Title: Proceedings of the Homicide Research Working Group Meeting,1997 and 1998 (pages 1-57). Series: Research Forum Author: NIJ Published: National Institute of Justice, May 1999 Subject: Homicide, domestic violence, juvenile violence, gun violence, gangs, crime patterns and future trends, and statistics 58 pages 139,000 bytes ------------------------------ This is an ASCII plain-text file. To view this document in its graphic format, download the Adobe Acrobat graphic file available from this Web site or order a print copy from NCJRS at 800-851-3420 (877-712-9279 for TTY users). ------------------------------ U.S. Department of Justice Office of Justice Programs 810 Seventh Street N.W. Washington, DC 20531 Janet Reno Attorney General Raymond C. Fisher Associate Attorney General Laurie Robinson Assistant Attorney General Noel Brennan Deputy Assistant Attorney General Jeremy Travis Director, National Institute of Justice Office of Justice Programs World Wide Web Site http://www.ojp.usdoj.gov National Institute of Justice World Wide Web Site http://www.ojp.usdoj.gov/nij ------------------------------ Table of Contents 1997: Policy, Practice, and Homicide Research Discussion 1997 Keynote Kick-Off: What Works? David Kennedy Building Bridges Between Research and Practice: Youth Violence Prevention Rich Rosenfeld and Troy Miles Anthony Braga, David Kennedy, and Tito Whittington Don Faggiani and Colleen McLaughlin Bridging the Gaps for the Virginia "Cradle-to-Grave" Homicide Project Colleen R. McLaughlin Donald Faggiani Jack Daniel Thomas J. Dover The Rewarding and Painful Process of Collaboration to Prevent Domestic Violence Jacquelyn Campbell Carolyn Rebecca Block Deborah Spungen Linda Langford Notes From a Collaborative Project: Risk of Serious Injury or Death in Intimate Violence Carolyn Rebecca Block The Houston Homicide Project: A Comparison of US and Canadian Findings on Uxoricide Risk for Women with Children Sired by Previous Partners Victoria E. Brewer Kelly R. Damphousse Derek J. Paulsen Partner Homicide in Massachusetts, 1991-1995: A Collaboration Between Advocates and Academics Linda Langford Methodological Challenges to Evaluating the Brady Handgun Violence Prevention Law Daniel W. Webster Evaluating the Brady Act and Increasing the Utility of BATF Tracing Data Gary Kleck Race, Class and Homicide: Looking Beyond Guns, Drugs, and Gangs Roland Chilton Demography and Lethal Violence Allan F. Abrahamse What Works? Using Firearm Tracing Information in Violence Reduction Intervention Projects John Firman Paul Blackman Joseph Vince Anthony Braga David Kennedy The Limitations on BATF Firearms Tracing Data for Policymaking and Homicide Research Paul H. Blackman Gangs Drugs and Youth Violence Recorder Notes Gangs, Race/Ethnicity and Houston Homicide in the 1990s Kelly R. Damphousse Victoria E. Brewer Cary D. Adkinson Today's Music and Youth Violence Kathleen M. Heide Area Research on Homicide: Discussion Notes Lin Huff-Corzine Appendix A Agenda 1998: Bridging the Gaps: Collaborations on Lethal Violence Research, Theory,and Prevention Policy 1998 Keynote Address School Shootings and School Violence: What's Going On and Why? Kathleen M. Heide Session 1: Essentials of Violence Surveillance Data Responses by Lois Fingerhut Injury Surveillance Using Data From the National Center for Health Statistics Lois A. Fingerhut On Intimate Partner Homicides in Massachusetts Linda Langford Nancy Isaac Stacey Kabat An Evaluation of the Completeness and Accuracy of SHR Data in Chicago, 1993 and 1994 Thomas D. Patterson Daniel Dick Carolyn Rebecca Block The Surveillance Value of "Bad" Data: Using Obliterated Serial Number Data in a Firearm Surveillance System Bill Sherlock David Krieghbaum Situational Factors Related to Public Mass Murder Incidents: 1965-1998 Thomas A. Petee D.C. Jail Gun Identification and Style Survey John May Khalid Pitts Earnest Williams Roger Oen Exploring Circumstances of Prior Gunshot Wounds Among Jail Detainees John May Roger Oen Khalid Pitts The Spousal SROK Revisited: A Comparison of Houston and Chicago Intimate Partner Homicide Derek J. Paulsen Victoria E. Brewer Spatial-Temporal Clustering of Chicago Homicides Margo Wilson Martin Daly Development of an Ultrasonic Method for Restoration of Obliterated Serial Numbers on Firearms Yari Yacobi Cohort Survival Projections of Homicide Victimization Allan F. Abrahamse Decreasing Violent Gun Crimes in New York City: A Result of Vigorous Law Enforcement Efforts, Other Variables, or Both? Steven F. Roth Homicide Trends 1947-1996: Short-Term Versus Long-Term Factors Thomas B. Marvell Session 3: Collaborations Among Academics, Practitioners, and the Community Recorder: Lin Huff-Corzine The SECURE Program: Safety Enhanced Communities Utilizing Resident Endeavors Richard Block David Katz Laura Herrin Youth, Firearms and Violence in Atlanta: A Problem-Solving Approach Dawna D. Fuqua-Whitley Arthur L. Kellermann Media Portrayal of Child Abuse: A Content Analysis of Australian Newspapers Ania Wilczynski Reducing Youth Violence in the 21st Century Kathleen M. Heide, Ph.D. Session 4: Integration: Theoretical Connections Appendix A Agenda Appendix B Meeting Participants ------------------------------ Please note that this document does not include all of the papers presented at the meetings, only those supplied to the National Institute of Justice by the authors. ------------------------------- 1997: Policy, Practice, and Homicide Research Proceedings of the 1997 Meeting of the Homicide Research Working Group ------------------------------ Discussion 1997 Keynote Kick-Off: What Works? Presenter: David Kennedy Recorder: Kathleen M. Heide, University of South Florida The keynote address generated a great deal of discussion from the audience. Participants noted that social policy involves ethical decisions and that policy analysis is an "art form," which ideally involves a collaborative effort from the "number crunchers" to the "people in the trenches." The challenge for researchers today is how to present practitioners and policymakers with useful information. It can be difficult to generate sound policy recommendations from research. However, sometimes researchers set their sights too high. Basic knowledge and recommendations that may seem simple and obvious may be helpful to practitioners and policymakers. In fact, sometimes "a non- answer," that is, clarifying what does not work, may be invaluable to practitioners and policymakers because it leads them in more productive directions. Social scientists possess extensive knowledge which can be helpful if they can package in a way that others can hear it. Information and contact is necessary. For example, putting a researcher in a police agency enables information to flow both ways. Researchers need to realize that today's resources appear more finite than in previous decades. Given the political realities, agencies and programs compete for limited dollars. If one program is demonstrated to work, practitioners from another may fear a loss of their funding. Accordingly, researchers may be dealing with "dueling providers," whose purpose ostensibly is to solve the same problem (e.g., youth violence), but who feel a need to compete rather than cooperate with one another. Researchers need to pay attention to funding cycles. Government leaders ask where money should be funneled when budgets are being prepared. They are not interested in empirical findings per se. Politicians simply what to know what to do. Researchers can inform them if they put information at an understandable level. Research, practice, and policy go hand in hand. The suggestion was made that researchers in criminal justice should consider starting with the policy question and designing research from there. ------------------------------ Building Bridges Between Research and Practice: Youth Violence Prevention (8:15 - 10:50 6/9) Recorded by John Jarvis Rich Rosenfeld and Troy Miles Anthony Braga, David Kennedy, and Tito Whittington Don Faggiani and Colleen McLaughlin Lois Mock and Bob Flewelling - Organizer Representatives Lois Mock opened this session by noting that the National Institute of Justice is supporting ideas in this area of building partnerships to prevent youth violence. The purpose of this panel is to report on these efforts. Bob Flewelling also noted that the National Institute on Drug and Alcohol Abuse is interested in more research on drug use prevention. Research ideas on the great disconnect between research and practice is needed. What answers does the research provide? He also asked about designing research to examine how research gets translated into practice. He also noted that emphasis is needed on school-based prevention. Finally, he suggested that there are reasons for this disconnect but feels that there are mechanisms to bridge the gap between research and practice. Youth Mentoring in St. Louis Rich Rosenfeld began this presentation by noting that the goal of their efforts was to evaluate community based intervention with community based partnerships. His ideas were critical of some partnerships, but not opposed to such relationships. His project was funded as an urban demonstration project by NIJ. He worked with the St. Louis Police department. The plan was to emphasize mentoring and conflict mediation and evaluate efforts over 6 months. They concentrated on violence among urban black youth. In particular, they focused on youths in the Andy Malone Youth Home. These youths were found to be intermediate risk to violence due to family conflicts and other risk factors- base line indicators of risk among both males and females. Their median age was 13-14. Other data showed elevated risk to violence and particularly gun violence. Preliminary results (4 months) suggest fewer identified needs for a gun for protection across both control and mentoring groups. Those who want a gun also declined but still remained high at 30% among those who do not possess a gun. Initial partnership effort failed due to structural reasons. Bringing the stakeholders to the table was the failure. Vested interests led to reluctance to share resources, information, etc. Each player was suspicious of the ends sought of the other. The whole was worse than the sum of its parts. The point was not that collaboration would not work. Rather, they suggest that collaborations will work when: 1) there are only a few players 2) prevention does not work when the demand for prevention is high but when the community resources become surrogate victims. 3) the fewer the resources the greater the cooperation. This is due to the fact that the availability of greater resources evolves into competition for larger slices of the pie. The partnership with Americorp worked for St. Louis. Troy Miles continued by noting that violence prevention can be achieved through community safety by designing a mentoring program that is 1-1. However, there is a necessity to develop volunteers. Perceptions are a source of problems. These kids are no different than other kids. Conflict mediation, listening, and cultural diversity are important. Recruitment is massive. Goal is to get 75 mentors. They have 15 this year and had 35 last year. There are not enough volunteers. The commitment is 2-3 hours a week. Troy Miles attributes the lack of volunteers to fear of these kids and parental fears of mentor due to jealousy. Some acceptance of liability needs to occur through brokering to other agencies. Some are getting sued due to their efforts. Its also about competition. Americorp develops mentors and deposits them with existing mentoring programs. Rich Rosenfeld continued noting that these partnerships are essential to do work- single most important thing is a federal not local presence in the project. Or at least not a state government representative that was some part of the initial problem. Americorp represents a possible solution to this problem since there is no stake in the local competition for services rather on finding a solution. Youth Violence Prevention Project in Boston David Kennedy began by noting a program entitled "Ceasefire Intervention" which is a 1 year city-wide treatment program aimed at gang violence. As a result, youth homicide has declined below pre-crack levels. This program began about a year ago. Federal, state, and local representatives consulted and promised to report any useful results. These individuals were also drawn upon for implementation. Much was going on but no one took ownership of violence problem. David Kennedy and Anthony Braga sought out this problem. It was directed at: 1) firearms traffic 2) gang violence and 3) direct communication with gang members emphasizing a crackdown on violence. The interagency work and crackdown were noted as the strongest success in this effort. Detective Tito Whittington of the Boston Police Department then narrated how this program worked and substantiated much of what David Kennedy noted. In particular, he pointed out that when David initiated this program he did not come into the Police Department with numbers but moved in physically. He also bought publicity into the police department. The embarrassment of a wealth of resources that were either underutilized or underidentified was again noted. Partnerships with the community were the key. Marrying these programs and resources with the community was the key. Through this effort they claim to have eliminated youth homicide by gun. Interagency forum was important to effort. Now they are looking at when these kids leave jail--looking at rehabilitation and the role of females. They are also looking at diversionary programs. What about resources- no new resources were garnered, rather they were aligned with law enforcement and the community was important. Change in philosophy of policing to look at social service issues rather than simply terminating at the end of the policing role. Single most important solution was the partnership that created publicity and defined resources that wee available. For the kids, the message the officers carried to the kids was important. More than busting kids but offering job opportunities, or other services as an option. The main thrust was coordinating law enforcement with probation. Youth Violence Injury Intervention Don Faggiani opened this session and described what he termed the "Cradle to Grave Project" which focused on aggregate trends noting that the Supplementary Homicide Report data was useful in the analysis of "Juvenile Murder in Virginia". This data showed large increases in juvenile murder from about 1988 until about 1993. Much of increase is in the category under 18 as noted in the report. Also many of the perpetrators were black males and firearms were used. They also examined MSAs and central cities and examined the circumstance surrounding these killings including arguments, robberies, and narcotics. They also focused on single victim- single offender incidents. These trends explained about 90% of the increase in overall homicide rates from 1986-1993. They also analyzed syndromes using the topology offered by the Blocks. Using this Instrumental versus Expressive classification by demographics they were able to gain a lot of response from the agencies responsible for delivering these services. Intervention strategies grew out of this response. Colleen McLaughlin then detailed the research effort which grew out of this and was funded by NIDA. This work entailed a review of 25 juvenile murders and attempted to merge the existing data with medical examiners data. The results showed offenders were predominantly black, male, and involved in drug selling. Other recent substance abuse was also present in the victim. Substance use on the part of both victim and offender correctly classified these cases about 85% of the time. In an attempt to answer the question of whether these victims were high risk victims further analysis showed prior firearms injuries were present. The medical records of the victim were crossed with the juvenile criminal records and found 65 matches. Of these victims, 2/3 were court involved. Court involvement was found to be a substantial risk factor for injury. Drug sellers even more so. Also increased penetration into juvenile justice system showed increase in likelihood of firearms injury. Injury patterns and patterns of offending were found to be more serious among court involved youth. Normalization, and routine activities associated with these injuries. Eradicating these attitudes was a goal. Education of high risk individuals was used an intervention to meet this goal. The California Wellness Foundation: A Violence Prevention Initiative Peter Greenwood opened this presentation with the identification of collaborators in this effort and gave some background information on the increases in youth violence experienced in California. The Wellness Program grew out of a conversion of health services that led to the Wellness Foundation. In this effort firearms were identified as a major factor. However, a lack of information on how to combat this violence persisted. Having identified a goal to reduce youth violence and to provide leadership on violence prevention, the Wellness Foundation also shifted the policy discussion from a criminal justice model to a public health model thru gun control, limiting access to alcohol by kids, etc. In order to do this, the Wellness Foundation was formed to focus upon: 1) Policy--and public relations focusing on education. 