You are viewing a Web site, archived on 09:38:48 Oct 16, 2004. It is now a Federal record managed by the National Archives and Records Administration.
External links, forms, and search boxes may not function within this collection.
Public Health Seal report title shim
contents search order press resources links home curve end shim

Chapter 2:
The Magnitude of Youth Violence

Measuring Youth Violence

The Violence Epidemic

Arrests for Violent Crimes

Prevalence of Violent Behavior

Differences by Sex and Race/Ethnicity

Violence at School

Gangs and Violence



Appendix 2-A

Chapter 2


Prevalence refers to the proportion of American youths involved in one or more violent behaviors. UCR arrest rates, as discussed earlier, cannot be used to calculate prevalence. The only national youth survey from which long-term trends in self-reported violent behavior can be gleaned is the MTF,6 which was begun in 1975 and is conducted annually by the University of Michigan's Institute for Social Research. The longest-running survey of youths, MTF asks a nationally representative sample of high school seniors about a wide range of social attitudes and behaviors.7 Although the survey is administered at school, it asks about violent behavior and victimization across all community settings.

It is worth reiterating that self-reports, whether by offenders or victims, are an essential research tool for determining the extent of youth violence. They furnish a window into violent behavior that never reaches the police. For example, the National Crime Victimization Survey reveals that the majority (58 percent) of serious violent crimes committed by youths are not reported to the police (Snyder & Sickmund, 1999). A large fraction of the crimes that are reported never result in an arrest. Estimates indicate that only 6 to 14 percent of chronic violent offenders are ever arrested for a serious violent crime (Dunford & Elliott, 1984; Elliott, 2000a; Huizinga et al., 1996; Loeber et al., 1998).

The MTF gathers data about five acts of violence and from them compiles a violence index (see Figure 2-8 for the specific offenses included). This violence index is not the same as the UCR violent crime index, which aggregates the four types of arrests covered in this chapter. According to the MTF's violence index, about 3 out of 10 high school seniors reported having committed a violent act in the past year, an annual prevalence rate of about 30 percent. The MTF's violence index has been relatively stable for almost 20 years, in sharp contrast to the dramatic increase in arrests.

Figure 2-8. Trends in prevalence of serious violence among 12th graders, 1980-1998

Figure 2-8. Trends in prevalence of serious violence among 12th graders, 1980-1998
Figure 2-8 (click to enlarge)

Although the prevalence rate of self-reported violent behavior is relatively constant, it is still strikingly high, partly because high school seniors age 17 and 18 are at the peak ages of violent offending and partly because the violence index includes some less serious violent behaviors as well as some very serious ones.

Because this report focuses on violent behavior carrying the potential for serious injury or death, Figure 2-8 also includes the prevalence rates of assault with injury and robbery with a weapon, the two most serious acts in the MTF violence index. An assault with injury could lead to an arrest for aggravated assault; likewise, a robbery with a weapon could lead to an arrest for armed robbery. Therefore, assault with injury and robbery with a weapon may be used as proxy measures for aggravated assault and armed robbery, respectively.

Over the past two decades, the MTF's prevalence rates for assault with injury ranged from 10 to 15 percent ( 1.3 to 1.8). A small but significant increase took place between 1979 and 1998. About half of this increase occurred between 1983 and 1993, but rates remained fairly constant after 1993 (the increase from 1993 to 1998 shown in Figure 2-8 is not statistically significant). The prevalence of robbery with a weapon ranged from 2 to 5 percent ( 0.7 to 1.1) between 1983 and 1993 and remained constant thereafter. Thus, unlike arrest data, MTF data show no evidence of a downward trend in self-reported assaults or robberies after 1993.

Prevalence rates of this magnitude—10 to 15 percent of high school seniors8—for the most serious types of violence are confirmed by other self-report surveys described in Chapter 3. For example, an average prevalence rate of 9 percent ( 2.0) was reported for 17-year-olds between 1976 and 1982 in the National Youth Survey, whose measure of violence includes aggravated assault, robbery, gang fights, and rape. This rate is similar to the MTF's, but the National Youth Survey measure includes more serious violent offenses. Two general city surveys—the Denver Youth Survey and the Rochester Youth Development Survey, which use the same measure of violence as the National Youth Survey—report somewhat higher prevalence rates among 17-year-olds: 12 percent ( 1.6) and 14 percent ( 2.0), respectively.9

International Prevalence
Are U.S. youths unique in reporting a high prevalence of violent behavior? How do they compare to their European counterparts? The answers can be found by comparing the MTF findings with the International Self-Report Delinquency Study (Junger-Tas et al., 1994), a study of delinquent behavior conducted in several European countries.

