National Academies Press: OpenBook

Understanding and Preventing Violence: Volume 1 (1993)

Chapter: 2 Patterns of Violence in American Society

« Previous: 1 The Diversity of Violent Human Behavior
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 42
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 43
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 44
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 45
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 46
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 47
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 48
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 49
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 50
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 51
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 52
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 53
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 54
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 55
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 56
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 57
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 58
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 59
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 60
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 61
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 62
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 63
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 64
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 65
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 66
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 67
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 68
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 69
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 70
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 71
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 72
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 73
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 74
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 75
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 76
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 77
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 78
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 79
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 80
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 81
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 82
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 83
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 84
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 85
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 86
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 87
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 88
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 89
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 90
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 91
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 92
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 93
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 94
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 95
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 96
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 97
Suggested Citation:"2 Patterns of Violence in American Society ." National Research Council. 1993. Understanding and Preventing Violence: Volume 1. Washington, DC: The National Academies Press. doi: 10.17226/1861.
×
Page 98

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 42 2 Patterns of Violence in American Society The mass media daily trumpet the most recent violent events in the world, and violence is an essential ingredient for many viewers of video screen drama. News of a relatively few violent crimes attracts a disproportionate amount of media attention compared with the more numerous violent events that are known only to the families, friends, and acquaintances of particular victims and perpetrators. Accounts of the more dramatic episodes of actual homicides, rapes, and robberies command special attention, as do the shootings of well-known people, gangsters, and police officers. To what extent do the violent criminal events reported in the media represent national patterns of violence in everyday life? How do patterns of violence in the nation today compare with patterns in other countries and at other times? This chapter describes patterns discernible in violent behavior and trends in its occurrence or reporting, focusing on comparative levels of violence, the risks of becoming a victim of violence, the characteristics of violent offenders, and the circumstances surrounding violence. DIFFICULTIES OF MEASUREMENT There is no single way to define, classify, and measure the domain of violent events, because each counting system involves some evaluation of people's observations and reports of what they

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 43 perceive as violent events. Any set of crime statistics, therefore, is based on events that are defined, captured, and processed by some institutional means of collecting and counting. Violent events that are socially constructed as crimes consequently are counted more accurately and completely than those that are not. In addition, society's interpretation of violent events as crimes changes over time and differs across segments of the population. There is no national profile of all the violent events with which this panel was concerned. Rather, three nationwide measurement systems count and classify various components. We describe the strengths and limitations of these national systems here, and then indicate the variety of subnational data bases that are used in violence research. First, for events that police classify as crimes, the Uniform Crime Reports (UCR) system collects basic information about the most serious crime committed. The UCR also records supplementary information about the circumstances of homicides and basic descriptions of arrestees. Second, for homicides, data are also tabulated annually from death certificates by the National Center for Health Statistics (NCHS). Third, for nonfatal violent victimizations, the National Crime Survey (NCS, later, the National Crime Victimization Survey) is a primary data source. The survey periodically asks national samples of all persons in households who are age 12 and older to recall and describe their recent victimizations. Survey questions are worded so as to direct respondents' attention to events that they personally regard as crimes. The task of describing the national pattern of violence from these three systems is complicated by the fact that they differ in terms of (1) the domain of events that they attempt to capture, (2) the unit of count on which their statistics are based, (3) the timing of the counting and tabulation, and (4) the sources of discretion and error in recording and counting events. Table 2-1 provides a summary comparison of these characteristics of the three systems. To help readers interpret national data with the appropriate caution, we summarize these differences and their implications before reporting what is known about national patterns and trends. The difficulties of measuring violence are explained more fully in Appendix B. Differences In Domain Both NCS and UCR data produce national counts of three types of nonfatal violent crimes: forcible rape,1 robbery, and aggravated

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 44 TABLE 2-1 Comparison of Violence Measurement Systems System National Crime Uniform Crime National Mortality Characteristic Survey Reports Statistics Source • Sample of • Incident reports of • Death certificates U.S. participating police households agencies Domain • Nonfatal • Violent crimes • Homicides violent known to police victimizations including homicides of persons aged and violence during 12+ crimes against organizations Unit of Count • Most serious • Most serious crime • Deaths victimization during event during event Timing • Events • Collected • Collected occurring in 6- contemporaneously contemporaneously month reference period • Published • Published annually • Published annually, annually with with 2-year time lag 10- to 12-month lag Sources of • Respondent • Victim/witness • Medical examiner Discretion/ recall, decision to report to judgment Error construction as police and victimization, discretionary police and choice to detection recount • Interviewer • Police judgment determination that crime occurred • Agency rules • Counting rules for counting

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 45 assault. Both count attempted as well as completed crimes. However, within each type, the two sets of statistics are not strictly comparable because they capture different categories of event. The NCS is designed to record nonfatal victimizations of persons age 12 and over. Of these events, the UCR records only those incidents that become known to police as crimes—some 40 to 50 percent, according to NCS estimates. However, as shown in the figure, UCR counts include police-recorded incidents of two categories of events that fall outside the NCS domain: crimes against children younger than 12 and robberies of organizations. Simple assaults, the largest and least serious category of violent victimizations in NCS data, are not counted as index crimes by UCR. Homicides are counted in both UCR and NCHS national mortality statistics. Usually, UCR homicide counts are slightly smaller than NCHS counts; homicides that occur outside police jurisdictions (e.g., in prisons) account for some of the discrepancy. No national system reports threats to commit violent acts. Differences in Unit of Count Our concern is with violent events, but a single violent event may include both a number of violent crimes and a number of offenders and their victims. Differences between the NCS and UCR units of count in these events must be considered in interpreting the two sets of statistics. The event described in Chapter 1 involving Mr. Waldermarian, for example, in which an attempted robbery ended in a shooting death, is counted only as a homicide by UCR (because UCR counts only the most serious crime in the event) and is uncounted by NCS because the victim's death eliminates him from the sample frame. Thus, the attempted robbery is also omitted from both systems. However, if the event involved multiple offenders, both of whom were later arrested, the event would be represented twice in arrest statistics. A related discrepancy occurs, for example, in a robbery of multiple victims simultaneously. The UCR counts such a robbery as a single crime, as does NCS in its estimated count of incidents.2 NCS victimization counts are weighted, however, to reflect multiple victimizations. Differences in Timing The systems that measure violence also differ in terms of their timing relative to the events they measure. Crimes are recorded

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 46 in UCR more or less as they occur. Statistics are published annually with a lag of about six months. In contrast, NCS interviews take place throughout the year, with the respondents asked to recall all victimizations they experienced during the preceding six months, and data preparation and publication require about a year. Therefore, at times some events represented in the most recent NCS report have occurred at least a year before some events counted in the most recent UCR report. Not only does this discrepancy complicate comparisons of trends in the two series, but the comparative recency of UCR data tends to give them more visibility in the eyes of the public and policy makers. National mortality statistics are published with a lag of approximately two years. Sources of Discretion and Error The statistics produced by all three systems reflect not only actual events but also the discretion of the authorities that collect and publish the data. With homicide, for example, the systemic differences noted above probably account for some of the discrepancy between UCR and NCHS homicide counts. But even the NCHS counts are influenced by medical examiners' complex and subtle judgments, which are themselves subject to error. Counts of nonfatal violent events are affected by the interactions of errors and discretion by both the statistical agencies and the victims of violence. Reporting a victimization to an NCS interviewer requires the respondent to accomplish a series of tasks: recall the event, recognize it as a victimization, decide whether it occurred during the six-month reference period, and decide whether it is worth mentioning, given the effort and possible embarrassment or fear that describing it to the interviewer may entail. In turn, the interviewer makes judgments about how to record the responses, and the Census Bureau and the Bureau of Justice Statistics set rules for interpreting, coding, and tabulating the recorded information. Before a violent event is counted in UCR statistics, it must have been reported to or discovered by a police agency, where someone is responsible for following a fairly detailed set of instructions for counting the crimes that occurred during the event. Thus, in both sets of data, complex activities are translated into statistics of violent events. One clear example of how judgments affect national profiles involves what NCS classifies as “series victimizations." Whenever a respondent recalls more than three fairly similar victimizations and is unclear about the precise number or details of each

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 47 one, the policy is to publish series victimization statistics separately from the annual statistics, ascribing the characteristics of the most recent victimization to all events in the series. Each event in the NCS series would be counted in UCR statistics provided that the police had knowledge of the incident. This NCS policy for reporting of series victimizations is probably the most appropriate, given existing knowledge about recall of series victimizations. While the count of single victimizations underestimates total victimizations by at least 9 percent, any arbitrary rule for combining series and single victimizations would significantly raise the count. Counting each series victimization as three incidents (the minimum threshold for designating a series victimization) would increase the total by 28 percent, and weighting each reported series by the respondent's estimate of the count in the series would increase it by a factor of 10. Moreover, assuming that all victimizations in a series are identical to the last one introduces anomalies, especially when very small numbers of respondents report series victimizations of a given type. The social construction of crime—the informal rules of society and the official policies of organizations that designate some acts but not others as crimes—no doubt affects national statistics on violent crime, but these effects have not been measured. As explained further in Appendix B, one effect of recent changes in the position of women in society has been to criminalize sexual and physical assaults against women in the eyes of many. Therefore, more NCS female respondents than in the past can be expected to construe assaults by their husbands and "date rapes" by acquaintances and intimates as victimizations that should be mentioned to an NCS interviewer. There has also been considerable moral and legal pressure for police agencies to treat more acts of domestic violence as criminal assaults, which would be counted in UCR statistics. Without knowing how widely these attitudes have spread, how significantly embarrassment and fear of retribution still inhibit women from reporting such incidents to NCS interviewers and to police, and how responsive police agencies are to such reports, we cannot know how much of the recent rise in overall assault rates is due to these changes in social construction. Two comparisons explained more fully in Appendix B indicate that even after reporting a forcible rape, robbery, or assault to the police, some respondents fail to recount them to NCS interviewers. First, the UCR forcible rape count substantially exceeds the number that "should" have been reported, according to NCS data on the number of forcible rapes and the fraction of these reported

