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.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 404 B Measuring and Counting Violent Crimes and Their Consequences People often ask, "How much crime is there?" "How many violent crimes are committed each day, month, or year in the United States?" as if these questions admitted of some reasonably precise answer. They do not, and perhaps they cannot, because of the institutional nature of counting. Any counting system involves some valuative institutional processing of people's observations and reports of what they perceive as events. Any set of crime statistics, therefore, is not based on some objectively observable universe of behavior. Rather, violent crime statistics are based on the events that are defined, captured, and processed as such by some institutional means of collecting and counting crimes (Biderman and Reiss, 1967). There is consequently no single way to define, classify, and measure the domain of violent behavior or its subset of behavior that constitutes violent crimes. Rather, there are a multiplicity of ways of doing so, each of which has its own counting and classification errors. The institutional nature of counting requires that we have multiple means of counting, each of which compensates for some of the errors in classification and counting in the others. At best, however, a national system of crime counts produces multiple and divergent estimates for which only some errors in measurement are estimated. As we probe the undercounted domain of violent crime, it is apparent that current institutional means of counting and accounting do not cover satisfactorily the kinds of violent behavior that are required for its understanding, and that the domain
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 405 of violent crime is not covered by any current institutional means of counting. New means of counting are not easily compared with existing systems, because each has its own rules for selecting, classifying, and counting crime events. The example of parental assault of children illustrates some of the problems. If we try to estimate how many and how frequently parents assault their children, neither Uniform Crime Reports (UCR) nor the National Crime Survey (NCS) can currently supply accurate information. Both systems rely primarily on family members for their information. At present they rarely receive information on parental assaults on children, but if either were to attempt a systematic count of parental assaults, they would encounter substantial underreporting. Reports to police involve self-incrimination for the assaulting parent(s), making it unlikely that the violence would be reported in single-parent families or in families in which both parents are assaultive. Likewise, the parent who does not commit the violence may fail to report the incident because of intimidation or the fear that it will jeopardize the family economy or unity. In addition, there are problems of separating peer and sibling assaults from parental assaults and of assault from accidental injuryâformidable problems to accurate classification and counting of such events. And counts of parental assault obtained by means independent of the UCR and NCS estimates are not merged with them to obtain a more comprehensive estimate of assaults. The current institutional system of acquiring information on physical assaults by parents relies primarily on reporting by professionals who have contact with children. The school and the medical clinic or office are the main places for these contacts by physicians, nurses, and teachers. Preschool children are the least likely to be observed by professionals. Their contacts with medical personnel and social workers are sporadic. Nonmedical professionals, moreover, rely on superficial visual examination, which risks missing evidence of sexual assault, for example. The observational injury data supplied by teachers or medical personnel do not provide information to indicate if the injury is the result of violent behavior, if the behavior constituted a criminal act by a parent, or even whether a parent is the source of the injury. Establishing these facts requires investigation by others, an investigation that generally must rely on home visits and testimony either by parent and child or by persons in a position to observe the family's behavior. Moreover, many of these observations of injury are investigated and institutionally processed through
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 406 social rather than police service organizations. Hence, their criminal status remains undetermined. Although this loosely structured system increases the chances of detecting, counting, and processing injuries of children as the result of violent behavior, the status of many remains indeterminate and only the most serious cases are detected and processed as violent crimes. Any system for counting parental physical assaults that respects parental rights, is loosely structured around professionals' contact with children, and relies on visual evidence that must be verified by still others is likely to encounter serious problems of misclassification and underreporting. Social Constructions of Violent Behavior And Violent Crimes It is no simple matter to determine which violent behaviors are to be regarded as violent crimes for purposes of measurement and counting. What constitutes a violent crime is the result of societal determinations of what is violent behavior and which violent behavior is violent crime. Social constructions of violent behavior are dynamic and change both temporally and organizationally. Social constructions of violent behavior that are institutionalized in the criminal law as violent crimes can be institutionalized differently in other systems, such as state systems that process behavior or crimes and their accounting systems (e.g., family courts or mental health systems). Whether violent acts will be institutionally processed as crimes depends on the processing of events by legal agents empowered to make decisions to arrest, charge, prosecute, and adjudicate criminal matters. Not all social constructions of violent behavior and violent crime are formal; they emerge from informal processes as well. The domain of violent behavior includes some behaviors that are not legally and institutionally processed as violent crimes. That is, not all violent behavior is socially construed as criminal. The physical punishment of children, for example, may be regarded as discipline rather than as violent behavior on the part of the disciplinarian. Similarly, most violence in competitive contact sports such as ice hockey is regarded as acceptable. Only rarely does intentionally inflicted harm in sports fall within the sphere of violent crime, even when it results in death. Perhaps the most highly institutionalized acceptance of violent behavior in a sport is the prize fight. What constitutes an incident of violent behavior or violent crime is sometimes subject to the discretion of decision makers. Some
