Lost in Translation? Validity and Reliability in Bilingual Political Surveys.

Are survey items comparable across linguistic groups? As the largest minority in the U.S., Latinos are the focus of increasing attention from survey researchers interested in understanding their political attitudes and behavior. To that end, scholars have relied on a growing repository of survey data to measure a wide array of constructs among Latinos, including political identity, political sophistication, policy attitudes, and political values. However, because these constructs and their attendant measures were primarily developed and tested on English-speaking populations, it is reasonable to expect discrepancies in the performance of these items among a bilingual population like Latinos. In particular, survey items may not capture the same political construct across English-speaking and Spanish-speaking Latinos because the same items may mean different things to both language groups. Using structural equations modeling (SEM), this paper tests the invariance of political constructs across English-speaking and Spanish-speaking Latinos. By relying on commonly used data sets, such as the Latino National Political Survey (1989-1990), the Multi-City Survey of Urban Inequality (1992-1994), and the National Survey of Latinos (2004), this paper tests whether survey measures in English are equivalent to survey measures in Spanish. In this way, the paper aims to facilitate the development of constructs and survey items that more faithfully reflect the linguistic parameters of an increasingly diverse U.S. society. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]