Moorpark College Safety Concerns in the Community Memorandum

579825ANN
research-article2015
The Annals of the American AcademyCrime Decline and Neighborhoods
Violence and
Neighborhood
Disadvantage
after the Crime
Decline
By
Michael Friedson
and
Patrick Sharkey
Violent crime is known to be concentrated in the same
urban neighborhoods as poverty and other forms of
disadvantage. While U.S. violent crime has declined at
an unprecedented rate over the past two decades, little
is known about the spatial distribution of this decline
within cities. Using longitudinal neighborhood crime
data from six U.S. cities during the national crime
decline, this article examines changes in (1) crime rates
of neighborhoods grouped by their initial crime levels,
poverty rates, and racial/ethnic makeups; (2) the neigh?
borhood exposure to violence of urban residents classi?
fied by race/ethnicity and poverty status; and (3) the
relative distribution of violent crime across urban
neighborhoods. We find that crime levels declined the
most in the initially most violent and disadvantaged
neighborhoods and that exposure to violence fell the
most among disadvantaged urban residents. None?
theless, crime remained concentrated in cities? initially
most violent and disadvantaged locales.
Keywords: crime decline; neighborhood change; con?
centrated disadvantage; violent crime;
urban inequality
O
ver the past 20 years, the United States
has experienced the most sustained
decline of violent crime in its modern history
(Zimring 2008). The national rates of homicide
and of all violent index crime have been cut
roughly in half since the early 1990s. The drop
in violence is visible in every source of data
available, from official police statistics on homi?
cides to victimization surveys conducted with
Michael Friedson is a postdoctoral fellow at New York
University. His research examines crime and policing in
disadvantaged urban areas, with a longitudinal focus.
Patrick Sharkey is an associate professor of sociology at
New York University. His research considers the role of
neighborhoods and cities in generating and maintaining inequality across multiple dimensions. He is currently examining how the decline of violent crime has
affected urban life and urban inequality in America.
DOI: 10.1177/0002716215579825
ANNALS, AAPSS, 660, July 2015
341
342
THE ANNALS OF THE AMERICAN ACADEMY
ordinary Americans. No matter the data source, the decline of violent crime is
staggering in its scale and duration.
Despite the evidence for how much violence has declined, little is known
about where it has declined. The dearth of evidence on how the crime decline
has been distributed across urban neighborhoods is a major gap in the literature.
One of the most unique aspects of interpersonal violence is that it is geographi?
cally concentrated. Research conducted in multiple cities has shown that a dis?
proportionate share of all violent crime takes place within an extremely small
number of city blocks and neighborhoods (Braga, Papachristos, and Hureau
2010; Braga, Hureau, and Papachristos 2011; Weisburd et al. 2004). A converging
strand of empirical research argues that the spatial concentration of violence may
be a core mechanism leading to the reproduction of neighborhood inequality
(Sharkey and Sampson, in press; Peterson and Krivo 2010; Sharkey 2010;
Burdick-Will et al. 2011).
This research suggests that to explain and interpret the crime decline, it is
necessary first to have a clear sense of the degree to which communities have
been affected by it. If the decline of violent crime never reached the nation?s
most disadvantaged neighborhoods, then ?The Great American Crime Decline?
(Zimring 2008) might be seen as yet another trend that has exacerbated neigh?
borhood inequality. Alternatively, if the decline of violent crime was concentrated
in the most disadvantaged neighborhoods, then the crime drop may have weak?
ened one of the central mechanisms by which neighborhood inequality is main?
tained and reproduced.
To address these questions, this study analyzes trends in neighborhood-level
violent index crime rates for six cities?Chicago, Cleveland, Denver, Philadelphia,
St. Petersburg (Florida), and Seattle. These municipalities are not representative
of U.S. cities but were selected because all have available data on neighborhoodlevel violent crime over a period of at least a decade. The data allow us to
describe how the crime drop was distributed across neighborhoods within each
city, and to answer questions about which neighborhoods and which populations
experienced the greatest declines in violent crime.