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 48 to police. Second, while intensive "reverse-record comparisons" between NCS and UCR records in two jurisdictions found, as expected, that many violent victimizations reported to NCS interviewers had not been reported to police, they found, for example, that nearly half the aggravated assaults reported to San Jose, California, police were not reported to NCS interviewers. Reiss (1985:166-167) computed a crude estimate that the true San Jose aggravated assault rate was approximately double the NCS rate and triple the rate estimated from police reports. One type of violence that is especially poorly counted is assaults of children by their parents. Some of these go unrecognized because they are socially construed as discipline rather than violence. The NCS counts such incidents only if the victim is aged 12 or older and overcomes the fear and embarrassment associated with recounting the event to an interviewer. A UCR record would require either a self-incriminating parental report to police or a report by a courageous victim, other family member, or an institutional official. The current institutional system for reporting and counting such incidents relies largely on reporting by school, social service, or medical professionals, who may miss physical evidence on many parts of the child's body, who often have no basis for judging whether an assaultive injury was inflicted by a parent or by another family member, and who may justifiably anticipate a confrontation and potential legal action if they report their suspicion. In recent years, the NCS has begun publishing special tabulations of intrafamily violent victimizations. However, the counts are lower than counts based on several special-purpose surveys conducted since 1975. National estimates based on these surveys are discussed in Chapter 5. Secondary Information The primary violence measurement systems also differ in terms of the secondary information they record and report beyond occurrence of a violent event. The NCS reports information on characteristics of the victim, preexisting relationships between the victim and offender, the victim's perceptions of offender characteristics, and the aftermath of the victimization (e.g., extent of injury and financial loss, reporting of the victimization to the police). The NCS data on victim ethnic status are limited, however, and are collected according to current Census Bureau definitions. The program has only recently begun calculating separate

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 49 victimization rates for Hispanics, and the national sample is too small to compute stable estimates for Native Americans or Asians. Because the NCS sample is designed to produce statistically efficient national estimates, it cannot produce reliable estimates for states or local jurisdictions. For nonfatal violent crimes, the UCR program currently reports no information beyond occurrence of the event and place of occurrence in the years for which national data are available.3 The Supplementary Homicide Reports tabulate limited additional information on victims' and offenders' demographic characteristics (if the offender is known), on victim-offender relationships, on circumstances of the event (e.g., committed in the course of a crime), and on the type of weapon. Because this information is tabulated before extensive police investigation in a coding scheme that does not accommodate multiple circumstances or relationships, its validity has been questioned. For city and suburban arrestees, UCR arrest reports tabulate gender and ethnic status (using the categories whites, blacks, Native Americans, and Asians/Pacific Islanders, with no separate category for Hispanics). As a census of all crimes and arrests, the UCR program publishes crime and arrest statistics for states and localities with their own police agencies. Subnational Data Bases Quantitative research on the causes of violence uses not only these national data bases, but also other data that describe subnational areas or selected samples of individuals. Examples of subnational data bases in government agencies include: individual arrest records maintained by police departments and reports of child abuse maintained by local authorities. Individual investigators also assemble research data bases in a variety of ways—by collecting new data (e.g., surveys of violence in the home, longitudinal data on individuals' violent crimes and related life events), by coding data from agency record systems (e.g., hospital emergency department injury admissions, police incident reports), or by linking aggregate-level data (e.g., merging murder statistics and socioeconomic statistics at the census tract level). These kinds of data, and their strengths and limitations for research purposes, are discussed throughout the rest of this book, in conjunction with the findings from research in which they are used. Recognizing the limitations of existing national measurement systems, this chapter describes patterns and trends in violence.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 50 Although our national systems of measuring and counting violent crimes are imperfect and not entirely comparable, they can answer a number of basic questions. COMPARATIVE LEVELS OF VIOLENCE Common perceptions are that American society is more violent than most societies, that it is more violent today than ever before, and that violence is ever increasing. The facts are, unfortunately, more complex and simple generalizations are quite misleading. Historical Trends in Homicide Violent crime rates have fluctuated considerably throughout this century. The homicide rate has peaked twice; each peak was followed by a decline. From 1900 to the early 1930s, there was a substantial increase in homicide. The rate then fell for the next 30 years, to reach a low in the early 1960s. A rise in the 1960s and 1970s peaked in 1980 and was followed by a decline in the early 1980s and higher rates since 1985. The 1990 homicide rate reported by UCR (9.4 per 100,000 population) is somewhat below that of the two previous peaks in the twentieth century (a mean of 9.5 in 1931-1934 and 10.4 in 1979-1981) (Holinger, 1987). Substantial differences in gender and ethnic homicide victimization rates have persisted since the late nineteenth century. Fragmentary evidence for the nineteenth century and more substantial time series for the twentieth show that nonwhite men and women have always been more likely to be murdered than white men and women (see Figure 2-1; Holinger, 1987). Although the ratio of black to white homicide rates varies from 1910 onward, in any year the ratio has never been lower than 5 to 1, and in the 1950s and 1960s the ratio was as great as 11 to 1 (calculated by Gurr, 1989, from Holinger, 1987); in 1989 the ratio was roughly 7 to 1. The homicide pattern for blacks has diverged historically from that for other immigrants to the United States. In the nineteenth and early twentieth centuries, first-generation European immigrants to U.S. cities had generally higher homicide rates than those of the native-born white population (Lane, 1979; Monkkonen, 1989). Although it is difficult to trace the ethnic status of their descendants in contemporary America, the available evidence points to few, if any, differences in homicide rates of later generations of these European immigrant populations (Gurr, 1989).

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 51 Figure 2-1 Age-adjusted homicide rates, by sex and race, United States, 1929-1989. SOURCE: Adapted from Holinger (1987) The historical evidence for blacks suggests that their homicide victimization rates in the twentieth century are generally above those in the nineteenth century. Fragmentary data for some eastern U.S. cities suggest that, from 1860 to 1910, the ratio of black to white homicides varied between 2 and 3 to 1. After 1910, when more complete data are available, the ratio is greater. The 1931-1934 overall peak coincided with peaks for both white and black males. However, the 1979-1981 peak followed a tripling of the white male rate over the preceding three decades; the rate for black males had peaked about a decade earlier and, except for occasional year-to-year deviations, continued to fall until the mid-1980s. The explanation for this historical divergence for blacks may involve the residential segregation and the structure of opportunities for occupational and social mobility that blacks have experienced. According to a recent review (Jencks and Mayer, 1990), however, available evidence does not permit us to separate the effects of family, school, and neighborhood. Trends since 1979-1981 in homicide and nonfatal violent crime rates are examined in more detail later in this chapter.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 52 International Comparisons Despite national differences in counting systems, it is clear that the developed nations of the world vary substantially in their rates of violent crime. The United States has significantly higher rates for most violent crimes than almost all other developed nations. The more serious the crime, the greater the difference between the U.S. rate and those of other developed nations. Comparisons among countries are most valid for the crime of homicide (Figure 2-2). Based on reporting to the World Health Figure 2-2 Crude homicide rates for selected countries, most recent year for which data are available. Data from the World Health Organization, 1987.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 53 Organization during the period 1981-1986, the homicide mortality rate in the United States (8.3 per 100,000 population in 1984) was exceeded only by rates in the Bahamas and Ecuador and was nearly twice that of any European nation. Spain had the next highest rate, 4.3. Most European and British Commonwealth countries report rates on the order of 1 to 2 per 100,000. It is worth noting that recent U.S. homicide mortality rates for whites exceed the overall rates for developed countries generally; however, we do not have comparable homicide rates for nonminority populations in those countries. For nonfatal violence, detailed comparisons are hampered by differences in the ways that countries define violent crimes, victims report crimes to the police, and police detect them and follow information, reporting, and counting practices. In an effort to achieve greater comparability, an International Crime Victim Survey was undertaken by telephone in 1989 (van Dijk et al., 1990). It allows comparisons between the United States and 15 other countries for the violent crimes of robbery, forcible rape, and assaults. However, because of small samples and low response rates, the data are more useful for relative rankings of countries and ranges of difference among them than for precise statistical comparisons. For robbery, the 1988 personal prevalence victimization rates were highest in Spain and the United States. For serious sexual assaults (forcible rape, attempted rape, indecent assaults), the United States had the highest prevalence rate, with Canada, Australia, and West Germany also reporting rates well above those of most of the other countries. Assaults show a somewhat different pattern, perhaps owing to the nature of the question asked to elicit responses. The differences among countries were less marked; the rates for Australia and Canada were close to those of the United States. What is puzzling in these data is that the homicide rates of the respective countries do not seem closely related to their assault rates. Especially noteworthy are the relative high rates of assault with force of the Netherlands and West Germany compared with their very low homicide rates. Some of the difference may be due to country differences in the ownership of handguns, which have greater lethality in assaults. Household possession of handguns was substantially greater in the United States than in any European country—a reported rate of 29 percent compared with less than 7 percent in any other country surveyed except Switzerland. The United States has considerably more serious violent crime than does Canada, a country that classifies most offenses under

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 54 rules comparable to ours. Although violent crime has been increasing in both countries since 1960, substantial differences persist between the rates reported by the UCR for both countries. The 1988 U.S. rate for homicide was four times that of Canada; the U.S. rate for robbery was two and a half times that of Canada. For 1982, the last year for which Canadian data are available, the U.S. rate for forcible rape was on the order of three times that of Canada. A satisfactory explanation of these disparities would have to take into account differences in the composition and distribution of the population as well as differences in culture and firearm availability. VIOLENT CRIME IN THE UNITED STATES How much crime is violent? The technical problems discussed above preclude any precise estimate of how much of all crime in the United States is violent. We report below some crude estimates, which are useful mainly to establish relative magnitudes among the types of crime and to discern overall trends. Comparisons focus primarily on the 1990 reporting year, the latest available at this writing. National Estimates In 1990 the National Crime Survey reported an estimated 34,403,610 personal and household crime victimizations.4 Of these, 17 percent were attempted or completed violent crimes—8 percent if one excludes simple assault.5 In considering only the 18,984,120 attempted or completed personal victimizations reported, about 32 percent were violent—just over 6 million. Of these, just over half were simple assaults, the least serious violent offense (Bureau of Justice Statistics, 1992: Table 1a). The bulk of personal victimizations reported are theft from persons, which includes larceny with and without contact with the victim.6 Together with the violent act of simple assault,7 these nonviolent thefts constitute the large majority of personal victimizations. In 1990 UCR estimated that the nation's police departments received reports of 14,475,600 index crimes, excluding arson; of these, about 1.8 million, or 13 percent, were attempted or completed violent crimes.8 Police departments report a smaller proportion of all offenses as violent than the NCS reports, excluding simple assault (13 compared with 15 percent). This is in part due

