Fingerhut LA, Ingram DD, Feldman JJ. Homicide Rates Among US Teenagers and Young AdultsDifferences by Mechanism, Level of Urbanization, Race, and Sex,1987 Through 1995. JAMA. 1998;280(5):423-427. doi:10.1001/jama.280.5.423
From the Office of Analysis, Epidemiology, and Health Promotion, National Center for Health Statistics, Hyattsville, Md. Dr Feldman is retired.
Context Homicide rates for persons 15 through 24 years old began to decline
between 1993 and 1994, but recent trends in homicide rates by mechanism of
homicide and urbanization group have not been described.
Objective To examine homicide trends from 1987 through 1995 for persons 15 through
24 years old by urbanization level.
Design Homicide rates by urbanization level were analyzed using the Compressed
Mortality File, a county-level mortality and population database maintained
by the National Center for Health Statistics, Centers for Disease Control
and Prevention, and the rural-urban continuum codes developed by the Economic
Research Service, US Department of Agriculture.
Setting United States, 1987 through 1995, according to 5 urbanization strata:
core, counties with the primary central city of a metropolitan statistical
area (MSA) of 1 million or more; fringe, remaining counties within an MSA
of 1 million or more; medium, counties within an MSA of 250,000 to 999,999;
small, counties in an MSA of less than 250,000; and nonmetropolitan, counties
not in an MSA.
Subjects All persons 15 through 24 years old by race whose cause of death was
homicide (International Classification of Diseases, Ninth
Revision codes E960-E969).
Main Outcome Measures Firearm and nonfirearm homicide rates and average annual percentage
changes by 5 urbanization levels, race, and sex.
Results From 1987 through 1991, the average annual firearm homicide rates among
persons 15 through 24 years old among all 5 urbanization strata increased
between 10.7% in small counties and 19.8% in fringe counties. From 1991 through
1993, the rates increased between 3.3% in core counties and 11.7% in small
counties. From 1993 through 1995, the rates declined between 4.4% in fringe
counties and 15.3% in medium counties. By 1995, firearm homicide rates among
persons 15 through 24 years old ranged from 6.5 and 7.3 per 100,000 in the
nonmetropolitan and small counties, respectively, to 9.6 and 13.3 per 100,000
in the fringe and medium strata, respectively, to 33.5 per 100,000 in the
core stratum. During 1987 through 1990, nonfirearm homicide rates either were
stable or increased, and from 1990 through 1995, nonfirearm homicide rates
declined in all 5 strata, on average 3.7% to 8.0% per year, with rates in
1995 ranging from 2.1 to 4.7 per 100,000 across the strata.
Conclusions After increasing since 1987, firearm and nonfirearm homicide rates began
declining between 1993 and 1995 among persons 15 through 24 years old. These
declines are taking place across all urbanization strata and among white and
black males and females.
HOMICIDE RATES for persons 15 through 24 years old in the United States
increased between the mid-1980s and the early 1990s. Since 1990, the homicide
rates for persons 15 through 24 years old have been higher than those for
persons who are younger or older.1 As for most
other age groups, homicide rates for these young people began to decline between
1993 and 1994. The rate in 1994 was 3% lower than the rate in 1993, and the
rate of 20.0 deaths per 100,000 population in 1995 was 10% lower than in 1994.
Provisional data2 indicate that the national
decline continued in 1996 and 1997. Despite the declines, the homicide rate
for persons 15 through 24 years old remains high; the rate in 1995 was 71%,
higher than a decade earlier.
In 1995, homicides accounted for 21% of all deaths among persons 15
through 24 years old compared with 12% of all deaths in 1985. Firearms continue
to be the mechanism with which most homicides are committed, with the proportion
of homicides resulting from the use of firearms increasing from 66% in 1985
to 84% of all homicides in 1995.3
In an earlier analysis of homicide data, we reported urbanization level
differences and trends from 1979 through 1989 in firearm and nonfirearm homicides
among white and black males and females 15 through 19 years old.4
In this analysis, we focus on trends and differences by urbanization from
1987 through 1995 for persons 15 through 24 years old to further our understanding
of the changing trends in homicide.
This analysis of homicide among persons 15 through 24 years old examines
trends from 1987 through 1995. The analysis begins with 1987—the year
marking the beginning of a period of rapid increases in firearm homicide rates.4Figure 1
shows homicide rates for the longer period, 1979 through 1995, so that the
recent trends can be viewed in relation to the longer-term trends.
The data for this analysis are from the Compressed Mortality File, a
county-level national mortality and population database maintained by the
Office of Analysis, Epidemiology, and Health Promotion, National Center for
Health Statistics (NCHS), Centers for Disease Control and Prevention, Hyattsville,
Md. The Compressed Mortality File has age-specific, race-specific, and sex-specific
data on underlying cause of death by decedent's county of residence compiled
from all death certificates that are a part of the NCHS National Vital Statistics
System. Thus, all homicides in the United States are included in this analysis.
Cause-of-death information is coded in accordance with the International Classification of Diseases, Ninth Revision (ICD-9).5 The population estimates were
provided by the Bureau of the Census.