The Nexus of Concentrated Disadvantage and Violence
Multiple forms of disadvantage, from poverty to family structure to disease to
homicide, tend to come bundled together in urban neighborhoods (Kasarda
1993; Sampson and Morenoff 2006; Sampson 2012). This multifaceted disadvan?
tage tends to be durable, with a rank ordering of urban neighborhood status that
is remarkably stable over time (Sampson and Morenoff 2006; Sampson 2012).
NOTE: This research was funded by a grant from the William T. Grant foundation. The
authors would like to thank audience members at the Penn State Stratification Conference on
Residential Inequality in American Neighborhoods and Communities as well as Mikaila
Arthur, the special issue editors, and anonymous reviewers for excellent feedback on an earlier
draft of the article.
Crime Decline and Neighborhoods
343
The multiple dimensions of neighborhood disadvantage are mutually reinforcing,
constituting what some have referred to as a ?poverty trap? (Bowles, Durlauf, and
Hoff 2006; Sampson and Morenoff 2006; Sampson 2012).
From the 1960s through the 1990s, the rise of violent crime and the emer?
gence of mass imprisonment led to a new concentration of violence and a
strengthened spatial link among poverty, segregation, violence, and the criminal
justice system (Hagan and Peterson 1995; Krivo and Peterson 1996; Peterson and
Krivo 2010; Sampson 2012). Crime and neighborhood disinvestment reinforced
each other to constitute a ?spiral of decay? in high-poverty areas (Skogan 1990;
see also Bursik 1986; Liska and Bellair 1995). An extensive literature documents
how the spatial concentration of poverty, violence, aggressive police oversight,
and incarceration erode public life, compromising the capacity of neighborhood
residents to achieve social cohesion and community organization (Clear 2009;
Fagan and Meares 2008; Klinenberg 2003; Sampson, Raudenbush, and Earls
1997).
Considering the tight link between violent crime and urban disadvantage, the
dramatic drop in violent crime that has occurred since the 1990s has potentially
important implications for our understanding of urban poverty and neighbor?
hood inequality. Pope and Pope (2012) find a highly significant, negative relation?
ship between change in property values and change in crime rates during the
1990s in U.S. urban zip codes. Ellen and O?Regan (2008) find that the greatest
rates of economic growth during the 1990s occurred in urban neighborhoods
with the highest proportions of black and poor residents, while these neighbor?
hoods experienced the greatest economic losses as crime grew in the prior two
decades. These authors hypothesize that the concentration in these neighbor?
hoods of both the crime decline and the prior crime increase may help to account
for these patterns. Prior research bolsters this hypothesis, demonstrating that
increasing crime rates lead to losses of relatively affluent populations at both the
neighborhood and city levels (Cullen and Levitt 1999; Morenoff and Sampson
1997).
To understand whether the decline of violent crime has reversed these pat?
terns of disinvestment, decline, and out-migration, it is essential to analyze where
crime has dropped within urban areas. Much of the existing literature on the
magnitude of the crime decline has focused on measuring changes in crime com?
mission or victimization rates among groups classified by age, gender, socioeco?
nomic status, or race/ethnicity, or in large areas such as states or regions (Cook
and Laub 2002; Zimring 2008). Less attention has been devoted to the spatial
dimensions of the crime decline within cities. Studying crime trends in Cleveland
and Denver, Ellen and O?Regan (2009) establish that neighborhoods in these
cities with the greatest proportions of minority residents experienced the greatest
crime declines in the 1990s, and that neighborhood-level exposure to violent
crime fell most among these cities? poor and minority residents. A recent line of
research?conducted on the level of street segments and intersections (so-called
microplaces)?has made groundbreaking advances in our understanding of the
degree to which the crime decline and the prior incidence of violent crime were
localized within cities (Braga, Papachristos, and Hureau 2010; Braga, Hureau,
344
THE ANNALS OF THE AMERICAN ACADEMY
and Papachristos 2011; Weisburd et al. 2004). This research finds that a majority
of violent incidents in Boston and Seattle were concentrated in a relative handful
of microplaces. A paradoxical pattern of change and stability characterized the
crime decline in Boston?s most violent microplaces. Even as much of the city?s
declines in gun violence and robbery occurred in these locales, Boston?s gun vio?
lence and robbery incidents remained concentrated in these microplaces after
the declines.
