Title: Proceedings of the Homicide Research Working Group Meetings,
1997 and 1998 (pages 114-194).
Series: Research Forum
Author(s): NIJ
Published: National Institute of Justice, May 1999
Subject: Homicide, domestic violence, juvenile violence, gun violence,
gangs, crime patterns and future trends, and statistics
63 pages
148,000 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).

-------------------------------

1998: Bridging the Gaps: Collaborations on Lethal Violence Research,
Theory, and Prevention Policy

Proceedings of the 1998 Meeting of the Homicide Research Working
Group

-------------------------------

1998 Keynote Address
School Shootings and School Violence: What's Going On and Why?

Kathleen M. Heide, Professor, Department of Criminology
University of South Florida, Tampa

Abstract

Recent data, although limited, suggest that school shootings and school
violence have increased in the U.S. Social factors that are converging in
the 1990s appear to be contributing to the youth violence problem. These
factors include institutional change, societal influences, and situational
factors that affect some youths more than others. These variables make
children and adolescents more vulnerable to behaving inappropriately and
coping maladaptively than youths who have different life experiences.
After the discussion of these social factors and their effects, individual
factors that identify youths at greater risk of choosing violent solutions to
life's challenges are highlighted. 

Introduction

The massacre at Jonesboro Middle School on March 24, 1998, happened
in a small rural Southern community and became a global event. Time and
Newsweek, weekly news magazines that sell millions of copies in the U.S.
and across the world, prominently displayed the two young killers, ages
11 and 13, on their covers. These two "kids" were accused of killing four
girls and a teacher and wounding 10 more in an ambush of their fellow
students. 

These two boys did more than allegedly take the lives of five people,
however. In a matter of seconds, armed with semiautomatic weapons, they
forever changed the reality of millions of children in the United States
who now wonder when they go to school each morning whether they too
will be shot by a fellow classmate.

This fear of annihilation by "kids" in the hall or on the playground is a
new phenomenon. I can honestly say that when I was growing up the
thought that I would be shot and killed when I went out to recess or
responded to a fire alarm in my school building never even entered my
mind -- not even for a split second. I have asked audiences of
professionals and university students did they ever fear as a child that they
would be shot by a fellow classmate. Always, my question is met with
denial and disbelief, typically followed by anguish and sadness.

Multiple School Shootings in U.S. Schools

Data indicate that the type of school shooting exemplified by the
Jonesboro massacre has become more common in recent years. These
shootings are characterized by multiple victims, often randomly selected.
The killings seem to be an explosion of feeling dumped on an amorphous
target rather than directed at a particular individual whom the assailant
perceives has injured him.

I found 14 incidents involving multiple shootings of this type from 1993
through 1998. A very different picture emerged when I examined these
incidents by calendar years (1/1-12/31) and by academic years (8/1-6/30).
As revealed in Table 1, 8 of the 14 incidents of multiple shootings in
schools occurred during the most recent school year, 1997-1998.

Table 2 provides additional data on these multiple school shootings.
Perusal of these incidents leads to four observations. Drawing firm
conclusions at this point is ill-advised given the small number of cases and
the possibility that the academic year 1997-98 might be an anomalous
year. With these caveats in mind, it behooves us to note the following:


First, the killings and wounding have become more concentrated over the
time frame. Over the five to six year period, 30 people were killed and 57
were wounded in these 14 incidents of multiple school shootings. During
the academic year 1997-1998, 18 of the 30 dead (60 percent) were killed
and 52 of the 57 wounded (91 percent) were injured. 

Second, the number of victims killed or wounded per incident has
increased. Although the numbers of incidents are very small to permit
conclusions, an increasing trend is noticeable. The number of victims
killed or wounded in the two incidents in 1992/93 averaged 2; for the
three incidents in 1995/96, it averaged 3; for the one incident in 1996/97,
it was 4; and for the 8 incidents in 1997/98, it averaged 9.

Third, the age of the assailants appears to be getting younger over time.
Of 15 youths involved in the 14 incidents over the five to six year time
frame, 8 were 14 or under. Of the 9 youths involved in the academic year
1997/98, 7 were 14 or under. The high proportion of youths under 14
involved in the school homicides is in stark contrast to figures on juvenile
homicide offenders in general. Approximately 88 percent of juveniles
arrested for murder in the U.S. are ages 15, 16, and 17 years old (Heide,
1999). 

School Violence in U.S. Schools

School shootings and mass killings are among the most extreme acts of
violence in our nation's schools. There are less deadly violent
victimizations that, unlike multiple school shootings, occur daily in
schools across the United States and leave students, parents, and
communities in the grips of fear. A 1993 national school-based survey of a
representative sample of high school students revealed, for example, that
more than 4 percent of responding students missed a day of school each
month because they feared for their public safety at school or while
traveling to or from school. In a 1994 national survey of parents with
children in public school, 40 percent of parents of high school students
related that they were worried about their child's safety in school or on
their way to, and from, school. In the 1994 National League of Cities
survey of 700 communities across the nation, 80 percent of respondents
reported that violence was a serious problem in classrooms, hallways, and
playgrounds; 40 percent indicated that violence in schools had risen
noticeably during the past five years. Moreover, of the schools
participating in the survey, 25 percent related that students had died or
sustained injuries for which they were hospitalized as a result of violence
(Arnette and Walsleben, 1998). 

Recently released data from the 1989 and 1995 School Crime Supplement
to the National Crime Victimization Survey indicate that criminal violent
victimizations in school have increased (U.S. Department of Justice,
1998). These data provide snapshots of students' reports of crime in school
during the six months preceding the surveys taken in 1989 and 1995.
Nationally representative samples of approximately 10,000 students
between the ages of 12 and 19 were surveyed during the two time periods.
To be eligible to participate, respondents had to have been in school at
some point during the six months preceding the interview. School crime
was defined as occurring in the school building, on school grounds, or on
a school bus. Criminal violent victimization was defined as involving
physical attacks (assaults) or the taking of property by force, weapons, or
threats (robberies).

Compared to 1989, students in 1995 were more likely to report being the
victim of a criminal violent incident. In 1989, 3.4 percent of students
surveyed indicated that they had been physically attacked or had property
taken by actual or threatened violence; in 1995, the percentage increased
to 4.2 percent. While those percentages may seem small, when
extrapolated to the population at risk, the magnitude of the problem of
school violence is more apparent. Caution is advised when extrapolating
from a sample of 10,000 to the universe, in this case, 21,554,000. With
this caveat in mind, extrapolation would suggest that more than 1,000,000
students were the victims of a violent crime in a school building, on
school grounds, or on a school bus sometime between the beginning of
July 1994 and the end of June 1995. 

Comparison of the 1989 and 1995 data revealed several important
findings with respect to gender, age, and grade level. Violent
victimizations rose for both boys and girls. However, while boys were
more likely to be the victims of violent school crimes, the increase in the
percentage of girls who reported violent victimizations (from 2.0 percent
to 3.3 percent) was higher than that for boys (from 4.8 percent to 5.1
percent). 

The percentage of youths reporting violent victimization in school was
inversely related to age for youths ages 12 through 17 during both time
periods. In 1995, the percentages of youths who reported being a victim of
violent crime decreased consistently from 6.8 percent of 12-year-olds to
1.9 percent of 17-year-olds. Similarly, the percentage of youths who
indicated that they had been a victim of violent crime was negatively
associated with grade level. In 1995, the percentages of school children
who reported violent criminal victimization decreased continuously from
6.7 percent of six graders to 1.7 percent of twelfth graders.

Compared to 1989, students in 1995 were more likely to report that street
gangs were in their school (15.3 percent vs. 28.2 percent) and that drugs
were available at their school (64.8 percent vs. 67.2 percent).
Interestingly, close inspection of these data revealed that these problems
appear to be correlated with criminal violent victimization at school.
Student reports of having been a victim of a violent crime at school were
related to street gang presence in their schools and to student reports of
drug availability in their schools. 

Data on the presence of guns in school, only available for 1995, were
alarming. Of those surveyed, 12.7 percent reported that within the last six
months they knew someone who brought a gun to school and 5.3 percent
indicated that they had seen a student with a gun at school during this time
period. These figures, when extrapolated, would suggest that more than
3,000,000 students knew someone who brought a gun to school and more
than 1,000,000 students had actually seen a student with a gun in school
within a six month period (U.S. Department of Justice, 1998).

Social Factors Contributing to School Violence And Shootings[1]

The data on school shootings and school violence indicate that there is a
problem in our schools with respect to violent criminal victimization and
suggest it is getting worse over time. For generations, the 3 R's have stood
for "reading, writing, and arithmetic." If we don't want the 3 R's to
become "reading, writing, and run for cover" for this generation of school
children, we need to understand why we are seeing more violence by
youths in the 1990s in schools and other settings. 

There are social factors that are converging in the 1990s that appear to be
contributing to the youth violence problem. These factors include
institutional change, societal influences, and situational factors that affect
some youths in American society more than others. These variables make
children and adolescents more vulnerable to behaving inappropriately and
coping maladaptively than youths who have different life experiences.
After discussing these social factors and their effects, I will turn to
individual factors that identify youths at greater risk of choosing violent
solutions to life's challenges.

Institutional Change

Dramatic changes have occurred in societal institutions over the last
generation that seriously impact on the socialization of children.
Historically, families, religious institutions, schools, and communities
played major roles in helping children become productive citizens and
contributing members of society. Today, these institutions have undergone
major changes that have severely curtailed their sphere of influence. 

The change in family structure is by far the most serious. More than in
other generations, children and adolescents today are growing up in an era
beset by "an overall decline of the extent and influence of the family from
the extended multigenerational family, to the nuclear family, to the single
parent family, to the 'no parent' family of street children" (Friedman 1993,
p. 509). With the decline of the family in the United States, the task of
socializing children has become more difficult. More children today,
compared to the past, are being raised by single mothers. Due to the
demands placed on single parents, many children today, relative to their
counterparts of even a generation ago, are not learning acceptable ways of
behaving. In many households, again due to changes in the family,
positive male role models are not available. As a result, appropriate ways
of acting as a man are not being taught. In addition, values are not being
reinforced; codes of right and wrong are not being effectively transmitted.

The declining power of religious institutions has also taken place over the
last generation. Churches and synagogues have traditionally reinforced
acceptable codes of behavior and standards of right and wrong. They have
played an important role in value transmission. As Church attendance by
families declined, many youth in America lost an important voice in the
call to behave as law-abiding and moral people.

The loss of authority of teachers and other school personnel has coincided
with the deteriorating influence of the family and the churches. Teachers
report today that much of their time is spent trying to maintain order in
their classrooms. Respect for the teacher's authority, once freely accorded
by students and their parents, is no longer a given. For example, seventy
percent of the high school students surveyed in the Public Agenda's report,
Getting by: What American teenagers really think about their schools,
indicated that disruptive students were a serious problem in their schools
(Sloan, 1997). Teachers complain that when they try to discipline
students, they frequently are challenged by parents who want to know
why they are "picking on" their children.

Not surprisingly, this generation of youths, more than other generations, is
beset by a loss of communities. The decline in adult authority at the level
of families, religious institutions, and schools has clearly impacted on the
role of adults in community. For generations, neighbors frequently kept an
eye out on children playing in the schoolyard or congregating on the street
corner. Adults would step in to offer guidance and even correct children
for minor transgressions, such as smoking or using bad language. But
rarely do adults get involved anymore in the 1990s. Some maintain that
they are too busy with work and other commitments. Others, however,
insist that they are afraid that they will be rebuffed by some adults for
daring to correct others' children or that they will be harmed by youths
who resent their intrusion into the youths' lives. 

Societal Influences

In addition to fundamental institutional change, youths growing up in the
1990s are subjected to societal influences that send confusing messages to
them. These include the crisis in leadership and the lack of heroes
available today to serve as role models. The saturation of violence in
society and changes with respect to firearms are also perplexing to many.

In contrast to previous generations, children and adolescents today are
living in a country that has been experiencing a crisis in leadership and
lack of heroes. In the past, U.S. Presidents, successful entertainers, and
legendary sports figures were presented to the youth of America as people
to emulate. In the 1990s, the personal ethics and behavior of many of
these individuals have been seriously questioned. Government leaders
who break campaign promises and involve themselves in money and sex
scandals have shown that many politicians today deny responsibility for
their behavior and their decisions. When leaders of our country are no
longer expected to keep their word and are not held accountable, some
youths become cynical about following societal dictates. When police
officers are viewed on nationwide television repeatedly beating an
African-American in their custody and are proven to be lying on the
witness stand in the case of another African-American man, adolescents,
particularly those from minority groups, increasingly lose faith in a
criminal justice system that is supposed to protect them and to dispense
equal justice. When world class athletes and notorious gangsta rappers are
accused of violent criminal acts, some adolescents feel free to adopt
similar courses of behavior.

At a time when heroes and moral figures appear to be sorely lacking,
American society has become saturated with violence. Witnessing
violence has been correlated with lessened inhibition to use violence
(Prothrow-Stith and Weissman, 1991). Over the last two decades, TV,
including the evening news, and films have become increasingly more
violent (Levin and Fox, 1985; Prothrow-Stith and Weissman, 1991; Fox
and Levin, 1994). Violent videogames and gangsta rap music provide
graphic scenes and messages of violence. Scores of youths have seen
violence in their own homes and in their neighborhoods. To many youths
today, the world is a violent place. Accordingly, many youths feel
compelled to carry guns and are prepared to use violence when they
perceive the situation as warranting it (Heide, 1997, 1998).

Not only do our youths grow up in a world that encourages violence, those
in the United States are increasingly finding themselves surrounded with
the tools which make acts of violence quick and easy (Sheley and Wright,
1995). Changes in the absolute number of guns in society, the availability
of guns to juveniles, the increased firepower of today's firearms, and the
attitudes towards the appropriate use of guns are factors that affect
children and adolescents in our culture. Recent research has demonstrated
that youth involvement in violence has been associated with the frequency
of carrying a weapon (Resnick et al., 1997). Moreover, the increase in
murders by juveniles in recent years in the U.S. has been tied directly to
their use of firearms, particularly handguns (Blumstein, 1995; Fox, 1996;
Kennedy, 1997; Snyder, Sickmund, and Poe-Yamagata, 1997). Recent
studies have also shown that juvenile homicide offenders like to equip
themselves with newer and more powerful weapons (Kennedy, 1997),
which they appear to be acquiring illegally from firearms dealers. The
proliferation of guns in American society and the advent of firearms with
increased firepower has been accompanied with a change in attitudes
towards firearms. In past generations, youths were taught that guns were
to be used for self-defense and/or hunting. In the 1990s, cultural messages
embedded in music, television, and movies emphasize firearms as a
symbol of power and the instrument to use to redress grievances, no
matter how trivial. 

Situational Factors

The above discussion has suggested that youths growing up in the 1990s
are more vulnerable to choosing antisocial means to resolve difficulties
and fulfill needs because they live in a society where major institutions of
socialization have declined. In addition, they are more vulnerable than
their counterparts of the past to respond in violent ways because they are
subjected to societal influences that are demoralizing and that also
promote violence. Against this background are youths who are subjected
to certain situations spared to others, which make particular children and
adolescents even more vulnerable to behaving maladaptively. These are
youths who are abused, neglected, raised in poverty, using drugs and/or
alcohol, and involved in gangs. The prevalence of each of these situational
factors has increased in the 1990s, meaning that the numbers of youths in
the United States who are exposed to these conditions have also risen
(Heide, 1999).