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 55 to the fact that victims were somewhat more likely to report property crimes than crimes of violence to the police. In 1988 victims reported a higher fraction of household victimizations of burglary, household larceny, and motor vehicle theft (40 percent) than violent personal victimizations (36 percent) to the police (Bureau of Justice Statistics, 1990: Table 92).9 Prototype Pattern The NCS program reported 6,008,790 violent victimizations in 1990, and the UCR program counted 1,820,127 violent index crimes reported to police agencies. Despite their differences, the prototype violent event that emerges from the two systems is an assault, either aggravated or simple (Table 2-2). Of the violent victimizations, aggravated and simple assault account for nearly 8 of every 10. By excluding homicide from the UCR and simple assault from the NCS counts to achieve rough comparability, aggravated assaults account for almost 6 of 10 violent crimes reported in both systems. Robbery, which is an assault or threat of assault for the purposes of taking valuables, accounts for most of the rest—which means that at least 9 of every 10 acts reported as violent crimes or victimizations in the United States are either assaults or robberies. Sexual assaults account for most of the remaining violent offenses, and homicides account for just over 1 percent—23,438 of the violent crimes counted by the UCR. As we said earlier, estimates of the frequency of criminal events are influenced by the decisions of victims and witnesses to report them, of officials to process them, of police detection practices, and of institutions to devise rules for counting them. Just how closely institutional reporting accords with the occurrence of events is a matter of disagreement, but, for example, it seems clear that murders are counted far more accurately than forcible rapes. Despite the errors, we can conclude that generally in 1990, the more serious the violent crime, the fewer the occurrences recorded. Historically, societal sanctions for crimes yielded the rank order of seriousness reflected in the rows of Table 2-2: murder, forcible rape, robbery, aggravated assault, simple assault. The UCR column of the table illustrates the inverse seriousness/frequency relationship for the four most serious violent crime types. The NCS column illustrates it for forcible rape, robbery, aggravated assault, and simple assault victimizations. In both columns, less serious acts are reported more frequently than more serious acts. Historically, completed crimes have been considered more serious

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 56 TABLE 2-2 Offense and Victimization Rates for Violent Crimes, United States, 1990 UCR Offensesa NCS Victimizationsb Rate per 100,000 Type of Number Rate per Number All Age 12 Violent Crime 100,000 Persons & Older Persons All violent 1,820,127 731.8 6,008,790 2,415.9 2,956.0 crimes All violent 1,820,127 731.8 2,880,660 1,158.2 1,417.1 crimes, excluding simple assault All violent 1,796,689 722.4 2,880,660 1,158.2 1,417.1 crimes, excluding simple assault and homicide Murder and 23,438 9.4 — — — nonnegligent manslaughter Forcible rape 102,555 41.2 130,260 52.4 64.1 Forcible rapec (80.0) (104.7) (123.5) Robbery 639,271 257.0 1,149,710 462.3 565.6 Aggravated 1,054,863 424.1 1,600,670 643.6 787.4 assault Simple assault — — 3,128,130 1,257.7 1,538.9 Total persons 248,709,873 US Total persons, 203,273,870 age 12 and older a Source: Federal Bureau of Investigation (1991: Table 2). b Source: Bureau of Justice Statistics (1992: Table 1a). c Forcible rape rates calculated for female population only (Federal Bureau of Investigation, 1990: Appendix III p. 329; Bureau of Justice Statistics, 1992: Table 5).

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 57 than attempted crimes, and crimes that involve weapons or cause injury are considered more serious than those that do not. Table 2-3 distinguishes between completed and attempted victimizations and reports the counts of assaults and victimizations that involve weapons and injuries. With a few exceptions, the data in the table conform to the expectation that less serious victimizations occur more frequently than more serious acts. The greater frequency of completed than attempted robberies appears to contradict expectations regarding frequency and seriousness; in 1990 the NCS reported more than twice as many completed robberies as attempted ones. However, this proportion may not reflect the true pattern of occurrence. Attempts may be reported less frequently to NCS interviewers, even if they occur more frequently. Evidence on school crime and other studies suggests that attempted as well as actual robberies of children under age 12 and of schoolchildren at all ages are quite common. An alternative conjecture is that a higher proportion of all forcible rapes and robberies may be brought to completion because the victim is more compliant when the apparent risk is greater in these crimes and less likely to take forcible means of resistance. Research is needed to learn more about the relationships that link reporting and self-protective behavior to characteristics of violent events for all violent crimes. Especially for the crimes of forcible rape and robbery, such studies should investigate how victim compliance under threat affects injury, completion, and reporting rates. The various types of assault show the expected inverse relationship between frequency and seriousness. In 1990, 67 percent of assaults were attempts. Of all aggravated and simple assaults, 46 percent were attempted simple assaults without a weapon and without injury; 21 percent were attempted aggravated assaults with a weapon, but without injury; 20 percent were simple assaults with injury; and 13 percent were aggravated assaults with injury (Bureau of Justice Statistics, 1992: Table la). Assaults with injury are least frequently reported; attempted simple assaults without a weapon and without injury constitute almost half of all reported assaults. Violent victimizations are more likely than nonviolent ones to involve multiple simultaneous victims. Therefore, on average, every 100 aggravated assault victimizations occurred in 80 events and every 100 simple assaults in about 90 events.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 58 TABLE 2-3 Number and Percent Distribution of Victimizations, by Sector and Type of Crime, 1990 Sector and Type of Number Percentage of Crimes Total Persons Crime Within Sector Personal sector 18,984,120 100.0 203,273,870 Crimes of violence 6,006,790 31.7 Completed 2,421,530 12.8 Attempted 3,587,260 18.8 Rape 130,260 0.7 Completed 62,830 0.3 Attempted 67,430 0.4 Robbery 1,149,710 6.1 Completed 800,510 4.2 With injury 286,020 1.5 From serious assault 123,740 0.7 From minor assault 162,280 0.9 Without injury 514,480 2.7 Attempted 349,190 1.8 With injury 110,380 0.6 From serious assault 43,930 0.2 From minor assault 66,440 0.4 Assault 4,728,610 24.9 Aggravated 1,600,670 8.4 Completed with injury 627,000 3.3 Attempted with weapon 973,660 5.1 Simple 3,128,130 16.5 Completed with injury 931,170 4.9 Attempted without 2,196,960 11.6 weapon Crimes of theft 12,975,320 68.3 Completed 12,154,550 64.0 Attempted 820,760 4.3 Personal larceny with 637,010 3.4 contact Purse snatching 165,490 0.9 Completed 124,010 0.7 Attempted 41,470 0.2 Pocket picking 471,520 2.5 Personal larceny without 12,338,319 65.0 contact Completed 11,559,010 60.9 Less than $50 4,592,470 24.2 $50 or more 6,452,940 34.0 Amount not available 513,590 2.7 Attempted 779,290 4.1

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 59 Consequences of Violent Crime The major consequences of violence involve physical injury, psychological consequences, monetary and other costs, and lethality (the probability of death). There are no national data on psychological consequences, and those on economic costs are limited. Even the information on physical injury is limited since there is no assessment of the injury independent of the victim's self-report on the kind of injuries and the kind of medical treatment obtained for them. Physical Injury Physical means of violence include a person's body (hands, fists, feet), instruments such as firearms and knives, flammable liquids and explosives, poisons, and animals (such as attack dogs). Physical injury can range from minor harm, such as bruises, lacerations, and scrapes, to death. During 1990, the NCS reported that the victim sustained some physical injury in 33 percent of robbery and assault victimizations (Bureau of Justice Statistics 1992: Table 80). The rate of physical injury was somewhat greater for aggravated (39%) than simple (30%) assault and for completed (36%) than attempted (32%) robberies. The Bureau of Justice Statistics (1989a: Table 4) reported the distribution of injury severity and level of medical care received by injured victims for violent crimes in the United States, aggregated for the years 1979 to 1986.10 The bulk of the injuries—83 percent—were considered minor in severity. Over half—52 percent—did not require any medical treatment. One-fifth required hospital care, but only 4 percent required an overnight stay. In any given year, only a small proportion of those with an overnight stay required more than a single day of hospitalization. Nevertheless, of the injured 1990 victims, 38 percent received hospital care for physical injuries from violent crime. Of the injured, 23 percent were hospitalized for less than a day, 7 percent were hospitalized from one to three days, and 8 percent for four days or more (Bureau of Justice Statistics, 1990: Table 88). Costs Physical and emotional trauma are usually assessed initially in terms of medical standards of physical or emotional impairment. Subsequently, these traumas and resulting impairment may be

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 60 evaluated in monetary terms, as the cost of rehabilitation. Physical injury imposes costs on the injured person of emergency and continuing medical treatment. Violent deaths impose economic and psychological costs on surviving family members and loved ones. An impaired capacity to work, to continue in school, or to maintain one's quality of life are less often recognized, but consequential, costs of violent victimization. Loss of property in violent crimes is another way in which violence imposes economic harm. Cost estimates prepared for the panel are $54,100 per attempted or completed forcible rape, $19,200 per robbery, and $16,500 per assault (Cohen et al., Volume 4). We emphasize that these estimates involve assumptions about what costs to include, who bears them individually and collectively, and the value of a life lost. Only about 15 percent of these costs are monetary, for such things as money losses, lost productivity, emergency medical services, and administration of compensation; 85 percent are imputed costs for nonmonetary losses such as pain, suffering, psychological damage, and the risk of death or reduced quality of life. There are additional costs to society arising from the discretionary collective response to violent victimization. Law enforcement, adjudication, victim services, and correctional expenditures add thousands of dollars of cost to each criminal event. These costs are estimated by Cohen et al. (Volume 4). Less direct consequences of violence may be delayed or cumulative. The stress induced by violent acts, especially when repeated within an intimate relationship, may culminate in severe emotional trauma or physical illness. Stresses may also accumulate for communal organizations. Although the cumulative effects of violent crime on neighborhood cohesion and public services are not easily measured, they are substantial for some communities. Delayed effects occur in even less obvious ways. The intergenerational transmission of violent behavior by physically abusive parents, siblings, and caregivers, for example, is a delayed cost, as new generations of violent persons inflict physical harm on their families and others in the community. Lethality Violent crimes vary considerably in their lethality—the probability that they will cause the victim's death. There are no routine published estimates of lethality because, as we explained above, the counts of homicides and victimizations come from different

