The section that follows systematically describes the findings about crimes for which this study's incidence estimates go beyond those of the NCVS.

Fatal crimes. Including victims of drunk drivers, crime took 31,079 lives in 1990. This count largely derives from Vital Statistics data, which recorded 4,500 crime deaths, primarily negligent manslaughter cases, excluded from the Federal Bureau of Investigation's Uniform Crime Reports. The number excluded deaths due to legal intervention, as well as 2,376 deaths of unknown intent, many of them poisonings and firearm deaths. Some of the latter deaths probably were homicides. For example, child abuse deaths are likely to be undercounted because some children who allegedly die of Sudden Infant Death Syndrome (SIDS) are actually abuse victims. Vital Statistics data also exclude victims whose bodies are never recovered. The drunk driving count excludes intoxicated drivers and nonoccupants. As described below, the number further reduces the victim count in an attempt to remove crash deaths that would have happened even if everyone involved had been sober.

Child abuse. Existing estimates of child abuse vary considerably. The two main sources of national information on child abuse are the National Incidence and Prevalence Survey of Child Abuse and Neglect (NIS) (Sedlak, 1991) and the National Family Violence Surveys (Gelles and Straus, 1985; Straus and Gelles, 1976). The latter produce estimates far greater than the former, mainly due to differences in the samples employed. Gelles and Straus used a probability sample of adults with children, and Sedlak relied on a sample of child protective services (CPS) and other government agencies to generate counts of actual cases that came to their attention. The CPS-based estimate undoubtedly represented only serious cases and is thus an underestimate of the total number of child abuse victims. The Gelles and Straus estimates were based on self-reported behaviors that are difficult to validate with external measures.

This study's researchers base estimates on data from the 1986 NIS but use a statistical technique, capture-recapture modeling, that estimates the number of people unknown to the system from the number detected by various sources (Miller, Kilpatrick, and Resnick, 1994). The estimates are conservative. They largely exclude the 1.3 million cases annually that child protective services agencies are unable to substantiate or do not classify as abuse and neglect using NIS definitions.

The unduplicated count of 1990 abuse victims is at least 794,000. In order of precedence for victims of multiple types of abuse, 185,000 children were sexually abused, 308,000 were physically abused, and 301,000 were emotionally abused. Unlike the victim count, the victimization count treats a child who was the victim of both physical and sexual abuse as being victimized twice.

The estimate is thus higher than Sedlak's estimate of 590,800 total abuse cases (of which 133,600 were sexual abuse victims) but lower than Straus and Gelles' estimates. Straus and Gelles (1986, 1987) estimated about 1.3 million children are physically abused annually, including children targeted with thrown objects that missed their mark.

This project's estimate equates to about 13-17 million adults who were physically abused by parents or adult caretakers as children and 355,000 new physical abuse cases annually. The Commonwealth Fund (1993) estimated that 20 million adults were "physically abused" by someone as children. The Family Violence Prevention Fund (EDK Associates, 1993) estimate is that 15 million adults were "beaten by their parents or witnessed a sibling's beating." This estimate probably is low since people who had never witnessed a parent beating a child were not asked if they personally had been beaten.

More recently, Finkelhor and Dziuba-Leatherman (1994) used a telephone survey of youths to estimate the criminal victimizations of children ages 10-16 during 1991. They estimated that 0.5 percent of children ages 10-16 were victimized by a parent or parent-substitute. Based on Sedlak's (1991) estimate that 89 percent of child physical abuse was by a parent or substitute, and an estimated 24.3 million children ages 10-16, the Finkelhor and Dziuba-Leatherman estimate translates into 246,000 child abuse victims (with 95 percent confidence limits of 137,000 to 355,000). By comparison, the present study estimates 170,000 child physical abuse victims ages 10-16 (out of 308,000 child abuse victims).

Rape.Until 1992, the NCVS did not ask directly about rape, but left it to the victim to mention (and define), a method that resulted in criticism for undercounting rape. After an extensive redesign process, a new NCVS questionnaire was fielded in 1992 that, among other things, directly asks about rape. Preliminary data released in the fall of 1994 suggest