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 407 types of violent behavior that laws have recently criminalized lie at the margin of institutional processing as violent crimeâdate rape is an example. Other forms of coerced sexual behavior may lie outside the criminal law, as is currently the case in most states with marital rape. Recording of assaults involving custodians and their wards, in prisons and mental hospitals, for example, is often discretionary and fragmentary. Domestic Assaults The significant changes in the last 40 years in the role of women in society have influenced social constructions of domestic assault. Historically, the police dealt with reported domestic assaults on women as crimes only if the bodily injury was serious enough to qualify as an aggravated or felony assault. As a consequence of some major tort liability suits against police departments and the results of the experiments in Minneapolis (described in Chapter 7) concluding that arrest was a deterrent to repeated domestic assault (Sherman and Cohn, 1989), there has been a substantial change in their reporting. Eleven states now mandate arrest in all cases of domestic assaultsâ misdemeanor as well as felony. Mandatory arrest policies appear to affect the determination of the threshold between a simple (misdemeanor) assault and an aggravated (felony) assault. Most important, however, mandatory arrest substantially increases the official reporting of domestic violence as it reduces police discretion to make threshold decisions. In the state of Connecticut, for example, the official reporting of domestic assaults more than doubled following the passage of a mandatory arrest law. Unfortunately, we cannot determine just how much of the recent rise in the assault rate nationwide to attribute to an increase in domestic violence reporting. Our systems of national reporting do not adequately classify and count repeat incidents of domestic violence by the same and different offenders. It seems reasonable to conclude, nevertheless, that changes in the social construction of domestic violence and police response to such violence account for some of the increase in the reported number of assaults and domestic assaults. Bias Crimes Another recent change in social construction involves bias crimes. A minority of states have enacted bias crime legislation; among the first to do so were Maryland, Pennsylvania, and Connecticut.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 408 The Hate Crimes Statistics Act of 1990 mandates development of a national reporting system for bias crimes. Statistics on bias crimes committed in 1990 or earlier are available from only a few of the police reporting agencies in UCR. The statistics that are available are collected by national organizations such as the National Gay and Lesbian Task Force Policy Institute and B'nai B'rith. The U.S. Department of Justice also collects information on grand jury indictments and convictions involving the criminal violation of the civil rights of ethnic minorities. The Civil Rights Division of the department collects information on complaints against police. The National Gay and Lesbian Task Force Policy Institute has collected information on statistics from gay service organizations in six major metropolitan areas and reports a substantial increase from 1989 to 1990 in these metropolitan areas. Just how much of this increase is due to a greater awareness and willingness of victims to report their victimization is undetermined because the information is not available. What is clear is that the Federal Bureau of Investigation, the U.S. Attorneys, and local police departments have shown increased willingness to bring indictments and prosecute bias crimes under federal civil rights statutes. Sexual Assaults The social construction of sexual assaults has undergone considerable change in part as a result of efforts by the women's and gay rights social and political movements. Two effects are particularly noteworthy: the redefinition of rape and other sexual assaults and the redefinition of their reporting. Although the effects of the redefinition of rape and its reporting to the police are still being assessed, it appears that our society's conception of rape has changed in several respects. First, social movements have been reasonably successful in removing the social stigma surrounding the reporting of rape. Of special importance is the emphasis now placed on ensuring that women who have been assaulted will be treated with dignity and sensitivity by law enforcement officials. Whenever possible, initial investigations of sexual crimes against women that require the victim's cooperation are handled by female police officers. Social support groups now aid women during this process to prevent their being socially stigmatized during or after its reporting. The second major change involves what constitutes a reportable sexual assault. There now is a concept of date rape and of
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 409 the moral obligation to report all coerced as well as completed nonconsenting sexual intercourse as rape regardless of the victim's relationship to the offender. In years past, few rapes or attempted rapes that occurred on a date were reported because of the youth of the women involved and the social stigma occasioned through its reporting. Women are now more likely to define themselves as victims when coerced into sexual intercourse while on a date. With this change has also come a redefinition of the necessity for equal consent to sexual relations and a redefinition of coercion. This redefinition of what constitutes rape or sexual assault undoubtedly has a greater impact on self-report surveys of sexual assault victimizations than on incidents of rape reported to the police. A number of surveys report a substantial prevalence rate of rape by an acquaintance or friend. The stigmatization of males who report being victims of sexual assault is also attenuatingâleading to increased reporting of male rape in the National Crime Survey. Some states also have forged a redefinition of rape at law. Historically, the law of evidence for conviction on a charge of rape required two independent sources of evidence that it had occurred. Typically, this required some evidence of coerced penetration or the recovery of semen, procedures often experienced as degrading by the victim. Now some states require only the sworn testimony of the victim for conviction, providing there is no evidence to impeach that testimony; the victim's moral integrity is ordinarily not permitted as grounds for impeachment. The nature of these changes is to redefine what behavior comprises rape, to ensure that the victim will be treated with respect when reporting the rape, and to encourage greater victim reporting by reducing the stigma of victimization. Assaults on Children With a growing concern for child abuse and neglect, there is increased reporting of simple and aggravated assaults of children. Special attention has been given to child sexual abuse (Bolton et al., 1989). Substantial changes in its reporting as well as redefinition of permissible behavior toward children are under way. Other major changes are affecting the status of children as victim witnesses. Recent case law, for example, now permits children who are victims of sexual and other forms of abuse to testify under certain circumstances without having to submit to cross-examination or to give testimony in the presence of the accused.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 410 Systemic Influences: Drug Markets Systemic changes in recent decades appear to have affected the composition and volume of violent crime in ways that raise new counting and classification problems. The most noteworthy of these is the growth of illegal drug markets and the population of drug users. The growth of drug markets is well documented, although it is difficult to estimate the size of local markets (Johnson et al., 1990). The violent crime of armed robbery has increased with the demand for drugs in a cash market. Robbery and burglary have become the principal means of meeting this individual demand for drugs. Of the two, robbery is preferable in that it provides ready cash for their purchase (Cromwell et al., 1991; Rengert and Wasilchick, 1989). The increase in armed robbery may also have increased the felony murder rate somewhat.1 A second major systemic change is in the territorial organization of drug markets. There has been a proliferation of small illegal drug markets, often organized around a particular illegal drug. Until recently, illegal markets were controlled by a syndicated structure that used violence only sparingly to control the size and scope of its market. Normally the markets were stabilized by explicit or tacit agreements with other syndicates; violence was used only occasionally, when such agreements broke down. Traditionally illegal market syndicates had less need for violence to control their territory because they could co-opt and corrupt the police to permit their operations. These syndicate conditions do not prevail in current urban drug markets. The marketing of at least some drugs requires different ethnic and national connections and a local territorial base, so that small syndicates have arisen around these connections. The marketing of drugs is somewhat disorganized, then, because it is fragmented and lacks central coordination. The presence of valuable drugs and large amounts of cash is an incentive for robberies in and around drug markets, but practice varies about whether to classify these events as robberies or drug related. Similarly, violence is a principal means for opening new markets, expanding the scope of existing local markets, and defending a small local market against encroachment from other entrepreneurs. That violence is armed, often with high-calibre weapons that are particularly deadly. In drive-by shootings to disrupt a local market, discharges from a automatic weapons injure or kill some bystanders (Sherman et al., 1989), but because such victimizations are not premeditated, their classification as crimes, or perhaps as accidents, is problematic.