CONFERENCE VERSION
SOCIAL CONTAGION OF VIOLENCEa
Jeffrey Faganb
Deanna Wilkinsonc
Garth Daviesd
a. This research was supported by grants from the Centers for Disease Control, the National
Institute of Justice, and the Robert Wood Johnson Foundation, in addition to support from the
Sloan Working Group on Social Contagion of Youth Violence. All opinions and errors are mine
alone. Data on homicide characteristics and locations were generously provided by the Injury
Prevention Program, New York City Department of Health. Excellent research assistance was
provided by Jay Galluzzo, Garth Davies, Deanna Wilkinson, Tamara Dumanovsky, Marlene
Pantin and Carolyn Pinedo. Thanks also to interviewers Richard McClain, David Tufino, Davon
Battee, and Whetsel Wade, for creating a unique and very rich source of data on streetcorner life in
New York City.
b. Professor, School of Public Health, Columbia University, and Visiting Professor, Columbia Law
School.
c. Assistant Professor, Department of Criminal Justice, Temple University.
d. Assistant Professor, Department of Criminology, Simon Fraser University.
Presented at the
Urban Seminar Series on Children's Health and Safety
John F. Kennedy School of Government
Harvard University
May 11, 2000
SOCIAL CONTAGION OF VIOLENCE
1
Like many large American cities, New York City experienced a sudden and
dramatic increase in homicides beginning in 1985. The homicide runup was highest
for adolescents, but rates increased quickly for older persons as well (Fagan, Zimring
and Kim, 1998). Unlike other cities, however, where homicides declined gradually
and have yet to recede to their pre-1985 levels, the increase in New York was
followed by an even larger decline over the next five years. By 1995, homicide in
New York City had returned to its 1985 level; by 1996, it was 15% lower than the
previous low in 1985; by 1997, homicide was about 25% lower than its 1985 level.
Both in New York and nationally, this decline was unprecedented not just in the
present era, but represents a decline of greater magnitude than any observed in large
American cities since the 1950s.
Explanations of this rollercoaster pattern have tended to partition the periods
of increase and decline as distinct phenomena with unique causes. Moreover, these
causes are typically regarded as exogenous to the people or areas affected. For
example, the onset and severity of the homicide trend was attributed to the sudden
emergence of unstable street-level crack markets, with high levels of violence
between sellers (Fagan and Chin, 1990; Baumer, et al., 1996). Others suggested that
drug markets created a demand for guns that in turn trickled down from drug sellers
into the hands of adolescents (Fagan, 1992; Blumstein, 1995). Structural theorists
implicated race-specific economic deficits in inner cities (Peterson and Krivo, 1996)
or racial residential segregation (Massey, 1995). There have been many claims
regarding the sources of the decline, including changes in police strategy (Kelling and
Cole, 1996), demographic changes (Eckberg, 1995; Cook and Laub, 1998),
incarceration (Blumstein and Beck, 1999), and lower demand for illegal drugs
(Curtis, in press).
None of the explanations of either the increase or the decline are fully
satisfying. Moreover, the gap between the scale of demographic and policy changes
and the scale of the crime decline suggests that there are processes at work other than
SOCIAL CONTAGION OF VIOLENCE
2
these usual suspects. Some have used the term epidemic metaphorically to describe
the homicide runup and decline, but with little precision and often conflating several
features of epidemics.1
Epidemic is a term widely used in the popular and scientific literature to
describe two quite separate components of a phenomenon: an elevated incidence of
the phenomena, and its rapid spread via a contagious process within a population in
a short period of time.2 Epidemics need not be contagious. Consider an outbreak of
food poisoning from contaminated materials, or a cancer cluster nearby a polluted
water supply. These medical problems may occur at a rate well above an expected
base rate, but are not spread from person to person through physical contact or an
infectious process. In contrast, an outbreak of influenza, the adaptation of cultural
fads, medical or industrial innovation, or changes in the rates of antisocial behavior,
all reflect spread through interpersonal exposure to an “infectious” agent. While
disease spreads through a host and agent,3 social contagion involves the mutual
influence of individuals within social networks who turn to each other for cues and
behavioral tools that reflect the contingencies of specific situations (Burt, 1987;
Bailey, 1967; Coleman et al., 1966).
1. Others use the term more literally, but usually to conflate several different processes: social
concentration, spatial diffusion, and temporal spikes. See, for example, Norman T. Bailey, The
Mathematical Theory of Infectious Diseases and Applications (1976).
2. For example, Gladwell (1996: 33-4) describes how the incidence of an ordinary and stable
phenomenon such as a seasonal flu can become epidemic when its incidence increases in a very
short time from a predictable base rate to an elevated rate of infections.
3. See, for example, Kenneth Rothman, Modern Epidemiology (1986); Leon Robertson, Injury
Epidemiology (1990).
SOCIAL CONTAGION OF VIOLENCE
3
The contagious dimension is especially salient during the upswing of an
epidemic, when physical or social contact is critical to spread pursuant to exposure.
Epidemics end, also, as the rate of new incidence of the phenomenon declines. This
may occur because the density of contacts may decline, or because some form of
resistance develops that reduces the odds of transmission from one person to the next,
even in the presence of exogenous contributing factors (Burt, 1992; Bailey, 1967).
In this paper, we assess whether the rollercoaster pattern of homicides in New
York City from 1985-96 fits a contagious model, and identify mechanisms of social
contagion that predict its spread across social and physical space. This framework
for interpreting the homicide trends as an epidemic includes two perspectives. First,
the sharp rise and fall is indicative of a non-linear pattern where phenomenon are
spreading at a rate far beyond what would be predicted by exposure to some external
factor. And, the phenomenon declines in a similar pattern where the reduction from
year to year exceed what might be expected by linear regression trends. This leads
to the second perspective: the factors leading to its spread are not exogenous factors,
as in the case of contamination or disaster. Instead, the non-linear increase and
decline suggest that the phenomenon is endemic to the people and places where its
occurrence is highest. And, this behavior may be effectively passed from one person
to another through some process of contact or interaction.
We assess the validity of these assumptions in three ways. In Part I, we use
Vital Statistics data from the New York City Department of Health to present simple
time series data that describe the increase and decline in homicides. We concentrate
on homicides involving adolescents and young adults, populations who experienced
the sharpest rise and decline in homicide, both in New York and nationally (Fagan
et al., 1998; Cook and Laub, 1998). We supplement these data with police records
of complaints and arrests to raise and evaluate competing explanations of youth
homicide. In Part II, we examine the spatial and social trends in youth homicide by
disaggregating the data by neighborhoods over the 11 year period, and fitting models
SOCIAL CONTAGION OF VIOLENCE
4
to demonstrate the spatial diffusion of youth homicide in specific areas of the city.
By covarying neighborhood social and economic characteristics with temporal
homicide trends, we are able to show that the diffusion of homicide in this era was
specific to the most socially isolated areas of the City. In this section, we isolate gun
homicides as the contagious agent, showing that it is gun homicides that diffused
across New York City neighborhoods, and gun homicides that retreated just as
quickly. In Part III, we present data from interviews with young males active in gun
violence during this time. Their reports of the role of guns in violent events further
specify how diffusion may in fact be the results of a dynamic process of social
contagion. We conclude by integrating these perspectives into a framework for
theory, research and policy development.
I. THE EPIDEMIOLOGY OF YOUTH HOMICIDE
IN NEW YORK CITY
A. Historical and Current Homicide Trends
The epidemic of youth violence in New York City from 1985-95 is best
understood in a social and historical context that spans nearly 35 years. Like the
nation’s largest cities, New York experienced a sharp increase in homicide and other
violence rates beginning in the mid-1960's. The homicide rate rose from 4.7 per
100,000 population in 1960 to 31.0 in 1995. By 1996, the rate had receded to 13.9
per 100,000, a level unseen since 1968. Figure 1 shows the homicide counts for
1968-96, disaggregated by gun and non-gun methods.
The 1960 rates were typical of homicide rates in New York City for the first
half of this century. With the exception of the decade influenced by the Volstead Act
and the Great Depression, homicide rates in New York City varied between 3.8 and
SOCIAL CONTAGION OF VIOLENCE
5
5.8 per 100,000 population.4 From 1965 to 1970, the average annual homicide rate
rose from 7.6 to 12.6, and rose again to 21.7 by 1975. Thus, homicide counts and
rates in New York nearly tripled within a decade. The rates have remained for the
past 35 years at the plateau first reached in 1968. Accordingly, Figure 1 suggests that
homicides in New York have been normalized at an elevated rate, and has become
characteristic of the city’s social landscape. Thus, the escalation in killings is
cumulative, with each new peak building on the elevation of the base rate established
in the previous peak. One interpretation of the recent decline may simply be the
recession of this longer-term social and historical trend.
Figure 1 also shows that this long-term trend involves three sub-epidemics.
The first of these peaked in 1972, the second in 1981, and the third in 1991. Each
coincided temporally with drug epidemics and the growth of retail drug markets:
heroin in the early 1970's, powder cocaine in the late 1970's, and crack beginning in
1985 (Johnson et al., 1990). Moreover, successive epidemics were cumulative in
their trends, not distinct. To use a term introduced in Part I, the pattern of killings in
particular does rather resemble the shape of a rollercoaster, with an ascent through
the late 1970s to a relatively low peak, a return to near the previous low point, a sharp
increase to a high peak in 1990 and a precipitous drop thereafter.
Finally, Figure 1 also shows the growing importance of guns in homicides in
each of the three peaks. Increases in both gun and non-gun homicides contributed
to the tripling of homicide rates through 1972. In 1972, the ratio of gun to non-gun
homicides was 1.23. By the next peak in 1981, the 1,187 gun deaths were nearly
1.76 times greater than the 673 non-gun homicides. In 1991, the modern peak, the
1,644 gun homicides were 3.16 times greater than the 519 non-gun homicides.
In addition to sharp increases the number of gun homicides, the gun:non-gun
4. In its 1996 report, the Office of Vital Statistics and Epidemiology reports homicide rates prior
to 1985 in five year intervals. Homicide rates rose from an average of 4.9 in 1916-1920 to 7.6 in
1931-1935, and declined to 4.5 by 1936-1940 (NYCDOH, 1997).
SOCIAL CONTAGION OF VIOLENCE
6
ratio also rose sharply because of a long-term decline in the number of non-gun
homicides. Since 1980, the number and rate of non-gun homicides has declined by
nearly 50 percent, from 735 to 335 non-gun killings in 1996. There are thus two
dynamic and different pattern in the data on homicide by weapon. Gun killings
follow the rollercoaster pattern of steadily increasing peaks beginning in 1972.
Nongun killings trend down from 1980, to rates unseen since 1960. This long-term
secular trend in nongun killings is substantial, but it has not previously been noticed.
Figure 1 also shows that the recent 1985-96 cycle was qualitatively different
from the preceding peaks in four important ways: (1) its starting point was lower than
the starting point for previous (1981) peak, (2) its peak was about 15% higher than
the preceding peak, (3) it had a far greater share of gun killings, and (4) its decline
was far steeper than any previous decline (Fagan et al., 1998). To illustrate the extent
of the differences between the 1985-96 cycle and its predecessors, Figure 2 presents
the data from Figure 1 normed to the 1985 base.5
Gun killings accounted for all of the increase in homicides in this period, and
most of the decline. Non-gun killings declined steadily since 1986, and by 1996 were
about half the 1985 rate. Gun homicides doubled from 1985 to 1991, and then
declined by 1996 to about 70% of the 1985 starting point. From the peak in 1991,
gun homicides declined by over 60 percent. While the declines after 1992 in nongun
killings are a continuation of the eight years of previous decreases, the increase and
decline in gun killings is evidence of a homicide spike that is unique from its
predecessors.
B. The Social Structure of Homicide
5. The nongun total for 1990 has been adjusted by deleting 89 of the 90 killings from the
Happyland Social Club fire, in effect counting that episode as one homicide. This is done to
smooth out the long-term trend curve.
SOCIAL CONTAGION OF VIOLENCE
7
The demographic patterns of gun and nongun homicide victimization during
this period tell two interesting stories, and two predictable stories. First, the
homicide trends for women differ from the patterns for men. Second, changes in
adolescent homicide rates were accompanied by parallel but less dramatic changes
among older populations. This trend varies from the national picture of steadily
declining rates among older groups. Third, as we saw before, the homicide runup
and decline were concentrated in gun killings. Fourth, the homicide epidemic was
concentrated among non-whites. We observed these trends for both homicide
victims and offenders.
1. Sex
Nearly all the increase and decline in killings from 1985-95 were gun
homicides of males.6 Table 1 shows that the rate of gun killings among males
doubled from 21.8 per 100,000 in 1985 to 44.5 in 1991. Non-gun killings of males
declined steadily throughout this period, and by 1995 were less than half the 1985
rate. Killings of males were increasingly gun events: the ratio of gun to non-gun
homicide victimizations of males increased from about 1.5:1 in 1985 to 3.23:1 in
1995.
Table 1 shows that the temporal patterns were similar for females. The rate
of gun homicides of females peaked in 1991, the same year as males, and sustained
their peak rate for approximately three years before dropping sharply in 1994. By
1995, gun homicides for females had dropped 5% below their 1985 levels. Non-gun
homicides of women declined steadily throughout this period, from 4.7 per 100,000
in 1985 to 3.8 per 100,000 in 1995.
But unlike males, the rates of non-gun
homicides of females were higher than the rates of gun homicides. Throughout the
6. The 1990 spike for male nongun homicides most likely reflects the 89 arson homicide deaths in
the Happyland Social Club fire. We could not adjust the age-, race-, or gender-specific rates for
these homicide deaths since data were not available on their characteristics.
