ABSTRACT
Do the geographic contexts in which disadvantaged children are raised influence
whether they have difficulties in elementary school? We address this question by
estimating Cox proportional hazard models with instrumental variable measures of
context, using data for 410 low-income, Latino and African American children who lived
in Denver public housing before age six. The Denver Housing Authority’s procedure for
allocating families to dwellings mimics random assignment, thus offering an unusual
natural experiment for measuring context effects isolated from geographic selection bias.
We find that several socioeconomic and demographic contextual indicators are
statistically and substantively important predictors of low-income, Latino and African
American children’s difficulties in elementary school, though sometimes in nonlinear and
interactive ways. Generally, the hazard of being assigned to special education classes,
suspended or forced to repeat a grade is greater in neighborhoods with higher
occupational prestige and percentages of immigrants and lower in those with higher
percentages of African American residents.
Abstract word count: 154
JEL codes: I21, R2, R29
Keywords: neighborhood effects, natural experiments, elementary school
outcomes, instrumental variables
1
"Elementary School Difficulties of Low-Income, Latino and
African American Youth: The Role of Geographic Context"
INTRODUCTION
Improving the educational outcomes for low-income, minority youth has been a
longstanding goal of urban educational policy, given the importance of these outcomes in
shaping future life chances (Newburger, Birch and Wachter, 2011; Tate, 2012). Though
the importance of home environment is undeniable, increasing social scientific attention
has been devoted to investigating the degree to which geographic context (“context”
hereafter) also exerts a substantial, independent influence upon educational outcomes; for
recent reviews, see: DeLuca and Dayton (2009); Johnson (2010); Sastry (2012);
Niewenhuis and Hooimeijer (2014); Sharkey and Faber (2014).
The investigation of context effects on educational outcomes is complicated by
numerous significant empirical challenges; see: Galster (2008); Harding et al. (2011);
Sharkey and Faber (2014). These principally include: (1) geographic selection bias; (2)
limited variability of context experienced by low-income families; and (3) multidimensional heterogeneity of context.1 As amplified below, our study arguably meets all
three challenges, through leveraging a natural experiment involving the Denver (CO)
Housing Authority (DHA), which since 1969 has operated public housing units located in
a wide range of neighborhoods throughout the City and County of Denver. Because the
initial assignment of households on the DHA waiting list to dwellings (and, thus, to
1
Other challenges include measuring the timing and duration of exposure to a particular dimension of
context, the appropriate geographic scale(s) of “context,” the measurement of various aspects of context
(especially causal processes). and potential heterogeneity of contextual effects across different kinds of
children.
2
associated geographic context) mimics a quasi-random process, this program represents
an unusual opportunity for reducing geographic selection bias and observing the unusual
combination of low-income children raised in high-quality spaces. We further control for
potential geographic selection bias that may emerge post-DHA assignment by
instrumenting for context experienced during elementary school with the characteristics
associated with the place where the family was first offered a dwelling by DHA. We
marshal census and administrative data in devising a wide array of indicators measuring
context.
In this study we analyze data from administrative sources and data we have
collected from telephone surveys with Latino or African American current and former
DHA tenants who were raising children when they lived in DHA housing. Our surveys
provide retrospective information about academic performance and temporally
corresponding characteristics of children, their caregivers2 and their households.
Our research question involves identifying the magnitude of context effects:
Are there significant differences in the hazards of low-income Latino and African
American children encountering academic and behavioral difficulties during
elementary school (i.e., special education assignment, repeating a grade, being
suspended) that can be attributed to differences in their geographic environments,
all else equal?3
Though secondary school outcomes are undoubtedly important, we are unapologetic in
2
We will use the term caregivers instead of parents in this paper because a substantial share of children in
our study were not being raised by their biological parents.
3
More specifically, we answer this question in our study by focusing on low-income Latino and African
American children whose families were assigned DHA housing before they were age six, who spent at least
two years living in DHA housing and who were at least age 12 by the time of our Denver Housing Study
survey.
3
our focus because the impact of early childhood context may produce lasting effects
(Slavin, Karweit, and Wasik 1993; Bouchard 1997; Duncan et al. 1998; Shonkoff and
Phillips 2000; Sampson, Sharkey, and Raudenbush, 2008). Framing our investigation in
terms of urban policy, we are asking what happens to low-income Latino and African
American students in elementary school when their families are assigned to public
housing in radically different parts of Denver.
HOW GEOGRAPHIC CONTEXT MIGHT AFFECT SCHOOL PERFORMANCE
Our theoretical framework for studying links between geographic contexts and
children and youths’ outcomes draws most heavily from accepted ecological models of
human development. This perspective sees children’s development being shaped by the
proximal (e.g., family) as well as distal (e.g., neighborhood, school) contexts in which
children live and interact (e.g., Bronfenbrenner and Morris, 1998). Geographic context
may affect children through a variety of causal mechanisms that can occur either through
social, institutional, or biological processes; for extended discussions see Jencks and
Mayer (1990), Duncan, Connell and Klebanov (1997), Gephart (1997), Sampson (2001),
Dietz (2002), Sampson, Morenoff, and Gannon-Rowley (2002), Odis (2007), Johnson
(2010); Harding et al. (2011), Galster (2012) and Sharkey and Faber (2014). The
potential mechanisms relevant for school performance of young children include: peer
influences; socialization and social control; violence and social disorder; institutional
resources; environmental health; parental mediation; and school practices:
Peer Influences: Children may develop and modify attitudes, values, behaviors and
expectations about school as a result of interactions with proximate peers in
4
neighborhoods and schools (Anderson, 1990; Case and Katz, 1991). These peer effects
may be transmitted among children in a contagion-like fashion (Crane, 1991; South,
Baumer and Lutz, 2003).
Socialization and Social Control: Children’s attitudes, values, behaviors and
expectations about school may be shaped by local youth and/or adult role models in the
environs and norms enforced by the residential and/or school community (Wilson, 1987;
Klebanov, Brooks-Gunn and Duncan, 1994; Connell and Halpern-Felsher, 1997; McCoy,
Roy, and Sirkman, 2013).
Violence and Social Disorder: Exposure to violence in the neighborhood or school
may lead to adverse physical responses (like ill health from stress), psychological
responses (like post-traumatic stress syndrome) and inhibitions to speech communication,
all of which impede academic performance (Massey, 2001; Hurt et al., 2001; Henrich et
al., 2004; Ratner et al., 2006; Sampson, Sharkey and Raudenbush, 2008).
Institutional Resources: Public and private institutions controlling services and
facilities (especially schools, after-school tutoring, etc.) vary spatially in their quantity
and quality of resources, thereby differentially affecting youths’ perceptions of the value
of good school performance (Entwisle et al. 1997; Bennett, 2011). Variations in the
physical, curricular, pedagogical and student-body quality of elementary schools is of
obvious relevance here (Jargowsky and El Komi, 2011). Some components of local
institutions may be endogenous to context. For example, neighborhoods may affect the
socioeconomic and behavioral composition of local schools’ students to the extent that
they are strongly linked to attendance zones, thereby shaping the peer influences to which
5
children will be exposed in the classroom (Feinstein and Symons, 1999; Hoxby, 2001;
Lavy, Silva, and Weinhardt, 2009; Sykes and Musterd, 2011).
Environment and Health: Geographically-based variations in exposure to ambient
noise, toxins, lead, or other pollutants can affect mental and behavioral development and
the severity of school absences due to asthma and other diseases, thereby affecting
children’s academic performance (Aneshensel and Sucoff, 1996; Acevedo-Garcia et al.
2003).
Parental Mediation: Variants of the foregoing mechanisms may also affect the
physical and mental health, attitudes, behaviors, and resources of parents. These indirect
contextual effects may be transmitted to children inasmuch as they affect the parents’
willingness and ability to assist, monitor, and enrich their children’s educational
experiences (Connell and Halpern-Felsher, 1997; Furstenberg et al, 1999; Leventhal and
Brooks-Gunn, 2000; Oh, 2005) or deter other behaviors that interfere with their school
performance (Bellair and Roscigno, 2000).
School Practices: Schools may have de jure and/or de facto practices about how
certain types of students will be treated by teachers, staff and administrators. Some
schools may be more prone to, for example, identify low-income students with
developmental disabilities, assign them to special education classes, or reactive punitively
to shortcomings in discipline (Oswald et al., 1999; Drame, 2002; Ferri and Connor, 2005;
Tajalli and Garba, 2014).
Though there is evidence to support several of the context effect mechanisms
above, there is no agreement on which mechanism(s) may dominate for elementary
school outcomes. Indeed, this remains a critical realm of future research (Harding et al.
6
2011; Galster, 2012; Sharkey and Faber, 2014).4 Below we will reflect on what our
findings might imply for these causal processes.
MEASURING THE INDEPENDENT, CAUSAL EFFECT OF CONTEXT:
CHALLENGES AND RESPONSES IN THE EMPIRICAL LITERATURE
From a developmental perspective it is clear theoretically that the academic
performance of children will be jointly influenced by characteristics of: the children and
their caregivers, households, schools and neighborhoods. Attempts to measure the causal
contribution of neighborhood confront many difficult empirical questions, however.
What are the appropriate neighborhood characteristics to measure? Can adequate
variation in neighborhood context among low-income youth be obtained from nonexperimental datasets? Given that the analyst cannot, by definition, operationalize and
measure all caregiver characteristics, what can be done to minimize potential bias from
these omitted variables that may be strongly associated both with caregivers’
neighborhood selection process and the child’s school performance?
No single study has addressed all of these questions satisfactorily; for a fuller
discussion see Galster (2008, 2012); Johnson (2010); Niewenhuis and Hooimeijer (2012).
Virtually all the U.S.-based work has used as indicators of context census tracts measures
of socio-economic status (rates of poverty, female headship, welfare receipt, educational
and occupational attainment) and, less often, ethnic and nativity composition as proxies
for the underlying causal processes described above. Few studies have, however, also
4
Two other theoretical aspects about causal process that are beyond the scope of our review: contingency
(Bennett, 2012; Galster, 2012; Small and Feldman, 2012; Francois, Overstreet and Cunningham, 2012;
Sharkey and Faber, 2014) and nonlinear relationships (Crane, 1991; Galster, 2014; Galster, Andersson and
Musterd, 2014).
7
operationalized the aforementioned exposure to violence, environmental health, or
institutional mechanisms affecting school performance, and thus the specter of omitted
variable bias must be acknowledged. Similarly, most studies either measure context of
school or residential neighborhood (for exceptions, see: Cook et al., 2002; Owens, 2010;
Sykes and Musterd, 2011; Carlson and Cowan, 2015). Moreover, few non-experimental
studies have focused on potential context effects upon low-income minority youth
(arguably the most relevant group from the perspective of urban policy) because of
inadequate sample sizes and/or variation in neighborhood contexts (DeLuca and Dayton,
2009). Many studies have employed surveys that permit the statistical control of a wide
range of parental and youth characteristics, though no survey can avoid with certainty the
potential problem of geographic selection based on caregiver unobservables, of course.
