The feasibility of employing survey data that uses the states of the United States as the societal units in macro-level sociological research was studied. The range of issues available for research opened to investigation would be greatly expanded if it were possible to aggregate national survey data to produce state-level statistics for variables such as approval of violence and gender role attitudes. If such data were to be available, they could be used to investigate issues such as whether state-to-state differences in the degree of approval of violence explain part of the huge differences among states in the rate of violent crimes. The validity analyses reported in this paper were conducted using data from the 1975 National Family Violence Survey, the 1985 National Family Violence Resurvey, and the 1972-84 cumulative General Social Survey. Analyses provide information on concurrent validity, construct validity, validity of specific variables, and multi-indicator indexes. It is concluded that aggregate survey data should be avoided unless there are strong reasons to use such data despite the problems. (TJH)