Ross, Mirowsky / NEIGHBORHOOD

URBAN DISORDER
AFFAIRS REVIEW / January 1999

DISORDER AND DECAY

The Concept and Measurement of
Perceived Neighborhood Disorder
CATHERINE E. ROSS
JOHN MIROWSKY
The Ohio State University

The authors develop and assess a scale of perceived neighborhood disorder. The scale of neigh-
borhood disorder has high reliability, external validity, and shows interesting distinctions, and
overlaps between physical and social disorder. It also shows that order and disorder are two ends
of a single continuum.

Despite growing recognition of the importance of neighborhood disorder,

few systematic attempts have defined and measured the concept. Neighbor-
hood disorder may have consequences for individuals, reducing individual
well-being and increasing fear, mistrust, isolation, anger, anxiety, and de-
moralization. It also may have consequences for communities, reducing so-
cial ties among neighbors, which further undermines social control and leads
to more disorder in the neighborhood. Without intervention, Skogan (1990)
argued, disorder ultimately will lead to the decline of the neighborhood.
In Disorder and Decline, Skogan (1990) wrote eloquently of the process
of disorder and decline, but he never systematically defined or measured
community disorder. His work stimulates a number of questions, that we at-
tempt to answer. Is there one coherent disorder factor? Are physical and so-
cial disorder separate aspects of disorder, or do they both indicate one under-
lying concept? Are order and disorder separate concepts, or are they two ends
of a single continuum? Are perceptions of crime in the neighborhood distinct

AUTHORS NOTE: This research was supported by a grant from the National Institute of Men-
tal Health, Community, Crime and Health (RO1 MH51558) to Catherine E. Ross (principal
investigator) and Chester Britt (coprincipal investigator). We thank Karlyn Geis for her help and
Paul Bellair and Sung Joon Jang for their comments.
URBAN AFFAIRS REVIEW, Vol. 34, No. 3, January 1999 412-432
1999 Sage Publications, Inc.
412

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 413

from disorder? The goal of this research is to (1) systematically define the
concept of disorder and (2) measure and empirically assess the relationships
between perceptions of crime and disorder, order and disorder, and physical
and social disorder.1

THE CONCEPT OF PERCEIVED

NEIGHBORHOOD DISORDER

Our concept of neighborhood order and disorder derives from, and ex-
pands on, Skogans (1990) research. Perceived neighborhood disorder, as
conceptualized here, refers to visible cues indicating a lack of order and so-
cial control in the community. Order is a state of peace, safety, and obser-
vance of the law, and control is an act of maintaining this order. Order and
control are indicated by visible cues that residents perceive.
These cues are both social and physical (Skogan 1986, 1990; Skogan and
Maxfield 1981; Taylor and Hale 1986; Taylor and Shumaker 1990). We de-
fine social disorder as signs indicating a lack of social control that involve
people. Visible signs of social disorder include fights and trouble among
neighbors and the presence of people hanging out on the streets, drinking,
taking drugs, panhandling, and creating a sense of danger. Physical disorder
refers to the overall physical appearance of a neighborhood. Places with high
levels of physical disorder are noisy, dirty, and run down; many buildings are
in disrepair or abandoned; and vandalism and graffiti are common. Physical
disordersuch as litter, graffiti, and vandalismalso indicates that social
control has broken down.
Social and physical disorder are conceptualized on a continuum, with high
levels of order on one end and disorder on the other. Visible signs of order in-
clude a clean, safe, quiet neighborhood where buildings are in good repair,
police protection is good, crime is low, neighbors watch out for one another,
and not a lot of young people are hanging out.
The concept of perceived neighborhood disorder is related to concepts of
incivilities, lack of social control in the community, and neighborhood prob-
lems (Garofalo and Laub 1978; Lewis and Maxfield 1980; Lewis and Salem
1986; Moore and Trojanowicz 1988; Rohe and Burby 1988). Communities
with high levels of disorder are characterized by deviance, noise, vandalism,
drug use, trouble with neighbors, and other incivilities associated with a gen-
eral breakdown of social control (Lewis and Salem 1986). Crime is high in
these neighborhoods, but peoples concerns go beyond crime and victimiza-
tion (Garofalo and Laub 1978). Few residents are directly victimized, but
people see the cues of disorder each time they walk past a group of teenage

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

414 URBAN AFFAIRS REVIEW / January 1999

boys hanging out on the street, a boarded-up building or vacant lot, and
drunks or panhandlers (Skogan 1986).
Perceived neighborhood disorder refers to conditions and activi-
tiesboth minor and major, noncriminal and criminalthat residents per-
ceive to be cues or signs of the breakdown of social control (Skogan and Max-
field 1981; Taylor and Hale 1986). Although some aspects of neighborhood
disorder fall into the realm of criminal activity, other aspectssuch as teen-
agers hanging out on the street, buildings that are abandoned or in disrepair,
noise, litter, and grimedo not. These activities and conditions may violate
norms without violating the law (Bursik and Grasmick 1993). Moreover,
most of the criminal activities that Skogan and others discuss are relatively
minor and without specific victimscrimes such as graffiti and public drink-
ing. Although some researchers attempt to separate neighborhood disorder,
problems, and incivilities from major crimes such as murder, robbery, bur-
glary, mugging, and rape (Skogan 1990), it is difficult to separate crime from
disorder (Bursik and Grasmick 1993). According to Lewis and Salem (1986),
perceived increases in crime are among the clearest indicators of social dis-
order in an area. Major crimes are rare and often unseen, but they form the
extreme end of the disorder continuum. Crime is included in related concepts
such as neighborhood problems. Lee (1981) included a general measure of
crime, and Rohe and Stegman (1994) included burglary and muggings as in-
dicators of neighborhood problems. We examine whether reports of crime in
the neighborhood indicate disorder.
We conceptualize disorder as a characteristic of the neighborhood in
which an individual lives. It is a description of a place by its residents in terms
of the appearance of disorder. To describe a neighborhood, a person must be
aware of it and perceive it, and although two people in the same neighborhood
might describe it somewhat differently, judgments of any two individuals
will be positively correlated because both are describing an objective place.
Correlations between respondentsreports of disorder in their neighborhoods
and independent assessments by researchers are moderate to high. Perkins
and Taylor (1996) found correlations that ranged from a low of .20 between
independent observations of young men hanging out on the streets and re-
spondents reports of physical disorder to a high of .76 between observations
of residential physical disorder and respondentsreports of physical disorder.
Because both researcher and resident assessments are reports of what they
observe, however, it is not the case that the first is more objective. One is an
outsiders subjective view, and the other is a residents. Furthermore, the con-
struct validity of respondents assessments is typically higher than those of
the researchers (Perkins et al. 1990; Simcha-Fagan and Schwartz 1986),