The vast majority of U.S.-based studies in this field5 have reported nontrivial
partial correlations between various measures of the socioeconomic composition of
neighborhood residents and several measures of school performance; cf. Datcher (1982),
Corcoran et al. (1990), Crane (1991), Garner and Raudenbush (1991), Clark (1992),
Duncan (1994), Duncan, Brooks-Gunn, and Klebanov (1994), Connell et al. (1995),
Chase-Lansdale and Gordon 1996; Ensminger, Lamkin and Jacobson (1996), ChaseLansdale et al. (1997), Kohen et al. (2002), Ainsworth (2002), South, Baumer and Lutz
(2003), Ceballo, McLoyd and Toyokawa (2004), Sampson, Sharkey and Raudenbush
(2008), Bennett (2011) and Francois, Overstreet and Cunningham (2012). Recent metaanalysis of the international literature by Niewenhuis and Hooimeijer (2014) and a review
5
For an exception, see Plotnick and Hoffman (1995). The European-based evidence is considerably more
mixed in conclusions; cf. Kauppinen (2007), Nicoletti and Rabe (2010) and Niewenhuis and Hooimeijer
(2014).
8
of the U.S. literature by Sharkey and Faber (2014) conclude that there can be nontrivial
neighborhood effects on school outcomes.6
The causal interpretation of context-educational performance relationships
measured by these studies is subject to challenge, however, due to potential geographic
selection bias.7 There have been several types of methodological responses to this
challenge in the education-context effects literature:
Propensity Score Matching: Outcomes for children receiving some neighborhood
“treatment” are compared to those for a synthetic sample of non-treated matched via a
probability model of their similar likelihoods of being treated (e.g., Harding, 2003)
Sibling Models Based on Longitudinal Data: The biases from unobserved, timeinvariant caregiver characteristics are eliminated by measuring differences in educational
outcomes and neighborhood experiences between two siblings (e.g., Aaronson, 1998).
Fixed-Effect Models Based on Longitudinal Data: Unobserved, time-invariant
characteristics of caregivers that may affect both neighborhood selection and their
children’s educational outcomes are measured by dummy variables (e.g., Vartanian and
Gleason, 1999; Jargowsky and El Komi, 2011).8
6
They conclude that the strength of the relationship is contingent on outcome, youth and context being
investigated, however.
7
The basic issue is that caregivers may have certain (unmeasured) behaviors, values, and skills related to
their children’s school performance and move to / remain in certain types of neighborhoods as a
consequence of these same attributes. Any observed statistical relationship between their children’s
neighborhood conditions and school outcomes may therefore be biased because of this systematic
spatial selection process, even if all the parents’ and children’s observable characteristics are controlled
(Manski 1995, 2000; Duncan et al. 1997; Duncan and Raudenbush 1999, Dietz 2002). The direction of the
bias has been the subject of debate, with Jencks and Mayer (1990) and Tienda (1991) arguing that measures
of neighborhood impacts are biased upwards, and Brooks-Gunn, Duncan, and Aber (1997) arguing the
opposite. Gennetian, Ludwig, and Sanbonmatsu (2011) show that these biases can be substantial enough to
seriously distort conclusions about the magnitude and direction of neighborhood effects.
8
Here we do not consider the techniques related to fixed-effects that have been employed in the
voluminous literature related to peer effects, such as value-added modeling, since they are not directly
related to controlling for geographic selection. For the seminal example, see Hanushek (1971); Reardon
and Raudenbush (2009) provide a critique; Rothstein (2010) provides a recent application.
9
Instrumental Variables for Neighborhood Characteristics: Proxy variables for
neighborhood characteristics are devised that only vary according to attributes exogenous
to the individual children and their caregivers (e.g., Duncan, Connell, and Klebanov,
1997; Crowder and South, 2003; Galster et al. 2007).
Timing of Events: Weinhardt (2014) has argued that youth moving into deprived
social housing-dominated neighborhoods after taking a school achievement test are likely
to share common unobservable characteristics with youth moving into the same places
before taking the test, so the short-term effect of the neighborhood can be measured by
comparing the two groups’ outcomes. Sharkey and Sampson (2010) and Sharkey et al.
(2012, 2014) address the selection bias problem by exploiting the variation in the timing
of local homicides compared to interview assessments for a sample of children in
families that have previously selected the same neighborhood. Carlson and Cowan
(2015) take advantage of the institution of an open-enrollment system.
Natural Quasi-Experiments: Observations are produced by idiosyncratic
interventions, such as the Gautreaux and Yonkers public housing desegregation programs
(e.g., Rosenbaum, 1995; Fauth, Leventhal, and Brooks-Gunn, 2007; DeLuca et al. 2010),
public housing revitalization programs (Jacob, 2004; Clampet-Lundquist, 2007) or
inclusionary zoning (Schwartz, 2010; Casciano and Massey, 2012) that create exogenous
variation in contexts for assisted tenants.9
Random Assignment Experiments: Data are produced by an experimental design
whereby households are randomly assigned to different neighborhoods; the only relevant
example here is the Moving To Opportunity (MTO) demonstration (Ludwig, Ladd and
9
Often there is still room for some geographic selection by participants and/or by program staff in these
programs, however.
10
Duncan, 2001; Leventhal and Brooks-Gunn, 2004; Leventhal, Fauth and Brooks-Gunn,
2005; Kling, Liebman and Katz, 2007; Ludwig et al. 2008; Sanbonmatsu et al. 2011).
Unfortunately, there is no consensus about the unbiased magnitude of context
effects on educational outcomes from this set of studies that attempt to overcome
selection bias (Burdick-Will et al., 2010; Niewenhuis and Hooimeijer, 2014; Sharkey and
Faber, 2014). The studies noted above that use the first five, non-experimental methods
for overcoming selection bias find strong residential neighborhood effects on educational
outcomes, with one exception (Weinhardt, 2014). Similarly, the natural quasiexperiments provide only one example of no effect: Jacob (2004). The only randomassignment experiment in the field, however, identified generally small effects on
educational outcomes (Ludwig, Ladd and Duncan, 2001; Kling, Liebman and Katz, 2007;
Sanbonmatsu et al. 2006; 2011), though results from Chicago and Baltimore may be
more substantial (Sharkey and Faber, 2014). However, MTO-based studies’ conclusions
regarding context effects have been challenged for many reasons (cf. Sanbonmatsu et al.
2006; Clampet-Lundquist and Massey, 2008; Sampson, 2008; Briggs, Ferryman, Popkin,
and Rendon; 2008, DeLuca and Dayton, 2009; Burdick-Will et al., 2010; Briggs, Popkin
and Goering, 2010; Galster, 2011).10
Thus, the existing literature has not provided definitive evidence about the
potential educational benefits to low-income, Latino and African American children from
10
The debate focuses on four domains. First, although MTO randomly assigns participants to treatment
groups, it neither randomly assigns characteristics of neighborhoods initially occupied by voucher-holders
(except maximum poverty rates for the experimental group) nor characteristics of neighborhoods in which
participants in all three groups may move subsequently. Thus, there remains considerable question about
the degree to which geographic selection on unobservables persists. Second, MTO may not create adequate
duration of exposure to contextual conditions by any group at any location to observe much treatment
effect. Third, MTO overlooks the potentially long-lasting and indelible developmental effects upon adult
experimental group participants who spent their childhoods in disadvantaged places. Fourth, it appears that
even experimental MTO movers rarely moved out of predominantly minority-occupied neighborhoods near
those of concentrated disadvantage and achieved only modest changes in school quality.
11
residence in geographic contexts that vary on multiple dimensions. Our study aims to
shed light on this crucial urban issue by leveraging a natural experiment related to the
Housing Authority of the City and County of Denver (DHA). As explained below, our
unique contribution is that we simultaneously address all three empirical challenges
above. We use the DHA allocation process to introduce exogenous variation into
household location; low-income, minority children are quasi-randomly assigned to a wide
variety of neighborhoods throughout Denver County. We further control for potential
geographic selection after initial assignment by using instrumental variables for
contextual characteristics experienced during elementary school.
THE NATURAL EXPERIMENT INVOLVING PUBLIC HOUSING IN DENVER
In addition to its large-scale, conventional public housing developments, the DHA
has operated since 1969 a program providing approximately 1,500 low-income families
with opportunities to live in scattered-site, single-family and small-scale, multi-family
units. These units are located in a wide range of neighborhoods throughout the congruent
City and County of Denver, whereas the modestly scaled, well-maintained conventional
developments are typically located in less-advantaged neighborhoods. The generally
high-quality public housing dwellings located in a majority of neighborhoods, wellmanaged by a consistently high-performing housing authority, distinguishes the situation
in Denver from those in the MTO sites and in other cities with more problematic public
housing contexts. The wide variety of neighborhood conditions that DHA offered tenants
in our study are shown in the middle panel of Table 1 and maps in online Appendix C. In
comparing neighborhoods where DHA dwellings are located to Denver County as a
whole (cf. middle and right-hand panels in Table 1) it is clear that while, on average,
12
DHA units are in less-advantaged, greater-minority, higher-crime areas, there is
substantial variation in all indicators.
[Table 1 about here]
From 1987 onwards, as applicants (who met certain basic eligibility criteria) came
to the top of the public housing waiting list they were offered a vacant DHA unit (in
either conventional or scattered-site programs) with the number of bedrooms appropriate
for their family size and gender of children. If they did not accept this unit they were
offered the next similarly sized unit that became available (typically after a nontrivial
wait). Applicants who did not accept this second unit dropped to the bottom of the
queue, creating a wait of a year or more.
Evaluating the Randomness of Initial Assignment
As detailed in Appendix A [intended as online], we conducted a variety of
statistical tests to ascertain whether the initial assignment of households to a DHA
dwelling unit (and neighborhood thereby) mimicked random assignment of household to
neighborhood. These tests not only involve conventional “balancing tests” of
correlations between observable tenant and neighborhood characteristics but also an
original test for ascertaining whether typically unobserved tenant characteristics were
uncorrelated with neighborhood characteristics. We concluded that, conditioned on
ethnicity (which we control in our models), the DHA allocation process produced a quasirandom initial assignment of households across neighborhood characteristics.
The quasi-randomness of this initial DHA assignment potentially erodes over
time, however, as some residents selectively leave their initial locations while others
selectively stay. Thus, three potential sources of geographic selection based on
13
parent/caregiver unobservables might arise after initial assignment. First, DHA
households can voluntarily transfer between scattered-site and conventional public
housing developments. This occurred rarely, however, as documented in Appendix A.
Second, a substantial part of our information comes from households no longer residing
in DHA housing, and their subsequent locations were likely not randomly chosen.11 A
third potential source of selection relates to those who do not move out of their DHA
housing for an extended period. Perhaps their unwillingness or inability to move out of
DHA is related to some unobservable caregiver characteristics that also may be
connected to the educational outcomes being investigated.
Identification Strategy
The aforementioned potentials for geographic selection in the natural experiment
led us to adopt an identification strategy that employs the year and contextual
characteristics associated with the dwelling first offered by the DHA to a household when
it rose to the top of the waiting list. We would argue that the offered spatial
characteristics and the year of such offer provide valid instruments for the actual context
experienced by children during elementary school. There is no plausible reason why the
offer of a place or its timing should: (1) influence a child’s educational outcomes other
than through its influence on actual geographic context experienced; or (2) be related to
unobservable characteristics of the caregiver related to neighborhood preferences and the
child’s educational performance. We would further argue that the offered spatial
characteristics likely provide strong instruments for the actual characteristics experienced
by children during elementary school because (as noted above) three-quarters of
households ended up residing in the neighborhood first offered and a majority of them
11
Slightly more than one-third of all caregivers interviewed in the study were former DHA residents.
14
were still residing there when we interviewed them to gather information about their
children’s educational outcomes. The timing of the offer may also be a strong instrument
for characteristics that evinced clear overall trends during the study period, such as rising
densities of immigrants and Latinos, and declining percentages of dwellings built before
1940 and crime rates. As we demonstrate later, these claims are verified empirically.