Many of today's youth grow up in families that foster violent and
destructive behaviors. Despite a decrease in the number of young
Americans, reports of child abuse have greatly increased in recent years
(United States Advisory Board on Child Abuse and Neglect, 1993;
Snyder, Sickmund, and Poe-Yamagata, 1997). Although the majority of
children who are victims or witnesses of family violence do not grow up
to victimize others (Smith and Thornberry, 1995), a growing body of
research indicates that these children are at greater risk of engaging in
delinquent behavior (see Heide, 1999). There is a growing body of
evidence indicating that exposure to parental violence is also related to
violent behavior (Thornberry, 1994; Howell et al., 1995; Heide, 1999).
Some youths who are abused do not bond with others. Consequently, they
develop no values or empathy to insulate them from killing innocent
human beings. Other abused juveniles are angry and in pain, and vent
their rage by destroying others (Magid and McKelvey, 1987).

Neglect frequently accompanies abuse, but it can also exist independently,
often manifesting itself as the common failure of parents to supervise their
children (Heide, 1992). During the last 25 years, several significant
changes in family structures have contributed to decreasing levels of child
supervision and have placed adolescents at greater risk of getting into
serious trouble. These changes include a rise in the number of children
born to single mothers, the increase in the number of children raised by a
single parent due to illegitimacy or the subsequent divorce of the parents,
and the increase in the number of working mothers (Heide, 1997). Given
these familial changes, the time that youths spend with their parents and
the amount of guidance that they receive have significantly decreased
during the past several decades (Carnegie Council on Adolescent
Development, 1995).

The percentage of children living in poverty in the 1990s has also
increased as a byproduct of the changes in family structure over the last
two decades. The escalation in single-female-headed households
occasioned by the rise in births to unwed females and in divorce has
resulted in more children being raised in poverty (Garfinkel and
McLanahan, 1986; Wright and Wright, 1995). Research indicates that
about three out of four households headed by single females live in
poverty at least some of the time and one third are chronically poor. As
we approach the millennium, it appears that one out of every three
children under age 6 lives below the poverty line (Stephens, 1997). The
rise in the number of children living in poverty means that more children
today have limited access to health care, including mental health services,
and lack other resources to improve the quality of their lives. Some of
these youths will fare well; others will cope maladaptively by engaging in
criminal behaviors and by using drugs and alcohol.

Drug use surveys indicate that the rates of illicit drug use by adolescents,
which had declined during the 1980s (Osgood, 1995; White House Office
of National Drug Control Policy, 1997) are again rising in the 1990s, and
are much higher than they were a generation ago. This increase has been
observed among younger, as well as older adolescents ("Drug use up,
study shows," 1997). The percentage of youths reporting past month use
of marijuana, stimulants, hallucinogens, and inhalants rose from 1991
through 1994 (Office of National Drug Control Policy, 1995). A 1993-
1994 survey of junior high (grades 6 through 8) and high school students
(grades 9 through 12) conducted by the Parent Resource Institute for Drug
Education (PRIDE) found a strong link in both groups between use of
alcohol and marijuana and several measures of violent behavior, including
carrying a gun to school and threatening to harm another person (Office of
National Drug Control Policy, 1995). These findings are consistent with
those from a growing number of studies indicating a substantial
relationship between adolescent violence and substance abuse (See, e.g.,
Elliott et al., 1989; Johnston et al., 1993; Office of National Drug Control
Policy, 1995; Osgood, 1995). Although using alcohol or drugs does not
cause youths to commit crimes or be violent, it is likely that chemical
abuse affects their judgment about engaging in criminal activity and their
perceptions during the event. In addition, it is highly probable, in light of
prior research, that the use of alcohol and drugs by many children and
adolescents is "more a reflection of shared influences on a wide variety of
deviant behavior than of any causal relationship" (Osgood, 1995, p. 32).
Several researchers have found that various types of deviant or illegal
behaviors are positively related to one another (See, e.g., Osgood et al.,
1988; Elliott, Huizinga, and Menard, 1989; Gottfredson and Hirschi,
1990; Dembo et al., 1992; Resnick et al., 1997).

The numbers of gangs and of gang memberships have increased
significantly during the last decade in the United States. Although most
noticeable in large urban areas, gangs have also started to appear in
suburban areas and smaller towns across America (Howell, 1994, 1995;
U.S. Department of Justice, 1997). Substantial evidence exists that gangs
have become increasingly responsible for a disproportionate amount of
violence (See, e.g., Howell, 1995; Thornberry and Burch, 1997), which
appears to be largely due to the ready availability of firearms and more
sophisticated weaponry (See, e.g, Block and Block, 1995; Sheley and
Wright, 1995). The presence of gangs in schools, which almost doubled
from 1989 to 1995 (U.S. Department of Justice, 1998), has also been
associated with increased victimization in schools.

Youths at Higher Risk of Behaving Violently

We have identified the social factors impinging on American youth today
that make growing up in the 1990s more difficult for them than for
children and adolescents growing up a generation ago. We have looked
beyond the major changes in our institutions and societal influences to
situational factors that affect some children in the United States and
consequently put added strain on them. The question becomes, can we do
better at isolating the youths who are most vulnerable to behaving
violently in society? The number of school shootings, and even the
number of juvenile homicides, is far too small to predict which youths
will respond in such a fashion. However, my clinical experiences
evaluating approximately 100 violent adolescents, mostly murderers,
suggest that the possession of certain personality characteristics,
particularly when combined with involvement in certain types of
activities, put youths at higher risk of choosing violent solutions. 

I have frequently observed certain personality characteristics present in
the case histories of youths who engage in homicidal behavior. These
include low self esteem and an inability to deal with strong negative
feelings. These youths are often angry, depressed, alienated from
mainstream society, and chronically bored. Their judgment is poor and
their behavior is impulsive.

These personality characteristics are not necessarily indicative of youths
who will behave violently. The "ante" is raised, however, when these
youths are drawn to nihilistic activities and become heavily involved in
them. Youths who are preoccupied with fantasy games, such as Dungeons
and Dragons, can blur fantasy with reality. Those who become extremely
involved in cult and fringe groups that promulgate separation and hate
may come to see "outsiders" as the enemy or as inferior human beings that
are expendable. Kids who become absorbed in violent media (movies,
music, videogames) and are obsessed with acts of dying and
destructiveness are aligning themselves with destructive and violent
themes. Children and adolescents who are cruel to animals and to more
vulnerable human beings (younger children, smaller kids, the mentally ill,
homeless people, elderly adults) are displaying sadistic behavior that
warrants concern. Youths who are drawn to malignant acts of aggression
lack empathy and may see death and destruction as the means to achieve
power and control and to fulfill other existential needs in maladaptive
ways (Fromm, 1973).

The Cumulative Effect in Context

In summary, changes in major institutions, societal influences, and
situational factors affecting some children in the 1990s appear to be
significant factors influencing the behaviors of youths, including those
who resort to violence. These variables likely interact with the personality
characteristics of particular adolescents, making some youths more likely
to engage in violent behavior than others.

In essence, for many youths, the effect of these factors is cumulative. Put
succinctly, many youths growing up in the 1990s feel that they have little
or nothing left to lose. These are the kids who are angry, frequently in
pain, and often unattached to other human beings due to experiences in
their home and neighborhood environments. Many of these youngsters
lack self-esteem and the resources to improve their lives. They are living
in a society experiencing increases in youths having sex and babies
outside of marriage (Friedman, 1992), using drugs, participating in
criminal violence, and dying violently whether through homicide or
suicide. As a result, many young people today are severely alienated
(Lerner, 1994). They do not hold conventional values or dreams. Often
chronically bored, they use drugs, alcohol, and sex to numb themselves
and commit crimes for fun. They live in the moment. To them, thrills --
and lives -- are cheap. 

Biological factors also may be intricately entwined in the homicidal
equation in many cases. A growing body of research suggests that
criminal behavior may be linked at least in some cases to genetics,
neurological factors, and biochemical reactions (Widom, 1991; Lewis,
1992; Pincus, 1993; Heide, 1999). 

Sociobiologists have maintained that criminal behavior is influenced by
both individual biological factors and social and environmental conditions
(Jeffrey, 1979). Lewis's extensive studies on juvenile murderers led her to
conclude that genetic factors and biological vulnerabilities, particularly
when severe, predispose certain individuals to respond violently. Her
research suggests that if these individuals are subjected to intense
psychological, social, and environmental stressors that exceed their ability
to cope, violent expression is more likely to result, particularly among
males (Lewis et al., 1989; Lewis et al., 1991; Lewis, 1992). Lewis' theory
of neuropsychiatric vulnerability also received support in a larger study
involving urban delinquents in Chicago (Hughes et al., 1991).

What can be done to reverse the upward surge in violence by today's
juveniles? Neutralizing or eliminating the variables that contribute to
youths becoming involved in violent incidents may require a generation or
more to accomplish (Heide, 1999). My clinical experiences with violent
youths have convinced me that change must include parents, the
educational system, communities, government leaders, medical and
mental health professionals, the media, and individuals joining together to
raise a healthier next generation and to build a more peaceful society. I
will provide a blueprint to reduce youth violence in the United States
during my talk on Friday. (This paper is contained later in these
proceedings.)

References

Arnette, J.L., and Wasleben, M.C. (1998). Combating fear and restoring
safety in schools. Washington, D.C.: U.S. Department of Justice, Office of
Juvenile Justice and Delinquency Prevention. 

Block, C.R., and Block, R. (1995). Street gang crime in Chicago. In M.
W. Klein, C.L. Maxson, and J. Miller (eds.), The modern gang reader (pp.
202-211.). Los Angeles, CA: Roxbury Publishing Company.

Blumstein, A. (1995, August). Violence by young people: Why the deadly
nexus? National Institute of Justice Journal, pp. 2-9. 

Carnegie Council on Adolescent Development. (1995, October). Great
transitions: Preparing adolescents for a new century. New York: Carnegie
Corporation of New York. 

Dembo, R., Williams, L., Wothke, W., Schmeidler, J., Getreu, A., Berry,
E., and Wish, E.D. (1992). The generality of deviance: Replication of a
structural model among high-risk youths. Journal of Research in Crime
and Delinquency, 29, 200-216.

Drug use up, study shows. (1997, Aug. 14). Tampa Tribune,
Nation/World, p. 4.

Elliott, D.S., Huizinga, D., and Menard, S. (1989). Multiple problem
youth: Delinquency, substance use, and mental health problems. New
York: Springer-Verlag. 

Fox, J.A. Trends in juvenile violence. (1996). Washington, D.C.: Bureau
of Justice Statistics, U.S. Department of Justice.

Fox, J.A., and Levin, J. (1994). Overkill: Mass Murder and Serial Killing
Exposed. New York: Plenum Press.

Friedman, H.L. (1992). Changing patterns of adolescent sexual behavior:
Consequences for health and development. Journal of Adolescent Health,
13, 345-350.

Friedman, H.L. (1993). Promoting the health of adolescents in the United
States of America: A global perspective. Journal of Adolescent Health, 14,
509-519.

Fromm, E. (1973). The anatomy of human destructiveness. Greenwich,
CT: Fawcett Publications, Inc.

Garfinkel, I., and McLanahan, S.S. (1986). Single mothers and their
children: A new American dilemma. Washington, D.C.: Urban Institute
Press.

Gottfredson, M.R., and Hirschi, T. (1990). A general theory of crime.
Stanford, CA: Stanford University Press. 

Heide, K.M. (1992). Why kids kill parents: Child abuse and adolescent
homicide. Columbus, Ohio: Ohio State University Press (Cloth); (1995)
Thousand Oaks, CA: Sage Publications (paper).

Heide, K.M. (1997). "Juvenile homicide in America: How can we stop the
killing?" Behavioral Sciences and the Law, 15, 203-220. 

Heide, K.M. (1999). Young killers: The challenge of juvenile homicide.
Thousand Oaks, CA: Sage Publications.

Howell, J.C. (1994, April). Gangs. Office of Juvenile Justice and
Delinquency Prevention (Fact Sheet #12). Washington, D.C.: U.S.
Department of Justice. 

Howell, J.C. (1995). Gangs and youth violence. In J.C. Howell, B.
Krisberg, J.D. Hawkins, and J.J. Wilson (Eds.). A sourcebook: Serious,
violent, & chronic juvenile offenders (pp. 261-274). Thousand Oaks, CA:
Sage. 

Howell, J.C., Krisberg, B., and Jones, M. (1995). Trends in juvenile crime
and youth violence. In J.C. Howell, B. Krisberg, J.D. Hawkins, and J.J.
Wilson (Eds.). A sourcebook: Serious, violent, & chronic juvenile
offenders. Thousand Oaks, CA: Sage Publications.

Hughes, J.R., Zagar, R., Arbit, J., and Busch, K.G. (1991). Medical,
family, and scholastic conditions in urban delinquents. Journal of Clinical
Psychology, 47(3), 448-464.

Jeffrey, C.R. (1979). Biology and crime. Beverly Hills, CA: Sage
Publications.

Johnston, L.D., O'Malley, P.M., and Bachman, J.G. (1993). National
survey results on drug use from the Monitoring the Future Study, 1975-
1992: Vol. 1 Secondary school students. Rockville, MD: National Institute
on Drug Abuse. 

Kennedy, D.M. (1997). Juvenile gun violence and gun markets in Boston.
National Institute of Justice Research Review. Washington, D.C.: U.S.
Department of Justice, National Institute of Justice.

Lerner, R.M. (1994). America's youth in crisis. Thousand Oaks, CA: Sage
Publications.

Levin, J., and Fox, J. (1985). Mass murder: America's growing menace.
New York: Plenum Press.

Lewis, D.O. (1992). From abuse to violence: Psychophysiological
consequences of maltreatment. Journal of the American Academy of Child
and Adolescent Psychiatry, 31(3), 383-391.

Lewis, D.O., Lovely, R., Yeager, C., and Femina, D.D. (1989). Toward a
theory of the genesis of violence: A follow-up study of delinquents.
Journal of the American Academy of Child and Adolescent Psychiatry,
28(3), 431-436.

Lewis, D.O., Yeager, C.A., Cobham-Portorreal, C.S., Klein, N.,
Showalter, C., and Anthony, A. (1991). A follow-up of female
delinquents: Maternal contributions to the perpetuation of deviance.
Journal of the American Academy of Child Psychiatry, 30(2), 197-201.

Magid, K., and McKelvey, C.A. (1987). High Risk: Children Without a
Conscience. New York: Bantam Books.

Mahoney, A.R. (1987). Juvenile justice in context. Boston, MA:
Northeastern University Press.

Office of National Drug Control Policy. National Drug Control Strategy:
Executive Summary. (1995). Washington, D.C.: Author.