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 411 Measuring and Counting Violent Crimes Statistics about violent crimes and their victims are usually reported in terms of incidentsâspecific criminal acts or behaviors involving one of more victims in a single event. An offense is specific behavior that is legally prohibited; the major violent offenses are homicide, rape, robbery, and assault. The perpetrator of the offense is the offender; if more than one is involved in an offense, they are co-offenders. The victim of an offense can be a person, a household, an organization, or the publicâbut only persons are regarded as victims of violent crimes. The newspaper accounts of homicides in New York City that began this report illustrate some of the complexities involved in counting crime incidents. A single event or situation may result in several different types of violent crime. Heriberto Altreche, the son who beat his mother and sister, was arrested for an assault for punching his sister and, depending on the results of the autopsy, may also be held liable for the homicide of his mother. In the shootings at the Bedford Park Christmas party, the 34-year-old man was a homicide victim, and the 28- year-old man wounded in the shooting was undoubtedly a victim of an aggravated assault. An unreported number of other people were also shot at the party but fled before police arrived. Similarly, the number of offenders in a single incident can also vary. Two men dressed in black clothing and hoods were involved in the killing of Mr. Rodriguez, but neither will contribute to statistics on murder arrests unless police have evidence as to which one or both fired the fatal shot. A crime incident can also be described in terms of the place and circumstances of its occurrence and the consequences to offenders or victims, such as injury or arrest. When the incident includes more than a single victim or offender, not all offenders may have committed the same offense and not all victims may have experienced the same victimizationsâall of which can be separate counts. This variation in the complexity of crime incidents is the basis for differences in crime counts among classification and counting systems for violent crimes. Four Different Systems The counting of violent behavior in our society is organized around four independent systems, each with its own constructions of which violent behavior is to be measured and counted as violent crime.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 412 (1) Violent crimes reported to law enforcement agents and their investigation are the basis of the Uniform Crime Reports system of classifying and counting crimes. (2) Deaths from violent causes processed by medical examiner's offices, which also report the cause of death, are the basis of the National Center for Health Statistics (NCHS) death registration system. (3) Sample surveys of a population to secure information for some interval of time on criminal victimizations are the basis for the National Crime Survey system of classifying and counting crime; some information can also be obtained from specialized sample surveys such as the Survey of Occupational Injuries and Illnesses and the National Hospital Discharge Survey, but these ordinarily are not used for counting crime incidents. (4) Information systems infer violent crime from the consequences of violent behavior; the most common are injury surveillance systems based on reports of persons seeking compensation for injury or seeking treatment in emergency and other facilities. These are all voluntary reporting systems, and only the first three compile national statistics. None of them is developed solely for the purpose of counting violent crime, and none is developed to measure and count all forms of violent behavior. Only the UCR and the NCS are organized exclusively around measuring and counting crimes and some of their consequences, and each has classification and counting rules that limit what is counted as a crime of violence. Information on homicides derives primarily from the UCR system and the Office of Vital Statistics reports on deaths. None of these systems covers all forms of violent crime, although UCR has the most comprehensive coverage. As summarized in Chapter 2, differences among the systems derive primarily from the population covered, the means for collecting information, and the rules for classifying and counting behavior. The UCR system traditionally classified and counted crime occurrences or events in terms of only the most serious offense committed in an incident and arrests of offenders within the incident.2 The Federal Bureau of Investigation is moving to a system in which at least the larger police agencies will use a National Incident Based Reporting System (NIBRS) that will classify and count all offenses, offenders, and victims in incidents and their surrounding circumstances. The National Crime Survey counts victimizations from crime and their circumstances with the minimal information victims can report about their offenders. One of the reasons for developing
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 413 the National Crime Survey was that a great many violent crimes are not reported to the police, and hence a substantial proportion of violent crimes are not measured and counted in UCR. The NCS provides another institutional means for measuring and counting violent crime by securing reports of victimizations from a national sample survey of households and their members who are age 12 and older. Consider classification and counting for an incident in which an offender attempts an armed robbery of two victims, shooting both as they flee, killing one, and subsequently being arrested. The UCR would count this as a murder under its hierarchical rule of counting the most serious offense; the offense would be cleared by the arrest; a Supplemental Homicide Report would be filed on the victim's being killed in a felony homicideâthe robbery. There would be no reporting of the assault on the second victim nor of the attempted robbery of the two victimsâinformation that will now be reported as part of the NIBRS. The NCS would obtain information on the incident only from the surviving victim and count an attempted robbery and an aggravated assault for the surviving victim, weighting the robbery by one-half for the presence of a second victim. There would be no count of the homicide. Reconciling UCR and NCS Counts As we illustrate in Chapter 2, when the trends for NCS and UCR are compared, they produce divergent results. In recent years this difference has generated considerable discussion in the popular media, Congress, and research literature (see Blumstein et al., 1991; Jencks, 1991). One reason offered for the difference between the two data seriesâ essentially no change in NCS offenses while UCR offenses have increased, especially during the 1980sâis that the rate at which victims report crimes to the police has increased. Yet NCS data on victim reporting rates to the police have been remarkably stable, remaining between 50 and 60 percent for robbery and aggravated assault throughout the period 1973 to 1989 (Jencks, 1991). Moreover, reporting rates are not a factor in accounting for year-to-year variations in UCR crime counts (Blumstein et al., 1991). Jencks (1991:103) offers a more cogent explanation for the difference between NCS and UCR trends: â[T]he police are recording more of the violence that citizens report to them. In 1973, for example, citizens reported about 861,000 aggravated assaults to the police. The police recorded only 421,000. By 1988 citizens