A complimentary way to describe our approach is in terms of an “intent to treat”
analysis. DHA could be viewed as offering to the waiting list household a “treatment”
consisting of a bundle of geographically attached attributes: dwelling, residential
neighborhood, school, environment, public service package, etc. Most accept this
treatment but some hold out for another package they deem preferable to that originally
offered. Others may subsequently move from the initially occupied DHA context.
Regardless, our identification strategy can be thought of as a way to evaluate causally the
impact of DHA’s public housing program (having both conventional and scattered-site
components) on minority children’s elementary school performance.
Generality
The use of natural experiments inevitably raises questions about the generality of
results. We believe that our findings can fairly be generalized to low-income, Latino and
African American families who apply for and remain on the waiting list long enough to
obtain public housing. As such, it may not be fully generalizable to the population of
minority families who obtain subsidized rental housing (like vouchers), and certainly may
not be to the larger population of minority families who qualify for housing assistance.
Nevertheless, it is similar to--yet considerably more general than--the populations
forming the samples for the oft-cited MTO-based scholarly studies noted above.
15
DATA COLLECTION IN DENVER
Denver Child Study Survey of Current and Former DHA Households
Most information we analyze here come from the Denver Child Study survey that
we developed and fielded during 2006-2008. Study eligibility criteria were: (1) presence
of children in the home between ages 0 and 18 years when they moved into DHA; (2)
family remained in DHA housing for at least two years; (3) family first entered DHA in
1987 or later (when DHA’s current quasi-random assignment process came into
operation); and (4) were of Latino or African American ethnicity. We estimate an overall
participation rate of 56.5 percent, with most non-participation due to our inability to
locate the household; less than six (6) percent refused to participate once contacted.
Further details of this survey, are presented in Appendix B [intended as online]. Our
team successfully completed 710 interviews with the primary caregivers of eligible
households whose surveys subsequently passed our rigorous data verification and
reliability checks. The 410 children analyzed here come from 246 of these households
and meet the following criteria: (1) were assigned to DHA housing before they were age
six; (2) spent at least two years living in DHA housing; (3) were at least age 12 by the
time of our survey; and (4) had complete information on all relevant variables. We
emphasize that in all cases analyzed here, assignment to DHA housing preceded the
elementary school period during which outcomes were measured.
Caregiver and Household Characteristics
Our Denver Child Study survey collected information on a wide variety of
caregiver and household characteristics that we employed as controls; those ultimately
employed are listed in Table 2. Covariates included: whether the caregiver was born
outside of the U.S., proportion of the elementary school period that both parents were
16
cohabiting, whether the caregiver had attained a high school diploma or higher, the
cumulative number of household moves the child had experienced and the natural
logarithm of average household income while the child was in elementary school. Our
survey also asked questions that permitted us to measure a series of stressful household
events from which we created an annual average “household economic stressors index”
(scaled 0-5) experienced while the child was in elementary school. Caregivers were
asked whether and when they experienced any of the following events for each residence:
a. Unemployed a month or more?; b. Have a major illness or injury?; c. Have too little
money to buy enough food for your family?; d. Have your electricity, gas, or phone
service cut off?; or e. Get evicted from your home? This index was incremented by one
for each of the above circumstances experienced by the household. Previous research has
indicated that acute economic shock to a household can have seriously disruptive effects
on children’s’ mental and physical health that, in turn, can impair a variety of outcomes
over the longer term (Shonkoff et al., 2012). The survey also asked questions that
permitted us to estimate the presence of depressive symptomatology exhibited in primary
caregivers at the time of survey (CES-D).12
Because we employ Cox Proportional Hazard models, the temporal measurement
of the aforementioned time-varying characteristics depended on the year when each child
first experienced the particular time-varying educational outcome being investigated (if
ever). 13 Cohabiting, income and stressor variables were measured as averages from age
6 until the first occurrence of the given outcome, or until age 11 if the outcome did not
12
We use a dummy variable indicating whether the parent exhibited sub-clinical or clinical depressive
symptomatology (score at least 16 on the CES-D scale); details are available in Appendix B.
13
The exception was the CES-D scale, which was only measured once at time of survey and is intended as
a control for respondent affect at that time.
17
occur by then. The cumulative number of lifetime moves and education of caregiver
were measured at first occurrence of the given outcome, or at age 11 if the outcome did
not occur by then.
We believe that this battery of characteristics adequately controls for household
role modeling, child supervision, parenting behavior, attitudes, norms, and economic
resources that shape students in ways that would affect their educational performance
(Hotz and Pantano, 2014). Indeed, these covariates are conventionally employed in
educational outcome models (e.g., Crowder and South, 2003). In preliminary models we
experimented with a much wider range of covariates. Given our relatively modest
sample sizes, however, we omitted from our reported models covariates that never proved
statistically significant in preliminary trials.14
[Table 2 about here]
Denver Child Study caregiver and household characteristics as portrayed in Table
2 clearly reflect their disadvantaged circumstances. While their child was in elementary
school their average annual income was below $11,700 and both parents were cohabiting
only 33 percent of the years. Over half the caregivers had no high school diploma; 17
percent were immigrants. Caregivers often faced economic challenges: 1.26 incidents of
acute financial crisis while their children were in elementary school, on average annually.
Compared to most low-income households they were remarkably stable, however, due to
the low turnover among DHA tenants: children had moved only 1.28 times in their lives
by the first instance of the given elementary school outcome (or age 11).
14
In preliminary models our additional caregiver controls included measures of: disability, fertility and
employment history, alcohol and drug use, and whether the household had health insurance. We also
assessed caregiver gender, but since virtually all were female this was not included as a covariate. In
preliminary models we also controlled for birth order of the child and family size, but such did not prove
predictive.
18
Child Characteristics
The Denver Child Study survey asked caregivers to supply information about all
of their children with whom they had lived in DHA public housing for at least two years.
The only characteristics that proved important were our gender/ethnicity dummy control
variables; see Table 2. Our analysis sample consists of 28 percent Latinas, 29 percent
Latinos, 21 percent African American girls and 22 percent African American boys.
We ascertain difficulties in elementary school on the basis of the Denver Child
Study survey caregiver responses. It was beyond the scope of this study to gather data
from school administrative records. Placement in special education classes was based on
the question, “Has __ [child] __ ever been classified by school personnel as needing
special education? [If yes, at what age?].” Twelve percent of our sample had experienced
such special education placements during elementary school, with the mean age of first
placement at 7.8 years old. Grade repetition was determined on the basis of the question
“Has __ [child] __ ever repeated a grade? [If yes, which grade(s) was/were repeated?]?”
Nine (9) percent of our sample repeated at least one grade during elementary school, with
an average age of 8.4 years at time of first repetition. Disciplinary problems were
measured by caregivers’ responses to the question, “Was __ [child] __ ever suspended
from elementary school [If yes, during which grades?].”15 In our sample seven (7)
percent of students had been suspended, with a mean age at the first such disciplinary
action of 9.4 years.
We note that caregiver reports of student outcomes are common in the vast
neighborhood effects literature that employs PSID or NLSY data; see the meta-analysis
15
There were too few expulsions during primary school to model this outcome.
19
by Nieuwenhuis and Hooimeijer (2014). Nevertheless, we recognize the potential
shortcomings of our school performance indicators. They are subject to recall error by
the caregiver survey respondent, though we intentionally chose outcomes for which this
likely would be minimal.16 They cannot distinguish unambiguously among outcomes
that are produced by the student’s academic abilities and behaviors and those that are
produced by school programs, facilities, policies and actions by teachers, counselors and
administrators. As illustration, a student may be sufficiently under-achieving to warrant
repeating a grade but is not required to do so because of an institutional culture of social
promotions in the school.
Geographic Context Indicators
We used census tract level data from the 1980, 1990 and 2000 censuses. We
employed the Neighborhood Change Data Base (a Geolytics proprietary product) for this
information because it adjusts data to account for potential changes in tract boundaries
between decennial censuses. For estimates of non-census year data, we used linear
interpolation or extrapolation. We gathered indicators that have been widely employed in
prior research on neighborhood effects,17 including percentages of: households moving in
during the prior year, female-headed households, families below the poverty line,
unemployed adults, non-Hispanic African American population, Hispanic population,
foreign-born population, homes that are renter-occupied, homes built before 1940, and
mean occupational prestige based on the General Social Survey prestige score weighted
by the observed proportional distribution of occupations of employees in the tract. Given
16
Random recall errors would drive down the statistical significance of measured neighborhood effects,
however.
17
Cf. Carlson and Cowan (2015). Our neighborhood indicators are measured as averages for the
elementary school period.
20
high correlations among several of these variables, we conducted a principal component
analysis that consistently across census years produced a single component comprised of
the roughly equally weighted sum of census tract percentages of: poor, unemployed,
renters, and female householders. We call this our neighborhood social vulnerability
index.
We also acquired data on geographic context from the Denver-based Piton
Foundation’s Neighborhood Facts Database, which provided small area-based, annually
measured information culled from Denver administrative databases that are not provided
by the Census. In this study, we used violent crimes and property crimes reported to
police per 1,000 population that were averaged over the child’s elementary school years.
These crime data are aggregated by the Piton Foundation to 77 named neighborhood
areas consisting of two census tracts, on average, and thus are measured at a larger spatial
scale than our census-based data.
To get a sense of the sorts of places where children in this study resided when
attending elementary school, we present the mean characteristics of their geographic
context in the left-hand panels of Table 1. We emphasize again, however, that it is the
substantial variation around the mean that gives our study unusual power, and provides
implicit testimony to the effectiveness of the DHAs scattered-site program; cf. the middle
panels of Table 1. Table 1 indicates that our average child experienced during
elementary school many census tract indicators suggesting moderately deprived contexts
compared to Denver: mean social vulnerability index of about 142 (while the minimum
value was 44), and modest occupational prestige (mean 36 while the minimum value was
32). The average tract was ethnically diverse: 18 percent African American, 53 percent
21
Latino and 29 percent other (Anglo and Asian), but neighborhoods varied from virtually
none to almost all of one group or another. Further, 19 percent of the average
neighborhood’s residents were foreign-born (3 percent minimum). Thirty percent of the
housing stock was built before 1940, on average (minimum zero). In our children’s Piton
neighborhoods, mean violent crime rates were 14 per thousand population (one
minimum) and property crime rates were 77 per thousand population (15 minimum).
The multiple geographic attributes we employ raises the specter of
multicollinearity, however. As shown in Appendix Table 1, most of our indicators are
indeed correlated to a statistically significant degree. Preliminary diagnostic regressions
revealed, however, that the conventional Variance Inflation Factor limit (5.0) was
exceeded only by the percentage of Latino residents.18 With the exclusion of this
indicator from our analyses we can be confident that our findings were not unduly
influenced by multicollinearity.
Despite the fact that we employ more contextual indicators than virtually all prior
studies, we acknowledge that we have only indirect and incomplete measures of the
various potential causal mechanisms that might connect geographic context and
elementary school outcomes, as discussed above and in Galster (2008). Our two crime
measures serve as rather direct measures of violence and disorder. The concentration of
old housing serves as a proxy for interior environmental deficits. Our neighborhood
social vulnerability index, occupational prestige score, ethnic and nativity composition
percentages serve as proxies for aspects of the peer effect and socialization / social
control mechanisms, but they likely are also correlated with other mechanisms such as
the quality of the natural environment and local institutions and public services such as
18
The VIF for percentage Latino was about 8 across our outcomes.
22
schools.