This study adds to this evidence base by considering how the crime decline has
been experienced across neighborhoods and across populations characterized by
economic status and by race and ethnicity. We describe the degree of change in
violent crime in six cities, tracking violent crime at the neighborhood level for at
least a decade in all cities. As described in the following section, we consider the
absolute and relative decline of violent crime, changes in different types of neigh?
borhoods and for different segments of urban populations, and change and stabil?
ity in the spatial distribution of violence. We explore implications of these
patterns for future research on broader social conditions in disadvantaged neigh?
borhoods and for policies to reverse cycles of decay and disinvestment from
which these neighborhoods have previously suffered.
Data and Methods
Analytic approach
The analysis is carried out in three sections. The first section consists of a
neighborhood-level analysis examining where the decline in violent crime was
concentrated within cities. Each city?s neighborhoods are first divided into quin?
tiles based on their initial violent crime rates. For each quintile in each year of
the data, a violent index crime rate?expressed as the number of crimes per
10,000 residents?is calculated by dividing the total number of such crimes
occurring in the quintile?s neighborhoods by the total number of residents of
these neighborhoods, and multiplying the resulting quotient by 10,000. The
amount of change in this rate, from the data?s initial year to its final year, is com?
pared for each city?s most violent quintile and its remainder. Each city?s neighbor?
hoods are then divided into a group of ?poor? and a group of ?nonpoor?
neighborhoods, based on whether at least 30 percent of their residents were in
poverty during the data?s initial year.1 A violent index crime rate is calculated for
each of these groups in each year of the data. The amount of change in this rate,
from the data?s initial to its final year, is compared for each city?s poor and non?
poor neighborhoods. Finally, each city?s neighborhoods are divided according to
whether blacks, Hispanics, or non-Hispanic whites constituted a majority of their
populations in the data?s first year, or are designated as ?other? if none of these
groups made up a majority. Annual violent index crime rates are calculated for
each set of neighborhoods, and the amount of change, from the data?s initial to
its final year, is compared for each set.
Crime Decline and Neighborhoods
345
The second section considers change in the rate at which different groups of
individuals were exposed to violent crime within their neighborhoods from the
data?s initial year to its final year. Changes in exposure to neighborhood violent
crime are examined based on poverty status (poor and nonpoor residents), and
on race/ethnicity (non-Hispanic black, Hispanic, and non-Hispanic white popula?
tions). The measure of exposure is defined as the sum of the violent crime rates
of all of the city?s neighborhoods, with each neighborhood rate weighted by the
number of group members residing in that place. This definition can be expressed
in symbolic form as S nj =1 (V j * (P j / Pt )) , where each Vj is the violent crime rate of
one of the n neighborhoods in the city; Pj is the population in that neighborhood
of the group for which the exposure rate is being calculated; and Pt is the group?s
population in the city as a whole.
The final section analyzes how the crime decline affected the distribution of
violence across the neighborhoods of each city. The correlation is measured
between the initial and final violent crime rates of the neighborhoods in each city.
These rates are logged to prevent extreme or outlying values from having a dis?
proportionate influence on results. The correlation coefficient provides one
measure of the stability or change in the distribution of these neighborhoods?
crime rates. Each city?s neighborhoods are then divided into quintiles by their
violent crime rates in the data?s initial year and final year, respectively. Transition
matrices are used to display the degree of change in the relative position of
neighborhoods within a city in terms of violent crime. Specifically, we focus on
the proportions of neighborhoods that began in the most or least violent quintile
and remained there at the end of the timeframe under study.
Data
We analyze data on the violent index crime rates of the neighborhoods within
six cities?Chicago, Cleveland, Denver, Philadelphia, St. Petersburg, and
Seattle.2 These cities were selected because neighborhood-level crime data cov?
ering a period of at least a decade during the national crime decline are publicly
available for each. Violent index crimes, as defined by the FBI, consist of inten?
tional homicides, robberies, rapes, and aggravated assaults. A neighborhood?s
violent index crime rate consists of the number of such crimes occurring within
its boundaries per 10,000 neighborhood residents. Because each city?s crime data
are derived from a different local source, there are inconsistencies between cities
in terms of the years and crimes for which data are available, as well as the defini?
tions of ?neighborhoods.?