SOCIAL CONTAGION OF VIOLENCE
8
period, the changes in rates for females were quite small, and not far from the
expected rates historically. The same is true of male non-gun killings. Accordingly,
this epidemic is confined to gun killings among males.
2. Race
Nationally, virtually all increases in homicide rates from 1985 to 1990 among
people 10-34 years of age were due to deaths of African American males. Most of
these were firearm fatalities that were overwhelmingly concentrated demographically
and spatially among African American males in urban areas (Fingerhut, Ingram, and
Feldman, 1992a; 1992b.) The trends in New York mirror these national trends.
Unfortunately, none of the data sources permitted detailed disaggregation of the
homicide trends by ethnicity over the entire 1985-1995 period. Detailed data were
available only for African Americans; whites and Hispanics were not distinguished
in the police or Vital Statistics data until after 1990. Our analysis is limited to
comparisons between whites and non-whites; non-whites are primarily persons of
African descent, including some Hispanics.
Table 2 compares gun and non-gun homicides by race for each year. The
within-race ratio of gun to non-gun homicide rates illustrates the concentration of the
homicide epidemic in gun homicides among non-whites. For whites, the ratio rises
from 1.26:1 in 1985 to a peak of 3.23:1 in 1992, before receding to 2.1:1 in 1995.
For non-whites, the ratio rises from 1.20:1 in 1985 to a peak of 4.05:1 in 1992, and
recedes to 2.32:1 in 1995.
The narrow difference between whites and non-whites in Table 2 is due to the
inclusion of Hispanics among the whites in the population and homicide counts. The
extent of this bias can be seen in 1993 data from the New York City Department of
Health injury surveillance system. The mortality and morbidity rates of gunshot
wounds for Hispanics is 228 per 100,000 persons, compared to 302 for African
SOCIAL CONTAGION OF VIOLENCE
9
Americans and 60 for Whites in that period.7
3. Age
Much of public and scholarly attention on violence in the past decade has
focused on the increase in gun homicides by adolescents (Blumstein, 1995; Cook and
Laub, 1998). Trends nationwide show that gun homicide rates for adolescents
increased during this period while gun homicide rates for persons over 25 years of
age were declining. In New York City, homicides were not confined to younger age
groups, but were a serious problem across a wide age range from 15 to 34 years of
age.
Table 3 shows that gun homicide rates were higher than non-gun rates for all
age groups. For each year, gun homicide rates were highest for persons 20-24 years
in all years. Gun homicides by adolescents ages 15-19 rose more sharply over this
period than other older population groups. Nevertheless, Table 2 shows that while
adolescent participation in gun homicide did rise sharply from 1985-1991, rates for
other age groups also continued to rise during this period.
Gun homicide rates declined sharply for all three age groups from 1992-95,
to about 50% of their peak rates in 1991, and were about the same as their 1985 rates.
Although post-1991 decline was precipitous for adolescents 15-19 years of age, their
1995 gun homicide rates remained exceeded 25% above their 1985 base rate.
Table 3 also shows that nongun homicide rates declined steadily for all age groups,
and by 1995 were 50% or more lower than their 1995 rates.8 Similar to gun homicide
rates, the non-gun homicide rates were highest for persons ages 20-24 years.
4. Victim-Offender Homogeneity
7. See, for example, New York City Department of Health, Injury Mortality in New York City,
1980-90 (1992); New York State Department of Health. Injury Facts for New York State (1994).
8. New York City Department of Health, id. at __.
SOCIAL CONTAGION OF VIOLENCE
10
Most homicides are within-group events, especially with respect to gender,
race and ethnicity (Sampson and Lauritsen, 1994; Cook and Laub, 1998).
(Exceptions include domestic homicides, and homicides that follow rape.) For
homicides in New York City from 1985-95, the aggregate age, gender and race
characteristics of homicide victims also applied to homicide offenders. At the case
level, we observed within-group homogeneity with respect to our limited categories
of race.
However, despite aggregate age similarities, the age composition of homicide
events may differ for gun and non-gun killings, and these differences may have
widened during the most recent homicide cycle. Age stratification of peer groups has
traditionally created age-specific social networks. These rigid age boundaries offered
few opportunities for cross-age social interactions among delinquent groups. For
example, age grading is a hallmark of street gangs (Klein, 1995) and adolescent
cliques (Schwendinger and Schwendinger, 1985). Recent changes in the social
contexts of inner cities, where homicides were concentrated throughout this period,
may contribute to a breakdown of traditional age-grading. Factors such as the
emergence of street drug markets and dense streetcorner groups of males not in the
workforce contribute to a mixing of the ages on the street. Among adolescents and
young adults, competition for street status through violence contributes to a process
of “status forcing” that promote cross-age interactions (Wilkinson, 1998).
Using data from police reports, we analyzed the within-age distribution of
homicide events for each year in the recent homicide cycle. Table 4 reports the
percent of homicide offenders whose victims are within their own age group. In
1985, at the outset of the latest homicide runup, age homogeneity for gun and nongun homicides was comparable, but very low. For adolescents and young adults,
about one gun homicide in four involved persons within the same age categories. For
adults ages 25-34, about four in ten gun homicides were within-age killings. Once
the gun homicide runup began in 1986, age homogeneity rose as homicide rates rose
SOCIAL CONTAGION OF VIOLENCE
11
generally. For example, within-age gun homicides for young adults rose from 22.9
percent of gun homicides in 1985 to 35.8 percent in 1990; for offenders ages 25-34
years, within age group homicides rose from 38.5 percent in 1985 to 54.9 percent in
1989. Even with these increases, however, the majority of gun killings involved
persons from different age groups. During the same period, within-age non-gun
homicide rates varied from year to year in an inconsistent pattern.
Comparing age-groups, age homogeneity for gun homicides was highest for
homicide offenders 25-34 years old, and were lowest for 20-24 year olds. The low
rates for the 20-24 group reflects their age status between the two other groups and
the higher likelihood of cross-age interactions.
5. Contextual Effects
Both popular and social science explanations of the homicide epidemic in
New York and elsewhere have focused on important social trends, particularly
changes in drug markets (Blumstein, 1995). Fagan et al. (1998) discuss the appeal of
these explanations First, homicide and drug epidemics have been closely phased,
both temporally and spatially, in New York and nationwide, for nearly 30 years
(Fagan, 1990, 1997). Homicide peaks in 1972, 1979 and 1991 mirror three drug
epidemics: heroin, cocaine hydrochloride (powder), and crack cocaine. These longterm trends predict that trends in drug use would occur contemporaneously with
trends in homicide. Second, the emergence of volatile crack markets in 1985 is cited
as one of the primary contextual factors that have driven up homicide rates in New
York(Goldstein et al., 1989; Johnson et al., 1990; Bourgois, 1995). Competition
between sellers, conflicts between buyers and sellers, and intra-organizational
organizational conflict were all contributors to lethal violence within crack markets
(Fagan and Chin, 1990; Hamid, 1994). Crack also is implicated in the decline of
homicide since 1991 (Curtis, 1998).
Figure 3 compares trends in gun homicides for three age groups with trends
SOCIAL CONTAGION OF VIOLENCE
12
in drug overdose deaths. Drug overdose deaths follow a pattern of short cycles, with
relatively brief periods of increase and decline. The rates increase from 1986 to
1988, decline through 1990, and increase again for three years before leveling off.
The runup of gun homicide rates in 1985 to 1988 matches the increase in drug
overdose deaths, but the homicides continue to increase even as drug overdose deaths
decline. Drug overdose death rates increase from 1992 to 1994, even as gun
homicide rates decline. Accordingly, there appears to be little mutual influence of
drug overdose deaths and gun homicide trends for any of the three age groups.9
Changes in drug use patterns may explain this disjuncture, with drugs such as heroin
and ketamine supplanting crack as the favored street drug (Curtis, 1998) . These
drugs are more likely to cause overdose deaths.
An alternative though imperfect indicator of drug market activity is drug
arrests. Arrests reflect both strategic decisions by police as well as drug market
characteristics. In conjunction with other indicators, arrests are a useful marker of
drug trends. The trend lines in Figure 4 for age-specific homicide victimization rates
and felony drug arrest rates also show little relationship between gun homicides and
drug arrests. Both drug arrests and gun homicides rates increase from 1986 through
1989, but the trend lines move in different directions after that. Homicides increase
through 1991 for adolescents and 1992 for young adults. Drug arrests decline from
1990 through 1993, and begin to rise again in 1994. Most of these felony drug
arrests were for sale or possession with intent to sell, and most were either crack or
cocaine arrests, the two drugs that were most actively traded in street markets. The
portion of felony drug arrests that involved crack or cocaine rose from 57 percent in
1986 to 64 percent in 1988, and declined steadily to 48 percent in 1995.
Overall, drug selling activity fails to adequately explain the runup and decline
9. Other indicators, such as drug use among arrestees recorded in the Drug Use Forecasting
System (DUF), also show little relationship with trends in gun homicide rates. Fagan et al. (1998)
show that the incidence of drug-positive arrestees remains unchanged throughout the period, and is
unrelated to both firearm and non-firearm homicide trends.
SOCIAL CONTAGION OF VIOLENCE
13
in gun homicides. Violence associated with drug use remained relatively infrequent
during the onset of the crack crisis (Fagan and Chin, 1990; Fagan, 1992). Moreover,
the share of homicides due to drug selling did not rise during the homicide runup
(Goldstein et al., 1989). Drug selling accounts for an unknown proportion of
homicides, with estimates ranging from about 10 percent in the FBI’s Supplemental
Homicide Reports,10 to 50 percent in local studies in New York (Goldstein et al.,
1989) or Los Angeles (Maxson et al., 1991). Thus, a decline in street-level drug
selling activity may have reduced, to some unknown extent, the types of social
interactions that lead to gun killings. But drug selling alone is unlikely to have
produced the unprecedented runup or decline in gun killings so consistently across
time, social groups and areas.
10. Analysis by authors, data available on request.
SOCIAL CONTAGION OF VIOLENCE
14
Finally, demographic changes also offer weak explanations of the homicide
increase or decline. The population for the highest risk groups, non-white males ages
15-29, declined by about 10 percent from 1985-95 (Fagan, Zimring and Kim, 1998),
a far smaller scale of change than the change that could produce the observed
declines in gun homicides. While is tempting to dismiss demography as a correlate
of the homicide decline, the relationship of population to a changing behavioral
pattern may be non-linear, though (Gladwell, 1996). In other words, did the
population decline reach a threshold where it could lead to a decline in the incidence
of gun homicides?11 This is a plausible but unfalsifiable explanation. Like the
effects of declining drug markets, the contraction in the highest risk population is an
important but unknowable influence on the decline in firearm homicides from 199296.
In other cities, the emergence and evolution of street gangs has been linked
to homicide increases. From 1980 to 1992, over 80 percent of the nation’s 250
largest cities reported new or intensified gang activity, much of it linked to drug
selling. In New York, however, gangs provide a very limited explanation of the
homicide trends, especially for the increase and decline in gun homicides. Street
gangs of adolescents in New York City are largely a thing of the past. Contemporary
street gangs are weak and disorganized structures (see, for example, Sullivan, 1989).
Current gangs either are extensions of multistate prison gangs such as the Latin
Kings or their rivals (Brotherton, 1998), tied to adult crime groups (see, for example,
Chin, 1995, on Chinese gangs), or are drug business organizations (Fagan, 1994;
11. According to Burt (1987), the relationship of network density to social contagion is non-linear.
Thus, epidemiologists discuss thresholds, or tipping points, where behavioral change accelerates
and spreads through a population before beginning its process of decline (see, for example, Crane,
1991; Gladwell, 1996).
SOCIAL CONTAGION OF VIOLENCE
15
Curtis, 1998). There are nascent gang structures in middle and secondary schools,
but they have emerged only recently and well after the decline in gun homicides.
II. NEIGHBORHOOD EFFECTS ON SOCIAL CONTAGION
OF YOUTH HOMICIDE
We begin the task of dissagregation of “contagion effects” into dynamic
processes across social aggregates. The first discussion sets out a framework for
understanding the hierarchical structure of contagion, from social circumstance to
individual interactions. Next, we present analyses to show the transmission of
behaviors across social areas or neighborhoods. We use census tract as the boundary
for “neighborhood,” based on the size (area) of tracts in New York and their
isomorphism with important social units such as public housing developments and
feeder school patterns.12 These analyses set the stage for the analysis in Section III
of micro-social interactions where social contagion take place.
12. Tracts are commonly used to represent neighborhoods in sociological research, due to their
size and robustness in predicting variation in a variety of social interactions. See, for example,
Kenneth Land, Patricia McCall, and Lawrence Cohen, Structural covariates of homicide rates: Are
there any invariances across time and space, 95 American Journal of Sociology 922 (1990). Most
cities lack well organized data sets that are aggregated to meaningful social areas. For exceptions,
see, for example, Robert J. Sampson, Stephen W. Raudenbush, & Felton Earls, “Neighborhoods
and violent crime: A multilevel study of collective efficacy,” 277 SCIENCE 918 (1997), analyzing
neighborhood differences in crime rates and informal social control in Chicago using tract-level
data; Jonathan Crane, The epidemic theory of ghettos and neighborhood effects on dropping out
and teenage childbearing. 96 American Journal of Sociology 1226-59 (1991), analyzing the spread
of teenage births across Chicago census tracts.