Osgood, D.W. (1995, January). Drugs, alcohol, and violence. Institute of
Behavioral Science, Regents of the University of Colorado, Boulder.

Osgood, D.W., Johnston, L.D., O'Malley, P.M., and Bachman, J.G.
(1988). The generality of deviance in late adolescence and early
adulthood. American Sociological Review, 53, 81-93.

Pincus, J.H. (1993). Neurologist's role in understanding violence.
Archives of Neurology, 8, 867-869.

Prothrow-Stith, D., and Weissman, M. (1991). Deadly Consequences.
New York: Harper Collins.

Resnick, M.D., Bearman, P.S., Blum, R.W., Bauman, K.F., Harris, K.M.,
Jones, J., Tabor, J., Beuhring, T. Sieving, R.E., Shew, M., Ireland, M.,
Bearinger, L. and Udry, J.R. (1997). Protecting adolescents from harm:
Findings from the National Longitudinal Study on Adolescent Health.
JAMA, 278(10), 823-832. 

Sheley, J.F., and Wright, J.D. (1995). In the line of fire: Youth, guns, and
violence in America. New York: Aldine de Gruyter, Inc.

Sloan, J. (1997, May 4). Students prefer school of hard knocks. The
Tampa Tribune, Nation/World, p.1, 10.

Smith, C., and Thornberry, T.P. (1995). The relationship between
childhood maltreatment and adolescent involvement in delinquency.
Criminology, 33(4), 451-481.

Snyder, H.N., Sickmund, M., and Poe-Yamagata, E. (1997). Juvenile
offenders and victims: 1996 update on violence. Washington, D.C.: Office
of Juvenile Justice and Delinquency Prevention.

Stephens, G. (1997). Youth at risk: Saving the world's most valuable
resource. Bethesda, Maryland: World Future Society.

Students' reports of school crime: 1989 and 1995. Washington, D.C.: U.S.
Department of Justice, Bureau of Justice Statistics.

Thornberry, T.P. (1994). Violent families and youth violence. Office of
Juvenile Justice and Delinquency Prevention, U.S. Department of Justice,
Fact Sheet #21.

Thornberry, T.P., and Burch, J.H. (1997, June). Gang members and
delinquent behavior (Juvenile Justice Bulletin). Washington, D.C.: U.S.
Department of Justice, Office of Justice Programs, Office of Juvenile
Justice and Delinquency Prevention. 

U.S. Advisory Board on Child Abuse and Neglect (USABCAN). (1993).
Neighbors helping neighbors: A new national strategy for the protection
of children. Washington, DC: National Clearinghouse on Child Abuse and
Neglect.

U.S. Department of Justice, Office of Juvenile Justice and Delinquency
Prevention. (1997). 1995 National youth gang survey. Washington, D.C.:
Government Printing Office.

U.S. Department of Justice, Bureau of Justice Statistics (1998). Students'
reports of school crime: 1989 and 1995. Washington, D.C.: Government
Printing Office.

White House Office of National Drug Control Policy, Drug Policy
Information Clearinghouse. (1997). Fact sheet: Drug use trends.
Rockville, MD: National Criminal Justice Reference Service.

Widom, C.S. (1991). A tail on an untold tale: Response to 'Biological and
genetic contributors to violence' -- Widom's untold tale. Psychological
Bulletin, 109(1), 130-132.

Wright, K.N., and Wright, K.E. (1995). Family life, delinquency, and
crime: A policymaker's guide. Research summary. Washington, D.C.:
U.S. Department of Justice, Office of Juvenile Justice and Delinquency
Prevention.

NOTE

1. Some of the material contained in the remainder of this paper
previously appeared in Heide, 1997, 1999.

-------------------------------

Session 1: Essentials of Violence Surveillance Data 

Responses by Lois Fingerhut

What's the difference in the mortality vs. morbidity e-codes for injuries by
firearms?

Lois: The code structures are basically the same, the question is how
they're used. They're generally a pretty good match.

What about the sibling crime idea, questions on homicide as the outcome
of something else? So these e-codes, since they don't reflect specific
crimes, can't be used to trace sibling crimes?

Lois: The forms only have good information on this if the physician puts
the description of how the event happened on the form. The ICECI, a new
classification coding from Injury Violence has a very detailed violence
module which might help.

My experience using ICD-9 was difficult, it was hard to access firearms
data from this.

Lois: ICD-10 is even harder. Using CDC Wonder (interactive software)
it's pretty simple, however, to get to any mortality data.

Do the differences in ICD-10 provide for better descriptions of types of
firearms?

Lois: It is better in ICD-10, but is still not complete.

Are there materials for doing historical analysis?

Lois: Yes, there are materials on which the codes are comparable to each
other across time. 

How do you determine "undetermined intent"?

Lois: When there isn't enough information to classify, we classify as
undetermined intent.

How good are the surveillance systems on mortality and morbidity for
looking at homicide and lethal violence?

Lois: Better than they were. For example, in the emergency room data,
very often due to malpractice and confidentiality issues, information on
intent does not get into the record. But, there is a big movement to get ER
personnel to enter the full information. The national data aren't the best
source or the most ideal source, but they are all we have access to. I think
it will be a good source eventually on lethal data. There is more money
going into the training of individuals for coding on these items, much
more extensive training for physicians. Physicians have to become
convinced that prevention is part of their job.

Regarding discharge data - this discharge data is only overnight. In New
York state, if someone doesn't stay overnight they weren't considered a
discharge.

Lois: I will check on that.

Hargarten: Increasingly, people are not even being admitted. They are
being observed for 8, 10, or 22 hours but not admitted, so researchers
need to look carefully at state data. And, having practiced in an ER, it is
very hard to get information in the ER on the incident. The Crash
Outcome Data Evaluation made links between medical and law
enforcement data. That's an exciting possibility, since there are inherent
limits to hospital or morbidity data.

When you have information on children, do you also have data on
parents?

Lois: Not from what we have in the mortality or morbidity data, we might
have that in the National Health Interview Survey.

What's available on tape versus CD-ROM?

Lois: Everything on data tapes is available on CD-ROM. I haven't used
the CD-ROM, but my understanding is that the level of the data on the
CD-ROM is the same. I understand you can get these free from ICPSR, if
not call or e-mail me to get them.

Are there specific reliability issues regarding child homicides?

Lois: In the details on cause it may be a problem. Child abuse is
underrepresented on death certificates. Much more is reported in other
sources.

-------------------------------

Injury Surveillance Using Data From the National Center for Health
Statistics

Lois A. Fingerhut, Special Assistant for Injury Epidemiology
Office of Analysis, Epidemiology and Health Promotion
National Center for Health Statistics
6525 Belcrest Road, Room 750, Hyattsville, Maryland 20782

Abstract

The mission of the National Center for Health Statistics (NCHS) is to
provide statistical information that will guide actions and policies to
improve the health of the American people. Many of the data systems at
NCHS can be used for injury surveillance. Most of these collect reliable
data that are E-coded. In this presentation, I will explain what is meant by
national injury surveillance and why E-codes are crucial for good
surveillance. In addition, the individual data systems will be explored in
terms of their strengths and weaknesses for injury surveillance.
Detailed information on data availability and documentation can be found
on the NCHS home pages: http://www.cdc.gov/nchswww/.

What is an injury surveillance system and why is it important?

Very broadly and very simply, a surveillance system is a data driven
system that continually or periodically collects information and data for a
particular purpose. In this case we are focused on injuries - optimally the
system will be able to be used to describe the circumstances, the person
injured, the timing, the location and perhaps most importantly both the
diagnosis of the injury and the external cause of the injury. The external
causes of injury are most often identified in the data classification system
by the World Health Organization's internationally accepted classification
scheme, the International Classification of Disease. Currently, the Ninth
revision is in use in this country. The chapter of the Ninth revision of the
ICD on the external causes is commonly referred to the E-code chapter. 

Why are E-codes important for surveillance?

E-codes help identify and prioritize injury problems. E-codes provide
information about both the event during which the injury took place and
about the individuals who were injured. When used correctly, E-codes
recreate a picture of the specific circumstances of an injury including the
"how" and the "where" the event occurred. (Children's Safety Network,
1998).

Value of E-codes: [based on unpublished material prepared by State of
Minnesota]

1-Calculating costs associated with treatment of injuries related to specific
causes

2-Development of community education programs to address injuries
from particular causes

3-Evaluation of impact of local laws and regulations 

4-Health system planning for program development

5-Identification of patterns in injury

The data that are captured with E-codes are used to measure trends, detect
patterns and identify risk factors for injury. The data systems can be
national, state or local. My role this morning is to focus on national
systems.

My remarks are limited to national data systems from the NCHS- despite
the fact that there are many other national sources for injury data . First
and very importantly, NCHS data systems are designed to be useful as
general purpose surveillance tools. Injury is but one component, and quite
frankly, a small one, but an exceedingly important one. There are different
ways one can classify NCHS surveillance systems, and for the purposes of
this presentation, they will be dichotomized into ones that collect data on
mortality vs. those that collect data on morbidity.

There are fundamental differences between E-code guidelines for
mortality vs. those for morbidity.

Mortality:
--Single event
--Underlying cause
--Single code
--Explains death event
--"Undetermined" intent requires official confirmation
--No updates in between revisions

Morbidity:
--Multiple episodes of care/different settings
--Refers to proximal/principal diagnosis
--Multiple codes
--Explains cause of injury
--"Undetermined" defaulted-unintentional (pre-10/96)
--Annual updates

An example: An elderly female "accidentally" falls down the stairs and
strikes her head on the wall. If she dies, the E-code on the death certificate
is for an "Unintentional fall" as the fall was the underlying cause or
initiating event. The injury to her head will be included on the death
certificate, not as the underlying cause, but as a contributing cause. It will
be coded from the death certificate but can only be found on the multiple
cause of death data tapes. If she survives the fall and is taken to the
emergency room, the principal diagnosis is related to her head injury and
the relevant E-code on the medical record is for "striking against an
object," as that is the proximal cause. If the information on the certificate
is questionable as to the intent of the fall, that is if it could not be
determined that the fall was unintentional or intentional (as in she was
pushed), the code would be for a fall of undetermined intent. Until
October 1996, the emergency department record would have been coded
to unintentional as there was no code guideline allowing for undetermined
intent; the default was to unintentional. For official morbidity coding
guidelines, see: http://www.cdc.gov/nchswww/datawh/ftpserv/
ftpicd9/ftpicd9.htm#guide.

Mortality

E-code guidelines

Use an E-code for the underlying cause when the morbid condition is
classifiable to ICD-9 800-999; assign E-code to initiating event (except
when initial event was "trivial" leading to a more serious injury or if
initial event was a slight injury (one that rarely causes death).

"Accident" due to disease condition: when a disease condition such as
heart attack or alcoholism is indicated as the underlying cause of the
injury event (accident), code to the injury event unless there is evidence
that the death occurred prior to the event with few exceptions (one of
which is accidents resulting from epilepsy).

For more detail, see: Instructions for Classifying the Underlying
Cause-of-Death, 1992, at http://www.cdc.gov/nchswww/about/major/
dvs/im.htm.

National Vital Statistics System

The National Vital Statistics is used by most people to describe the
epidemiology of injury mortality in this country. While most NCHS data
systems are sample based, the national vital statistics system is universal
in its coverage. NCHS mortality data can be analyzed at least 3 levels:
national, state and county level surveillance. In the decentralized vital
statistics of the U.S., death certificates are legal and statistical documents
of the states, not of the Federal government. Some degree of
standardization in the structure and content of the various death
certificates used by the states is achieved by their willingness, for the most
part, to adhere to a "model" certificate promulgated by NCHS.

Death Certificates

In the United States, two persons complete the information on the death
certificate. The bottom half of the certificate is the medical certification of
death which is completed by the attending physician, and in the case of
injury generally by a medical examiner, or coroner; and the top half,
which contains the demographic information, is completed by the funeral
director, who also has the ultimate responsibility for filing the certificate
with the appropriate state registration officials, who are custodians of the
original records. The state registration officials also have the authority and
responsibility to conduct queries for questionable or incomplete
information (such as follow up for death whose cause is pending
investigation), or where the particulars of an 'accident' or injury are not
adequately described.

For injury-related deaths, the U.S. Standard Certificate of Death has a
number of items including the date and time of the injury, whether the
injury occurred at work, a description of how the injury occurred, the
place of injury, and the actual street location of the injury. Clearly, the
death certificate is a potentially rich source of statistical information on
injuries. It is also instructive to note what the standard death certificate
does not ask regarding injuries. It does not, for example, ask explicitly
about drug or alcohol involvement; and it does not clearly specify the
degree of detail that is acceptable when describing how the injury
occurred. Moreover, it does not include prompts specific for accidents that
would encourage the medical provider to provide useful information in an
automobile accident for example as to whether the decedent was the driver
or a passenger.

Definition of underlying cause of death: 

The disease or injury which initiated the train of morbid events leading
directly to death, or the circumstances of the accident or violence which
produced the fatal injury.

Additional detail on the National Vital Statistics System can be found at:
http://www.cdc.gov/nchswww/about/major/dvs/mortdata.htm.

National Mortality Followback Survey (NMFS 1993):

The 1993 survey samples individuals aged 15 years or over who died in
1993. Forty- nine of the 50 State vital registration areas granted approval
to sample their death certificates, as well as the independent vital
registration areas of the District of Columbia and New York City. (South
Dakota declined to participate in the NMFS due to State law restricting
the use of death certificate information.) A sample of 22,957 death
certificates from the Current Mortality Sample was drawn. To meet
specific research needs, the sample included 9,636 death certificates
selected with certainty. There is an over-sample of death certificates to
obtain reliable numbers for important population subgroups; persons
under age 35, women, and the black population. 

The 1993 NMFS focused on five subject areas:
--Socioeconomic differentials in mortality
--Associations between risk factors and cause of death
--Disability
--Access and utilization of health care facilities in the last year of life
--Reliability of certain items reported on the death certificate

The 1993 NMFS is different from the five previous mortality followback
surveys in several ways. It emphasizes deaths due to homicide, suicide,
and unintentional injury. The subject areas are considerably broader.
However, many previously-surveyed subject areas are included for trend
analysis. The survey is the first to acquire national-level information from
medical examiners and coroners. The complexity of the questionnaire
necessitated telephone or in person interviews. 

The 1993 NMFS was designed in collaboration with other agencies of the
Public Health Service, Department of Health and Human Services, and the
National Highway Traffic Safety Administration. Several of these
agencies provided funding through NCHS's Reimbursable Work Program.
Results from the first release of data from the 1993 NMFS are available
on the FTP server. 

Additional detail on the survey can be found at:
http://www.cdc.gov/nchswww/about/major/nmfs/nmfs.htm.

General Mortality Issues: (Rosenberg and Kochanek, 1995)

Completeness of Death Certificate Information

Completeness of reporting is a critical element in the effective use of
death certificate information for injury prevention and control. For
example, it is important to know who the person injured is in motor
vehicle injuries, and to completely specify falls, and the type of weapon
when a firearm is the cause.

