X
6
m1
a
mjk
X
mijk
!
D
tijk
p
tjk
X
5
m1
d
mjk
W
mijk
!
D
rijk
p
rjk
X
3
m1
t
mjk
Z
mijk
!
;
Table 3
Neighborhood indicators derived from systematic social observation of 1135 streets in 57 study neighborhoods
a
Factor/item Yes N (%) No N (%)
Physical incivilities
Grati present 444 (39.1) 691 (60.9)
Resident grounds poorly maintained 94 (9.8) 870 (90.2)
Resident buildings poorly maintained 110 (11.0) 891 (89.0)
Moderate or signicant litter 280 (24.7) 853 (75.3)
Vacant/burned residences present 351 (30.9) 780 (68.7)
Vacant/burned commercial establishments present 103 (9.1) 1009 (90.7)
Public spaces poorly maintained 370 (32.9) 738 (65.0)
Territoriality
Crime watch/security/no trespassing signs visible 836 (73.7) 299 (26.3)
Residents react to presence of raters 502 (61.1) 320 (38.9)
One third or more of homes with borders/hedges 405 (40.8) 587 (59.2)
One third or more of homes with security bars 251 (25.2) 746 (74.8)
One third or more of homes with decoration 610 (61.2) 386 (38.8)
Sign visible denoting neighborhood name 25 (2.5) 968 (97.5)
Available play resources
One third or more of homes with yards 764 (76.6) 234 (23.4)
Street is not a busy thoroughfare 802 (70.8) 330 (29.2)
Public playground in good condition present 20 (1.8) 1098 (98.2)
Children visible playing 162 (14.3) 973 (85.7)
a
The balance between the sum of the yes and no responses and 1135 represents missing data.
M.O. Caughy et al. / Health & Place 7 (2001) 225236 230
where D
pijk
is an indicator coded 1 if the item is part of
the physical incivilities scale and 0 if otherwise, D
tijk
is an
indicator coded 1 if the item is part of the territoriality
scale and 0 if otherwise, and D
rijk
is an indicator coded 1
if the item is part of the play resources scale and 0 if
otherwise. X
mijk
, W
mijk
, and Z
mijk
are a series of dummy
variables representing 6 of the 7 physical incivilities
items, 5 of the 6 territoriality items, and 3 of the 4 play
resources items, respectively. These dummy variables are
centered around their grand means, allowing the
intercept variables, p
pjk
, p
tjk
, and p
rjk
, to be interpreted
as the log-odds of observing a positive response on a
typical item on the physical incivilities scale, territori-
ality scale, and play resources scale, respectively. These
log-odds are allowed to vary at both the street and
neighborhood levels. The remaining parameters, a
mjk
,
d
mjk
, and t
mjk
, when multiplied by 1, represent the item
diculties or severities for item m in the physical
incivilities scale, the territoriality scale, and the play
resources scale, respectively. These item diculties are
considered as xed eects at the street and neighbor-
hood levels. For complete information on the analytic
method, please see Raudenbush and Sampson (1999).
The parameters for the three-level item response model
were estimated using MLwiN BETA version 1.10.0001
(Rabash et al., 1999).
Results
The estimated item diculties at the street level are
displayed in Table 4, and the variancecovariance
components are displayed in Table 5. For the item
diculties, one item per scale is chosen as the reference
item, usually the item with the median frequency, and
the other item diculties should be compared to it. The
item diculties show a nice bit of variation, one of the
characteristics of a good scale, and their ordering is
consistent with what one would expect from the
frequency distributions of the items. Items which were
observed more frequently are considered less severe or
dicult than items observed less frequently.