The race-specific analyses are limited to white and black persons because
age-specific county population estimates for Asian and American Indian persons
are only available for census years. Similarly, data for Hispanic persons
could not be analyzed separately because of the lack of age-specific county
population estimates prior to 1990. Thus, white and black persons include
those persons of Hispanic origin. Race-specific and ethnic-specific national
homicide rates show that Hispanic homicide rates for males are about half
the rate for black males and about 8 times the rate for non-Hispanic white
males.1 Consistency of race identification
between the death certificate (numerator of death rate) and from the census
estimates of race (denominator of death rate) is high for white and black
Homicides are defined in this analysis by the ICD-9 E codes, E960 through E969. This is consistent with the current standard,
as proposed by the 1997 recommendations of the Injury Control Community.7 In the earlier analysis, homicide included legal intervention, ICD-9 codes E970 through E978. Deaths from legal intervention
between 1979 and 1995 among persons 15 through 24 years old have been as few
as 50 per year (in 1988) and as many as 128 (in 1979); firearms (E970) were
the mechanism used in 95% to 100% of the deaths. Firearm homicide includes ICD-9 codes E965.0 through E965.4; nonfirearm homicide
includes ICD-9 codes E960 through E964 and E965.5
For this analysis, counties are categorized into 5 urbanization groups:
(1) core metropolitan—counties containing the
primary central city of a metropolitan statistical area (MSA) with a population
of 1 million or more; (2) fringe metropolitan—remaining
counties in an MSA with a population of 1 million or more; (3) medium metropolitan—counties in an MSA with a population of 250,000
to 999,999; (4) small metropolitan—counties
in an MSA with a population of less than 250,000; and (5) nonmetropolitan—counties not in an MSA.8
Classification of counties into these 5 strata by the Economic Research Service,
US Department of Agriculture, Washington, DC, is based on a modification of
the 1993 urban-rural continuum codes (available from the authors). Note that
the population size given for each urbanization stratum refers to the population
of the entire MSA, not the population of the individual counties that constitute
the MSA. The individual counties usually have much smaller populations. For
example, core and fringe counties often have populations of less than 1 million.
In this analysis, total and sex-specific firearm and nonfirearm homicide
rates for white and black persons 15 through 24 years old were calculated
for the 5 urbanization strata. Unless otherwise stated, all comparisons between
death rates made in this article are statistically significant at P≤.05. Random variation in death rates is addressed fully in the
technical notes of reports of mortality statistics from the NCHS.9
Visual examination and change-point analyses of the homicide rates indicated
that 3 trend lines should be fit to the firearm homicide rates, 1987 through
1991, 1991 through 1993, and 1993 through 1995, and 2 trend lines should be
fit to the nonfirearm homicide rates, 1987 through 1990 and 1990 through 1995.
A few exceptions to these trend lines included the following: (1) For white
females in the fringe stratum and black females in the medium metropolitan
stratum, only 1 trend line was fit to the firearm homicide rates from 1987
through 1993. (2) For all the race-sex groups in the medium metropolitan stratum,
except black females, the second firearm homicide trend line was fit to 1987
through 1994, and the percentage decline between 1994 and 1995 was calculated.
(3) For black females in the nonmetropolitan stratum, trend lines for nonfirearm
homicide were fit for 1987 through 1991 and 1991 through 1995. (4) Finally,
for some of the race-sex-urbanization strata, trend lines could not be fit
because of small numbers of deaths. Weighted least squares regression was
used to fit a line to the natural logarithm of the death rates for each period.
The weights used were the number of deaths. A χ2 test was used
to determine whether the slopes were significantly different from zero. The
average annual percentage change in the death rates for a given period is
100(eb−1), where b is the slope estimate of the regression line for the particular period.
In the trend figures, death rates are plotted on a logarithmic scale
rather than on a linear scale because the logarithmic scale emphasizes relative
rather than absolute change. For a given rate change, eg, from 2 to 3 deaths
per 100,000 population or 20 to 30 deaths per 100,000 population, the percentage
change is the same, 50%.
From 1987 through 1991, firearm homicide rates in the 5 urbanization
strata among persons 15 through 24 years old increased an average of 10.7%
to 19.8% per year; from 1991 through 1993, the rates increased at 3.3% to
11.7% per year (in the small metropolitan stratum, the average rates of firearm
homicides in the 2 periods were similar); and from 1993 through 1995, they
declined an average of 4.4% to 15.3% per year. The declines during 1993 through
1995, the first declines observed since the early 1980s, returned the firearm
homicide rate to the level observed in 1990 through 1991 (Table 1 and Figure 1).
Trends in nonfirearm homicide rates have been different than the trends
in firearm homicide. From 1987 through 1990, nonfirearm homicide rates were
stable in the core, medium, and small metropolitan strata, and they increased
in the fringe and nonmetropolitan strata. From 1990 through 1995, nonfirearm
homicide rates declined in all 5 urbanization strata, on average 3.7% to 8.0%
Firearm Homicide. In the core metropolitan county stratum, firearm homicide rates for
all 4 race-sex groups increased an average of 9.2% to 20.9% per year from
1987 through 1991 (Figure 2). They
were generally stable from 1991 through 1993 and, from 1993 through 1995,
firearm homicide rates for black males and females decreased an average of
12.0% and 16.2% per year (Table 1).