How can this be addressed? For one thing, better education of medical
certifiers is needed on how to complete the death certificate. NCHS has
initiated a number of efforts directed at physicians to improve cause-of-
death reporting beginning with two national workshops, one in 1989 and
the other in 1991. These initiatives are continuing. A second approach to
addressing this problem is querying at the state level. Death certificates
with incomplete information on injuries should not be permitted to pass to
the stage of processing without asking the medical certifier for sufficiently
complete information to make it useful for injury surveillance. These
initiatives need to be national in scope if they are to result in good
information on which to base injury prevention programs. 

Information Augmentation

It needs to be recognized that even if all the items on the death certificate
were answered completely and accurately, there would still be need for
additional information on injuries that is not routinely captured on the
death certificate, or, if captured, not in a standard, uniform, and
dependable way. Examples include whether drugs or alcohol may have
been involved in the accident. Without a direct question to the certifier
asking about substance abuse, one can expect as many studies have shown
that the impact of substance abuse on injuries cannot be adequately
measured using information on the standard death certificate. Additional
information from another source is needed to augment the information
routinely collected on the death certificate. 

What kinds of augmentation are possible? One type is what NCHS calls
"follow back" surveys. These are surveys using death certificates as a
sampling frame that can be used to get additional information on deaths
for a special subset of the decedent population, based on demographic
characteristics or on causes of death. The 1993 National Mortality Follow
back Survey (NMFS) was designed to provide national estimates of
important characteristics of the 2,218,940 people aged 15 years and older
who died in 1993. Last conducted in 1986 (21); the National Mortality
Follow back Survey focused on obtaining socioeconomic information
such as income, and information on health care in the last year of life.

Another approach to augmenting information reported on the death
certificate is by linking information reported on the death certificate with
that from another source. For example, the 1993 national mortality follow
back survey includes a component to link with abstracts of
coroner/medical examiner records. This will not only augment
information on the death certificate but will also be a useful basis for
checking the reliability of the cause of death reported by the same medical
examiner or coroner who completed the death certificate. 

The death certificate can be linked to a variety of other sources including
hospital records, health examination survey records, health interview
records, and administrative records -- each of which can potentially enrich
the mortality data base for injury research.

Validity and Reliability

The question of validity and reliability is one that suffuses information
from the vital registration system. The death certificate, and in particular
cause of death, is always a prime suspect in these investigations. Many
studies have been published on the validity of cause of death reflected in
the NCHS annotated bibliography of 128 such studies carried out over a
period of 23 years, with an update published in 1991.

Some of these studies raise troubling questions regarding the medical
certification of death, but these have been largely in the area of natural
causes, or deaths related to disease processes of relatively long duration.
For injuries, the cause of death tends to be more clear-cut and immediate
in its fatal action. Nevertheless, questions of validity do often arise
regarding manner of death, that is, whether the injury was accidental,
suicidal, or homicidal. Only in-depth studies can shed light on this, and,
even in some cases, the basic records will not reveal what the medical
certifier has chosen not to report. 

E-coded mortality data are useful to the extent that they reflect accurate,
specific information about the circumstances surrounding the fatal injury-
causing event that are recorded on the death certificate. The specificity of
these data could be improved by:

---encouraging greater specificity in reporting and avoiding use of
generalized codes, such as "fracture, cause unspecified", "unspecified
accident" and "assault by unspecified means;"

---providing sufficient narrative detail in the item "how the injury
occurred" on the death certificate in order to enhance the information on
the cause-of-death section of the certificate. It is particularly important to
identify when an agent (e.g., consumer product, type of motor vehicle) is
involved in an injury, as well as specific information about the agent and
the injury scenario, since that information is not routinely captured in the
E code.

---educating medical certifiers about the usefulness of E-coded mortality
data. (McLoughlin, Annest, Fingerhut, et. al., 1997)

ICD 9 vs. ICD 10 for mortality (Fingerhut, Rosenberg, Kochanek and
Pickett, 1998)

The United States will begin coding its national mortality data using ICD-
10 beginning in 1999. Major changes have been made from ICD-9 to
ICD-10 in terms of both diagnostics codes as well as external cause of
injury codes. For example, the external cause of injury codes are no longer
a supplementary chapter of the ICD. All ICD chapters are divided into an
alphanumeric coding scheme of one letter and two numbers at the 3-digit
level with decimal subdivisions for the 4th digit. Codes for external causes
of injury are found in Chapter 20 and use letters V, W, and X- and thus
are definitely not "E-codes". Injury diagnostic codes are found in Chapter
19 and use letters S and T. The codes in ICD-10 are multi-axial in
concept, in that there are requisite codes for injury incidents for place of
occurrence and for activity the victim was involved in when the death
occurred. Transportation related mortality codes have undergone a major
revision. The letter "V" is used for transportation related injuries with the
first subdivisions being for the victim's mode of transport (for example,
pedestrian, occupant, pedal cyclist); the third character identified the
victim's counterpart or the circumstance of the accident (collision with
vehicle, noncollision). The fourth character identifies the activity of the
victim (driver, passenger) and whether the incident occurred in traffic or a
non-traffic situation. Other examples of significant changes in this
chapter: "fracture not otherwise specified" which was classified with Falls
in ICD-9 is now classified with "exposure to unspecified factors".
Homicide codes will now include more detailed codes for abuse, neglect
and abandonment and contain codes for perpetrator. Late effects codes are
now combined in one section rather than being placed with relevant
sections of unintentional, suicide or undetermined intent. 

Morbidity

National Surveys: 

National Hospital Discharge Survey

In this survey, data are collected from a sample of records from a sample
of hospitals. In 1995 data were collected for about 263,000 discharges
from 466 hospitals. Data from this survey can be used to make national
estimates of hospital discharges. The injury diagnostic information that is
derived from this survey related primarily to the diagnosis, that is ICD 9
CM codes 800-999.

In 1994, only about half of the medical records for which an injury was
the principal diagnosis had an accompanying E-code.- This proportion has
increased remarkably to 64% in 1996 as the number of states mandating e-
codes has increased. Information is likely in the detailed patient record but
not on the summary face sheet from which the abstract information is
gathered.

Hospital records are for identification of severe nonfatal injury discharges
not people. There can be multiple discharges for the same person.

For more information on the survey, see:
http://www.cdc.gov/nchswww/about/major/nhcs/nhcs.htm#nhds.

National Hospital Ambulatory Medical Care Survey (NHAMCS)
Emergency Department component

NCHS also conducts a family of surveys related to outpatient care. The
one that is most useful for injury surveillance is the NHAMCS that was
begun in 1992. In this survey, both ICD codes for injury diagnoses as well
as for external causes of injury can be captured. In this survey,
information is abstracted for a systematic random sample of visits during a
randomly assigned 4-week reporting period. 

In 1995, patient record forms were completed for 22,000 visits to the
emergency department and for 28,000 visits to outpatient clinics. Overall,
injuries represented about 40% of visits to the emergency department;
approximately 85-90% of those visits had an associated E-code.

For more information on the survey, see:
http://www.cdc.gov/nchswww/about/major/ahcd/ahcd1.htm.

For detailed injury data from the NHAMCS-ED, see:
http://www.cdc.gov/nchswww/products/pubs/pubd/series/sr13/pre-131/sr1
3_131.htm.

National Health Interview Survey (NHIS)

Historically, the NHIS has not been a rich source of data on cause of
injury. Data were collected in a national sample of households on
conditions that either caused some kind of restricted activity or resulted in
medical attention. These are the injuries at the base of the injury pyramid.
Beginning in 1997, the core questions of NHIS were redesigned including
an entire section on injury- including the verbatim text of how the injury
occurred. There are as yet no data, but we are looking forward to seeing
some preliminary results in the next few months.

For more information on the survey, see:
http://www.cdc.gov/nchswww/about/major/nhis/nhis.htm.

State role in Morbidity Surveillance

While E-codes for mortality have more or less been taken for granted in
this country and elsewhere, E-codes for morbidity are considerably rarer,
although things are improving quickly. As of this year, 34 States and the
District of Columbia have been collecting statewide hospital discharge
data with cause of injury coding as a part of each record. Fewer have
systems that are based on emergency departments. A survey was
conducted and completed recently that polled states on the kinds of
external cause injury data they were collecting. Highlights include:
Approximately 80% of states have a statewide Hospital discharge data
system. Most of those routinely collect some level of E-codes. Not all,
however, collect them as part of a mandate. RI and Wash were the first
states to mandate E-coding in 1989; most other states implemented in the
early 1990's. (APHA, 1998)

There is a fair amount of variation across States in the level of detail
collected. Fewer states (12) have ED systems and 11 routinely collect
some level of E-codes. State role- number of states has been increasing
rather sharply during the 1990's.

General morbidity data quality issues

The issues of concern with regard to E-codes that I spoke of in terms of
mortality are equally relevant for morbidity: completeness of the
information (better education of medical certifiers; of hospital personnel);
instead of querying at the state level - query at the hospital level; augment
the information routinely collected by, for example, linking information
reported on the hospital record with that from another source (for example
emergency medical services records)

ICD 10 CM (Fingerhut, Rosenberg, Kochanek and Pickett, 1998)

The clinical modification is due to be implemented in 2001. The current
draft of ICD-10-CM contains a significant increase in the number of codes
over ICD-10 and ICD-9-CM. Notable improvements in the content and
format include: the addition of information relevant to ambulatory and
managed care encounters; expanded injury codes; the creation of
combination diagnosis/symptoms codes to reduce the number of codes
needed to fully describe a condition; the addition of a sixth character;
incorporation of common 4th and 5th digit subclassifications; laterality;
and greater specificity in code assignment. The new structure will allow
further expansion than was possible with ICD-9-CM. In ICD-10, and
more importantly in 10 CM, the external cause of injury codes are no
longer a supplementary chapter of the ICD. 

ICD-9 was often criticized for its single axial approach to external causes
of injury as it was not very effective for injury prevention initiatives. In
ICD-10 and 10-CM there are requisite codes for all fatal and nonfatal
injury incidents for place of occurrence and for the activity the
victim/patient was involved in when the injury occurred. Transportation-
related ICD codes have undergone a major revision to focus on the victim
and the type of vehicle involved, rather than the other way around; the
letter "V" is used for transportation related injuries with the first
subdivisions being for the victim's mode of transport (for example,
pedestrian, occupant, pedal cyclist); the third character identified the
victim's counterpart or the circumstance of the accident (collision with
vehicle, noncollision). The fourth character identifies the activity of the
victim (driver, passenger) and whether the incident occurred in traffic or a
non-traffic situation. Other examples of significant changes in this
chapter: homicide/assault codes will now include more detailed codes for
abuse, neglect and abandonment and contain codes for perpetrator; and
late effects codes are now combined in one section rather than being
placed with relevant sections of unintentional, suicide or undetermined
intent.

Diagnosis codes have also been revised; the major subdivisions for
diagnosis codes are by body part rather than by type of injury as in ICD 9.
For example, they are for head, neck, hip and thigh, knee and lower leg-
rather than fracture, open wound, or superficial injury. Each of these type
of injury categories is specified with body part. 

In ICD-10 CM, poisoning codes have undergone a major change in that
there will no longer be external cause codes; rather the intent
(unintentional, suicide or undetermined) will become an additional digit to
the poisoning diagnostic codes. ICD 10 CM will contain many more codes
than ICD 10 for mortality.

References

APHA, Data Committee of the Injury Control and Emergency Health
Services Section. 1998. How States are Collecting and Using Cause of
Injury Data (in press).

Children's Safety Network. 1998. See:
http://www.edc.org/HHD/csn/ecodes/misslink.html.

Fingerhut, L.A., Rosenberg, H.M., Kochanek, K.D., and Pickett, D. 1998.
The transition to ICD 10 and 10-CM: implications for injury research.
Abstract for 1998 APHA meetings.

McLoughlin, E., Annest, J.L., Fingerhut, L.A., et al. 1997. Recommended
framework for presenting injury mortality data. Morbidity and Mortality
Weekly Report, Recommendations and Reports, vol.46. No. RR-14 CDC,
Atlanta, Georgia.

Rosenberg, H.M ., and Kochanek, K.D. 1995. "The death certificate as a
source of injury data" Pages 8-1:17 in Proceedings of the International
Collaborative Effort (ICE) on Injury Statistics, Volume I. National Center
for Health Statistics. Hyattsville, Maryland.

-------------------------------

On Intimate Partner Homicides in Massachusetts

Linda Langford, Harvard School of Public Health -- 
Nancy Isaac, Northeastern University School of Law
Stacey Kabat, Peace at Home

Abstract

The FBI's Supplementary Homicide Report (SHR) is often used by
researchers to study intimate partner homicide since it is the only national
data source that includes information on the victim-offender relationship.
One problem confronting these studies is the extent of missing data in the
SHR, meaning that information on intimate partner homicide incidence is
incomplete. 

The current study examines the extent of this problem in Massachusetts
over a five-year time period. Multiple data sources were used to construct
a database of all intimate partner violence-related homicide (IPVH) cases
in Massachusetts from 1991 through 1995. These data were compared
with the SHR for the same time period. Results show that the SHR
identified only 71.1% of the intimate partner victims, incidents involving
multiple victims were vastly underreported in the SHR during these years,
and cases involving unmarried former partners were less likely to be
reported as partner homicides in the SHR. An evaluation of one
methodology designed to adjust for flaws in the SHR showed that
adjusting the SHR data on intimate partner homicides using this
methodology overestimated the actual rate of intimate partner homicides
during the study years.

Background

Numerous studies of intimate partner homicide have been conducted,
many of which have documented national trends. These studies employed
the only national homicide data source that includes information on the
victim-offender relationship, the Supplementary Homicide Report (SHR)
(Browne & Williams, 1993; Bureau of Justice Statistics, 1994; Fox, 1994;
Mercy & Saltzman, 1989; Plass, 1993; Straus, 1986). While the SHR has
important strengths, particularly the collection of homicide data nationally
and the inclusion of the victim-offender relationship, some investigators
have described shortcomings of the SHR system and cautioned against
uncritical use of the data. Because it is a voluntary system, some law
enforcement agencies do not take part in the UCR system and therefore
some homicides are never reported (Williams & Flewelling, 1987). Even
those agencies that do participate sometimes fail to submit reports for a
given month, and the extent of this nonreporting may vary across agencies
and types of cases. In addition, homicides tabulated by the UCR may not
be followed up by a SHR form, which is revealed in the slight discrepancy
between the number of homicides reported by the UCR and SHR (Brewer,
1993; Williams & Flewelling, 1987). In addition, there is substantial
missing data on victim-offender relationships within reported SHR cases,
which limits detection of intimate homicide cases. Some investigators
have compensated for the limitations of the SHR, or indeed of any single
data collection system, by using multiple data sources to ensure
identification of the greatest number of homicide cases and to increase the
amount of contextual information collected (Keppel & Weis, 1992; Rand,
1993). The current study uses multiple data sources to identifyand
characterize intimate partner violence-related cases. In so doing, an
attempt was made to detect the maximum number of cases and compile
detailed data about the circumstances of each homicide.

