separate race responses and approximately 600 different American Indian tribes. The Hispanic-origin item elicited 70 different origin write-in responses, and the ancestry item, an open-ended item, elicited over 600 ancestry groups. Use of a computer-automated coding operation in the 1990 census greatly improved the quality of operation in terms of decreased clerical errors and reduced costs. Even with this automated system, however, the Census Bureau encountered difficulties in aggregating some 250,000 race write-ins that indicated mixed racial or ethnic identities or nationalities and write-ins without a clear racial connotation.
The 1990 census produced unprecedented amounts of data on racial groups and people of Hispanic origin, documenting the transformation of the United States to a multiethnic and multiracial society with 1 of 4 being black, Hispanic, Asian and Pacific Islander, or American Indian and Alaskan Native. It also documented the growth of new immigrants with significant impact on the Asian and Hispanic populations. Of the total foreign-born population, nearly 44 percent arrived between 1980 and 1990. One-third of population growth for this period was due to immigrants. If current levels of immigration continue—with the latest projections of 880,000 net immigrants a year—another 10 million or so immigrants will be added to the U.S. population by the year 2000.
The Census Bureau's evaluation of data on race and ethnicity from the 1990 census revealed that broad generalizations about the accuracy of the race and ethnicity data are not appropriate. Various statistical measures suggested that the overall quality of the data was good; however, the evaluations also showed persistent problems related to question wording and respondent's understanding of the questions. These problems are of major concern because: (1) they have a substantial impact on the quality of the data for small geographical areas and for such relatively smaller racial and ethnic groups as American Indians; (2) these data are used for drawing political boundaries for very small geographical areas and funding allocations; (3) there is some evidence that a higher proportion of the population had difficulties with answering the race and ethnicity questions in 1990 than in 1980 (Cresce et al., 1992:6); and (4) some studies suggest that the current concepts may not be the most appropriate for a changing population (Hirschman, 1993; Lieberson, 1993).
If a respondent did not provide answers to each question to the census, special procedures were made by the Census Bureau to impute (or allocate) the response. Not all items were allocated, including ancestry. The Hispanic-origin item continued to have the highest allocation rate (10 percent) of any item collected on the short form (see Table K.1). By contrast, the rate was much lower for the race item at 2.0 percent. The allocation rates for the 100 percent items for 1990, particularly for Hispanic origin, were significantly higher than those for