The Urban Institute undertook a comprehensive
assessment of communities approaching decay to provide public
officials with strategies for identifying communities in the early
stages of decay and intervening effectively to prevent continued
deterioration and crime. The assessment consisted of three parts: (1)
a survey of innovative local programs designed to combat neighborhood
decay and crime conducted with the Police Executive Research Forum,
(2) a review of the literature to identify research findings on
preventing decay and crime and promising areas of future research, and
(3) the analysis of the predictive validity of alternative indicators
of community decay. Out of the third segment came the data prepared
for this data collection. Existing theories reflect considerable
disagreement over the temporal sequence between decay and crime.
Although community decline is a dynamic spiral downward in which the
physical condition of the neighborhood, adherence to laws and
conventional behavioral norms, and economic resources worsen, the
question of whether decay fosters or signals increasing risk of crime,
or crime fosters decay (as investors and residents flee as reactions
to crime), or both, is not easily answered.
Using specific indicators to identify future
trends, predictor models for Washington, DC, and Cleveland were
prepared, based on data available for each city. The models were
designed to predict whether a census tract should be identified as at
risk for very high crime and were tested using logistic regression.
The classification of a tract as a "very high crime" tract was based
on its crime rate compared to crime rates in other tracts in the same
city. The models use crime as the dependent variable, i.e., the
outcome of social and economic distress at prior time-points and
breakdowns in public order and violations that undermine the physical
maintenance and quality of life in the neighborhood, controlling for
earlier crime rates. The eight predictors of high crime risk used in
this study comprise four general groups. The first group is the prior
crime rate for the offense. The second group includes indicators of
breakdown in public order and the presence of illegal activity harmful
to neighborhood environment, including the drug arrest rate, the
delinquency rate, and the rate of confirmed or suspected arson
incidents. The third group reflects factors related to the maintenance
of social control by a stable population with sufficient resources and
a common interest in protecting the area. This includes an index of
family poverty, the presence of public housing, and home
ownership. The fourth group is comprised of the percentage of lots
zoned for commercial use, representing access to situations that
increase opportunities for certain crimes. Dichotomous variables were
created to compare the highest-rate tracts to all others and used as
dependent variables in the prediction equations. Because the accuracy
of prediction would vary depending on the number of tracts in the
"very high crime" group, alternative definitions grouped the worst
10 to 30 tracts in multiple tests of each model. To control for
differences in population and to facilitate cross-tract comparisons,
counts of crime incidents and other events were converted to rates per
1,000 residents. In Washington, DC, tract population for interim years
was estimated by using the change in city population from 1980 to 1990
to revise the 1980 tract population (average per year proportion
change in population times the number of years since 1980). Also, the
Washington, DC, data were aggregated by 1970 tract boundaries because
some agencies did not shift their geographic coding to 1980 Census
tract boundaries for several years and so did not reflect instances
where a single tract was divided into two tracts as the population
expanded. Tracts with less than 100 residents were considered
nonresidential or institutional and were deleted from the analysis. In
Washington, DC, 12 tracts were deleted and in Cleveland, 10 tracts
were deleted.
Washington, DC, and Cleveland, Ohio, were selected because
data could be provided on multiple indicators for multiple years
between 1980 and 1990.
census tract
District of Columbia Office of Planning, Police
Department, Office of Criminal Justice Plans and Analysis, and
Division of Research and Statistics of the Commission of Public
Health, and the Center for Urban Poverty and Social Change, Mandel
School of Applied Social Sciences, Case Western Reserve University
aggregate data, and census/enumeration data
Washington, DC, variables include rates for arson
and drug sales or possession, percentage of lots zoned for commercial
use, percentage of housing occupied by owners, scale of family
poverty, presence of public housing units for 1980, 1983, and 1988,
and rates for aggravated assaults, auto thefts, burglaries, homicides,
rapes, and robberies for 1980, 1983, 1988, and 1990. Cleveland
variables include rates for auto thefts, burglaries, homicides, rapes,
robberies, drug sales or possession, and delinquency filings in
juvenile court, and scale of family poverty for 1980 through 1989.
Rates for aggravated assaults are provided for 1986 through 1989 and
rates for arson are provided for 1983 through 1988.
Not applicable.
None