Method

Data Sources

A victim-based database of all partner homicide cases from 1991 through
1995 was compiled from news articles, Supplementary Homicide Reports,
lists assembled by District Attorney's offices, and reports from domestic
violence advocacy agencies. This database will be referred to as the "study
database." Because SHR reports do not contain names, cases identified
through the SHR were matched with death certificates to identify
individuals by name. The study database is intended to capture all cases of
Massachusetts residents killed in IPV-related incidents from 1991 through
1995. Five out-of-state residents were killed in IPV-related incidents in
Massachusetts during the five-year study period, and these cases are
excluded from the present analysis. Our sample does include four
Massachusetts residents killed in adjacent states. There may have been
Massachusetts residents killed in IPV-related incidents in more distant
locations, but we did not identify any cases of this nature.

Data Analysis

The study database was compared with SHR files to examine the number
and types of IPV-related cases that are unreported or miscoded in the
SHR. First, we compared the number of Partner Victims in the study
database with the number of Partner Victims reported in the SHR. Cases
in the study database were then individually matched to SHR cases using
city and county; victim-offender relationship; victim's age, race, and sex;
offender's age and sex; and weapon. Although SHR records always
contained data on the city, month, and year of the homicide, at times
missing victim or offender data in the SHR made it difficult to establish a
definitive match. In those instances, we compared the SHR record with
the full list of death certificates to determine whether the SHR case could
conceivably be a match for a different homicide case. If there was no
other homicide case among the death certificates that matched the SHR
record, it was assumed to be a match for our case. Finally, we compared
the number of multiple victim homicides found in the study database with
the number reported in the SHR. Note that the SHR files were also used to
help detect cases, so these analyses do not compare the completeness of
other sources with that of the SHR. Rather, we are comparing a list
compiled from all available sources (as close to 100% case finding as we
could accomplish) with the SHR files.

We also evaluated one methodology that has been developed to adjust for
known flaws in the SHR. Williams and Flewelling (1987) created a
methodology to compensate for unreported cases and missing data
whencalculating homicide rates based on the SHR. To adjust for
nonreporting agencies, a weighting factor is calculated based on the
difference between the FBI's overall victim count in the UCR and the
number of victims reported in the SHR. To compensate for missing
victim-offender relationship data,Williams and Flewelling suggest two
possible procedures whereby the homicide rate can be adjusted by
extrapolating the characteristics of the known data to the unknown cases.
Both procedures involve starting with the number of intimate partner
homicides among the cases in which the victim-offender relationship was
reported and then adding a percentage of the cases in which the victim-
offender relationship was unknown. The rate is then calculated based on
the estimated total number of intimate partner homicides among cases
with both known and unknown relationships. We applied these procedures
to the SHR for 1991-1995 intimate partner homicide cases and compared
them to the rate of Partner Victims only (excluding Other Victims)
calculated from the study database. We applied these procedures to the
SHR for 1991-1995 intimate partner homicide cases and compared them
to the rate of Partner Victims only (excluding Other Victims) calculated
from the study database.

Results

Comparison of Study Database With SHR

There were 149 Partner Victims killed during the five year period, with a
range from 25 to 40 victims per year. A comparison of the study database
with the SHR shows that, compared to the 149 Partner Victims in the
study database, the SHR reports only 106 (71.1%) of the Partner Victim
cases in our study. Examining the yearly data reveals that the SHR data
steadily improve over the 5-year period, from reporting 56.0% of intimate
partner cases compiled in the study database in 1991 to reporting 85.0%
of these cases in 1995. Despite the improvement, in every year the SHR
underestimates the true number of victims, which suggests that the
problem with underreporting in the SHR is persistent.

We then examined each case in our study to determine whether there was
a corresponding case in the SHR. Each one of our cases was classified as
"matched" (matched an SHR case and the victim-offender relationship
recorded there was correct), "miscoded" (matched an SHR case but the
victim-offender relationship was not coded there as a partner relationship),
or "unreported" (no SHR case matched the victim, offender, and incident
information in the study database.) In some cases, the victim-offender
relationship was coded differently than our record but did indicate an
intimate partner relationship (e.g., a relationship that we had coded as
"girlfriend" was coded "wife" in the SHR.) For this analysis, a relationship
was only considered miscoded if we had recorded the victim-offender
relationship as an intimate partner relationship but the victim-offender
relationship of the matching SHR case was coded as a non-partner
relationship.

Of 149 Partner Victim cases, 59 (39.6%) were either unreported or
miscoded in the SHR. This number is larger than the discrepancy of 43
cases reported above, due to erroneous and duplicate entries in the SHR,
plus intimate partner homicide cases in the SHR that did not meet our
inclusion criteria. These errors result in an inflation of the SHR totals.

Of the 59, 34 cases (22.8% of Partner Victim cases) had no match in the
SHR. In 25 cases (16.8% of Partner Victim cases), the relationship was
coded as a non-intimate relationship. Therefore, these cases would not be
identified in an analysis using SHR data that counted intimate partner
homicide victims based on victim-offender relationship. Examination of
the actual coding of the 25 miscoded cases reveals that, in the majority of
cases, the incorrect relationship was recorded as "unknown" (n=10) or
"acquaintance" (n=7). The remainder were coded as "other known" (n=4),
"friend" (n=2), and "daughter" (n=2). The two cases in which partners
were coded "daughter" results from a systematic problem in multiple
victim homicide data.

Examining the unreported cases more closely reveals that these cases
often occurred in smaller jurisdictions. Of 34 unreported cases, 21 (62%)
occurred in cities with populations less than 30,000. These communities
may not have full-time law enforcement personnel and agencies may not
be familiar with reporting procedures, as homicides occur less often in
these localities. In contrast, larger jurisdictions were more likely to report
cases with miscoded victim-offender relationships. Of the 25 miscoded
cases, 72% had occurred in cities with populations greater than 90,000,
with the majority of these cases being reported by Boston.

We investigated the coding of different types of relationships in the SHR
to determine whether certain relationships were more likely to be
miscoded. These data show that, among cases that were reported in the
SHR, ex-girlfriends and ex-boyfriends were coded as intimate partners
59% of the time, while current boyfriends and girlfriends were coded as
intimate partners 83% of the time. These findings should be interpreted
with caution, due to the fact that they are based on a relatively small
number of cases, but they do suggest that the absence of codes for ex-
girlfriend and ex-boyfriend in the SHR may contribute to the problem of
miscoding. There was also a higher proportion of unreported cases among
ex-girlfriends and ex-boyfriends (39%) than among current girlfriends and
boyfriends (16%).

Statistical Adjustment for Non-Reporting Agencies and Missing Data

The five-year rate of intimate partner homicide based on the study
database (the "true" rate) is 3.10 per 100,000 persons. Calculating this
same rate based on SHR reports, without using any adjustment
procedures, yields a five-year rate of 2.45 per 100,000 persons, an
underestimate of the true rate. Application of Williams and Flewelling's
first adjustment procedure results in a five-year rate of 5.02 and the
second adjustment yields a figure of 5.75 per 100,000 persons, both of
which overestimate the true rate substantially. 

Multiple Victim Homicides

Many studies using the SHR exclude multiple victim homicides because
investigators theorize that these events are different from single victim
incidents and constitute a very small proportion of cases. We examined
the reporting of multiple victim homicides in Massachusetts during the
study years. Our data show that 15 of 175 incidents (8.6%) had more than
one victim. There were 34 victims in these 15 incidents (18% of the total
194 cases). The SHR for the same time period reports only 4 incidents
involving 9 victims.

References

Brewer, V. (1993). Issues in the use of Supplemental Homicide Reports.
In C.R. Block & R.L. Block (Eds.), Questions and Answers in Lethal and
Non-Lethal Violence: Proceedings of the Second Annual Workshop of the
Homicide Research Working Group (pp. 87-88). Washington, DC: U.S.
Department of Justice.

Browne, A., and Williams, K.R. (1993). Gender, intimacy, and lethal
violence: trends from 1976 through 1987. Gender and Society, 7(1), 78-
98.

Bureau of Justice Statistics. (1994). Violence between Intimates (NCJ-
149259). Washington, D.C.: Bureau of Justice Statistics, U.S. Department
of Justice.

Fox, J.A. (1994). Domestic homicide in America: trends and patterns for
1976-1992 (unpublished ). Boston, MA: Northeastern University College
of Criminal Justice.

Keppel, R.D., and Weis, J.G. (1992). Improving the Investigation of
Murder: An Analysis of the Effect of Time and Distance Relationships in
Murder Investigations (NCJ-138620). Washington, D.C.: National
Institute of Justice.

Mercy, J.A., and Saltzman, L.E. (1989). Fatal violence among spouses in
the United States, 1976-85. AJPH, 79(5), 595-599.

Plass, P.S. (1993). African American family homicide: Patterns in partner,
parent and child victimization, 1985-1987. Journal of Black Studies,
23(4), 515-538.

Rand, M.R. (1993). The Study of Homicide Caseflow: Creating a
Comprehensive Homicide Dataset. Washington, D.C.: U.S. Bureau of
Justice Statistics.

Straus, M.A. (1986). Domestic violence and homicide antecedents. Bull
NY: Acad Med, 62(5), 446-465.

Williams, K.R., and Flewelling, R.L. (1987). Family, acquaintance, and
stranger homicide: alternative procedures for rate calculations.
Criminology, 25(3), 543-560.

-------------------------------

An Evaluation of the Completeness and Accuracy of SHR Data in
Chicago, 1993 and 1994

HRWG 1998 Intensive Seminar

Thomas D. Patterson, Daniel Dick and Carolyn Rebecca Block
Illinois Criminal Justice Information Authority

Discussant: Margo Wilson, McMaster University

At the request of the Bureau of Justice Statistics, the Illinois Criminal
Justice Information Authority conducted a case-by-case comparison of
homicide cases in the Supplemental Homicide Reports to Chicago
Homicide Dataset cases booked in 1993 or 1994, the SHR Quality Project.
This project posed two methodological challenges, database organization
and comparative analysis. Now it has been completed, however, the
results provide information about the quality of SHR data in Chicago, and
by extension, in other cities with similar situations. This presentation will
review the methods and the results.

Chart 1 is a schematic overview of the method we used for case-by-case
matching, given the constraint that the SHR files did not contain an ID
number. Using key variables and an intermediate-stage "transfer" file, we
were able to match each of 1,750 victim records in the Chicago Homicide
Dataset (CHD) with one and only one SHR victim record (Chart 2). These
1,750 records accounted for 1,674 separate incidents.

With the 1,674 matched cases, we were able to look at missing and
incomplete information in the SHR, relative to the CHD. Some key results
regarding missing offender information are in Table 4.12. The 93
incidents with missing offender information differed significantly from
the 1,581 in which offender information was not missing in mean
offender's age (slightly younger), offender's gender (male), weapon (more
handguns), gang motive (more victims were a rival gang member and
more motives were a gang altercation). However, there was no significant
difference in the race/ethnicity of the offender or in the time lag between
injury and death.

Although CHD cases missing completely from the SHR differed from
those that were not missing in that almost all of the missing cases had a
lag time between injury and death, this was not true of the homicides in
which the case was present in the SHR but the offender information was
missing. However, the 93 offender-missing cases were significantly more
likely to have a lag between the date of injury and the date of arrest. This
suggests possibilities for improving the quality of SHR data.

-------------------------------

The Surveillance Value of "Bad" Data: Using Obliterated Serial Number
Data in a Firearm Surveillance System

Bill Sherlock, Illinois State Police
David Krieghbaum, Bureau of Alcohol, Tobacco and Firearms

Purpose

To initiate a project through the Crime Gun Analysis Branch to study
current and potential methods to be utilized in the restoration of
obliterated serial numbers on crime guns.

The Crime Gun Analysis Branch (CGAB) has recently initiated a project
aimed at the restoration, collection, and analysis of data relative to
firearms that have been recovered with their serial numbers removed by
methods of obliteration.

Historically, obliterated serial numbers on crime gun recoveries are not
raised unless there is a specific investigative requirement, i.e., homicide.
When there numbers are raised or even partially raised, NCIC and TECS
are used to research the status of the weapon and subsequently report
positive results to the requester. This information provides a firearms
trafficking investigative lead which should be proactively pursued because
any firearm that has had the serial number intentionally obliterated has
only one purpose--crime. Unfortunately, because of the narrow scope of
jurisdictional enforcement in most city and State police departments, these
trafficking leads are ignored, often deferring investigative efforts to the
local case at hand. If this information is incorporated into Project Lead
(which is available in the upgraded version), queries can be made to link
information on persons involved in trafficking crime guns using methods
of serial number obliteration. The mere fact that a serial number has been
obliterated is an absolute "intent to traffic" indicator.

Unfortunately, there has been no coordinated effort to proactively target
leads on obliterated serial numbers seized by Federal, State, and local
agencies. Labs are swamped and unable or unwilling to handle the
potential number of obliterated crime guns now sitting in evidence vaults
across the country. These guns are "one step from the burner," yet the
information that could be generated from just a few restored numbers is
still untested, and intelligence on firearms traffickers remains
uninvestigated, because the firearms are being destroyed along with
valuable serial number information. Also, due to lack of resources and
manpower in the field, firearms with obliterated serial numbers are very
seldom investigated because ATF does not routinely try to raise these
numbers despite the fact that criminal intent is absolute when serial
numbers are obliterated. These are the best leads available to start
investigating and removing crime gun traffickers from the streets.

The NTC has determined that raising serial numbers is relatively easy to
learn and inexpensive to render. For nonmagnetic guns, generally the
cheap handguns (Raven, Lorcin, Davis, etc.) that predominately turn up as
crime guns, the investment is approximately $60 of chemicals that can be
mixed in individual batches that have a 30-day shelf life.

Magnetic guns require a different chemical mixture with a similar price
and shelf life. Additionally, magnetic techniques can also be applied in
combination with chemical techniques for greater success in restoring
obliterated markings on magnetic surfaces. This opens opportunities in
geographic areas that recover magnetic crime guns of a higher quality
(Smith & Wesson, Colt, etc.). The training for certification to restore
obliterated serial numbers takes only two days and is a relatively simple
concept.

The NTC would like to propose the initiation of an obliterated serial
number research program to be housed at the NTC. This program would
consist of one special agent, currently detailed to the NTC as a Project
Officer, and administrative support. The purpose of this program would
be to train special agents in the field on the importance of raising serial
numbers in relation to firearms trafficking. In order to provide this
information to the field, the special agent detailed to the NTC would
travel to each field division and obtain information on possessors
associated with obliterated serial numbers and raise serial numbers from
crime guns in State and local police department vaults. Once this
information is obtained, these firearms will be traced and this information,
along with possessor information, will be entered into Project Lead for
link analysis.

The NTC recognizes the insufficient manpower in the field and would, in
addition to entering this information into Project Lead, research
obliterated gun information to build the foundation of firearms trafficking
investigations to be referred to the field. 

As a basis for this proposal, the NTC has received a list of some 67
restored serial numbers from obliterated serial number firearms sent to the
NTC from the Boston Field Division. With the first raised serial number,
the NTC traced the firearm and queried the possessor information. The
NTC discovered that the individual identified as the purchaser has
purchased more than 430 firearms by way of past multiple sales and has
since referred the information to the New Orleans Field Division, which
has opened a substantial firearms trafficking investigation. This was just
one firearm, and more than 430 other firearms were found to be associated
to the individual.