Unfortunately, our study has no measures of the school environment, which other
work has identified as an important conduit of context effects (Cook et al., 2002; Owens,
2010; Sykes and Musterd, 2011; Carlson and Cowan, 2015).19 It is therefore essential to
consider how our lack of information about local school quality shapes the validity and
interpretation of our results as measures of residential neighborhood or more general
context effects. We believe that our estimates of the magnitude of geographic context
effects remain valid and unbiased, though we cannot distinguish impacts arising from the
residential neighborhood and the associated local school.
Consider two possibilities of how school and residential neighborhood
characteristics might be correlated. First, school characteristics might be strongly linked
to geography (i.e., via rigidly applied school attendance zones), just as numerous other
spatial attributes (Galster, 2001; Owens, 2010). Indeed, this is part of the institutional
resources mechanism we discussed above: neighborhood embodies a “bundle” of
contextual attributes, one of which is the local school. If school quality indeed were
correlated with socio-demographic, physical and/or safety characteristics of the
surrounding neighborhood, then an educational outcome model like ours could not
distinguish whether, for example, the apparent effect from neighborhood occupational
prestige on the hazard of a particular educational outcome transpired due to social
processes among neighbors or its associated variations in school practices and resources.
19
Our research project had insufficient resources to access school administrative records for surveyed
students or to generate aggregate measures of school quality or composition of student body. Such would
have required doing so retrospectively to 1988 when our oldest students began elementary school. Few
studies of neighborhood effects on educational outcomes have tried to control for school characteristics; see
Cook et al. (2002), Owens (2010), Sykes and Musterd (2011), Sanbonmatsu et al., (2011), Jargowsky and
El Komi (2011), and Carlson and Cowan (2015).
23
Nonetheless, we see this as a limitation of interpreting unambiguously how the context
effect occurs, not whether it occurs. Indeed, this case is likely in our study since 91
percent of surveyed students attended the neighborhood elementary school.20
Second, the geographic link between school characteristics and neighborhood
characteristics might be weak if caregivers can send their children to schools other than
those servicing the neighborhood, whether they might be public magnet or charter
schools, open-enrollment systems in nearby public school systems, or parochial school
options. In such a situation, our educational outcome model indeed would have omitted
school variables of likely importance. Yet, such would not bias the estimate of the effect
of the neighborhood indicators included in the model because, by construction in this
hypothetical, the school and neighborhood variables are not correlated.21
A final consideration is that school selection may be a joint process with
residential neighborhood selection. Obviously, if households move out of DHA,
caregivers can potentially select a neighborhood because it has a good local school
associated with it (1st case above). Less obviously, they may still select it if they can opt
into another school if the neighborhood does not have a good local school (2nd case
above). This could cause a potential econometric problem of neighborhood and school
selection based on caregiver unobservable characteristics. We believe that our
identification approach overcomes the potential confounding effects of this joint selection
process. If the “treatment” that DHA provides is the initial offer of a quasi-randomly
As Carlson and Cowan (2015: 40) note, “In a context where school assignment rules produce an
inextricable link between students’ school and neighborhood environments, the analytic
approach…[of]…examining one context without explicitly accounting for the other—is unproblematic. In
an important sense, these neighborhoods and schools are truly extensions of one another and can validly be
treated as such analytically.”
21
In our case this proves to be the case, as explained below. When we control for whether children
attended the local elementary school or not, there were virtually no differences in our estimates of
neighborhood effects.
20
24
assigned dwelling/neighborhood/school, then subsequent selection of
neighborhood/school packages can be viewed as a potential effect of the treatment. Thus,
our model can be seen as producing unbiased, reduced-form estimates of context effects
working (invisibly by us) through potential residential neighborhood and/or school
characteristics, some of which may be selected endogenously in response to the DHA
geographic “treatment” offered.
ANALYTICAL APPROACH
Cox Proportional Hazard Specification
Our analytical approach models the timing of a particular elementary school
outcome at time t for an individual ij with covariate vector χ using a Cox proportional
hazards model:22
λ(t|χij) = λ0(t) exp(β1χ1ij + … + βnχnij ) = λ0(t) exp(χij β)
where λ(t|χij ) is the observed time of outcome (or the censoring time of age 11) for youth
ij and λ0(t) is the baseline hazard.
We intentionally omit from our models any variables describing children’s mental
and physical health, exposure to violence or risky behaviors, inasmuch as many of these
may themselves be affected by context. In this fashion we avoid “over-controlling” and
thus minimizing the apparent influence of geographic context. We can therefore interpret
our models as akin to yielding “reduced form” estimates of the degree to which
neighborhood variables correlate with elementary school difficulties through unspecified
intervening causal pathways.
22
For these outcomes global chi-square tests failed to reject the null hypothesis that the hazards were
proportional across groups, thus we estimate a Cox hazard instead of accelerated failure time model here.
25
We test for nonlinear effects using a spline specification. Each continuous
context indicator X is supplemented by a corresponding spline variable that takes on the
value of X less its mean for values greater than the mean and zero otherwise. A
statistically significant spline coefficient indicates that the slope differs across above- and
below-mean ranges.
Instrumental Variables
The results of our first-stage OLS regressions of elementary school context variables on
the above instruments (and all the covariates in our second-stage model) are presented in Table 3.
Results were encouraging: the R-squares ranged from .33 to .52 and all chi-square statistics were
highly significant. Characteristics of the context first offered by DHA proved to be strong
instruments for their corresponding characteristics during our children’s elementary school
period, and often for other characteristics as well. Calendar year at time of first offer also was a
strong instrument for all but the percentage of African American residents and the property crime
rate. We thus can place confidence in the IV estimates reported below.
[Table 3 about here]
RESULTS
Core Cox Proportional Hazard IV Models
We present in Table 4 estimated parameters for Cox proportional hazard models
where non-dichotomous predictors are normalized to aid cross-variable comparability of
coefficients, and all neighborhood indicators are IV estimates. Overall the models
perform satisfactorily using chi-square and log pseudo-likelihood criteria.
Consider first results for covariates. The most consistent finding was that
children who moved more evinced inferior elementary school outcomes in all three
26
measures, a result confirmed by other studies (see review in Ziol-Guest and McKenna,
2014). Children who moved one standard deviation more often during early childhood
would be predicted to have a 45 percent greater hazard of being placed in special
education, a 49 percent greater hazard of repeating a grade, and a 127 percent greater
hazard of being suspended. Higher average household income was associated with fewer
academic difficulties in elementary school, also as expected. Children whose families
had a standard deviation-higher income during childhood would be predicted to have a 33
percent smaller hazard of being placed in special education and a 51 percent smaller
hazard of repeating a grade. A standard deviation-higher number of average household
stressors during childhood was associated with a doubling of the hazard of being
suspended. Finally, Latinas were rarely suspended; their hazard was 98 percent lower
than that of African American boys. Surprisingly, caregiver nativity, education, and
cohabitation status were not strongly related to our three elementary school outcomes.23
More importantly for our research question, Wald chi-square tests of the null
hypothesis that all seven neighborhood indicators were zero in each individual model
could be rejected for the suspension outcome at p<.01, for special education at p<.02, and
for repeating a grade at p<.08; the first two remained statistically significant even when a
Bonferroni correction was applied.24 There were four coefficients of neighborhood
variables that were significant at p<.05 or better across the models. A Bonferroni test of
23
The review 13 social experiments by Duncan, Gennetian and Morris (2007-08) also revealed, however,
similar patterns of insignificance as predictors of educational outcomes. We also note that caregiver
reporting of educational outcomes does not appear to be affected by their depressive symptomatology.
24
Bonferroni corrections to critical values for statistical inference are suggested when so many parameters
are estimated across multiple models that some statistically significant results might emerge by chance. In
this Bonferroni test the null is rejected if any of the three chi-square statistics associated with the three
estimated equations is significant with a p value less than the selected alpha level divided by the number of
equations (tests). In our case, the critical value was .05/3 = .017, a threshold that was met by the chi-square
values associated with all but the repeat grade outcome.
27
the null hypothesis that all 21 neighborhood coefficients across the three outcomes jointly
equaled zero was rejected at p<.05.25 We conclude that the observed pattern of several
statistically significant coefficients of context variables was unlikely to have occurred by
chance. Thus, we have confidence in presenting the individually statistically significant
results; more holistic interpretation and discussion follows our robustness checks.
[Table 4 about here]
The nativity composition of the neighborhood proved a strong predictor of two
outcomes. A child spending elementary school in a neighborhood having a standard
deviation-higher average percentage of foreign-born residents would be predicted to have
a 84 percent greater hazard of repeating a grade and a 175 percent greater hazard of being
suspended. Ethnic composition was related to placement in special education classes. A
child spending elementary school in a neighborhood having a standard deviation-higher
average percentage of African American residents would be predicted to have a 70
percent smaller hazard of such placement. Finally, the occupational prestige score was
positively related to placement in special education classes. A child spending elementary
school in a neighborhood having a standard deviation-higher average prestige would be
predicted to have a 51 percent greater hazard.
We were surprised by the lack of statistical significance of other neighborhood
indicators. In the case of our social vulnerability index, we think this is likely due to the
idiosyncrasies of Denver where concentrated disadvantage (especially among African
Americans) is rare. Neither measure of crime proved predictive; we think this is likely
25
In this Bonferroni test the null is rejected if any t-statistic associated with the 21 coefficients estimated
across equations is significant with a p value less than the selected alpha level divided by the number of
coefficients (tests). In our case, the critical value was .05/21 = .0024, a threshold that was met by the p
value associated by the coefficient of percentage of foreign born residents in the suspension equation. For
an analogous application in a neighborhood effects test, see Kling, Liebman and Katz (2007).
28
due to the larger geographic scale (on average, two census tracts) at which these
indicators are measured, not because exposure to violence is not an important aspect of
developmental context. Finally, we must note that even the strongest instrumental
variable estimates add additional imprecision, tending to drive our contextual coefficients
toward insignificance.
Robustness Tests
We explored several extensions of the core model to explore how robust the
results in Table 4 were to alternative specifications. First, to probe differences in
exposure to the DHA-assigned context, we experimented with two sorts of interaction
terms allowing the impact of a context indicator to vary: (1) with child age at initial DHA
assignment; and (2) with years a child spent in DHA during elementary school. These
experiments revealed three interesting patterns, all of which indicated interaction effects
that countered (but did not completely offset) the main effects noted above.26 Both the
negative relationship between percentage of African American residents and the hazard
of special education placement as well as the positive relationship between percentage of
foreign-born residents and the hazard of being suspended were attenuated in absolute
magnitude. The younger the child when first resided in DHA housing, the smaller the
hazard though the main effects persisted whenever the child entered DHA. The positive
relationship between average neighborhood occupational prestige and the hazard of
special education placement fell to virtually nil if the child spent all of elementary school
residing in DHA housing. Implications of these interactions are discussed below.
26
Detailed results are available from the authors.
29
Second, we conducted investigations employing spline regressions and uncovered
some noteworthy non-linear relationships.27 Regarding the hazard of special education
placement, the positive relationship with occupational prestige score demonstrates
considerable diminishing returns at above-mean levels of prestige. The negative
relationship with percentage African American only appertains at above-average
percentages, suggesting a threshold point (Galster, 2014). Moreover, the percentage of
foreign born evinced a positive relationship, but only for below-average ranges of this
percentage, suggesting that this strong nonlinearity cloaked this relationship in the core
model. Regarding the hazard of suspension, the positive relationship with percentage of
foreign born residents persisted over all ranges, but an unusual nonlinear relationship
emerged that was not seen in the core model. Occupational prestige was associated with
lower hazards of suspension at below-average levels of prestige, but just the opposite at
above-average levels. Explanations and implications of these nonlinear findings are
discussed below.