Like the MTF, this study relies on self-reported behavior. Of the countries included, only England/Wales, the Netherlands, Spain, and Italy used a probability sample that provided national estimates of violence comparable to the violence index used in the MTF survey. Self-reported serious violence among young people age 16 to 17 in these countries in 1992 or 1993 ranged from 16 to 26 percent (Table 2-1). These prevalence rates are lower than the U.S. rate of about 30 percent for the MTF's violence index. Thus, while the questions in the international study may be somewhat different, the findings show that while a higher proportion of U.S. youths commit violent acts, youth violence is not unique to the United States.

Table 2-1. International comparison of the annual and cumulative prevalence of self-reported violent behaviora by youths, 1992-1993b

Table 2-1. International comparison of the annual and cumulative prevalence of self-reported violent behaviora by youths, 1992-1993b
Table 2-1 (click to enlarge)

A major difference between the United States and several other industrial countries is the ease of access to firearms. From 1990 to 1995, the United States had the highest rate of firearm-related deaths among youths in the industrialized world (CDC, 1997). The rate for children below age 15 was five times higher than that of 25 other countries combined.

In summary, youth violence, although international in scope, is greater in the United States, more likely to involve firearms, and more lethal in its consequences. According to self-reports, both the prevalence and incidence (volume) of assault and robbery increased among U.S. high school seniors between 1983 and 1993. This finding is consistent with an epidemic of violence among U.S. youths, although self-reports point to a more modest upsurge than arrest trends do. However, both self-reports and arrest rates for aggravated assault point to an ongoing problem of youth violence after the apparent end of the violence epidemic. Thus, the rise and fall in arrest rates for most violent offenses is set against more enduring rates of violent behavior.


Self-reported violence and arrest rates for violent offenses can also be compared by sex and by race/ethnicity. Ratios based on these two sources of data show similar findings with respect to sex but remarkably different findings with respect to race/ethnicity—differences that have yet to be fully explained.

Differences in Self-Reports
Self-reported rates of serious violent behavior differ widely by sex but considerably less by race. Table 2-2 compares the violent incident rate (the number of robberies and assaults per 1,000 high school seniors) and the violence index prevalence rate (the prevalence of the five serious acts of violence described in Figure 2-8) by sex and by race. The table focuses on two critical periods, 1983 to 1993 and 1993 to 1998. In general, there was little change in those periods, with one exception.

Table 2-2. Differences in youths' self-reported serious violent behavior, by sex and race, 1983, 1993, and 1998

Table 2-2. Differences in youths' self-reported serious violent behavior, by sex and race, 1983, 1993, and 1998
Table 2-2 (click to enlarge)

In 1983 and 1993, the ratios of male to female youths committing violent acts were 7.4 to 1 and 7.0 to 1, respectively. This means that for every violent act committed by female youths in these years, at least seven violent acts were committed by male youths. By 1998, this ratio had closed to 3.5 to 1, indicating that females are closing the gap. The difference in prevalence rates changed little over the same period, but at a ratio of 2 to 1, it was much smaller to begin with. Taken together, the trends show that the proportions of males and females involved in violence (the prevalence rate) have not changed but that the relative number of violent acts by males and females (the incident rate) has changed, with females committing more violent acts in 1998 than in earlier years.

Differences by race are also presented in Table 2-2. The only available national comparisons for serious violence are for white and African American youths (see Chapter 3 for local longitudinal studies that include rates for Hispanic youths). Overall, incident rates are lower for white than African American youths over these years; the gaps are largest in 1993 and 1998, when approximately 1.5 violent acts were committed by African Americans for every 1 violent act by whites. The racial gap appeared to increase somewhat during the violence epidemic and has remained higher through 1998. There are essentially no differences by race in the prevalence rates for serious self-reported violent behavior.

Differences in Arrest Rates
Arrest rates differed widely by sex and by race/ethnicity between 1983 and 1998 (Table 2-3). Overall, the difference was greater by sex than by race/ethnicity and was most evident in regard to homicide arrests: In 1998, 11 times as many males were arrested as females. A similar male-female gap was evident for robbery, but the gap for aggravated assault was considerably smaller.

