In addition, we assessed other physical
features of the block groups: garbage accumulation,
graffiti, abandoned cars, billboards
and signs, and general upkeep of nonstructures
such as parks, playgrounds, vacant lots,
and institutional properties. To collect these
data, we walked the block group areas and
evaluated each block or street segment (e.g.,
the 1600 block of Canal St). For each street
segment, we looked for the presence of
garbage, graffiti, and abandoned vehicles. We
developed dichotomous variables to evaluate
these conditions (e.g., presence or absence of
accumulated garbage). The score for each
block group was an aggregate of the percentage
of street segments that had accumulated
garbage, graffiti, or abandoned (apparently
inoperable) cars.
Investigating the physical conditions
of neighborhood public high schools was
another aspect of our survey of neighborhood
conditions. We obtained physical plant
inspection reports of public schools conducted
by the Office of Sanitation Services in
New Orleans. These provided assessments of
the schools’ sites and play areas, buildings,
toilet facilities, handling of solid wastes, and
water supply from drinking fountains. To
code the reports, we totaled the number of
reported problems for each category. We gave
each individual problem of whatever type
highlighted on the report a score of 1, thus,
for example, giving equal weight to a broken
toilet and missing floor tiles. When the sanitation
inspector noted a problem or code violation
in specific rooms or locations in the
school, we gave a point to each room or area
listed on the report. When the report did not
give exact numbers for the observed problems
but cited problems in a “few,” “many,”
or “most” areas of the school or “throughout”
the school, we gave such descriptions the
following values: a few or some = 3 problems;
many or several = 5 problems; most or
throughout = 7 problems.
To link the schools to the block groups,
we obtained school “catchment area” maps
from the New Orleans School Board. Using
MapInfo and a block group boundary map,
we identified those schools whose catchment
areas encompassed our study block groups.
We then gave the schools’ scores (total number
of physical plant problems) to each block
group within each school catchment area.
We also calculated the density of retail
outlets that sold alcohol for consumption off
the premises (“off-sale”) by geocoding all
off-sale licenses obtained from the State
Alcohol Beverage Control Agency. Off-sale
alcohol outlet density has been shown to be
geographically related to gonorrhea rates.24
We created an index that reflects neighborhood
deterioration, the “broken windows”
index, which is the sum of the percentage of
homes with major structural damage, minor
structural damage, or cosmetic damage; the
percentage of streets with trash, abandoned
cars, or graffiti; and the number of physical
problems and building code violations in public
high schools, as documented by the Office
of Sanitation Services. Each of these variables
was normalized, with a mean of 0 and a standard
deviation of 1, so that each contributed
equally to the broken windows index. Using
1990 US Census data with 1995 updates, we
created an index of poverty for each block
group that consisted of the sum of the percentage
of households with incomes of less
than $15000, of individuals with less than a
high school education, and of persons older
than 18 years who were unemployed.
Bivariate relationships among several
variables were examined, including the sum
of rates of gonorrhea (1994–1996), the broken
windows index, the poverty index, and
the percentages of residents who were married,
Black, home owners, between the ages of
15 and 24 years, or female. Since many of the
independent factors measured similar variables
(e.g., broken windows and percentage of
home ownership are related to poverty), correlations
were performed to determine the
extent of possible confounding.
Backward elimination linear regression
was used to examine the relationship of broken
windows to gonorrhea and other correlated
variables. The following variables were
included: the percentage of the population
between the ages of 15 and 25 years, the poverty
index, marital status, home ownership,
broken windows index, off-sale alcohol outlet
density, and race. In addition, it was necessary
to determine whether gonorrhea rates
were related to poverty in different ways,
depending on the level of neighborhood deterioration.
To test this, an interaction term was
added to the regression model. Results
showed no significance, so the term was
omitted. Using ANOVA, we calculated the
difference in mean gonorrhea rates among
high- and low-poverty block groups and
block groups with high and low broken windows
indexes.