SOCIAL CONTAGION OF VIOLENCE
16
A. Background
The spread of ideas and practices is contingent on the way in which social
structure brings people together in close physical proximity within dense social
networks (Burt, 1987: 1288).13 Through a process of mutual influence involving
contact, communication and competition, adoption of behaviors occurs when
information is transmitted that communicates the substance of the innovation and the
consequences of adoption. The consequences can be econometric or intrinsically
pleasurable, and may be reinforced through the benefits of a vicarious experience or
a trial use. In addition, these behaviors acquire social meaning that is communicated
through repeated interactions within social networks (Lessig, 1995: 947).14
Contagious epidemics involve the transmission of an agent via a host through
susceptible organisms whose resilience is weakened by other conditions or factors
(Bailey, 1967). Susceptibility is critical to the ability of an agent to exert its process
on a host. This medical rendering of contagion can be analogized to social
contagion. Thus, the fundamental social causes of disease -- primarily social
structural, or ecological -- can be seen as pathways along which more micro-level
causes can exert their effect (Gostin, Burris and Lazzarini, 1999: 74). According to
Gostin et al., these fundamental social causes reflect inequalities that work in two
ways. First, these conditions increase exposure to the more proximal causes, whether
microbic or behavioral. Second, they compromise the resistance or resilience of
social groups to these proximal causes. That is, their exposure and their behavior in
those structural circumstances both have social roots.15
13. See, also, David C. Rowe and Joseph L. Rogers, A Social Contagion Model of Adolescent
Sexual Behavior: Explaining Race Differences, 41 Social Biology 1, 16 (1994), showing that an
epidemic model combining social contagion through social contacts among adolescents of within a
narrow age band explains the onset and desistance of adolescent sexual behavior.
14. See, also, Dan M. Kahan, Social Influence, Social Meaning, and Deterence, 83 Virginia Law
Review 349, 367-73 (1997).
15. Lawrence O. Gostin, Scott Burris, and Zita Lazzarini, The Law and the Public’s Health: A
Study of Infectious Disease Law in the United States, 99 Columbia Law Review 59, 75 (1999).
SOCIAL CONTAGION OF VIOLENCE
17
Memetics provides a complementary framework for understanding how
beliefs, ideas and behaviors spread throughout society. Memes are singular ideas that
evolve through a process of natural selection not unlike the evolution of genes in
evolutionary biology.16 The principal law governing the birth and spread of memes
is that of the “fittest ideas,” defined as those ideas that are the best at self-replication
rather than those that may be truest or have the greatest utilitarian value.17 In the
present analysis, violence may be the “fittest” behavior, even when it contradicts
more socially useful normative values imported from the dominant society. Memes
achieve high-level contagion through a variety of social interactions across social
units such as families and social networks, and each mode increases the “host”
population for that meme. The meme is then reproduced within networks and
transmitted across interstitial network boundaries. Replicated memes become what
Balkin refers to as “cultural software” that is expressed in language, behavior and
normative beliefs,18 creating a set of narrative expectations, or behavioral “scripts.”19
Script theory can explain contagion in several ways: (1) scripts are ways of
organizing knowledge and behavioral choices;(2) individuals learn behavioral
repertoires for different situations; (3) these repertoires are stored in memory as
scripts and are elicited when cues are sensed in the environment; (4) choice of scripts
varies between individuals and some individuals will have limited choices; (5)
individuals are more likely to repeat scripted behaviors when the previous experience
16. Aaron Lynch, Thought Contagion: How Beliefs Spread Throughout Society (1996); Jonathan
Balkin, Cultural Software (1998).
17. Lynch, id __.
18. Balkin at 42-57.
19. Jeffrey Fagan, Context and Culpability in Adolescent Violence, 6 Virginia Review of Social
Policy and Law 101 (1999). The script framework is an event schema used to organize
information about how people learn to understand and enact commonplace behavioral patterns. A
“script” is a cognitive structure or framework that organizes a person’s understanding of typical
situations, allowing the person to have expectations and to make conclusions about the potential
result of a set of events Robert P. Abelson, “Script processing in attitude formation and decisionmaking,” In Cognition and Social Behavior (J.S. Carroll and J. W. Payne, eds.) (1976); Robert P.
Abelson, “Psychological status of the script concept,” 36 American Psychologist 715 (1981).
SOCIAL CONTAGION OF VIOLENCE
18
was considered successful;and (6) scripted behavior may become “automatic”
without much thought or weighing of consequences.20
Accordingly, social contagion is convergence of transmission of behaviors
and beliefs that motivate or sustain them. Social contagion arises from people in
proximate social structures using one another to manage uncertainty of behavior.21
It requires an interaction where information, behavioral innovation, belief, or meme,
is transmitted across a social synapse. At its core, contagion occurs when two people
interact where one has adopted a construct and the other has not.
Contact,
communication or imitation, are influential processes that make transmission
possible.22
Synapses themselves are situated within social networks, and the
adoption of innovation or a meme triggers the adoption by another person. Burt
(1987) suggests that adoption has less to do with the cohesion of people within social
structures, or networks, and more to do with the structural equivalence – the social
homogeneity – of the network. That is, transmission is more likely to occur between
similarly situated persons – siblings, fellow graduate students, streetcorner boys –
than persons simply because they are closely bonded.23
Within structurally equivalent networks, similarly situated people are likely
to influence or adopt behaviors from one another that can make that person more
attractive as a source of further relations. The importance of structural equivalence
– or placement within a socially homogeneous interpersonal network – is that it
fosters interconnected patterns of relationships that makes contagion efficient.
In the remainder of this section, we show how transmission of violence
occurs across neighborhoods whose social structures of densely packed networks are
20. Abelson, supra n. 19
21. Burt, at 1288. Thus, the diffusion of innovation takes place in a parallel structural context as
does the diffusion of drug trends or sexual behavior. See, Rowe and Rogers, supra n. __;
Lawrence Gostin, The Interconnected Epidemics of Drug Dependency and AIDS, 26 Harvard
C.L.-C.R. Law Review, 113, 125 (1991).
22. Burt, id at 1288-89.
SOCIAL CONTAGION OF VIOLENCE
19
vulnerable to rapid contagion. In the next section, we show how the memes of
toughness and the valued status from violence are the object of transmission and
exchange. The implications for a social influence model of contagion are discussed
in the concluding section.
B. Susceptibility: Neighborhood Risk
We draw on the literature of neighborhoods and violence to construct a
framework of structural risk that simultaneously compromises resilience against
transmission while increasing susceptibility. Both theory and empirical research
suggest that neighborhoods are susceptible to the spread of violence when
structurally weakened (Taylor and Covington, 1988; Patterson, 1991; Massey,
1995).24 Wilson (1987) argues that there has been both an economic and a social
transformation of the inner city, where the exodus of manufacturing jobs beginning
in the 1970's has changed the social and economic composition of inner cities,
leading to a concentration of resource deprivation.
Wilson (Wilson, 1987: 58) goes on to suggest that the concentration of
resource deprivation in specific areas led to dynamic changes in the processes of
socialization and social control in those areas. As middle- and working class African
American families moved away from the inner cities when their jobs left, there
remained behind a disproportionate concentration of the most disadvantaged
segments of the urban populations: poor female-headed households with children and
chronically unemployed males with low job skills. The secondary effects of this
exodus created conditions that were conducive to rising teenage violence: the
weakness of mediating social institutions (e.g., churches, schools), and the absence
23. Id at 1291.
24. For example, recent studies suggest that violence shares several explanatory variables with
concentrated poverty (Wilson, 1987, 1991; Sampson and Lauritsen, 1994) or resource deprivation
(e.g., Williams and Flewelling, 1988; Land et al., 1990). These constructs describe the lack of
sufficient means, including income poverty and inequality, to sustain informal social control
SOCIAL CONTAGION OF VIOLENCE
20
of informal social controls to supervise and mentor youths.25 Wilson (1987) refers
to these conditions of weak social control as social isolation.
The concept of social isolation suggests an ecological dynamic where the
components of poverty and structural disadvantage are interconnected with the
dynamics of social control and opportunity structures. The decline of manufacturing
jobs increased unemployment among adult males, primarily African Americans,
whose job skills limited them to unskilled labor. Other economic transformations,
including the rise of service and technical jobs outside central cities, motivated the
exodus of middle-class families to the outer rings and suburbs surrounding the inner
cities. Remaining within the abandoned central cities were unskilled males whose
"marriage capital" was low, giving rise to an increasing divorce rate and declining
marriage rate. Changes in the composition of central city neighborhoods also
weakened the social institutions that were critical to the informal social control and
collective supervision of youths. The weakening of social controls had their strongest
effects in transactional settings of neighborhoods and in places like schools and
(Sampson and Wilson, 1995).
25. The male divorce rate also is a consistent predictor of violence and homicide rates, and effects
are greater for juveniles than for adults. For example, Messner and Sampson (1991) showed that
Black family disruption was substantially related to rates of murder and robbery involving Blacks.
These findings are consistent with the consistent findings in the delinquency literature on the
effects of broken homes on social control and guardianship. The effects of male divorce can be
interpreted either as a consequence and correlate of the rise of female-headed households, or as an
indicator of weak social control of children who then are raised primarily by women. Whatever its
meaning, the male divorce rate has positive, clear cut effects on robbery, assault, rape and
homicide (Sampson and Lauritsen, 1994).
SOCIAL CONTAGION OF VIOLENCE
21
church where adolescent development takes place. And, the exodus of middle class
families from inner cities weakened the political strength of the remaining families,
leading to physical deterioration (Wallace, 1991), lower housing values and in turn
increased residential (spatial) segregation (Massey and Denton, 1993).
In turn, the social isolation of people and families was extended to
institutions. The rise in poverty and weak social institutions also undermined the
presence of and institutional support for conventional behaviors. As a result,
conventional values and behaviors were attenuated because they were not salient and
had little payoff for one's survival or status (Elliott et al., 1996; Wilson, 1987). These
dynamics in turn attenuated neighborhood social organization, increasing the
likelihood that illegitimate opportunity structures would emerge: illegal economies
including drug distribution or extortion, gangs, and social networks to support them.
These structures competed with declining legal work opportunities both as income
sources and as sources for social status. As these networks flourished, the systems
of peer and deviant social control replaced the controls of social institutions and
conventional peer networks (Fagan, 1992).
Accordingly, structural change that violence and homicide are more likely
to occur in an ecological context of weak social control, poorly supervised adolescent
networks, widespread perceptions of danger and the demand for lethal weapons, and
the attenuation of outlets to resolve disputes without violence. It is in this ecology
of danger that violence becomes transmittable through weapons and their impact on
perception and decision making in social interactions.
C. Guns and Social Contagion
In these social contexts, several processes contributed to the epidemic of
lethal violence. The growth in illegal markets heightens the demand for guns as basic
tools associated with routine business activity in illegal markets (Johnson et al., 1990;
Blumstein, 1995). In turn, the increased presence of weapons and their diffusion into
SOCIAL CONTAGION OF VIOLENCE
22
the general population in turn changes normative perceptions of the danger and
lethality associated with everyday interpersonal disputes, giving rise to an "ecology
of danger" (Fagan and Wilkinson, 1998). Thus, we hypothesize that guns were
initially an exogenous factor in launching an epidemic of gun homicide, but became
endogenous to socially isolated neighborhoods and came to dominate social
interactions (Wilkinson and Fagan, 1996). Everyday disputes, whether personal
insults or retributional violence, in turn are more likely to be settled with lethal
violence (Fagan and Wilkinson, 1998).
Whether viewed in social, medical or memetic frameworks, guns can be
constructed as a primary agent of violence contagion over the most recent epidemic
cycle.
Guns are a form of social toxin (Delgado, 1985) in everyday social
interactions, altering the outcome of disputes and changing the developmental
trajectories of young males whose adolescent development took place in contexts of
high rates of gun use (Fagan, 1999).26
The development of an ecology of danger reflects the confluence and
interaction of several sources of contagion. First is the contagion of fear. Weapons
serve as an environmental cue that in turn may increase aggressiveness (Slaby and
Roedell, 1982). Adolescents presume that their counterparts are armed, and if not,
could easily become armed. They also assume that other adolescents are willing to
use guns, often at a low threshold of provocation.
Second is the contagion of gun behaviors themselves. The use of guns has
instrumental value that is communicated through urban “myths,” but also through the
incorporation of gun violence into the social discourse of everyday life among preadolescents and adolescents. Guns are widely available and frequently displayed.
They are salient symbols of power and status, and strategic means of gaining status,
26. Jeffrey Fagan, Context and Culpability in Adolescent Violence, Virginia Review of Social
Policy and Law (in press).
SOCIAL CONTAGION OF VIOLENCE
23
domination, or material goods.
Third is the contagion of violent identities, and the eclipsing or devaluation
of other identities in increasingly socially isolated neighborhoods. These identities
reinforce the dominance hierarchy built on “toughness” and violence, and its salience
devalues other identities. Those unwilling to adopt at least some dimensions of this
identity are vulnerable to physical attack. Accordingly, violent identities are not
simply affective styles and social choices, but strategic necessities to navigate
through everyday dangers.27 The social meanings of violent events reach a broader
audience than those immediately present in a situation. Each violent event or
potentially violent interaction provides a lesson for the participants, first-hand
observers, vicarious observers, and others influenced by the communication of stories
about the situation which may follow.
D. Analytic Models: Diffusion and Contagion
We estimated models of contagion of gun violence and its diffusion across
New York City neighborhoods beginning in 1985. We tested two distinct conceptual
models for the spread of gun violence from one neighborhood to the next. An
outward contagion model posits that adolescent homicide spreads out from a central
census tract (T) or the immediate neighborhood, to adjacent census tracts (X, Y and
Z) or the surrounding community. In this model, the incidence and prevalence of
adolescent homicide in a given neighborhood exerts a significant influence over the
incidence and prevalence of adolescent homicide rates in the adjacent community.
