The model views differences in adolescent violence between Whites and other racial-ethnic groups as a function of variation in community contexts, family socioeconomic well-being, and the social capital available to adolescents and families. The model was tested with an analysis of data drawn from three waves of the National Education Longitudinal Survey (NELS) and United States zip-code data on local communities. The NELS is a large, nationally representative data set with student, parent, teacher, and principal components. During the first wave in 1988, the survey drew random samples of about 25 eighth graders in each of about 1,000 middle schools. Students were then traced to high school in 1990 (Wave 2) and 1992 (Wave 3), with high follow-up response rates. There were 16,489 respondents with completed questionnaires in the base year and first and second follow-ups. Independent variables that pertained to racial-ethnic background, gender, family income, and family structure were drawn from the first wave of the NELS (1988). Indicators of family social capital were drawn from the students and parent files of the second and third waves. The dependent variable, fighting, was the indicator of adolescent violence. This indicator was derived from two questions in the third wave of the NELS that asked the respondents how many times they had been in a physical fight at or on the way to or from school over the previous half year. Racial-ethnic groups were distinguished as Whites, Asians, American Indians, Blacks, Hispanics, and respondents of "other" backgrounds. Consistent with the proposed model, the White-Black disparity in adolescents' fighting was explained by higher levels of disadvantage in the communities in which Black children often lived. The effect of concentrated disadvantage on fighting was mediated by more proximate processes that were linked to family well-being. Variation in structural disadvantage was apparently less central in the White-Latino difference in fighting, but it was still consequential. The primary explanation was variation in family well-being between Latino and White families, particularly parental education. American Indians were the only group for whom differences relative to Whites were not fully explained. Theoretical and public policy implications of these findings are discussed. 3 tables, 114 references, and appended supplementary information and data