We use data from the 2006 American Community Survey to examine race and ethnic differences in the effects of marital status and co-residence of the middle generation on the likelihood of poverty among grandfathers who have primary responsibility for co-resident grandchildren (N = 3,379). Logistic regression results indicate that race/ethnicity and household composition are significant predictors of poverty for grandfather caregivers: non-Hispanic white grandfathers, those who are married, and those with a co-resident middle generation are the least likely to be poor. The effects of race/ethnicity, marital status, and the presence of a middle generation are, however, contingent upon one another. Specifically, the negative effect of being married is lower among grandfathers who are Hispanic, African American, non-Hispanic, and non-Hispanics of other race/ethnic groups compared to whites. In addition, having a middle generation in the home has a larger negative effect on poverty for race/ethnic minority grandfathers than for non-Hispanic whites. Finally, the combined effects of marriage and a middle generation vary across race/ethnic group and are associated with lower chances of poverty among some groups compared with others. We use the theory of cumulative disadvantage to interpret these findings and suggest that race/ethnicity and household composition are synergistically related to economic resources for grandfather caregivers.