that estimates of completed and attempted rape are about 2.7 times higher than those based on the older questionnaire. In addition, the new form developed estimates of nonrape sexual assault for the first time. Together, 607,000 rapes and other sexual assaults were estimated to have occurred in 1992 (BJS, October 1994). This amount excludes about 114,000 rape or sexual assault victims who were subject to series rapes in excess of six or more times and an additional 113,000 victims of "unwanted sexual contact without force." If these estimates are added together, there were approximately 834,000 victims in 1992.

Because the redesigned NCVS instrument had not yet been implemented at the time of this research (i.e., as of May 1995, the 1992 data files had not yet been released), alternative national estimates of rape were used. The best data the researchers could find that coincided with the time period under study were produced from the National Women's Study (NWS) developed by Kilpatrick et al. (1992). The NWS surveyed women age 18 years and older in 1990 and found approximately 683,000 women had been raped or sexually assaulted in the previous year. Followups of the same sample (providing 1-year bounded reference periods) in 1991 and 1992 produced annual estimates averaging 632,000, an amount that closely matched the original findings (Kilpatrick et al., 1994). Kilpatrick's questions about rape were clear and direct. They explicitly included vaginal intercourse, oral and anal sex, and/or other sexual penetration by force, threat of force, or lack of consent. The Kilpatrick numbers are similar to the preliminary estimates from the 1992 NCVS, even though the questions were worded differently. Using the Kilpatrick estimates for 1990 is thus not an unreasonable approach to avoiding the undercount in the pre-1992 NCVS.

Several adjustments to the Kilpatrick estimates are necessary for comparability with the rest of this study's incidence data. This research started with the 1991-92 Kilpatrick estimate of the number of adult female rape victims (632,000). Because Kilpatrick surveyed only women, the ratio of female to male rapes in the NCVS (17 percent of victims are male) was used to inflate the estimate. Next, because persons under 18 were not included in the NWS, an adjustment for child rapes was made. The NWS asked women if they were raped as children. From these responses, the number of children that would have been raped per year was estimated in order to report such a response as adults. Although the NWS did distinguish attempts from completed rapes, it is not clear whether respondents excluded attempts; it is also possible that respondents to the pre-1992 NCVS considered oral sex and other sexual assaults as attempted assaults, events that Kilpatrick defines as rape. To be conservative, this study did not, therefore, increase the NWS completed rape estimate to account for attempts. Finally, because the Kilpatrick estimate is a prevalence estimate (a count of victims) and not an incidence estimate (a count of victimizations), the NWS figure was multiplied by 1.27, the ratio of victimizations to victims from NCVS. The result is an estimated 1.1 million rape and sexual assault victims--very close to the recently released redesigned NCVS preliminary estimates of 834,000 victims noted above.

AssaultThere is no single source of data on assaults. Based on NCVS data, an estimated 7.8 million assaults and attempted assaults occur annually, representing about 4.4 million victims. About 30 percent of these assaults involve injury. The NCVS excludes children under age 12 and undercounts domestic assaults.

To supplement the data, this study estimates that nonfatal assaults against children under age 12 number 139,000 annually (97,000 victims). To avoid double counting, this estimate excludes physical child abuse, defined as assault by adult caretakers or by others who parents let assault their children (Sedlak, 1991). This estimate was computed by multiplying the NCVS count of nondomestic assaults on children ages 12-17 times the ratio of medically treated nondomestic assaults for children ages 0-11 versus children ages 12-17. That ratio came from proprietary health care system data on 13,528 hospitalized crime victims and 65,555 crime victims treated in emergency rooms and released. (By comparison, the unweighted 1987-91 NCVS data included fewer than 250 overnight hospital admissions and 1,000 other victims treated in hospital emergency or outpatient departments.) The hospitalized data cover all victims admitted in California, Vermont, and Washington at times when those States required coding of injury causes. The nonhospitalized data are from a convenience sample of 21 hospital emergency departments spread across 9 cities. They include all but one of the emergency departments this study could identify nationally that coded injury cause and type and maintained data on extractable computer files. However, since these emergency departments are not nationally representative samples, the assault count for children under age 12 should be viewed as preliminary and is reported separately.