27. One important development is a breakdown in the age grading of behaviors, where traditional
segmentation of younger adolescents from older ones, and behavioral transitions from one
developmental stage to the next, are short-circuited by the strategic presence of weapons. Mixed
age interactions play an important role in this process. Older adolescents and young adults provide
modeling influences as well as more direct effects. We found that they exert downward pressure
on others their own age and younger through identity challenges which, in part, shape the social
identities for both parties. At younger ages, boys are pushing upward for status by challenging
boys a few years older. See, Deanna Wilkinson, The Social and Symbolic Construction of
Violent Events among Inner City Adolescents (Doctoral dissertation, Rutgers University) (1988).
SOCIAL CONTAGION OF VIOLENCE
24
This influence is hypothesized to operate in at least two different ways. First,
a threshold effect is expected concerning adolescent homicide counts, such that the
presence of at least one adolescent homicide in a given neighborhood will
substantially increase the probability of experiencing at least on adolescent homicide
in the surrounding community. Second, with respect to adolescent homicide rates,
positive covariation is anticipated whereby increases or decreases in the adolescent
homicide rates in a neighborhood are reflected in concomitant increases or decreases
in the surrounding community’s adolescent homicide rate.
It is also possible that the contagion effect of adolescent homicides is, in fact,
reversed. Accordingly, the inward contagion model asserts that the adolescent
homicide level in an immediate neighborhood is at least partially contingent upon the
level of adolescent homicides in its broader community. Again, the two distinct
relationship forms (threshold effect and positive covariation) are possible.
By
considering the simultaneous influence of adjacent spaces, we address the problem
of spatial autocorrelation by effectively controlling for mutual influences within and
over time.
In addition to corresponding adolescent homicide rates, both the outward and
inward contagion models incorporate relevant structural and demographic features
of neighborhoods and communities as key explanatory constructs. Thus, for the full
outward contagion model, the presence and rate of adolescent homicide in the
surrounding community is function of the relevant characteristics of the community
as well as the presence or rate of adolescent homicide in the neighborhood. In
contrast, the full inward contagion model suggests that both relevant neighborhood
features and the presence or rate of adolescent homicide in the community predict the
presence or rate of adolescent homicide in the neighborhood. Although these factors
are presumed to play a significant, independent role in the prediction of adolescent
homicide rates, it is nonetheless hypothesized that effects of homicide rates as
dependent variables will remain substantial even once controls for relevant
SOCIAL CONTAGION OF VIOLENCE
25
neighborhood and community characteristics have been introduced.
Models were estimated using mixed effects regression models (Singer, 1999).
Mixed effects regression models approximate multilevel models where data are
hierarchically organized. For example, this class of models is useful in cases such
as estimating the simultaneous effects of school climate and individual student family
background on standardized test scores, or neighborhood characteristics and
household composition on crime rates.28 In these examples, we might specify the
ecological effects of school or neighborhood as fixed effects and individual
household or family influences as random effects. This specification approximates
the presumed hierarchy of influences. One may reverse the specification, as well,
comparing estimates to assess reciprocal effects between the two sets of predictors.
Mixed effects models simulate the hierarchy of effects by estimating the differences
in error-covariance matrix structures for each set of effects (Singer, 1999: 3).29
28. Mixed effects models also are useful in estimating individual growth curves, or within-subject
change over time. See, for example, Anthony Bryk and Stephen Raudenbusch, Hierarchical
Linear Models (1992), T.A.B. Snijders and R.J. Bosker, Modeled Variance in Two-Level Models,
22 Sociological Methods and Research 342 (1994).
29. A second advantage of the mixed models approach is that it allows for greater flexibility in
specifying the covariance structure of the data. Specifically, mixed models allow for the analysis of
data where the requisite assumptions of OLS regression concerning error term independence are
violated. This is particularly important in research involving aggregate units. Because the
neighborhoods and communities are comprised of geographically contiguous census tracts,
autocorrelation is inherent in the data structure and it would be inappropriate to assume a simple
covariance structure for these analyses. All of the models are instead analyzed with an
autoregressive covariance structure.
SOCIAL CONTAGION OF VIOLENCE
26
In this analysis, the contagion models specifies the characteristics of the
“sending” neighborhood as fixed effects, and characteristics of the adjacent
neighborhood as random effects. To estimate contagion effects on prevalence, we
use the logistic form of the equation. To estimate changes in rates, we use the linear
regression form of the equation. To estimate changes over time, we use a repeated
measures design where year is included as a random effect that approximates a
developmental growth curve (see, Goldstein, 1995).
E. Data
Dependent Variables. There are four dependent variables, two for each of the
contagion models. In the outward contagion model, the first dependent variable is a
dichotomous indicator of whether or not at least a single adolescent homicide had
occurred in at least one of the census tracts that comprise the surrounding
community. This will allow for the testing of a threshold effect hypothesis. The
second dependent variable is the aggregate adolescent homicide rate for the
surrounding community and is directly related to a positive covariance hypothesis.
The dependent variables for the inward contagion model are: (1) a dichotomous
indicator noting whether at least a single adolescent homicide had occurred in the
neighborhood, and (2) the adolescent homicide rate for the neighborhood. Again,
these dependent variables are used to address the threshold effect and positive
covariance propositions respectively.
Independent Variables. In addition to the variables based on adolescent
homicide rates for each neighborhood by year, we operationalized the model of
neighborhood risk or susceptibility described earlier. Data reflecting the structural
and demographic composition of neighborhoods and communities are predictors the
contagion models.
Following Land et al. (1990), we selected 14 tract-level variables from the
1990 Census (STF3A and 3B files) to characterize social areas. Means and variances
SOCIAL CONTAGION OF VIOLENCE
27
for the 14 initial variables are presented in Appendix B. Principle components
analysis was used to eliminate autocorrelation among the 14 variables, and identify
three orthogonal and conceptually distinct factors. The results in Table 5 show that
neighborhoods and communities are characterized along three dimensions:
deprivation, population characteristics, and social control. Because communities are
actually composites of census tracts, their factor scores are actually weighted factor
score composites.30
Lag Effects. The “contagion effect” of adolescent homicide is not expected
to be immediate. Rather, it is more reasonable to assume that some period of time
must elapse between the occurrence of an adolescent homicide (threshold effect) or
change in the adolescent homicide rate (positive covariation) and the realization of
a related occurrence or rate change. For this study, the time elapsed is estimated to
be one year.31 Thus, all the of models have been run with data lagged at one-year
intervals. For example, with the outward contagion model, the adolescent homicide
rate in a neighborhood for 1990 is used to predict the rate in the surrounding
community for 1991. Conversely, the adolescent rate for a neighborhood in 1993
under the inward contagion model is estimated using the community adolescent
homicide rate from 1992.
F. Results
1. Epidemic Effects: Neighborhood Trajectories. Table 6 simply shows the
prediction of homicide rates over time from the three neighborhood factors in each
census tract. Recall from Figures 1 and 2 that the distribution of homicides over
time is curvilinear, reflecting the epidemic pattern. Accordingly, we applied
30. The weighting variable was 1990 population.
31. It is worth noting that models with two year time lags produced results very similar to those
reported here. Thus the results as reported to not appear to be an artifact of the lag time chosen.
SOCIAL CONTAGION OF VIOLENCE
28
individual growth curve models to neighborhoods to estimate models for change over
time by neighborhood. Table 6 shows the estimation of a growth curve model with
covariates. We assume neighborhood structure is fixed at its 1990 status, midpoint
in the time period.32
The results show that homicides were concentrated in the city’s poorest
neighborhoods over time. The effects of neighborhood factors on the likelihood of
experiencing an adolescent homicide in that area consistently significant, but vary
greatly in their magnitude. Repeated measures logistics regression shows that
homicide rates were highest over time in neighborhoods characterized by economic
deprivation (poverty) and population risk factors. Anonymity and social control are
not significant predictors of homicide trends over time.
2.
Reciprocal Contagion Effects.
We next estimated models of the
reciprocal effects of homicides between neighborhoods. Table 7 shows the results
of both repeated logistic regression and repeated standard (least squares) regressions
to estimate the spread of homicide over time from one neighborhood to the next.
Models 1 estimates the spread of homicide from a neighborhood to the adjacent
neighborhoods without covariates; Model 2 adjusts these estimates for the effects of
covariates of the source neighborhood. Model 3 estimates the inward spread from
surrounding areas to a surrounded neighborhood, and Model 4 adjusts these estimates
for the effects of covariates of the target neighborhood.
Together, the significant coefficients in both parts of Table 7 suggest
reciprocal contagion effects. That is, adolescent homicide is characterized by both
outward and inward contagion processes. In the upper half of Table 7, results of all
four models show that the presence of at least one adolescent homicide in a census
tract significantly increases the likelihood of at least one adolescent homicide
32. We estimate models with covariates where both slopes and intercepts vary, and residual
observations within census tracts are correlated through the within-tract error-covariance matrix.
SOCIAL CONTAGION OF VIOLENCE
29
occurring in the surrounding neighborhood. The introduction of neighborhood
characteristics reduces the effect size for both inward and outward contagion, but the
contagion effect remains significant even when controlling for relevant neighborhood
characteristics. The lower half of Table 7 shows a similar pattern of results for
models of changes in the rates of homicide.
3. Guns and Social Contagion of Homicide. To assess the effects of guns on
social contagion, we next estimated models that separated the contagious effects of
gun homicides from non-gun homicides. Models of outward contagion of gun
homicides are shown in Tables 8 and 9. Model 1 estimates the effects of gun
homicides in one neighborhood on gun homicides in a surrounding neighborhood;
Model 2 adjusts these estimates for the effects of the covariates of the source
neighborhood. Models 3 and 4 repeat these procedures to estimate the contagious
effects of gun homicides on non-gun homicides. The upper half shows results for
changes in homicide prevalence, and the lower half shows results for changes
homicide rates.
Tables 8 and 9 both show that the contagious effects of gun homicides in one
area are limited to gun homicides in adjacent areas. Table 8, Model 1 shows that gun
homicides are contagious, both for prevalence and increases in rates. Although
neighborhood characteristics diminish the effect sizes, Model 2 shows that the results
remain the same. The presence of gun homicides in one neighborhood significantly
increases the likelihood of a gun homicide in any of the surrounding neighborhoods
in the subsequent year, after independent of neighborhood characteristics. Once
again, a similar results were obtained for changes in the homicide rate.
No such effect was found for the effects of gun homicides on non-gun
homicides. Models 3 shows evidence of contagion, but Model 4 shows that the effect
is not statistically significant when the model is adjusted for neighborhood
covariates. And once again, a similar pattern of results is evident for changes in the
homicide rate.
SOCIAL CONTAGION OF VIOLENCE
30
Table 9 repeats this analysis for inward contagion: the spread of homicides
from surrounding neighborhoods to a central neighborhood. The same pattern of
effects was observed: gun homicides were contagious only to produce other gun
homicides, after controlling for neighborhood characteristics; gun homicides had no
such effects on the prevalence or rates of non-gun homicides in the surrounding
areas.
Tables 10 and 11 reverses this analysis, estimating the contagious effects of
non-gun homicides on gun and non-gun homicides. The models in Table 10 show
no outward contagious effects of non-gun homicides on either gun or non-gun
homicides. Table 11 shows one significant effect: inward contagion of non-gun
homicide rates.
Non-gun homicide rates predicted higher non-gun rates in
surrounded neighborhoods, after controlling for neighborhood characteristics.
However, the relatively high proportion of gun to non-gun homicides
throughout this period raises the possibility that estimates of non-gun contagion
might be unstable. We addressed this by re-estimating the models in Tables 10 and
11, but including total homicide as the predicted measure. The results in Tables 12
and 13 confirm the earlier pattern for gun homicides: gun homicides show effects of
both outward and inward contagion, for both prevalence and rates, after controlling
for neighborhood characteristics. Non-gun homicides appear to have contagious
effects on total homicides, but only in models of inward contagion. There is no
outward contagion effect of non-gun homicides on total (gun and non-gun)
homicides.
G. Summary
The contrast in findings of non-gun contagion for non-gun versus total
homicide illustrates the importance of guns in the dynamics of social contagion at the
population level. In addition, the consistent finding of socio-economic risk as a
contributor to the spread of violence captures both the significance of susceptibility
SOCIAL CONTAGION OF VIOLENCE
31
and the importance of structural equivalence in shaping the trajectory of diffusion.
That is, the adoption at the population level of gun violence was facilitated by the
social concentration of poverty, and of the close social synapses intrinsic to poor
neighborhoods. Accordingly, the social and spatial clustering of homicide suggests
that it is concentrated within overlapping social networks in small areas (Fagan and
Wilkinson, 1998).
Social contagion theory suggests that individuals are likely to mutually
influence the behaviors of others with whom they are in frequent and redundant
contact (Burt, 1987; Bovasso, 1996, at 1421). The social interactions underlying
assaultive violence suggest its spread by social contact (Loftin, 1986). Moreover, the
dissolution of social networks from attrition -- death, incarceration, or declining birth
rates -- would reduce opportunities for social transmission. We explore these themes
in the next section.
III. SOCIAL IDENTITY, YOUTH VIOLENCE,
AND SOCIAL CONTAGION
We turn next to an analysis of the individual- and group-level processes of
social contagion. We identify dynamic social processes that fuel the social contagion
of youth violence.
At the heart of this process is the development and
communication of social identities among males. Violence plays a central role in the
formation and maintenance of social identities among adolescents in inner city
neighborhoods, where structural conditions often make other social identities
unavailable. The contagious element in the epidemic of youth violence is the process
of making, maintaining and exchanging status through the use of violence. The
strategic value of guns in this process intensifies the dynamics that fuels the
epidemic. Thus, the presence of guns in events where there is “status forcing” or
exchanges of identities links together persons and events across time, and sustains
SOCIAL CONTAGION OF VIOLENCE
32
the processes of social contagion.