Data spanning just over a decade are given for four cities: Chicago,
Philadelphia, St. Petersburg, and Seattle. The time periods covered by these
municipalities? data are, respectively, 2001 to 2012, 1998 to 2009, 2000 to 2012,
and 1996 to 2007. For the remaining two cities, Cleveland and Denver, data span
the period from 1990 to 2010. This study?s data tables and graphs display changes
in Cleveland and Denver crime rates over the period from 1990 to 2010 and each
of its constituent decades. This enables comparison of changes occurring in
Cleveland and Denver and elsewhere over periods of similar duration. By
346
THE ANNALS OF THE AMERICAN ACADEMY
default, results for Cleveland and Denver that are discussed in the text concern
1990 to 2010.
Neighborhoods are defined as census tracts for all cities except Denver. The
source for the Denver data (the Piton Foundation) divides this city into 77 neigh?
borhoods, whereas the 2000 U.S. Census divides Denver into 136 tracts. We
believe that the benefits of adding this city to our dataset outweigh any inconsist?
encies introduced into our analysis by the larger size of the neighborhood units
for which data are available.
Homicides and rapes are not included in the Philadelphia crime counts pro?
vided by our data source. This omission is not expected to introduce substantial
inconsistencies or distortions into our analysis, given that the omitted crimes
constituted just 6 percent of all violent index crimes in Philadelphia during the
period under analysis,3 and given the typically high degree of correlation between
tract-level counts of the omitted crimes and of all violent index crimes. For
instance, the correlation equals 0.9 (indicating an extremely strong linear rela?
tionship) between the total number of homicides and all violent index crimes
occurring in Chicago tracts from 2001 to 2012, while that of the number of rapes
and all violent index crimes is even stronger.
Demographic variables?namely population counts disaggregated by poverty
status, race, and ethnicity?are derived from the same sources as the crime data
for neighborhoods in Denver and Cleveland. For tracts in the other cities, these
variables are derived from the U.S. censuses and American Community Survey
five-year averages. These averages are treated as applying to the middle year in
their five-year span. Annual values of demographic variables, for years between
those for which they are provided by these data sources, are calculated via linear
interpolation.
Longitudinal analysis of neighborhood-level data over multiple decades poses
challenges due to changes in tract boundary definitions in each decennial census.
These challenges do not apply to the Cleveland and Denver data, because the
sources from which they are derived use consistent neighborhood boundaries
throughout the periods they cover. Tract boundary changes in Seattle and St.
Petersburg are minor over the periods covered by their crime data. Data for each
of these cities was manually adjusted to conform to a uniform set of tract bounda?
ries. Challenges posed by tract boundary changes in Chicago and Philadelphia
are more substantial. See the online methodological appendix for a detailed dis?
cussion of adjustments made to these cities? data to facilitate the longitudinal
analysis of their neighborhood crime rates.4
Results
Where did violent crime decline? A neighborhood-level analysis
The analysis presented in Table 1 shows the absolute and proportional changes
in violent crime rates for the most violent quintile of neighborhoods (as of the
baseline year) and the remainder of each city?s neighborhoods. Absolute declines
347
Crime Decline and Neighborhoods
Table 1
Change in Violent Crime Rates of Neighborhoods Grouped by Initial Violent Crime
Levels
City
Chicago
Cleveland
Cleveland
Clevelanda
Denver
Denver
Denver
Philadelphia
Seattle
St. Petersburg
Time
Period
2001?2012
1990?2010
1990?2000
2000?2010
1990?2010
1990?2000
2000?2010
1998?2009
1996?2007
2000?2012
Absolute Change
Relative Change
Highest
Entire
Quintile Remainder City
Highest
Entire
Quintile Remainder City
?109.67
?175.83
?177.00
1.17
?95.42
?83.47
?11.95
?62.65
?67.32
?202.31
?28.92
?43.28
?43.57
0.51
?47.54
?41.58
?10.19
?22.91
?28.54
?42.94
?32.31
19.27
?7.38
26.65
?10.77
?16.19
5.42
?2.00
?10.47
?41.31
?51.36
?20.22
?41.75
21.53
?25.75
?28.57
2.82
?15.40
?21.01
?74.50
?32.57
18.39
?7.05
27.36
?22.51
?33.84
17.12
?2.25
?23.80
?46.72
?33.10
?12.31
?25.42
17.57
?33.14
?36.76
5.73
?12.28
?25.55
?45.88
a. Although the count of violent incidents in Cleveland was roughly stable from 2000 to 2010,
the city experienced a major loss of population in these years that resulted in an increase of its
overall violent crime rate.
