UC Riverside

The objective of this dissertation is to evaluate the quality of the decennial Census through the use of Demographic Analysis (DA), which is a methodology that estimates population size by using data on births, deaths, in-migration, and out-migration. Using DA, the quality of the 2000 Census will be assessed by using estimates of Census undercount for children aged 0 - 9, as well as measures of relative undercount for the 1990 and 2000 Censuses for Hispanics of older ages. In cleaning the data that measure the components of the Hispanic population, the chapters of this dissertation address topics on the sociology of the Hispanic population.

The first chapter critiques the measurement of the Hispanic population in general, arguing that the Census Bureau's use of Hispanic identifiers is politically charged and ambiguous. Logistic regression is used to demonstrate that the personal identification with the Hispanic ethnicity is affected by dynamics of assimilation, race, and social context. The second chapter argues that the quality of the 1990 Census was compromised by the political atmosphere in which the Census was embedded, while producing alternate estimates of emigration using algebraic models informed by recent research. The third chapter re-visits the Hispanic Mortality Paradox by accounting for its counter-explanations (i.e. death undercount and emigration). Using life tables, it will be shown that the existence of the Mortality Paradox largely depends on the assumptions that one makes regarding the quality of the data used, although it continues to hold when the most plausible data are used. Chapter four demonstrates that interethnic childbirth is largely influenced by a mother's race and ethnicity, although social factors such as education and the marriage market serve to "break down" ethnic divisions and stir the melting pot. Chapter five tests the success rates of a number of imputation methodologies for parental ethnicities and predicts the number of Hispanic births that have occurred throughout the 90s. Putting all these pieces together, the conclusion presents a range of estimates of undercount of Hispanic children aged 0 - 9, as well as relative undercount that compares the counts of the 1990 and 2000 Censuses.