2) Community Action Programs-outreach, mentoring, etc. 3) Leadership Programs-education to professionals, fellowship program to extend projects 4) Research Program-topic analyses firearms, alcohol use, etc. Also, evaluation component is present. 5) Academic advisory committee. Having established such an entity some conflicts have emerged. Among these are: 1) The public health model conflicts with traditional criminal justice system models. Most of money and resources are in criminal justice system so conflict emerges since criminal justice practitioners are seen as the bad guys. 2) Sorting through the empirical work? What works is difficult to determine. 3) Difficulties in keeping up with the new trends. 4) Sorting through national versus state and local data and trends. These are difficult to make sense of and sift through. 5) Politics, or pork barrel, problems in terms of how to carve up the pie. 6) Alcohol rather than drug use as emphasis? 7) Evaluation and perpetuation of programs as a problem? What happened and will it keep going on? The majority of effort has been primarily been in gun control. There has been a clear strategy, well connected, and the players are known. Passing local ordinances has also played an important role. However, a weak link between policy research and advocacy functions remains. Some conclusions regrading public education issues are offered. In particular, the public education effort has been effective in getting the message out about handgun violence against kids. Additionally, videoconferencing may have assisted in passing ordinances. There has also been recognized utility of behind-the-scenes involvement with activist organizations. With regard to community action programs, the main goal is to keep the doors open and keep the services going. There is no time to advocate. All in all, the annual meeting provides the glue to keep this together and all research areas defined have been covered with research findings being generated. Finally, while there is a commitment to mentoring, interaction within the group has been spotty at best. Yet, most activities that were initiated are ongoing, many accomplishments have resulted, and there has been strong movement toward the overall goals. Nonetheless, more needs to be done with less to continue these directions and integrate research and policy. In the meantime, public education has been the principal success in this effort. Questions, Answers, and Comments: Garen Wintemute pointed out the need for the research component and the need for evaluation. One area of collaboration that thrives is the development of limitations on small firearms (Saturday nite specials). Cheryl Maxson, drawing from her gang analyses suggested that identification of risk and prevention factors among gang members may be very important and advocates translations of research efforts into policy development. Allan Abrahamse, on the other hand, suggested that the poor quality of public health data are firing the conflict between the criminal justice model and the public health model. Following on Allan's comments Lois Mock asked for information on resource competition among community organizations and conflict between criminal justice and public health. David Kennedy then offered that their work had been mostly a policing project at beginning but shifted under their feet by attention to working relationships that were exploited privately. Community interaction happened but not formally. These things do not evolve as they should but draw on existing links and strengths. Rich Rosenfeld furthered this point by noting that competition over resources used to be there but has since been transcended by movements. However, entrepreneurship has become the movement. Peter Greenwood then noted that community service organizations have a goal of continuing service rather than a specific issue. Wedding these to violence is the key. Bob Flewelling argued that community organizations and researcher partnerships rather than state agency cooperation is probably more effective than the traditional consensus building among agency players. On a different note, Jianfang Chen asked about rates versus totals in Boston data, but also noted the role of demographics. David Kennedy responded that such trends are not attributable to demographics or base population changes. Dick Block returned to partnership issues and raised questions about foundations, community groups, and local and state governmental agencies. The evolution of efforts may not be coordinated in the same directions. Who should take the lead? Dick advocated the community group as defining the lead. Rich Rosenfeld commented that the federal presence has incredible influence on local politicians. Peter Greenwood pointed out that deciding how to do this depends on both learning what organizations can do in this effort and on coordinating these efforts. Roland Chilton then returned to the issue of interventions and asked "Are the kids being told what is right or what is wrong in their behavior? The answer from the research (St. Louis and Virginia) suggests that emphasizing what is right rather than what is wrong to inform their decision making in a nonconfrontational but educational way may be the correct course. Everett Lee then suggested that determining the optimal age for these interventions is important. Some say as early as age 8 but this is inconclusive. Others note that its not only the child, but the levels of intervention that are also important. Gary Kleck then returned to issues relative to the Wellness Foundation and gun control and research as an afterthought. Gary Kleck argued that such bans are in the direct opposition to known research. Enthusiasm over process may be misplaced if such efforts are demonstrated to be a failure. Peter Greenwood responded by suggesting that only about 1/7th of effort goes to this. Garen Wintemute supported this and noted that Saturday nite special purchasers are at a higher risk for future involvement in violence and injury. Peter greenwood also went further by noting that any movement is good movement. Wintemute also noted that research does not guide policy but substantiates policy. Final comments also noted that emphasis on schools and school structure as a place to quarter many efforts to intervene. ------------------------------ Bridging the Gaps for the Virginia "Cradle-to-Grave" Homicide Project Colleen R. McLaughlin Virginia Department of Juvenile Justice and Department of Surgery, Medical College of Virginia Donald Faggiani Criminal Justice Research Center of the Virginia Department Criminal Justice Services Jack Daniel Office of the Chief Medical Examiner, Commonwealth of Virginia, Department of Health Thomas J. Dover Henrico County Division of Police Abstract The researchers will address the process and barriers involved in bridging the gaps between research, theory, and practice; and the critical role played by the HRWG in the formation of this collaborative effort. We begin by discussing two parallel, though complementary studies of juvenile murderers in Virginia. One study resulted in a research-based violence prevention program aimed at breaking the cycle of violence for those with a high risk for intentional injuries. We will also discuss how we are overcoming some of the barriers to form a collaboration between two state agencies that will contribute to our understanding of the victims and perpetrators of lethal violence. Juvenile Murder in Virginia: a Study of Arrests and Convictions In 1993, the number of juveniles arrested for murder was the highest in Virginia's history. Most of the growth occurred between 1988 and 1993. Not only are the numbers of murders increasing but the weapons used in homicides are changing as well. During the early 1980's approximately 65% of juvenile homicide arrests involved the use of firearms. By 1994 this figure had increased to over 87%.To address this growing problem the Criminal Justice Research Center of the Virginia Department of Criminal Justice Services undertook a project to (1) to consolidate information on juveniles arrested for murder, juveniles convicted of murder, and the victims of juvenile perpetrated homicides; (2) to examine the relationships, trends, and situations underlying juvenile arrests for homicide; and (3) to categorize juvenile homicide to aid in identifying strategies to combat the growing trends in juvenile homicides. The information on juveniles arrested for murder is extracted from the Uniform Crime Reports Supplemental Homicide Reports (SHR). Juvenile correctional center data from the Virginia Department of Juvenile Justice and Circuit Court Pre-Sentence Investigation reports from the Virginia Department of Corrections provided information on the convictions of juveniles for murderer. The data from the SHR show that between 1986 and 1994 92% of juveniles arrested for homicide are male. Eighty percent of those arrested are black with white, Mexican American and Asian/Pacific Islander accounting for the remaining 20% of arrests. The average age of juveniles arrested for murder is 15.9 years. Firearms were used in 81.5% of the incidents in which a juvenile was arrested as the suspected offender. The gender and race characteristics of the victims closely follow those of the individuals arrested. Eighty-six percent of the victims were male, 68% of the victims were black. The average age of victims of incidents in which a juvenile is arrested is 28.5 years. In our examination of the trends that can be extracted from the available data several factors become readily apparent. First, urban and suburban areas account for the majority of juvenile arrests for homicide. From 1990 through 1994, 90% of all juvenile arrests for murder were within urban and suburban areas. Second, beginning around 1988 the percentage of incidents involving the use of a firearm began to grow rapidly. In fact, between 1989 and 1990 there was a 115% growth in incidents involving the use of a firearm. Third, males, and especially black males, account for the increase in arrests of juveniles for murder and non-negligent manslaughter from 1988 through 1994. Homicides begin as a form of confrontation. Research has shown that homicides can be characterized by the type of circumstances involved, the victim-offender relationship and the primary motive of the offender. Once characterized this information can be used to develop intervention strategies to combat the growing trends in juvenile perpetrated homicides. Our analysis shows that about 49% of incidents involving the arrest of a juvenile offender for homicide have as the primary motive the acquisition of money, property or drugs. In addition, 48% begin as an act of aggression where the primary motive of the offender is the aggression itself. Less than 25% of incidents where a juvenile is the suspected offender involve the killing of a stranger. The majority (60%) of victims of juvenile perpetrated murder involve the killing of an acquaintance or friend. Finally, we examined the sentencing of juveniles convicted for murder or non-negligent manslaughter. Most juveniles (83%) charged with some form of homicide are transferred to the circuit court. Once in circuit court it takes, on average, slightly less than one year to receive a sentence. Of those convicted and sentenced in circuit court, 98% receive some form of incarceration. Twenty-one percent of juveniles convicted of murder in circuit court receive a life or death sentence. In addition, juveniles convicted of first degree murder and incarcerated in the Department of Corrections prison system serve on average about 2.8 years longer incarcerated than their adult counterparts. Re-injury Prevention Program for Victims of Intentional Injuries Violent crime has been characterized as a public health crisis; demanding the combined efforts of social service, health care and criminal justice professionals serving at the forefront of this epidemic. In response to this problem, criminal justice and health care professionals in the Commonwealth of Virginia have initiated a major collaborative effort for the study, treatment and prevention of violence. Members of the Virginia Department of Juvenile Justice (DJJ) and the Office of the Chief Medical Examiner (OCME) have recently completed a study on the relationship between substance use, drug selling and lethal violence in 25 adolescent male homicide offenders and their victims. The results of this study indicated that 28% of the homicides perpetrated by adolescent males were related to the sale or distribution of illegal drugs, while juvenile drug sellers comprised a significant percentage (52%) of those incarcerated for murder. Moreover, recent victim drug use and perpetrator substance use emerged as important predictor variables in identifying drug-related juvenile homicides. These results are consistent with the view of homicide as a behavioral interaction where both victim and perpetrator variables play a role in the final outcome. Additional research done in collaboration between members of DJJ, the OCME, and the Medical College of Virginia Hospitals further documents a strong relationship between substance use, juvenile offending and assault-related firearms injuries. Retrospective review of the male, pediatric admissions for assault-related firearms injures (n = 65) revealed that 66% of the victims had documented involvement in juvenile offending. Recent substance use and/or court-involvement was identified in 82% of the cases, suggesting that substance use and criminal offending may represent important risk factors for intentional injuries in adolescent males. The constellation of individual consequences associated with intentional injuries includes social maladjustment, depression, substance use, and promiscuity, as well as subsequent violent acts and criminal offending. Moreover, the high-risk behaviors linked to initial injuries may be exacerbated by the violent event, significantly increasing the risk for re- injury if not changed. Further compounding the problems associated with intentional injuries, treatment providers increasingly are encountering normalization, and even glamorization or expectation of violent victimization. Thus completing the cycle of violence, the negative sequelae following an intentional injury may increase the risk for re-injury substantially. Though the causes and consequences of violence are multifaceted and complex, the identification of at-risk populations and characterization of putative risk factors provides an opportunity to develop focused violence prevention programs and break the cycle of violence. Research indicates that certain groups are at increased risk for sustaining intentional injuries. This population includes those previously injured, and individuals involved in substance use, drug selling or other crimina1 offending. Two programs resulting directly from our ongoing research on the relationship between drugs and violence are the Intentional Injury Prevention Program for high-risk youth, and the Re-Injury Prevention Program. The goals of these treatment and prevention programs are to target individuals at high- risk for intentional injuries or re-injury, attenuate the negative sequelae associated with an assault, and identify and link possible involvement in high-risk behavior to an increased risk for violent victimization through the use of a brief intervention model. The Intentional Injury Prevention Program has been included in a suburban Richmond, Virginia Juvenile and Domestic Relations Court assault diversion program since May of 1996. It has been expanded recently to include four additional counties in the Richmond, Virginia metropolitan area as part of a jail-based violence prevention program. The Re-Injury Prevention Program will be initiated with pediatric firearms injury victims at the Medical College of Virginia Hospitals during the summer of 1997. Where Are We Going? The Virginia "Cradle-to-Grave" Homicide Project The Virginia "Cradle to Grave" Homicide Project will provide the first opportunity to combine traditionally disparate data resources in an effort to characterize juvenile homicide from a victim, perpetrator and event perspective. These data resources will include complete perpetrator information (criminal, social, and psychological histories), victim information (cause, manner and mechanism of lethal injury; coincident substance use), and crime scene characteristics. The results of this study will significantly enhance ongoing and new treatment and prevention programs by facilitating judicious allocation of scare programmatic resources, and the identification of putative risk factors for lethal violence. This emerging collaborative effort has the potential to contribute significantly to our understanding the victims and perpetrators of lethal violence; underscoring the need for bridging the gap between research and practice in youth violence prevention. Acknowledgments This work was supported, in part, by the National Institute on Drug Abuse at the National Institutes of Health (DA10000). The authors thank Dr. Marcella F. Fierro, Chief Medical Examiner for the Commonwealth of Virginia, for permission to access the Medical Examiner Records; and Ms. Heather P. Sanderson for assistance with data collection and analysis. ------------------------------ The Rewarding and Painful Process of Collaboration to Prevent Domestic Violence Jacquelyn Campbell, Carolyn Rebecca Block, Deborah Spungen, and Linda Langford Judy Bonderman, Recorder This Workshop focused on the advantages, disadvantages, problems, and opportunities of collaborative intimate violence research. Jackie Campbell from Hopkins opened with some of the challenges in research collaboration with community agencies including differences in priorities, paychecks, attitudes, ideologies, and ethnic background. She acknowledged that there is often appropriate mistrust on the part of community agencies dealing with violence because of past paternalistic attitudes of researchers. Other concerns include client safety and confidentiality, lack of appropriate outcome measures, and inappropriate use of research results. Some of the solutions suggested and discussed for building and maintaining true partnerships were: --Take time to build and maintain collaborative relationships by spending time in each other's worlds (homicide department, domestic violence shelter); --Negotiate mutual gains through honest discussions. Develop multiple, creative, real world outcome measures by asking what agencies want to find out. Use results for mutually agreed on policy change efforts. List agency contacts as co-authors; --Use combinations of qualitative and quantitative culturally competent research; --Use action or empowerment research models which include use of survivors as well as student research assistants who can also assist agency programs. The presenters then discussed four collaborative projects, which brought together researchers and practitioners, academics and policy makers, public health and public safety agencies, and government entities. 