We begin the discussion by examining the role of identity in adolescent
development, and decomposing identity into several components. Then, we review
specific social identities, and show how identities are linked across persons to
provide a motivational context for violence. Next, we describe the process of “status
forcing” and its role in fueling identity-based violence. The importance of guns is
evident in these interactions.
A. Youth Violence and Social Identity
Issues of respect, honor, and pride are repeatedly described as central features
of male identity formation beginning in early adolescence.33 “Toughness” also has
been central to the development and maintenance adolescent masculine identity in
many social contexts of American life. Physical prowess, emotional detachment, and
the willingness to resort to violence to resolve interpersonal conflicts are hallmarks
of adolescent masculinity. While these terms have been invoked recently to explain
high rates of interpersonal violence among nonwhites in central cities, “toughness”
has always been highly regarded and a source of considerable status among
adolescents in a wide range of adolescent subcultures, from street corner groups to
gangs.34
While changing over time with tastes, these efforts at “impression
33. See, for example, Messerschmidt, 1993; Anderson, 1994; Sullivan, 1989; Majors and Billson,
1992; Oliver, 1994; Luckenbill and Doyle, 1989; Toch, 1969; Horowitz and Schwartz, 1974;
Nisbett and Cohen, 1996; Connell, 1995; Canada, 1995.
34. Whyte, 1943; Goffman, 1959; 1963; 1967; Wolfgang and Ferracuti, 1982; Canada, 1995;
SOCIAL CONTAGION OF VIOLENCE
33
management” through violence have been evident across ethnicities and cultures
(Katz, 1988).35 Violence often is used to perpetuate and refine the pursuit of
“toughness,” and to claim the identity of being among the toughest.
Hagedorn, 1997.
35. In some cases, displays of toughness are aesthetic: facial expression, symbols and clothing,
physical posture and gestures, car styles, graffiti, and unique speech are all part of “street style”
that may or may not be complemented by physical aggression. The ‘tough’ status requires young
males to move beyond symbolic representation to physical violence
SOCIAL CONTAGION OF VIOLENCE
34
The process of self-preservation through displays of toughness, nerve, or violent
behavior is necessary part of day to day life for inner-city adolescents, especially
young males.36 Acquiring fighting skills (and perhaps more importantly shooting
experience) is considered important as a means of survival in the inner city. The
status and reputations earned through violent means provide inner city adolescent
males with positive feelings of self worth and “large” identities especially when other
opportunities for identity development are not available (Messerschmidt, 1993;
Hagedorn, 1997). Those individuals who have prestigious identities are granted
respect while the stigmatized face attacks on self.37 Group formation crystalizes
one’s personal and social identity.38
Social identity often shapes the extent to which violence occurs.39 Social
identity has a particularly strong influence because “individuals have little control
over situations and especially going outside of the expected role for their particular
social identity” (Goffman, 1963: 128). Strauss (1996: 57) explained that “face-toface interaction is a fluid, moving, running process; during its course the participants
take successive stances vis-a-vis each other. ...The initial reading of the other’s
identity merely sets the stage for action, gives each some cues for his lines.”
In addition, Strauss po
who are tough, who have gained respect by proving their toughness, and who reenact
36. See, Fagan and Wilkinson, 1998, and Sullivan, 1989:113.
37. The concept of respect or honor refers to granting deferential treatment to what Goffman
called one’s “personal space.” One who grants another respect would acknowledge and esteem the
other’s individuality and personal space (or least not attack it). The adolescent male is looking to
others to reflect back (“looking glass self” phenomenon) aspects of his own self-image which is
constantly shaped and reshaped within the context of social interaction with others.
38. Many of the vital functions of adolescent social life operate through these groupings whether
they are loosely or tightly connected (e.g. social learning and mentoring, play, nurturing, social
support, and economic opportunity). Researchers have described the dynamic nature of social
identity during different periods of adolescent development. Kinney (1993) for example, found
that “identity recovery” was possible through increased opportunity and diversity of peer groups.
These school-based studies show that the school setting offers a myriad of opportunities and
constraints to identity development. The street, as a social context, offers a similar opportunity for
adolescent identity formation, trials, and maintenance. For specific examples, see: Goffman, 1963;
Schwendinger and Schwendinger, 1985; Eder, 1995; Kinney, 1993.
SOCIAL CONTAGION OF VIOLENCE
35
their appropriate role in public.
B. Status Forcing and the Communication of Social Identity
Social identity may be built, earned, applied, or assigned (Strauss, 1996:
145). During the course of social interaction, individuals “may force each other into
such statuses” (79). Status forcing has different consequences for the identity of
actors depending upon the duration and permanence of that status. Strauss argued
“it makes a difference whether the placement is temporary (banishment), permanent
(exile), or of uncertain duration (idolization)” (Strauss, 1996: 83). He argued that
some forced statuses were “reversible” and therefore would have less long term
impact on identities.
In this section, we analyze narrative reconstructions of violent events reported
by 125 adolescents and young men from two inner-city neighborhoods in New York
(Fagan and Wilkinson, 1998). The events have been classified as erupting out of
some type of challenge or test to one’s social identity or status. Three distinctive
types of challenges to identity were described: personal, material, and social.
Personal attacks challenge who a person has a right to be (projected self-image),
material affronts contest what possessions a person has a right to have, and finally,
relational assaults who a person has a right to be in relationship with. The initial
interaction sparking many of these events involved some type of insult, degradation,
violent threat, bump, slight, ice grill (hard stares), domination, cunning, or
unprovoked physical attack. These violent event accounts for 66.5% of all of the
events described by respondents.
The parameters of a character contest are
determined by social interactions among actors of differing status (“mixed
interactions”) in specific contexts.
We describe different types of situational identities that our respondents
39. Supra, n. 16 and supra, n. 18.
SOCIAL CONTAGION OF VIOLENCE
36
reported, how those identities were won and lost, and how young men compete for
desirable identities. The analysis explores what appears to be a hierarchy of social
identities with regard to violent behavior. Young inner-city residents must learn to
negotiate the street world by building a social identity, projecting a reputation, and
developing a protective peer group in the neighborhood. The process of finding a
niche and forming a “safe” identity typically includes engaging in violent behavior.
The data also suggest that guns play a significant role in forming and sustaining
“positive” social identities within the neighborhoods.
1. Socialization and Identity
The social construction of male identity in the context of the street world
follows specific age-graded tracks. Violence plays a central role in “defining” social
identities during different age periods in the inner-city street context. Development,
age specific expectations, and ritualized “rites of passage” add legitimacy to gaining
and withholding respect through violent means. Violence appears to be a resource
for passage from one status or identity to another. Making a powerful social identity
is a critical tool for survival in the inner-city context.
The socialization process into this dynamic is quite clear according to our
respondents. At an early age, males frequently experience violent attacks and must
learn “how it was in the street.” One respondent reported: “Alright say your small,
probably let people pick on you and stuff like that, just let them do what they want
but as you get older you start fighting back. You stop letting people take advantage
of you.” (SBN35). Developing a desirable social identity for the street appears to
become important for males beginning between the ages of seven and eleven.
Defining one’s status in comparison to others involves a number of staged plateaus
or scheduled. According to several respondents (SBN27) “everybody’s a herb when
they’re --- in the beginning, everybody. ...Everybody I know who’s keeping it real
[being honest] has gotten fucked up so bad that they just don’t wanna get fucked up
SOCIAL CONTAGION OF VIOLENCE
37
no more. So that’s why they act the way they act.”
Two themes emerged in this analysis: (1) individuals start status-deprived and
(2) establishing status requires public performance. The process was described as a
kind of “who’s who” of the street by our sample, in essence a way of identifying
potential threats and resources within neighborhood associations. Participation in
violent social interactions, provides young men with information about the abilities
and potential of others with whom they share social space and time. This public
performance allow others to classify and categorize males in terms of threat, power,
“heart,” and status.
Peer groups play a significant role in defining social identities. Belonging to
a clique or street corner group may fulfill a variety of needs for the young men
including: protection, income generation, adventure, companionship, love, identity
affirmation, partying, and drug/alcohol consumption in a social atmosphere. The
social network, among other contexts, enables hegemonic masculine views to take
shape. Groups also take on social identity and group affiliation brings with it
privileges and obligations. According to our sample, criminals and males who
exhibit tough qualities and behavior are the “populars” and get the most attention
from others.
The pressure among peers to “be part of the scene” or to “prove that you are
capable of using violence to fit in” to the street life was very great. Group influence
seemed especially important in establishing one’s identity. One respondent
(ENYN17) described how peers made assumptions about him based upon the type
of household he lived in. His belief that others felt that “shit was sweet in a private
house” made him feel like an outsider. Boys from the projects intensified this
distinction by teasing and threatening the respondent. The respondent had to prove
he was “status worthy” and did so by being the first kid on the block with a gun. The
gun was used strategically to demonstrate the respondent’s capacity for violence.
Identity challenges occurred when the status of others was unknown or when
SOCIAL CONTAGION OF VIOLENCE
38
the situation calls for definition and classification. Shows of disrespect or “dissing”
are often defined by young males as an intentional attack or attempt to downgrade
someone else’s identity. These challenges were addressed aggressively. All parties
in this type of interaction defined the situation from their reference point which
included the collective meanings attached to the action. This identity negotiation or
testing process is clearly central to the making, remaking, and breaking of identities.
Respondents described the testing process as a necessary part of social development
which “made me stronger” and “to this day nobody can’t do that shit to me.”
C. A Hierarchy of Social Identities
In this context, identities based on violence were conflated with social status.
Gun use was an important part of the status hierarchy. Three broad types of social
identities were described by the respondents: being “crazy/wild” (frequent unstable
fighter/shooter), “holding your own” (functional fighter/shooter), and being a punk
or herb (frequent victim struggling for survival).40 Social identities become more
salient through repeated performance. The social meanings attached to each
performance determines when and how an actor will be known to others in the
neighborhood context, and in turn, subsequent interactions will be defined.
Thus an individual’s social identity can both prevent violence from coming
(he won’t get picked on) and promote additional violence (other young men will
attempt to knock him off his elevated status). The individual who performs poorly
becomes known and labeled as being a “punk” or “herb.” The person who has a
40. The three types of social identities described in this paper were most prominent among this
sample. Most of the interactions were defined in terms of avoiding being classified as a punk or
herb. Respondents did describe other violence-related social identities including: “the avoider”
“the nice guy” “the beef handler,” “too cool” for violence, etc.
SOCIAL CONTAGION OF VIOLENCE
39
“successful” performance gains status and becomes known for “holding his own.”
The young man who gives an “extraordinary” performance is labeled as being “wild”
or “crazy.” These social identities may be temporary or permanent. This section
briefly describes the characteristics of three ideal identity types. The majority of
respondents would classify themselves as being someone who “holds his own” at the
time of the interview. A small number would be described as fitting into the “crazy,”
“wild” or “killer” identity at the time of the interview.
Few, if any, of the
respondents would classify themselves as a punk or herb during the period of the
interview. Looking back over their life histories however, most respondents, 78%
of those queried (71 of 96 respondents), reported experiencing one or more situations
during childhood or adolescence of feeling like a punk or herb as direct result of
violence perpetrated against them by older more powerful males. All of the 125
respondents described the importance of using violence to gain social status and
personal security.
1. Being Known as “Crazy” “Wild” or “Killer”
At the top of the identity hierarchy of the street is the “crazy” “wild” or
“killer” social identity. Individuals who perform extraordinary acts of violence are
frequently feared and granted a level of respect that others cannot easily attain. A
small number of respondents in our sample described themselves or others as being
“wild,” “crazy,” or a “killer.” Some took on this identity temporarily or situationally
while others described themselves as always that way. The performances are often
socially defined as shocking or judged to be beyond what was necessary to handle a
situation.
Once an individual gives an extraordinary performance he may notice
changes in the way others relate to him. He may also start viewing himself
differently. This status brings with it a certain level of power and personal
fulfillment that may be reinforced by projecting this identity. Future violent
SOCIAL CONTAGION OF VIOLENCE
40
performances would enable him to maintain the image of the most violent or toughest
on the street.
A person who has an identity as someone who is crazy, wild or a killer, gives
off the impression that he has extreme heart, is untouchable, and does not care about
what happens. He has the capability to use extreme violence and gets respect for
dominating others. Others may want to associate with him to benefit from his high
status on the street. The identity itself carries privileges, expectations, and obligations
which may open the individual to additional opportunities for violence. The
powerful identity may be forced downward by someone else’s extraordinary
performance.
2. Being Known as “Holding Your Own”
Many respondents described the process of “holding their own” in violent
situations and how personal identities formed around displays of “doing what you got
to do” are generally positive on the street. The majority of our respondents would be
classified as “holding their own.” Individuals who “hold their own” are respected on
the street although they will eventually face challenges to their ability to do “what it
takes” in heated situations and in all likelihood faced numerous challenges on the
way up to that status (Strauss, 1996: 90). A person who has an identity as someone
who holds his own, gives off the impression that he has the capability to use extreme
violence but does so only when necessary. This person will face a challenge directly
and is respected for that position. This identity allows an individual to be considered
an “insider” with the street world, however, this status can be unstable and may
require acts of violence when faced with public attacks on identity.
A person who ‘holds his own’ has used violence as a resource for obtaining
that status. These young men face the same type of testing process as the punk or
herb however it is expected that this class of men will handle their conflicts with
violence and it will be effective. If violence was not effective, someone who is
SOCIAL CONTAGION OF VIOLENCE
41
known to “hold his own” will be granted respect for putting up a good fight or taking
a bullet “like a man.” If this character is situationally “punked” or “herbed” by
someone with a lower status, his identity could face a downward slide.
