Victimization surveys are useful as a complement to the Uniform Crime Reports (UCR), which remains the basic statistical body of crime data. While the UCR counts all events, its data range is narrow. In contrast, victimization surveys count only a representative sample of criminal events but assemble a wide scope of information concerning each incident. Such surveys are not replacements for UCR; they are expensive and cannot be conducted in all cities and in time series. They are limited by sampling errors and difficult to use for small-area estimates because of the relative statistical rarity of certain crime forms (e.g., serious crimes). The first national victim survey was conducted in 1966. In the 1970's, LEAA and the Census Bureau jointly undertook a national survey of 10,000 households and businesses (22,000 individuals) through personal interviews repeated at 6-month intervals and a survey of 26 cities where respondents were interviewed only once. Interview schedules queried household characteristics, individual characteristics associated with victimization, details of criminal events, and the victim-offender relationship. A reverse record check was done to verify respondents' lapses of recall, and quality control was maintained through reinterviews with a random sample of respondents. These surveys revealed that almost half of all robberies and a quarter of all larcenies were not reported to the police. Household nonreporting was twice as high as that of businesses. Furthermore, different types of crime appeared to have different nonreporting patterns, frequently associated with whether or not the criminal event was completed. Victimization data are generalizable to unsurveyed locales because study findings have been consistent nationally and across all Standard Metropolitan Statistical Areas. These surveys provide systematic and detailed information which is available nowhere else.