1. "Risk Factors for Femicide in Violent Intimate Relationships" Jackie Campbell's seven city case control study of the relative risk factors for femicide in a battering relationship is truly interdisciplinary. It is funded by NIJ, CDC, and NIH, with eight primary investigators representing the disciplines of nursing, public health, criminology, and medicine. In each city there is a collaboration between law enforcement, the courts (for orders of protection), universities, domestic violence shelters, and bereavement centers. The study uses the public health case control model; the 250 cases (femicide victims) are represented by police homicide files as well as by a proxy (surviving mother, sister, or best friend.) The study is also collecting data from 250 attempted femicides and 250 controls (battered women) surveyed by a research firm. 2. "Partner Homicide in Massachusetts, 1991-1995" Linda Langford discussed a collaborative project between Peace at Home, a human rights agency, and the Harvard School of Public Health. Peace at Home was founded in 1991 to help free women who were in prison for killing their batterers. It started by tracking cases of femicide and publicizing cases of intimate homicide. When they decided they wanted to become involved in serious research, they contacted Nancy Isaak at Harvard who put them in touch with Linda. Together they constructed a database of all spouse and intimate partner related homicides in Massachusetts using multiple data sources. The study has two objectives: 1) to count and characterize accurately all partner homicides in the state over the 5-year period; and 2) to contrast the number of partner homicides in the new database with the number in the Supplemental Homicide Report. The study is in the final stages of data collection. The study began with a definite advocacy objective. Peace at Home wanted the numbers to show policy makers the seriousness of the problem. They have a lot of political connections and will be able to get the report out to a broad audience. The researchers are interested in how different databases mesh and what the SHR might be missing. An early issue arose when Peace at Home was surprised by how long the project would take. Linda suggested that a realistic time frame for completion should be laid out in the beginning. 3. "The Anti-Violence Partnership of Philadelphia" Deborah Spungen discussed two programs. The Anti-Violence Partnership of Philadelphia (AVP) began as a support group for the parents of area homicide victims and then evolved into a full-service victim advocate agency known as Families of Murder Victims (FMV). As Special Projects Director for AVP, Deborah is working with the Philadelphia Women's Death Review Committee to analyze the circumstances leading up to the death of all women murdered in Philadelphia between the ages of 15 and 60. This research project sprung from the concerns of the family of a woman killed by her husband at her place of work. The women had secured a protection order against her husband and the family wanted to go back through the events of their violent relationship to see if there were other points at which an intervention could have prevented this death. The Philadelphia DA has put together an interdisciplinary group composed of the medical examiner, the police, medical and public health personnel, an individual therapist, and the sheriff's department. Linda noted that there is lots of resistance in Massachusetts to death review but it is a good way to identify what may have gone wrong in the system. Deborah's second project is the Domestic Violence Workplace Partnership, a collaboration between AVP and Women Against Abuse. They studied employer procedures for protecting abused women and found that only 21% had policies in place. The partnership brings Employee Assistance Programs together with domestic violence shelters to develop a training package to proactively respond to the intrusion of domestic violence in the workplace. Their primary obstacle so far is getting the project funded. 4. "Risk of Serious Injury or Death in Intimate Violence" Becky Block described her project as a collaboration, not a partnership, that grew organically over the last 3 years. Chicago medical, public health, and criminal justice agencies are working together to identify factors that place women abused by an intimate partner in danger of life- threatening injury or death. Rather than starting with a hypothesis, the project began with the concept that we, as researchers, know who is at risk of being abused, but that doesn't help a 911 operator/dispatcher, a police officer, or an emergency room attendant decide who is at risk of dying from a battering relationship. The point of the project was to link non- lethal and lethal risk factors. Jackie's study is a partner to Becky's project. The goal is to get information to those who can help. Practitioners want to know who is at most risk of dying among battered women. Aside from the collaborating agencies, the study benefits from the contributions of a number of active participants, including "site advisory boards: of community members. The study is based on a point of service sample at six clinics and hospital sites, with the sampling process integrated into the site intake process. In a sense, all interviewers and service providers are collaborators in the project. The project is currently piloting the study and hiring and training interviewers. Some major legal and ethical issues have surfaced concerning respondent safety and confidentiality. There is a need for a consultant psychologist or counselor on staff at each site where interviews take place to handle any psychological trauma that may arise. The project is also sensitive to the need to avoid disrupting service or delivery of treatment at the clinic or hospital. The need to translate the questionnaire into Spanish for the Hispanic community developed into an unintended benefit for the project. A group of 20 women got together to translate each question and make the language culturally relevant. The group enjoyed the process so much that it still meets regularly as a site advisory board. Discussion In the group discussion that followed the presentations, several people commented on the emotional overload that researchers sometimes feel when working on homicide issues. Quite a few present seemed relieved to know that others had the same problems. For example: During a study of female homicide cases at the University of North Carolina, the data collectors, who had been interviewing the police officers and getting graphic detail of murder from the sheriffs, needed to get help from a counselor to overcome the cumulative effect of the information they were processing. The question arose: Does the project have a responsibility to get emotional help for its interviewers and researchers? Although the group did not take a position on this, many commented that such help was an integral part of their work environment. Susan Wilt from New York City agreed that homicide work is very stressful, whether it's reviewing medical examiner records or interviewing family members. Her projects benefitted from monthly group staff meetings with a counselor. They also limited case load where necessary and had frequent debriefing sessions. David Kennedy also agreed that interviewers burn out and feel guilty because they can't help the people they are interviewing. Jackie Campbell uses researchers who have to look at color photos from homicide files. She warns researchers about how they may react. Jackie thought it was important to look at the complete files before interviewing surviving family members to develop empathy. Others commented that researchers should not look at photos if they are going to be objective in their research. Linda suggested that researchers go back to their collaborators who are service providers to get help for the interviewers. Garen Wintemute asked if any of this research ever leads to an escalation of violence for the interviewee? Jackie pointed out that batterers have the latest telephone technology and could listen in on telephone interviews. Becky said that her Board of Advisors is developing a protocol for what to do when the interviewer uncovers a case where there is a risk of immediate harm. Lois Mock suggested that interview researchers should have phone numbers of where the interviewee can go for help. ------------------- Notes From a Collaborative Project: Risk of Serious Injury or Death in Intimate Violence Carolyn Rebecca Block, Illinois Criminal Justice Information Authority 120 South Riverside Plaza, Chicago, Illinois 60645 Abstract Chicago's "Risk of Serious Injury or Death in Intimate Violence" project, a study seeking to identify risk factors for life-threatening injury or death in situations in which a woman is being abused by an intimate partner, could never have happened without the close and continuing collaboration of local medical, public health and criminal justice agencies; experts in research methods, practical interventions, and the epidemiology of intimate violence; plus the funding agency, the National Institute of Justice. This Work in Progress reviews the role that collaboration plays in the project, discusses the difficulties and benefits of collaboration, and outlines lessons that similar projects might learn from this collaborative experience. Who, What, Where, When and Why? When field-level practitioners (e.g.: nurses, police officers) encounter an abused woman, they need to know how factors such as leaving the situation, pregnancy and firearm availability could affect the risk of a lethal outcome. Earlier research tells us who in the general population is most likely to be abused, but not who among abused women is most at risk of serious physical injury. This study will produce high-risk intimate violence profiles (statistically established relative risk factors) for danger of life-threatening injury or death at the hands of either partner within a year of hospital or medical center contact, for African American, Latino and non-Latino white abused women. So that the study results will be relevant to decisions made by field practitioners, we are using a "point of service" sample of women identified as they come into a hospital, community health center or clinic for any kind of contact (regular checkup, traffic accident, and so on). The sampling procedure at each of the six point-of-service sites has been developed in close cooperation with staff at the site, and is an integral part of the site's standard intake process.[1] The Chicago Department of Health "Abuse Screener," routinely administered by clinic or hospital staff, identifies women in the "abused" sample and women in a "comparison" sample. We are making a special effort to include women in the sample who may be in high risk situations but who do not contact shelters or support networks, women who are high-risk but underserved or unknown to official agencies. To further reflect the complexity of field-level decisions, we are using lengthy face-to-face interviews covering changing circumstances and interventions over a retrospective year, and are re- interviewing the sampled women after six months and a year. But the purpose of the "Risk" study is to develop risk profiles for serious injury or death. In order to do this, we must compare situations in which abuse ended in death to situations in which it did not. This study was designed to accomplish such a "lethal/nonlethal" analysis within a finite period of time and with a finite budget (table 1). It is built upon a comparative analysis of the longitudinal point-of-service sample and a sample of everyone (women and men) killed in Chicago by an intimate partner over a two-year period. We will interview two proxies (knowledgeable friends or family) of each victim, with the interview covering a retrospective year. To the extent possible, the same interview schedule is being used for both samples. Role of Collaboration The study was designed around goals and guiding principals developed jointly by the collaborative team over a three-year period. It is an open and flexible collaboration built on consensus, more organic than a legalistic formal partnership. Although additional agencies and individuals are being added to the list of participants as the study progresses, current collaborators include the Chicago Department of Public Health, Erie Family Health Center, Cook County Hospital, Cook County Medical Examiner's Office, Illinois Criminal Justice Information Authority, the Chicago Police Department, and the staff of the six point-of-service sample sites. An active Site Advisory Committee of Erie Family Health Center community members has been instrumental in developing the Spanish translation of the questionnaire and other study materials, and evaluating their cultural sensitivity. Staff at the study sites are vital participants. The sample screener is administered by site staff as part of their usual routine, and a protocol of procedures for each site was developed in close collaboration between project and site staff. The collaborators have found the following research goals to be crucial: 1) to focus sample selection and data analysis on information available to helping agencies, so that the results of the study will be useful for practical decisions; 2) to link data on abused women with similar data on people who have been killed by an intimate partner; 3) to investigate the effect of multiple factors, changing over time, including interventions, individual help-seeking and support networks; 4) to have an adequate sample so that intimate violence profiles will be valid for African American and Latino women as well as nonLatino white women; and 5) to make every effort to sample high-risk but understudied populations (such as expectant mothers and battered women unknown to helping agencies). The collaborators also consider the following guiding principles to be perhaps even more important than the research goals: 1) respondent safety and confidentiality are to be maintained as crucial and primary considerations throughout all aspects of the study; 2) the research is to be integrated with the procedures and culture of each study site, so that site staff are involved in site decisions and problem-solving and encouraged to become project collaborators; and 3) ethical and safety issues have a strong vote, if not a veto, in cases when they conflict with research goals. Unanticipated Benefits and Problems As the project has developed over the years, the maintenance of the collaborative team has required some care and feeding, as we had expected it would. We had not expected, however, some of the benefits and advantages that grew out of the collaboration. The difficulties are the usual team-building issues. It is necessary to devote a great deal of project resources (measured in time spent) to maintaining group communication and in building smaller work groups