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 415

presumably because residents have more experience with conditions in their

neighborhood and can describe them at least as well as an outsider.
We measure perceived disorder with the amount of agreement with state-
ments about ones neighborhood such as, there is a lot of graffiti in my
neighborhood, my neighborhood is noisy, vandalism is common in my
neighborhood, there are a lot of abandoned buildings in my neighborhood,
there are too many people hanging out on the streets near my home, there
is a lot of crime in my neighborhood, and there is too much drug use in my
neighborhood. On the other end of the continuum, order is measured with
the amount of agreement with statements such as, my neighborhood is safe,
my neighborhood is clean, and people in my neighborhood take good care
of their houses and apartments. The appendix shows items in previous mea-
sures of neighborhood incivilities, problems, and disorder from which we de-
rive disorder items. Notice, however, that each measures some aspects of dis-
order, but none measures social or physical order. To measure order, we
sometimes used the semantic opposites of disorder, such as safe versus dan-
gerous and clean versus dirty, and sometimes more directly operationalized
the concept of informal social order with statements such as, in my neigh-
borhood, people watch out for each other.
Our measure of perceived disorder attempts to operationalize the concept
described earlier and has the following characteristics. It includes both social
and physical disorder, reports of crime in the neighborhood, and both ends of
the continuumorder and disorder. This brings up the following empirical
questions. Do physical and social disorder indicate one underlying concept,
or are they distinct concepts? Are order and disorder two ends of a single con-
tinuum or distinct concepts? Is perceived crime in the neighborhood an as-
pect of disorder, or is it distinct?

SAMPLE AND DESCRIPTION

OF INDEPENDENT VARIABLES

To address these research questions, we use the 1995 Survey of Commu-

nity, Crime and Health (CCH), a probability sample of Illinois households.
Respondents were interviewed by telephone. They were selected using a pre-
screened random-digit dialing method that increases the rate of contacting
eligible numbers (i.e., decreases the rate of contacting business and nonwork-
ing numbers) and decreases standard errors compared to the standard
Mitofsky-Waksberg method while producing a sample with the same demo-
graphic profile (Lund and Wright 1994; Waksberg 1978). The survey was

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

416 URBAN AFFAIRS REVIEW / January 1999

limited to English-speaking adults. The adult (age 18 or older) with the most
recent birthday was selected as the respondent, which is an efficient method
to select a respondent in the household randomly. Up to 10 callbacks were
made to select and contact a respondent, and up to 10 callbacks were made to
complete the interview once contact had been made. Interviews were com-
pleted with 73.1% of the eligible persons who were contacted. The final sam-
ple has 2,482 respondents, ranging in age from 18 to 92, with an average age
of 45. The mean education level is 13.8 years, and the mean family income is
$48,274 (median is $40,000). Of the respondents, 41% are male, 84% are
white, and 56% are married. This profile is comparable to the population of
Illinois, where 78.3% of the population were white, 14.8% were black, and
the median family income was $38,664 in 1990. However, the sample over-
represents women: In Illinois in 1990, 48.5% of the population were male.
Of the respondents, 16% live in Chicago, 30% in the suburbs of Chicago,
12% in small cities, 27% in small towns, and 14% in rural areas of Illinois.
Place of residence is based on respondentsreports. We contrast living in Chi-
cago with all other areas of residence. Age and education are measured in
years. Family income is measured in thousands of dollars. Race and sex are
coded as dummy variables. Race is coded 1 for blacks, 0 for whites. Sex is
coded 1 for females, 0 for males.

THE MEASUREMENT OF
PERCEIVED NEIGHBORHOOD DISORDER

Table 1 lists all the items in our perceived neighborhood disorder scale. It
shows the theoretical distinction between physical and social disorder and or-
der. It also shows an exploratory factor analysis.
The exploratory factor analysis shows that all items indicating physical
and social order and disorder load on one factor. The one item with a loading
under .4 on the first factor is police protection, our only indicator of formal
social control. A scale comprising all the items has an alpha reliability of
.921. However, two factors were extracted, and they distinguish between dis-
order (or negatively worded items) and order (or positively worded items).
Although the order items load on the first factor over .4, they have higher
loadings on the second factor. Unfortunately, exploratory factor analysis can-
not distinguish methods factors (in which positively and negatively worded
items load on distinct factors because of response tendencies such as the ten-
dency to agree) from substantive factors (in which order and disorder load on
distinct factors because they are conceptually distinct). Covariance structure

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 417

TABLE 1: Items in the Perceived Neighborhood Disorder Scale, Theoretical Dis-

tinction Between Physical and Social Disorder and Order, and Ex-
ploratory Factor Analysis