Turning now to the variancecovariance components
in Table 5, the estimated variation in physical incivilities
between streets in the same neighborhood is 0.657 while
the between neighborhood variation is 1.919, yielding an
intraclass correlation for this scale of 0.74. That is, 74%
of the variation in physical incivilities is estimated to be
between neighborhoods. This is substantially higher
than the corresponding amount reported by Rauden-
bush and Sampson (1999) of 39% for their physical
disorder scale. For the territoriality and play resources
scales, the within neighborhood variances as well as the
scale covariances were estimated to be zero. This is
Table 4
Item diculty at the street level
Scale/item Coecient SE
Physical incivilities
Intercept 1.375 0.195
Grati 0.966 0.111
Condition of public spaces 0.640 0.112
Vacant residences 0.450 0.113
Litter 0 (reference)
Condition of resident buildings 1.290 0.148
Condition of resident grounds 1.446 0.148
Vacant commercial establishments 1.547 0.142
Territoriality
Intercept 0.388 0.079
Crime watch/security/no trespassing signs 1.478 0.095
Reactions to raters 0.874 0.099
Residences with decoration 0.862 0.094
Residences with borders/hedges 0 (reference)
Residences with security bars 0.741 0.099
Sign denoting neighborhood name 3.343 0.214
Play resources
Intercept 1.805 0.139
Yards 3.245 0.121
Busy street 2.944 0.115
Children playing 0 (reference)
Presence of usable public playground 2.267 0.243
M.O. Caughy et al. / Health & Place 7 (2001) 225236 231
similar to the ndings of Raudenbush and Sampson
(1999) for their social disorder scale, who concluded that
the zero variance estimate resulted from the rarity of a
positive response on their social disorder scale rather
than from a true lack of variance in social disorder
within neighborhoods. It is unlikely this is a similar issue
for the data reported here. Whereas the proportion of
positive responses was extremely low for the items on
Raudenbush and Sampsons social disorder scale (range
0.075.7%), the proportions of positive responses for
the territoriality and play resources scales in this study
were much higher. Therefore, given the estimated
variances, it is not possible to reject the null hypothesis
that the true within neighborhood variances for
territoriality and play resources are zero.
Contrary to our prediction, the Territoriality scale
was not negatively correlated with the physical incivi-
lities. The territoriality scale was also not correlated with
the play resources scale. However, consistent with our
prediction, the correlation between physical incivilities
in a neighborhood and play resources was 0.29, which
was just short of statistical signicance (p 0:05),
suggesting that neighborhoods with a high degree of
trash, grati and abandoned buildings are also less
likely to have resources available for childrens play.
Raudenbush and Sampson (1999) provide methods
for calculating street-level as well as neighborhood-level
reliabilities. The neighborhood-level reliability for each of
the scales was very high at 0.96, 0.93, and 0.94 for the
physical incivilities, territoriality, and play resources
scales, respectively. At the street-level, the reliability of
the physical incivilities was 0.74, which is more than
adequate. However, the street-level reliability of the
territoriality and play resources scales was low at 0.33
and 0.42, respectively. These results indicate that
although there is sucient variation in territoriality
and play resources to distinguish between neighbor-
hoods, there is insucient variation to distinguish
between streets within the same neighborhood. In
contrast, the variability in physical incivilities was
sucient for distinguishing not only between neighbor-
hoods but also between streets within the same
neighborhood.
Construct validity was examined by estimating the
correlation between the three observational scales,
neighborhood structural characteristics and residents
perceptions of their neighborhood. These correlations
are displayed in Table 6. As predicted, impoverished
neighborhoods with higher crime density also had higher
rates of observed physical incivilities. Territoriality was
not correlated with neighborhood impoverishment but
was positively correlated with crime density. Although
this was not predicted, it may have been that selected
elements of the territoriality scale such as security bars
and/or crime watch signs may have contributed to the
positive association between territoriality and crime
because the presence of these items in a neighborhood
may have been a response of the residents to higher
crime rates. Also surprising was the positive correlation
between play resources and neighborhood impoverish-
ment.
The physical incivilities scale was correlated as
expected with residents perceptions of neighborhood
conditions. Physical incivilities were positively corre-
lated with residential mobility, fear of retaliation, and
perceived physical and social disorder and negatively
correlated with willingness to intervene in acts of
delinquency and/or misbehavior or provide aid to
children in need as well as with overall neighborhood
quality. Territoriality appeared to have the opposite
association with perceived neighborhood conditions,
having a negative association with residential mobility
and fear of retaliation, and positive associations with
perceived willingness of residents to collectively monitor
youth. As such, it conrms our hypothesis that observed
signs of territoriality may be an outward manifestation
Table 5
Variance-covariance components
Variancecovariance component Estimate SE
Between streets w/in neighborhoods Variance of physical incivilities 0.657 0.083
Variance of territoriality } }
Variance of play resources } }
Covariance of physical incivilities and territoriality } }
Covariance of physical incivilities and play resources } }
Covariance of territoriality and play resources } }
Between neighborhoods Variance of physical incivilities 1.919 0.384
Variance of territoriality 0.212 0.051
Variance of play resources 0.759 0.165
Covariance of physical incivilities and territoriality 0.001 0.099
Covariance of physical incivilities and play resources 0.358 0.184
Covariance of territoriality and play resources 0.057 0.065
M.O. Caughy et al. / Health & Place 7 (2001) 225236 232
of social cohesion in the community. In contrast, the
correlations between the play resources scale and the
neighborhood perceptions measures, while signicant,
were counterintuitive, with negative associations with
neighborhood monitoring of youth as well as overall
quality. It is unclear whether these associations may be
an artifact of the restricted range of neighborhoods for
which perceptions data are currently available compared
to the full range of neighborhoods providing observa-
tional data.
