Kiss my ass!!
of Management and Budget (1997) for additional details]. This creates a problem in our analysis because it is not at all clear that someone who responds affirmatively to the Hispanic origin question and selects white as her race is mixed, at le ast not mixed in the same way that someone who selects white and Asian is mixed. For example, a person who identifies as Hispanic and white may believe that she and her parents are of Mexican heritage, and that none of her ancestors mixed with the indigenous population of Mexico (i.e., she is white). Alternatively, another person who provides these same answers to the Hispanic origin and race questions may have a mestizo mother from Mexico and a white father from Ireland. These two cases represent very different situations. The latter would probably be considered mixed-race by most people while the former would not. A second limitation of Add Health is that 24 percent of nonhispanic adolescents were only interviewed at home because some schools agreed to le t their students be selected into the in-home sample, but refused to allow in-school interviews. This missing data is problematic because it means that for a substantial minority of students we cannot compare answers to the home and school race questions. 6 Third, the racial identity of biological parents is only ascertained if adolescents live with their biological parents at the time of the home interview. Due to high rates of out-of-wedlock births and marital dissolution (Farley 1996; Smith, Morgan, and Koropeckyj-Cox 1996), a large share of adolescents do not live with both biological parents. The result is that further restricting analyses to adolescents who live with both biological parents reduces the sample by an additional 57 percent. This reduction in sample size is distressing because it limits the number of mixed-race people for analysis, but more importantly because it results in a biased sample of adolescents and thus eliminates one of the key reasons for using Add Health for this study. Given the limitations inherent in restricting the sample to adolescents who live with both biological parents, associations between other variables and parent race will seldom be examined with Add Health data, and when they are it should be understood that our findings might be biased. A fourth limitation of Add Health is that the interviewer is asked his/her perception of a respondent’s race immediately after the respondent answers the home race and best race questions. As a result of this unfortunate ordering of race questions it is highly likely that interviewer’s perception of the respondent’s race is biased by the respondent’s expressed racial identity. This bias is of no consequence for the vast majority of people whose expressed and perceived racial identities are identical, but for the mixed-race population bias in the interviewer perception of race questions limit their usefulness. One final limitation of Add Health is that despite its large initial sample size, once Hispanic adolescents and those who did not complete school surveys are excluded, there are only enough cases to support an assessment of three mixed-race groups—white/black, white/American Indian, and white/Asian adolescents.