Finally, we considered how robust relationships would be when attending a
neighborhood elementary school is controlled. As discussed earlier, families may make
neighborhood and school selections jointly or independently, depending on the
availability of non-local school options. We wished to test whether controlling for this
selection altered our estimates of context effects; were they to do so it would imply that
our identification strategy failed to remove bias. Reassuringly, we found that a dummy
denoting that the child attended the neighborhood school (91 percent did) was never
statistically significant and coefficients of neighborhood indicators were never altered.
27
Detailed results are available from the authors.
30
DISCUSSION
We have identified several aspects of geographic context that are strongly
associated with the hazards of low-income Latino and African American children having
difficulties in elementary school. Because in our analysis residential neighborhood and
school dimensions of context are so closely linked, however, our discussion will offer
alternative (not mutually exclusive) interpretations of potential causal processes
associated with these two dimensions that may underlie the statistical relationships
observed. In each case the processes associated with residential neighborhood context
will relate to the likelihood that the child objectively displays the problematic
performance (e.g., low scores on standardized tests) or behavior (e.g., violent acts). The
processes associated with schools will relate both to the likelihood that an administrative
decision will be rendered (e.g., assignment to special education, suspension, or grade
repetition) given a particular student’s performance or behavior, and to the possibility
that a student’s performance will be altered by differences in school climate, resources
and student body.
Nativity Composition
The hazard for low-income, minority children being assigned to special education
classes, suspended, or required to repeat a grade generally is directly related to the share
of immigrants in their geographic context, controlling for the nativity of their
caregivers.28 The fact that the majority of such immigrants in Denver originate from
Mexico likely is significant for interpreting these results. From a residential context
perspective, immigrant-dense neighborhoods may offer our Denver Child Study children
access to adults and extended family and friendship networks that better monitor
28
Complementary results were found by Tajalli and Garba (2014).
31
neighborhood children’s behavior as well as enhance their sense of neighborhood safety
and security – both of which may foster environments that are more conducive to student
learning and academic success. However, if these same networks of adult neighbors,
family members and student peers themselves have limited English proficiency, low
levels of literacy in either or both English and Spanish, and/or low levels of educational
attainment, children residing in such neighborhood contexts are particularly vulnerable to
not being cognitively ready for elementary school. Children who may only be exposed to
adults and student peers with limited formal education also may be lacking the role
modeling of social skills and behaviors necessary for learning. Lack of school readiness
either cognitively or socially greatly increases the likelihood of being placed into special
education classes and being retained in grade (Farran, 2000). Concern about school
readiness has been the catalyst for the development of early childhood interventions such
as Early Head Start, Head Start, and neighborhood learning communities for caregivers,
grandparents, neighbors and friends of at-risk, disadvantaged young children.29 This
mechanism is supported by our finding that the effect of foreign-born neighbors is
strongest for children who first occupied DHA housing earlier in their childhoods,
presumably during these critical formative years of cognitive, social, behavioral, and
language development.
From a school context perspective, immigrant-dense neighborhoods will translate
into immigrant-rich classrooms. Immigrant elementary school students may provide
fewer positive social externalities for their classmates (e.g., American cultural capital)
and may indeed divert teachers’ attention from native-born students. Our findings also
may be the result of schools in immigrant-dense areas diverting limited curricular
29
See, for example, Shonkoff and Meisels (2000) for an overview of these programs and their outcomes.
32
resources to compensatory or English as a Second Language classes, to the detriment of
other students. Moreover, these schools may systematically be associated with faculties
whose social and pedagogic practices within the classroom tend to make them more
likely to view aggressive or non-normative behaviors by all students as sufficient grounds
for labelling them and exacting penalties (Drame, 2002).
Occupational Prestige
The hazard of being assigned to special education classes is substantially lower in
the least-prestigious areas of Denver, compared to places with above-average prestige
(among which there is no variation), but this relationship disappears if the child has spent
all of elementary school residing in DHA housing. From the perspective of schools, this
finding suggests that low-income, minority children may appear to perform less-well by
some relativistic indicators (e.g., grades) when they are initially exposed to a more
competitive group of (higher-status) students, resulting in increased hazards for special
education designation.30 The longer they persist in the same neighborhood and school,
however, the better they apparently adapt to the new local culture and their performance
improves to the point where they are indistinguishable from other students’. The
academic and cultural learning that takes place in such higher-prestige contexts likely
occurs jointly in school and residential spheres, abetted by peer interactions, greater
parental involvement in schools that spills over to enhanced levels of cohesion with other
30
Analogous adaptation-to-school processes were observed in the Gautreaux and MTO program
evaluations; see Briggs, Popkin and Goering (2010: 176) and Sanbonmatsu et al. (2011). Compounding
this effect may be the oft-observed designation of minority students in predominantly white schools as
mentally or emotionally retarded; see reviews in Oswald et al. (1999) and Ferri and Connor (2005). Such
designations may be “objective” or based on prejudices of administrators in schools serving more
prestigious neighborhoods.
33
neighborhood parents, and collective pro-educational norms in both (Casiano and
Massey, 2012).
It is in places with average levels of occupational prestige, however, that our
Denver Child Study sample demonstrates the lowest hazard of suspension. We think that
this is explicable by considering the interplay of countervailing factors: one related to the
probability of children evincing disciplinary problems and the other related to the
probability that school personnel will view a given breach of discipline as so egregious as
to warrant suspension. Suspension hazards are highest in the lowest-prestige areas both
because there are the greatest synergisms among peers generating many disciplinary
problems and there is the least tolerance among teachers of such behaviors (Drame,
2002). Suspension hazards may also be elevated in the highest-prestige communities if
the generic behaviors and modes of social interaction of the (relatively rare) low-income,
minority students are labelled by school staff as deviant and heavily sanctioned (Briggs,
Popkin and Goering, 2010: 176; Tajalli and Garba, 2014).
Ethnic Composition
The hazard for low-income, minority children being assigned to special education
classes is lowest in places where the percentage of African American residents is highest,
though this relationship weakens the older the child was when first occupying DHA
housing and is only manifested for above-average ranges of this ethnic concentration.
We think that this result can be explained primarily through the school aspect of context.
Heavily African American neighborhoods translate into similarly composed
elementary school student bodies, meaning that our Denver Child Study children will be
less distinguishable physically, culturally and socioeconomically. This also likely means
34
a shift in school context and climate that reduces the hazard of class or cultural bias in
assignment to special education, i.e., White teachers singling out the exceptional minority
student and/or views about children’s behavior at school (Drame, 2002).
Additionally, what may be at play here (and in the results regarding occupational
prestige) is a change in students’ perceptions about themselves and their academic
abilities relative to their student peers. As described in detail in Owens (2010), relative
deprivation theory posits that minority students from lower SES backgrounds and
neighborhoods are less likely to compare themselves and their abilities favorably with
student peers from more advantaged backgrounds and neighborhoods. When lowincome, minority students attend schools in more advantaged, White neighborhoods, they
metaphorically have become “small frogs in a big pond.” When this occurs, there is a
heightened risk that these less advantaged students begin to feel incompetent as well as
incapable of competing with their more advantaged and perceived higher ability peers.
This sense of incompetence may dampen motivations and lead not only to lower
academic expectations but actually to lower academic performance (see Espenshade,
Hale and Chung, 2005). However, if these same students attend schools with peers from
similar SES backgrounds and neighborhoods, they have the opportunity to become “big
frogs in the small pond” of a less advantaged, primarily minority school. Although these
schools may be under-resourced, attending such schools with similar-ability or lowerability peers has long been associated with higher academic performance (see for
example, Davis, 1966; Owens, 2010) because of this “big frog” effect.
35
Implications for Urban Policy
Our findings related to neighborhood occupational prestige and African American
composition hold important implications for policies aimed at enhancing opportunity as
well as socioeconomic mixing of neighborhoods and schools. On the one hand, if lowincome, minority students are able to claim their status as “big frogs” within their local
neighborhood schools, we need to enhance opportunities to strengthen the “ponds” in
which they learn. This may include working with students, parents and school personnel
to redefine the normative expectations for academic performance for the entire student
body and to change the local school and neighborhood climate to one that encourages
academic excellence for all. In order for this to occur, however, these schools need to be
equipped with the educational resources and personnel that will enable such schools and
their student bodies to flourish and serves as the foundation for the bipartisan Strong Start
for America’s Children Act (S. 1697/H.R. 3461) currently being deliberated in the U.S.
Congress.
On the other hand, if we continue to believe in expanding the educational options
available to low-income, minority students through the promotion of social mixing in
both schools and neighborhoods, we need to look at the ways in which we promote such
mixing across different groups of students, classrooms and neighborhoods. While the
extant evidence suggests the potential benefits of ethnic and socioeconomic diversity on
children’s educational attainment (for recent reviews, see Michelson and Bottia, 2010;
Kalenberg, 2012), merely physically changing the neighborhoods where low-income,
minority children live and/or the schools they attend is insufficient to eliminate the
deleterious effects associated with coming from less favorable neighborhood contexts.
36
Just as we need to consider carefully how we integrate low-income, minority families
into the larger social fabric and networks operating in more advantaged neighborhoods
when they move to new residential locations, social mixing at the school level must
specifically address the processes whereby these students are integrated into their new
classrooms and schools. How do schools actively incorporate and engage students in
these new settings? How do schools encourage the involvement of their parents? Critical
to successful integration is the recognition of all involved of the positive contributions
that all children can make in enhancing their school environment.
Caveats
In closing, several shortcomings of our study should be acknowledged. The first
is that, as noted above, we have no measures of school resources or environment. We do
not think that this biases our results for contextual indicators (because virtually all
children analyzed attend neighborhood schools), but unfortunately it inhibits the parsing
of distinct influences from the residential neighborhood and school components of
context. Second, we recognize that Denver may not be a generalizable location in which
to study context effects, inasmuch as it has below-average levels of ethnic segregation
and concentrated poverty (Iceland, Weinberg and Steinmetz, 2002; Jargowsky, 2013).
Third, we have not investigated contextual impacts at the transitional point between
childhood and adolescence, a period where Graber and Brooks-Gunn (1996) have
claimed context has a stronger impact. Finally, we have not attempted to probe
statistically the potential paths through which context may affect elementary school
performance via exposure to violence, risky behaviors, nutrition, and health, which might
37
reveal more about underlying causal mechanisms. These latter shortcomings will be
addressed in future work.
CONCLUSION
Researchers and policymakers have struggled with the daunting challenges of
estimating the true causal impact of geographic context on elementary school
performance. An innovative public housing program instituted by the Denver Housing
Authority provides a unique opportunity to explore this issue because the DHA mimics a
random assignment to a wide range of locations for families with children who apply for
DHA housing.
Using a Cox proportional hazard model with instrumental contextual variables,
we find that several socioeconomic and demographic indicators are statistically and
substantively important predictors of low-income, Latino and African American
children’s difficulties in elementary school, though sometimes in nonlinear and
interactive ways. Though we cannot distinguish whether these effects are produced
primarily in the residential neighborhood or school spheres, we can conclude that urban
geography matters. Our natural experiment and identification strategy permits the
confident assessment that findings are not influenced by selection on unobserved
characteristics of caregivers.
REFERENCES
Aaronson, D. (1998). Using sibling data to estimate the impact of neighborhoods on
children's educational outcomes, Journal of Human Resources 33: 915-946.
