Counting Non-Hispanic Blacks in America
The Bureau of the Census provides different ways of identifying these black populations, depending on the data source that is used.
For data on individuals the 1990 5% Public Use Microdata Sample (1990 PUMS) data files and the Census 2000 1% Public Use Microdata Sample (2000 PUMS) allow us to count the number of African Americans, Afro-Caribbeans, and Africans by combining information on their race, birth and ancestry.Among non-Hispanic blacks, we classify those reporting their ancestry and/or country of birth in the predominantly black islands of the Caribbean (including such places as Jamaica and Trinidad, but not Guyana) as “Afro-Caribbean. " We classify people reporting their ancestry and/or country of birth as a specific sub-Saharan African country as “African." We classify the remainder of the black population, including those who report their ancestry as “African" without a specific country reference but whose place of birth is not Africa, as “African American."
The 1990 and 2000 Censuses also provide aggregate data in STF4A (1990) and SF3 (2000) through which we can determine more precisely where members of these three black populations lived (in terms of metropolitan regions or even census tracts).Afro-Caribbeans are defined by ancestry in the predominantly black and non-Hispanic islands of the Caribbean (again including such ancestries as Jamaican and Trinidadian).However the available tabulations force us to define “Africans" solely by country of birth (sub-Saharan African). This means that our counts at the national level, from the 1990 and 2000 PUMS, include group members of all generations, but our analyses at the metropolitan or tract level only include first-generation African immigrants. Readers should be aware that this implies a substantial underestimate of Africans in metro-specific tables.Based on national data, the “true" African population in each metro area including immigrants and their descendants might be 20% higher than our count.The African American population may be slightly overestimated for this same reason.
These data are sample estimates (based on census returns from one of every six households), rather than population enumerations.At the national level they provide a very close approximation of group characteristics.Also when we combine information from many census tracts to calculate metro-level measures, error from sampling has only a small effect.