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 61 sources. Nevertheless, by combining UCR estimates of the fraction of homicides that occur in crime-related circumstances with NCS counts of violent victimizations, the panel calculated a crude estimate of lethality: in 1990, about 1 in every 257 violent victimizations was fatal to the victim. This estimate reflects an increase from about 1 in every 287 in 1988. Homicide occurs more frequently in conjunction with some crimes than others. Rough estimates combining the same data sources are that death occurred in about 1 in every 500 forcible rapes and robberies, 1 in every 120 assaults with injury, and about 1 in every 400 arsons. These 1990 estimates are roughly unchanged from 1988 and, for forcible rape and robbery, are approximately the same as 1987 estimates prepared by Cohen et al. (Volume 4). VICTIMS OF VIOLENCE Although the rhetoric of campaigns against violence suggests that victims and offenders are distinct populations, in fact there is great similarity in their demographic profiles. Generally, both groups—those at the highest risk of violent offending as well as those at highest risk of violent victimization—tend to be young, black males of low socioeconomic status who live in the nation's central cities. The most insightful explanations of this victim-offender overlap are based on studies of local communities and samples of individuals, which we discuss in Chapter 3. Victim characteristics are discussed in this section, and offender characteristics in the next. Risks of Violent Victimization What is the risk of becoming a victim of a violent crime? We generally lack the information to calculate national estimates of risks that are contingent on the prior states, activities, and conditions that theory or practical experience suggests as possible causes of violence. Evidence about such determinants of risk, which is usually developed in special studies of small samples, is discussed in later chapters. Here we focus on the most commonly available national estimates of risk: rates for demographically defined subpopulations. The annual risk of becoming a victim of personal violence—of homicide, forcible rape, robbery, or assault—is well below that of victimization from property crimes. As judged by victim reports

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 62 to NCS, the risk of becoming a victim of personal violence in 1990 was 1 in 34 for people age 12 and older. That risk is less than half the risk of becoming a victim of a personal theft, which was 1 in 16 in 1990 for people age 12 and older (Bureau of Justice Statistics, 1992: Table 1a). In 1990 the risk of household victimization through burglary, household larceny, or motor vehicle theft was 1 in 6 households (Bureau of Justice Statistics, 1992: Table 1a). For many purposes, the lifetime risk of violent victimization would be a more informative description than the annual risk for any single age range. The Bureau of Justice Statistics, drawing on the annual prevalence rates from the National Crime Survey, estimated that about 83 percent of the people now age 12 will be violently victimized in their lifetimes (Koppel, 1987). There is good methodological reason to conclude that this is a substantial overestimate (Lynch, 1989). However, because information is unavailable on the proportion of victims in any year who have also been victimized in previous years, we were unable to determine the degree to which it is overestimated. The panel recommends that ways be developed to measure the lifetime risk of nonfatal violent victimizations, especially risks of injury by serious assault, for different groups of people in the population. Homicide Usable estimates of lifetime risk have been calculated for homicide (Loftin and Wiersma, 1991). In 1987 and 1988, the annual risk of becoming a victim of a homicide was about 1 in 12,000, although by 1990 it had risen close to 1 in 10,600. The lifetime risk of being a homicide victim is of course much greater (Figure 2-3). Of the six demographic subgroups shown in the figure—male and female whites, blacks, and American Indians—black males are at the highest lifetime risk: 4.16 per 100 black males, which is equivalent to a 1 chance in 24.1 of dying by homicide. American Indian males are also at high risk (1.75 per 100 American Indian males, 1 chance in 57), as are black females (1.02 per 100 black females, 1 in 98.1). For white males and females and for American Indian females, chances of dying by homicide are substantially less than 1 in 100.11 This figure also makes clear that, despite media attention to the killings of adolescents and young adults, less than one-fourth of one's lifetime homicide risk is experienced before the twenty-fifth birthday. For five of the six subgroups (the exception is

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 63 American Indian males), the ratio of homicide risk by age 24 to lifetime risk lies in a narrow range, between 0.21 and 0.26. Murders of American Indian males occur later in life on average; the corresponding ratio is only 0.14. Figure 2-3 Cumulative homicide rate in five-year age intervals by race and gender, 1987. Recent media attention has focused on homicides of young black males. The rate for black males ages 15-24 rose during the late 1980s, approaching levels not observed since a previous peak around 1970 (see Figure 2-4). However, Figure 2-3 makes clear that, although a sizable fraction of black males die of murder by age 24, the high homicide rate for young black males must be viewed in light of the high homicide rate for black males at all ages. One major unresolved question is why the homicide death rate is so high for blacks at all ages, especially black males (see Griffin and Bell, 1989). Another is to understand why, for this age category, the trends for black and white males have moved in different directions during several periods since 1940, as Figure 2-4 shows. We return to these issues in the context of firearms in Chapter 6 and in terms of community-level factors in Chapter 3.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 64 Figure 2-4 Homicide rates, persons ages 15-24 years, by race and sex, 1940-1988. There is considerable variation in homicide mortality by gender and age. In 1988, for example, the homicide mortality rate ranged from 4.2 for females to 14.2 for males, and from 5.3 for whites to 34.4 for blacks. The variation by age and sex within these groups was, for example, from 2.7 for black males age 5 to 9 to 128.2 for black males age 20-24. The range for white males was from 0.8 for age 5 to 9 to 14.6 for white males 20-24 (National Center for Health Statistics, 1991). The UCR Supplementary Homicide Report discloses that most victims in single offender-single victim homicides are slain by an offender of the same ethnic status:12 86 percent of white victims and 93 percent of black victims in 1990. In this kind of homicide, only 6 percent of black victims were slain by white offenders and 12 percent of white victims were slain by black offenders (calculated from Federal Bureau of Investigation, 1992:11). Multiple homicides by one offender in a single event (so-called mass murders) account for only a very small proportion of all homicides. Similarly, in projecting from data on homicides by known offenders, instances of serial homicide by one offender also appear to be quite infrequent—on the order of 1 percent of all homicides, according to estimates prepared by Fox and Levin for the panel (J. Fox and J. Levin, personal communication, 1990). Importance of Homicide as a Cause of Death The major causes of traumatic death in the United States are accidents, suicide, and

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 65 homicide. Of these, homicides contribute the smallest number: from 1945 to 1988, the risk of death by homicide was about two-thirds that of death by suicide, one-third the risk of death by motor vehicle accident, and about one-fourth the risk of death by all other accidental causes. As Table 2-4 indicates, however, the overall homicide death rate per 100,000 population has increased since 1950, while the suicide rate has remained roughly constant and the motor vehicle accident death rate has decreased. Overall, in the United States, homicide accounts for about one-fifth of all deaths due to suicide, homicide, and motor vehicle accidents. Adjusting annual death rates by age,13 however, discloses significant variations in risk by gender and ethnic status. In the United States, the relative risk of death by motor vehicle accident varies primarily by gender, and the risks of homicide and suicide vary primarily by ethnic status. Table 2-4 shows that, among black males and females, homicide (and legal intervention) is the leading cause of death by traumatic means, whereas among white males and females, homicide is substantially less common than the other causes. For both whites and blacks, the age-adjusted traumatic death rates are three to four times greater for males than for females; although the male-female differences fluctuate over time, the differences between black males and females are greater than those between white males and females. Blacks are more likely to be victims of homicide; whites are more likely to be victims of suicide. Years of Potential Life Lost Although homicide accounts for a relatively small proportion of deaths each year, a different calculation of its cost tells a somewhat different story. Figure 2-5 shows the years of potential life lost before age 65 (YPLL) (Haenazel, 1950) to death from traumatic causes, heart disease, and malignant neoplasms (i.e., cancers).14 Because homicide victims are younger on average than suicide victims, total YPLL for both causes of death are nearly identical, even though the homicide death rate is only 75 percent of the suicide death rate. For analogous reasons, the homicide YPLL is more than 40 percent of the YPLL for heart disease and about 36 percent of the YPLL for malignant neoplasms, despite much larger differences in death rates. Other Sources of Violent Death Comparisons of homicides with other sources of violent death beyond the panel's scope provide another perspective for our findings.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 66 TABLE 2-4 Age-adjusted Death Rates for Selected Causes of Death: United States, Selected Years by Gender and Ethnic Status (per 100,000 resident population) Year Cause 1950a 1960a 1970a 1980a 1987a 1988b 1989b All ethnic status and both genders Homicide and 5.4 5.2 9.1 10.8 8.6 9.0 9.4 legal interventionc Suicide 11.0 10.6 11.8 11.4 11.7 11.4 11.3 Motor vehicle 23.3 22.5 27.4 22.9 19.5 19.7 18.9 accidents White male Homicide and 3.9 3.9 7.3 10.9 7.7 7.7 8.1 legal intervention Suicide 18.1 17.5 18.2 18.9 20.1 19.8 19.6 Motor vehicle 35.9 34.0 40.1 34.8 28.4 28.5 26.8 accidents Black male Homicide and 51.1 44.9 82.1 71.9 53.8 58.2 61.5 legal intervention Suicide 7.0 7.8 9.9 11.1 12.0 11.8 12.5 Motor vehicle 39.8 38.2 50.1 32.9 28.5 29.6 29.4 accidents White female Homicide and 1.4 1.5 2.2 3.2 2.9 2.8 2.8 legal intervention Suicide 5.3 5.3 7.2 5.7 5.3 5.1 4.8 Motor vehicle 10.6 11.1 14.4 12.3 11.4 11.6 11.5 accidents Black female Homicide and 11.7 11.8 15.0 13.7 12.3 12.7 12.5 legal intervention Suicide 1.7 1.9 2.9 2.4 2.1 2.4 2.4 Motor vehicle 10.3 10.0 13.8 8.4 8.7 9.2 9.1 accidents a Source: Adapted from U.S. Department of Health and Human Services (1990: Table 23, p. 121-122). b Source: National Center for Health Statistics (1991: Part A). c Legal interventions includes injuries inflicted by the police or other law enforcing agents including military on duty in the course of arresting or attempting to arrest law breakers, suppressing disturbances, maintaining order, and other legal actions including legal execution.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 67 Figure 2-5 Total years of potential life lost for selected causes of death, 1987. Every year a relatively small number of deaths are classified as justifiable homicides—persons killed in self-defense or the justifiable defense of others. And periodically the casualties of military actions are a major cause of violent injury and death. Estimated American battle deaths for the Korean conflict are 34,000; for the Vietnam conflict, 47,365 (Bureau of the Census, 1990: Tables 556 and 557). In a year of intense conflict, the number of deaths in battle can equal or exceed those by homicide. In 1968, for example, the year when casualties in the Vietnam conflict were greatest, there were 14,623 American battle deaths and 13,648 reported deaths from murder and nonnegligent manslaughter in the continental United States (Bureau of the Census, 1990).15 Other Violent Crimes For violent crimes that do not end in death, Table 2-2 shows that victimization rates from the two major reporting sources were substantially different in 1990. The aggregate rate for violent victimizations calculated from the NCS count was more than triple the UCR count of violent crimes known to the police. Previously discussed differences in the way the two systems classify and