Imagine the firearms trafficking cases sitting in vaults just waiting to be
investigated. This is a major avenue that has not even been touched.
Imagine the impact this could have to enhance ATF's Firearms Trafficking
Strategy. The results could be phenomenal because never before has a
Federal law enforcement agency chosen to specifically target those
individuals associated with the obliteration of serial numbers on crime
guns.

-------------------------------

Situational Factors Related to Public Mass Murder Incidents: 1965-1998

Thomas A. Petee, Department of Sociology, Auburn University, Auburn,
AL 36849-5209

Abstract

This project analyzes 124 public setting mass murder incidents that
occurred in the United States between 1965 and 1998. Situational factors
such as the primary location of the murders, weapon choice, duration of
the event, and how the incident ended were examined to determine if any
patterns exist in mass murder events. 

Analyzing Mass Murder

Recent events in Pearl, Mississippi, West Paducah, Kentucky, Jonesboro,
Arkansas, and Springfield, Oregon, have resulted in a renewed interest in
public episodes of mass homicide. The present study examines 124
incidents of mass murder, defined as the murder of three or more people
in one place at one time (see Petee, Padgett and York, 1997 for a detailed
discussion of what constitutes mass murder), that occurred in public
settings in the United States between January 1965 and May 1998. 

Despite concerns over the recent murders that have taken place in school
settings, there appears to be no particular pattern to the specific place
where public mass murder incidents occur (see Figure One). While
restaurants were the most frequent place where mass murders transpire
(16.9% of all incidents), there was no single dominant setting for these
events. In some cases, the murders occur at an "at-risk" location (e.g.,
retail/grocery stores for felony-related mass murders). In other cases, the
location has some significant meaning for the offender (e.g., the law firm
targeted by Gian Luigi Ferri in San Francisco in 1993), or is situationally-
related (e.g., in direct conflict situations- see Petee et al., 1997). Given the
differences found in offender motivation, the variation in location for
mass murder is not all that surprising.

Frequently, discussion of mass murder tends to focus on issues related to
firearms. Typically, mass murder incidents such as that perpetrated by
Colin Ferguson on a commuter train in New York City in 1993, or by
Joseph Wesbecker in a printing plant in Louisville, Kentucky in 1989 will
generate a heated debate over the merits of gun control (Petee, York and
Padgett, in press). Indeed, an examination of the weapons used in public
incidents of mass murder (see Figure Two) reveals that firearms are the
weapons of choice for this type of homicide. In only 12.1% of the cases
did the offender(s) use a weapon other than some form of firearm. While
it is hard not to conclude that the presence of firearms makes for a more
lethal encounter (the exception being explosives or arson), the actual
impact of gun control on the occurrence of mass murder is debatable. A
closer examination of these incidents indicates that in most cases the
weapons used were legally obtained, and were not of the variety
commonly banned by gun legislation (i.e., "assault" weapons).

The response of law enforcement agencies to these types of incidents is
another commonly cited concern pertaining to mass murder. Figures
Three and Four examine the duration of mass murder incidents and how
these incidents ended respectively. In the majority cases, the homicide
episode lasted 5 minutes or less (58.9%), and the incident ended with the
offender fleeing the scene (57.3%). These findings certainly call into
question what impact, if any, law enforcement could have on these
incidents. 

References

Petee, T.A., K.G. Padgett, and T.S. York. 1997. Debunking the
Stereotype: An Examination of Mass Murder in Public Places. Homicide
Studies 1 (4): 317-337.

Petee, T.A., T.S. York, and K.G. Padgett. In Press. "Constructing Mass
Murder: The Stereotypical Representation of Mass Homicide." In Deviant
Behavior: A Case Study Approach to Understanding Social Problems,
Dennis L. Peck, ed. Dix Hills, NY: General Hall Publishers.

-------------------------------

D.C. Jail Gun Identification and Style Survey

John May, Khalid Pitts, Earnest Williams, Roger Oen
Central Detention Facility Health Services
1901 D Street S.E., Washington, DC 20003

Introduction

Urban jail inmates report much familiarity with firearms. This project
examined and compared the ability of different age groups of inmates to
match the name of a gun to its photograph. Additionally, the project asked
the different age groups to discuss features that they admired about
particular guns and to identify the gun that they would most want.

Method

The study was conducted during two weeks of July, 1997, at the
municipal jail for the District of Columbia. Pictures of nine different guns
on a worksheet were shown to 135 pre-trial detainees during voluntary
health education classes. The pictures were chosen from a handgun
catalog by three randomly selected male detainees as guns which they
believed would be recognized by fellow detainees. (The guns were the
Uzi, Tec-9, AK-47, Mac-11, Phoenix Arms Raven, Calico M110, Glock
19, Jennings J-25, and Colt Detective Special). See Table 1. Detainees
were asked to privately match the name of the gun to the picture of the
gun. On a separate worksheet, detainees were asked to identify the gun
they would want to have. Additionally, they were asked to list nicknames
for each gun and features which they admired, if any, about each gun. The
detainees indicated their ages on the worksheets, but no name or
identifying information.

Results

Ninety-six percent (96%) of the 15-19 year old males could correctly
match the manufacture's name to the gun. Each successive age group was
less successful in correctly identifying the guns in nearly a linear
distribution.

When asked to identify the gun that they would most like to have, younger
inmates chose weapons with a menacing appearance and reputation for
rapid fire and multiple rounds (the Calico was their first choice). Older
inmates preferred guns which were smaller (the Glock was their first
choice). Younger detainees listed admired gun features which represented
speed, destruction, or style. Older detainees listed admired gun features
which highlighted utility or practicality. The results suggest that younger
detainees are attracted to the manufacturing style and power of guns,
while older detainees are more attracted to the utility of the weapon.

-------------------------------

Exploring Circumstances of Prior Gunshot Wounds Among Jail Detainees

John May, Roger Oen, Khalid Pitts
Central Detention Facility Health Services
1901 D Street S.E., Washington, DC 20003

Introduction

Many detainees in urban jails have sustained prior gunshot wounds. The
purpose of this study was to explore the circumstances surrounding these
shootings and how the shootings impacted their lives.

Methods

Detainees entering the city jail in Washington, D.C., from April through
June, 1997, were screened for a history of gunshot wounds. One in 4 of
the males had been shot before. Extensive interviews were then done with
every third male detainee who reported a prior gunshot wound within the
past 5 years until 92 interviews had been completed.

Results

A whole continuum of injuries were represented, from single superficial
extremity injuries to extensive multiple wounds to the head and torso.
Two different clusters emerged based upon the seriousness of the
shooting, the circumstance of the shooting and the perceived intention of
the shooter. The first cluster involved injuries sustained during arguments,
retaliation, or assaults. These wounds tended to have the greatest potential
for death. Victims of these shootings were 2-3 fold more likely to have
been shot with multiple bullets. They also were more likely to sustain
injuries to the head or torso. See Table 1. 

This group also were more likely to believe that their shooter meant to kill
them. They were more likely to carry a gun after the shooting, more eager
to retaliate, and could identify more ways in which the shooting changed
their lives including substance abuse and "vicious" behavior.

The second cluster consisted of injuries sustained during robberies,
carelessness, crossfire, or police fire. These wounds tended to be less
threatening. Victims were more likely to sustain single wounds. They
were more likely to be injured in an extremity. See Table 2.

Members of this group were either uncertain of their shooter's intention or
did not believe their shooter meant to kill them. They were less likely to
carry a gun after the shooting, less likely to retaliate, and experienced less
life-altering sequela.

Further exploration of these findings might be useful in designing
interventions and responses to survivors of firearm injuries, particularly in
interrupting a response of gun-carrying and retaliation. 
-------------------

The Spousal SROK Revisited: A Comparison of Houston and Chicago
Intimate Partner Homicide

Derek J. Paulsen, M.A.
Victoria E. Brewer, Ph.D.
The Houston Homicide Project
College of Criminal Justice, Sam Houston State University

Funded by the National Institute of Justice and the National Consortium
on Violence Research. 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.

In a provocative cross-cultural examination of intimate partner homicide,
Wilson and Daly (1992) found that the spousal "sex ratio of killing"
(SROK) was approximately twice as great in the U.S. as other Western
nations, including Canada, Australia, and Great Britain.

In this analysis of Intimate Partner Homicide in Chicago and Houston, we
find both notable similarities and differences in the SROK when
disaggregated by race/ethnicity, weapon used, and coresidency.

Wilson, Margo I., and Martin Daly. 1992. "Who Kills Whom in Spouse
Killings? On the Exceptional Sex Ratio of Spousal Homicides in the
United States." Criminology 30, 2: 189-215.

-------------------------------

Spatial-Temporal Clustering of Chicago Homicides

Margo Wilson and Martin Daly
Department of Psychology, McMaster University
Hamilton, Ontario, Canada \L8S 4K1

Abstract

There was significant (P.01) nonrandom spatial-temporal clustering (on a
scale from 0 to 5 km and 0 to 30 days) of those homicide incidents in
Chicago, Illinois, (1965-1990) in which men killed unrelated men in the
context of gang activities and other social conflicts, but not in robbery-
burglary homicide incidents. The GIS spatial point patterning algorithms
(Splancs ) scale for the number of incidents that would be expected
given the extent of spatial and temporal clustering, considered separately.

Between 1965 and 1990 in the city of Chicago, 19,335 people were
homicide victims. There was marked variation in the number of victims
from year to year (Block and Christakos, 1995) and from neighborhood to
neighborhood (Wilson and Daly, 1997), but was there clustering of
homicides in space and time? Asynchronous deterioration or improvement
of different neighborhoods would be expected to cause nonrandom
spatial-temporal clustering on a relatively long (e.g., annual) time scale,
but we were interested in whether there might also be nonrandom spatial-
temporal clustering on a scale of days or weeks, and blocks. Our interest
derives from the expectation that any such clustering might be a signature
of one or more of the following social processes: (1) social contagion,
including copy-cat murders, power struggles, revenge, and retaliation; (2)
nonrandom distribution in space of temporal changes in factors that
elevate the risk of homicide; (3) changes in market dynamics.

How to Measure Clustering in Space and Time?

Peter Diggle and colleagues (Diggle et al., 1995; Gatrell et al., 1996) have
devised GIS spatial point patterning algorithms which treat each incident
(here a homicide) as a focal point for computing observed and expected
numbers of additional incidents within a continuously varying radius (s)
and time interval (t), in order to determine the degree of space-time
clustering in excess of that which would be expected given the extent of
spatial clustering and the extent of temporal clustering, considered
separately. The resulting numerical value for each s/t combination is a
standardized index denoted as D/SE. If there were no space-time
clustering, the surface of the 3-dimensional graph (with axes of s, t, and
D/SE) would be relatively flat, even if temporal trends and spatial
clustering existed. In order to assess whether the 3-dimensional surface
pattern is statistically improbable, Monte Carlo simulations are run
utilizing the same degree of spatial clustering and temporal clustering. If
the sum of the standardized indexes (sum of all D/SE values) is greater
than 99% of the obtained simulation values, then P less than .01.

We limited analysis to distances of 0-5 km between any two homicide
incidents and intervals of 0-30 days. Incidents are the units of analysis.
We asked whether there was nonrandom spatial-temporal clustering of
men killing unrelated men in gang-related homicides, other "social
conflict" homicides, and robbery-burglary homicides in Chicago over the
period 1965-1990.

Results and Discussion

Figure 1 shows the 3-dimensional surface pattern of the standardized
index, D/SE, of spatial-temporal clustering. Index values greater than 2.0
are likely to be statistically improbable as assessed with Monte Carlo
simulations. The probability of the overall surface pattern was statistically
significant (P less than .01) for gang-related incidents and for other social
conflicts, but the robbery/burglary incidents was not significant (P=11). 
Note that the surface of the robbery/burglary graph is relatively flat
compared with the other two kinds of homicide incidents.

Diggle's method allows one to detect space-time clustering at whatever
scale such clustering may actually exist. Interpretation of the observed
patterns still requires theory, appropriate control comparisons, and
imagination. Interpretation of the three graphs in Figure 1 requires
consideration of the social processes and structural forces that underlie the
different kinds of homicide incidents.

Acknowledgments

We thank the Chicago Police Department for the data, Richard Block for
the geo-coded locations, and Patrick DeLuca, Peter Diggle, Barry
Rowlingson, Anthony Gatrell, and Steve Reader for their expertise and
assistance with GIS. This work was supported by the Social Science and
Humanities Research Council of Canada and by the John D. and Catherine
T. MacArthur Foundation.

References

Block, C.R., and A. Christakos. 1995. Chicago homicide from the sixties
to the nineties. Pp. 17-50 in Trends, Risks, and Interventions in Lethal
Violence, C. Block and R. Block, eds. Washington, DC: National Institute
of Justice.

Diggle, P.J., A. G. Chetwynd, R. H„ggkvist, S.E. Morris. 1995. Second-
order analysis of space-time clustering. Statistical Methods in Medical
Research 4: 124-136.

Gatrell, A., T.C. Bailey, P.J. Diggle, B.S. Rowlingson. 1996. Spatial point
pattern analysis and its application in geographical epidemiology.
Transactions of the Institute of British Geographers, New Series 21: 256-
274.

Knox, G. 1984. Epidemiology of childhood leukemia in Northumberland
and Durham. British Journal of Preventative and Social Medicine 18: 17-
24.

Wilson, M., and M. Daly. 1997. Life expectancy, economic inequality,
homicide, and reproductive timing in Chicago neighborhoods. British
Medical Journal 314: 1271-1274.

------------------------------

Development of an Ultrasonic Method for Restoration of Obliterated
Serial Numbers on Firearms

Yari Yacobi
Guest Research in the National Institute of Standards and Technology
Gaithersburg, MD

Special presentation sponsored by the Bureau of Alcohol, Tobacco and
Firearms through an Innovations in Government Award from the Ford
Foundation and Harvard University

Introduction

There is a recognized need for a method which will be able to restore an
erased serial number on firearms.

One of the instruments that might help to achieve this goal is the acoustic
microscope. The basic concept that stands behind this attempt is that the
action of pressing the numbers on the surface of the firearm produces a
local stress concentration below the numbers. These stresses still exist
after the numbers have been erased. 

The cause for the stresses comes from the local strains caused by the
action of pressing the numbers into the material.

When the ultrasonic parameters are appropriate, then the acoustic
microscope is able to see these stresses. Mapping of the stress on the
firearm can differentiate between areas with and without stress. As a result
of this property, we might get a picture of the erased numbers. 

The results of the scanning probably won't give us the image of the full
numbers, but further processing of the picture might give us a clue and
even more of the original numbers.

Goals

--To find a method in which a firearm's serial number, that has been
erased, will be restored.

--To try to define the optimum dimensions and geometry for the letters,
which will give us the best capability for restoration of the numbers.

--To look for new methods to mark the serial number that will help the
restoration of it later, in the acoustic microscope.

The Acoustic Microscope Scanner

The scanner is based on a computerized ultrasonic system in which a
sample is held in a bath of a liquid, usually water, and scanned. As a
result, we a re receiving an image of the area, in which each pixel
represents the change in amplitude in that place. 