Acevedo-Garcia, D., Lochner, K., Osypuk, R., and Subramanian, S. (2003). Future
directions in residential segregation and health research: A multi-level approach.
American Journal of Public Health 93: 215-221.
Ainsworth, J. (2002). Why does it take a village? The mediation of neighborhood effects
on educational achievement. Social Forces 81: 1 17-52.
Anderson, E. (1990). Streetwise: Race, class and change in an urban community.
Chicago: University of Chicago Press.
Aneshensel, C. and Sucoff, C. (1996). The neighborhood context and adolescent mental
health. Journal of Health and Social Behavior 37: 293-310.
Bellair, P.E. and Roscigno, V.J. (2000). Local labor market opportunity and adolescent
delinquency. Social Forces 78: 1509-1538.
Bennett, P. R. (2011). The relationship between neighborhood racial concentration and
verbal ability: An investigation using the institutional resources model. Social
Science Research 40(4): 1124-1141.
Briggs, X., Ferryman, K., Popkin, S. and Rendon, M. (2008). Why did the Moving to
Opportunity experiment not get young people into better schools? Housing Policy
Debate 19: 53-91.
Briggs, X., Popkin, S. and Goering, J. (2010). Moving To Opportunity. NY: Oxford
University Press.
Bronfenbrenner, U., and Morris, P. (1998). The ecology of developmental processes. In
Lerner, R.M., (Ed., Handbook of child psychology: Vol. 1. Theoretical models of
human development, 5th ed. (pp. 993-1028). New York: Wiley.
Brooks-Gunn, J., Duncan, G. and Aber, J.,Eds,. (1997). Neighborhood poverty: vol. 1
Context and consequences for children. New York: Russell Sage Foundation.
Bouchard, T. (1997). IQ Similarity in twins reared apart: Findings and responses to
critics. In Steinberg, R. J. and Grigorenko, E., Eds., Intelligence, heredity, and
environment (pp. 126-62). New York: Cambridge University Press.
Burdick-Will, J., Ludwig, J., Raudenbush, S., Sampson, R., Sanbonmatsu, L.,
and Sharkey, P. (2010). Converging evidence for neighborhood effects on
children's test scores: An experimental, quasi-experimental, and observational
comparison. Washington, DC: Brookings Institution.
Carlson, D. and Cowan, J. (2015). Student Neighborhoods, Schools, and Test Score
Growth: Evidence from Milwaukee, Wisconsin. Sociology of Education 88(1):
38–55.
Casciano, R. and Massey, D. (2012). School Context and Educational Outcomes: Results
from a Quasi-Experimental Study, Urban Affairs Review 48: 180-204.
Case, A. and Katz, L. (1991). The company you keep: The effects of family and
neighborhood on disadvantaged youth. NBER Working Paper 3705. Cambridge,
MA: National Bureau of Economic Research.
Ceballo, R., McLoyd, V. and Toyokawa, T. (2004). The influence of
neighborhood quality on adolescents’ educational values and school effort.
Journal of Adolescent Research 19: 716-739.
Clampet-Lundquist, S. (2007) No more ‘bois ball: The impact of relocation from public
housing on adolescents. Journal of Adolescent Research 22: 298-323.
Clampet-Lundquist, S., and Massey, D. (2008). Neighborhood effects on economic selfsufficiency: A reconsideration of the Moving To Opportunity experiment.
American Journal of Sociology 114:107-143.
Connell, J.P., Halpern-Felsher, B.L., Clifford, E., Chichlow, W. and Usinger, P. (1995).
Hanging in there: Behavioral, psychological and contextual factors affecting
whether African American adolescents stay in high school. Journal of Adolescent
Research 10: 41-63.
Cook, T., Herman, M., Phillips, M. and Settersten, R. (2002). ‘Some Ways in Which
Neighborhoods, Nuclear Families, Friendship Groups, and Schools Jointly Affect
Changes in Early Adolescent Development. Child Development 73: 1283-1309.
Cornell, J. and Halpern-Felsher, B. (1997) How neighborhoods affect educational
outcomes in middle childhood and adolescence. In Brooks-Gunn, J., Duncan, G.
and Aber, J., Eds., Neighborhood poverty: Vol. 1, Context and consequences for
children (pp. 174-199). New York: Russell Sage Foundation.
Corcoran, M., Gordon, R., Laren, D., and Solon, G. (1992). The association between
men’s economic status and their family and community origins, Journal of
Human Resources 27: 575-601.
Crane, J. (1991). The epidemic theory of ghettos and neighborhood effects on dropping
out and teenage childbearing. American Journal of Sociology 96: 1226-1259.
Crowder, K. and South, S. (2003). Neighborhood distress and school dropout: The
variable significance of community context. Social Science Research 32:
659–698
_____. (2011). Spatial and temporal dimensions of neighborhood effects on high school
graduation. Social Science Research 40: 87-106.
Datcher, L. (1982). Effects of community and family background on achievement.
Review of Economics and Statistics 64: 32-41.
Davis, J. A. (1966). The campus as a frog pond: An application of the theory of relative
deprivation to career decisions of college men. American Journal of Sociology
72: 17-31.
DeLuca, S. and Dayton, E. (2009). Switching contexts: The effects of housing mobility
and school choice programs on youth outcomes. Annual Review of Sociology 35:
457-491.
DeLuca, S., Duncan, G., Mendenhall, R. and Keels, M. (2010). Gautreaux mothers and
their children. Housing Policy Debate 20: 7-25.
Dietz, R. (2002). The estimation of neighborhood effects in the social sciences. Social
Science Research 31: 539-575.
Drame, E. (2002). Sociocultural context effects on teachers' readiness to refer for
learning disabilities. Exceptional Children 69(1): 41-53.
Duncan, G. (1994). Families and neighbors as sources of disadvantage in the schooling
decisions of White and Black adolescents. American Journal of Education 103:
20-53.
Duncan, G., Connell, J. and Klebanov, P. (1997). Conceptual and methodological issues
in estimating causal effects of neighborhoods and family conditions on individual
development. In Brooks-Gunn, J., Duncan, G. and Aber, J., Eds., Neighborhood
poverty: Vol. 1, Context and consequences for children (pp. 219-250). New
York: Russell Sage Foundation.
Duncan, G., Gennetian, L. and Norris, P. (2007-08). Effects of welfare and antipoverty
programs on participants’ children. Focus 25 (2): 3-11.
Duncan, G. and Raudenbush, S. (1999) Assessing the effect of context in studies of child
and youth development, Educational Psychology 34: 29-41.
Duncan, Greg, Yeung, W., Brooks-Gunn, J., and Smith, J. (1998). How much does
childhood poverty affect the life chances of children? American Sociological
Review 63: 406-23.
Ensminger, M., Lamkin, R., and Jacobson, N. (1996) School leaving: A
longitudinal perspective including neighborhood effects. Child Development 67:
2400-2416.
Espenshade, T., Hale, L., and Chung, C. (2005). The frog pond revisited: High school
academic context, class rank, and elite college admission. Sociology of Education
78: 269-293.
Farran, D. (2000). Another decade of intervention for children who are low income or
disabled. What do we know now? Pp. 510-548 in J. Shonkoff and S. Meisels,
Eds., Handbook of Early Childhood Intervention. Cambridge, UK: Cambridge
University Press.
Fauth, R., Leventhal, T., and Brooks-Gunn, J. (2007). Welcome to the neighborhood?
Long-term impacts of moving to low-poverty neighborhoods on poor children's
and adolescents' outcomes. Journal of Research on Adolescence 17: 249-84.
Feinstein, L. and Symons, J. (1999). Attainment in secondary school. Oxford Economic
Papers 51: 300-321.
Ferri, B. and Connor, D. (2005). In the Shadow of Brown Special Education and
Overrepresentation of Students of Color. Remedial and Special Education 26(2):
93-100.
Ferryman, K, Briggs, X., Popkin, S., and Rendon, M. (2008). Do better neighborhoods
for MTO families mean better schools? Washington, DC: Urban
Institute, Metropolitan Housing and Communities Center Brief Number 3, March.
Francois, S., Overstreet, S., & Cunningham, M. (2012). Where We Live: The
Unexpected Influence of Urban Neighborhoods on the Academic Performance of
African American Adolescents. Youth and Society 44(2): 307-328.
Furstenberg, F., Cook, T., Eccles, J. and Elder, G. (1999). Managing to make it. Chicago,
IL: University of Chicago Press.
Furtado, B. (2011). Neighbourhoods in urban economics: Incorporating cognitively
perceived urban space in economic models, Urban Studies 48: 2827-2848.
Galster, G. (2001). On the nature of neighborhood. Urban Studies 38 (12): 2111-2124.
_____ (2008). Quantifying the effect of neighbourhood on individuals: Challenges,
alternative approaches and promising directions. Journal of Applied Social
Science Studies [Schmollers Jahrbuch / Zeitscrift fur Wirtschafts- und
Sozialwissenschaften] 128: 7-48.
_____ (2011) Changing the Geography of Opportunity by Helping Poor Households
Move Out of Concentrated Poverty: Neighborhood Effects and Policy Design,”
Pp. 221-234 in H. Newburger, E. Birch and S. Wachter, eds., Neighborhood and
Life Chances: How Place Matters in Modern America. Philadelphia: University
of Pennsylvania Press.
_____. (2012). The mechanism(s) of neighborhood effects: Theory, evidence, and policy
implications. In Van Ham, M., Manley D., Bailey N., Simpson L., and Maclennan
D., Eds., Neighbourhood effects research: New perspectives (pp. 23-56).
Dordrecht, NL: Springer.
_____. (2014) Nonlinear and threshold aspects of neighborhood effects,” Pp. 117-133
in Soziale Kontexte und soziale Mechanismen [Social Contexts and Social
Mechanisms]. Edited by J. Friedrichs and A. Nonnenmacher. Wiesbaden, GER:
Springer. doi: 10.1007/s11577-014-0268-3
Galster, G., Andersson, R. and Musterd, S. (2010). Who is affected most by
neighbourhood income mix? Urban Studies 48: 2915-2944.
_____. (2014) “Are males’ incomes influenced by the income mix of their male
neighbors? Explorations into nonlinear and threshold effects in Stockholm,”
Housing Studies 30; doi: 10.1080/02673037.2014.931357
Galster, G., Marcotte, D., Mandell, M., Wolman, H. & Augustine, N. (2007). The impact
of childhood neighborhood poverty on young adult outcomes. Housing Studies
22: 723-752.
Galster, G. & Santiago, A. (2006). What’s the ‘Hood Got to Do with It? Parental
Perceptions about How Neighborhood Mechanisms Affect Their Children,
Journal of Urban Affairs 28 (3): 201-226.
Galster, G., Tatian, P., Santiago, A., Pettit, K. and Smith, R. (2003). Why NOT in My
Back Yard? The Neighborhood Impacts of Assisted Housing. New Brunswick,
NJ: Rutgers University Center for Urban Policy Research / Transaction Press.
Gennetian, L., Ludwig, J. and Sanbonmatsu, L. (2011). An overview of Moving To
Opportunity. In Newburger, H., Birch, E. and Wachter, S., Eds., Neighborhood
and life chances: How place matters in modern America (pp. 163-178).
Philadelphia: University of Pennsylvania Press.
Gephart, M. (1997) Neighborhoods and communities as contexts for development. In
Brooks-Gunn, J., Duncan, G. and Aber, J., Eds. Neighborhood poverty: Vol. I.
Context and consequences for children (pp. 1-43). New York: Russell Sage
Foundation.