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 68 count violent crimes—especially the counting by NCS of events not reported to police and of simple assault—largely account for these discrepancies. Two differences in the rates are especially noteworthy. The rate for forcible rape reported in the NCS, although greater than that reported by UCR, is well below what one would expect, given the rates at which forcible rape victims identified by NCS claim to have reported their victimizations to police. Research reviewed by Loftin and MacKenzie (1990) and others suggests that the NCS rate is an underestimate, perhaps because victims—even some who have reported the crimes to the police—may be inhibited by fear or embarrassment from discussing their victimization with the NCS survey interviewers. Also, they may not characterize forced intercourse by intimates as a victimization to be reported. The UCR is also likely to undercount incidents involving intimates and appears to undercount forcible rape attempts. The other difference worth noting is that for robbery and aggravated assault in 1990, the ratios of UCR to NCS counts are about 7 percentage points higher than the fractions of NCS respondents who claim to have reported those crimes to the police. The “excess" UCR counts are probably explained by the broader coverage of UCR—commercial robberies and victimizations of children younger than 12 that are outside the NCS scope. Individual Risks of Nonfatal Violent Victimizations In this section we examine how the risks of violent victimization differ according to people's age, gender, ethnic status, marital status, and socioeconomic status. Two- and three-way associations for age, gender, and ethnic status suggest that the relationships among these factors are generally additive (Sampson and Lauritsen, Volume 3). For some subgroups of the population, the relative risk of violent victimization is higher than expected because of interaction effects. Age has the largest independent effect, followed by gender; ethnic status has the smallest effect of the three. The rates reported are based on statistical information for the 1990 National Crime Survey (Bureau of Justice Statistics, 1992) unless another source is specified. Age Age is one of the most important single predictors of an individual's risk of violent victimization (Hindelang et al., 1978). The annual

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 69 risk of victimization by violent crime peaks at age 16 to 19 for both men (95 per 1,000 population) and women (54 per 1,000) and declines substantially with age, to 3-4 per 1,000 at age 65 and older. This declining risk of victimization by age holds for the major violent crimes of forcible rape, robbery, aggravated assault, and simple assault, with substantially higher rates below than above age 25. Although the risk of violent victimization is highest at the younger ages, there is evidence that juvenile victimization is less likely to involve serious injury (Garofalo et al., 1987). Gender Except for forcible rapes and partner assaults, the risk of a woman becoming a victim of a violent crime is lower than that of a man. Among all female murder victims in 1990, however, 30 percent were slain by husbands or male friends compared with 4 percent of male victims killed by wives or women friends (Federal Bureau of Investigation, 1992:13). The lifetime risk of homicide is three to four times greater for men than for women. Gender differences are much smaller for robbery (Bureau of Justice Statistics, 1992: Table 3). Women are substantially less likely than men to report being victims of aggravated or simple assault or attempted assault with a weapon. The risk of injury for assault victims is somewhat greater for women than men (38 per 1,000 for women and 30 for men). Women have a higher rate of both simple and aggravated assault by relatives than do men. Their vulnerability to assault by relatives is greater for simple than aggravated assault: the 1990 rate of simple assault by relatives was six times greater for women than men; the rate for aggravated assault was only roughly twice that for men (Bureau of Justice Statistics, 1992: Appendix V). The reported forcible rape rate for women16 in 1990 was 1.0 for every 1,000 women age 12 and over (Bureau of Justice Statistics, 1992:Table 3), well below their rates of aggravated assault (4.5) and simple assault (12.7). Female children are three times more likely than male children to be sexually abused (Sedlak, 1991:5–7). Ethnic Status Americans of minority status are at greater risk of victimization by violent crime than are those of majority status. The overall 1990 violent victimization rate reported by the NCS was 39.7 for blacks and 37.3 for Hispanics compared with a rate of 28.2 for whites (Bureau of Justice Statistics, 1992: Tables 6 and 8).

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 70 Simple assaults are a substantially larger proportion of all violent crimes for whites (56%) than for blacks or Hispanics (35%), but the risk of simple assault is about the same for these ethnic groups (16 per thousand whites, 14 for blacks, 13 for Hispanics). Excluding simple assaults from the violent crime rate, the rate of violent crime (forcible rape, robbery, and aggravated assault) for blacks and Hispanics is roughly twice that of whites—13 for whites, 26 for blacks, 24 for Hispanics. For reasons discussed early in this chapter, it is not possible to report victimization rates for other ethnic groups. As noted above, blacks, especially black males, are disproportionately the victims of homicide. Although variations have occurred over time in rates by ethnic status and sex, black rates of homicide have exceeded white rates since at least 1910. In 1990 half of all homicide victims were black, and blacks were homicide victims at a rate six times that of whites (Federal Bureau of Investigation, 1992). Several studies using subnational data have found that the black-white homicide differential is attenuated substantially at high income levels. We discuss this interaction, and others that may help to explain this differential, in Chapter 3. American Indians and Alaska natives are also at greater risk of homicide than are white Americans, though exact comparisons are lacking. In a recent special report, the Indian Health Service (1991) placed their age-adjusted rate at 14.1 per 100,000 in 1988—above that for the total population at 9.0 but half the 28.2 rate for all groups other than white. According to the Indian Health Service (1988: Chart 4.21) the rate for Native Americans has substantially exceeded the white rate, but has not exceeded the rate for the entire nonwhite population, since at least 1955. Socioeconomic Status Family income,17 the primary indicator of socioeconomic status measured by the NCS, is inversely related to the risk of violent victimization. In 1988 the risk of victimization was 2.5 times greater for individuals in families with the lowest income (under $7,500) as the highest ($50,000 and over). Of all violent crimes, this negative relation is strongest for robbery. The net effect of family income is less than that for age, gender, race, and marital status (Sampson and Lauritsen, Volume 3). Its contribution relative to these other factors may be negligible; consequently it remains unclear just how much and in what ways

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 71 poverty contributes to the risk of violent victimization—an issue we take up in Chapter 3. VIOLENT OFFENDERS There is more uncertainty about the perpetrators of violent crimes than about their victims because of measurement errors in arrest records and sampling errors in surveys of offenders' self-reports. Because the two data sources are subject to different sources of error, one can be fairly confident about conclusions on which they converge. The panel cautions readers against interpreting annual statistics on arrestees as an indicator of the distribution of the people actually committing crimes: because persons arrested more than once in any year are disproportionally represented in arrest statistics, there are doubts that the arrest population is representative of the offender population. Personal Characteristics Ethnic Status Blacks are disproportionately represented in all arrests, and more so in those for violent crimes than for property crimes. In terms of violent crimes, blacks constitute 45 percent of all arrestees. They are most overrepresented in the most serious violent crimes of homicide, forcible rape, and robbery (Table 2-5). Particularly striking is their substantial overrepresentation in the crime of robbery, a crime that is both a person and a property crime. Other minorities are also overrepresented among all arrestees and among those arrested for violent crimes. Particularly striking is the relatively high representation of American Indians and Alaska natives, especially for aggravated and other assaults, given their proportions in the U.S. population (Federal Bureau of Investigation, 1990: Table 38). It is not possible to calculate annual arrest rates for violent crimes for most demographic categories; however, rates of arrest can be calculated separately for whites, blacks, and others.18 Thus: 1 white was arrested for every 576 whites in the population; 1 black for every 94 blacks; and 1 "other" for every 739 "others." This arrest rate for violent crimes is about six times greater for blacks than whites. However, because data are not available on repeat arrests during the year, these arrest incidence figures do not reflect the annual prevalences of arrest for the different subgroups.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 72 TABLE 2-5 Percentage Distributions of All Persons Arrested for Violent Crimes, by Ethnic Status, 1990 Ethnic Status Offense White Black American Asian or Total Charged American American Indian or Pacific Alaskan Islander Native Murder and 43.7 54.7 0.7 0.9 100.0 nonnegligent manslaughter Forcible rape 55.1 43.2 0.8 0.9 100.0 Robbery 37.7 61.2 0.4 0.8 100.0 Aggravated 59.9 38.4 0.9 0.8 100.0 assault Other assaults 64.1 33.9 1.2 0.8 100.0 Source: Federal Bureau of Investigation (1991: Table 38). Gender and Age Men make up 89 percent of all people arrested for violent crimes (Federal Bureau of Investigation, 1991: Table 37). Women accounted for only 10 percent of all arrestees for murder and nonnegligent manslaughter, 8 percent of those for robbery, and 1 percent of arrestees for forcible rape. They accounted for a higher proportion of those arrested for assaults: 13 percent for aggravated assault and 16 percent for simple assaults. In 1990, arrestees for violent crimes were somewhat older on average than victims, with more falling in the age range of 25–29 than in any other. The age distribution for female arrestees is similar to that for males. Males under age 18, who constituted 16 percent of the U.S. population in 1988, represented roughly comparable portions of male arrestees for murder and nonnegligent manslaughter (14%), forcible rape (15%), aggravated assault (14%), and other assault (15%); they were overrepresented among arrestees for robbery at 24 percent (Federal Bureau of Investigation, 1990: Table 34). The only major gender-age interaction of note is that the male arrest rate declines beginning at ages 45–49, while the female arrest rate remains fairly constant after age 45. After age 65, the female arrest rate approaches half that of males with both rates at their lowest for any age—comparable to their arrest rate for murder at age 14 and under. Figures 2–6 A, B, C, and D present 1988 gender and age profiles for people arrested for violent crimes.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 73 Figure 2-6A Arrest rates for murder and nonnegligent manslaughter, 1990. Figure 2-6B Arrest rates for forcible rape, 1990. For females, there are too few cases cited to be reliable.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 74 Figure 2-6C Arrest rates for robbery, 1990. Figure 2-6D Arrest rates for aggravated assault, 1990.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 75 Patterns of Offending Violent Co-offending Nearly three-fourths of all violent crimes are committed by lone offenders. Forcible rape was most likely to be committed alone: in 1990 only 12 of every 100 forcible rapes involved co-offenders. Robbery was most likely to be committed with others: in 1990 about 48 of every 100 completed robberies involved co-offenders. Although the majority of violent crimes involve a single offender, co- offending substantially increases the number of people involved in violent victimizations. For example (using Figure 2-7 and Bureau of Justice Statistics, 1992: Table 70), for every 100 completed robberies, there will be a minimum of 182 offenders. That means that 182 people would have to be apprehended to clear all 100 robberies by arrest. Because a fair number of these offenders are also involved in other robberies or/and in other offenses with other co-offenders, they are linked in a much larger offending Figure 2-7 Violent crimes: percentage distribution of co-offenders by type of offense, 1990.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 76 network that recruits and selects offenders (Reiss and Farrington, 1991). Criminal Careers A question asked by policy makers and others is whether criminals specialize in committing violent crimes or if "careers" in crime show a pattern of escalation from nonviolent to violent crimes. If either of these situations were the case, it would have implications for the effectiveness of selective incarceration strategies to reduce levels of violence. The evidence, however, shows otherwise. Studies of various American and European cohorts have found that no more than 1 in 5 persons ever arrested had an arrest for a violent crime. Furthermore, such arrests were likely to be embedded in long careers dominated by arrests for nonviolent crimes, so that arrests for violence accounted for no more than 1 in 8 of all arrests in the group studied (see for example Farrington, 1991). There is a clear tendency for adult criminals to specialize in various kinds of property crimes or in various kinds of violent crimes rather than to switch between the two types (Blumstein et al., 1986; Farrington et al., 1988). Although few criminal careers begin with a violent crime, most lengthy careers contain at least one. This pattern is sometimes erroneously interpreted as evidence of escalation from nonviolent to violent crime, or as a demonstration that certain arrest patterns predict subsequent violent crimes. In fact, by most measures, predictions of future violent behavior from arrest records have proven highly inaccurate (Monahan, 1988; Piper, 1985). This is true in part because officially recorded violent crimes are committed largely in the course of lengthy, versatile criminal careers rather than by specialized violent career criminals who could be easily targeted in a strategy of incarceration. Victim-Offender Relationships Overlap Between Offenders and Victims National-level estimates are available on social and personal relationships between victims and offenders that may affect the distribution and consequences of violent events. By social relationship, we mean whether the offender and victim are of the same or different categories—defined in such terms as gender,

