This report examined the participation of at-risk children in preschool programs, after controlling for selected individual, family, and geographic characteristics. A multivariate statistical technique, a logistic regression, was applied to data from the 1990 Decennial Census. This allowed an analysis of the relative effects of each of the following variables on preschool participation: income, education status of most educated parent, race, immigrant status, linguistic-isolation status, employment status of parent, family type, the urbanicity of residence, and state of residence. With Head Start funds reaching less than half of the eligible 3- and 4-year-olds, children living in low-income families were 16 to 20 percent less likely to attend preschool than their middle-income counterparts, even after controlling for ethnicity, family type, immigrant status, parent education, and other family characteristics. By contrast, children in higher-income families were 25 to 50 percent more likely to go to preschool than children from middle-income families. The education level of children's parents has a large influence on participation; the higher the level of education of the parent(s), the more likely that the child would participate in preschool. The report also found that Black and Native American children were more likely than white children to attend preschool, after controlling for individual, family, and geographic characteristics. Three appendices contain the objectives, scope, and methodology of the study, data points for report figures, and General Accounting Office (GAO) contacts and staff acknowledgements. (MDM)