within the collaborative team. Although we found it helpful to create computerized mailing lists and use modern communication technologies, nothing substitutes for face-to-face meetings and person-to-person letters, FAXes, EMails and phone calls. Since decisions are not made without thorough discussion, they take longer than in a non-collaborative project. However, the final decision is something that all of the project participants can support, and it benefits from the expertise and differing perspectives of all of the team members. For example, the questionnaire was developed by more than 20 people working together over many months, but the end product is an extremely innovative and carefully-developed instrument that, we anticipate, will be used in other research studies. Similarly, the translation of the instrument into Spanish, which is being done interactively by members of the Erie Site Advisory Board, has taken quite a long time to accomplish. Again, however, we expect that the process will produce a Spanish questionnaire that reflects a concern for cultural sensitivity as well as a "correct" translation. Some of the benefits of our collaboration were more surprising. The Erie Site Advisory Board, for example, was neither planned nor anticipated, but grew from the enthusiasm of community members asked to attend a translation "focus group," and continued to develop through the sensitive encouragement of Eva Hernandez, one of original project collaborators. Also, the collaborative team was in large part responsible for the success of the extensive training class for interviewers, which spanned two weeks. In a third example, the original study design called for the use of the Department of Health "standard screener" for abuse. However, when we began working with the staff of each site, we discovered that the standard screener was not actually being used. Project staff and site staff worked closely together in each site over a number of weeks to develop a screening procedure that the site staff could use and support. Thus, the "Risk" project has served as a pilot for universal abuse screening in Chicago hospitals and health clinics. Lessons We Have Learned About Collaboration This project has taught us some techniques for building a collaborative team of diverse groups - public health and criminal justice agencies; researchers, practitioners and community members. I believe, however, that the most important lesson we have learned is that, as David Kennedy pointed out in his keynote discussion, "What Works?" is not the only criterion for a successful research project. "Works" should not be defined narrowly as research results evaluating an intervention or testing a hypothesis. The process of research can not only analyze or describe a problem, but can also be part of the solution. For example, the time and effort devoted to integrating the study within each hospital and health center site was rewarded by the project being a catalyst for each of the sites to develop effective procedures for screening all incoming women for abuse. Though it might have been the "official" procedure, none of the sites had actually been doing universal screening prior to this project. Thus, because it has been so highly collaborative, the existence of the "Risk of Serious Injury or Death in Intimate Violence" project has developed ties among community agencies and individuals working to reduce lethal violence, and given them new problem-solving tools. This benefit of the project was unanticipated, but may turn out to be among the most important products the project produces. NOTES 1. The point of service sample includes women sampled at six sites - the Chicago Department of Health's Roseland Clinic. Two facilities of Erie Family Health Center. The Chicago Women's Health Center, and the Trauma Unit and the Ambulatory Screening Clinic of Cook County Hospital. ------------------------------ The Houston Homicide Project: A Comparison of US and Canadian Findings on Uxorcide Risk for Women with Children Sired by Previous Partners (with Additional Findings on Overall Domestic Homicide) A study of homicide in Houston, 1984-1994. Victoria E. Brewer, Ph.D. Sam Houston State University Kelly R. Damphousse, Ph.D. University of Oklahoma Funded by the National Institute of Justice and the National Consortium on Violence Research.[1] Victoria E. Brewer Kelly R. Damphousse Derek J. Paulsen Daly, Wiseman and Wilson (1997) found, in a sample of 20 cases of uxoricide (wife killing) of women who were mothers of coresident minor children, 55% were women with minor children sired by a previous partner, although comparable women represented only 7.3% of the population at large. --Approximately 20% of Houston married couple households with coresident children under the age of 18 include stepparents/children. --The Daly, Wilson, & Wiseman research highlights the increased risk of homicide to women in such families in Hamilton, Ontario. --Research Question: Are women in Houston, TX, with coresident minor children, at greater risk of being murdered when their present partner is not the father of those children? Daly, Martin, Karen A.Wiseman, and Margo I. Wilson. "Women with Children Sired by Previous Partners Incur Excess Risk of Uxorcide." Homicide Studies, Vol. 1 No. 1, February 1997:61-71. --"Male sexual proprietariness, aroused by women's efforts to leave unsatisfactory marriages or by adulterous or potentially adulterous interactions with other males, is by far the leading ostensible factor in violence against wives, especially lethal violence. An evolutionary psychological perspective suggests that this motive is an especially powerful one because the fitness of our male ancestors depended crucially on sexual and reproductive control of women in a social milieu of rival men. The effect of male rivalry that matters in evolutionary time is differential paternity, and it is therefore not surprising that marital conflict is in general reduced by the presence of children but exacerbated when those children are the products of prior unions." (Daly & Wilson, 1996). Daly, M., & Wilson, M.I. "Evolutionary Psychology and Marital Conflict: The Relevance of Stepchildren." In D.M. Buss & N. Malamuth (Eds.), Sex, Power, Conflict: Feminist and Evolutionary Perspectives pp. 9-28. New York: Oxford University Press. --In Hamilton, Ontario . . . 34% of all uxoricides were of women with coresiding minor children sired by a previous partner. comparable women represent only 7.3% of the population at large. --In Houston, Texas . . . 41% of all uxoricides were of women with coresiding minor children sired by a previous partner. Yet comparable women represent less than 20% of the population at large. --In Hamilton, Ontario . . . Female-initiated separation was a motivational factor in 56% of the cases of uxoricide in Hamilton. --In Houston, Texas . . . Female-initiated separation or jealousy were motivational factors in 56% of the cases of uxoricide. "Mother of 4 shot to death....by an estranged boyfriend, as the woman's 10-year-old child looked on...forced his way in after she came home from a date about 2 a.m......." 3/11/86 "A capital murder defendent, accused of killing his ex-girlfriend and her 9-year-old daughter, .... her 7-year-old son also was wounded...her other children hid in a closet during the shooting..." 1/6/93 "A man who spent months harrassing his ex-common-law wife was sentenced Friday to die by injection for a 14-gunshot shooting spree inside her home that killed her and her brother and wounded two bystanders....." 11/13/93 Source: Houston Chronicle --17% of all victims of spousal homicide had minor children. --The male partner was the offender in 79% of the cases where there were coresident minor children. The male partner was the offender in 50% of the cases where there were no children. --70% of offenders in couples with minor children were charged with murder or capital murder. 45% of offenders in couples with adult children were charged with murder or capital murder. --In 18% of the cases of spousal homicide, the offender (predominantly male) also committed, or attempted to commit, suicide. This was the case in 49% of the couples with adult children. --In 22% of the cases of spousal homicide, there was a 5-10 year age gap between partners. In 27% of the cases, there was an age gap of +10 years. --62% of victims were in registered or common-law marriages. --4.7% of victims were in coresiding homosexual relationships. NOTES 1. Findings and conclusions of the research reported here are those of the authors and do not necessarily reflect the official position or policies of the U.S. Department of Justice or the National Consortium on Violence Research. ------------------------------ Partner Homicide in Massachusetts, 1991-1995: A Collaboration Between Advocates and Academics Linda Langford Harvard School of Public Health, Department of Health and Social Behavior 677 Huntington Avenue, Boston, MA 02115 Project Description "Partner Homicide in Massachusetts, 1991-1995" is a collaboration between Peace at Home (PAH), a human rights agency focused on violence against women, and the Harvard School of Public Health (HSPH). Using multiple data sources, we are constructing a database of all homicides related to intimate partner violence in Massachusetts from 1991 to 1995. This study has two objectives: 1. to count accurately and characterize homicides related to partner violence in the state over the 5- year period, and 2. to contrast the number of partner homicides found by our study with the number reported by the FBI's Supplementary Homicide Report during the same time period. Multiple data sources are used to identify and verify cases and to gather additional case information. Sources used for case identification and verification include news articles, correspondence with district attorney's offices and advocacy agencies, Supplementary Homicide Reports, and discussions with police. Data sources used to gather information about the victims, offenders, and homicide circumstances include death certificates, news articles, the criminal justice information system, the registry of civil restraining orders, district attorney's offices, and police officers. How Did the Collaborative Project Come About? Unlike many collaborative research studies, this project was initiated by advocates. Borrowing a strategy from her past work as a human rights activist, Stacey Kabat, director of Peace at Home, began to record the domestic violence homicides in Massachusetts. By documenting the lethality of battering relationships, she hoped to help policy-makers and the public understand the dire situation of battered women. Three years into the process of recording the homicides, Ms. Kabat sought the help of researchers at HSPH to help systematize, organize, and present the data. Since HSPH joined the effort, we have expanded the project to collect two additional years of data and access additional data sources to increase the amount of information available on each case. Benefits of the Collaboration It is very gratifying to work on a project that was initiated by the advocacy community. Although this project is not the first productive advocate-academic collaboration concerning domestic violence in Boston, it provides an additional opportunity for the advocacy and research communities to work together on a project of mutual interest. PAH is very well respected in the community and has the contacts and clout to use the report to advocate for policy changes that will make a difference for battered women in Massachusetts. Both groups have much to gain from this project. PAH gets research expertise, the stamp of legitimacy of the Harvard name, and a completed report. Harvard gets access to PAH's data, connections, and political and substantive expertise. Because of its mutually beneficial nature, motivation has been very high on both sides, which has helped tremendously in overcoming obstacles, making cooperative decisions, and finding the time for collaborative work. Problems with the Collaboration On the whole, this collaboration has been very successful. Although we were asked to discuss "problems," it would be more accurate to characterize the following as issues that emerged during the course of this particular collaboration that may provide insights for future efforts. Initially, there were unexpected "culture clashes" between the research and advocacy worlds. For example, the majority of victims on Peace at Home's list were intimate partners, but they had also included some cases of homicide among other family members. Thus we were faced with the need to decide on a consistent case definition of "domestic" for the purpose of the project. This process was surprisingly painful. It is distinctly "research-like" thinking to rule cases in or out based on a study definition--although second nature to the researchers, this way of thinking was not familiar to PAH staff members. Their perspective was that each victim had died a horrible death at the hands of a family member, and excluding anyone from the list seemed to trivialize their death. Before we could come to an agreement about the case definition we would use, both parties had to come to understand the other's way of thinking about the cases and extensive discussion was needed to define our population of interest. Another issue that emerged was a difference in expectations about how much time it takes to do research. Originally PAH staff thought that HSPH could produce a report in a few months. The question of project duration was further confused by the mutual decision early on to expand the study beyond the original parameters. Even with the project expansion, the expected time to completion was far greater than PAH had expected and was the cause of some frustration, as they were understandably anxious to release the report. Again, this topic required discussion and negotiation to resolve. There were numerous nitty gritty issues that were also important to the success of the collaboration. For example, in constructing the database, we took care to choose software that is compatible with the computer systems at PAH so they will be able to use the final database without purchasing new software. Lessons Learned For any given project, it is obviously best to discuss and clarify as many issues as possible up front. For example, our initial meeting included an explicit agreement about what each party expected to get out of the collaboration. Inevitably, unanticipated issues will arise. It is important to pick your battles carefully. While methodological choices must meet the standards of good research, other decisions are not as critical. Flexibility by both sides is important. When areas of mutual misunderstanding occur, it is important to devote sufficient time to discussion. Often these issues can bring up strong feelings, and many meetings may be required for resolution. A commitment to listening and patience with the process are crucial. Successful resolution of early problems helps build trust, so later issues are handled more easily. I wish to give special thanks to my collaborators, Nancy Isaac, formerly with the Harvard School of Public Health and currently with Northeastern University School of Law, and Stacey Kabat of Peace at Home. ------------------------------ Methodological Challenges to Evaluating The Brady Handgun Violence Prevention Law Daniel W. Webster Johns Hopkins Center for Gun Policy and Research Contamination of Law Effects To estimate the effects of public policies such as the Brady Law on violent crime, it common to use quasi-experimental designs that compare intervention groups with no-intervention groups. Although there are known threats to validity associated with these designs, such threats can often be evaluated, and in some instances controlled, with appropriate statistical procedures. When the intervention being tested is a law, one validity threat that cannot easily be remedied by sophisticated analytic techniques is contamination, i.e., when likely comparison groups, who are not necessarily the target of the law, are nonetheless affected by it. In the case of the Brady Law, it may seem logical to compare violent crime trends in states in which gun dealers were forced to begin conducting background checks of potential handgun purchasers (referred to hereafter as "Brady-affected") with states that had been screening handgun purchasers prior to Brady. But there are reasons to believe that states that were not targeted by the new screening provisions of the law could nevertheless benefit as much as, or possibly even more than, states that were the target of these provisions. Why? Because violent crime is highly related to population density and tends to be much higher in urban than in rural areas. Because handgun availability tends to increase the lethality of violent altercations, the effects of handgun regulations (or lack thereof) are likely to be most salient within urban areas. This is obviously an important reason large cities and more urban states tend to enact stricter gun control laws. Most of the Brady-affected states are more rural than the