Factor Loadingsa
Factor 1 Factor 2

Physical disorder
There is a lot of graffiti in my neighborhood. .811 .472
My neighborhood is noisy. .750 .516
Vandalism is common in my neighborhood. .834 .509
There are lot of abandoned buildings in my neighborhood. .739 .419
Physical order
My neighborhood is clean. .617 .712
People in my neighborhood take good care of their houses and
apartments. .573 .663
Social disorder
There are too many people hanging around on the streets near
my home. .748 .533
There is too much drug use in my neighborhood. .810 .512
There is too much alcohol use in my neighborhood. .744 .489
Im always having trouble with my neighbors. .436 .370
There is a lot of crime in my neighborhood. .845 .577
Social order
In my neighborhood, people watch out for each other. .429 .766
The police protection in my neighborhood is adequate. .374 .657
My neighborhood is safe. .601 .738
I can trust most people in my neighborhood. .507 .756

Alpha reliability .921

Mean 1.816

NOTE: All items are scored so that a high score indicates disorder. Disorder items are scored
strongly disagree (1), disagree (2), agree (3), and strongly agree (4). Order items are scored
strongly agree (1), agree (2), disagree (3), and strongly disagree (4). All items load over .4 on the
first factor, except for police protection, but all the order questions have higher loadings on factor 2
than factor 1.
a. Factor loadings from principle components analysis specifying a minimum eigenvalue of 1,
structure matrix, oblimin rotation.

models that explicitly model the tendency to agree with statements regardless
of content are necessary to untangle the two possibilities.
Thus we next specified a covariance structure model using EQS (Bentler
1989).2 The measurement equations correspond to a confirmatory factor
model and allow us to test whether disorder is one coherent factor or whether
physical and social disorder, order and disorder, and crime and disorder are

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

418 URBAN AFFAIRS REVIEW / January 1999

distinct. The structural model allows assessment of construct validity. By ex-

amining the antecedents of disorder, we can see whether the measure works
as theory would suggest. For instance, do individuals who live in Chicago ex-
perience more neighborhood disorder than people in nonurban areas?
We began by specifying one neighborhood disorder factor that did not fit
well. In the initial measurement model, we specified one substantive factor
(neighborhood disorder) and one methods factor (agreement bias). The
specification of one disorder factor never produced a fit statistic of .9 (the
standard cutoff for a good fit), and Lagrange Multiplier (LM) tests indicated
many correlated residuals. The fit statistics for a one-factor model are near
.88 for all three fit indexes (Bentler-Bonett Normed, Nonnormed, and
Comparative Fit), and the chi-square equals 2,088 with 208 degrees of free-
dom, p < .001. The one-factor model indicates many correlations with unique
components, including those between abandoned buildings and cared-for
homes and apartments, between abandoned buildings and a clean neighbor-
hood, and between abandoned buildings and graffiti. Freeing too many of
these produced convergence problems. The loadings of all items on disorder
are significant, but the three lowest are cared-for homes and apartments
(.674), abandoned buildings (.557), and trouble with neighbors (.414). (We
return to trouble with neighbors later, as indications are that this is a poor
item.) Both the correlations among unique components and the loadings
point to a second factor indicating physical decay.
Therefore, we specified a second substantive factor, which we call decay.
Decay, theoretically and empirically, consists of the purest physical disorder
items. For the model to be identified, we need at least one item that loads on
decay alone, and it is preferable to have twoone positively and one nega-
tively worded. Thus we chose the two items suggested earlier that relate to
buildingsa lot of abandoned buildings and, on the other end of the contin-
uum, well-maintained and cared-for buildings. Once two substantive factors
(disorder and decay) and one methods factor (agreement bias) are specified,
the final model fits well, about .95 for all three fit statistics.
To arrive at this final model, we began by specifying a baseline model with
the following properties. The measurement model specified all metric load-
ings as +1 or 1. It also assumed no correlations of the unique components
(measurement errors) with each other, with equation residuals, with factors,
or with other indicators. The baseline structural model specified all correla-
tions between residuals as zero. It specified all direct effects as nonzero if
consistent with the order assumptions and as zero otherwise. (Gender had no
significant effects in any regression equation, so it was eliminated [all effects
set to zero].) The fit of the model was improved by making the following ad-
justments (in the order shown): fix to zero any structural coefficient that is not

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 419

significantly different from zero, free any correlation between residuals that
the LM test suggests is not zero, free any factor loading that the LM test sug-
gests is not +1 or 1, and free any correlation between unique components
that the LM test suggests is not zero. Thus, the final model is pruned of insig-
nificant effects and relaxes restrictions on the loadings and error correlations
that appear to be inconsistent with the observed covariances. The final mea-
surement model is shown in Table 2, and the final structural model is shown
in Table 3. Both show metric coefficients. Figure 1 shows standardized coef-
ficients from the final model of theoretical interest (not shown are the agree-
ment factor and correlations among exogenous variables).
The final measurement model shown in Table 2 and Figure 1 shows that
most indicators load on the disorder factor, four indicators load on both disor-
der and decay, and two indicators load on decay alone. Reports of people
hanging out on the streets, crime, drug and alcohol use, trouble with neigh-
bors and on the other end of the continuum, reports of lack of police protec-
tion, and the lack of agreement that people watch out for each other, that the
neighborhood is safe, and that neighbors can be trusted signify disorder. Re-
ports of graffiti, noise, vandalism, and lack of cleanliness indicate disorder
and decay. Reports of abandoned buildings and the absence of well-
maintained buildings indicate decay.
Disorder includes social disorder and physical cues that social order is
weak, such as graffiti, noise, dirt, and vandalism. Physical decay includes
physical problems with buildings, and these same cuesgraffiti, noise, dirt,
and vandalismalso indicate social disorder. Graffiti, noise, vandalism, and
dirt indicate disorder and decay.3
Once the tendency to agree to statements, regardless of content, is mod-
eled explicitly and adjusted, there is no evidence of distinct order and disor-
der concepts. All items, both negatively worded disorder items and positively
worded order items, load positively on agreement bias.
Reports of crime in the neighborhood load highly and significantly on
disorder.
The measurement model also shows a number of correlated residuals. The
latent factors of disorder and decay are specified as a block on the same
causal level. Thus the association between them is shown as a correlated dis-
turbance. This is high.777. The residuals of a number of indicators are also
significantly correlated: the residuals of drug and alcohol use, of people
watching out for each other and trustworthy neighbors, and of a clean neigh-
borhood and cared-for houses and apartments.
The final structural model shows determinants of living in neighborhoods
characterized by perceived disorder and decay. Individuals who live in Chi-
cago are much more likely to report that they live in a neighborhood with a lot