NOTE: Neighborhoods are grouped into quintiles of roughly equal population sizes by their
initial violent crime rates. The ?highest quintile? consists of the neighborhoods with the highest
initial rates, while the ?remainder? consists of those outside the highest quintile. Absolute
change is the difference between the crime rates in the last and first years of the specified time
period. Relative change is this difference as a percentage of the first year?s crime rate. To
reduce the impact of anomalous annual crime rates on results, multiyear averages of initial
crime rates are used to divide the neighborhoods into quintiles. See the online appendix
(http://ann.sagepub.com/supplemental) for graphs of change in cities? neighborhood crime
rates, for neighborhoods grouped by initial crime rates, poverty levels, and racial/ethnic
makeup.
in violent crime in each city?s most violent neighborhoods far outstripped the
changes in violent crime occurring in their remainders. For example, from 2001
to 2012 the violent crime rate in Chicago?s most violent neighborhoods dropped
by 110 crimes per 10,000 residents, whereas the rate of violent crime dropped by
32 crimes per 10,000 residents in the remainder of the city.
Due to the large declines in the absolute levels of violent crime in each city?s
most violent neighborhoods, there was a convergence in violent crime rates
between the most violent neighborhoods of each city and the rest of its neighbor?
hoods. The absolute difference in violent crime rates between each city?s most
violent neighborhoods and all other neighborhoods shrunk by between 28 per?
cent for Chicago and 65 percent for Cleveland during the years covered by the
data.
348
THE ANNALS OF THE AMERICAN ACADEMY
Table 2
Change in Violent Crime Rates of Neighborhoods Grouped by Poverty Status
Absolute Change
City
Chicago
Cleveland
Cleveland
Cleveland
Denver
Denver
Denver
Philadelphia
Seattle
St. Petersburg
Relative Change
Time Period
Poor
Nonpoor
Poor
Nonpoor
2001?2012
1990?2010
1990?2000
2000?2010
1990?2010
1990?2000
2000?2010
1998?2009
1996?2007
2000?2012
?104.74
?52.45
?88.05
35.60
?78.52
?73.25
?5.27
?32.34
?120.21
?250.66
?35.38
14.26
?2.92
17.18
?14.50
?20.28
5.78
?6.11
?15.23
?59.79
?32.64
?20.32
?34.11
20.93
?44.75
?41.75
?5.16
?15.71
?35.89
?43.55
?30.65
15.89
?3.26
19.79
?25.12
?35.13
15.43
?6.79
?22.60
?46.93
NOTE: Neighborhoods are classified as ?poor? if their poverty rates are at least 30 percent in
the data?s initial year and as ?nonpoor? otherwise. See the Table 1 note for definitions of abso?
lute and relative change.
The columns of results focusing on absolute levels of violent crime give equal
weight to every incident of violent crime.5 The second set of columns shows the
declines in violent crime rates in proportional terms, or relative to initial rates. In
Chicago, Seattle, and St. Petersburg, the proportional decline of violent crime
was roughly equivalent in the cities? most violent neighborhoods and in the
remainder of the cities? neighborhoods. In Cleveland, Denver, and Philadelphia,
the proportional decline in the most violent quintile of neighborhoods was sub?
stantially larger than that in the remainder of neighborhoods. Indeed, in both
Cleveland and Philadelphia, all or nearly all of the drop in violent crime was
concentrated in the cities? most violent neighborhoods.
The analysis presented in Table 2 repeats that of Table 1, but for neighbor?
hoods classified by their initial poverty status rather than their initial violent
crime rates. The absolute decline in violent crime in each city?s poor neighbor?
hoods far exceeded that in its nonpoor neighborhoods. For example, the decline
was more than 120 crimes per 10,000 residents in Seattle?s poor tracts, compared
with only about 15 crimes per 10,000 residents elsewhere. Proportional declines
in violent crime were roughly similar in poor and nonpoor tracts in Chicago and
St. Petersburg. In contrast, the percentage drops in violent crime in the remain?
ing four cities were substantially greater in poor neighborhoods.