states that required background checks of handgun purchasers prior to Brady. Thirty-two percent of the population of Brady-affected states reside in rural areas compared with 18% in the other states. Gun density is likely to be higher within Brady-affected states because these states are more rural and gun ownership is twice as high in rural areas than in the largest cities (Cook and Ludwig, 1996). Brady-affects states are also in regions of the U.S. in which gun ownership is highest and, in some states guns are rather ubiquitous. The number of licensed gun dealers per-capita in Brady affected states is 81% higher than in other states (79 vs. 44 FFLs per 100,000 population). The higher density of firearms in Brady-affected states should make it easier than in other states for proscribed handgun purchasers to obtain handguns through non- regulated secondary sales, gun shows, or theft. Supporting this claim is the fact that street prices for guns are significantly higher in states with the greatest restrictions on gun purchases (Cook, Molliconi, & Cole, 1995). By requiring background checks in all states, dramatically reducing kitchen-table gun dealers, and enhancing the regulation of gun dealers nation-wide, the Brady Law (along with a more concerted effort by ATF to combat illegal gun sales that began close in time to the implementation of Brady), should curtail both intra- and interstate gun trafficking. It is such trafficking that has plagued states with high-population-density by weakening the effectiveness of their state and local gun control laws. Thus, these more urban states that were not the target of Brady's handgun screening provisions stand to gain from reductions in illegal gun transactions. Problems with Other Comparisons: Inequivalence of Gun and Nongun Homicides Theoretically, gun-crime vs. non-gun-crime comparisons are appealing strategies for isolating gun policy effects and examining potential weapon substitution effects. But these comparisons may need to be narrowed in order to avoid an apples vs. oranges problem. Non-gun homicides differ somewhat from gun homicides in terms of ages of offenders and victims, victim-offender relationships, and circumstances surrounding the homicide. Demographic Differences Compared with non-firearm homicides (NFHs), firearm homicides (FH) offenders and victims are more likely to be male (84% vs. 66%), Black (54% vs. 42%), and young (55% vs. 30% are 15-29 years of age) (Table 1). The age distribution of NFH victims is much flatter than that of FH victims. The primary difference in the offender age distributions of FHs and NFHs is that FH offenders are almost twice as likely to be in their teens and NFH offenders are more likely to be in their thirties. Differences in Type of Homicides NFHs are twice as likely as FHs to involve intimate partners or family members (24% vs. 12%) and somewhat less likely to have the victim- offender relationship be unknown [1] (Table 2). FHs and NFHs are similar in that about 4 of 10 for which the circumstances are known are related to "arguments." But FHs are twice as likely as NFHs (36% vs. 15%) to involve either robberies, gang feuds, or involvement in illegal enterprises (e.g., drugs, gambling, prostitution). Nearly all homicides involving gangs and/or illegal enterprises are committed with firearms. Differences in Trends Because of the differences outline above, it is not surprising that homicide trends have varied by weapon type. Since the mid 1980s, age-adjusted rates of FHs and NFHs have followed two distinct patterns in the United States. Age-adjusted FH rates rapidly declined from 1980 to 1983 and then rose sharply from 1985 to 1993. In contrast, age-adjusted NFH rates have been on a downward trend since the early 1980's. (Figure 1). There is actually a slight negative correlation (r=-.37, p=.11) between age-adjusted rates of FHs and NFHs within the U.S. from 1975 through 1994. This negative correlation (r=-.59, p=.07) is much stronger for the period 1985- 1994. But looking a correlation coefficient can obscures the somewhat more complex temporal relationship between FHs and NFHs. The year-to- year change, i.e., difference between the age-adjusted homicide rate in a given year and that of the previous year (Yt-Yt-1), for FHs and NFHs were positively correlated (r=.45, p=.05) during 1975-1994. Confounding Interventions One can statistically control for any pre-law differences in trends. However, because gun and nongun homicides differ in the ways just described, interventions targeting certain types of crime (e.g., gang- related, drug-related, domestic violence) that occur near in time to the gun law could make it harder to attribute any observed differences in weapon- specific changes solely to gun policies. Identifying confounding interventions is a common problem in the evaluation of gun policies such as the Brady Law. Deciding which policies to account for in the analysis is not clear-cut. Just determining which policies were implemented in a state at what time is no small task. This is even more difficult at the local level, and it is probably at the city- or even the neighborhood-level that most crime-fighting interventions occur. These local initiatives are often the result of changes in police practices and neighborhood revitalization rather than legislation, and therefore, can be difficult to track. Interactions between policies could also be important, however, testing interactions introduces its own complexities. Specifying Law Effects Estimating the appropriate relationship between policy implementation and full policy effect can also be problematic. Is the effect assumed to be immediate or delayed, a yes/no step function, linear or logarithmic over time? There are reasons to believe that the full effect of the Brady Law may not have been realized immediately upon its implementation. First, there was great room for improvement in the systems for searching criminal records in Brady-affected states when the law went into affect, but progress has been made on that front. Second, many people who were proscribed from legally possessing a handgun when Brady was implemented, already had access to one or more handguns through other means. Over time, these individuals may be less able to replace those guns through due to the provisions and enforcement of the Brady Law, in addition to other gun regulations. If there was a surplus of guns available in illicit markets when Brady was implemented, it may take some time before the law, (in conjunction with other efforts targeting illegal gun markets) is able to reduce the supply, and ultimately the "consumption" of handguns by juveniles and convicted felons. Thus, it may be several years before the full effects of the law are realized, but estimating the appropriate function for assessing the law's effects is no easy task. Suggested Direction for Evaluating Brady Given the difficulties mentioned above, we should consider more targeted studies as alternatives to attempts to estimate overall effects on violent crime. For example, one could examine changes in time lags from retail sale-to-crime using tracing data because the waiting period is intended, in part, to prevent impulsive crimes, and because the law should primarily affect new guns. Increasing lags between dealer sales and crime involvement would suggest that greater controls over handgun sales due, in part, to Brady have reduced the availability of new handguns to high- risk groups. Using survey data from state prisoners, one could also look for changes in the number of persons convicted of committing violent crimes that had a previous felony conviction as some evidence that the background checks are having a preventive effect. These suggestions, of course, have their own limitations. Tracing data is generally unrepresentative of all guns used in crime. Reductions in crimes committed by convicted felons could be due to increasing prison sentences and parole policies. It may be that these evaluation alternatives are no better than the traditional quasi-experimental approach using regression in attempt to control for confounders. As scientists we be honest about our ability to accurately answer certain questions. Because the effects of Brady on violent crime are not expected to be large, immediate, or highly targeted, I suspect that any "state-of-the-art" evaluation of the Brady Law will still leave objective scientists quite uncertain about the effects of the law on violent crime. NOTE 1. Usually, the relationship is unknown in homicide cases because the police were unable to make an arrest. Because crimes involving strangers are more difficult to solve, homicides which do not lead to an arrest are probably more likely than other homicides to involve strangers. References Cook PJ, Ludwig J. Guns in America: Results of a Comprehensive National Survey of Firearms Ownership and Use. Washington, DC: Police Foundation, 1996. Cook PJ, Molliconi S, Cole TB. 1995. "Regulating Gun Markets." Journal of Criminal Law and Criminology 86(1):59-92. ------------------------------ Evaluating the Brady Act and Increasing The Utility of BATF Tracing Data Gary Kleck School of Criminology and Criminal Justice Florida State University Tallahassee, FL 32306-2025 Abstract Pooled cross-sections analyses provide a model for evaluating a particular intervention, one that reduces some of the serious shortcomings of univariate interrupted time series case studies. Annual county-level data on crime and crime-determinants are available for over 3,000 U.S. counties. Dependent variables would include rates of total homicide, gun homicide, and nongun homicide, as well as total, gun and nongun robbery and aggravated assault, and total rape. The main independent variable of interest would be a dummy variable identifying county-years that are located in "Brady states," i.e. those that had new background checks as a result of the Brady Act, in 1994 or later. To check to see if the law's effect is being confused with other unmeasured variables changing around 1994, analyses should also be performed on property crime rates. Comparison of data from the 1991 Survey of State Prison Inmates and a post-1994 survey could also provide information on whether fewer criminals acquired guns from licensed dealers after Brady. BATF trace data could be used in future, though not in connection with a Brady evaluation, to monitor trends in the share of crime guns that were purchased by their criminal users from a retail dealer. This could be done by getting law enforcement agencies requesting gun traces to note the name of any criminal suspect found in possession of the firearm, and looking for name-age-sex-race matches with the gun's previous retail purchasers, as indicated in ATF Form 4473 records. Introduction The Brady Handgun Violence Prevention Act (hereafter, the Brady Act), which became effective on February 28, 1994, is the most significant piece of federal firearms control legislation passed since the Gun Control Act of 1968. The Brady Act, during its first four and a half years, imposes a waiting period of five business days before a handgun may be purchased from a licensed dealer, and requires gun dealers to check with law enforcement authorities to see if the prospective buyer was disqualified under federal law from buying a gun, especially whether they had been convicted of a crime. After November 29, 1998, the waiting period requirement is dropped and the law's central gun control mechanism becomes an instant background check on persons seeking to purchase guns of any kind, not just handguns, from licensed gun dealers (U.S. Congressional Research Service 1994). By way of full disclosure, I should note that this is basically the primary gun control measure I endorsed in my 1991 book, Point Blank (Kleck 1991, pp. 432-440). The Brady Act exempts those 24 states that already had their own gun purchase background checks in place before 1994, and thus introduced new background checks into the remaining 26 states, which included about 39% of the U.S. population. The main limitations of Brady are twofold. First, it only restricts gun acquisitions through licensed gun dealers. The best available evidence indicates that about 73% of gun acquisitions by felons are made via routes other than purchases from retail outlets, such as theft or purchases from friends and relatives (U.S. Bureau of Justice Statistics 1993, p. 19). Second, with respect to the 27% of criminals who, pre-Brady, acquired guns from retail dealers, an unknown but presumably nonnegligible share could also obtain guns from nondealer sources. The potential for success of any background check limited to dealer purchases depends heavily on how large this share is. One purpose of this paper is to outline some ways to evaluate the Brady Act's impact on violence. Given the location of these meetings, and who our hosts are, I also use this opportunity to outline how gun tracing activities of the Bureau of Alcohol, Tobacco and Firearms (BATF) could have been used in an evaluation, had tracing data been gathered differently. The point is not to cry over spilt milk, but to suggest how improvements in these data might still be useful in evaluating laws not yet passed, as well as enlightening us with respect to patterns of criminal acquisition and use of guns. How the Brady Act Is Intended to Work According to proponents, the purpose of the Brady Act is to reduce violence, and more specifically, to reduce gun violence. This presumably implies that in any given violence category, e.g. homicide, there should be a reduction in both total violence and gun violence. Few would regard it as a success if the law produced 500 fewer gun homicides but 500 more nongun (e.g. knife) homicides as a result of weapon substitution. Proponents stress that benefits are most likely to be evident in the homicide category, since many gun control advocates will concede that gun laws may not reduce the frequency of crime, but can reduce the lethality of crimes, reducing the share of violent crimes that result in death, thereby reducing the homicide rate (Cook 1991). Further, given that both fatal gun accidents and suicides disproportionately involve persons with criminal records (Kleck 1991, Chapters 6 and 7), it also possible, though given less emphasis by proponents, that the Brady Act could reduce these kinds of deaths as well. Figure 1 illustrates the most straightforward way in which the Brady Act could achieve these effects. By requiring gun dealers to seek a background check on prospective gun buyers, a mechanism is provided by which the dealers can know whether the prospective buyer is legally forbidden from purchasing a gun under previously existing gun law. The record check would at minimum scan computerized criminal history files and can also make use of records, of varying degrees of completeness, concerning other categories of persons prohibited from acquiring guns. In practice, however, few permanent denials, as distinct from temporary "administrative" denials due to delays in accessing records, are for any reason other than a criminal conviction. During the first year of Brady's operation, about 49% of denials were for a criminal record, 39% for administrative reasons, 8% for traffic offenses (presumably erroneous denials), and less than 5% were for other ineligible categories (fugitives from justice, "mental defective," mental patients, dishonorably discharged veterans, persons who have renounced their U.S. citizenship, and illegal immigrants) or due to restraining orders (U.S. General Accounting Office 1996). Excluding the presumably temporary administrative denials and erroneous traffic offense denials, 91% of denials were for criminal convictions. Thus, to the extent that Brady will be effective, it should work largely because it denies some convicted criminals access to guns through licensed gun dealers. In the law's first year of operation, there were an average of about 3,725 total denials for all reasons per month in Brady states, with about 1,814 of these due to criminal records, implying about 22,000 denials for criminal record per year. The estimated share of background checks resulting in some kind of denial was about 4.3%, and the denial rate based on a criminal record was about half that (U.S. General Accounting Office 1996; see also U.S. Bureau of Justice Statistics 1997). Critics have often noted that the Brady Act's impact on the extent to which criminals are denied guns can be overstated if one assumes that all those denied a dealer purchase fail to get a gun from any source. It can, however, also be overestimated if one assumes that, when criminals are completely prevented from getting a gun from any source, that they are left without a gun. Many, probably most, criminals (and, for that matter, noncriminals) who seek to buy a gun already have at least one other gun. This is implied by two facts. First, about 77% of gun-owning households in the U.S. own more than one gun, with an average of at least four guns per gun-owning household (Kleck 1991, pp. 54-55). Second, among felons who stole guns and kept one for themselves in Wright and Rossi's (1986) ten-state prison