The changes in amplitude are usually caused by defects in the sample,
changes in the materials due to thermal treatment, changes in geometry,
stresses, etc.

Shear and Longitudinal Waves

The waves that penetrate the material split into two modes: one is
longitudinal waves, and the second is shear waves. They differ by their
polarity. The longitudinal waves vibrate in the direction of the beam, and
the shear waves are polarized perpendicular to the direction of the beam.

We are using the shear waves because they are more sensitive to stresses.

Leaky Waves

The leaky waves are waves that we are producing by working at a distance
much closer to the material, in that often we are making surface waves
that leak through the surface of the material. These waves are sensitive to
defects of the surface of the material, including surface roughness and
stresses at the surface. As we mentioned before, stresses that were caused
by the serial numbers can be detected by the stresses that are left after they
have been removed.

Experimental

The experiments have been done on a few metallic samples. We used steel
plates that were cut from firearms made by New England Firearms Co.
and steel plates that we got from the ATF.

The first plates were machined from their backside and then scanned in
the SCM (Scanning Acoustic Microscope), from their back using shear
waves. As a result, we got a mirror picture of the numbers.

The first step was to polish the numbers, in controlled steps, and then to
scan them again, until we didn't detect the numbers anymore. In this stage,
we measured the maximum depth in which the number can be detected by
this method. These experiments were done on both plates.

Results

We have tried to find a method to restore the numbers after they have
been erased, but we didn't succeed. We suspect that the problem lay on the
transducers that we have. We suspect that the field of stresses near and
below the numbers is small and for that we need a special transducer, that
at this stage we don't have.

Future Research

Our goals for future research are:

--To find a simple way to restore the serial numbers from a firearm after
they have been erased. In this research, we will need to find a good fit
between the material in use, the transducer, and the ultrasonic scanner. In
the meantime, we are working with the existing equipment.

--To define a standard for the serial number (the size and depth of the
letters), that will help us in the future to restore the numbers after they
have been erased.

--To define the location for the serial number in a place that will make the
restoration of it easy, but still the grinding of it will be difficult.

-------------------------------

Cohort survival projections of homicide Rates: Victimization Types

Allan F. Abrahamse, RAND Corporation, 1700 Main, Santa Monica, CA
90407

ABSTRACT

Cohort survival projections of homicide victims multiply the projected
number of people of a given age by the projected victimization rate for
people that age. Projecting age-specific victimization rates looks
complicated, because these rates do not merely rise and then fall, they fall
(from infancy to about age 12), rise (from age 12 to about age 25), fall
(from age 25 to about age 70) and then rise again. One approach may be to
partition homicides into a small number of different types, each of which
has a relatively simple relationship to age. This paper describes a set of
homicide types that meets this objective. 

Figure 1 shows the arrest rate for homicide by age.[1] While this figure
displays rates for all persons for a particular period of time, it resembles
what age specific arrest rates for a birth cohort would look like if we had
the data to draw such a curve. Rates rise sharply after about age 10, to a
peak at about age 20, and then decay steadily from that age on. It seems to
suggest a simple process of initiation and desistance. Because this
distribution can be described with just a few parameters, it is relatively
easy to project age specific homicide arrest rates for a birth cohort. 

As Figure 2 shows, the corresponding risk of being a homicide victim has
a much more complicated shape. It begins in the first year of life at a
relative high rate, falls quickly to about age 7 or 8, rises sharply during the
teen years, falls steadily until about age 75, then begins to rise again. It is
difficult to view exposure to homicide risk as a simple process of initiation
followed by desistance, and we cannot characterize the shape of this curve
with just a few parameters.

Figure 3 show victimization rates by age and sex, with rates for females
"pointing down". The figure is like a population pyramid on its side. Both
male and female rates have the same shape: a fall, followed by a rise,
followed by a fall, followed by a rise. For most homicide types, rates for
males are higher than for females.

Victimization types

While we can imagine that activities that affect the probability of getting
arrested for homicide might follow a simple lifetime trajectory of initiation
followed by resistance, we cannot do so for victimization. Instead,
perhaps, we can think of the net exposure to victimization as a
consequence of exposure to several different trajectories, each with its own
relatively simple characteristics. Once victimizations are so partitioned, it
may be possible to make projections for each type with some simple
model, and then combine them.

To explore this idea, I have partitioned homicides into ten mutually
exclusive classes. I present, one after the other, roughly in the order in
which each class seems to pose the greatest risk, from youngest to oldest. I
make no claim for originality; others have discussed much that will appear
below.[2]

In what follows, I will display the types roughly in the order of the age at
which they present the largest risk of victimization. I give each type a
rather simple title, but because the types are mutually exclusive, this title
does not fully describe the definition of the type. See the appendix for this
definition. 

As Figure 4 shows, infant victims are most likely to be killed by unarmed
relatives (here and everywhere else in this paper, by "unarmed " I mean
"unarmed with a gun"). The top graph shows the victimization rate
(victims per 100,000 at risk) by age and sex (males above the
horizontal axis, females , below it). The lower graph shows the percent of
all homicides in which the offender was known to be an unarmed relative,
by age and sex (males above the axis, females below). It is hard to
understand why the rate for male infants is nearly twice that of female
infants.

As Figure 5 shows, the other big risk faced by children is homicides by
unarmed friends and acquaintances (but not relatives). Homicides of this
type account for about a quarter of all children around the age of four.

In Figure 6 see that, while an arson homicide is a very rare event, it
accounts for a distinct fraction of all homicide victims ages about 5 to 10
years old. 

Figure 7 shows that homicides by armed friends and acquaintances
accounts for about a quarter of all homicides of children around the age of
10.

As Figure 8 shows, homicides involving rape are mostly females. The age
at which rape accounts for the largest fraction of female homicides is
around 12. That there are any rapes for infants, particularly that of males,
is probably an artifact of the way homicides are coded. In a multiple
victim homicide, some of the coded facts are those that characterize the
first victim listed. If a mother is killed in the course of a rape, and her
child is also killed, if the mother is the first listed victim the child will be
coded has have been killed during a rape, but the child itself may not
have been raped. 

As seen in Figure 9 the class of homicides that I call arguments and
disputes accounts for most homicide victims. The rate profile by age looks
very similar to the age profile of arrests, and perhaps we have here a broad
class of homicides that, like arrests, can be characterized by initiation
during the teen-age years followed by a steady desistance for the rest of
life.

Armed robbery, as shown in Figure 10, has a pattern similar to arguments
and disputes, but the rate peaks at a slightly later age, and accounts for the
largest fraction only in old age.

Getting killed by your spouse, ex-spouse, boy/girl friend, etc. is a risk you
face relatively late in life, as Figure 11, and later for men than for women.
Such homicides account for nearly half of women killed around age 40.

As Figure 12 shows, being a victim of a homicide committed during the
course of a non-gun robbery, or a burglary or theft, is an increasing risk in
the later years of life. One could hypothesize two different reasons for this
phenomenon: (1) older victims are more frail and thus less likely to
survive injuries received in the course of an attack, (2) older victims are
perceived by predators as being weaker, and thus older folks are more
likely to be targets of such attack.

All the homicides whose nature I couldn't figure out I call uncoded
homicides, instead of using a word like unknown which leads to
confusion.[3] Figure 13 shows the distribution of uncoded homicides.
People who code homicide files seem to know more about homicides of
children than homicide of older people. However, it may be the case that
when faced with the death of a child, if little can be determined about the
circumstances, the death is less likely to be recognized as a homicide.

NOTES

1. All charts in this report were drawn using data describing 100,784
victims of non-justifiable murder and non-negligent manslaughter in the
Supplemental Homicide Reports for the five years 1988 through 1992, and
estimates of the resident population obtained from the Census Bureau.

2. Nelsen, Candice and Lin Huff-Corzine (1998). "Strangers in the Night:
An Application of the Lifestyle-Routine Activities Approach to Elderly
Homicide Victimization." Homicide Studies, 2: 130-159.

3. Riedel, Marc (1998). "Counting Stranger Homicides: A Case Study of
Statistical Prestidigitation." Homicide Studies, 2:206-219.

"Appendix: Definition of the Homicide Types" is not included in this
ASCII plain-text file.

-------------------------------

Decreasing Violent Gun Crimes in New York City: A Result of Vigorous
Law Enforcement Efforts, Other Variables, or Both?

Steven F. Roth, New York State Division of Criminal Justice Services
4 Tower Place, Albany, New York 12203

For the last several years, reported violent crime has been decreasing
across the United States. New York City has been notable concerning this
trend. Since 1991, while the number of violent crimes reported within the
City has decreased significantly, violent crimes involving firearms,
especially handguns, have decreased by an even greater percentage. But
Why? Some point to the New York City Police Department's "Gun
Strategy," implemented to decrease the proliferation of guns and gun-
related crimes. Since the Strategy's initiation, reports of shooting incidents
and shooting victims have dramatically decreased. However, the Gun
Strategy was not introduced until March 1994, and the decreases in gun-
related violent crimes in NYC began decreasing following 1991,
decreasing to a greater extent during 1994 and thereafter. Thus, looking at
the statistics alone, it is unclear whether, and to what extent, the efforts of
the NYPD can be attributed to decreasing violent firearm offenses in
recent years. 

Decreasing Violent Crime in New York City

Decreasing Numbers of Reported Crimes

Starting in 1991, and more dramatically following 1993, reported violent
crime -- including homicide -- has been decreasing in New York City,
following a steady increase since 1985. Between 1991 and 1996 (the latest
year available for determining violent crimes involving firearms), while
the number of reported violent crimes (murder, rape, robbery and
aggravated assault) has decreased from 170,496 to 98,733 (- 42.1%),
violent crimes involving firearms, especially handguns, has decreased by
an even greater extent, from 53,621 to 19,495 (- 63.6%). Figure 1 shows
that the decrease in violent offenses involving handguns, as compared with
the decrease in violent offenses overall, has been even more dramatic after
1993 -- percentage change in reported violent crimes ('94: -11.1%, '95:
-16.4%, '96: -13.5%), versus percentage change in reported violent crimes
involving handguns ('94: -22.2%, '95: - 34.4%, '96: -21.2%). In addition, it
could be argued that the annual increases in overall violent crimes during
the late 1980s ('88: +9.8%, '89: +4.0% and '90: +3.0%) were being driven
by the larger increases in violent crimes involving handguns ('88: +17.3%,
'89: +11.6%, and '90: +14.2%).

Extent of Year-to-Year Decreases

The percentage decrease from the prior year in the number of violent
crimes involving handguns, as compared with violent crimes overall, has
been even more dramatic after 1993. The annual percentage change in
reported violent crimes overall were as follows -- 1994: - 1.1%, 1995: -
16.4%, and 1996: - 3.5%. This compares to the percentage annual change
in reported violent crimes involving handguns -- 1994: - 22.2%, 1995: -
34.4%, and 1996: - 21.2%. The annual percentage decreases in these two
categories of reported crimes are shown in Figure 2.

New York City Police Department Gun Strategy

Many have attributed the New York City Police Department's "Police
Strategy No.1: Getting Guns off the Streets of New York" (hereafter "Gun
Strategy"), initiated in March 1994, as the "cause" of the decrease in
violent firearm crimes, as well as decreases in arrests for illegal possession
of a loaded handgun (which is also classified as a Violent Felony Offense,
but not included in the aforementioned "violent crimes"). In most basic
terms, the Gun Strategy has involved targeting and expanding
investigative and patrol services, revising and expanding training,
obtaining and deploying technologically advanced equipment, and
securing public support to reduce crimes involving firearms.

Targeting and expanding investigative resources involves efforts to: 1)
identify and pursue all accomplices involved in the commission of violent
crimes; 2) identify and pursue gun traffickers inside and outside New
York whenever a gun is used in the City; 3) utilize detectives in the
precincts to strengthen all gun arrests made by uniformed and plainclothes
police officers; and 4) assign case responsibility at the precinct level to
encourage results-oriented police and investigative work. Refocusing
patrol resources concerns targeting and expanding the Street Crime Unit
of the Patrol Services Bureau to attack specific areas of high gun violence.
Training revision and expansion was developed to teach detectives case
development of gun investigations, apprehension of illegal firearm
traffickers, confidential-informant registration, search warrant procedures,
and to teach police officers to better recognize the signs of weapons
possession and to make stronger cases against people they arrest.
Technologically advanced equipment is obtained and deployed to match
guns used in more than one outstanding crime, and to trace firearms to
illegal sellers throughout the country. Public support is secured to identify
and help locate those who possess, use or sell illegal guns in their
neighborhoods.[1]

Particular means by which illegal guns are seized include a "radio run" --
usually turned in by a citizen as "a man with a gun." Another method is a
search warrant (more common for narcotics than firearms in buy and bust
operations lasting months to garner evidence for a search warrant) when
illegal guns are discovered in the search for illegal drugs. Other means by
which the police seize illegal firearms are calls to crimes in progress in
which a perpetrator is fleeing a crime scene with a gun. Another effective
means by which the police have seized guns is apprehending subway turn-
style jumpers who are caught with illicit firearms. When an arrest is made,
the officer has the constitutional right to search the arrestee for the
weapons in order to ensure the officer's person protection, a process by
which illegal guns are sometimes found and seized.[2]

On a limited basis, a "model block" program has been conducted where
barriers are placed at each end of streets in high crime areas -- with police
stopping people based on reasonable suspicion and asking about their
business, and to see identification. If a positive i.d. is not produced, the
officer has the right to detain, arrest and search the person for the officer's
protection -- which may produce an illegal firearm. Another specific
example is if a schoolyard closes at certain time, people have no
legitimate reason to be there after hours. Someone being present under
these circumstances, by itself, is an acceptable reason for an officer to stop
the individual and ask questions. In addition, the "bulge in the pocket," or
reasonable suspicion of carrying a gun determined by examining walking
style or other gestures ( as developed by an NYPD detective with
hundreds of illegal gun arrests) has produced numerous other illegal
firearm confiscations.[3]

A vital aspect of the effort to get guns off the streets was the February
1997 expansion of the NYPD's Street Crime Unit by 300 officers, and its
reorganization into citywide and borough units to combat violent street
crime and firearm offenses.[4] City areas are targeted based on pattern and
trend analyses. A computerized repeat calls-for-service application allows
identification and tracking of high volume crime locations that need
increased enforcement. Precinct commanders, given more direct decision-
making by the COMPSTAT Program, may also use MAPINFO crime
mapping techniques to deploy resources most effectively as they see fit.[5]
Detectives debrief all prisoners charged with unlawful possession of a
firearm in order to identify the weapon's source, and all firearms are
subject to gun tracing through NYPD Ballistics in cooperation with the
Federal Bureau of Alcohol, Tobacco and Firearms (ATF).