A further step was taken to examine the utility of the
observational methods of neighborhood conditions by
examining the variability of the indicators within
neighborhoods that are relatively homogeneous with
respect to socioeconomic status. We have argued else-
where that measures of socioeconomic status alone mask
a considerable amount of variability within neighbor-
hoods (Caughy et al., 1999). To examine this, the study
neighborhoods were stratied according to their quartile
of average household wealth with respect to the city as a
whole, and the distribution of each neighborhood
observational indicator within each wealth quartile is
displayed in the Fig. 1. As expected, there appears to be
a linear and negative association between physical
incivilities and average household wealth, with observed
physical incivilities decreasing with increasing average
household wealth. Notable, however, is the signicant
amount of variation within wealth quartiles in the
amount of physical incivilities that were observed. As
can be seen in the gure, there is a signicant amount of
overlap in the distributions, with some neighborhoods in
the lowest quartile of wealth having rates of physical
incivilities which are lower than neighborhoods in the
highest quartile. Thus, socioeconomic status is insu-
cient as a measure for discriminating between neighbor-
hood residential contexts.
The distributions of territoriality and play resources
by average household wealth are also displayed in the
gure. The correlation between territoriality and average
household wealth is non-signicant (r 0:218, p
0:12 after excluding outlier cases), and the variation
appears more restricted within wealth categories than
physical incivilities. Counterintuitively, availability of
play resources is negatively correlated with average
household wealth (r 0:345, p50:05 after excluding
outlier cases). The degree of variation in the play
resources scale in neighborhoods in the lowest quartile
of wealth is notable.
Discussion
We presented the results of the development of an
assessment tool for collecting information on neighbor-
hood characteristics of relevance to studies concerned
with the well-being of families and children. Our
instrument draws from previous eorts (NORC, 1995;
Perkins et al., 1992) and includes additional items
important for understanding neighborhood inuences
on parents and children. Our protocol can supplement
information on neighborhoods obtained from more
routine secondary sources of data as it is easy to
implement and yields factors with high reliability.
Previous contextual research which rely on secondary
sources of data for information on neighbor-
hood characteristics has narrowly conceptualized
Table 6
Correlations between observational measures of neighborhood, neighborhood structural characteristics, and residents perceptions of
neighborhood climate
a
Physical incivilities Territoriality Availability of
play resources
Neighborhood structural characteristics
Impoverishment 0.293** 0.120 0.279***
Crime density 0.407*** 0.346*** 0.109
Neighborhood perceptions scale
Residential mobility 0.341*** 0.183** 0.087
Social interaction 0.070 0.017 0.104
Fear of retaliation 0.389*** 0.193** 0.133*
Intervene delinquency 0.133* 0.131* 0.150**
Intervene misbehavior 0.189** 0.128* 0.030
Intervene assist children in need 0.253*** 0.174** 0.164**
Physical/social disorder 0.375*** 0.129 0.029
Fear of victimization 0.078 0.090 0.048
Overall neighborhood quality 0.175** 0.134 0.148**
a
Outliers on the territoriality and play resources scales were excluded, respectively, in the calculations of correlations between each
observational scale and perceptions scale. The correlations for the play resources scale were virtually unchanged, but the correlations
for the territoriality scale changed substantially. See gure for graphic display of outliers for each scale. *p50:10, **p50:05,
***p50:01.
M.O. Caughy et al. / Health & Place 7 (2001) 225236 233
neighborhood inuences on the health and well-being of
residents. Collection of primary data using rapid
assessment methods such as the observational method
presented here oers the opportunity to expand our
understanding of the relevant characteristics that aect
families with relatively little outlay of resources.
Collecting data on each block face took anywhere from
10 to 30 min depending on whether residents stopped to
talk to the neighborhood assessors. Data entry of the
instrument was straightforward as all items were close
ended or of a check box format.