Table 2-3. Differences in youth arrests for serious violent crimes, by sex and race/ethnicity, 1983, 1993, and 1998

Table 2-3. Differences in youth arrests for serious violent crimes, by sex and race/ethnicity, 1983, 1993, and 1998
Table 2-3 (click to enlarge)

Trends in the male-female gap vary, depending on the crime for which youths are arrested. From 1983 to 1993, the male-female disparity in homicide arrests doubled: In other words, the violence epidemic was driven by arrests of males. During the same period, the male-female gap in arrests for both robbery and aggravated assault shrank. More recently, from 1993 to 1998, the male-female disparity in all three types of arrests has held constant or declined further.

Differences in arrest rates by sex are similar in magnitude to differences in self-reported violent incidents. Combining aggravated assault and robbery arrest data yields male:female ratios of 6.8 to 1, 5.7 to 1, and 4.3 to 1 for 1983, 1993, and 1998, respectively.10 The ratios for self-reported incidents were 7.4 to 1, 7.0 to 1, and 3.5 to 1 (Table 2-2). Thus, both self-report and arrest rates attest to a difference by sex in the volume of violence but also to a narrowing of that gap between 1983 and 1998—except for homicide arrests. Possible reasons for the male-female gap are discussed in Chapter 4.

Self-reports and arrest rates provide different pictures of violent offending by race. Self-reports, as noted above, reveal small differences between African American and white youths. Arrest records, on the other hand, reveal large differences, even though these gaps narrowed between 1993 and 1998 (Table 2-3). The narrowing of the gap was particularly noteworthy for homicide arrests: Whereas about nine African American youths were arrested for every white youth in 1993, only about five were arrested for each white youth in 1998. Even at 5 to 1, the ratio of African American to white youths arrested for homicide remains greater than that of Native American or Asian youths to white youths.

Ratios cannot be calculated for Hispanic youths because data for this ethnic group are not broken out in the UCR or other systematic data collection systems (Soriano, 1998). A few regional and city studies suggest that homicide arrest rates for Hispanic males are substantially higher than those for non-Hispanic white males and that African American males typically have the highest rates (Prothrow-Stith & Weissman, 1991; Smith et al., 1988; Sommers & Baskin, 1992; Zahn, 1988). The difference between homicide arrests of Hispanic and non-Hispanic white youths is substantial in these studies, but it is not as great as the difference between African American and white youths.

The existence of much larger racial and ethnic differences in arrest rates than in self-reported violence is a matter of great concern. On the one hand, there is no reason to expect similar distributions, because these measures were designed to assess different aspects of violence. But if both measures are valid and reliable, the discrepancy suggests that the probability of being arrested for a violent offense varies with race/ethnicity. Explanations for this discrepancy focus on selective reporting of offenses to the police, different patterns of police surveillance, racial/ethnic biases in self-report measures, and racial/ethnic bias on the part of police, victims, and witnesses. Some studies have explored these explanations, but their findings are not definitive (Austin & Allen, 2000; Blumstein et al., 1986; Hawkins et al., 1998; Sampson & Lauritsen, 1997). This complex issue will also be discussed in Chapter 3, which considers other dimensions of violent offending.

Arrest ratios of Native American11 to white youths are similar, except for the homicide ratio in 1998. Similarly, arrest rates of Asian Americans for homicide and robbery differ little from those of whites, but at least two whites are arrested on charges of rape or aggravated assault for every Asian American. Possible reasons for these differences have not been well studied.

In sum, racial and ethnic differences in rates of violence are greater in arrest statistics than in self-reports of violent behavior. The reasons are not well understood, with conflicting evidence from various studies. Self-reports and arrest records produce similar estimates of trends in violence by sex: Violent behavior still occurs more often among male than female youths, but the gap has been narrowing.


Recent shootings at schools have galvanized public concern about school safety, but studies described here find that schools nationwide are relatively safe. In contrast to public perceptions, schools have fewer homicides and nonfatal injuries than homes and neighborhoods. However, some students are at greater risk of being killed or injured at school than others—specifically, senior high school students from racial or ethnic minorities who attend schools in urban districts (Kachur et al., 1996).