3. Being Known as a “Punk” or a “Herb”
At the bottom of the status hierarchy of the street is the punk or herb. Like,
the school-based “nerd” or “dweeb” the “punk” or “herb” identity is assigned to those
who do not fit into the deemed high status or tough identities. In the inner-city, those
who cannot fight or prove of their toughness may be instead stigmatized either
temporarily or permanently. Other guys in the neighborhood will act upon that
stigma.
The process of punking or herbing someone, as respondents called it,
closely resembled the process of ‘fool-making’ described by Klapp (cited in Strauss,
1996).
If someone has the punk or herb identity he is considered “fair game” for
attacks and robberies. The attacks are motivated both by the need to restate the
dominance hierarchy and as a sort of punishment for not living up to group norms.
If a young man does not have a tough identity or at least have close associates or
relatives who can protect him either by association or literally, he is a punk. Others
in the setting degrade, dominate, and victimize those individuals who have punk or
herb characteristics.
The degradation typically involves a direct or implicit
emasculation of the “weaker” males. Punks and herbs are also called “soft”
“suckers” “wimps” “pussy” “bitch” “ass” and “chumps.” Given the intensified
acceptance of hegemonic masculinity in the inner city context, these messages would
have a strong negative impact on a punk or herb’s self-image. Most young men
assume that “outsiders” in the neighborhood (and relevant social network) are punks
or herbs and the presumed punk or herb must prove otherwise.
SOCIAL CONTAGION OF VIOLENCE
42
C. Status Forcing and Situational Identities
Negotiating the street requires tests of character, knowledge of the rules of
respect, knowledge of the players on the street, and open displays of violence. Since
the social identities of actors in the setting are ever changing the testing process
appears to be continuous until at least age twenty. The “mixed contacts” between
young males of differing social standing provides the clearest context for observing
status forcing --thus seeing identity on the way up, the way down, and holding steady.
For example those who are attempting to transcend a punk or herb identity
must interact with those who “hold their own” or typed as “killers” in order to avail
an opportunity for “identity recovery” (see Kinney, 1993). Punks and herbs take all
sort of abuse in our inner-city neighborhoods. They get used by more powerful street
guys to test their nerve. A young male who “holds his own” may face threats from
punks who are attempting to transcend into a higher social identity. The identity shift
is also publicly constructed through the reinforcement and praise by those observing
and/or hearing about the performance. Observers and others within the neighborhood
context offer rewards in the currency of respect for these performances. Violent
behavior motivated by other issues may also have side benefits for social identity
especially among members of the peer group.
Gun use may involve “crossing a line”or giving what an extraordinary
performance that shifts one’s view of oneself from a “punk” or even “cool/holding
your own” to “crazy” or “wild.” Guns were used by many as a resource for
improving performance. The abundance of guns in these neighborhoods increased the
severity of violent performances. For the majority of our sample, guns became
relevant for conflict resolution around the age of fourteen. The process of identity
formation, loss and recovery creates a dynamic of status forcing. Violence is
conflated with manhood, and manhood with status and identity. The result is a
constant churning of a finite reservoir of status that produces violent events, many
SOCIAL CONTAGION OF VIOLENCE
43
involving guns.
D. The Micro-Processes of Social Contagion
The dynamics of violent events reflect several processes that illustrate the
spread of violence within social structures that exert a contracting influence on social
networks of adolescents: (1) achieving a highly valued social identity occurs through
extreme displays of violence, (2) achieving a “safe” social identity may also require
the use of extreme forms of violence, (3) the ready availability of guns clearly
increases the stakes of how one achieves status, (4) much behavior is motivated by
avoiding being a punk or herb (sucker or weakling), (5) identities can change from
being a punk or herb into a more positive status such as “hold your own,” (6) guns
equalize the odds for some smaller young men through the process of “showing
nerve,” (7) one can feel like a punk for a specific situation but not take on a punk
identity, (8) one can feel like a “crazy” killer in a specific situation but not take on
a “crazy” or killer identity. If “compulsive masculinity” or Anderson’s “street
orientation” is dominant in public spaces and personal safety as our data suggest, then
those who do not conform will be victimized.
The reproduction of social identities constructed through violent behavior,
and the eclipsing or devaluation of other identities spreads in increasingly socially
isolated networks. These identities reinforce the dominance hierarchy built on
“toughness” and violence, and its salience devalues other identities. Those unwilling
to adopt at least some dimensions of these norms either overtly or symbolically are
vulnerable to physical attack. Accordingly, these identities are strategic necessities
to navigate through everyday dangers of inner city life.
The maintenance and reinforcement of identities supportive of violence is
made possible by an effective socio-cultural dynamic that sets forth an age grading
pathway to manhood that includes both behaviors and the means of resolving
SOCIAL CONTAGION OF VIOLENCE
44
violations of respect. The illustrations in this chapter show the strong influence of
street code, similar to the codes identified by Anderson (1994), or the code of honor
described by Toch (1969), over the behaviors of young children, adolescents, and
young adults. The absence of alternative means of attaining valued masculine
identities further compound the problem. The transmission of these social processes
occur on both the micro and macro level. Children growing up in this environment
learn these codes, or behavioral-affective systems, by navigating their way through
interpersonal situations which oftentimes involve violence encounters.
One effect of “danger” as a dominant ecological marker is the difficulty that
adolescents have in maintaining that duality of behavior and of orientation. The
street code has a functional purpose for attaining status and avoiding danger, even for
adolescents who harbor conventional attitudes and goals. Negotiating safety within
this context is extremely difficult especially when much of the social activity
available to young men often involves expressing dominance over others. The effects
are a hardening of street codes, and an eclipsing of other avenues for social status and
respect. Attempts to reverse this cycle must address its manifestations at both the
individual, group, and societal levels.
IV. SOCIAL CONTAGION AND SOCIAL NORMS
The dynamics of social contagion can be accommodated within concepts of
social influence and social norms. The social influence concept of behavior borrows
from both economics and sociology.41 Its economic component suggests that people
will act to maximize their utility, while its sociological dimension suggests that
41. Bernard E. Harcourt, Reflecting on the Subject: A Critique of the Social Influence
Conceptions of Deterrence, the Broken Windows Theory, and Order-Maintenance Policing New
York Style, 97 Michigan Law Review 291, 305 (1998).
SOCIAL CONTAGION OF VIOLENCE
45
conduct is shaped through direct and vicarious social interactions. A simple version
of this nexus suggests that the choice of conduct is influenced by observation and
practice of the most effective options. Choices are contextualized as well, reflecting
both the range of available options and the specific contingencies in which they are
applied.42
In the contagious dynamics of violence, the social meaning of violence is
constructed through the interrelationship of its action and its context.43 The social
meaning in this case involves actions (violence) that have both returns (identity,
status, avoidance of attack) and expectations that, within tightly packed networks, are
unquestioned or normative. Conduct impregnated with social meaning has influence
on the behaviors of others in immediate proximity. The social meaning of violence
influences the adaptation of behavioral norms, expected responses (scripts), and even
beliefs (memes) about systems of behavior. Social norms are the product of repeated
events that demonstrate the meaning and utility of specific forms of conduct. Social
influence thus has a dynamic and reciprocal effect on social norms.44 In poor
neighborhoods, social interactions are dominated by street codes, or local systems of
justice, that reward displays of physical domination and offer social approval for
antisocial behavior.45
The endogeneity of social contagion to networks and neighborhoods
illustrates the differences in the two types of epidemics. The origins of a contagious
epidemic that travels through a population become distal influences on the pathway
and dynamics of transmission through populations over time. The setting or context
of contagion reflects the susceptibility of populations to the transmission of a socially
meaningful behavior, and its exposure to the behavior that has acquired meaning.
42. Fagan, Context and Culpability in Adolescent Violence, at __.
43. Lessig, at 949.
44. Harcourt at 307; Lessig at 1040.
45. Fagan, Context and Culpability in Adolescent Violence, at __; Anderson, Violence and the
Inner City Code of the Streets (1999).
SOCIAL CONTAGION OF VIOLENCE
46
This can be true both for fashion and art,46 as well as for problematic social behaviors
such as drug use,47 teenage pregnancy,48 and violence.49
Recent applications of social influence models to crime control emphasize the
seminal role of exogenous influence of “disorder,” where minor crimes signal to
would-be criminals that crimes in that area will be tolerated and not reported.50 At
first glance, Broken Windows suggests that there is a spread of norms supporting
crime that overwhelm norms of orderliness. The spread comes from the continuing
signals from disorder. Apart from the problematic nature of this dichotomous
categorization of persons,51 Broken Windows medicalizes the conditions of disorder
and criminality. It assumes that exposure to the disorder is a constant and recurring
process that signals to the motivated offender that crime can succeed. Withdrawal
of the signs of disorder will change social norms by allowing the social influence of
orderliness to flourish. But this theory is limited by focusing only on the introduction
of the original cues or sources of crime, and rely on the causal effects of these
exogenous factors. This is analogous to the food poisoning model of epidemics.
The dynamics of social contagion instead suggest an endogenous process,
where the spread of social norms occurs through the everyday interactions of
individuals within networks that are structurally equivalent and closely packed.
Here, the ill grows and spreads from the inside, often long after the origins have
subsided. This is analogous to influenza contagion, or to the spread of cultural or
political thought.
46. Malcolm Gladwell, Cool Hunting, The New Yorker __(1997); James Servin, Cool Hunting
with Jane, Harpers Bazaar, February 1999 at 90.
47. Rowe and Rodgers
48. Jonathan Crane
49. Fagan, Context and Culpability; Anderson, Violence and the Inner City Code of the Streets
50. See, Wilson and Kelling, Broken Windows, The Atlantic Monthly, March 1982, at 29; George
Kelling and Susan Coles, Fixing Broken Windows (1986), analogizing to water supply
contamination as the source of the cholera epidemic in nineteenth century London. See, Edwin
Tufte, Visual Displays of Quantitative Information (1995).
SOCIAL CONTAGION OF VIOLENCE
47
The concept of contagion neutralizes the categorizations of disorder and order
that theoretically inform the new path of deterrence. A literal translation of contagion
would emphasize guns as a recurring source of violence, and as an agent in the
transmission of violence norms. Because the recent epidemic cycle of violence was
in reality a gun homicide epidemic, the case for gun-oriented policing52 strategies is
much stronger than practices based the more diffuse and unsupported theory of
disorder control and order-maintenance strategies.
While disorder embraces
orderliness, cleanliness, and sobriety,53 violence appears to travel on vectors quite
unrelated to that particular set of social norms.
25
V. REFERENCES
American Academy on Pediatrics (1993). Firearms and adolescents. Pediatrics 89(4): 78487.
Anderson, Elijah (1994). The Code of the Streets. The Atlantic Monthly, May, 81-94.
Berman, Alan L. (1995). Suicide Prevention: Clusters and Contagion.” Pp. 25-55 in Suicide
Prevention: Case Consultations, edited by Alan L. Berman. New York: Springer.
Bovasso, Gregory (1996). A network analysis of social contagion processes in an
organizational intervention. Human Relations 49 (11): 1419-35.
Brooks-Gunn, Jeanne, Gregory J. Duncan, P.K. Klebanov, & N. Sealand (1993). Do
neighborhoods influence child and adolescent development? American Journal of Sociology.
99: 353-95.
Burt, Ronald S. (1987). Social contagion and innovation: Cohesion versus structural
equivalence. American Journal of Sociology 92: 1287-1335.
Cavalli-Sproza, L.L., and M.W. Feldman (1981). Cultural Transmission and Evolution: A
Quantitative Approach. Princeton NJ: Princeton University Press.
51. Harcourt at 343.
52. Fagan, Zimring and Kim, 1998.
53. Harcourt at 345.
SOCIAL CONTAGION OF VIOLENCE
48
Cook, Philip J., and Kristin A. Goss (1996). A selective review of the social-contagion
literature. Mimeo. Durham, NC: Duke University.
Cook, Philip J., and John H. Laub (1998). The unprecedented epidemic in youth violence.
Forthcoming in Youth Violence – Crime and Justice: An Annual Review of Research, edited
by Michael Tonry and Mark H. Moore. Chicago: University of Chicago Press.
Coulton, Claudia J., Jill E. Korbin, M. Su, & J. Chow. (1995). Community-level factors and
child maltreatment rates. Child Development 66:1262-76.
Crane, Jonathan. 1991. The epidemic theory of ghettos and neighborhood effects on dropping
out and teenage childbearing. American Journal of Sociology 96(5): 1226-59.
Eckberg, Douglas L. (1995). Estimates of early twentieth-century U.S. homicide rate: An
econometric forecasting approach. Demography 32(1): 1-16.
Eckberg, Douglas L. (1995). What homicide epidemic? Long-term vital statistics trends.
Presented at the Annual Meeting of the American Association for the Advancement of
Science, Atlanta GA, February.
Eckland-Olsen, Sheldon (1982). Deviance, social control and social networks. Research on
Law, Deviance and Social Control 4: 271-99.
Fullilove, Mindy Thompson, Veronique Heon, Walkiria Jimenez, Caroline Parsons, Lesley
L. Green, and Robert E. Fullilove (1998). Injury and anomie: Effects of violence on an innercity community. American Journal of Public Health 88 (6): 924-927.
Fullilove, Robert E., Mindy Thompson Fullilove, Mary E. Northridge, Michael L. Ganz,
Mary T. Bassett, Diane E. McLean, Angela A. Aidala, Donald H. Gemson, and Colin
McCord (1999). Risk factors for excess mortality in Harlem: Findings from the Harlem
Household Survey. American Journal of Preventive Medicine 16(3S): 22-28.
Gladwell, Malcolm. 1996. The Tipping Point. The New Yorker, June 3, 1996, 32-38.