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 77 ethnic status, and sexual preference. By personal relationship we mean the connections between victim and offender as individuals—strangers, drinking partners, lovers, spouses, custodian and inmate, and others. Social Relationships Clear-cut statements can be made about social relationships defined in terms of characteristics such as gender and ethnic status, which are easily observable and traditionally recorded in counting systems. When characteristics are more ambiguous, there is greater error in classification. As an example of the latter, violent bias crimes, which have recently attracted substantial public attention, are subject to greater error in reporting. Gender Violence frequently crosses gender lines. Sexual assaults are disproportionally committed by males against females. Homicides by both men and women are more likely to involve male victims. Assaults, in contrast, are more likely to involve an offender and a victim of the same gender. Like assaults against women generally, cross-gender violence has more serious consequences for female victims. Readers are referred to Chapter 3 and Kruttschnitt (Volume 3) for a discussion of sexual violence and Chapter 5 for a discussion of family violence. Ethnic Status For violent crimes that involve blacks and whites, one can construct a "chance-encounter" race mix by assuming that each individual's chances of violent offending and victimization are independent of race, so that the probability of any offender-victim race combination depends only on the prevalence of each race in the U.S. population. In these hypothetical circumstances, 78 percent of all violent events would involve a white offender and victim, 21 percent would cross racial lines, and only 1 percent would involve a black offender and victim. According to the 1987 National Crime Survey (calculated from Bureau of Justice Statistics, 1989b: Table 43), in single-offender victimizations, whites assault whites at about the chance-encounter rate, blacks assault whites at about 72 percent of that rate, and whites assault blacks at about 56 percent of the chance-encounter rate. In contrast, blacks assault blacks at about 800 percent of the chance-encounter rate. Violent Bias Crimes One type of violent behavior that has been

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 78 recently defined in the criminal law is referred to as hate or bias crimes. Bias crimes are distinguished from other crimes by the presumed role of social relationships in their motivation. The Hate Crime Statistics Act of 1990 requires the UCR program to begin counting bias crimes and specifies that violent attacks, intimidation, arson, and property damage “that manifest evidence of prejudice based on race, religion, sexual orientation, or ethnicity" are all considered to be violent bias crimes. New state and local statutes that prescribe enhanced sentences for bias crimes define them in fairly similar terms. Documenting and analyzing patterns of violent bias crimes is difficult, because these crimes are sometimes hard to recognize. Absent such signals as graffiti, organizational identity, or lifestyle of victims, classifying some violent act as a bias crime makes it necessary to determine the perpetrator's motivation —a difficult task subject to uncertainty, especially when the prejudice serves only to aggravate a conventional robbery with a gratuitous shooting, beating, or mutilation, for example. The available statistics have generally been developed by advocacy organizations. Such organizations often lack the resources and infrastructure for regularly counting incidents and classifying them according to rigorous criteria but, by increasing awareness of bias crimes, they may encourage the designation of ambiguous events as bias crimes. Thus, for example, Montgomery County, Maryland, reported 196 bias crimes during 1989. This count constitutes between 14 and 81 percent of various advocacy groups' recent national counts (compiled by Ellis, 1990, for the panel), a share that is severely disproportionate to Montgomery County's 0.2 percent share of the U.S. population. Most of the available data do not distinguish between violent and nonviolent bias crimes. An exception is a synthesis of 10 available victimization surveys of gay men and lesbians (Berrill, 1990). Between 24 and 48 percent of the gay men and lesbians surveyed reported having been threatened by violence related to their sexual orientation at some time in their lives. Similarly, between 9 and 23 percent reported having been punched, hit, or kicked, and between 4 and 10 percent reported having been assaulted with a weapon. In most of the surveys, the victimization rates for gay men exceed the rates for lesbians by factors of 2-4 to 1—slightly greater than the difference by gender for assault victimization in the general population, according to the National Crime Survey (Aurand et al., 1985; Gross et al., 1988).

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 79 Personal Relationships About half of all homicide victims are murdered by neither intimate family members nor total strangers, but rather by people with some kind of preexisting relationship: friends, neighbors, casual acquaintances, workplace associates, associates in illegal activities, or members of their own or a rival gang (Table 2-6). The high prevalence of such preexisting relationships between victims and their killers suggests that most people's fears of being killed by strangers overestimate the risk; by the same token, people underestimate the probability of being killed by someone with a close or a known relationship to them. As discussed earlier in this chapter, women face only about one-third the homicide risk faced by men (4.2 and 14.2 per 100,000, respectively). However, among homicide victims, women are about four times as likely as men to have been killed by intimate partners, and 50 percent more likely to be killed by other family members. For violent crimes that do not end in death, a preexisting relationship between victim and offender is less likely, yet there is variation by type of crime. Of all nonfatal violent crime types, forcible rapes are most likely to involve intimates or acquaintances (61%), and attempted robbery is least likely (14%). COMMUNITY VARIATIONS IN VIOLENT CRIME Community characteristics affect the probability of violent crime. In this section we examine data describing variation in the rates of specific crime types over time, by community size, and from area to area within cities. Variation by Community Size The size of a community is related to its rate of violent crime. In line with many people's perceptions, smaller cities and rural and suburban areas on average have lower rates than larger cities and metropolitan areas. Figure 2-8A shows aggregate UCR violent crime rates from 1973 to 1990 for cities of different sizes. In general, the smaller the place, the lower the rate of violent crime. By 1990, the aggregate rate varied from 359 per 100,000 residents for cities with less than 10,000 population to a rate of 2,243 for those in cities with a million or more. Although not shown in the figure, outer suburban county areas not included in other groupings and rural counties were relatively safe places: in 1990, the comparable

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 80

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 81 crime rate in suburban areas was 450 per 100,000 population and in rural counties was 209—the lowest rate of all. Figure 2-8A Total violent crime rate by city size. The figure also shows that, over a 17-year period, differences in the relative safety of communities of different sizes persisted through an overall increase in violent crime reported to the police. Annual changes in violent crime rates vary from year to year by size of community, yet overall the rise in rate is greater for large places than for small ones. Outer suburban and rural areas do not follow

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 82 the pattern, however: the increase they have experienced is somewhat higher than that of cities with less than 25,000 residents. Figure 2-8B Murder and nonnegligent manslaughter rate by city size. In Figures 2-8B to 2-8E we present the rates of specific violent crimes over time for cities of different sizes. Forcible rape represents an exception to the general relationship between city size and crime rates—forcible rape rates in cities with populations between 250,000 and 999,999 exceed those in larger cities through almost the entire period. Otherwise, with occasional exceptions

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 83 for a year or two, specific crime rates are highest for large cities and lowest for small cities. Only assault and forcible rape rates have increased over time in cities of all sizes over the past 17 years. Homicide increases have been greatest in large cities, especially in recent years. In most size ranges, the rate for robbery was fairly stable, except for the increases in largest cities, especially since 1988. Just how much of these changes over time are due to changes in reporting behavior is unclear. Homicide The 15-year trends for homicide (Figure 2-8B) show roughly similar proportional fluctuations for cities of all sizes. The homicide rate rose to a peak in 1980 or 1981, then declined to a low in the mid-1980s. The rate has been rising since then, and at a somewhat greater rate for cities over than under 1 million. By 1991 the homicide rate was approaching that for 1980-1981. The perception that recent homicide rates in the 1980s are the highest ever is unfounded. Of course, as the U.S. population increased over time, the annual number of homicides grew to historically high levels even with slightly lower rates. Within any city size group, individual cities vary considerably in their annual homicide rate. Even some fairly sizable cities have no homicides in a given year. In Massachusetts in 1990, for example, 65 of the 86 cities with 10,000 or more population reporting to the UCR had no homicides. Two cities with over 100,000 residents—Scottsdale, Arizona, and Irvine, California—reported no homicides in 1990. Forcible Rape Because of UCR reporting practices, the reported rates for forcible rape are underestimated by at least a factor of two. Nevertheless, a striking feature of the forcible rape pattern is the relatively low rate for the largest cities, and its restricted range of variation by city size (Figure 2-8C). The rate for each city size group differs significantly from that of the others but, unlike the pattern for all crimes, the annual rate does not vary directly with city size: among the three largest city size groups, those with 250,000 to 499,999 population have the highest rate and those with over 1 million the lowest. Annual variation in the rate differs somewhat from that for all violent

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 84 crimes. In the 15-year period 1975 to 1990, there was a rise in the rate for all city size groups. In large cities, the fluctuations showed two peak periods followed by a decline. The short-term fluctuations around a rising secular rate in the large but not the smaller cities suggest there may be greater variation in reporting practices for the larger jurisdictions since it seems unlikely that such perturbations reflect changing local causal chains. Figure 2-8C Forcible rape rate by city size. Cities of all sizes showed an increase over time in the forcible

