratified RDD, "top market," and surname designs. In addition, some studies obtain interviews in English only, while others offer both English and Spanish but make choices regarding the use and allocation of bilingual interviewers. Finally, there are a range of considerations in the weighting of Hispanic survey data. Utilizing data from a national omnibus survey, the General Social Survey, and the Pew Hispanic Center National Survey of Latinos, this article explores these three foci: sampling, interviewing language, and weighting. We report on what we find to be best practices and the implications of failing to enact these practices, as measured by bias and variance in survey estimates of Hispanics.