Gould, Madelyn S., Sylvan Wallenstein, and Marjorie Kleinman (1990). Time-Space
Clustering of Teenage Suicide. American Journal of Epidemiology 131(1): 71-78.
Gould, Madelyn S., Sylvan Wallenstein, Marjorie Kleinman, Patrick O’Carroll, and James
Mercy (1990). Suicide Clusters: An Examination of Age-Specific Effects. American Journal
of Public Health 80(2): 211-212.
Hunt, Leon Gibson, and Carl D. Chambers (1976). The Heroin Epidemics. New York:
Spectrum Publications.
Jones, Marshall, B. (1997). Behavioral contagion and official delinquency: Epidemic course
in adolescence. Social Biology 45 (1-2): 134-42.
SOCIAL CONTAGION OF VIOLENCE
49
Jones, D.R., and M.B. Jones (1995). Preferred pathways of behavioral contagion. Journal
of Psychiatric Research 29: 193-209.
Jones, D.R., and M.B. Jones (1994). Testing for behavioral contagion in a case-control
design. Journal of Psychiatric Research 28: 149-64.
Kennedy, David M., Anne M. Piehl, and Anthony A. Braga. “Youth Gun Violence in
Boston: Gun Markets, Serious Youth Offenders, and a Use Reduction Strategy.” Program
in Criminal Justice Policy and Management Working Paper, August 23, 1996.
Kennedy, David M. (1994). Can We Keep Guns Away From Kids? The American Prospect,
Summer, p. 74-80.
Loftin, Colin (1986). Assaultive violence as a contagious process. Bulletin of the New York
Academy of Medicine 62(5): 550-555.
Lynch, John W., George A. Kaplan, Elsie R. Pamuk, Richard D. Cohen, Katherine E. Heck,
Jennifer H. Balfour, and Irene H. Yen (1998). Income inequality and mortality in
metropolitan areas of the United States. American Journal of Public Health 88(7): 1074-80.
Massey, Douglas, & Nancy Denton (1993). American Apartheid. Cambridge MA: Harvard.
May, Robert M., Roy M. Anderson, Sally M. Blower (1990). The Epidemiology and
Transmission Dynamics of HIV-AIDS. In Living With AIDS, edited by Stephen R.
Graubard. Cambridge, Ma.: The MIT Press.
Messner, Steven , & Kenneth Tardiff (1986). Economic inequality and levels of homicide:
An analysis of urban neighborhoods. Criminology 24: 297-318.
Morenoff, Jeffrey D., and Robert J. Sampson. “Violent Crime and the Spatial Dynamics of
Neighborhood Transition: Chicago, 1970-1990.” Social Forces, forthcoming.
Morenoff, Jeffrey D., and Marta Tienda (1997). Underclass neighborhoods in temporal and
ecological perspective. The Annals of the American Association of Political and Social
Science 551: 59-72.
Murray, James D. “Modeling the Spread of Rabies.” American Scientist, Vol. 75, pp. 280284.
Peterson, Ruth D., & Krivo, Laurie J. (1993). Racial segregation and urban Black homicide.
Social Forces 71: 1001-26.
Rodgers, Joseph and David C. Rowe (1993). Social contagion and sexual behavior: a
developmental EMOSA model. Psychological Review 100: 479-510.
SOCIAL CONTAGION OF VIOLENCE
50
Roncek, D., & Maier, P.A. (1991). Bars, blocks, and crimes revisited: Linking the theory of
routine activities to the empiricism of hot spots. Criminology 29: 725-54.
Rose, H., & McClain, P. (1990). Race, Place, and Risk: Black Homicide in Urban America.
Albany NY: SUNY Press.
Rothman, Kenneth. 1986. Modern Epidemiology. Boston: Little Brown.
Rothstein, Edward (1998). Not quite a virus, but a contagion brain to brain. New York Times,
May 25, p. E2.
Rowe, David C. (1994). A social contagion model of adolescent sexual behavior: Explaining
race differences. Social Biology 41: 1-18.
Rowe, David C., Joseph L. Rodgers, and S. Meseck-Bushey (1994). An “epidemic” model
of sexual intercourse prevalences for black and white adolescents. Social Biology 36: 127-45.
Rowe, David C., and Joseph L. Rodgers (1991). Adolescent smoking and drinking: Are they
“epidemics”? Journal of Studies on Alcohol 52: 110-17.
Sampson, R.J. (1985). Race and criminal violence: A demographically disaggregated analysis
of urban homicide. Crime and Delinquency 31: 47-82.
Sampson, R.J. (1986). Crime in cities: The effects of formal and informal social control. Pp.
271-311 in Communities and Crime, edited by A.J. Reiss, Jr., & M.Tonry. Chicago:
University of Chicago Press.
Sampson, R.J. (1987). Urban Black violence: The effect of male joblessness and family
disruption. American Journal of Sociology 93: 348-82.
Sampson, Robert J. (1993). Linking time and place: Dynamic contexutalism and the future
of criminological inquiry. Journal of Research in Crime and Delinquency 30(4): 426-44.
Shihadeh, Edward S., & Darrell J. Steffensmeier (1994). Economic inequality, family
disruption, and urban Black violence: Cities as units of stratification and social control.
Social Forces 73: 729-751.
Swidler, Ann (1986). Culture in action: Symbols and strategies. American Sociological
Review 51: 273-86.
Tienda, Marta (1991). Poor people in poor places: Deciphering neighborhood effects on
behavioral outcomes. Pp. 244-262 in Macro-Micro Linkages in Sociology, edited by J.
Huber. Newbury Park, CA: Sage.
Wacquant, Loic D., & William J. Wilson (1989). The costs of racial and class exclusion in
the inner city. The Annals of the American Academy of Political and Social Science
SOCIAL CONTAGION OF VIOLENCE
51
501:8-25.
Wallace, Roderick (1991). Expanding coupled shock fronts of urban decay and criminal
behavior: How U.S. cities are becoming "hollowed out." Journal of Quantitative
Criminology 7: 333-356.
Wilkinson, Deanna L., and Jeffrey Fagan. (1996). Understanding the role of firearms in
violence ‘scripts’: The dynamics of gun events among adolescent males. Law and
Contemporary Problems 59 (1): 55-90.
Wilkinson, R.G. (1996). Unhealthy Societies: The Afflictions of Inequalities. London:
Routledge.
Williams, Kirk R., & Robert L. Flewelling (1987). Family, acquaintance, and stranger
homicide: alternative procedures for rate calculations. Criminology 25: 543-560.
Williams, Kirk R. & Robert L. Flewelling (1988). The social production of criminal
homicide: a comparative study of disaggregated rates in american cities. American
Sociological Review 53: 421-431.
Wilson, William J. (1987). The Truly Disadvantaged. Chicago: University of Chicago Press.
Wilson, William J. (1991). Studying inner-city social dislocations: The challenge of public
agenda research. American Sociological Review 56: 1-14.
SOCIAL CONTAGION OF VIOLENCE
52
Table 1. Gun and Non-Gun Homicide in New York City, by Sex, 1985-1995a
Gun
Non-Gun
Male
Year
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
N
747
846
980
1220
1282
1470
1526
1457
1387
1065
722
Female
Rate
21.8
24.7
28.6
35.6
37.4
42.9
44.5
42.5
40.4
31.0
21.0
N
87
102
120
111
107
127
146
138
142
95
96
Male
Rate
2.2
2.6
3.1
2.9
2.7
3.3
3.8
3.5
3.6
2.4
2.5
N
493
491
417
416
376
474
350
302
309
290
223
Female
Rate
14.4
14.3
12.2
12.1
11.0
13.8
10.2
8.8
9.0
8.5
6.5
N
182
183
166
184
190
183
170
138
157
142
146
Rate
4.7
4.7
4.3
4.7
4.9
4.7
4.4
3.5
4.0
3.6
3.8
a. Rate per 100,000 persons within each population group
Table 2. Gun and Non-Gun Homicide in New York City, by Race, 1985-1995a
Gun
White
Year
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
N
386
416
377
506
580
722
742
738
659
492
365
Rate
10.1
10.9
9.8
13.2
15.1
18.8
19.4
19.3
17.2
12.8
9.5
Non-Gun
Nonwhite
N
Rate
448
12.8
532
15.2
723
20.7
825
23.6
809
23.2
875
25.1
930
26.6
849
24.3
856
24.5
658
18.9
446
12.8
a. Rate per 100,000 persons within each population group
White
N
306
278
232
246
257
342
246
229
212
219
171
Rate
8.0
7.3
6.1
6.4
6.7
8.9
6.4
6.0
5.5
5.7
4.5
Nonwhite
N
Rate
369
10.6
396
11.3
351
10.1
354
10.1
309
8.9
315
9.0
274
7.8
209
6.0
251
7.2
201
5.8
191
5.5
SOCIAL CONTAGION OF VIOLENCE
53
Table 3. Adolescent Gun and Non-Gun Homicide in New York City, by Age, 1985-1995a
Year
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
Under 14 Years
N
Rate
8
0.6
15
1.1
17
1.2
21
1.5
19
1.3
31
2.2
14
1.0
18
1.3
22
1.6
10
0.7
12
0.8
15-19 Years
N
Rate
96
20.8
92
19.9
153
33.1
210
45.5
199
43.1
235
50.9
265
57.4
223
48.3
224
48.5
183
39.6
120
26.0
Gun
20-24 Years
N
Rate
191
34.1
247
44.1
283
50.6
318
56.8
336
60.0
371
66.3
371
66.3
381
68.1
357
63.8
266
47.5
194
34.7
25-34 Years
N
Rate
290
21.1
351
25.6
347
25.3
460
33.5
520
37.9
579
42.2
579
42.2
587
42.8
514
37.5
411
30.0
248
18.1
Over 35 Years
N
Rate
249
7.1
243
6.9
300
8.5
322
9.2
315
9.0
381
10.8
443
12.6
386
11.0
412
11.7
290
8.3
244
6.9
25-34 Years
N
Rate
167
12.2
194
14.1
174
12.7
169
12.3
184
13.4
202
14.7
150
10.9
108
7.9
122
8.9
121
8.8
90
6.6
Over 35 Years
N
Rate
331
9.4
296
8.4
275
7.8
278
7.9
246
7.0
261
7.4
227
6.5
218
6.2
229
6.5
200
5.7
180
5.1
Non-Gun
Year
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
Under 14 Years
N
Rate
41
2.9
48
3.4
26
1.8
43
3.0
37
2.6
48
3.4
44
3.1
34
2.4
58
4.1
44
3.1
37
2.6
15-19 Years
N
Rate
48
10.4
37
8.0
29
6.3
28
6.1
30
6.5
59
12.8
37
8.0
32
6.9
18
3.9
28
6.1
24
5.2
a. Rate per 100,000 persons within each age group
20-24 Years
N
Rate
88
15.7
99
17.7
79
14.1
82
14.6
69
12.3
87
15.5
62
11.1
48
8.6
39
7.0
39
7.0
38
6.8
SOCIAL CONTAGION OF VIOLENCE
54
Table 4. Age Homogeneity of Gun and Non-Gun Homicides, 1985-95a
Gun
Non - Gun
Victim and Offender Age
Victim and Offender Age
Year
1985
1986
15 to 19
20 to 24
25 to 34
15 to 19
20 to 24
25 to 34
26.5
22.9
38.5
25.4
23.0
38.5
25.0
27.1
40.7
15.5
22.5
28.7
1987
1988
1989
1990
1991
1992
1993
1994
1995
28.5
26.8
43.5
17.9
20.9
45.3
37.7
28.6
43.5
9.3
13.0
41.5
33.3
29.7
54.9
11.6
13.9
41.6
26.4
35.8
44.9
19.7
23.2
31.0
34.2
27.0
38.2
29.7
17.1
35.4
34.4
29.4
46.7
24.0
19.0
32.1
35.8
27.5
41.3
18.2
10.0
41.5
37.2
30.7
41.7
19.2
17.6
32.7
15.5
30.9
42.4
18.8
24.2
26.1
a. Percent of homicides within age group. Includes only cases where victim and offender ages are both known. Annual rates adjusted
for missing months of data. Source: Supplemental Homicide Reports, 1976-94; New York City Police Department, 1995.