In the fringe stratum, trends in the firearm homicide rates differed
only slightly from those observed in the core stratum (Figure 3). Rates for black and white males increased during 1987
through 1991, followed by slower annual increases during 1991 through 1993.
Rates for white females increased steadily during 1987 through 1993. During
1993 through 1995, the only significant decline in firearm homicide was observed
for black males.
In medium metropolitan counties, firearm homicide rates among black
males increased from 1987 through 1991 an average of 26.1% per year and among
white males and females an average of 10.4% and 11.5% per year (Table 1). From 1991 through 1994, firearm homicide rates for black
males and white females were stable and the rates for white males increased
at about the same pace as earlier. From 1987 through 1993, the rate among
black females increased an average of 13.9% per year. From 1994 through 1995,
the firearm homicide rate for white females and for black males and white
males declined an average of 12.5% to 17.8% per year (Figure 4). In the small metropolitan counties, most of the race-sex
specific firearm homicide rates did not change significantly.
In nonmetropolitan counties, firearm homicide rates among black males
increased at average annual rates of 24.5% and 11.2% during 1987 through 1991
and 1991 through 1993, respectively, followed by declines of about 16.5% per
year during 1993 through 1995. The firearm homicide rates for the other race-sex
groups were relatively stable during 1987 through 1995 (Table 1 and Figure 5).
Nonfirearm Homicide. From 1987 through 1990, the race-sex specific nonfirearm homicide rates
were generally stable in each of the urbanization strata. However, rates for
white males in the core and fringe strata increased an average of 10.0% to
16.1% per year and rates for black females in the core stratum decreased an
average of 8.9% per year. From 1990 through 1995, nonfirearm homicide rates
in the core stratum decreased for all race-sex groups. In the fringe stratum,
rates also declined for black males and white males, and, in the medium stratum,
rates declined for black males and females. The only decline in the nonmetropolitan
counties was observed for white males.
Among black males 15 through 24 years old, in 1995, firearms were the
mechanism in 92% of all homicides (88%-93% across the 5 urbanization strata).
The black male firearm homicide rate of 174.5 deaths per 100,000 population
in the core metropolitan counties was about 2 to 4 times the rate of the other
strata (Table 2).
For white males, 82% of all homicides were caused by firearms (72%-87%
across the strata). The firearm homicide rate for white males of 29.8 deaths
per 100,000 population in the core stratum was about 3 to 4 times the rates
of the fringe and medium strata, and about 6 times the rates of the small
county and nonmetropolitan strata (higher relative differences than in the
rates for black males).
For black and white females 15 through 24 years old, the proportions
of homicides caused by firearms were lower than for males, 70% and 56% for
black and white females, respectively. Among black females, the core county
firearm homicide rate of 14.2 deaths per 100,000 population was 1.6 times
the rates of the fringe and nonmetropolitan strata and was similar to the
rate in the medium stratum. Among white females, the firearm homicide rate
in the core stratum of 3.5 per 100,000 population was 2.5 times the rates
of the fringe stratum and 1.6 to 1.8 times the rates of the medium and nonmetropolitan
strata (Table 2).
The nonfirearm homicide rates for black and white males in the core
metropolitan stratum were about 1.5 to 3 times the rates of the other strata.
Among females of both races, nonfirearm homicide rates were similar across
each of the strata.
Homicide rates have recently begun declining among persons 15 through
24 years old, the age group where rates have been the highest and in which
the most rapid increase in rates from 1987 through 1991 was experienced. These
declines, which are occurring for both firearm and nonfirearm homicides, are
taking place in both metropolitan and nonmetropolitan counties and among white
and black males and females. The most rapid declines have been among black
females and among black males—the group that has always had the highest
homicide rates. Provisional data for 1996 and 1997 indicate that the national
declines in homicide are continuing.
To extend these analyses, the homicide trends should be examined at
smaller geographic levels, ie, at the regional, state, or individual county
level. Examination of the homicide trends at various geographic levels may
aid in understanding the nature of the recent national declines. Preliminary
analyses of homicide trends in individual counties reveal interesting patterns.
For example, within the core stratum, the downturn apparently began in 1993
in some large, urban counties and in 1994 in others. Despite the extremely
sharp declines that many core counties had shown by 1995, there were a number
of other core counties showing no decline at all. Will all the core counties
eventually experience the decline? It is unclear what factors have to come
together to produce the downturn. The fact that each of the classes of counties
examined in the present article showed appreciable aggregate declines between
1993 and 1995 suggests that the responsible forces are not unique to a few
localities. Given normal leads and lags between areas, the downturn might
very well have by now, in 1998, spread to many of the individual geographic
areas that had not shown any sign of a decline as of 1995. (Final data for
1996 and 1997 were not available for analysis at the time of writing.) However,
as we had been unaware that we had reached a trough in the mid-1980s or that
we had reached a peak in 1993, the next inflection point is certain to surprise