The rapid assessment method presented here has a
number of advantages compared with other objective
assessment methods of neighborhood conditions re-
cently reported in the literature. Compared with
measures using videotaped data collection (Cohen
et al., 2000; Raudenbush and Sampson, 1999), our
method is quick and easy to implement. Despite this ease
of implementation, however, the factors it yields show a
high degree of reliability at a relatively small unit of
observation, the census block group. Furthermore, we
have identied two additional neighborhood constructs,
territoriality and availability of play resources, which
expand the variety of neighborhood characteristics that
are measured beyond a sole focus on physical and/or
social disorder. Future research will be needed to
determine if these neighborhood characteristics are
important correlates of individual health and well-being.
There are limitations of the proposed assessment
methodology that should be noted. A number of
limitations arise from the use of a three-level
item response model for the analysis. For one, the
requirement that all indicators be dichotomized
may have sacriced important variation in the
indicators. Hopefully, further advances in analytic
methods will permit us to capture more fully the
variation in these measures. Secondly, unlike other
methods of latent variable analysis, a three-level item
response model does not provide an index of model t or
other criteria to assist the investigator in determining
whether the factor structure chosen is the one which best
ts the data.
Fig. 1. Neighborhood observation indicators by average household wealth.
M.O. Caughy et al. / Health & Place 7 (2001) 225236 234
A second limitation of the measure was the lack of
items reecting social disorder such as observations of
illegal activity (i.e., drug dealing, prostitution). This
limitation was partly the result of conducting all
observations in the late morning and early afternoon,
thereby reducing the likelihood of observing illegal
social interactions. Finally, the generalizability of the
measure may have been compromised by its focus on an
older urban center in the northeastern United States.
Because the physical structure of neighborhoods varies
greatly by geographic region, features that were salient
for observation in Baltimore may be irrelevant in other
parts of the United States or in other countries.
Another limitation of the measure is a lack of
dierentiation between those neighborhood character-
istics which result from the behavior of residents from
those which are a function of municipal policies or
actions. In truth, the distinction between the two is often
dicult to make. For example, the presence of trash and
grati has been attributed primarily to the behavior of
residents and interpreted as evidence that residents
dont care about themselves or their community
(Cohen et al., 2000). However, the level of physical
incivilities in a neighborhood is not only a function of
the will of neighborhood residents but also that of
municipal agencies which may not provide the same
level of services through such things as garbage
collection or property owners who live outside the
neighborhood who do not maintain their properties to
an acceptable standard (i.e., slum landlords). In our
own experience doing neighborhood studies, both
processes have been observed operating in select
neighborhoods. Although some characteristics of neigh-
borhoods, such as trac patterns, are more likely to be
under the control of municipal entities while others, such
as resident decoration and plantings, are under the
control of residents themselves, the distinction between
the two is not always clear. More research is needed to
tease apart how the relationship between residents and
broader municipal processes give rise to neighborhood
physical and social processes.
All observational methods of neighborhoods are limited
in their application as cross-sectional snapshots of a
community. Conducting a single observation does not
capture the dynamics of the neighborhood that may
change on a daily basis, based on season of the year, or
over the course of a number of years. Conducting multiple
observations in a strategic manner could help to capture
the dynamic characteristics of community life.
The appropriate application of any neighborhood
observational assessment requires that it be imbedded
within a larger neighborhood assessment protocol.
Interviews with community leaders and/or community
organizations, for example, would provide important
information for the valid interpretation of observational
data. Without a broader understanding of the commu-
nity context, there is a real risk that observational data
will be misinterpreted. Furthermore, an observational
assessment method should be combined with a resident-
report measure such as that reported by Coulton et al.
(1996) to yield the most comprehensive assessment of
the neighborhood environment. There are some aspects
of neighborhoods such as social interaction patterns and
collective ecacy which are most eectively tapped by
resident report although some proxies for them, such as
our measure of territoriality, may be evident through
observation.
The increasing number of studies in the literature
which examine the inuence of neighborhood context on
individual health outcomes represents an exciting devel-
opment which promises to improve our understanding
of the complex interplay between individuals and their
communities and its contribution to the production of
health. This area of inquiry is still in its infancy, and
researchers are struggling with how best to conceptualize
neighborhoods and to measure their most important
features. A careful consideration of theory when
developing measures and interpreting ndings will
provide a critical foundation as the eld progresses. The
results of such research would help us to design eective
community revitalization eorts and other policies to
improve the lives of families and their children.