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 414 said they reported 940,000 aggravated assaults to the police, and the police recorded 910,000. The same pattern recurs for robbery and rape." The stark increase from 49 to 97 percent of victim-reported assaults being recorded by the police provides compelling evidence that changes in police recording practices account for differences between NCS and UCR trends. It is possible that changes in citizen willingness to report crimes to the police might account for some of the observed difference, but it seems unlikely that such changes would fit the observed pattern in reporting. The same distinctive upward trend in police recording rates characterizes violent and property crimes alike. The trends for the various crime types, however, suggest the need for caution in relying entirely on the police recording hypothesis. By 1989, police recording ratesâcalculated from the ratio of UCR crime counts to NCS counts of offenses reported to the policeâexceeded 100 percent. For burglary, the recording rate has continued to rise steadily from 100 percent since 1981, and the recording rate for rape is highly variable and often well above 100 percent. If police crime counts are indeed becoming more accurate through more complete recording of citizen-reported offenses, then the trend in recent years for UCR crime counts to exceed NCS counts of reported crimes raises the possibility of crime undercounts by the NCS. Victimization of populations that are typically underrepresented in NCS samplesâespecially transient, homeless, and offender populations, all of whom are at substantially greater risk of violenceâmay be an important factor in NCS crime undercounts. The extent to which these special populations have increased over time or become more difficult to survey would contribute to an absence of upward trends in NCS counts of violent crimes. Information About Offenders and Offending The current systems of collecting information on violent crimes provide substantially more information on the violent crime incidents and their victims than they do on offenders and their patterns of offending. Both UCR and NCS provide only limited information on offenders. The only other major source of information on offenders is the National Prisoner Statistics. UCR currently provides information only on arrested offenders, and it is limited. Information is provided on ethnic status, age, and sex of arrestees and the most serious crime for which arrested.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 415 Until the incident-based NIBRS reporting system is fully implemented, ethnic status-sex-age distributions cannot be obtained. Moreover, the same offender may be arrested more than once in a given year, so that it is impossible to calculate prevalence rates for arrest or to determine the characteristics of an offender population. The information available amounts to a population of arrests and the characteristics of persons arrested. Arrest counts are available by size and type of jurisdiction and place, but this is an extremely limited data base. The NIBRS system will provide considerably more information on arrested persons, such as whether or not they were armed or had drugs in their possession. It remains unclear how soon it will be possible to calculate national rates based on that information. The Supplementary Homicide Report provides somewhat more information on persons arrested for homicide and permits a determination of the characteristics of multiple offenders in homicides. The information is quite limited and incomplete on both the homicide and the offender. For example, the relationship of the offender and victim was unknown in one-third of all homicides (Federal Bureau of Investigation, 1989:13). Even the circumstances of the murder were not reported for 22 percent of all murders in 1988 (Federal Bureau of Investigation, 1990:14). NCS provides only a limited amount of information on offenders, and it is based on the victims' unverified descriptions of their offenders. Information is solicited on the number of offenders in the incident; their ethnic status, sex, and age; and whether or not the offender is known to the victim and if so what the relationship is. When there is more than one offender, the information is sought on their aggregate composition by ethnic status, age, and sex (or for example the age of the oldest and youngest offender). Information is also sought on whether any was believed to be drinking or on drugs. There are no validation studies that permit an assessment of the accuracy of victim accounts of offenders for the NCS. Considerably more information could be obtained for offenders convicted of violent crimes, but there are few national tabulations for uniquely identified individuals. National prisoner statistics are also limited to major type of violent crime for which incarcerated and ethnic status, sex, and age characteristics of inmates. These characteristics of offenders can be obtained by jurisdiction and size of place from which committed. There are no national statistics on the criminal careers of persons who commit violent crimes. Available knowledge is derived
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 416 from studies of special incarcerated populations or by tracing the careers of cohorts of arrested offenders using the arrest records available through state and federal identification bureaus. Some information on criminal careers also is available from small-scale longitudinal follow-ups of cohorts sampled from the general population (e.g., Elliott et al., 1989) or from self-report studies of juveniles. The dearth of information on violent crimes in offenders' careers is a severe limitation on developing an understanding of violent criminal activity and of interventions to prevent its occurrence. One way to develop that information is investigation based on longitudinal studies of cohorts who are followed from birth to the middle years, as we describe in Chapter 3. Barriers to Obtaining Complete Information Measuring the Risk of Violent Victimization Two questions are often asked: What are my chances of becoming the victim of a rape, a robbery, or an assault? Are there ways to alter my behavior to lower my chances of being a victim? To answer them and indeed to understand the nature of violent behavior and its consequences, it is necessary that one understand how changes in conditions surrounding the violent behavior affect its occurrence. To understand a robbery murderâsuch as that of Mr. Waldermarian recounted at the beginning of this bookâimplies understanding a chain of events, each of which could change the outcome. These are referred to statistically as conditional probabilities, the probability associated with a given condition, such as being in a particular place or situation. The chain of conditional probabilities for the robbery murder would include the chances of being approached in a given place and situation by a robber and, if approached, by one who is armed with a gun. Given the robbery attempt, what is the likelihood that one will be shot by the robber and then, if shot, that one will be injured? Moreover, given the injury, what is the likelihood that it will be life- threatening and, if life-threatening, that it will result in deathâthe fate of Mr. Waldermarian? Looked at from the perspective of the effectiveness of interventions, to know that some intervention reduces robbery murder levels implies that it reduces at least one conditional probability in the chain without causing offsetting increases in the others.