The quality of race and ethnicity information is compromised for many reasons. Observer bias is a significant source of error because an interviewer or data collector may incorrectly classify an individual as belonging to a race or ethnicity other than the self-identified ones. When comparing self-reported to interviewer-generated generated race and ethnicity information using an earlier version of the National Health Interview Survey (NHIS), Massey found that the 6% of individuals who self-identified as black, 29% as Asian or Pacific Islander, 62% as American Indian and 80% as other, were classified as white by their interviewer (Massey, 1980). Demographic data that is collected via self-reported information is considered to be the “gold standard” in disparities research.
Another example of observer bias relates to the National Death Index. Race/ethnicity data on death certificate is inaccurate because of inferred information on the deceased. Scott and colleagues found that only 63% of medical examiners, 50% of coroners and 37% of funeral directors communicate with family members to obtain a decedent’s race/ethnicity (Williams, 1999).
Respondent reliability is also a major source of error for race/ethnicity data. Researchers have estimated that up to one-third of the U.S. population has reported different race or ethnicity information from one year to the next (Johnson, 1986). There are also opportunistic self-identification shifts that can occur within the U.S. population. For example, from 1960 to 1990 there was a dramatic increase in the Native American population in the U.S. that could not be explained by increased reproductive rates or international migration. Instead, individuals who previously self-identified as white began to self-identify as Native American, most likely due to economic incentives and decreased societal discrimination (Passel & Berman, 1986).