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

420 URBAN AFFAIRS REVIEW / January 1999

TABLE 2: Measurement Equations of the Latent-Factor Covariance Structure

Model

Latent Factor
Observed Measure Disorder Decay Agreement

Hanging out 1.345 1.000

(10.781)
Crime 1.345 1.000
(10.781)
Drug use 1.345 1.000
(10.781)
Alcohol use 1.221 1.000
(31.419)
Trouble with neighbors .482 1.000
(17.267)
Police protection OK .859 1.000
(4.029)
People watch out 1.000 1.000

Trustworthy neighbors 1.000 1.000

Safe 1.194 1.000

(5.879)
Graffiti .181 1.000 1.000
(28.241)
Noise 1.000 .231 1.000
(17.477)
Vandalism 1.000 .247 1.520
(19.815) (3.377)
Clean .162 1.000 1.000
(36.435)
Cared-for houses and apartments 1.000 1.000

Abandoned buildings 1.000 1.000

NOTE: Metric coefficients are presented with t-values in parentheses. The t-values represent the
gap measured in standard errors between the estimated coefficient and the hypothesized one. In
the measurement model, all hypothesized metric coefficients are 1 or 1, depending on whether
the item is taken as a positive or negative sign of the factor. The model assumes that the residuals
of equations and the unique components of measures are uncorrelated unless a Lagrange Multi-
plier (LM) test suggests otherwise (Bentler 1989). The LM test shows that four possible residual
correlations are significantly different from zero: residuals for disorder and decay correlate .777,
residuals for alcohol and drug use correlate .387, residuals for trustworthy neighbors and neigh-
bors who watch out for each other correlate .167, and residuals for a clean neighborhood and
cared-for houses and apartments correlate .218. There is a significant direct effect of Chicago on
graffiti, b = .341, t = 8.327. Coefficients of disorder on people hanging out, crime, and drug use
are constrained to be equal. Originally, they were all fixed to 1 or 1, but the LM test suggested
that they were different from 1. When freed, they were all larger than 1 and within each others
standard errors, so they were constrained to be equal.

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 421

TABLE 3: Construct Equations of the Covariance Structure Model

Dependent
Independent Disorder Decay Agreement

Age .003 .002

(5.048) (6.372)
Black .350 .249
(8.608) (6.430)
Education .031 .040 .004
(6.982) (9.358) (2.311)
Household income .001 .002
( 4.648) (6.230)
Chicago .649 .396
(19.776) (12.752)

NOTE: Metric coefficients are presented with t-values in parentheses. The probability value for
the overall fit of the model is .001 (chi-square = 939, df = 181). The fit indexes are .949 (Bentler-
Bonett Normed), .947 (Bentler-Bonett Nonnormed), and .958 (Comparative Fit Index). Blanks
represent coefficients fixed to zero. Measurement equations and the correlations of measurement
errors and construct residuals are given in Table 2.

of disorder and decay than are residents of suburbs, small cities, small towns,
and rural areas. Blacks are more likely than whites to live in neighborhoods
characterized by disorder and decay, and persons with high household in-
comes and educational attainment are less likely. Older people are less likely
to live in disordered neighborhoods but equally as likely as younger people to
live in decaying neighborhoods.4 Apart from its association with the overall
level of disorder and decay, Chicago residence also increases the likelihood
of reporting that one lives in a neighborhood with a lot of graffiti.
Our model specifies two distinct factors. Next we test whether there are
two factors or one. Because disorder and decay are highly correlated and be-
cause four indicatorsgraffiti, noise, vandalism, and a clean neighbor-
hoodload on both, we end by revisiting our general research question of
whether there is one disorder factor. Once we have a model with two factors,
we can modify that model to test whether the two factors are really distinct.
We test whether there is a perfect correlation between the two. That is, if we
regress one on the other, do we explain all of the variance (given that EQS
models distinguish measurement error from residual variance)? To answer
this question, we first respecify the covariance of disorder and decay to zero
and instead model a direct effect of disorder on decay. (The path coefficient is
.803, the fit statistics are the same as those presented earlier, and the degrees

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 422

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

Figure 1: Structural and Measurement Model of Neighborhood Disorder and Decay
NOTE: Not shown is the agreement tendency factor. Table 2 shows that all disorder and decay items load positively on agreement, and all but one (vandalism)
have unstandardized coefficients set to 1. Also not shown are two significant structural effects: The effect of age on agreement is .202, and the effect of educa-
tion on agreement is .074. Not shown are the correlations among exogenous variables, which follow: age with black = .043, education with black = .089,
income with black = .073, Chicago with black = .348, education with age = .178, income with age = .124, Chicago with age = .085, income with educa-
tion = .286, Chicago with education = .007, and Chicago with income = .018.
 Ross, Mirowsky / NEIGHBORHOOD DISORDER 423

of freedom are the same.) We test the hypothesis that disorder and decay are
the same factor (differing only by a scaling constant) by setting the residual
variance of decay to zero. The increment in chi-square associated with this
restriction is 281.437, df = 1, which is highly significant at p < .001, indicat-
ing that the restriction significantly worsens the fit. Disorder and decay are
distinct although highly related factors.