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 417 For example, interventions that reduced the supply of guns in the population would reduce the chance that an injury kills the victim, one of the conditional probabilities in the chain. It would reduce the number of robbery murders provided there are not offsetting conditions that would increase other chances of a fatal injury during a robbery, such as more frequent victim resistance. One of the barriers to assessing the risk of violent behavior for ordinary citizens, of testing hypotheses about violent behaviors, or of verifying claims about preventive interventions is that one lacks the data needed to calculate the necessary conditional probabilities. Clearly most of that information is lacking for calculating the conditional probabilities for robbery murder, for at least three reasons: (1) Lack of denominators for rate calculations: One can often ascertain many elements of circumstances surrounding an event such as an attempted robbery (e.g., Mr. Waldermarian was in the driveway of his house on Christmas Eve carrying Christmas presents). Unless one knows how many other men in similar situations were not victims of attempted robberies, one cannot calculate the first conditional probability in the chain or know how important time and location are as risk factors. One can assemble information on theoretically relevant circumstances from police or prosecutors' records, but it is different because these data are not routinely tabulated in statistical reports. (2) Probabilities are usually unavailable for the population at risk : Two of the succeeding conditional probabilitiesâof being approached by two men attempting robbery and of a gunshot injuryâcould be obtained from victimization surveys, but only for surviving victims. One cannot tell from victimization data if victims who die are somehow different from those who live. Police data potentially can provide information on those who report attempts to rob with a shotgun and of how many homicides are deaths from shotgun wounds, but we would lack information on how many people are wounded by shotguns in attempts to rob, since it is possible that, despite the legal obligation to report, some of these would not be reported to the police or require emergency treatment and the hierarchy rules result in underreporting. (3) Inability to link data from different sources: The last probability, of death given the injury, is well established for many epidemiological injury categories, but that information cannot be linked to the rest of the chain without data that distinguish violent from unintentional injuries.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 418 Progress in understanding violent crime is severely hampered by the lack of information to calculate conditional probabilities. This is a severe limitation not only for understanding causal chains leading to violent crimes and their outcomes but also, as noted above, for our capacity to assess the effect of interventions to reduce violent crime. Such assessments depend substantially on securing information on the base for a rate that is integral to the calculation of a conditional probability. Discretionary Decisions After a crime incident occurs, the initial decisions whether to report and what to report are made by the immediate participants, far removed from the government agencies with official responsibility for classifying and counting crimes. Police agencies must depend on the discretionary decisions of victims or witnesses of violent crimes. Unless they call, an incident that might qualify as a violent crime will not be accessioned in official police statistics. The police likewise have thresholds for determining what constitutes a violent crime. Although governed by the criminal law, police officers have discretion whether to classify and record incidents reported to them by citizens. What constitutes a violent crime, then, will vary among legal jurisdictions, depending on citizen and police construction of events and the decisions they make regarding them. As we said, the NCS was developed to address the problem that many violent crimes are not reported to the police. However, as a sample survey of people reporting victimizations to interviewers, it too depends on discretion: people's willingness to be interviewed and to respond to the survey questions as they understand them. Despite an estimated compliance rate of about 96 percent of all eligible housing units participating in the 1988 survey (Bureau of Justice Statistics, 1990:120), there are reasons to conclude there is a sample selection bias in the NCS. That is, respondents at high risk of violent behaviorâas both victims and offendersâare underrepresented, for example, young unemployed black males and homeless people. The age limitation on the NCS results in underestimates of assaults against minors by family members and peers, as well as underestimation of strong-arm and even armed robberies against those under age 12. Some, but by no means even a majority, of those offenses will be counted by UCR because adults report them to the police. In classifying and counting the information from respondents,
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 419 the NCS survey organization follows a particular procedure to determine which of the behavioral incidents are to be considered violent crimes; this procedure substitutes for the discretionary decisions of citizens and the police. The NCS uses an algorithm to classify violent and other crimes based on the answers to individual survey questions. For example, the NCS has not asked respondents whether they were victims of a completed or attempted forcible rape. Rather, individuals are first asked screen questions; if they respond to one or more indicating they were a victim of force or threat to use force, they are asked for information on a crime incident form (Bureau of Justice Statistics, 1990: Appendix II). That form asks more specific questions about the nature of the threat or attack. If the victim describes the attack or describes the injury in terms of a sexual assault that the interviewer considers forcible rape, the interviewer then exercises discretion to code it as "tried to rape" or "raped." The procedure thus depends on victims' associative recall when asked more general questions about threats and attacks on the person or on the nature of injuries.3 Similarly, the NCS does not ask direct questions about family violence, although it provides estimates by special tabulation of the family relationship between the victim and the offender, the location of the victimization, and other information considered relevant to classifying victims of intrafamily assaults. This procedure does not include partner or cohabitant assaults in the estimate. By contrast, the current UCR classification procedure depends on both the discretionary decision of officers when taking a report and the subsequent review of that classification by trained staff. The new UCR incident-based reporting system will more closely resemble that of the NCS, in that officer responses to individual items about the incident will enable classification by algorithm. Also, the NCS is entirely dependent on the victim as the source of information, whereas police departments derive their information from a variety of sources. Witnesses and others bring violent crimes to the attention of the police that victims would fail to report; this is especially the case for reporting assaults. A substantial number of incidents of violent crime are reported neither to the police nor to survey interviewers. Some information on the nature of such events can be gleaned from reverse-record check studies 4 conducted by the NCS in Washington, D.C. (Bureau of the Census, 1970a), Baltimore, Md. (Bureau of the Census, 1970b), and San Jose, Calif. (National Institute of Law Enforcement and Criminal Justice, 1972). These studies concluded
