Methods: Self-classifications on National YTS changed over time as race/ethnicity questions changed. In early cycles (American Legacy Foundation), one question allowed a multiple response for race/ethnicity combined, with Hispanic as a response options; another question asked for a single best-classification for combined race/ethnicity. Beginning in 2004 after NYTS transitioned to CDC, one question asked about Hispanic origin, and a second question allowed multiple responses for race (to comply with federally-mandated OMB Directive 15). The analysis looks at pairs of race/ethnicity categories allowed by multiple race responses and computes conditional probabilities for a single-race response given each combination of multiple-race responses. It also presents models for the probability of each response as a function of other student characteristics.

Results: Proportions of students reporting each single race, combination of race/ethnicity, and patterns of missing data for the race or ethnicity response will be presented. Conditional probabilities will provide a guide for imputation of missing data and for assigning a single race for when multiple races are reported. Prevalence trends by racial/ethnic subgroup will be examined.

Conclusions: Analyses inform discussion of multiple sampling, survey design and analytic issues. For example, the percentages of self-reported Hispanic students have grown in time as a function of both of actual population increases and an increased tendency for Hispanic self-classification. This increase mitigates the need for over-sampling Hispanics.