DISTINCTIONS AND OVERLAPS BETWEEN

PHYSICAL DISORDER, SOCIAL DISORDER, AND DECAY

Disorder and decay are highly correlated, and they share many indicators.
Disorder and decay have a correlated disturbance of .777. Furthermore, re-
ports of graffiti, noise, and vandalism and reports that the neighborhood is not
clean (i.e., dirt, trash, and garbage litter the streets) indicate both disorder and
decay. These four items are indicators of physical disorder in that they refer to
the physical look and sound of a neighborhood. But they are also cues that
people were there. Noise is created by people, their stereos, and their cars.
Vandalism and graffiti are caused by people. Maybe the neighbors did not see
the acts, but they see the consequencesbroken windows, charred rubbish in
a vacant lot, uprooted bushes, and symbols spray-painted on the sides of
buildings. These cues are physical, but they indicate the presence of people.
They indicate social disorder and physical decay.
Abandoned buildings and their oppositecared-for and well-maintained
buildingsmost purely indicate physical decay. Maybe people understand
that an abandoned, vacant, or run-down building in their neighborhood, espe-
cially if it is a large apartment building, is often a consequence of an absentee
landlord who never visits the neighborhood, and the presence of these build-
ings is not a direct indicator of the breakdown of social control in their neigh-
borhood. Unlike the local teenagers who hang out on the street drinking or
taking drugs and engaging in vandalism and graffiti, building repair may have
its causes outside the neighborhood. Unless people own their own homes,
some (although not all) of the problems with buildings are outside neighbor-
hood control.
For practical purposes, researchers can combine disorder and decay, or
they can separate the two items that are pure physical decay indicators. The
total scale has an alpha reliability of .921. The scale with abandoned build-
ings and well-maintained houses and apartments eliminated has an alpha reli-
ability of .913, slightly lower. Both have very high reliabilities. Part of the de-
cision depends on the researchers purposes. If the concept of interest refers
only to cues that social order has broken down in the neighborhood, then a

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

424 URBAN AFFAIRS REVIEW / January 1999

researcher could eliminate the two physical decay items from the disorder
scale and examine them separately.
Distinguishing disorder and decay shows an interesting finding relating to
age. We find that older persons are less likely than younger persons to live in a
neighborhood characterized by disorder. However, older persons are no less
likely to report that they live in a neighborhood characterized by physical de-
cay. Maybe older adults help maintain social order in their neighborhoods be-
cause they are home during the day, are more likely to visit and maintain so-
cial ties with their neighbors that help maintain control in the community, and
do not have teenage children. However, older adults are no less likely to live
in a neighborhood characterized by physical decay. Older persons have lower
incomes than middle-aged adults, and their houses are older (Siegel 1993).
This may be part of the reason they are equally likely as others to live in places
that are not well maintained. Finally, although Chicago has more physical de-
cay than smaller cities, suburbs, small towns, and rural areas, it may be the
case that older persons who live in decaying neighborhoods live in small
towns and rural areas (see the negative correlation between age and living in
Chicago in the notes to Figure 1). Some rural communities may have a lot of
decay because of economic history rather than social disorder.
On the whole, social and physical aspects of perceived disorder indicate
one underlying concept. Many of the physical aspects of a neighborhood,
such as graffiti, noise, vandalism, dirt, and grime, are indicators of the break-
down of social control. They are clear cues to residents that people are in-
volved. Thus the distinction between social and physical disorder is not
clear-cut. Only two items relating to building disrepair and vacancy are pure
indicators of physical decay.
Order and disorder are not separate concepts. Although the exploratory
factor analysis gave some indication that they might be, this resulted from a
methods or response bias factor. The tendency to agree with items regardless
of their content gives the misleading appearance that there might be distinct
order and disorder concepts. Once the tendency to agree is modeled explicitly
and accounted for statistically in the covariance structure models, there is no
evidence that order and disorder are separate concepts. Order and disorder
form two ends of a single continuum.
Exploratory factor analysis as a method to assess psychometric properties
of a scale is susceptible to cross-cutting factors. As it does in this case, ex-
ploratory factor analysis often produces methods factors based on positively
and negatively worded questions. Some people tend to agree, irrespective of
the content of the question. The agreement factor crosscuts the concept of in-
terest and must be modeled explicitly with covariance structure models to
distinguish methodological response tendencies from theoretical concepts