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 85 rape rate from 1975 to 1990. For cities with less than 250,000 population, the rate in 1990 was equal to or above that of the peak rate for the 15-year period. For the cities with 250,000 or more, the 1990 rate was still below the peak period. What is surprising is that after 1975 the rate for large metropolitan cities was significantly below that of cities between 250,000 and 1 million and remained so to 1990. Indeed, cities in the size group 250,000 to 500,000 generally showed the greatest annual increase. Although direct evidence to account for this pattern of variation is lacking, some of the rise seems attributable to changes in the willingness of women to report forcible rape to the police and to changing practices of police departments in sympathetic processing of forcible rape victims (see Chapter 3). Robbery The basic pattern of variation by city size for robbery is similar to that observed for all violent crime combined (Figure 2-8D). What is most striking is the substantial variation at the beginning and the end of the 15-year period. The range in 1990 was from a low of 49 in cities under 10,000 to a high of 1,138 in cities over 1 million. Rural counties (not included in the figure) reported a rate of only 16 per 100,000 inhabitants. What is also apparent is that the cities of more than 1 million have rates substantially above that of even the next largest cities. This difference prevailed throughout the period, and in 1990 the range was from around 600 for cities of between 250,000 to 1 million to 1,138 for cities over 1 million. Clearly, the nation's eight largest cities in 1990—New York, Los Angeles, Chicago, Philadelphia, Houston, Detroit, San Diego, and Dallas—are far more dangerous as measured by robbery incidents than are smaller metropolitan central cities on average. What may seem somewhat surprising is that, despite fluctuations in the robbery rate from 1973 to 1990 for each of the city size groups, the trend is distinctly upward only for the cities with more than 1 million and those with 100,000 to 250,000 population. Other trend lines are fairly flat. Aggravated Assault The rates for aggravated assault have increased in cities of all sizes since 1973 (Figure 2-8E). The rates peaked in 1980 for all groups, declined slightly, then rose to new highs in 1990. By 1990

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 86 the range of rates had widened substantially—from lows of 282 and 283, respectively, for cities with under 10,000 and 10,000 to 25,000 population to a high of 1,011 for cities of over 1 million. There was, on the whole, a fairly steady increase in the aggravated assault rate. Figure 2-8D Robbery rate by city size. Worth noting is the fact that the aggravated assault rate was greater in cities of 250,000 to 500,000 than in cities of 500,000 to 1 million population. Indeed, over time, the aggravated assault rate of cities in the quarter to half million size diverged from that

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 87 for other cities over 100,000. This class of cities does not generally fit the pattern of a direct relation between city size and violent crime. Why these 32 cities should experience higher aggravated assault rates than all but our largest central cities is unclear. Their rate of 842 in 1990 was well above the rate of 602 for cities of 100,000 to 250,000 and that of 621 for cities of 500,000 to 1 million. Not surprisingly, the lowest rate of aggravated assault—164 per 100,000 population in 1990—is reported for the rural counties. Figure 2-8E Aggravated assault rate by city size.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 88 Variation Within Cities Violent crime rates vary, not only according to community size and over time, but also from neighborhood to neighborhood within cities. This spatial distribution of violent crime within cities can be dramatic, equaling and sometimes far exceeding the variation across city-size groups. Where one lives within even the highest crime rate cities affects one's risk of violent victimization more than the size of the city. Figure 2-9 is a map of census tracts in the city of San Diego, indicating the rate of violent crime in each tract during the second quarter of 1991. While 7 census tracts recorded less than 1 violent crime per 1,000 population, 1 reported 300. Several clusters of adjacent San Diego tracts have roughly the same crime rates. Not surprisingly, higher rates generally occur in the central business district; however, these rates are difficult to interpret because the population of residents is smaller than the population of people at risk, who include many who commute there from outside the city or from other areas of the city. Rates are also high in low-income residential communities compared with those in above- average income communities. There is considerable evidence that residents perceive crime in terms of its spatial distribution and community organization (Warr, Volume 4). People's perceptions of the relative safety of different neighborhoods largely match the reality, leading them to prefer some neighborhoods to others and to avoid stopping or staying in areas they perceive as violent. They are also aware that in daily living the risks of victimization change as they move from one area to another—from home to work, from home to play, or obtaining business or professional services (see Figure 2-10). DATA NEEDS Examination of the systems for counting and classifying violent behavior leads to the inescapable conclusion that our present official and unofficial systems of counting and classifying violent criminal behavior result in a substantial underestimation of both the kind and the amount of violent behavior. Our official systems for collecting information on crimes largely fail to collect information on intrafamily violence. The systems also undercount most nonfamily crimes in which the victim survives. There appears to be less underestimation for armed robbery than for assaults. Both sexual assaults and interpersonal assaults are substantially underreported in our official systems as evidence from

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 89 Figure 2-9 San Diego violent crime rate by census tract, 1991. Source: San Diego Police Department Crime Analysis Unit.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 90 emergency treatment, other institutional records, and small sample surveys of selected populations (e.g., surveys of men and women for a city or college) make apparent. Figure 2-10 Contour map of perceptions in Summit County, Ohio. Source: Pyle (1980). As reflected in this chapter, there is a paucity of information about victimization risks for ethnic categories other than whites and blacks. The panel recommends that UCR, NCS, and NCHs explore and, to the extent possible, expand their capacities for collecting and reporting data for Hispanics, Asians, Native Americans, and other ethnic categories. There is also a paucity of information on the consequences of victimization by violent crime. The NCS collects information on injury and property loss, but as Cohen et al. (Volume 4) explain, the information is insufficient to assess the overall cost consequences

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 91 of violent crime. Moreover, there are few ways at present to merge information from diverse sources as, for example, from emergency or other medical treatment and crime victim data bases. The inability to merge data sources has implications not only for calculating the consequences and costs of crime but also for calculating conditional probabilities in the causal chain of crimes and their consequences. This lack of baseline conditional probabilities complicates the task of assessing the efficacy of interventions to prevent violent crimes or meliorate their consequences. Available information on violent bias crimes lags far behind the pace of legislative initiatives in this area. In implementing the Hate Crimes Statistics Act of 1990, high priority should be given to supporting research on the following questions: (1) how various punishment schemes (e.g., increments to basic sentences) affect the incidence of violent bias crimes; (2) the psychological consequences of bias crime victimization; (3) individual risk factors for victimization among members of groups that are frequent objects of hate crimes; and (4) validation of hate violence counts obtained through the FBI's Uniform Crime Reporting Program and analytical comparisons with the counts obtained from other sources, such as advocacy groups and community-based organizations. This chapter has pointed to various limitations of current information systems for measuring the national prevalence and incidence of violent crimes and their consequences, and for the victims of crimes. Those limitations imply a large number of recommendations, but the panel wishes to draw attention to five important information system needs at the local level to enhance our understanding of violent behavior and of interventions to control it. (1) There is a pressing need for statistical information systems that link information on injuries and their treatment to information about the violent event. The development of emergency and trauma center information to make it relevant to violent crime should be a high priority. In addition, more information could be obtained from current hospitalization and medical treatment surveys, and more detail on the extent of injury could be recorded in NCS and in the new incident-based UCR program. (2) There is an equally pressing need to develop information systems that will permit the calculation of probabilities of violent victimization conditional on circumstances. That information is useful public information for potential victims in assessing their personal risks, for researchers attempting to understand violent events, and for policy makers assessing the efficacy of

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 92 interventions to prevent violent crimes or meliorate their consequences. (3) Present information systems provide only limited information on many types of violent victimization in which there is considerable public interest and which are the major focus of public policy. These include intrafamily violence, violent sexual assaults of men, women, and children, and the persons confronted and occasionally killed in the course of commercial robbery. (4) Information on the consequences of violent crime and their costs is limited. There is almost no information on their cost to organizations and individuals. Efforts should be made to develop that information. At the present time there are no reliable ways to estimate the cost of the psychological consequences of crime such as fear of being victimized or the physical and psychological stress victims experience from past victimization. In addition, as Cohen et al. (Volume 4) conclude, there is little information on the nature and cost of victim services or of medical care. (5) There is a need for more reverse-record checks and other comparative analyses of events that are captured in systems for counting violent crimes, victimizations, injuries, and deaths. Such studies will facilitate more accurate estimates of the undercounts of violence inherent in each system. NOTES 1UCR and NCS count only forcible rapes (rape by force), excluding statutory rapes (Federal Bureau of Investigation, 1974:12; Bureau of Justice Statistics 1992: Glossary, p. 156). 2 Like the UCR, the NCS counts only the most serious event as an incident as defined by its hierarchy rules. Therefore, the assault of a victim during a robbery will be counted as only a robbery victimization. Neither system, consequently, offers an estimate of all violent crimes committed during an event. 3The UCR program is being changed to an incident-based reporting system, which will provide more information on victims and offenders and will also eliminate the hierarchy rule that counts only the most serious crime in a compound event. 4The NCS classifies all victimizations as either personal- or household-sector victimizations. Burglary, household larceny, and motor vehicle theft are household-sector victimizations. Personal-sector victimizations are classified as crimes of violence (rape, robbery, and assault) or crimes of theft (personal larceny with contact and personal larceny without contact). NCS violent victimizations are only roughly comparable to UCR index crimes of violence, for reasons discussed earlier in this chapter. 5 The calculation excludes an estimated 3,128,130 simple assaults

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 93 for crude comparisons with UCR. Their exclusion reduces the number of violent victimizations from 6,008,790 to 2,880,660. Aggravated assaults are attacks by one person on another for the purpose of inflicting severe or aggravated bodily injury. Attempts are included since it is not necessary that an injury result when a gun, a knife, or another weapon, including hands, fists, and feet, are used that could, and probably would, result in serious personal injury were the crime successfully completed. Both UCR and NCS classify injuries as serious when they result in broken bones, lost teeth, internal injuries, and loss of consciousness. The NCS also classifies any injury as serious if it requires two or more days of hospitalization. 6 Personal larcenies or personal crimes of theft include two subgroups. Personal larceny with contact involves personal contact between the victim and the offender and includes such crimes as purse snatching and pocket picking. Personal larceny without contact is theft of property of the victim without personal contact from any place other than the home or its immediate vicinity. The crime differs from household larceny only in the location in which the theft occurs. 7The panel recognizes that the division between simple and aggravated assault involves considerable classification error and also that there are valid reasons for including simple assault as a violent crime, especially given its prevalence in domestic violence. 8UCR violent index crimes are offenses of murder, forcible rape, robbery, and aggravated assault (Federal Bureau of Investigation, 1990: Table 1). 9The inclusion of commercial and nonresidential offenses in the UCR but not the NCS also accounts for some of the difference, as it disproportionally increases the base for property crimes. 10 Because the Bureau of Justice Statistics (1989a) report is based on the National Crime Survey, it counts injuries only among surviving victims. Because death occurs in only about 1 percent of all victimizations involving injury, however, this statistical artifact has a negligible effect on the statistics reported here. 11 Loftin and Wiersma (1991) could not calculate risks for Hispanics because, at the time their data were collected, death certificates in only five states provided for that demographic category. 12Single victim-single offender homicides account for only 54 percent of all homicides in which the ethnic status of victims was reported in 1990 (calculated from Federal Bureau of Investigation, 1990:11). 13 Age adjustment is a technique that applies the age-specific rates of a population to a standardized age distribution to eliminate the difference in observed rates that result from differences in population composition. For example, whites and females have lower infant mortality and longer life expectancies than blacks and males. Therefore, blacks and males are overrepresented in the ages of highest risk for criminality—a fact that would overstate gender/ ethnic status homicide mortality differentials if no adjustment were made. Adjustment for age is usually done