Homicides and Nonfatal Injuries
Two nationwide studies of school homicides have been conducted by the Centers for Disease Control and Prevention in collaboration with the U.S. Departments of Education and Justice. The first study covered a 2-year period from July 1992 through June 1994 and identified 68 students who were killed on or near school grounds or at school-related events (Kachur et al., 1996). Most of the victims were male and were killed with a firearm. These homicides represent less than 1 percent of all youth homicides in the period studied, and the estimated incidence of school-associated violent death was 0.09 per 100,000 student-years.12 Those at greatest risk of being killed were from racial or ethnic minorities, from senior high schools, and from urban school districts. The homicide rate in urban schools, for example, was nine times greater than the rate in rural schools. Most offenders and victims alike were male, under age 20, and from a racial or ethnic minority. The most common motives were an interpersonal dispute or gang-related activities.

The second study, using the same methodology, updated the figures through June 1999 (CDC, 2000a). It identified 177 students age 5 to 19 who were killed in this 5-year period; the vast majority of the homicides (84 percent) involved firearms. School-associated homicides remained at less than 1 percent of all homicides among students, but the frequency of homicides involving more than one victim increased. The three school years from August 1995 through June 1998 saw an average of five multiple-victim homicides or homicide-suicides per year. An average of one such event occurred in each of the 3 years from August 1992 through July 1995.

Thus, trends throughout the 1990s show that the number of school homicides has been declining. Yet within this overall trend, homicides involving more than one victim appear to have been increasing.

In regard to nonfatal injuries at school, the National Crime Victimization Survey found in 1998 that the rate of serious violent crimes against youths age 12 to 18 was one-half as great when they were at school as when they were not. At school, the highest victimization rates were among male students and younger students (age 12 to 14) (Kaufman et al., 2000). The rate was highest in urban schools in 1992, but by 1998 the rates at urban, suburban, and rural schools were similar. Overall, between 1992 and 1998, the rate of serious violent crimes at school remained relatively stable at about 8 to 13 per 1,000 students (Kaufman et al., 2000).

The stability of this trend is corroborated by the MTF survey, which asks high school seniors whether they have been victims of violence. The percentage of seniors reporting that they had been injured with a weapon at school remained stable at about 5 percent from 1976 to 1998 (Flanagan & Maguire, 1992; Maguire & Pastore, 1999) (Figure 2-9). The same victimization rate is reported by the National Study of Delinquency Prevention in Schools for 1998 (Gottfredson et al., 2000). However, the MTF trend masks large fluctuations in victimization reported by African American students (Figure 2-9). From 1980 to 1998, between 4 ( 2.8) and 13 ( 3.6) percent of African American students reported having been injured with a weapon at school.

Figure 2-9. Twelfth graders injured with a weapon at school, 1980-1998

Figure 2-9. Twelfth graders injured with a weapon at school, 1980-1998
Figure 2-9 (click to enlarge)

Weapons at School
Recent findings regarding students carrying weapons (a gun, knife, or club, for example) at school are encouraging. In 1999, the Youth Risk Behavior Survey (YRBS) found that about 7 percent of all high school students reported carrying a weapon on school property within the last 30 days (Kann et al., 2000) (Figure 2-5). In 1993, almost 12 percent of high school students reported carrying a weapon at school in the last 30 days (Kann et al., 1995), a 42 percent decrease (Brener et al., 1999; Kann et al., 2000). A somewhat less pronounced decline was apparent among high school seniors in the MTF survey (Kaufman et al., 1998). Both studies found the problem to be of roughly the same magnitude: In 1995, about 6 to 8 percent of 12th graders reported carrying a weapon at school at least once during the past month.

Evidence of an upsurge in the number of students carrying weapons at school before 1993 is less clear. The YRBS first asked this question in 1993, and the MTF did not ask until the 1990s. Nonetheless, smaller or less representative studies suggest a substantial increase in weapon carrying between the 1980s and the early 1990s (reviewed in Elliott et al., 1998).

Perceptions of School Violence
Although the overall risk of violence and injury at school has not changed substantially over the past 20 years, both students and their parents report being increasingly apprehensive about their schools. Studies reveal that, during the early 1990s, students grew more fearful about being attacked or harmed at school and that they were avoiding certain places within their schools (Kaufman et al., 1998). By 1999, these fears had subsided somewhat (Kaufman et al., 2000), but parents still say they are afraid for their children at school. A recent Gallup poll found that nearly half of the parents surveyed feared for their children's safety when they sent them off to school, whereas only 24 percent of parents reported this concern in 1977 (Gallup, 1999a). In May 1999, shortly after the shootings at Columbine High School in Littleton, Colorado, 74 percent of parents said that a school shooting was very likely or somewhat likely to happen in their community (Gallup, 1999b).