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 425

(Mirowsky and Ross 1996). Furthermore, exploratory factor analysis cannot

show highly related factors that share some indicators and not others. Covari-
ance structure models such as EQS solve both problems (Bentler 1989).
Some informal social control items appear to be good indicators of neigh-
borhood order. Safe neighborhoods, in which neighbors watch out for one
another and can be trusted, all have moderately high negative loadings on the
disorder factor (.739, .579, and .605). However, the worst item in our
scale is trouble with neighbors, an indicator of the lack of informal social
control. It has the lowest loading of any item (.331) that only indicates one
concept, although it is still significant. Researchers might consider other
ways to measure informal social control and its absence directly to see if bet-
ter indicators can be developed. For instance, Gates and Rohe (1987) in-
cluded being hassled by neighbors, and Covington and Taylor (1991) and
Perkins and Taylor (1996) included fights and arguments in their measures.
Conceptually, we think that watching out for others in the neighborhood is
the most direct indicator of informal social control. Of course, other indica-
tors are signs that informal social control is weakcues such as people hang-
ing out on the street, people taking drugs and drinking, and vandalism, graf-
fiti, litter, and noise in the neighborhoodbut these are more indirect signs.
Police protection, our only measure of formal social control, has a smaller
standardized loading than most informal control items but is still significant
(.499). Some researchers might want to examine it separately, depending on
their research questions. For instance, if researchers are interested in the
question of whether formal social control can compensate for a lack of infor-
mal social control in a neighborhood, proposing that police protection in an
otherwise disordered neighborhood decreases fear among residents, then
they would keep it separate.
In general, we recommend attention to whether individual indicators
should be eliminated, depending on whether they are potentially confounded
with independent or dependent variables of interest. For example, if the
research question of interest is whether neighborhood disorder affects indi-
vidual mistrust of others, then the item I can trust most people in my neigh-
borhood could be eliminated so as not to confound the independent and
dependent variables. If the research question of interest is whether neighbor-
hood disorder affects social ties among neighbors, then the item In my
neighborhood people watch out for each other could be eliminated. If the
research question is whether disorder promotes more serious crime, then the
item There is a lot of crime in my neighborhood could be eliminated.
Attention to potential confounders is an issue in any scale application.
Our scale of perceived neighborhood disorder has high reliability, exter-
nal validity, and shows interesting distinctions and overlaps between physical

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

426 URBAN AFFAIRS REVIEW / January 1999

and social disorder. It also shows that order and disorder are two ends of a sin-
gle continuum. For many purposes, researchers could use all 15 items.
For a short 10-item scale of perceived neighborhood disorder, we recom-
mend the 4 items that overlap the two scalesgraffiti, noise, vandalism, and
clean neighborhoodand the items with the largest loadings on disor-
derpeople hanging out, crime, drug use, alcohol use, trustworthy neigh-
bors, and a safe neighborhood. Crime has the highest loading of any indicator
on disorder. We recommend its inclusion. The 10-item scale has an alpha reli-
ability of .915.
Theoretically, the consequences of neighborhood disorder for individuals
are great (Garofalo and Laub 1978). Living in a neighborhood characterized
by disorderby crime, vandalism, graffiti, loitering, litter, noise, alcohol,
and drugsmay adversely affect individual well-being. In contrast to those
who live in safe, clean neighborhoods, where police protection is good and
neighbors watch out for one another, people who live in neighborhoods char-
acterized by disorder and a lack of social control may have high levels of fear,
mistrust, perceived powerlessness, isolation, restricted outdoor activity,
anger, anxiety, depression, and poor health (Ross 1993). Although the conse-
quences of neighborhood disorder are theoretically far reaching, empirically
there is little evidence, most of it showing that neighborhood incivilities
increase fear (Covington and Taylor 1991; Lewis and Maxfield 1980; Perkins
et al. 1990; Perkins and Taylor 1996; Rohe and Burby 1988; Taylor and Cov-
ington 1993). The negative consequences probably do not end with fear,
however. Fear may lead residents to believe that other people cannot be
trusted, to be suspicious, and to think that others are out to harm them
(Fischer 1982; Ross and Jang 1996). Fear and mistrust may increase social
isolation because people who mistrust others are unlikely to form social ties
with them, and people who are afraid rarely leave the house to visit with oth-
ers. Fear also may lead people to restrict their outdoor physical activities. The
likely consequences are anxiety, depression, and poor health on the individ-
ual level and further erosion of social control and neighborhood order on the
community level.
The potential consequences for individual well-being are negative enough
in their own right to warrant attention, and they have implications for further
neighborhood deterioration. Individuals who are mistrusting, fearful, and
isolated in their homes are unlikely to form social ties with neighbors that
could reassert social control in the neighborhood. Individuals who feel pow-
erless are unlikely to attempt to solve neighborhood problems. Individuals
who rarely go outside cannot watch the outdoor activities of teenagers. Thus
neighborhoods affect individuals, which in turn shape neighborhoods, in a
downward spiral (Skogan 1990).

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 427

APPENDIX
Items in Previous Disorder, Incivility,
and Neighborhood Problems Scales
Skogan (1990): Disorder
Loitering
Drug use
Vandalism
Gang activity
Public drinking
Harassment on street
Abandoned buildings
Garbage, litter, trash, and junk
Lewis and Maxfield (1980): Incivility
Teenagers hanging out on streets
Abandoned or burned-out buildings
Drug use
Vandalism, graffiti, and broken windows
Rohe and Burby (1988): Incivility
Drug use
Gangs
Abandoned houses
Poor building and grounds maintenance
Perkins et al. (1993): Incivility
Drug addicts
Prostitutes
Youth gangs
Homeless people loitering on the street
Graffiti
Vandalism
Litter
Unkempt housing
Taylor and Covington (1993): Incivility
Small groups on street
Males on street
Graffiti
Vacant houses
Litter
Housing density
Covington and Taylor (1991): Incivility
Public drinking
Drug use
Fighting
Arguing

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

428 URBAN AFFAIRS REVIEW / January 1999

Litter
Graffiti
Abandoned lots
Vacant housing
LaGrange, Ferraro, and Supancic (1992): Incivility
Trash and litter
Loose dogs
Bad neighbors
Graffiti
Vacant houses
Unsupervised youth
Noise
Abandoned cars
Gates and Rohe (1987): Neighborhood problems
Noisy neighbors
Stray dogs
Improper garbage disposal
Poor property maintenance
Hassled by neighbors
Landlords do not care about buildings
Lee (1981): Neighborhood problems
Noise
Crime
Trash, litter, and junk
Run-down housing
Inadequate police protection
Abandoned buildings
Taylor and Hale (1986): Neighborhood problems
Vacant lots
Empty houses
Negligent slumlords
Neglect of lawns and garbage by neighbors
Noisy neighbors
Drug use
Loitering teens
Rohe and Stegman (1994): Neighborhood problems
Run-down buildings
Litter
Harassment
Drug use
Burglary
Muggings
Perkins and Taylor (1996): Disorder
Loitering youths
Harassment

