suggest that violence against children accounts for a significant portion of our nationwide victim costs. Out-of-pocket costs for child victims are estimated to be more than 20 percent of all out-of-pocket crime victim costs and more than 35 percent of all costs (including pain, suffering, and lost quality of life).

The National Crime Victimization Survey (NCVS) is the government's main source of information about criminal victimization. To collect information for the NCVS, the Bureau of Justice Statistics polls people over age 12 about rape, robbery, assault, larceny, burglary, and motor vehicle theft. Partly because of the difficulty in obtaining certain information in a survey format, the NCVS does not collect data on certain crime categories (such as child abuse and drug abuse) and undercounts others (such as rape and domestic violence). The NCVS format does not attempt to comprehensively measure crime costs and consequences or document and assess crime-induced permanent disability and mental health treatment, which would cover such intangibles as the pain, suffering, fear, and lost quality of life that victimization brings. The survey documents numbers of crimes as reported by households and asks victims to quantify their short-term out-of-pocket losses due to victimization. This Research Report aims to add valuable information to the findings of the NCVS by estimating the full cost of crime to victims and by including many other types of crime.

The incidents of crime

Number of victimizations. It is difficult to know how much crime is committed in the United States. The Federal Bureau of Investigation's Uniform Crime Reports (UCR) collect data from police departments about the number of crimes reported or known to the police. The NCVS survey of households also obtains information about crimes not known to the police and thus identifies more criminal offenses than the UCR, but it still probably undercounts the total crimes, especially those regarding gunshot and knife assaults, domestic violence, and rape (National Research Council, 1993). As mentioned previously, the NCVS also excludes many crimes, among them murder, arson, drunk driving, child abuse and neglect, and crimes against children under 12. Also, the survey sample of U.S. households largely omits the homeless and others not attached to traditional households.

This project reviewed available estimates of the under-reported crimes, seeking more comprehensive incidence data. The data sources included national surveys and local studies. One difficulty with analyzing these studies was the variety of definitions used for the same crime. For example, some researchers defined rape as any unwanted sexual encounter, whether or not the victim believed she had been raped. Others were more restrictive, including only forced penetration that the victim characterized as rape. Another problem is that many studies have not used nationally representative samples; also, rape surveys have often polled only women. Ultimately, the researchers for this project focused almost entirely on data from nationally representative surveys where the crime definition was clearly reported. In some cases, they generated new estimates of crime incidence based on an analysis of earlier studies.

Total crimes. People and households in the United States faced more than 49 million crime attempts annually in 1987-1990. The annual toll includes more than 16 million violent crimes and attempted crimes (murder, rape, robbery, assault, child abuse, drunk driving, and arson). In 1990, 31,000 deaths resulted.

Table 1 distinguishes between victims and victimizations. The victimization counts in the last column of table 1 differ from most published NCVS counts. The reason is that the researchers included series victimizations, a chain of three or more similar crimes that respondents did not describe in detail. The NCVS asks series victims to estimate how many times they were victimized. Recent NCVS validation efforts suggest respondents can remember details of up to six related victimizations if pressed. This study accepted the serial victimization counts to a maximum of 10 per series but examined the effect of more and less aggressive choices in sensitivity analysis. Thus, if a man beat his wife every Saturday night, for example, table 1 records 1 victim and 10 victimizations annually (labeled "series=r"). Table 1 also shows a more conservative victimization count that treats a series of crimes which the victim is unwilling to describe individually as a single victimization (labeled "series=1 "). Even this count exceeds NCVS victimization counts, which completely exclude the series victimizations.

The victim counts in table 1 generally are not additive between crime types. The same person may be the victim of a rape and a burglary during one year, so the sum would double count. Duplication was eliminated within the NCVS data, however. The unduplicated annual victim count for NCVS crimes alone approaches 31 million.