SOCIAL CONTAGION OF VIOLENCE
55
_______________________________________________________________________________________
Table 5. Factor Analysis of 1990 STF3A Census Tract Variables (N=2,803)
________________________________________________________________________________________
Final Factors
Composition
Eigenvalue
% Variance
Explained
Deprivation
1. Household Gini Index
2. Median Household Income
3. Households Below Poverty
4. Education Ratio
5. Labor Force Participation Ratio
6. Unemployment Rate
7. Skilled Jobs Ratio
6.30
45.0
Population Structure
1. Proportion Nonwhite
2. Separated and Divorced
3. Never Married
4. Female Headed Households
1.52
10.8
Social Control
1. Supervision Ratio
1.09
7.8
2. Residential Mobility
3. Population
________________________________________________________________________________________
________________________________________________________________________________________
Table 6. Repeated Measures Analysis of Adolescent Homicide Rates by Census Tract,
New York City, 1985-1995
________________________________________________________________________________________
t-value
Deprivation
8.51 **
Population
9.90 **
Social Control
1.59
________________________________________________________________________________________
* p < .05
** p < .001
_________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
56
________________________________________________________________________________________
Table 7. Contagion of Adolescent Homicide from in New York City, 1985-1995
________________________________________________________________________________________
A. Any Homicide (Lag = 1 Year)
Outward Contagion
Model 1
t-value
At Least One Adolescent
Homicide in Neighborhood
13.58 **
Model 2
t-value
Model 3
t-value
Model 4
t-value
24.77 **
At Least One Adolescent
Homicide in Surrounding Community
Deprivation
Population Structure
Social Control
Inward Contagion
4.66**
7.31 **
29.53 **
50.40 **
2.82 *
20.76 **
27.23 **
7.56 **
________________________________________________________________________________________
B. Homicide Rate (Lag = 1 Year)
Outward Contagion
Model 1
t-value
Adolescent Homicide Rate
in Neighborhood
12.43 **
Model 2
t-value
Model 3
t-value
Model 4
t-value
5.56 **
Adolescent Homicide Rate
in Surrounding Community
Deprivation
Population Structure
Social Control
Inward Contagion
18.55 **
27.05 **
53.11 **
-4.78 **
5.48 **
14.35 **
25.23 **
-2.71 *
________________________________________________________________________________________
* p < .05
** p < .001
________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
57
________________________________________________________________________________________
Table 8. Outward Contagion of Adolescent Gun Homicide from Neighborhood to Surrounding
Community, New York City, 1985-1995
________________________________________________________________________________________
A. Any Homicide (Lag = 1 Year)
Gun Homicide in
Surrounding Community
Model 1
t-value
At Least One Adolescent
Gun Homicide
in Neighborhood
Deprivation
Population Structure
Social Control
14.96 **
Model 2
t-value
4.26 **
29.12 **
50.21 **
0.37
Non-Gun Homicide in
Surrounding Community
Model 3
t-value
7.42 **
Model 4
t-value
1.39
13.01 **
22.50 **
4.92 **
________________________________________________________________________________________
B. Gun Homicide Rate (Lag = 1 Year)
Gun Homicide in
Surrounding Community
Model 5
t-value
Model 6
t-value
Non-Gun Homicide in
Surrounding Community
Model 7
t-value
Model 8
t-value
Adolescent Gun
Homicide Rate
in Neighborhood
11.53 **
5.38 **
8.58 **
1.39
Deprivation
26.10 **
16.89 **
Population Structure
59.82 **
23.95 **
Social Control
-8.48 **
4.15 **
________________________________________________________________________________________
* p < .05
** p < .001
________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
58
________________________________________________________________________________________
Table 9. Inward Contagion of Adolescent Gun Homicide in New York City, 1985-1995
________________________________________________________________________________________
A. Any Homicide (Lag = 1 Year)
Neighborhood
Gun Homicide
Model 1
t-value
At Least One Adolescent
Gun Homicide in
Surrounding Community
Neighborhood
Non-Gun Homicide
Model 2
t-value
27.20 **
9.33 **
Deprivation
Population Structure
Social Control
Model 3
t-value
Model 4
t-value
2.20 *
20.63 **
29.37 **
6.72 **
1.01
3.80 **
-.028
.05
________________________________________________________________________________________
B. Homicide Rate (Lag = 1 Year)
Neighborhood
Gun Homicide
Model 5
t-value
Adolescent Gun Homicide Rate
in Surrounding Community
22.53 **
Deprivation
Population Structure
Social Control
Neighborhood
Non-Gun Homicide
Model 6
t-value
7.73 **
14.23 **
26.71 **
-4.30 **
Model 7
t-value
2.20 *
Model 8
t-value
1.05
3.70 **
-.028
.07
________________________________________________________________________________________
* p < .05
** p < .001
________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
59
________________________________________________________________________________________
Table 10. Outward Contagion of Adolescent Non-Gun Homicide in New York City, 1985-1995
________________________________________________________________________________________
A. Any Homicide (Lag = 1 Year)
Non-Gun Homicide
in Surrounding Community
At Least One Adolescent
Non-Gun Homicide
in Neighborhood
Gun Homicide
in Surrounding Community
Model 1
t-value
Model 2
t-value
Model 3
t-value
Model 4
t-value
.79
- .59
.70
.35
Deprivation
Population Structure
Social Control
18.73 **
31.14 **
12.39 **
11.92 **
22.78 **
-3.91 **
________________________________________________________________________________________
B. Homicide Rate (Lag = 1 Year)
Non-Gun Homicide
in Surrounding Community
Model 5
t-value
Adolescent Non-Gun
Homicide Rate
in Neighborhood
Deprivation
Population Structure
Social Control
- .14
Model 6
t-value
- .24
17.08 **
24.79 **
4.11 **
Gun Homicide
in Surrounding Community
Model 7
t-value
.74
Model 8
t-value
.59
26.44 **
61.59 **
-8.58 **
________________________________________________________________________________________
* p < .05
** p < .001
________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
60
________________________________________________________________________________________
Table 11. Inward Contagion of Adolescent Non-Gun Homicide in New York City, 1985-1995
_______________________________________________________________________________________
A. Any Homicide (Lag = 1 Year)
Non-Gun Homicide
in Neighborhood
Model 1
t-value
At Least One Adolescent
Non-Gun Homicide
in Surrounding Community
Deprivation
Population Structure
Social Control
2.55 *
Model 2
t-value
.69
4.67 **
.55
-1.43
Gun Homicide
in Neighborhood
Model 3
t-value
Model 4
t-value
-1.15
.46
8.50 **
15.58 **
-3.26 **
________________________________________________________________________________________
B. Homicide Rate (Lag = 1 Year)
Non-Gun Homicide
in Neighborhood
Model 5
t-value
Adolescent Non-Gun
Homicide Rate
in Surrounding Community
Deprivation
Population Structure
Social Control
2.27 *
Model 6
t-value
2.31 *
3.73 **
- .26
- .04
Gun Homicide
in Neighborhood
Model 7
t-value
8.65 **
Model 8
t-value
.98
15.89 **
31.81 **
-4.75 **
________________________________________________________________________________________
* p < .05
** p < .001
________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
61
________________________________________________________________________________________
Table 12. Contagion of Adolescent Gun to Total Homicide in New York City, 1985-1995
________________________________________________________________________________________
A. Logistic Repeated Measures Models (Lag = 1 Year)
Outward Contagion
Model 1
t-value
At Least One Adolescent
Gun Homicide
in Neighborhood
15.66 **
Model 2
t-value
Inward Contagion
Model 3
t-value
Model 4
t-value
4.73 **
At Least One Adolescent
Gun Homicide
in Community
25.84 **
Poverty Factor
Family/Ethnicity Factor
Anonymity Factor
8.34 **
29.48 **
49.91 **
2.84 **
20.51 **
26.45 **
7.72 **
________________________________________________________________________________________
B. Standard Repeated Measures Models (Lag = 1 Year)
Outward Contagion
Model 1
t-value
Adolescent Gun Homicide
Rate in Neighborhood
Adolescent Gun Homicide
Rate in Community
14.23 **
Model 2
t-value
Inward Contagion
Model 3
t-value
Model 4
t-value
5.81 **
20.76 **
6.38 **
Poverty Factor
27.03 **
14.06 **
Family/Ethnicity Factor
52.62 **
24.13 **
Anonymity Factor
-4.72 **
-2.53 *
________________________________________________________________________________________
* p < .05
** p < .001
SOCIAL CONTAGION OF VIOLENCE
62
__________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
63
__________________________________________________________________________________________
Table 13. Contagion of Adolescent Non-Gun to Total Homicide in New York City, 1985-1995
________________________________________________________________________________________
A. Logistic Repeated Measures Models (Lag = 1 Year)
Outward Contagion
At Least One Adolescent
Non-Gun Homicide
in Neighborhood
Model 1
t-value
Model 2
t-value
-0.09
-1.04
At Least One Adolescent
Non-Gun Homicide
in Community
Poverty Factor
Family/Ethnicity Factor
Anonymity Factor
Inward Contagion
Model 3
t-value
Model 4
t-value
12.84 **
3.19
21.82
30.17
7.34
30.15 **
52.14 **
2.88 **
**
**
**
**
________________________________________________________________________________________
B. Standard Repeated Measures Models (Lag = 1 Year)
Outward Contagion
Adolescent Non-Gun
Homicide Rate
in Neighborhood
Adolescent Non-Gun
Homicide Rate
in Community
Poverty Factor
Family/Ethnicity Factor
Anonymity Factor
Model 1
t-value
Model 2
t-value
-1.31
-0.52
Inward Contagion
Model 3
t-value
9.41 **
27.49 **
54.65 **
-4.87 **
Model 4
t-value
2.02
15.41
28.52
-2.95
*
**
**
*
________________________________________________________________________________________
* p < .05
SOCIAL CONTAGION OF VIOLENCE
64
** p < .001
Appendix A. New York City Mortality Data, 1968-1996
Counts
Rates (Base = 1985)
Year
Gun
Non-Gun
Total
Gun
Non-Gun
Total
1968
424
537
961
52.02
96.41
70.04
1969
490
521
1011
60.12
93.54
73.69
1970
628
513
1141
77.06
92.10
83.16
1971
806
701
1507
98.90
125.85
109.84
1972
936
756
1692
114.85
135.73
123.32
1973
889
791
1680
109.08
142.01
122.45
1974
811
725
1536
99.51
130.16
111.95
1975
886
707
1593
108.71
126.93
116.11
1976
846
742
1588
103.80
133.21
115.74
1977
918
636
1554
112.64
114.18
113.27
1978
891
660
1551
109.33
118.49
113.05
1979
1039
722
1761
127.48
129.62
128.35
1980
1109
735
1844
136.07
131.96
134.40
1981
1187
673
1860
145.64
120.83
135.57
1982
1056
671
1727
129.57
120.47
125.87
1983
1004
639
1643
123.19
114.72
119.75
1984
884
587
1471
108.47
105.39
107.22
1985
815
557
1372
100
100
100
1986
929
613
1542
113.99
110.05
112.39
1987
1085
565
1650
133.13
101.44
120.26
1988
1310
586
1896
160.74
105.21
138.19
1989
1378
513
1891
169.08
92.10
137.83
1990
1563
616
2179
191.78
110.59
158.82
1991
1644
519
2163
201.72
93.18
157.65
1992
1554
441
1995
190.67
79.17
145.41
1993
1488
468
1956
182.58
84.02
142.57
1994
1160
433
1593
142.33
77.74
116.11
1995
818
369
1187
100.37
66.25
86.52
1996
652
335
987
70.18
54.65
64.01
__________________________________________________________________________________________
SOCIAL CONTAGION OF VIOLENCE
65
Appendix B. Means and Standard Deviations for Neighborhood Characteristics (N = 2083 Census Tracts)
Variable
1. Population
2. Separated and Divorced
3. Never Married
4. Racial Composition
5. Residential Mobility
6. Poverty
7. Female Headed Households
8. Unemployment
9. Income
10. Income Equality
11. Supervision Ratio
12. Labor Force Participation Ratio
13. Education Ratio
Operational Definition
Means
Standard
Deviation
Valid
Cases
3474
11.66
35.55
55.06
63.49
18.11
10.16
9.68
$ 31,419
.38
284.90
59.62
2408
5.57
9.48
36.15
10.05
14.54
9.88
6.38
$14,287
.08
154.77
88.40
2083
2083
2083
2083
2083
2078
2078
2081
2083
2078
2082
2081
% Individuals over 15 who are separated or divorced
% Individuals over 15 who have never been married
% Individuals Nonwhite
% Households that have not moved in the past 5 years
% Households below poverty
% Households with children under 18 headed by female
Unemployment Rate
Median Household Income
Household Gini Index
Proportion individuals aged 25-64 to those aged 5-24
Proportion individuals in labor force to those not in labor force
Proportion individuals with some college education to those
without completed high school education
132.34
151.60
2081
14. Skilled Jobs Ratio
Proportion individuals with professional or managerial jobs to
those with non-skilled jobs
18.06
23.26
2072
_________________________________________________________________________________________________________________
19
68
19
69
19
70
19
71
19
72
19
73
19
74
19
75
19
76
19
77
19
78
19
79
19
80
19
81
19
82
19
83
19
84
19
85
19
86
19
87
19
88
19
89
19
90
19
91
19
92
19
93
19
94
19
95
19
96
19
97
19
98
1985 Base = 100
SOCIAL CONTAGION OF VIOLENCE
225
200
67
Figure 1 GUN AND NON-GUN HOMICIDE, 1968-1998
GUN
NONGUN
TOTAL
175
150
125
100
75
50
25
Year
Year
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
98
97
96
95
94
93
92
91
90
89
88
87
86
85
84
83
82
81
80
79
78
77
76
75
74
73
72
71
70
69
68
2000
19
19
19
19
19
19
19
Counts
SOCIAL CONTAGION OF VIOLENCE
68
Figure 2 GUN AND NON-GUN HOMICIDES, 1968-1998
2500
GUN
NONGUN
TOTAL
1500
1000
500
SOCIAL CONTAGION OF VIOLENCE
69
Figure 3. DRUG OVERDOSE DEATH AND AGE-SPECIFIC
GUN HOMICIDE VICTIMIZATION RATES, 1985-95
80
Drug Overdose Deaths
70
Gun Homicides 15 - 19
Gun Homicides 20 - 24
Age-Specific Rate per 100,000
60
Gun Homicides 25 - 34
50
40
30
20
10
0
1985
1986
1987
1988
1989
1990
Year
1991
1992
1993
1994
1995
SOCIAL CONTAGION OF VIOLENCE
70
SOCIAL CONTAGION OF VIOLENCE
71
Figure 4a. AGE-SPECIFIC FELONY DRUG ARREST RATES AND GUN
HOMICIDE VICTIMIZATION RATES, 1985-95
250
Gun Homicides 15 - 19
Gun Homicides 20 - 24
Gun Homicides 25 - 34
Age-Specific Rate per 100,000
200
Drug Arrests 10-19
Drug Arrests 20 - 24
Drug Arrests 25 - 34
150
100
50
0
1985
1986
1987
1988
1989
1990
Year
1991
1992
1993
1994
1995