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 429

Fights and arguments

Drug dealing
Vandalism
Vacant housing
Unkempt property
Litter
Trashed vacant lots

NOTES

1. Skogan (1990, 21) defined the concept of disorder as direct, behavioral evidence of dis-
organization, but this leaves unanswered the question of what is disorganization. He described
two components of disordersocial and physicaldefining social disorder as that which
involves specific events or activities and physical disorders as enduring, day-to-day aspects of
the environment, but it is not clear why a temporal distinction between events and ongoing con-
ditions is a distinction between social and physical disorder. Specifically, Skogan described
social disorder as loitering, drug use, vandalism, gang activity, public drinking, and street harass-
ment and physical disorder as abandoned buildings, garbage, litter, and junk, but he never
described psychometric properties or the external validity of an index. These indicators are
derived from residents descriptions of their major concerns with their neighborhood, although
noise is considered a major concern but not included. Furthermore, he called vandalism social
disorder in the introduction and in his description of disorder (pp. 4, 51), but he called vandalism
physical disorder when he described the difference between the two (p. 37). This may be an indi-
cation either that social and physical disorder are not distinct or that vandalism is a good measure
of both. Furthermore, Skogan emphasized the importance of order as the opposite pole of disorder
but made no attempt to measure it, possibly because he considers it a slippery concept (p. 4). On
the other hand, Skogan considered major crime and disorder separate concepts.
2. Covariance structure modeling combines the methods of path analysis with those of con-
firmatory factor analysis. Each model comprises a set of equations logically divided into two
subsets. The structural equations correspond to the regression equations in a path model. Each
equation describes the regression of a dependent variable on a set of independent variables. The
dependent variables form an interlocked set either because one appears as an independent vari-
able in the others equation or because their residuals covary (i.e., the partial correlation between
them is not zero). In our model, the structural equations predict disorder, decay, and agreement
bias from sociodemographic and socioeconomic status. The disorder and decay residuals covary
because the correlation between the two factors is not simply spurious because of the effects of
sociodemographic and socioeconomic status. The measurement equations correspond to the
loadings of indicators on factors in a confirmatory factor model. Each equation describes the
regression of the response to a specific question on the latent factors that may affect the response.
Our model posits three latent factors: disorder, decay, and agreement tendency. Some indicators,
such as graffiti or vandalism, may indicate both disorder and decay. Some, such as drug use and
people hanging out, may indicate disorder but do not indicate decay, except indirectly through
disorders correlation with decay. Others, such as abandoned buildings and dilapidated housing,
may indicate decay but do not indicate disorder, except through decays correlation with
disorder.

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

430 URBAN AFFAIRS REVIEW / January 1999

Covariance structure modeling programs such as EQS produce full-information maximum

likelihood estimates of the parameters in the equations. The program finds the set of coefficient
values most likely to have produced the entire set of observed variances and covariances. The fit
statistics measure how well the model reproduces the full set of observed variances and covari-
ances. The LM test identifies fixed parameters that may be causing poor fit. In our model,
parameters may be fixed to 0 (as when we assume that disorder has no direct effect on the pres-
ence of abandoned buildings), to 1 (as when we assume that public drug use and alcohol use are
equally signs of the presence of disorder), or to 1 (as when we assume that trustworthy neigh-
bors and people watching out for each other are equally signs of the absence of disorder). The
analysis seeks the model with the fewest free parametersand thus the fewest estimated coeffi-
cientsthat reproduces the variances and covariances adequately. The modeling process begins
with the simplest (most restricted) model that embodies the proposed relationships among indi-
cators, latent factors, and exogenous variables. A restriction is relaxed (a parameter is freed)
when the LM test indicates that the restriction implies variances and covariances that are incon-
sistent with those observed.
3. Interestingly, some of these indicators, which are physical aspects of the environment,
have higher loadings on disorder and some on decay. Graffiti loads .566 on the purely physical
aspects of neighborhood decay and .123 on social disorder. A clean neighborhood loads .607 on
decay and .118 on disorder. On the other hand, vandalism loads .129 on physical decay and .626
on disorder. Noise loads .106 on decay and .552 on disorder. Possibly, noise and vandalism are
greater perceptual cues of the presence of people.
4. Older persons are more likely than younger persons to agree to statements, regardless of
content, and the well-educated are less likely to argue than the poorly educated (see Table 3).

REFERENCES

Bentler, P. M. 1989. EQS structural equations program manual. Los Angeles: BMDP.
Bursik, R. J., Jr., and H. G. Grasmick. 1993. Neighborhoods and crime. New York: Lexington
Books.
Covington, J., and R. B. Taylor. 1991. Fear of crime in urban residential neighborhoods: Implica-
tions of between- and within-neighborhood sources for current models. Sociological Quar-
terly 32 (2): 231-49.
Fischer, C. S. 1982. To dwell among friends: Personal networks in town and city. Chicago: Univ.
of Chicago Press.
Garofalo, J., and J. Laub. 1978. The fear of crime: Broadening our perspective. Victimology: An
International Journal 3 (3/4): 242-53.
Gates, L. B., and W. M. Rohe. 1987. Fear and reactions to crime: A revised model. Urban Affairs
Quarterly 22 (3): 425-53.
LaGrange, R. L., K. F. Ferraro, and M. Supancic. 1992. Perceived risk and fear of crime: Role of
social and physical incivilities. Journal of Research in Crime and Delinquency 29 (3):
311-34.
Lee, B. A. 1981. The urban unease revisited: Perceptions of local safety and neighborhood satis-
faction among metropolitan residents. Social Science Quarterly 62 (4): 611-29.
Lewis, D. A., and M. G. Maxfield. 1980. Fear in the neighborhoods: An investigation of the
impact of crime. Journal of Research in Crime and Delinquency 17 (2): 160-89.