The violent crime rate was initially higher in each city?s poor neighborhoods
than in its remainder. The larger absolute decline in violent crime in each city?s
poor neighborhoods thus means that there was a convergence in crime levels
among poor and nonpoor neighborhoods. The absolute difference in violent
crime rates between each city?s poor and nonpoor neighborhoods shrunk by
between 23 percent for Philadelphia and 54 percent for Denver.
349
Crime Decline and Neighborhoods
Table 3
Change in Violent Crime Rates of Neighborhoods Grouped by Racial/Ethnic
Composition
Absolute Change
City
Chicago
Cleveland
Cleveland
Cleveland
Denver
Denver
Denver
Philadelphiaa
Seattle
St. Petersburg
Time Period White
2001?2012
1990?2010
1990?2000
2000?2010
1990?2010
1990?2000
2000?2010
1998?2009
1996?2007
2000?2012
Black
?18.80 ?73.55
20.23 ?55.78
2.68 ?88.23
17.55
32.45
?10.10 ?60.28
?17.93 ?53.67
7.83
?6.61
2.53 ?24.01
?15.11
?
?50.83 ?129.16
Hispanic Other
Relative Change
White
Black Hispanic Other
?48.84 ?37.46 ?32.85 ?25.66
?
?96.40
20.84 ?23.36
?
?104.30
2.76 ?36.95
?
7.90
17.60 21.55
?33.63 ?170.71 ?19.26 ?54.72
?33.84 ?131.59 ?34.19 ?48.72
0.21 ?39.13
22.69 ?11.69
?76.89
?9.42
3.92 ?13.91
?
?39.74 ?23.43
?
?
?
?51.15 ?35.62
?39.48
?
?
?
?31.08
?31.27
0.28
?28.36
?
?
?40.22
?40.39
?43.70
5.88
?64.98
?50.09
?29.84
?6.66
?27.04
?
a. Although the average crime rate of a city?s majority-black neighborhoods typically exceeds that of
its majority-Hispanic neighborhoods, the reverse holds in Philadelphia. The majority-Hispanic
neighborhoods in Philadelphia (n = 16) are exceptionally disadvantaged. Their average poverty rate
exceeded 50 percent in 2000, versus 29 percent in Philadelphia?s majority-black neighborhoods (n =
146).
NOTE: Neighborhoods are classified according to whether non-Hispanic whites, non-Hispanic
blacks, or Hispanics make up a majority of their populations in the data?s initial year, or as ?other? if
none constitutes a majority. Results are excluded for categories containing two or fewer neighbor?
hoods with available data.
The analysis in Table 3 repeats those in Table 1 and Table 2, but for neighbor?
hoods grouped by their majority racial/ethnic compositions. Results are excluded
for a racial/ethnic category when the city contains no more than two neighbor?
hoods (with available crime data) falling into that category. On this basis, results
are given for majority-black neighborhoods in all cities except Seattle. In these
cities, the absolute decline in violent crime in majority-black neighborhoods far
exceeded that in majority-white neighborhoods. For instance, Chicago?s majorityblack neighborhoods experienced a decline of 74 violent crimes per 10,000 resi?
dents, versus 19 per 10,000 in its majority-white tracts. In two cities, Cleveland
and Philadelphia, violent crime increased in majority-white neighborhoods while
falling in majority-black tracts. Results are provided for majority-Hispanic neigh?
borhoods in Chicago, Denver, and Philadelphia. In these cities, majority-His?
panic neighborhoods experienced greater absolute declines in violent crime than
majority-white neighborhoods.
The violent crime rate was initially higher in each city?s majority-black and
majority-Hispanic neighborhoods than in its majority-white neighborhoods. The
absolute difference in violent crime rates between each city?s majority-white and
majority-black neighborhoods subsequently shrunk by between 24 percent for
350
THE ANNALS OF THE AMERICAN ACADEMY
Chicago and 87 percent for Denver, while that between majority-white and
majority-Hispanic neighborhoods shrunk by between 38 percent for Philadelphia
and 45 percent for Chicago.