sample, only 37% reported keeping the gun because they did not have one at the time, suggesting that most gun thieves already had at least one gun at the time of their theft. Criminals, like other people, often purchase guns because they want yet another one, not because they do not have one. Thus, blocking the most recent attempt to get a gun does not necessarily imply a criminal without a gun. In sum, among criminals who (1) tried to buy a gun from a dealer, (2) were denied a gun because of a background check revealing a criminal conviction, and (3) did not already have a gun, (4) some unknown fraction would fail to circumvent the dealer controls by acquiring a gun from a nondealer source, hopefully resulting in a level of gun possession among convicted criminals that is lower than it would have been without the Brady Act. This "unknown fraction" will be large only to the extent that prospective criminal gun buyers do not know anyone with a gun who is willing to sell them one despite their criminal record, and to the extent that the actual price (including sale price and the cost of search time) exceeds that which the criminal is willing and able to pay. Reductions in criminal gun possession, in turn, will have a violence- reducing effect only to the extent that gun possession among criminals has a net positive effect on violence rates. No research has separately assessed the impact of criminal gun possession levels on violence rates, and the best available research that lumps criminal and noncriminal gun possession together indicates no net effect of gun levels on violence rates (Kleck 1991; 1997; Kleck and Patterson 1993). Nevertheless, mixed evidence still holds open the possibility that criminal gun possession levels do have the net violence-increasing effects that commonsense leads many to expect (Kleck 1991; 1997; Cook 1991). For example, evidence on individual incidents of violence indicates that offender gun possession has a net positive effect on the likelihood that the incident will result in the victim's death (Kleck and McElrath 1991; Kleck 1997; Cook 1991). Case-control research on households in high-crime areas, where a large share of the population has a criminal record, indicates that households with guns are more likely to experience a homicide victimization (Kellermann et al. 1993). The associations found in case-control studies may indicate only that the same factors that put people at higher risk of violent victimization (e.g. drug dealing or membership in a street gang) also motivate many people to acquire guns for self-protection, or they may indicate a genuine causal effect that is, however, limited to households with an unusually high potential for violence, in contrast with net victimization-reducing effects of gun ownership among other gun-owning households (Kleck and Hogan 1997). Other evidence indirectly suggests that criminal gun levels increase some violence rates, because it directly indicates some violence-reducing effects of state background check laws similar to the Brady Act. Mixed evidence indicates that gun laws requiring a license to own guns or a permit to purchase them, both of which entail background checks for criminal convictions, may reduce both homicide and suicide rates, though they show no apparent effect on rates of aggravated assault, robbery, rape, or gun accidents (Kleck and Patterson 1993, p. 274). This specific mix of findings supports the view that gun laws with background checks do not reduce the frequency of violent acts but may reduce the share that are fatal. Since these laws target only high-risk prospective gun buyers, primarily convicted criminals, the most likely mechanism by which these effects are achieved, assuming they are genuine, would seem to be a reduction in gun possession among criminals. It should be stressed, however, that evidence for an impact of background check laws is by no means consistent. Evaluation Strategies How might the hypothesized effects of the Brady Act be empirically detected? It may be worth addressing first the evidence that has most frequently been cited as bearing on this issue. President Clinton, Attorney General Reno, and many lesser proponents of Brady have cited estimates of the number of denials as measures of the law's effectiveness (Wall Street Journal 9-17-96, p. A18). This is at best misleading, since it is unclear whether a large number of denials is an indicator of success. Certainly, an effective Brady Act should lead to a large share (hopefully close to 100%) of convicted criminals who seek to buy guns from dealers being denied. A large number of denials, however, could be seen as a failure of the law's deterrence function, since it necessarily implies that large numbers of criminals attempted to get guns from dealers despite the law. Apparently the word has not yet gotten out to the entire criminal population that they cannot buy guns from dealers. Indeed, a completely effective Brady law might well be accompanied by no denials, if criminals were completely deterred from even attempting to get guns from dealers. And, as noted previously, the fact that a criminal was blocked from buying a gun from a dealer does not imply that he was prevented from acquiring a gun from any source. Proponents might consider the double-edged nature of their use of denial figures as indicators of effectiveness. These figures almost certainly will decline as awareness of the law's provisions reaches a larger share of the criminal population and fewer criminals try to get guns from dealers. Indeed, data already indicated decreasing denials by the second year after the law became effective (U.S. Bureau of Justice Statistics 1997, p. 1). Unless proponents want to have to explain why these declines do not indicate declining effectiveness of the background checks, they would do well to concede that the volume of denials indicates nothing about the law's effectiveness. Macro-level Analysis of Violence Data A well-established methodology has already been applied, with considerable sophistication, to evaluation of gun laws - pooled cross- sections/time-series analysis (PCTS). With this approach, analysts take advantage of data on variation in crime/violence rates both across space and time. Thus, Marvell and Moody (1995) studied annual crime rates over 24 years, for each of the states, to evaluate the impact of laws providing longer prison terms for felonies committed with guns. This type of law, favored by the National Rifle Association (NRA) as an alternative to gun control, was found to be ineffective. On the other hand, when Lott and Mustard (1997) evaluated another approach favored by the NRA, laws making it easier for noncriminals to get permits to carry guns in public places, they concluded that the laws reduced crime, presumably because criminals perceived greater risk from victimizing potentially armed victims. While I have my doubts about Lott and Mustard's claims of huge deterrent effects, I accept their empirical observation that passage of the laws was generally followed, for whatever reasons, by crime drops. Using univariate interrupted time series designs (ITSD) applied to fewer than ten counties, McDowall and his colleagues (McDowall et al. 1992; 1995a) had drawn precisely opposite conclusions with respect to both types of laws. They studied sentence enhancement laws in just three states, while Marvell and Moody studied them in all 49 states where they existed. Likewise, McDowall and his colleagues (1995a) studied nondiscretionary gun carry laws in just seven counties (grouped into five areas), without offering any rationale why they analyzed only these few areas, while Lott and Mustard analyzed all 3,000 counties for which requisite data were available. In addition, Lott and Mustard measured and statistically controlled for many other potential determinants of crime/violence rates. For these reasons, I place greater weight on the Marvell-Moody and Lott- Mustard findings than I do on the McDowall et al. findings. It is clear that PCTS and ITSD approaches have yielded diametrically opposed findings with respect to the effectiveness of these two sorts of gun controls. Critical flaws in the univariate ITSD approach have been identified, without effective rebuttal from ITSD proponents (Kleck et al. 1993; Polsby 1995a; 1995b; Britt et al. 1996a; McDowall et al. 1995b; 1996b; Britt et al. 1996b). Therefore, the PCTS approach should be preferred. A PCTS evaluation of the Brady Act could use annual, county-level crime and violence data covering virtually all 3,141 U.S. counties and county- equivalents, over the period from 1977 to 1995 (and later, as more recent data become available), available on public use computer tapes. The Inter- university Consortium for Political and Social Research (ICPSR) provides tapes with county-level FBI crime data, while Mortality Detail File tapes from the National Center for Health Statistics can provide county-level counts of total homicide, gun homicide, total suicide, gun suicide, and fatal gun accidents. The FBI's Supplementary Homicide Reports data on individual homicide incidents can even allow separate estimates of handgun homicide rates and longgun homicide rates, since the data make a gun type distinction not consistently maintained in mortality data. The unit of analysis would be the county-year, with crime/violence rates varying across both counties and years. Lott and Mustard (1997) have also identified sources of annual county-level data on some other potential determinants of violence rates, such as age, sex, and race distribution of the population, arrest rates, per capita income, and welfare expenditures. The main test of the Brady Act's impact on violence rates would be whether counties located in Brady states (i.e. those that did not already have their own pre-Brady background checks) should have, other things being equal, lower violence rates in 1994 and later years than either counties in nonBrady states or Brady counties prior to 1994. In a regression analysis, there would be a dummy variable indicating whether a county-year was in a Brady state and in 1994 or later. If the Brady Act was effective, the coefficient for this dummy variable should be negative and statistically significant. To check whether the law's effect was being confused with other unmeasured variables changing around 1994, analyses should also be performed on property crime rates. Since limits on guns should have little impact on crimes not involving guns, if the law is effective it should show impacts on gun crime rates but no impact on either nongun violent crime rates or on property crime rates. The chief shortcoming of the PCTS approach is that, although it clearly does better than the univariate ITSD approach in explicitly controlling for confounding factors, i.e. other factors that may have changed about the same time as the Brady Act went into effect and that may also have influenced violence rates, it still does not do a very good job. County-level data are not available, between dienniel Census years, for more than a handful of potentially confounding factors, and thus the best PCTS analyst has only limited ability to rule out alternative explanations of the violence patterns. This problem can be ameliorated to a modest degree by more narrowly specifying where the intervention's impact should be observed. Obviously gun homicide should be reduced if the law were effective, while nongun homicides should either be unaffected or increase somewhat as a result of weapon substitution, with similar patterns possible for aggravated assault, robbery, and suicide. Some further refinements are possible as a result of the two-phase aspect of Brady. During the February 1994-November 1998 period, the background checks apply only to handguns, while after November 1998, they apply to all guns. Thus, any impact should be evident only with handgun homicide during 1994-1998, and not with longgun (rifles and shotguns) homicide, while effects should be evident with violence involving any type of gun after 1998. Presumably there are a narrower set of confounding factors that could produce these specific patterns than could produce any old violence reduction. Therefore, if these patterns were observed it would increase our confidence that the Brady Act was responsible for them. Unfortunately, even these refinements will not rule out some alternative explanations of violence reductions. Homicide began to decline in the U.S. in 1991, and these declines were proportionally greater in the gun homicide category than in the nongun category. Yet, since they began years before the Brady Act's effective date in 1994, they could not be attributed to that law. Further, these declines disproportionately occurred among homicides linked with street gang combat, drug dealing and the robbery that so commonly is linked with drug market violence (U.S. FBI 1996, pp. 21, 58). About 90% of homicides committed in connection with drug dealing and street gang activity are committed with guns (p. 20), implying that changes in the frequency of such homicides will be observed almost exclusively in the gun homicide category, regardless of the causes of the changes. Thus, even if the Brady Act had no impact at all on homicide, if any nonBrady factors produced declines in these specific types of homicide, one would expect bigger drops in gun homicide than in nongun homicide. Indeed, since handguns claim an even larger share of these homicides than of other homicides (U.S. FBI 1996, p. 20), one would even more specifically expect larger drops in handgun homicides than in homicides with other types of guns. In sum, even the more refined focus on gun homicides or handgun homicides will not rule out alternative explanations that revolve around nonBrady factors reducing drug-related and gang- related violence (e.g., more effective and widely available drug treatment, community policing, a decline in drug market-destabilizing Drug War activity, or whatever else one might favor). A PCTS analysis can therefore provide relevant but far from decisive information on the impact of the Brady law. Individual-level Survey Data From Criminals There may also be some fairly direct individual-level evidence available on the way that criminals acquire guns, information that is, or should eventually be, available for periods both before and after February 1994. The Survey of State Prison Inmates (SSPI) was conducted in 1991, and asked a large nationally representative sample of prisoners whether they had owned guns before being sent to prison, and where and how they obtained their guns. The questions distinguished between such licensed sources as "gun shop or store" and "pawnshop" from unlicensed sources such as a crime victim, friend, family member, fence, black market source, or drug dealer. When and if similar data become available from a post-Brady survey, one could separately analyze recently incarcerated inmates who owned guns prior to imprisonment, and examine their sources of guns. If the Brady law worked as intended, the share of guns acquired from retail sources likely to be licensed should have declined. On the other hand, if no such decline were observed, it would make it less plausible that the law was responsible for any observed declines in gun violence, and certainly undercut the idea that it worked by discouraging retail gun acquisitions. Better still, an improved SSPI could even ask those felons who were denied guns as a result of Brady background check how they responded. Did they already have a gun, making the denial irrelevant to their armed status? If not, did they simply do without a gun, or did they get one from an unlicensed source? If the latter, what kind of source did they get it from? How Better BATF Gun Tracing Data Could Have Helped BATF performs over 90,000 "gun traces" a year. Many gun traces fit the following pattern. A police department recovers a gun connected with a crime, and asks BATF to trace it, providing them with the gun's manufacturer, model, serial number, and other identifying characteristics. BATF contacts the manufacturer (or consults their own archive of records from out-of-business manufacturers) to discover to whom the gun was sold. If it was sold to a wholesaler or distributor, this source is contacted to determine the licensed dealer to whom the gun was sold. If the trace is successful, BATF staff finally reach (sometimes after going through additional intermediate gun dealers) the retail dealer who sold the gun to its first retail buyer. BATF then asks the dealer to identify the individual purchaser, based on the Form 4473 Firearms Transaction Record that must accompany every transfer of a gun from a licensee to a nonlicensee. These forms are kept by the licensed dealers, but are available for inspection by the BATF. Normally, this is as far as the gun can be traced. BATF cannot trace guns first sold at retail before the Gun Control Act of 1968 went into effect, or guns lacking a serial number. At least two factors limit the utility of trace data for criminological purposes such as evaluating the Brady Act's impact. First, the trace data say nothing about the criminals