Producing Population Estimates Using Vital Statistics Data with Limited Race Detail

Amy S. Smith, U.S. Census Bureau
Nicholas A. Jones, U.S. Census Bureau

The recent changes in the standards for tabulating data on race and ethnicity greatly impact systems for producing population estimates. The Census Bureau shifted to the new Office of Management and Budget (OMB) race standards starting with Census 2000. Other federal agencies are mandated to adopt the new recommended standards by January 2003. This lag suggests that early in the decade the component data necessary to produce population estimates, such as vital statistics data, will likely have race categories that are incompatible with the population estimates. The Census Bureau produces annual population estimates with age, sex, race, and Hispanic origin detail for various geographic levels. Vital statistics data from the National Center for Health Statistics (NCHS) are used to produce these estimates by the cohort-component method. Birth data are used as input to indicate change in the population since the last census. Historically, the Census Bureau has incorporated birth data in the Population Estimates Program (PEP) by assuming that the race of the child is the same as the race of the mother. The Census Bureau’s Demographic Analysis (DA) produces estimates of the population assuming that the race of the child is the same as the race of the father. With the shift to the new race standards, and the option to indicate more than one race, it is no longer necessary to assign the race of the baby to be consistent with either the mother or the father’s single race report. In this paper we describe research currently underway to devise a model for assigning race for babies that is consistent with the new race standards and the option to indicate multiple races. The race of the mother and the race of the father as reported on the birth certificate, are both introduced in the model as factors in assigning the race of the baby. The model relies heavily on our earlier research suggesting that race specific patterns exist in the way that children and their parent(s) reported race in Census 2000.