Public perceptions about school safety seem at odds with the evidence that the risk for serious violence at school has not changed substantially over the past 20 years. But several indicators of violence did increase during the epidemic—school fights, gangs, drug use, and students carrying weapons to school. While gangs and weapon carrying have declined recently, the rates of drug use and physical fighting are high and have not changed between 1991 and 1999 (Brenner et al., 1999). Today's school bullies are still more likely to be carrying guns than those of the early 1980s, and the proportion of students reporting that they felt too unsafe to go to school has not changed since the peak of the violence epidemic in the mid-1990s. These findings add to the concern that the violence epidemic is not yet over.


Gang members, a relatively small proportion of the adolescent population, commit the majority of serious youth violence (see Spergel, 1990, for a review). In two major longitudinal studies in Denver and Rochester (discussed in more detail in Chapter 3), 14 to 30 percent of the youths surveyed were gang members at some time during the study, and they accounted for 68 to 79 percent of the serious violence reported (Thornberry, 1998). Similar findings have been reported in other studies using nonrandomized local samples (Battin et al., 1996; Fagan, 1990). In Rochester, 66 percent of chronic violent offenders were in gangs (Huizinga et al., 1995).

A high proportion of gang members are also involved in drug sales and possessing/carrying a gun, two behaviors closely linked to serious violence. The 1999 National Youth Gang Survey (a national survey of law enforcement agencies) estimates that 46 percent of youth gang members are involved in street drug sales (Egley, 2000). In the Rochester study, 67 percent of youths reporting they owned/carried a gun for protection were gang members and 32 percent reported they sold drugs. Only 3 to 7 percent of non-gun owners or sport gun owners were involved in drug selling. Further, 85 percent of youths who owned guns for protection were involved with peers who owned guns for protection (Huizinga et al., 1995).

Rates of violence are higher in schools where gangs are present. The rate of victimization in schools with gangs is 7.5 percent, compared to 2.7 percent in schools without gangs (Snyder & Sickmund, 1999). Gangs are present not only in inner-city schools, but in many suburban and rural schools as well. Between 1989 and 1995, the proportion of students reporting gangs at their school increased from 15 percent to 28 percent (Snyder & Sickmund, 1999). By 1999, however, that figure had dropped to 17 percent (Kaufman et al., 2000). A decline in the number of gangs in U.S. schools between 1996 and 1997 has also been reported by law enforcement agencies (National Youth Gang Center, 1999).

The National Youth Gang Survey reported more than 26,000 active youth gangs in schools and communities in 1999, down 15 percent from 1996 (Egley, 2000). Yet the same survey reported more than 840,500 active gang members in 1999, a decline of less than 1 percent from the peak level in 1996. Thus, from this source, it appears that the number of youths actively involved in gangs remains very high.

The racial/ethnic composition of gangs in 1999 was 47 percent Hispanic, 31 percent African American, 13 percent non-Hispanic white, and 7 percent Asian. These rates have been relatively constant since 1990.

In 1998, 92 percent of all gang members were male (National Youth Gang Center, 2000), although some evidence indicated that girls' involvement in gangs increased during the epidemic (Chesney-Lind et al., 1996; Chesney-Lind & Brown, 1999; Snyder & Sickmund, 1999). However, the National Youth Gang Survey reports a decline in female membership, with less than 2 percent of gangs nationwide reporting predominantly female membership.

The MTF prevalence estimates for both violent behavior and drug use have been confirmed by other studies in which there is overlap in years and ages. For example, see Elliott et al. (1989) and Menard and Elliott (1993).
About 16,000 high school seniors at 130 schools participate, although only about 3,000 of the students are asked questions about their violent behavior. Since the beginning of the survey in 1975, the participation rate among schools has ranged from 60 to 80 percent, and the student response rate has ranged from 77 to 86 percent (Kaufman et al., 1998).
The prevalence rates for assault with injury and robbery are not additive.
The rates for both the National Youth Survey and the city surveys were calculated by the senior scientific editor of this report (Elliott, 2000b) from gender-specific data in Elliott et al. (1998) and Huizinga et al. (1995).
10  Calculations by Elliott, senior scientific editor, from Snyder (unpublished).
11  The 1998 arrest rate was atypically high for the 1993-1999 period. This rate was twice the rate for every other year over this period and appears to be an anomaly.
12  Figure includes 63 homicides and 12 suicides.

Back to Top

Home | Contents | Previous | Next