Hispanic origin and race are problematic data elements in censuses and surveys. Most research to improve the quality of these data elements has been done in household surveys where a householder reports origin and race data on a few well-known correspondents. Very little research has been done in group quarters/establishment surveys where a third-person reporter may report origin and race data on many persons not personally known to him/her. The mix of factors affecting the accuracy and completeness of origin and race data in establishment and household surveys differs. This paper presents qualitative research results on third-person reporting of Hispanic origin and race for residents in juvenile facilities, using data from 33 cognitive interviews in 11 states. The paper summarizes the literature. It describes the successive results of testing, revising, and retesting experimental Hispanic origin and race questions in a multi-stage project to redesign the Children in Custody Census for the Department of Justice. Definitional, methodological, legal, and other factors affecting the quality and completeness of origin and race data provided by third-person respondents are discussed.