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 420 that in each city a substantial proportion of the rapes and attempted rapes, robberies, and assaults that occurred were not reported to the police; the NCS survey also failed to capture a sizable proportion of the violent crimes that were reported to the police. In San Jose, for example, just under half (48 percent) of all aggravated assaults included in the police case report subsample were reported by victims to survey interviewers. At the same time, only somewhat more than half of all persons who reported an aggravated assault to the survey interviewers said they reported it to the police. Each method of data collection thus fails to collect information on a substantial proportion of victims counted by the other, although overall the survey method captures more victims. Reporting Crimes to the Police Relatively little is known about how citizens decide to mobilize the police to deal with violent crimes. Recent work on domestic violence provides evidence that a sizable fraction of battered women do not call the police. Fear of retaliation from spouses is one of the more commonly stated reasons for calling a shelter rather than the police. There is likewise little information on why victims of violence do not report their victimization to survey interviewers. Again, what little information is available concerns domestic violence. In design interviews for the NCS, some of the men and women respondents indicated that they did not regard simple assaults at home, at work, or in bars to be crimes, suggesting that cultural constructions of men's and women's roles may account for failure to report these incidents either to the police or to victim interviewers. The main source of information on citizens' mobilization of the police when they are victimized by a violent crime is the NCS, which determines whether each victimization reported to the survey interviewer was also reported to the police,5 and if so why. For a victimization that was not reported to the police, the victim is queried as to the reasons. More than one reason may be given by a victim for each victimization so that there are more reasons given than there are victimizations. In 1988, for the estimated 5,909,570 violent victimizations, a total of 5,938,390 reasons were given for reporting (Bureau of Justice Statistics, 1990: Table 102) or not reporting (Bureau of Justice Statistics, 1990: Table 103) the victimization to the police. Fewer reasons apparently were given on the average for reported victimizations than for those not reported.6
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 421 Roughly 1 in 10 victims says that he or she reported the victimization to the police because "it was a crime" (Bureau of Justice Statistics, 1990: Table 102). The other reasons given for reporting vary with the type of violent victimization. For aggravated or simple assault, the main reason given for reporting 2 in every 10 victimizations is to prevent the offender from further violence toward the victim. These victimizations appear to be largely assaults in which the victim and offender are in a continuing relationship, primarily domestic assaults for female victims and acquaintance assaults for male victims. Evidence for this is also found in the fact that only about 5 percent of the reasons given for reporting assault victimizations to the police were to catch or find the offender. Another 1 in 10 assault victimizations is reported to the police because the victim hoped this offender would be prevented from committing a similar crime against any citizen; for an additional 1 in 10, the reason is to punish the offender. In about one-fourth of all completed robbery victimizations (Bureau of Justice Statistics, 1990: Table 102), the reason expressed for reporting was to seek the recovery of property. When recovery of property is not a motive, victims of attempted robbery are most likely to respond that they called because it was a crime. The number of rape victims is insufficient to do a reliable analysis of their reasons for calling the police, but the pattern that emerges is one of reporting in the hope the police will catch the offender, who will then be punished or prevented from further offendingâreasons given for more than half of all rape victimizations. Few injured victims state that they called the police because they needed help due to a violence-related injuryâperhaps because in most cases victims are still ambulatory and can seek medical help on their own. A variety of reasons are also given for not reporting victimizations to the police (Bureau of Justice Statistics, 1990: Table 103). In roughly 1 in 10 of the victimizations, the reason expressed for not reporting to the police was that it was reported to some other official. This reason seems to be offered more often for victimizations that include an injury. The two most frequently given reasons for not reporting a violent victimization to the police are that the matter was private or personal and that the offender was not successful in an attempted crime or that property was recovered in a completed property crime.7 That it was a private or personal matter is more likely to be the reason given for not reporting assault than for rape or robbery victimizations. Fear of reprisal is given for more than 10 percent of all completed robbery and rape victimizations. Fear of reprisal is given as a reason for
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 422 not reporting in fewer than 5 percent of the assault victimizations. For crimes not reported to the police, there are not only more reasons given but also much greater diversity. For 17 percent of such victimizations in 1988, some reason other than the 12 adopted by the NCS was given (Bureau of Justice Statistics, 1990: Table 103). Some other reason was given for not reporting 4 in 10 rape victimizations and for one-fifth of all completed violent crimes. There is apparently no simple explanation for not reporting violent crimes to the police from the perspective of their victims, and a sizable minority give more than one reason for not reporting. A minority of victims give reasons indicating they regard their victimization as lacking the seriousness worthy of police mobilization. In about 5 percent of all violent victimizations, the reason given is that the crime was not important enough to merit police attention; for almost 7 percent the reason was that the police would not want to be bothered. For another 6 percent, the reason was a lack of proof. For 3 percent, the reason given was that reporting is too time- consuming or inconvenient. Although these percentages are not additive because a victim could give more than one reason, perhaps as many as one-fifth of all victimizations are not reported to the police because citizens do not regard the incidents as worthy of police resources. Whether or not a crime is reported to the police also depends on the status of victims and offenders, the degree of violence in the event, and the consequences of violence: â¢ Victims are more likely to report each type of violent crime if the crime is completed, is a serious assault on the person, and involves injury (Bureau of Justice Statistics, 1990: Table 92). By way of example, 8 of every 10 completed robberies with injury from a serious assault were reported to the police in 1988, compared with fewer than 4 of every 8 simple assaults attempted without a weapon. â¢ Females are, with few exceptions, somewhat more likely to report their violent victimizations to the police than are males (Bureau of Justice Statistics, 1990: Table 94). â¢ Blacks who recount violent victimizations to NCS interviewers are more likely than whites to state that they reported the incidents to the police (Bureau of Justice Statistics, 1990: Table 95). â¢ Persons ages 12 to 19 are less likely to report their violent victimizations to the police, except that females in this age group