The NYPD continues its enhanced efforts to identify individuals who
illegally possess firearms, to track the sources of illegal guns trafficked
into the City with the assistance of ATF in Project Lead and contact with
law enforcement officers across the country, and use advanced technology
comparison systems in its Ballistics Unit to investigate firearms, bullets
and bullet casings involved in crimes. In addition, the NYPD has
participated with 16 major municipal police agencies across the country in
the Federal Youth Crime Gun Interdiction Initiative, a pilot program
developed by ATF to trace guns used by juveniles in the commission of
crimes to the location and dealer from which they were purchased. The
project is using ATF's National Tracing Center to produce information
about the sources of weapons, aggregate and analyze trace data, and share
information with state and local law enforcement officials. The objective
is to curtail the supply of illegal firearms used by youths to commit crime,
and to encourage arrests where appropriate.[6] In addition, another factor
possibly associated with decreasing homicides in the City is the overall
increase in NYPD full-time sworn personnel from 32,855 in 1990 to
37,090 in 1996.[7]

Hopefully, as a result of these and other criminal justice efforts, slightly
over 2,000 people were shot in New York City in 1997, as opposed to
over 5,000 people shot in the City in 1993. Based on firearm arrests, the
strong appearance is that fewer people in New York City are carrying
guns on their persons -- resulting in less opportunity to commit crimes
with firearms, including those that are not pre-meditated. 

New York City Mayor's Management Report

In March 1994, the NYPD introduced its Gun Strategy to reduce the
proliferation of guns and gun-related crimes in the City. The NYC
Mayor's Management Report: Preliminary Fiscal 1998 indicates that
shooting incidents in NYC have declined 62.5%, from 5,269 in calendar
1993 to 1,977 in calendar 1997. The number of shooting victims also
declined 61.4 % during this same time period, from 5,862 in calendar
1993 to 2,262 in calendar 1997. This trend continued during the first four
months of Fiscal '98, with a 15.3% drop in shooting victims -- to 834 from
985 during the same period of Fiscal '97. Since implementation of the Gun
Strategy, arrests for gun possession have declined by 43 %. From the first
four months of Fiscal '97 to the same period of Fiscal '98, arrests with gun
possession as the highest charge fell almost 18%, from 1,440 to 1,185.[8]
But what empirical research is there that might shed light on whether the
NYPD's recent strategy to reduce gun crime has indeed done so?

Study on Declining Homicide in New York City: Association With Police
Practices

Researchers Franklin Zimring, Jeffrey Fagan, and June Kim, in their draft
report "Declining Homicide in New York City: A Tale of Two Trends"
(which will appear in the Summer 1998 issue of The Journal of Criminal
Law and Criminology), examined whether NYPD's Gun Strategy is
associated with recent and dramatic declines in firearm homicides in the
City. With the authors' permission, the remainder of this paper uses graphs
and statements as contained in the draft report. As the authors state, "The
recent decline in homicides in New York City is an exception to the usual
pattern, the most celebrated example of crime-news-as-good-news in
decades." The drop in homicides was both large and abrupt -- the
homicide rate in the Nation's largest city fell 52 percent in five years. The
authors further state, " If this drop can be plausibly tied to enforcement
activities, it would be the most conspicuous success of city police
deployment policies in the 20th century." 

In an effort to empirically assess whether the homicide decline between
1993 and 1996 is significantly associated with police practices, the
researchers put the decline in a variety of statistical contexts, comparing
the drop to previous New York experience and to the experiences of other
cities in the United States. They also examined changes in homicide
patterns during the decline in search of clues about causes. They viewed
available police and other data about the scope of the decline in crime and
violence over the years when homicide dropped -- researching such
questions as: "Did many crime categories fall, and by how much?" and
"Was the decline concentrated in a few categories or spread evenly across
the spectrum of felony crime?"

Size of the Homicide Decline

Magnitude of Homicide Decline

In its relative and absolute magnitude, the homicide drops after 1992 were
by far the largest in the post-World War II history of NYC. The number
of lives involved is even more impressive, with more than 1,100 fewer
homicides in 1996 than in 1992. This reduction in homicide far exceeded
the total number of homicides the City experienced each year in the 1950s
and early 1960s.

New York City's homicide decline was also compared with homicide
declines experienced in other major U.S. cities for five-year periods from
1950-1996. (See Figure 3.) The five-year records for big cities vary from
a 61 percent decline in Pittsburgh through a 15 percent decline in
Chicago. The median decline was 40 percent, and seven cities reported
highest decline percentages of between 43 percent and 38 percent. The
New York decline is the third highest for major cities, behind Pittsburgh,
equal to Houston, and about 25 greater than the cluster of city records
around 40 percent. On the straight arithmetic, the New York experience is
not unprecedented, but is a higher percentage drop than 12 of the Nation's
15 biggest cities have experienced in a five-year span. Of the very largest
U.S. cities, NYC had the greatest percentage decline in reported homicide. 

Drop in Accordance With Previous Abnormally High Rate? 

The researchers tested the five cities with the highest five-year declines
against the possibility that an abnormally high homicide rate for a short
period of time is the reason for a large drop in the homicide rate --
negating the existence of a unique "trend". The low year in the largest
five-year decline is compared to the mean rate in that city for the prior 15
years. Three of the five largest declines (including NYC) represent a
substantial reduction from the average rate for the previous 15 years.
Zimring et al. concluded that because of the steep decline in NYC
homicides after 1993 (preceded by a temporary rise), it is difficult to argue
that the drops in recorded homicides were merely part of a cyclical
"roller-coaster" downward trend following abnormally high rates. 

In a time when very wide swings in criminal homicide rates occur, sharp
downward movements in the homicide rate might be expected after sharp
upward movements, akin to a roller-coaster that drops faster after
climbing to a great height. To use the roller-coaster metaphor, the
homicide climb in New York City was not as steep before the post-1991
drop, so that less of the decline appears likely to be merely cyclical.
However, as late as the end of 1995, New York's homicide rate was within
10 percent of its 1985 level. 

The rate of homicide broke important "new ground" only after 1995. In
sum, the percentage decline experienced by New York City in five years
is quite large, but in no means unprecedented in major cities.
Nevertheless, such a homicide decline is also rather uncommon. Thus, the
decline is large by historical standards and not merely an "echo" of a sharp
but temporary previous rise. This research dealt with the size of the
homicide decline.

Nature of the Homicide Decline

The homicide statistics discussed above involve an aggregation of many
different types of law enforcement, community areas and population
groups within New York City. The researchers further analyzed the types
of homicides that changed the most in order to detect probable causes of
the changes. Specifically, they performed cross-tabulations of NYC
homicide over time by borough, by weapon and by demographic groups in
the City, which were separately analyzed. 

Variations by Individual Borough 

The 1985-95 homicide patterns for the City's four largest boroughs were
examined to determine whether individual boroughs disproportionately
affected the City's steep decline in reported homicides starting in 1994.
The boroughs differ importantly in the patterns observed while homicide
rates grew in NYC, but are much less divergent during the years of
decline -- all running parallel (see Figure 5), discounting disproportionate
influence by individual boroughs.

Variation by Weapon

The significant break of homicides by weapon is between gun shot deaths
and all others. As shown in Figure 6, the patterns for gun and non-gun
homicides differing sharply between 1985 and 1996. Non-gun killings
drift steadily downward during this period (interrupted only by the
Happyland Social Club fire in 1990 -- 86 of the 87 killings were deleted
for this one incident so as not to skew the data). However, gun deaths
double between 1985 and 1991, and fall sharply from 1993. This reveals
that the decreases in homicides involving firearms has been driving the
reduction in overall homicide in the City since 1991, and the sharper
decreases in overall homicides since 1993. In addition, increases in
firearm homicides between 1985 and 1991 were also responsible for the
rise in overall homicides in this time period.

Demographics

The distinct patterns of gun and non-gun homicide raise questions about
whether these patterns were similarly distributed among the population of
homicide victims. Using Medical Examiner records, the researchers
disaggregated characteristics of gun and non-gun homicides by sex, age
and race to test the strength of association of each variable with the
decreasing homicide rates. 

Sex

Overall homicide rates for women were lower than for men from 1985-
1995, consistent with homicide trends historically. However, the trends in
gun and non-gun homicides for males and females differed in the
magnitude, timing and duration of change over the period. Figure 7 shows
that gun homicides by women increases by 68 percent during this period,
compared to an increase of over 105 percent for males. There were steady
declines in non-gun homicides for both men and women, but the decline
was greater for men. By 1995, non-gun homicide rates for males were
45% below their 1985 rates. However, the rates for women declined far
less. Actually, the overall trend for women is no change since 1992. 

Age

A great deal of public and scholarly attention on violence in the past
decade has focused on the increase in gun homicides by adolescents.
Trends nationwide indicate that gun homicide rates for adolescents
increased during this period while gun homicide rates for persons over 25
years of age were declining. Figure 8 shows that while adolescent
participation in gun homicide did rise sharply from 1985-91, rates for
other age groups also continued to rise during this period. Gun homicides
by adolescents ages 15-19 rose more quickly and sharply over this period.
By 1991, gun homicide rates for this group reached a peak of 176 percent
their 1985 rate. The other population groups also increased, peaking at
approximately the same time at about 100 percent above their 1985 rate.
Rates for the oldest population group, those 35 years of age or more, rose
slightly more slowly, peaking at 77 percent of their 1985 rate.
Accordingly, adolescents did not replace, but became an active part of a
homicide epidemic that spanned age groups.

Race

None of the data sources allowed detailed disaggregation of the homicide
trends by race over the entire 1985-95 period. Detailed data were available
only for African Americans; whites and Hispanics were not distinguished
in the police or Medical Examiner data until after 1990. Based on the
available data, there were no substantial changes in the racial composition
of NYC during this period. 

Concomitants

In addition to the aforementioned citywide homicide data, trends in crime
statistics and other social data widely believed to be associated to trends in
lethal violence should be examined for their possible influence on
decreasing homicide rates. The historical data for gun and non-gun
violence is presented separately below because of the finding that gun and
non-gun homicides have distinctly different histories in the City. 

Homicide and Assault Data

Figure 9 shows the pattern over time for firearm homicide and firearms
assault, with 1985 rates of shooting deaths and firearm assaults expressed
as 100 and all other values normed to that scale. In both timing and
relative magnitude, the firearm assault trends match the firearm homicide
trends. The increased level of firearm assaults, about 8,600 more known
cases by 1990 and 1991, appears to be the mechanism driving the increase
in gun fatalities until 1991. The sharp drop in firearm assaults after 1993
also seems to be the dynamic for the drop in firearm fatalities. 

The police data on non-firearm aggravated assault trends and NYC Health
Department records on homicides by all means other than firearms
are quite different from the firearm trends for assaults. The decline in
non-gun killings that starts in the mid-1980s is merely a shift from knife
and personal force attacks to gun attacks. It is widely believed that gun
incidents increased largely due to turf conflict in the crack trade. 

Gun and Non-Gun Robbery

The pattern of firearm robbery from 1985-96 is similar to the patterns of
firearm assault and homicide, peaking in 1991 before declining sharply.
The general, steady decline in non-firearm assault and homicide is
mirrored by the trend for property crimes and non-firearm robberies.

Population of Highest Risk Subgroups

Figures 12 and 13 represent the trends for firearm and nonfirearm
homicides arrayed with population estimates for the subgroups with the
highest homicide risk: males 15-29 years of age. Demographic trends
appear to be unrelated to firearm homicides. However, as shown in Figure
13, non-firearm trends appear to coincide with population declines among
white and black males.

Law and Policy

Significant changes occurred in law and criminal justice policy from
1985-96 which produced important changes in the likelihood of detection
and punishment of crime (with special emphasis on NYPD's gun strategy),
and included a rapidly expanding prison population.

Figure 14 shows the changes in policing over the 11-year period were
unrelated to the long-term decline in non-gun homicides. However, there
are some apparent links between police resources and firearm homicides,
as well as police strategy and homicide trends. Patrol strength increased
from its 1991 coundt of 6,647 officers to over 8,305 officers in 1995 -- an
increase of almost 25 percent. The increase in patrol strength was sharpest
from 1991 to 1994, the period of onset of the decline in firearm homicides.
Also, marked shifts in policing strategy began in 1994, concurrently with
even sharper declines in firearm homicides from 1994-96.

While the increased and aggressive patrol tactics should reduce
opportunities for visible or outdoor crime commission, the researchers
found no evidence of displacement from outdoor to indoor homicide in
these trends (Figure 15).

Increase in Patrol Strength

The authors found that the increase in patrol strength beginning in 1991
had a positive effect on reducing visible homicides. The second pressure
was tied to changes in the dynamics of drug distribution. The decline in
visible homicide was concurrent with the shift from outdoor to indoor
drug selling, reducing the volatility of drug transactions and the
opportunities for conflicts over money or turf. The dramatic 1994-96
decline in firearm homicides (greatly influencing overall homicides)
coincides with changes in NYPD strategy. Combined with earlier
downward pressures, the shift toward an aggressive enforcement strategy
targeted at firearms is believed to be significantly associated with sharp
declines in both indoor and outdoor homicides.

Incapacitation

Incapacitating potential homicide offenders through changes in
incarceration did not show an identifiable impact on homicide trends. The
NYC jail and State prison populations of NYC offenders rose in lockstep
with NYC homicide through 1991, and continued to rise even as firearm
homicides, and homicides overall, began their rapid decline after 1993.
(See Figure 16.)

Other Factors

Other factors besides the gun strategy likely have contributed to the
decline in gun homicides. It is difficult to exactly apportion credit between
demography, policing including NYPD's Operation FLARE which has
helped reduce the number of federally licensed firearm dealers in NYC
through police interviews of license applicants and enforcement of City
fire and zoning codes applying to retail business in firearms, and other
strategies being implemented by the NYPD including quality-of-life
offense enforcement. Changes in State and Federal gun laws and
enforcement efforts also may be influential. Cyclical variation as well as
social trends in risk and exposure may also be relevant -- such as people
avoiding areas of higher crime incidence. While the researchers concluded
that NYPD's firearm law enforcement efforts appear to be strongly
associated with decreases in NYC gun homicides in recent years, the
question remains: how long will this "trend" of declining firearm
homicides continue?

Conclusion

The authors concluded that the effects of both manpower and tactical
changes in policing are "no doubt" important and strong contributors to
the homicide decline. However, the size and timing of the decline in
firearm homicides are so great as to require additional, contemporaneous
explanations and effects that interact with policing to produce
unprecedented change, including the steady decline in non-gun homicides.

NOTE

1. Police Strategy No. 1: Getting Guns off the Streets of New York, New
York City Police Department, (March 7, 1994), pp.8-9.

2. Interview with Joseph Lovelock, Deputy Inspector. Office of
Management Analysis and Planning, New York City Police Department,
(March 9, 1998).

3. Ibid.

4. The Mayor's Management Report, Preliminary Fiscal 1998. Vol. I,
Agency Narratives, City of New York, p.13.

5. The Mayor's Management Report, Fiscal 1997 City of New York, p.11.

6. The Youth Crime Gun Interdiction Initiative: The Illegal Youth
Firearms Markets in 17 Communities, Department of the Treasury, Bureau
of Alcohol, Tobacco and Firearms (July 1997).

7. Crime and Justice Annual Report, 1990-1996, New York State Division
of Criminal Justice Services.

8. Mayor's Management Report, Preliminary Fiscal 1998, City of New
York, p. 13.