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 94 to compare two or more populations at the same point in time or one or more populations at two or more points in time. 14 A commonly used measure of premature mortality due to a given cause of death is years of potential life lost (YPLL) before age 65. This measure is useful for comparing different causes of death in terms of premature mortality. However, blacks and males have shorter life expectancies than do whites and females, even when homicides are omitted from the calculation. Therefore comparisons by race and gender of YPLL before age 65 due to homicide overstate the relative premature mortality effects of homicide on blacks and males. 15 The reader should bear in mind that the young male adult military population exposed to battle deaths is also that most at risk of homicide. Hence the homicide rate for the resident U.S. population in war years will be below that expected for years when there are no battle deaths (Federal Bureau of Investigation, 1969: Table 1). 16 Until 1990, the NCS had reports of fewer than 10 sample cases of sexual assaults on males classified as forcible rape. The 1990 rate was reported as 0.2 per 1,000 males. There is a substantial underreporting of sexual assault on both males and females, but especially so for males at younger ages. Efforts should be made to secure more reliable measures of sexual assaults for both men and women at all ages. 17 Family income in the NCS includes the income of the household head and all other related persons residing in the same housing unit. The income of persons unrelated to the head of household is excluded (Bureau of Justice Statistics, 1990: Glossary). 18 Numerators are counts of violent crime arrests in Federal Bureau of Investigation (1992: Table 38). Denominators are estimated populations, 12 years of age and over by race in Bureau of Justice Statistics (1992: Table 6). References Aurand, S.K., R. Adessa, and C. Bush 1985 Violence and discrimination against Philadelphia lesbian and gay people. (Available from Philadelphia Lesbian and Gay Task Force, 1501 Cherry Street, Philadelphia, PA 19102). Berrill, K.T. 1990 Anti-gay violence and victimization in the United States: An overview. Journal of Interpersonal Violence 5(3, September):274-294. Blumstein, A., J. Cohen, J.A. Roth, and C.A. Visher 1986 Criminal Careers and “Career Criminals." Vol. I. Washington, D.C.: National Academy Press. Bureau of the Census 1990 Statistical Abstract of the United States: 1990. Washington, D.C.: U.S. Government Printing Office.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 95 Bureau of Justice Statistics 1989a Injuries From Crime: Special Report. Washington, D.C.: U.S. Government Printing Office. 1989b Criminal Victimization in the United States, 1987. Washington, D.C.: U.S. Government Printing Office. 1990 Criminal Victimization in the United States, 1988. A National Crime Survey Report, December 1990, NCJ-122024. 1992 Criminal Victimization in the United States, 1990. Washington, D.C.: U.S. Government Printing Office. Ellis, W.W. 1990 Bias Crime. Commissioned paper for the Committee on Research on Law Enforcement and the Administration of Justice, National Research Council. Farrington, D.P. 1991 Childhood aggression and adult violence: Early precursors and later-life outcomes. Pp. 5-29 in D.J. Pepler and K.H. Rubin, eds., The Development and Treatment of Childhood Aggression. Hillsdale, N.J.: Erlbaum. Farrington, D., H. Snyder, and T. Finnegan 1988 Specialization in juvenile court careers. Criminology 26:461-488. Federal Bureau of Investigation 1969 Uniform Crime Reports: Crime in the United States, 1968. Washington, D.C.: U.S. Government Printing Office. 1974 Uniform Crime Reporting Handbook. Washington, D.C.: U.S. Government Printing Office. 1990 Uniform Crime Reports: Crime in the United States, 1989. Washington, D.C.: U.S. Government Printing Office. 1991 Uniform Crime Reports: Crime in the United States: 1990. Washington, D.C.: U.S. Government Printing Office. 1992 Uniform Crime Reports: Crime in the United States, 1991. Washington, D.C.: U.S. Government Printing Office. Garofalo, J., L. Siegel, and J. Laub 1987 School-related victimizations among adolescents: An analysis of national crime survey narratives. Journal of Quantitative Criminology 3:321-338. Griffin, Ezra E.H., and Carl C. Bell 1989 Recent trends in suicide and homicide among blacks. Special communication. Journal of the American Medical Association 282(16):2265-2269. Gross, L., S. Aurand, and R. Adessa 1988 Violence and discrimination against lesbian and gay people in Philadelphia and the Commonwealth of Pennsylvania. (Available from Philadelphia Lesbian and Gay Task Force, 1501 Cherry Street, Philadelphia, PA 19102) Gurr, T. R. 1989 Historical trends in violent crime: Europe and the United States.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 96 In T. R. Gurr, ed., Violence in America. Vol. 1: The History of Crime. Newbury Park, Calif.: Sage Publications. Haenazel, W. 1950 A standardized rate for mortality defined in units of lost years of life. American Journal of Public Health 40:17-26. Hindelang, M., M. Gottfredson, and J. Garofalo 1978 Victims of Personal Crime: An Empirical Foundation for a Theory of Personal Victimization. Cambridge, Mass.: Ballinger. Holinger, Paul C. 1987 Violent Deaths in the United States: An Epidemiologic Study of Suicide, Homicide, and Accidents. New York: Guilford Press. Indian Health Service 1988 Regional Differences in Indian Health. Washington, D.C.: U.S. Department of Health and Human Services. 1991 Regional Differences in Indian Health. Washington, D.C.: U.S. Department of Health and Human Services. Jencks, Christopher, and Susan E. Mayer 1990 The social consequences of growing up in a poor neighborhood. Pp. 111-186 in Laurence E. Lynn, Jr. and Michael G.H. McGeary, eds., Inner-City Poverty in the United States. Washington, D.C.: National Academy Press. Koppel, Herbert 1987 Lifetime likelihood of victimization. Bureau of Justice Statistics Technical Report NCJ-104274. U.S. Department of Justice. Lane, Roger 1979 Violent Death in the City: Suicide, Accident, and Murder in Nineteenth Century Philadelphia. Cambridge, Mass.: Harvard University Press. Loftin, Colin, and Ellen J. MacKenzie 1990 Building National Estimates of Violent Victimization. Draft paper presented at the Symposium on the Understanding and Control of Violent Behavior. Destin, Fla. April 1-4. Loftin, Colin, and Brian Wiersma 1991 Lifetime Risk of Violent Victimization from Homicide. Unpublished memo to the Panel on the Understanding and Control of Violent Behavior. Lynch, James P. 1989 An evaluation of lifetime likelihood of victimization. Public Opinion Quarterly 53:262-264. Maxfield, M.G. 1989 Circumstances in Supplementary Homicide Reports: Variety and validity. Criminology 27(4):671-695. Monahan, J. 1988 Risk assessment of violence among the mentally disordered: Generating useful knowledge. International Journal of Law and Psychiatry 11:249-257.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 97 Monkkonen, Eric H. 1989 Diverging homicide rates: England and the United States, 1850-1875. Pp. 80-101 in T.R. Gurr, ed., Violence in America. Vol. 1: The History of Crime. Newbury Park, Calif.: Sage Publications. National Center for Health Statistics 1991 Vital Statistics of the United States 1988. Volume II: Mortality. Washington, D.C.: U.S. Government Printing Office. Piper, E. 1985 Violent recidivism and chronicity in the 1958 Philadelphia cohort. Journal of Quantitative Criminology 1:319-344. Pyle, G.F. 1980 Systematic sociospatial variation in perceptions of crime location and severity. Pg. 226 in D. Georges-Abeyie and K.D. Harries, eds., Crime: A Spatial Perspective. New York: Columbia University Press. Reiss, A.J., Jr. 1985 Some failures in designing data collection that distort results. Pp. 161-177 in L. Burstein, H.E. Freeman, and P.H. Rossi, eds., Collecting Evaluation Data: Problems and Solutions. Beverly Hills, Calif.: Sage Publications. Reiss, A.J., Jr., and D.P. Farrington 1991 Advancing knowledge about co-offending: Results from a prospective longitudinal survey of London males. Journal of Criminal Law and Criminology 82(2):360-395. Sedlak, Andrea J. 1991 National Incidence and Prevalence of Child Abuse and Neglect: 1988. Washington, D.C.: Westat, Inc. (Revised September 5, 1991). U.S. Department of Health and Human Services 1990 Health United States, 1989. Hyattsville, Md.: Public Health Service. van Dijk, Jan J.M., Pat Mayhew, and Martin Killias 1990 Experiences of Crime Across the World: Key Findings from the 1989 International Crime Survey. Deventer, The Netherlands: Kluwer Law and Taxation Publishers.

PATTERNS OF VIOLENCE IN AMERICAN SOCIETY 98

Next: PART II UNDERSTANDING VIOLENCE »
Understanding and Preventing Violence: Volume 1 Get This Book
×
Buy Paperback | $75.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

By conservative estimates, more than 16,000 violent crimes are committed or attempted every day in the United States. Violence involves many factors and spurs many viewpoints, and this diversity impedes our efforts to make the nation safer.

Now a landmark volume from the National Research Council presents the first comprehensive, readable synthesis of America's experience of violence—offering a fresh, interdisciplinary approach to understanding and preventing interpersonal violence and its consequences. Understanding and Preventing Violence provides the most complete, up-to-date responses available to these fundamental questions:

  • How much violence occurs in America?
  • How do different processes—biological, psychosocial, situational, and social—interact to determine violence levels?
  • What preventive strategies are suggested by our current knowledge of violence?
  • What are the most critical research needs?

Understanding and Preventing Violence explores the complexity of violent behavior in our society and puts forth a new framework for analyzing risk factors for violent events. From this framework the authors identify a number of "triggering" events, situational elements, and predisposing factors to violence—as well as many promising approaches to intervention.

Leading authorities explore such diverse but related topics as crime statistics; biological influences on violent behavior; the prison population explosion; developmental and public health perspectives on violence; violence in families; and the relationship between violence and race, ethnicity, poverty, guns, alcohol, and drugs.

Using four case studies, the volume reports on the role of evaluation in violence prevention policy. It also assesses current federal support for violence research and offers specific science policy recommendations.

This breakthrough book will be a key resource for policymakers in criminal and juvenile justice, law enforcement authorities, criminologists, psychologists, sociologists, public health professionals, researchers, faculty, students, and anyone interested in understanding and preventing violence.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!