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 423 are more likely to report their victimization by rape than are older females (Bureau of Justice Statistics, 1990: Table 97). Other Barriers There are other reasons why crime counts differ depending on the institutional means of classifying and counting. The police information used in the UCR system to classify crimes is collected quite close in time to the event, and the information is secured from interviews not only with the victims but also from other informants, some of whom were witnesses to the event. The NCS relies solely on self-reports by persons who consider that they were victimized by an offender or offenders. Reporting depends entirely on recall of victimizations over the previous six-month period, which entails substantial inaccuracy in the recall and dating of events (Skogan, 1990). As Chapter 2 discusses, a major barrier to accurately estimating violent crime incidents is that, when three or more crime victimizations are similar if not identical, victims often are unable to recall the details of each incident so that they can be separately classified, or they cannot provide an accurate count of the number of such incidents. These are designated series incidents by the NCS. Because their number cannot be precisely measured, their count is excluded from NCS reports of victimizations and its estimates of incidents.8 The NCS provides some information on the effect this error has on estimates of crime. Table B-1 presents 1988 NCS counts for victimizations not in series, series victimizations, and the combined total, counting each series victimization as only a single victimization, based on the victim's report of the most recent one. In comparing the total combined count with the official count based on victimizations not in series, it is apparent that excluding violent series crimes from the count of violent victimizations underestimates the amount of violent crime. Even when counting a series victimization as a single additional victimization, the 1988 violent victimization count is underestimated by 9.2 percent. Multiplying each series incident by 3 would add an additional 1,638,690 violent crime victimizations to the 1988 count, resulting in an estimated increase of 27.7 percent. What is most evident from the table is that series victimizations are in the aggregate less serious crimes than are victimizations not in series. This is evident in a number of ways. Proportionately, more of the violent series victimizations are attempted
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 424 TABLE B-1 Crimes of Violence, 1988: Number and Percent Distributions of Series and Nonseries Victimizations by Type of Violent Crime Total Victimizations Series Victimizations Victimizations not in series Type of Number Percentage Number Percentage Number Percentage crime Crimes of 6,455,800 100.0 546,230 100.0 5,909,570 100.0 violence Completed 2,338,690 36.2 158,700 29.1 2,179,980 36.9 Attempted 4,117,100 63.8 387,520 70.9 3,729,580 63.1 Rape 137,350 2.1 9,970 1.8a 127,370 2.2 Completed 65,550 1.0 0 0.0a 65,550 1.1 Attempted 71,790 1.1 9,970 1.8a 61,810 1.1 Robbery 1,111,160 17.2 63,150 11.6 1,048,000 17.7 Completed 730,870 11.3 46,610 8.5 684,260 11.6 With injury 275,250 4.3 12,380 2.3a 262,870 4.5 From 133,920 2.1 3,820 0.7a 130,090 2.2 serious assault From 141,320 2.2 8,550 1.6a 132,770 2.3 minor assault Without 455,620 7.0 34,220 6.2 421,390 7.1 injury Attempted 380,280 5.9 16,540 3.1 363,730 6.1 With injury 118,540 1.8 8,260 1.5 110,270 1.8 From 56,850 0.9 6,360 1.2a 50,490 0.8 serious assault From 61,680 0.9 1,900 0.3a 59,780 1.0 minor assault Without 261,740 4.1 8,280 1.6a 253,450 4.3 injury Assault 5,207,290 80.7 473,090 86.6 4,734,190 80.1 Aggravated 1,842,100 28.6 100,710 18.4 1,741,380 29.5 Completed 610,720 9.5 40,130 7.4 570,580 9.7 with injury Attempted 1,231,380 19.1 60,580 11.0 1,170,800 19.8 with weapon Simple 3,365,180 52.1 372,370 68.2 2,992,800 50.6 Completed 931,540 14.3 71,950 13.2 859,580 14.5 with injury Attempted 2,433,640 37.8 300,410 55.0 2,133,220 36.1 without weapon a Estimate is based on about 10 or fewer sample cases and should not be considered reliable.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 425 rather than completed crimes of violence. Series victimizations are less often the more serious crimes of rape, robbery, and aggravated assault. Likewise, proportionately fewer of the series victimizations are reported with injury. It is also noteworthy that simple assaults are overrepresented among series victimizations compared with those not in series. This provides some additional evidence that the self-report victim survey is more likely to record the less serious crimes of violenceâthose that are less likely to be reported to the police. Even adding series victimizations as a single incident to the total count is still an underestimate, since a sizable number of victims estimate four or more incidents in their series. Loftin and MacKenzie (1990) estimated the effect of multiplying each series incident by the victim's estimate of the number of separate incidents in the series for 1987 incident-level data. The vast majority of incidents had a frequency of less than 10, but the distribution extends to 150 incidents in a series. High-frequency series victimizations have a considerable impact on the estimate when each series crime is weighted by the victim's estimate of the number of incidents in a series. Loftin and MacKenzie found that weighting by victim estimates of the number of victimizations in a series increased the estimated frequency of victimization 10 times over the estimate that excludes series victimizations. By this estimate, repeat victimization of a relatively small number of the total victimsâthe series victimsâaccounts for the bulk of victimizations in the weighted estimated rate. Loftin and MacKenzie caution against uncritical acceptance of a victim's estimate of the number of incidents in a series. They draw attention to the fact that response errors are necessarily large for series crimes because information is not collected for separate victimizations (incidents). Also, violent series crimes with high frequency are extremely rare, yet they have a substantial impact on the overall rate. Investigation of repeat victimization over time provides a third reason for caution. Multiple incidents reported as either series or nonseries victimizations persist for only short periods of time in the NCS population of victims. It seems likely that the sharp fall in repeat victims, especially of series victims, from one six-month interview to the next is an artifact of the survey procedure that does not measure repeat and continuing victimization adequately (Reiss, 1981:48).
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 426 Statistical and Other Compensating Mechanisms Just how seriously the national violent crime rate is underestimated given the fact that each of the two main methods misses a substantial proportion of victims captured by the other cannot be estimated precisely at the present time. Using a capture-recapture statistical model of estimation for the 1980 NCS data on aggravated assault and the data on aggravated assault reporting from the San Jose reverse-record check study, Reiss estimated that the estimated 1980 crude aggravated assault victimization rate using multiple capture was almost twice as high as that estimated by the NCS and over three times that estimated from police case reports (Reiss, 1985:166-167). These are crude estimates and should be regarded as illustrative of the problems faced in estimating the magnitude of violent crime rather than as definitive estimates. Although it is clear that each method fails to capture the whole picture, the magnitude of the loss from any one method in estimating the national violent crime rate is difficult to ascertain. There are a number of reasons for this conclusion. Multiple capture methods require that individuals or events be uniquely identified for accurate estimates, as that is the only reliable means of determining whether they are capturing the same events or individuals. There are significant institutional barriers to unique identification of individuals, particularly by government agencies such as the Bureau of the Census, which closely guards individual identity. Moreover, to increase the accuracy of matching, each information system would need to be redesigned, since it is no simple matter to achieve unique identity matching of either individuals or incidents. Unique identity of crime incidents reported by two or more systems is even more difficult to determine. The accuracy of any system of multiple capture would be enhanced by the use of dual frame or multiple frame sample selection, which again would require major institutional changes in data collection. It is useful at this point to note the substantial number of data collection systems, not currently taken advantage of, that acquire information on violent crimes. Use of that information could substantially increase our ability to estimate the nature of violent behaviors, including violent crimes, and their consequences. Emergency medical services are a major potential source of information on violent behavior and its consequent injuries. A number of studies in the United States and abroad provide evidence that emergency medical treatment services process a substantial number of victims of violent crime that are not reported to the police. A study of characteristics of victims of self-reported assaults admitted