Although absolute declines in violent crime were greater in majority-black
than majority-white neighborhoods in all cities, majority-white neighborhoods
saw greater proportional declines in two cities: Chicago and St. Petersburg. In
the remaining cities, proportional declines were greater in majority-black than
majority-white neighborhoods. Proportional declines were greater in majorityHispanic than in majority-white neighborhoods in all three cities that had results
for majority-Hispanic neighborhoods.
The most violent, disadvantaged neighborhoods within the cities experienced
declines in violent crime that were as large as, or larger than, those of the cities?
remainders. This is true regardless of whether neighborhoods are divided by
their initial violent crime rates, poverty levels, or racial/ethnic compositions. Still,
in all cities, the most violent quintile of neighborhoods continued to experience
higher crime levels, in the data?s final year, than had the second most violent
quintile in the initial year. The same applies to poor relative to nonpoor neighbor?
hoods and to majority-black (in all cities but Denver) and majority-Hispanic rela?
tive to majority-white neighborhoods. In other words, even after the crime
declines depicted here, the cities? worst-off neighborhoods had more violent
crime than other neighborhoods had before this decline.
Who experienced the crime decline? A group-level analysis
In this section, we analyze how these changes in violent crime rates translated
into shifts in individuals? exposure to violent crime in their neighborhoods.
As depicted in Figure 1, exposure to violent crime of each city?s poor and non?
poor residents declined over the period covered by the data. The decline experi?
enced by poor residents was, however, greater than that experienced by nonpoor
residents in each case. For example, exposure of the poor residents of Seattle
declined by 50 incidents per 10,000 residents, while that of its nonpoor residents
declined by only 17 incidents. The number of violent incidents by which the
exposure rate of each city?s poor residents declined is at least 1.7 times greater
than that by which the exposure rate of its nonpoor residents dropped.
The greater magnitude of the declines affecting poor residents means that
there was a convergence in the number of violent crimes to which poor and non?
poor residents were exposed. Poor residents of each city were initially exposed to
at least 54 (for Philadelphia) and at most 128 (for St. Petersburg) more violent
incidents per 10,000 residents than nonpoor residents. In the data?s final year, this
difference exceeds 50 incidents per 10,000 residents in just one city, St.
Petersburg. In four cities?Cleveland, Denver, Seattle, and St. Petersburg?this
difference shrunk by at least half of its initial level. In the remaining two cities,
Chicago and Philadelphia, it declined by substantial proportions, 41 percent and
26 percent, respectively. For all cities but Cleveland and Philadelphia, the poor
were exposed to about as much violent crime in the data?s final year as were the
nonpoor in the data?s initial year.
351
Crime Decline and Neighborhoods
Figure 1
Average Neighborhood Exposure to Violent Crime of Poor and Nonpoor City Residents
300
Violent Crimes per 10,000 Residents
250
200
150
100
50
0
Chicago
2001-2012
Cleveland
1990-2010
Poor, Ini?al Year
Denver
1990-2010
Poor, Final Year
Philadelphia
1998-2009
Non-Poor, Ini?al Year
Seale
1996-2007
St. Petersburg
2000-2012
Non-Poor, Final Year
Violent Crimes per 10,000 Residents
250
200
150
100
50
0
Cleveland
1990-2000
Poor, Ini?al Year
Cleveland
2000-2010
Poor, Final Year
Denver
1990-2000
Non-Poor, Ini?al Year
Denver
2000-2010
Non-Poor, Final Year
NOTE: The average neighborhood exposure rate to violent crime of a poor or nonpoor city
resident is the average amount of violent crime occurring in the neighborhood of such a resi?
dent. See the main text for the formula by which this average rate is calculated. When demo?
graphic data were not available for the initial or final years of crime data, the most proximate
available demographic data were used in weighting neighborhood crime rates to calculate
these averages.
Each city?s poor residents experienced a greater proportional drop in exposure
to violent crime (measured relative to its initial rate of exposure) than its nonpoor
residents. In three cities?Chicago, Philadelphia, and St. Petersburg?the pro?
portional ch