averages and include attempts that result in minimal consequences as well as extremely violent crimes that result in long hospital stays and devastate victims' quality of life for years to come. Thus, the fact that the average rape costs about $500 in medical expenses reflects the fact that only about 25 percent of all rape victims in the NCVS reported receiving any medical treatment; only about 2 percent of rape victims stayed overnight in a hospital. A brief explanation of what these categories are and how they were estimated follows.

Tangible losses

Property damage and loss: the value of property damaged and of property taken and not recovered, plus insurance claims administration costs that arise in compensating victims' property losses.

Most of these losses are taken directly from NCVS data. In cases where losses were insured, 16 percent was added to account for insurance claims processing costs, a published loss-adjustment expense ratio from Best (1993). Supplementary published sources were also used for drunk driving (Miller and Blincoe, 1994) and arson cases (Hall, 1993).

Property damage is typically a very small percentage of the cost of violent crime, often less than 1 percent of the total tangible costs. However, costs related to property damage are typically the bulk (60-80 percent) of the cost of property! household crimes such as burglary, larceny, and arson. The average arson victim incurs $15,500 in property damage; the average motor vehicle theft costs $3,300; while the average property value loss to robbery victims is $750.

Medical care: payments for hospital and physician care, as well as emergency medical transport, rehabilitation, prescriptions, allied health services, medical devices, coroner costs, premature funeral expenses, and related insurance claims processing costs. Also included in this category are victim legal expenses incurred in recovering medical costs from drunk drivers and their insurers.

Although the NCVS contains estimates of medical care costs, these estimates are self-reported by victims and only include medical costs up to a few months after the incident. This approach underestimates medical costs because the victim may not receive bills for the full cost of medical care, and it does not consider long-term costs. Although NCVS provides some details on the nature of injury, the categories are quite broad, e.g., broken bones or gunshot wounds and hospitalized versus nonhospitalized status.

In order to estimate lifetime medical costs, this study started with the distribution of injuries as reported in NCVS and made several adjustments. For hospitalized injuries, within the broad NCVS injury categories, an estimate was developed of the "ICD9" (International Classifications of Diseases, Ninth Edition, Clinical Modification Code) injury distribution by examining all hospitalized injuries in California, Vermont, and Washington, three States that have identified injuries by ICD9 code and cause (allowing us to isolate violence victims from unintentional victims and suicide attempts). Using the actual lengths of hospital stay from the State data and the medical payments per day (and other injury information) by ICD9 code from Miller, Pindus et al. (1995), the treatment costs were estimated for each hospitalized assault or rape victim. The average cost per injury by NCVS injury category was then computed. Although having a nationally representative sample of hospitalized intentional injuries would have been preferred, using all hospitalized rape and assault cases in only States that collected these data seemed reasonable. Since the research started with NCVS broad injury categories, the research team only used the three-State injury data to estimate the body parts and severity of injury within each category. It is unlikely that the typical "broken bone" in these three States is significantly more or less severe than the typical broken bone elsewhere in the United States. Data becoming available in 1994-1996 from a dozen other States will allow verification of this hypothesis.

For nonhospitalized injuries a similar approach was tried, but the estimated short-term medical costs by injury category were so similar to the NCVS estimates that the NCVS estimates of short-term costs were used. These estimates were multiplied by lifetime-to-short-term-cost ratios for nonhospitalized injuries by ICD9 code groupings (in the NCVS categories) computed from the ratios in Miller, Pindus et al. (1995) and the injury distribution from 21 hospital emergency departments that code ICD9 nature of injury and cause of injury for nonadmitted injuries.

The researchers know of no estimates of the medical costs associated with homicide victims. Instead, medical costs for fatal injuries were taken from Miller, Pindus et al. (1995), based on all fatal workers' compensation cases in 41 States. Similarly, virtually no estimates of medical costs are available for child abuse. Data on medical costs per child abuse case were obtained directly from the above-mentioned health care data and an inferred hospitalization rate. Thus,