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 427 to St. Vincent's General Hospital Accident and Emergency Center in Sydney, Australia, reported that 57 percent of the 60 percent of self-reported victims of assault who consented to be interviewed did not intend to report their assault to the police (Cuthbert et al., 1991:138, 143). In one northeastern Ohio county where rates for the victims of police assault incidents were compared with the rate for assaults known to hospital emergency departments, the emergency medical services incidence rates for assault with physical injury (including forcible rape) were nearly four times the rates reported in UCR (Barnachik et al., 1983). Notes 1 A felony murder is one committed during the course of committing any felony. Armed robbery is a felony. There is some evidence that persons who are under the influence of drugs and/or are young inexperienced robbers are more likely to fire their weapon during an armed robbery, resulting in a deadly assault. 2 An exception is the UCR Supplementary Homicide Report, in which age, race, sex, ethnic status, and relationship are reported for victims and known offenders. 3The Redesign Phase III of the National Crime Survey currently being implemented has added a screen question directly inquiring about âany rape, attempted rape, or other type of sexual attack" as well as a further probe about "incidents involving forced or unwanted sexual acts (that) are often difficult to talk about." Respondents who acknowledge such attacks are asked specific questions about them on the Crime Incident Report (Forms NCS 1X and 2X, OMB no. 1121-0111). 4 A reverse-record check study basically compares two methods of data collection by embedding a sample of those subject to one method of data collection on a population of events within a probability sample of individuals subject to the same event who are then exposed to a second method of data collection on the same population of events. These reverse-record check studies embedded a sample of victims of rape, robbery, and assault whose victimization was known to the police of the city within a probability sample of the households of that city in which all members were interviewed. 5 The term reporting to the police rather than calling the police is used for a number of reasons but primarily because the victim is not always the person who called the police. Additionally, some victimizations are reported in person rather than by calling the police. 6The total number of reasons for reporting is substantially less than the total reported victimizations (compare Bureau of Justice Statistics, 1990: Tables 92 and 102). That discrepancy may be the result of error in tabulation or reporting. 7 Unfortunately, the NCS combines the reasons of "offender unsuccessful
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 428 in the attempted crime" and "property recovered" for completed crimes. These would seem to be related to the type of violent victimization with the former reason more commonly given for rape and assault and the latter for robbery victimizations. 8 Table B-1 reproduces statistics on 1988 series victimizations published in a methodological appendix by the Bureau of Justice Statistics (BJS). The series counts for forcible rape illustrate the BJS dilemma. Ignoring series rape victimizations undercounts completed plus attempted rapes by at least 8 percent and probably far more. Yet, because the sample sizes for series rapes are too small to obtain valid counts, BJS might, by including these data in the body of its report, create a misimpression that only about 1 in 10,000 series rape attempts are successful. Placing the table in a methodological appendix discourages such misinterpretations while helping researchers to make adjustments to estimates based on their criteria. References Barnachik, J.I., B.F. Chatterjee, Y.E. Greene, E.M. Michenzi, and D. Fife 1983 Northwestern Ohio trauma study: I. Magnitude of the problem. American Journal of Public Health 73:746-751. Biderman, Albert D., and Albert J. Reiss, Jr. 1967 On exploring the "dark figure" of crime. Annals of the American Academy of Political and Social Sciences 374:1-15. Blumstein, Alfred, J. Cohen, and R. Rosenfeld 1991 Trend and deviation in crime rates: A comparison of UCR and NCS data for burglary and robbery. Criminology 29:237-263. Bolton, Frank, L. Morris, and A. MacEachron 1989 Males at Risk: The Other Side of Child Sexual Abuse. Newbury Park, Calif. : Sage Publications. Bureau of the Census 1970a Victim recall: Pretest household survey of victims of crime. Washington, D.C. (mimeo). 1970b Household surveys of victims of crime: Second pretest. Baltimore, Md. (mimeo). Bureau of Justice Statistics 1990 Criminal Victimization in the United States, 1988. A National Crime Survey Report, December 1990, NCJ-122024. Cromwell, Paul F., J.N. Olson, and D'aunn W. Avery. 1991 Breaking and Entering: An Ethnographic Analysis of Burglary. Newbury Park, Calif.: Sage Publications. Cuthbert, Marjorie, Frances Lovejoy, and Gordian Fulde 1991 Investigation of the incidence and analysis of cases of alleged violence reporting to St. Vincent's Hospital. In Duncan Chappell, Peter Grabosky, and Heather Strang, eds. Australian Violence: Contemporary Perspectives. Canberra, Australia: Australian Institute of Criminology.
MEASURING AND COUNTING VIOLENT CRIMES AND THEIR CONSEQUENCES 429 Elliott, D.S., D. Huizinga, and S. Menard 1989 Multiple Problem Youth. New York: Springer-Verlag. Federal Bureau of Investigation 1989 Uniform Crime Reports: Crime in the United States, 1988. 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. Jencks, Christopher 1991 Is the underclass growing? In Christopher Jencks and P. Peterson, eds., The Urban Underclass. Washington, D.C.: Brookings. Johnson, Bruce D., T. Williams, K. Dei, and H. Sanabria 1990 Drug Abuse in the Inner City: Impact on Hard-Drug Users and the Community. In Michael Tonry and J.Q. Wilson, eds. Drugs and Crime. Chicago: University of Chicago 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. National Institute of Law Enforcement and Criminal Justice 1972 San Jose Method Test of Known Crime Victims. Statistics Technical Report 1, Statistics Division. Washington, D.C.: U.S. Government Printing Office. Reiss, Albert J., Jr. 1981 Measuring repeat victimization in the National Crime Survey and the special case of series victimization. Social Statistics Section Proceedings of the American Statistical Association 41-50. 1985 Some failures in designing data collection that distort results. Pp. 161-177 in Leigh Burstein, Howard E. Freeman, and Peter H. Rossi, eds., Collecting Evaluation Data: Problems and Solutions. Beverly Hills, Calif.: Sage Publications. Rengert, George, and J. Wasilchick 1989 Space, Time, and Crime: Ethnographic Insights into Residential Burglary. Report to the National Institute of Justice, Temple University. Sherman, Lawrence W., and Ellen G. Cohn 1989 The impact of research on legal policy: The Minneapolis domestic violence experiment. Law and Society 23:117-144. Sherman, L.W., L. Steele, D. Laufersweiler, N. Hoffer, and S.A. Julian 1989 Stray bullets and "mushrooms": Random shootings of bystanders in four cities, 1977-1988. Journal of Quantitative Criminology 5(4):297-316. Skogan, W.G 1990 Criminal victimization. Pp. 7-22 in A. Lurigic and G.G. Skogan, eds., Crime Victims. Newbury Park, Calif.: Sage Publications.