who used the traced guns, and virtually nothing about the crimes in which the guns were used. Second, the guns traced are not representative of crime guns in general. In 1994 there were about 543,000 violent gun crimes known to the police (U.S. FBI 1995, pp. 18, 29, 32), and about 9,830 BATF traces of guns linked with violent crimes (U.S. BATF 1995), implying that less than 2% of violent gun crimes (homicides, assaults, and robberies) known to the police, and less than 1% of all violent gun crimes (reported to police or unreported), result in a trace. Further, this small sample of crime guns traced is not a random sample, nor is there anything in the method of "sampling" that can insure that it is representative of either all crime guns or of those recovered by police (U.S. Congressional Research Service 1992). Instead, the composition of the sample is determined by law enforcement agencies' preferences as to which guns they choose to have traced. Further, 57% of traces in 1994 were initiated by BATF itself (U.S. BATF 1995), which means that the composition of this sample is largely determined by the enforcement emphases and priorities of a single law enforcement agency. In this light, it is not surprising that samples of traced guns can be radically different from the entire population of crime guns recovered by police. For example, direct comparison of trace samples with local populations of crime guns recovered by police indicate that the trace samples overrepresented "assault weapons" by a factor of at least four. While these weapons accounted for, on average, only about 2% of all guns recovered by police, they claimed over 8% of those traced by BATF (Kleck 1991, Chapter 3; Kleck 1997, Chapter 4; U.S. Congressional Research Service 1992). It is likely that priorities in seeking gun traces are influenced by whether the guns fall into categories subject to a high degree of news media publicity and political attention. Thus, changes over time, or differences across areas, in the character of traced guns may reflect shifts or differences in media attention and political focus rather than actual changes or differences in the types of guns being used in crime. The forms that law enforcement agencies and BATF currently use to initiate traces include information largely confined to the gun itself. A single item asks for the type of crime to which the gun was connected. About 84% of traces are linked with weapons and drug offenses, possessory offenses in which the identity of a criminal would ordinarily be known to the police at the time the gun was recovered (U.S. BATF 1995). Either the gun was recovered at the time a suspect was arrested or it was seized as a result of a search of a premise linked with a known suspect. And certainly in some of the remaining 16% of the trace cases, a suspect had already been identified by the time the trace was requested. Therefore, police would ordinarily be able to provide identifying information about the suspect linked with most crime guns on which traces are requested. BATF, however, does not solicit this information. This information could be obtained through nothing more complicated than modifying the trace request form to include a few additional items. Immediately after the place where requestor identify the crime type to which the gun is linked, a question could appear: "At this time, has a suspect in this crime been identified?" If Yes: Suspect's Name Suspect's age at last birthday, Suspect's Sex, and Suspect's Race. Providing the information would probably entail no more than a few additional minutes of time for the trace requestor, an expenditure of time that would be a very infrequent one for most law enforcement agencies, given the infrequency of non-BATF-initiated trace requests. (Since only about 43% of traces are requested by law enforcement agencies besides BATF, this implies about 38,700 such traces per year, assuming 90,000 total traces [U.S. BATF 1995]. With over 16,000 law enforcement agencies [U.S. FBI 1995, p. 1], this implies only about two trace requests a year per agency.) What could be done with such information? Most significantly for present purposes, it could help determine whether the criminal user of a crime gun had acquired the gun new from a licensed dealer. All sales by licensed dealers to nonlicensees are supposed to be recorded on a Form 4473, which records the transferee's name, race, date of birth, and residence address, as well as height and weight. The form also uniquely identifies the gun purchased, by recording manufacturer, model, serial number, and other identifying characteristics. A trace could therefore be expanded to include just one additional step beyond those already taken now: seeing if the suspect and the gun identified on the proposed trace request form matches up with the person and gun described on the Form 4473 recording the dealer transfer of that gun. When there is such a match, it proves that the criminal user of the gun obtained the gun from a licensed dealer. Further, the information obtained in this way also indicates how long in the past the gun was obtained, and even indicates whether the gun crossed state lines in the interim. The major limit on the utility of such research would be that it would rarely be possible to match the traced gun with a buyer who was not the first retail purchaser, since traces usually "lose the trail" once the gun is in the hands of this purchaser. Although any later transfers involving dealers would usually be recorded on a Form 4473 somewhere, there would be no practical way to locate that form, which would typically be in the possession of an unknown licensee. Thus, matching to a Form 4473 would be practical mainly in cases where the eventual criminal user of a gun had purchased the gun when it was new. Had such data been available before as well as after 1994, it would have been possible to test the hypothesis that the share of crime guns purchased new from a licensed dealer decreased after the Brady Act mandated background checks. The other major problem with the trace data is that they do not cover representative samples of crime guns, but rather samples whose composition is determined largely by law enforcement agency priorities. Were there any serious interest in learning what crime guns in general are like, supplementary samples of guns, in addition to those currently traced, could be selected for tracing, using standard probability sampling procedures. While it might be prohibitively expensive to obtain a nationally representative sample of guns recovered by police, it might be practical to get a representative sample covering the nation's largest cities. BATF has about 24 Criminal Enforcement Field Divisions, each located in one of the nation's largest cities (for a listing, see U.S. BATF 1995, pp. 3-4). If each of these divisions were to obtain probability samples of guns from the police departments covering their home city and one other nearby big city, thereby including the nation's 48 largest cities, the resulting samples would cover cities with about 40% of the nation's violent crimes (U.S. FBI 1995, p. 196), and presumably a similarly large share of its crime guns. Samples could be obtained through systematic sampling of records in each police department's unit handling recovered property. All guns linked with homicides and rapes could be sampled, along with, say, every fifth gun linked with a robbery or aggravated assault, and every 20th gun linked with the more numerous weapons and drug offenses. The usual traces would then be conducted on the guns sampled from these cities. Cooperation from local agencies should be good, since it provides police with information about their local gun situation at the cost of little more than the time of a property room employee showing BATF personnel how to locate their gun records. While it is unlikely that an effort of this scale could be carried out frequently, it would seem worthwhile doing it at least once, so that we would finally have some information on a reasonably representative sample of crime guns, or at least big city crime guns recovered by the police. We could, with considerably more confidence than we have now, assess whether there is any truth to the claims that criminals in some sense prefer small, cheap "Saturday Night Specials," semiautomatic pistols over revolvers, military-style "assault weapons," large caliber guns, small caliber guns, or particular models of guns produced by particular manufacturers. We could also estimate, for as many as 48 different cities, subject to differing levels of state and local controls, the share of crime guns that came from out-of-state. This would permit a test of the oft-stated claim that criminals evade stricter controls in their own states by acquiring guns from interstate gun runners or by otherwise obtaining out-of-state guns. And by insuring a more representative sample, this effort, combined with matching of traced guns with Form 4473 information, would allow more meaningful assessment of the extent to which gun criminals get their guns from licensed dealers. References Britt, C., III, G. Kleck, and D. J. Bordua. 1996a. "A reassessment of the D.C. gun law: some cautionary notes on the use of interrupted time series designs for policy impact assessment." Law & Society Review 30:361-380. -------1996b. "Avoidance and misunderstanding: a rejoinder to McDowall et al." Law & Society Review 30:393-397. Cook, P.J. 1991. "The technology of personal violence." Pp. 1-71 in Crime and Justice, Volume 14, Michael Tonry, ed. Chicago: University of Chicago Press. Kellermann, A.L., F.P. Rivara, N.B. Rushforth, J.G. Banton, D.T. Reay, J.T. Francisco, A.B. Locci, J. Prodzinski, B.B. Hackman, and G. Somes. 1993. "Gun Ownership As A Risk Factor For Homicide in the Home." New England Journal of Medicine 329:1084-1091. Kleck, G. 1991. Point Blank: Guns and Violence in America. N.Y.: Aldine de Gruyter. ------- 1997. Targeting Guns: Firearms and their Control. N.Y.: Aldine de Gruyter. Kleck, G., C.B., and D.J. Bordua. 1993. "The emperor has no clothes: using interrupted time series designs to evaluate social policy impact." Paper presented at the annual meetings of the American Society of Criminology, Phoenix, October 30, 1993. Kleck, G., and M. Hogan. 1997. "A national case-control study of homicide offending and gun ownership." Revised version of a paper presented at the annual meetings of the American Society of Criminology, Chicago, November 21, 1996. Kleck, G., and K. McElrath. 1991. "The impact of weaponry on human violence." Social Forces 69:669-692. Kleck, G., and B. Patterson. 1993. "The impact of gun control and gun ownership levels on violence rates." Journal of Quantitative Criminology 9:249-288. Lott, J.R., Jr., and D. Mustard. 1997. "Crime, deterrence, and right-to- carry concealed handguns." Journal of Legal Studies 26:1-68. Marvell, T.B., and C.E. Moody. 1995. "The impact of enhanced prison terms for felonies committed with guns." Criminology 33:247-281. McDowall, D., C. Loftin, and B. Wiersema. 1992. "A comparative study of the preventive effects of mandatory sentencing laws for gun crimes." Journal of Criminal Law and Criminology 83:378-394. ------- 1995a. "Easing concealed firearms laws." Journal of Criminal Law and Criminology 86:193-206. ------- 1995b. "Additional discussion about easing concealed firearms laws." Journal of Criminal Law and Criminology 86:221-227. Polsby, D.D. 1995a. "Firearms costs, firearms benefits, and the limits of knowledge." Journal of Criminal Law and Criminology 86:207-220. ------- 1995b. "Daniel Polsby replies." Journal of Criminal Law and Criminology 86:227-230. U.S. Bureau of Alcohol, Tobacco and Firearms. 1995. The National Tracing Center 1994 Yearend Report. Washington, D.C.: Author. U.S. Bureau of Justice Statistics. 1993. Survey of State Prison Inmates, 1991. Washington, D.C.: U.S. Government Printing Office. ------- 1997. Presale Firearm Checks. Bureau of Justice Statistics Bulletin. Washington, D.C.: U.S. Government Printing Office. U.S. Congressional Research Service. 1992. 'Assault Weapons': Military- Style Semiautomatic Firearms Facts and Issues. Report 92-434 GOV. Washington, D.C.: Library of Congress. ------- 1994. Brady Handgun Violence Prevention Act. CRS Report for Congress, Report 94-14 GOV. Washington, D.C.: Library of Congress. U.S. Federal Bureau of Investigation. 1995. Crime in the United States 1994 - Uniform Crime Reports. Washington, D.C.: U.S. Government Printing Office. U.S. General Accounting Office. 1996. Gun Control: Implementation of the Brady Handgun Violence Prevention Act. GAO Report GAO/GGD- 96-22. Washington, D.C.: U.S. Government Printing Office. Wright, J., and P.H. Rossi. 1986. Armed and Considered Dangerous. N.Y.: Aldine de Gruyter. ------------------------------ Race, Class and Homicide: Looking Beyond Guns, Drugs, and Gangs Roland Chilton, University of Massachusetts at Amherst Abstract The proportion of U.S. homicides attributed to black males suggests a need for coordinated studies of urban homicide. Studies using race- and gender-specific homicide rates with race- and gender-specific measures of poverty and inequality are needed to assess the impact on homicide of racial separatism and the concentration of the poor in urban areas. The long-term overrepresentation of black males as homicide victims and offenders suggests this need to look beyond short-term changes in drug activity, the availability of guns, and the existence of gangs. Increased focus on racial and economic isolation should suggest programs that will reduce the number of black victims and offenders and thus U.S. homicide rates. Race and Homicide There is a great need for studies of the geographic distribution of urban homicide that go beyond a focus on guns, drugs, and gangs. This statement is not an attempt to tell others what to study. It is a call for increased focus on what is at least the third most salient aspect of homicide in the United States. Other than facts about the impact of gender and age on lethal violence, the most important facts about homicide in the United States are linked to race. Researchers looking at the distribution of homicide victims and offenders in specific cities should be, and to a certain extent are, working cooperatively as they examine the characteristics of census tracts with high rates of violence and homicide. (Blumstein, 1996) Such research is crucial for any assessment of the impact of racial and economic inequality on homicide rates in U.S. cities. The urgency of this issue becomes clearer as we look at data on race and crime in the United States. However measured, homicide victimization is a plague for black males in central cities. The National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC) reported that 39 percent of all 1995 homicide victims were black males. (Anderson et al., 1997) In contrast, individuals who identify themselves as black males make up, at most, six or seven percent of the U.S. population. (Census Bureau, 1996) Data from the National Incident-Based Reporting System (NIBRS), with data submitted from police agencies in only ten states, produces a figure for 1993 that is very close to the CDC figure. Forty-one percent of all homicide victims reported in the NIBRS program in 1993 were described as black males. (Chilton and Jarvis, 1995) Supplemental Homicide Reports (SHR) for 1995 suggest that 40 percent of all 1995 homicide victims were black males. (Snyder, 1997) Moving from victim to offender characteristics, a similar pattern emerges. Supplemental Homicide Reports for 1995, the National Incident-based Reporting System for 1993, and Uniform Crime Report arrest data for 1995 all suggest that black offenders are responsible for most homicides with black victims. The supplemental Homicide Reports for 1995 suggest that 48 percent of offenders in homicide cases are black males and that most homicides are intraracial. This figure was 51 percent for 1993 and 1994. (Snyder, 1997) Data from the not-yet-national, National Incident- based Reporting System for 1993 suggest a slightly higher percentage at 56 percent. (Chilton and Jarvis, 1995) National Uniform Crime Report arrest data for 1995 suggest that 52 percent of those arrested for homicide are black males. (FBI, 1996) Although these percentages are based on reports for 1993-1995, it is important to remember that black males have been over-represented in both the victimization figures and the offender figures for over 35 years. Figure 1 shows the percentage of homicide victims that have been identified as black males in the public health data (CDC) since 1960 and the percentage of victims that have been reported as black males in the Supplemental Homicide Reports since 1980. The CDC victim data indicate that the average percentage of victims of homicide reported as black males for the period 1960 to 1990 was about 39 percent. This figure was highest for 1969 at about 45 percent and lowest for 1984 at about 33 percent. For the period from 1990 to 1994 it was 40 percent. Averaging the Supplemental Homicide Report (SHR) data for victims for 1980 to 1995 produces very similar figures. The