Ethnicity is difficult to measure in complex, multi-ethnic societies. Classification distortions arise when using mutually exclusive but overlapping categories due to mixed ancestries and multiple social identities. In a web-based survey, we compared categorical classification with different measures of ethnicity by ancestral origins. Respondents first completed a traditional categorical measure of race/ethnicity (with a separate Hispanic item) and then asked about their and their parents‘ birth state/country. We randomly assigned one of 3 response format with the first asking about the birthplace of their grandparents. Two other response formats asked "Thinking about your own ancestry, about what percentage of your ancestors came from the following groups or locations?" and presented 7 alternatives (Spanish-speaking or Portuguese-speaking nation, Africa, Asia/South Pacific Island, Europe, Native American/Eskimo, Middle East, South Asia, plus "another place"). One response format had fixed categories (None, Less than 5%, 5% to 24%, 25% to 49%, 50% to 74%, 75% to 94%, 95% or more) while the other format used a constant sum task (0-100) to distribute ancestry proportions. While all three measures were correlated with cultural identification, the relationships between ethnic and racial identification, ancestry, and cultural identification were more complex than first expected.