Comparing Techniques to Identify Spanish-Speaking Households

In the United States, recruiting representative samples of the Hispanic population has been challenging due to social factors such as culture, language, and trust levels. However, reaching this population is crucial for ensuring the representativeness and accuracy of public opinion data, particularly because this group has unique perspectives on topics including immigration, education, family, and healthcare. Previous research has explored methods for identifying Spanish-speaking households and improving panel recruitment strategies targeting Hispanic populations. This study compares three techniques used to identify Spanish-speaking households for a 2023 nationally representative survey of over 35,000 respondents. Specifically, we assess the accuracy of using vendor-provided data, Bayesian Improved Surname Geocoding (BISG), and Census Spanish language isolation locations in identifying likely Spanish-speaking households. Using sensitivity, specificity, and predictive values screening tests, we show that the BISG flag is the most sensitive technique, identifying the highest proportion of households where a respondent completed the survey in Spanish (82%). We conclude by providing recommendations for future surveys on identifying likely Spanish-speaking households.