lowest percentage of victims described as black males in the Supplemental Homicide Report data was about 33 percent (1984) and the highest was about 42 percent (1994). Sixteen years of SHR data suggests that the average percentage of homicide victims reported as black males was 38 percent but that this percentage increased from 1984 to 1994. Figure 1 also shows UCR arrest counts for 1960-1995. They provide another indication of trends in offender characteristics and suggest that on average, about 49 percent of those arrested for homicide were black males. This figure varied from about 39 percent in 1984 to about 58 percent in 1969. However, the percentage of those arrested for homicide described as black males was 49 percent or higher from 1990 to 1995. The Supplemental Homicide Report information about offenders suggests that, on average, about 44 percent of the people reported as homicide offenders were described as black males. In the SHR program offender information can be provided--whether or not an arrest is made. The highest percentage of offenders described as black males in the SHR program was 51 percent (1993 and again in 1994). The lowest figure was 38 percent (1984). The percentage of offenders described in this data as black males was 49 percent or higher from 1991 through 1994. There is little doubt that black males are, and have for some years been, greatly over- represented as victims of homicide and greatly over-represented as homicide offenders. Social Class The traditional response to any discussion of this situation is the suggestion that these high homicide offending rates for black males are more a function of social class than race. In my view, this explanation is generally accurate. However, it is almost as traditional to suggest that we lack sufficient information on social class to claim empirical support for the social class explanation. One way to clarify this murky situation would be through coordinated studies of the geographic distribution of homicide in several U.S. cities. Homicide rates and data about other forms of violence are available for census tracts for a number of cities. This information could be used to test assertions and assumptions about the impact of persistent poverty and exclusion on homicide rates. By linking urban homicide data for census tracts to census data for the same tracts, it should be possible to produce a clearer picture of the impact of racial isolation and social conditions on homicide rates. Collaborative research efforts should also make it possible to link mapped homicide data to public health data and to public expenditures data. The Chicago data set comes to mind immediately as one that could be used to study the geographic distribution of homicides across census tracts while examining the economic and social characteristics of the same tracts. This might be a particularly useful data set because much of it is already in the public domain. (Block and Block, 1994) It appears that similar if not identical studies could be done for Atlanta, Baltimore, Houston, Milwaukee, Miami, Saint Louis, and San Antonio. There may be comparable data sets for other cities. A set of coordinated and parallel analyses carried out in these cities alone would be a very useful step forward. In such a coordinated approach, researchers working with data for specific cities would use the same variables, the same logic, and the same research procedures. Central to the effort proposed here would be the construction of race- and gender-specific homicide rates for census tracts for as many decennial censuses as possible--at least 1980 and 1990. It is equally important that these researchers create or capture and use the same set of race- and gender-specific measures of income, education, employment, occupation, family composition, and other measures of poverty and inequality for the same census tracts. Parker and McCall's (1997) city- level analysis of interracial and intraracial homicide provides an indication of the utility of race-specific data. Using race-specific independent variables for about 100 U.S. cities, they conclude that economic deprivation affects the intraracial homicide rates for whites and blacks. One outcome of a coordinated, tract-level, multi-city effort could be a much better understanding of the factors associated with high and low homicide rates in central cities. Moreover, the results of such studies would very likely suggest long-term, structural strategies for homicide reduction that could be used to complement a variety of short-term solutions suggested every year. Widespread and sustained efforts to increase and improve employment opportunities, to expand educational opportunities, but especially sustained and comprehensive efforts to reduce exclusion and segregation might be some of the strategies suggested by such studies. Even the partial list of variables presented above suggests that social class is too broad a term to use in this context. Moreover, social class is the term used in the self-report studies wherein young people are asked about their parents' educations, occupations, and possibly incomes--and their own delinquency. A large number of these studies have been assembled and reexamined to suggest that there is no connection between social class and delinquency. In these studies, the young people are almost never asked about homicide and only rarely asked about serious violence. Nevertheless, some criminologists have relied on such studies to convince themselves and perhaps the general public that there is no empirical link between social class and crime. Even a term such as "economic factors" is too vague to describe the ways in which vast differences in income and assets probably contribute to high homicide rates in US central cities. In my view, possibilities for understanding and reducing high black homicide rates in the US require that we move away from--or at least put much less emphasis on--self- report studies of delinquency and other individual level approaches. These approaches are simply not as useful as aggregate level approaches to the problem. For issue involving race, class and homicide, asking why the homicide rates are so high in specific areas of U.S. cities is probably more useful than asking why specific individuals commit violent offenses. Expectations When we raise such questions, we usually find that high homicide rates are closely linked to exclusion and segregation--economic, racial, and ethnic--but especially to the separation and isolation of large segments of the urban population based on income and assets. This separation is frequently based on race or ethnicity but it is increasingly linked to a combination of racial separatism and poverty. In the studies I am proposing, a concentration of the poor in areas with high homicide rates will probably be indicated by low median incomes, low educational attainment, higher proportions of low paying occupations, unemployment, and under employment. These indicators in turn will probably be closely related to housing conditions, living arrangements, and family composition. In these same areas, we should find reduced public service facilities (parks, pools, libraries, recreation centers) and reduced expenditures for schools and possibly even for police services. In short, coordinated studies of the geographic distribution of homicide rates will probably show that areas with high homicide rates are areas with concentrations of poor individuals and poor families, regardless of race or ethnicity. Such studies will probably also provide indications of the impoverishment of the community itself. The policy implications, the suggestions about what should be done to reduce homicide, of the most likely findings may be disturbing or at least daunting for many. However, recognition of the role of our economic and political institutions in the production of high homicide rates for black men could suggest more rational and more effective approaches to homicide reduction than our current attempt to punish our way out of the problem. Coordinated studies would permit criminologists to confront directly the staggeringly high black homicide rates by providing dependable empirical support for the assertion that these rates reflect the impact of exclusion, isolation, and impoverishment. Coordinated studies for multiple census years that are designed to assess the impact of concentrated poverty on both black and white urban populations should suggest new ways to reduce homicide. Nothing is gained by pretending that the rates of homicide victimization and homicide offending for black men do not exist or that the picture we have is the result of bad data. Nothing is gained by ignoring one the most salient aspects of homicide in the United States or acting as if the problem is too sensitive or too complex for empirical analysis. Given the composition of our prison populations and the probability that most Americans are aware of the linkage of homicide and race--but less aware of the extent to which our economic and political institutions contribute to this problem--criminologists and public policy experts who ignore these rates probably exacerbate and perpetuate the current situation. Just as nothing is gained by avoiding discussion of a set of very obvious trends, continued concentration on a set of closely related secondary problems may be equally unproductive. Fads and fashions occur in every aspect of social life--including crime, drug use, and law enforcement. The popularity of specific psychoactive substances, for example, changes over time. If a drug becomes fashionable at a time when particularly active law enforcement procedures are put in place, we can see the kinds of increases in homicide observed from 1985 to 1992. Such coalescence may occur at the same time as gang activity increases and more deadly firearms arrive on the market. This combination of gun availability, drug popularity, gang activity, and enforcement policy may produce limited, short-term increases in homicide rates in some cities. But this nexus is not a useful explanation of the high proportion of black males who have become victims and offenders in urban homicides every year since at least 1960. Continued focus on short-term trends leaves the extensive and persistent long-term differences unexamined and unexplained. It is almost certain that reducing the easy availability of guns, creating a more rational drug policy, and reducing or refocusing gang activity would reduce homicide. However, like homicide itself, gun possession, drug work, and gang involvement are not evenly distributed across society. Whatever impact drug policy, gun availability, and gang activity have on short-term trends, these influences will not explain the relatively stable and unusually high rates of homicide victimization and homicide offending reported for black males. For this we will have to look to widespread practices and procedures that persist over time and continue to exclude and isolate a large number of black males from full participation in the economic, political, and social life of American society. Only coordinated research focused on race, class, and persistently high homicide rates can test this assertion. Such research could also suggest programs and policies that will reduce the number of black victims and black offenders and in this way greatly reduce homicide rates in the United States. References Anderson, RN, KD Kochanek, and SL Murphy, 1997. "Monthly Vital Statistics Report: Report of Final Mortality Statistics, 1995." Monthly Vital Statistics Report, (45) 11, Sup. 2:48 Blumstein, A. 1996. "Initiatives of the National Consortium on Violence Research (NCOVR)," Presentation at the Fifth Annual Spring Symposium of the Homicide Research Working Group, Santa Monica, California. Block, Carolyn Rebecca and Richard L. Block and the Illinois Criminal Justice Authority. (1994) Homicides in Chicago, 1965--1994 [Computer File]. Illinois Criminal Justice Authority [producer]. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. Chilton, Roland and John Jarvis, 1995. "The National Incident-Based Reporting System (NIBRS) and the Characteristics of Offenders." Paper read at the 1995 meeting of the American Society of Criminology, Boston, Massachusetts. Federal Bureau of Investigation, 1960-1996. Crime in the United States. Washington, D.C.: U.S. Government Printing Office. Parker, Karen F. and Patricia L. McCall, 1997. "Adding Another Piece to the Inequality-Homicide Puzzle: The Impact of Structural Inequality on Racially Disaggregated Homicide Rates." Homicide Studies 1 (1): 35-60. Snyder, Howard and Terrence A. Finnegan (1996) Easy Access to the FBI's Supplementary Homicide Reports: 1980 - 1994 [Data Presentation and Analysis Package]. Washington, DC: Office of Juvenile Justice and Delinquency Prevention. U.S. Bureau of the Census, 1995. Statistical Abstract of the United States 1994. Washington, D.C.: U.S. Government Printing Office. ------------------------------- Demography and Lethal Violence[1] Allan F. Abrahamse, RAND Summary Because different demographic groups experience violence at different rates, changes in the homicide rate depend in part on changes in the demographic composition of the population at risk. Data describing the population and homicide experience for California from 1981 through 1995 suggests that demographic changes does not explain the large year-to-year changes we see in homicide rates. The Age Distribution It is well-known that the homicide rate depends strongly on age. Figure 1 shows the age-specific homicide rate for the State of California in 1981. As can be seen, in the first year of life, about 5 out of every 100,000 children is murdered. This risk falls rapidly until about age 10, which was statistically the safest year of life as far as homicide is concerned. During the teen-age years, the risk rose rapidly to a peak at about age 21, and then fell slowly for the next fifty years or so. Late in life, the rate rises, perhaps because older folks are more likely to be killed by events that would only have been an injury for a younger person. For the oldest of the old, the rate falls again. Over the last 15 years, the age composition of California's population has changed. Figure 2 shows the age-specific distribution of the population in 981 ("diamonds"), and in 1995 ("X's"). Basically, the population got older. Since generally older persons face lower risks of homicide, an aging population could lead to one that appears less violent. Race and Ethnicity Different racial and ethnic groups face sharply different risks of homicide, and the homicide rate for males is always much higher than for females. This chart shows the homicide rate in California in 1981 for four race/ethnic groups, by sex. African-American males faced highest risk, followed by Hispanic males, and then by African-American females. California has experienced a pronounced change in the ethnic make-up of its population. Figure 3 shows the fraction of its population that is either Hispanic (lower bar) or African-American (upper bar), between 1981 and 1995. As Figure 4 shows, the relative share of the population that is African-American has remained roughly constant at about 7%. The relative share that is Hispanic has grown. Since Hispanic face higher homicide risks, this rise in the fraction of the population that is Hispanic could lead to an apparent rise in violence. How Much Does Demography Matter? How much of the variation in the observed homicide rate can be accounted for by these changes in the composition of California's population? One way to answer this question is to ask: what would the homicide rate have been if the only thing that changed was the population composition? To answer this question, I calculate the homicide rate by age, race/ethnicity and sex specific rates as observed in 1981, and apply this historical rates to the actual population counts in each subsequent year. The formula for the estimated rate in 1995 is just where x varies over all age, sex and race/ethnicity classes. This estimated rate is plotted (line graph) in Figure 5. The homicide rate would have grown slightly over the rate seen in 1981, and then leveled off slightly around 1990. What actually happened (bar graph) was far different. After 1981, the rate dropped sharply in 1982 and 1983, grew rather slowly for the next five years, rose sharply between 1989 and 1993, reached a peak in 1993, and then fell dramatically in 1994 and 1995. Data recently released by the FBI indicate that rates fell again in 1996. The important point here, however, is that changes from year to year in the homicide rate are much greater than can be accounted for by mere changes in the composition of the population. The demographic composition of California has changed a lot in the last 15 years, but the changes have been relatively small from one year to the next. Homicide rates depend in part on demographic composition, but these calculations suggest that demographic change plays only a small part in the changes we see from one year to the next. NOTE 1. This paper was prepared with support from the James Irvine Foundation.