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

 Ross, Mirowsky / NEIGHBORHOOD DISORDER 431

Lewis, D. A., and G. Salem. 1986. Fear of crime: Incivility and the production of a social prob-
lem. New Brunswick, NJ: Transaction Books.
Lund, L., and W. E. Wright. 1994. Mitofsky-Waksberg vs. screened random digit dial: Report on
a comparison of the sample characteristics of two RDD survey designs. Paper presented at
the Center for Disease Controls 11th Annual BRFSS Conference, June, Atlanta, GA.
Mirowsky, J., and C. E. Ross. 1996. Fundamental analysis in research on well-being: Distress
and the sense of control. The Gerontologist 36 (5): 584-94.
Moore, M. H., and R. C. Trojanowicz. 1988. Policing and the fear of crime. In Perspectives on
policing, no. 3 (June) 1-7. National Institute of Justice and the Program in Criminal Justice
Policy and Management, Harvard University.
Perkins, D. D., P. Florin, R. C. Rich, A. Wandersman, and D. M. Chavis. 1990. Participation and
the social and physical environment of residential blocks: Crime and community context.
American Journal of Community Psychology 18 (1): 83-115.
Perkins, D. D., and R. B. Taylor 1996. Ecological assessments of community disorder: Their
relationship to fear of crime and theoretical implications. American Journal of Community
Psychology 24 (1): 63-107.
Perkins, D. D., A. H. Wandersman, R. C. Rich, and R. B. Taylor. 1993. The physical environment
of street crime: Defensible space, territoriality and incivilities. Journal of Environmental
Psychology 13 (1): 29-49.
Rohe, W. M., and R. J. Burby. 1988. Fear of crime in public housing. Environment and Behavior
20 (6): 700-20.
Rohe, W. M., and M. A. Stegman. 1994. The impact of home ownership on the social and politi-
cal involvement of low-income people. Urban Affairs Quarterly 30 (1): 152-72.
Ross, C. E. 1993. Fear of victimization and health. Journal of Quantitative Criminology 9 (2):
159-75.
Ross, C. E., and S. J. Jang. 1996. Community disorder, fear and mistrust: The buffering role of
social ties among neighbors. Paper presented at the annual meeting of the American Society
of Criminology, November, Chicago.
Siegel, J. S. 1993. A generation of change: A profile of Americas older population. New York:
Russell Sage.
Simcha-Fagan, O., and J. E. Schwartz. 1986. Neighborhood and delinquency: An assessment of
contextual effects. Criminology 24:667-703.
Skogan, W. G. 1986. Fear of crime and neighborhood change. In Communities and crime, edited
by A. J. Reiss and M. Tonry, 203-30. Chicago: Univ. of Chicago Press.
. 1990. Disorder and decline. Berkeley: Univ. of California Press.
Skogan, W. G., and M. G. Maxfield. 1981. Coping with crime: Individual and neighborhood
reactions. Beverly Hills, CA: Sage.
Taylor, R. B., and J. Covington. 1993. Community structural change and fear of crime. Social
Problems 40 (3): 374-95.
Taylor, R. B., and M. Hale. 1986. Testing alternative models of fear of crime. Journal of Criminal
Law and Criminology 77 (1): 151-89.
Taylor, R. B., and S. A. Shumaker. 1990. Local crime as a natural hazard: Implications for under-
standing the relationship between disorder and fear of crime. American Journal of Commu-
nity Psychology 18 (5): 619-41.
Waksberg, J. 1978. Sampling methods for random digit dialing. Journal of the American Statisti-
cal Association 73 (1): 40-46.

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015

432 URBAN AFFAIRS REVIEW / January 1999

Catherine E. Ross is a professor of sociology at Ohio State University. She studies the
effects of neighborhoods, work, and family on mens and womens health, mental health,
and their sense of control versus powerlessness. She is principal investigator on a grant
from NIMH, Community, Crime and Health, examining the ways in which neighbor-
hood affects subjective well-being. Her recent publications include The Links between
Education and Health, with Chia-ling Wu in the American Sociological Review (1995);
Education, Age and the Cumulative Advantage in Health, with Chia-ling Wu in Jour-
nal of Health and Social Behavior (1996); and A New Look at Urban Alienation: The
Effect of Neighborhood Disorder on Perceived Powerlessness, with Karlyn Geis, in the
Social Psychology Quarterly (1998).

John Mirowsky is a professor of sociology at Ohio State University and editor of the Jour-
nal of Health and Social Behavior. He studies the effects of mastery and equity on emo-
tional well-being and physical health. He and Catherine Ross are conducting a three-
year follow-up survey of American adults funded by the National Institute on Aging. The
project, titled Aging, Status and the Sense of Control, examines links between physio-
logical aging, age-related changes in social and physical activities, and the sense of
mastery and control. He is coauthor, with Catherine E. Ross, of Social Causes of Psycho-
logical Distress (Aldine de Gruyter). His recent publications include Fundamental
Analysis in Research on Well-Being: Distress and the Sense of Control, with Catherine
Ross in The Gerontologist (1996), and Physical Impairment and the Diminishing Ef-
fects of Income, with Paul Hu in Social Forces (1996).

Downloaded from uar.sagepub.com at UNIV OF MICHIGAN on June 16, 2015