U.S. racial and ethnic populations can be defined by a number of census questions-race/ethnicity, ancestry, place of birth, and/or language-but little is known about how using alternative definitions of identity affect the size and characteristics of different groups. This article examines this question using combined data from the 1 % and 5 % Public Use Microdata Samples in census 2000, using Mexicans and Arabs as case studies. The analysis uses the standard method of classifying these groups (Hispanic origin and Arab ancestry) as a baseline to explore differences across the range of possible permutations of ethnic identity. In the Arab case, persons captured using alternative definitions of identity (Arabic language at home and/or born in an Arab country) are lesser educated, more likely to be in poverty, and more likely to identify as non-white or multi-racial than the Arab population as a whole. In contrast, persons in the Mexican alternative definition group (Mexican ancestry and/or born in Mexico) are more highly educated, less likely to be in poverty, and more likely to identify as white than the Mexican population as a whole. The article concludes with research and policy implications of these findings. © 2013 Springer Science+Business Media Dordrecht.