Graber, J. and Brooks-Gunn, J. (1996). Transitions and turning points: Navigating the
passage from childhood through adolescence. Developmental Psychology 32:
768-776.
Hanushek, E. (1971). Teacher characteristics and gains in student achievement:
estimation using micro-data. American Economic Review, 61(2), 280-288.
Harding, D. (2003). Counterfactual models of neighborhood effects: The effect of
neighborhood poverty on high school dropout and teenage pregnancy. American
Journal of Sociology 109: 676-719.
Harding, D. (2009). Collateral consequences of violence in disadvantaged
neighborhoods. Social Forces 88(2): 757-784.
Harding D., Gennetian L., Winship, C., Sanbonmatsu, L., and Kling, J. (2011).
Unpacking Neighborhood Influences on Education Outcomes: Setting the Stage
for Future Research. In Whither Opportunity: Rising Inequality, Schools, and
Children’s Life Chances, ed. G. Duncan and R. Murnane, pp. 277-296.
New York: Russell Sage; Chicago: Spencer Foundation.
Henrich, C., Schwab-Stone, M., Fanti, K., Jones, S., and Ruchkin, V. (2004). The
association of community violence exposure with middle-school achievement:
A prospective study. Applied Developmental Psychology 25: 327-348.
Hotz, V. and Pantano, J. (2014). “Strategic Parenting, Birth Order and School
Performance.” Cambridge, MA: NBER Working Paper No. 19542.
Hoxby, C. (2001). Peer effects in the classroom: Learning from gender and race
variation. Cambridge, MA: NBER Working Paper No. 7867.
Hurt, H., Malmud, E., Brodsky, N., and Giannetta, J. (2001). Exposure to violence:
Psychological and academic correlates in child witnesses. Archives of Pediatrics
and Adolescent Medicine 155: 1351-1356.
Iceland, J., Weinberg, D. and Steinmetz, E. (2002) Racial and Ethnic Residential
Segregation in the United States: 1980-2000. Washington, DC: U.S. Government
Printing Office, U.S. Census Bureau, Series CENSR-3.
Jacob, B. (2004). Public housing, housing vouchers, and student achievement: Evidence
from public housing demolitions in Chicago. American Economic Review 94:
233-58.
Jargowsky, P. (2013). Concentrated Poverty in New Millenium. New York:
Century Foundation. Available at:
http://tcf.org/bookstore/detail/concentration-of-poverty-in-the-new-millennium
Jargowsky, P. and El Komi, M. (2011). Before or after the bell? School context and
neighborhood effects on student achievement. In Newburger, H., Birch, E. and
Wachter, S., Eds., Neighborhood and life chances: How place matters in modern
America (pp. 50-72). Philadelphia: University of Pennsylvania Press.
Jencks, C. and Mayer, S. (1990) The social consequences of growing up in a poor
neighborhood,. In Lynn, L. and McGeary, M., Eds., Inner-city poverty in the
United States (pp. 111-186). Washington, DC: National Academy Press.
Johnson, O. (2007). Ecology in educational theory: Thoughts on stratification, social
mobility and proximal capital. Urban Review 40: 227-46.
Johnson, O. (2010). Assessing neighborhood racial segregation and macroeconomic
effects in the education of African Americans. Review of Educational Research
80(4): 527-575.
Kahlenberg, R., Ed., (2012). The future of school integration: Socioeconomic diversity
as an education reform strategy. New York: Century Foundation Press.
Kauppinen, T. (2007) Neighborhood effects in a European city: Secondary education of
young people in Helsinki. Social Science Research 36: 421-444.
Klebanov, P., Brooks-Gunn, J. and Duncan, G. (1994). Does neighborhood and family
poverty affect mothers’ parenting, mental health, and social support? Journal of
Marriage and the Family 56: 441-455.
Kling, J., Liebman, J. and Katz, L. (2007). Experimental analysis of neighborhood
effects. Econometrica 75: 83-119.
Kohen, D., Brooks-Gunn, J., Leventhal, T. and Hertzman, C. (2002). Neighborhood
income and physical and social disorder in Canada: Associations with young
children's competencies. Child Development 73: 1844-60.
Lavy, V., Silva, O. and Weinhardt, F. (2009). The good, the bad and the average:
Evidence on the scale and nature of ability peer effects in schools. Cambridge,
MA: NBER Working Paper No. 15600, December.
Leventhal, T. and Brooks-Gunn, J. (2000). The neighborhoods they live in. Psychological
Bulletin 126: 309-337.
Leventhal, T., & Brooks-Gunn, J. (2004). A randomized study of neighborhood effects
on low-income children's educational outcomes. Developmental Psychology
40(4): 488-507.
Leventhal, T., Fauth, R., and Brooks-Gunn, J. (2005). Neighborhood poverty and public
policy: a 5-year follow-up of children's educational outcomes in the New York
City Moving To Opportunity demonstration. Developmental Psychology 41: 933952.
Ludwig, J., Ladd, H. and Duncan, G. (2001). Urban poverty and educational
outcomes. In Gale, W. and Pack , J.R., Eds., Brookings Wharton papers in urban
affairs: 2001 (pp. 147-201). Washington DC: Brookings Institution Press.
Ludwig, J. et al.(2008). What can we learn about neighborhood effects from the Moving
To Opportunity Experiment? American Journal of Sociology 114: 144-188.
Manski, C. (1995). Identification problems in the social sciences. Cambridge, MA:
Harvard University Press.
McCoy, D., Roy, A. and Sirkman, G. (2013). Neighborhood crime and school climate as
predictors of elementary school academic quality: A cross-lagged panel analysis.
American Journal of Community Psychology 52: 128-140.
Mickelson, R. and Bottia, M. (2010). Integrated education and mathematics outcomes:
A synthesis of social science research. North Carolina Law Review 87:
993–1089.
Muhajarine, N., Labonte, R., Williams, A., and Randall, J. (2008). Person, perception,
and place: What matters to health and quality of life. Social Indicators Research
85: 53-80.
Musterd, S., Galster, G. and Andersson, R. (2012). Temporal dimensions
and the measurement of neighbourhood effects, Environment and Planning A
44(3): 605–627.
Newburger, H., Birch, E. and Wachter, S., Eds. (2011). Neighborhood and Life Chances:
How Place Matters in Modern America. Philadelphia: University of
Pennsylvania Press.
Niewenhuis, J. and Hooimeijer, P. (2014). The Association between Neighbourhoods and
Educational Achievement: A Systematic Review and Meta-Analysis, pp. 12-37
in: Nieuwenhuis, J., Neighbourhood effects on youth’s achievements: the
moderating role of personality. Utrecht (NL) University: Department of
Sociology PhD Dissertation.
Oswald, D., Coutinho, M., Best, A. and Singh, N. (1999). Ethnic representation in special
education: The influence of school-related economic and demographic variables.
The Journal of Special Education 32(4): 194-206.
Owens, A. (2010). Neighborhoods and Schools as Competing and Reinforcing Contexts
for Educational Attainment. Sociology of Education 83: 287-311.
Ratner, H., Chiodo, L., Covington, C., Sokol, R., Ager, J. and Delaney-Black, V. (2006).
Violence exposure, IQ, academic performance, and children’s perception of
safety: Evidence of protective effects. Merrill-Palmer Quarterly 52(2):
264-287.
Reardon, S. and Raudenbush, S. (2009). Assumptions of value-added models in
estimating school effects. Education and Finance Policy 4: 492-519.
Rosenbaum, J. (1995). Changing the geography of opportunity by expanding residential
choice: Lessons from the Gautreaux program, Housing Policy Debate 6: 231-269.
Rosenbaum, J., DeLuca, S. and Tuck, T. (2005). New capabilities in new places: Lowincome families in the suburbs. In Briggs, X., ed., The Geography of Opportunity
(pp. 150-175). Washington, DC: Brookings Institution Press.
Rothstein, J. (2010). Teacher quality in educational production: Tracking, decay, and
student achievement. Quarterly Journal of Economics 125 (1): 175-214.
Sampson, R. (2001) “How do communities undergird or undermine human development?
Relevant contexts and social mechanisms.” In Booth, A. and Crouter, A., Eds.
Does it take a village? Community effects on children, adolescents, and family
(pp. 3-30). Mahwah, NJ: Lawrence Erlbaum Associates.
Sampson, R. (2008). Moving to inequality: Neighborhood effects and experiments meet
social structure. American Journal of Sociology 114(1): 189-231.
Sampson, R., Morenoff, J., and Gannon-Rowley, T. (2002). Assessing ‘neighborhood
effects’: Social processes and new directions in research. Annual Review of
Sociology 28: 443-478.
Sampson, R., Sharkey, P. and Raudenbush, S. (2008). Durable effects of concentrated
disadvantage on verbal ability among African-American children. Proceedings of
the National Academy of Sciences 105: 845-852.
Sanbonmatsu, L., Kling, J., Duncan, G. and Brooks-Gunn, J. (2006). Neighborhoods and
academic achievement: Evidence from the Moving To Opportunity experiment.
Journal of Human Resources 41: 649-691.
Sanbonmatsu, L., Ludwig, J., Katz, L., Gennetian, L., Duncan, G., Kessler, R., Adam, E.,
McDade, T., and Lindau, S. (2011). Impacts of the Moving to Opportunity for
Fair Housing Demonstration Program after 10 to 15 Years. Washington, DC:
U.S. Department of Housing and Urban Development, Office of Policy
Development and Research.
Sastry N. (2012). Neighborhood Effects on Children’s Achievement: A Review of Recent
Research. In Oxford Handbook on Child Development and Poverty ed. King. R.
and Maholmes, V. pp. 423-447. New York: Oxford University Press.
Schwartz, H. (2010). Housing policy is school policy: Economically integrative housing
promotes academic achievement in Montgomery County, MD. NY: Century
Foundation.
Sharkey, P. and Sampson, R. (2010). The acute effect of local homicides on children's
cognitive performance. Proceedings of the National Academy of Sciences of the
United States of America 107(26): 11733-11738. doi:10.1073/pnas.1000690107
Sharkey, P., Tirado-Strayer, N., Papachristos, A., and Raver, C. (2012). The effect of
local violence on children's attention and impulse control. American Journal of
Public Health 102(12); 2287.
Sharkey, P. and Faber, J. (2014). Where, when, why, and for whom do residential
contexts matter? Moving away from the dichotomous understanding of
neighborhood effects. Annual Review of Sociology 40:559–79.
Sharkey, P., Schwartz, A., Ellen, I. and Lacoe J. (2014). High Stakes in the Classroom,
High Stakes on the Street: The Effects of Community Violence on Students’
Standardized Test Performance. Sociological Science 1: 199-220.
Shonkoff, J. and Phillips, D., Eds. (2000). From neurons to neighborhoods: The science
of early childhood development. Washington, D.C.: National Academy Press.
Shonkoff, J. and Meisels, S., Eds. (2000). Handbook of early childhood intervention.
Cambridge, UK: Cambridge University Press.
Shonkoff, J., Garner, A., et al. (2012). The Lifelong Effects of Early Childhood Adversity
and Toxic Stress, Pediatrics. doi:10.1542/peds.2011-2663.
Slavin, R., Karweit, N. and Wasik, B. (1993). Preventing early school failure: What
works? Educational Leadership 50: 10-18.
Small, M. and Feldman, J. (2012). Ethnographic evidence, heterogeneity and
neighbourhood effects after Moving To Opportunity. In Van Ham, M., Manley
D., Bailey N., Simpson L., and Maclennan D., Eds., Neighbourhood effects
research: New perspectives (pp. 57-78). Dordrecht, NL: Springer.