Acknowledgements
This research was supported by the Maternal and
Child Health Bureau, Grant MCJ-240731-01, and by the
Smith-Richardson Foundation, Grant 9701-936. The
authors would like to thank Sherry Bazinet, Deborah
Brothers, Bennette Drummond-Fitzgerald, Crystal
Evans, Danielle Fagan, Yvette Flowers, Christopher
Jeter, Terese Lowery, Samara Mays, Elizabeth Mon-
tgomery, Karen Ricks, John Rogers, and Katy Schalla
for their assistance in conducting the neighborhood
observations. The authors would also like to thank Yi-
Hua Chen, Patricia Chore, and Crystal Evans for their
assistance with data management. Finally, the authors
are indebted to Drs. Ron Harrist and Louise Masse for
their patience in providing statistical advice.
References
Brodsky, A.E., 1996. Resilient single mothers in risky
neighborhoods: negative psychological sense of community.
Journal of Community Psychology 24, 347363.
Buckner, J.C., 1988. The development of an instrument to
measure neighborhood cohesion. American Journal of
Community Psychology 16, 771791.
M.O. Caughy et al. / Health & Place 7 (2001) 225236 235
Caughy, M.O., OCampo, P.J., Brodsky, A.E., 1999. Neighbor-
hoods, families, and children: implications for policy and
practice. Journal of Community Psychology 27, 615633.
Cohen, D., Spear, S., Scribner, R., Kissinger, P., Mason, K.,
Wildgen, J., 2000. Broken windows and the risk of
gonorrhea. American Journal of Public Health 90, 230236.
Coulton, C.J., Korbin, J.E., Su, M., 1996. Measuring neighbor-
hood context for young children in an urban area. American
Journal of Community Psychology 24, 532.
Duncan, G.J., Raudenbush, S.W., 1999. Assessing the eects of
context in studies of child and youth development. Educa-
tional Psychologist 34 (1), 2941.
Furstenberg, F.F., 1993. How families manage risk and
opportunity in dangerous neighborhoods. In: Wilson, W.J.
(Ed.), Sociology and the Public Agenda. Sage Publications,
Newbury Park, CA, pp. 231258.
Furstenberg, F.F., Hughes, M.E., 1994. The inuence of
neighborhoods on childrens development: a theoretical
perspective and research agenda. Paper presented at the
University of Wisconsin Institute for Research on Poverty
Conference on Indicators of Childrens Well-Being,
Bethesda, MD, November 1718.
Korbin, J.E., Coulton, C.J., 1997. Understanding the neighbor-
hood context for children and families: combining epidemio-
logical and ethnographic approaches. In: Brooks-Gunn, J.,
Duncan, G.J., Aber, J.L. (Eds.), Neighborhood Poverty:
Policy Implications in Studying Neighborhoods. Russell Sage
Foundation, New York.
McGuire, J.B., 1997. The reliability and validity of a
questionnaire describing neighborhood characteristics rele-
vant to families and young children living in urban areas.
Journal of Community Psychology 25, 551566.
NORC (National Opinion Research Center), 1995. Project for
Human Development in Chicago Neighborhoods; Harvard
Project 4709. SSO (Systematic Social Observation) Coding
Manual, June 1995.
OCampo, P., Wang, M., Xue, X., Caughy, M.O., 1997a.
Neighbourhood risk factors for low birthweight in Baltimore
City: A multi-level analysis. American Journal of Public
Health 87, 11131118.
OCampo, P., Caughy, M.O., Aronson, R., Xue, X., Wang, M.,
1997b. A comparison of two analytic methods for the
identication of neighborhoods as intervention sites for
community-based programs. Evaluation and Program Plan-
ning 20, 405414.
Perkins, D., Meeks, J., Taylor, R.B., 1992. The physical
environment of street blocks and resident perceptions of
crime and disorder: implications for theory and measure-
ment. Journal of Environmental Psychology 12, 2134.
Rabash, J., Browne, W., Healy, M., Cameron, B., Charlton, C.,
1999. MlwiN BETA version 1.10.0001. Multilevel Models
Project. Insitute of Education, University of London.
Raudenbush, S.W., Sampson, R.J., 1999. Ecometrics: toward a
science of assessment ecological settings, with application to
the systematic social observation of neighborhoods. Socio-
logical Methodology 29, 141.
Taylor, R.B., Gottfredson, S.D., Brower, S., 1984. Block crime
and fear: defensible space, local social ties, and territorial
functioning. Journal of Research in Crime and Delinquency
21, 303331.
Taylor, R.B., Shumaker, S.A., Gottfredson, S.D., 1985.
Neighborhood-level links between physical features and local
sentiments: deterioration, fear or crime,and condence.
Journal of Architectural Planning and Research 2, 261275.
M.O. Caughy et al. / Health & Place 7 (2001) 225236 236