disaggregated by the type of crime. These estimates are $376 for rape, $124 for robbery, $153 for assault, and $5,370 for murder. Nonfatal crime incidence and medical costs derive from the 1987 NCS, whereas the number of murders is taken from the Federal Bureau of Investigations (FBI) Uniform Crime Report (UCR).
NCS data are dependent on the self-report of individuals in sampled households, queried as to the extent of victimization experienced in the six months preceding the survey.14 Individuals are asked to recall if they were victimized within the defined time frame. Respondents report all injuries incurred due to victimization, given a listing of possibilities that includes rape; attempted rape; knife or stab wounds; gunshot or bullet wounds; broken bones or teeth knocked out; internal injuries; being knocked unconscious; bruises, black eye, cuts, scratches, swelling, and chipped teeth; and other (U.S. Department of Justice, 1989a). Victims reporting injury are asked if medical care, including self-treatment, was received and the location of such treatment (e.g., at the scene, in a hospital emergency room). Those individuals are further queried as to inpatient hospital duration, if any, and total medical costs, including doctor and hospital bills, medicine, therapy, braces, and other related expenses, regardless of whether such costs were covered by insurance.
The NCS data have several limitations. In addition to sampling and nonsampling errors, the NCS does not include victimization of organizations and commercial establishments, children under age 12, transients, the homeless, individuals in institutional settings such as nursing homes, and military personnel in the sampling frame.15 The medical costs estimated here do not include any victim injury costs borne by nonsampled populations.
A second limitation of the NCS is the implicit assumption that victims accurately recollect and report incidents that took place within the prescribed time frame (i.e., respondents report all incidents relevant to the time frame and, conversely, do not "telescope in" injuries from earlier time frames). Further, NCS is known to underreport both the incidence and the cost of certain types of crimes, such as domestic violence or assaultive behavior between nonstrangers.16 Efforts to strengthen NCS data in these areas were beyond the scope of the current analysis.
A closely related issue concerns individuals' abilities to accurately report the degree of injury and the extent of full medical costs. Self-reported injury and medical costs are suspect for several reasons. Victim selections of injury categories from the NCS listing may not accurately reflect actual medical diagnoses. Thus,