South, S., Baumer, E. and Lutz, A. (2003). Interpreting community effects on youth
educational attainment. Youth and Society 35: 3-36.
Sykes, B. and Musterd, S. (2011). Examining neighbourhood and school
effects simultaneously: What does the Dutch evidence show?
Urban Studies 48: 1307-1331.
Tajalli, H. and Garba, H. (2014). Discipline or Prejudice? Overrepresentation of Minority
Students in Disciplinary Alternative Education Programs. The Urban Review
46: 620-631.
Tate, W., Ed. (2012). Research on Schools, Neighborhoods and Communities. Lanham,
MD: Rowman and Littlefield Publishers, Inc.
Vartanian, T. and Gleason, P. (1999). Do neighborhood conditions affect high school
dropout and college graduation rates? Journal of Socio-Economics 28: 21-24.
Wilson, W.J. (1987) The truly disadvantaged. Chicago: University of Chicago Press.
Wodtke, G., Harding, D. and Elwert, F. (2011). Neighborhood effects in
temporal perspective: The impact of long-term exposure to concentrated
disadvantage on high school graduation. American Sociological Review 76:
713-736.
Ziol-Guest, K. and McKenna, C. (2014). Early Childhood Housing Instability and School
Readiness. Child Development 85 (1): 103–113.
Table 1: Descriptive Statistics for Neighborhoods of Denver Child Study Children,
Households and Denver County
Neighborhood Indicators
Social Vulnerability Index
Occupational Prestige Score
Percentage of African American Residents
Percentage of Latino Residents
Percentage of Foreign Born Residents
Percentage of Dwellings Built Pre-1940
Violent Crime Rate (per 1,000 population)
Property Crime Rate (per 1,000 population)
For Denver Child Study Children^
Mean Std. Dev. Min.
Max.
142.3
36.1
17.8
52.6
19.4
29.8
13.9
77.3
53.5
2.6
19.1
19.5
8.9
18.6
8.3
42.0
44.1
31.6
0.6
5.6
2.6
0.2
1.1
14.6
278.7
47.3
82.5
85.3
55.1
75.2
48.6
237.4
For Sample Households^^
Mean Std. Dev. Min.
Max.
163.4
35.1
15.6
55.2
16.2
31.5
15.0
97.7
58.8
3.1
17.8
18.4
8.6
19.1
10.0
56.8
18.8
30.3
0.0
4.6
0.0
0.0
0.8
2.2
284.4
46.9
91.6
86.5
45.1
75.0
51.1
357.3
^ Average measured during ages 6-11 (or from age 6 through age of outcome if occurred) for 410 children in Table 1
^^ Measured for year of offer of first dwelling by DHA
For Denver, 2000
Mean Std. Dev.
96.7
41.0
11.5
28.9
15.8
25.1
5.6
46.8
42.0
4.4
16.7
24.6
10.8
25.7
6.7
43.6
18
Table 2. Descriptive Statistics for Denver Child Study Children and their Households
Mean Std. Dev.
Outcomes During Elementary School
0.12
Placement in Special Education
7.84
Age When First Placed in Special Education
0.09
Repeated a Grade
8.42
Age When First Repeated a Grade
0.07
Suspended from School
9.42
Age When First Suspended
Covariates
Race/Ethnicity of Child (omitted=African American Male)
0.28
Latina Female
0.29
Latino Male
0.21
African American Female
0.25
Caregiver has Depressive Symptomatology^^^
0.17
Caregiver is Immigrant
0.48
Caregiver has HS Diploma or Better^
0.33
Proportion Residing with Two Parents^^
6.91
Household Income (natural logarithm)^^
1.26
Household Economic Stressor Scale^^
1.28
Cumulative Number of Household Moves^
Max
Min
0.33
1.69
0.29
1.95
0.25
1.30
0
6
0
6
0
6
1
11
1
11
1
11
0.45
0.45
0.41
0.43
0.37
0.50
0.41
4.24
1.14
1.17
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
10.75
5
7
N = 410 children in 246 families; source: Denver Child Study
^ At the time that outcome for focal child first occurred, or age 11 if never
^^ Average during ages 6-11 (or from age 6 through age of outcome if occurred)
^^^ At time of survey
Table 3. First Stage OLS Regression Results for Generating Instruments
Exogenous Predictors
Social Vulnerability
Covariates in Second Stage
Coeff. Std. Error
Race/Ethnicity of Child (omitted=African American Male)
Latina Female
1.47
(6.33)
Latino Male
0.32
(6.38)
African American Female
-6.68
(4.47)
Caregiver has Depressive Symptomatology^^^ 7.92
(5.29)
Caregiver is Immigrant
-2.17
(7.07)
Caregiver has HS Diploma or Better^
-3.93
(4.71)
Proportion Residing with Two Parents^^
-5.31
(5.82)
Natural Log of Household Income^^
-0.38
(0.57)
Household Economic Stressor Scale^^
-2.27
(2.29)
Cumulative # of Household Moves^
-1.35
(1.84)
Timing of DHA First Offer
offer1988
-1.85
(11.1)
offer1989
-1.22
(12.8)
offer1990
-3.73
(9.96)
offer1991
0.044
(15.0)
offer1992
-27.0*
(11.7)
offer1993
-10.6
(16.4)
offer1994
2.18
(10.8)
offer1995
-10.9
(9.69)
offer1996
19.3
(12.0)
offer1997
11.1
(11.4)
offer1998
10.1
(11.6)
offer1999
15.8
(12.3)
offer2000
17.5
(11.3)
offer2001-2003
24.5*
(11.4)
Neighborhood of DHA First Offer
Social Vulnerability Index
0.37***
(0.065)
Percent Foreign Born
-0.31
(0.37)
Percent African American
-0.18
(0.13)
Percent Dwellings Built Pre-1940
-0.12
(0.13)
Occupational Prestige Score
-0.56
(1.04)
Property Crime Rate
0.052
(0.068)
Violent Crime Rate
0.83*
(0.33)
Constant
100.7*
(45.4)
r-sq
0.37
Dependent Variables: Average Neighborhood Conditions During Elementary School
% Foreign Born
% African American % Dwellings Built Pre-1940 Occupational Prestige Property Crime Rate Violent Crime Rate
Coeff. Std. Error
Coeff.
Std. Error
Coeff.
Std. Error
Coeff.
Std. Error
Coeff.
Std. Error
Coeff. Std. Error
3.27**
3.32**
0.66
0.10
3.13**
-0.63
-2.29*
0.10
0.82*
-0.33
(1.02)
(1.05)
(0.85)
(0.86)
(1.09)
(0.83)
(0.89)
(0.096)
(0.38)
(0.30)
-11.7***
-12.4***
0.96
-0.071
-1.47
1.87
-0.71
0.041
-0.18
-0.97
(2.27)
(2.14)
(1.86)
(2.19)
(1.74)
(1.71)
(1.66)
(0.18)
(0.86)
(0.62)
0.37
1.50
2.76
-1.04
1.42
-4.00*
-2.15
0.060
0.054
-0.46
(2.24)
(2.31)
(1.81)
(2.03)
(1.95)
(1.85)
(1.95)
(0.19)
(0.81)
(0.62)
-0.70*
-0.73*
0.20
-0.46
0.022
-0.054
0.79**
-0.012
0.10
0.094
(0.31)
(0.32)
(0.29)
(0.24)
(0.34)
(0.26)
(0.30)
(0.030)
(0.11)
(0.092)
-4.77
-1.53
-4.04
11.5*
-7.93
-5.02
3.92
0.011
-2.76
-0.56
(5.45)
(5.72)
(4.30)
(4.97)
(5.77)
(3.79)
(4.76)
(0.46)
(1.95)
(1.70)
-2.16*
-2.29*
0.28
1.70
-2.07*
-0.67
1.31
-0.10
-0.041
-0.27
(1.03)
(1.03)
(0.90)
(0.90)
(0.90)
(0.76)
(1.01)
(0.10)
(0.32)
(0.28)
-2.26
0.77
2.49
-1.79
6.89**
9.78
1.95
1.30
2.89
2.68
2.09
1.02
7.23**
0.46
(1.62)
(1.67)
(1.59)
(1.53)
(2.43)
(5.05)
(1.68)
(1.87)
(2.25)
(1.97)
(1.86)
(2.24)
(2.45)
(2.82)
4.91
-0.48
2.31
-1.66
-5.93
-4.00
-2.40
-2.46
-3.68
-4.44
-5.82
-4.05
-2.91
-1.07
(3.45)
(4.42)
(4.27)
(4.33)
(3.38)
(5.00)
(2.74)
(3.49)
(3.45)
(2.77)
(3.24)
(3.24)
(3.28)
(3.78)
4.43
0.58
3.44
-1.40
1.89
11.7**
2.89
2.36
-2.91
0.56
-0.66
2.01
-2.47
0.052
(3.63)
(4.81)
(4.08)
(4.54)
(4.73)
(3.95)
(3.23)
(3.61)
(3.76)
(3.89)
(4.48)
(4.63)
(3.85)
(4.98)
-0.032
-0.66
-0.20
0.24
0.0032
-0.046
-0.14
0.057
-0.99
-0.69
-0.69
-0.60
-0.89
-1.77*
(0.61)
(0.56)
(0.51)
(0.72)
(0.61)
(0.76)
(0.52)
(0.46)
(0.51)
(0.53)
(0.72)
(0.61)
(0.91)
(0.75)
-5.98
-3.73
-8.04
-4.57
-23.1
-19.2
-6.01
-7.59
9.77
-6.20
1.83
3.23
4.74
11.3
(9.25)
(10.2)
(9.11)
(12.1)
(12.9)
(9.82)
(9.53)
(7.86)
(11.8)
(8.58)
(8.39)
(9.26)
(10.7)
(10.8)
1.45
1.30
0.65
-0.94
-3.37*
-2.03
-0.74
-1.92
1.40
-0.87
0.093
0.54
0.13
3.86
(1.82)
(1.90)
(1.59)
(1.85)
(1.65)
(1.89)
(1.51)
(1.55)
(1.74)
(1.56)
(1.93)
(1.66)
(1.66)
(2.82)
-0.0084
0.54***
-0.0028
0.0064
-0.23
-0.0028
0.086
15.0*
0.40
(0.011)
(0.088)
(0.023)
(0.023)
(0.17)
(0.0098)
(0.049)
(7.25)
0.037
0.043
0.60***
-0.043
0.60
-0.011
-0.15
-6.23
0.52
(0.028)
(0.11)
(0.050)
(0.049)
(0.48)
(0.025)
(0.15)
(18.8)
0.053*
0.13
0.0064
0.50***
0.63
-0.047*
0.16
-17.2
0.37
(0.027)
(0.13)
(0.047)
(0.056)
(0.54)
(0.023)
(0.12)
(21.6)
-0.0023
0.018
0.0037
0.016*
0.44***
0.0013
-0.030
20.9***
0.33
(0.0038)
(0.022)
(0.0082)
(0.0066)
(0.072)
(0.0036)
(0.016)
(2.94)
-0.094
-0.32
-0.10
-0.081
0.14
0.43***
0.66*
54.0
0.38
(0.057)
(0.29)
(0.14)
(0.12)
(0.85)
(0.060)
(0.29)
(34.6)
-0.0035
-0.0040
-0.030
0.025
0.10
0.023*
0.39***
4.10
0.42
(0.010)
(0.051)
(0.020)
(0.024)
(0.21)
(0.010)
(0.063)
(8.41)
Clustered standard errors in parentheses; * p<.05; ** p<.01; *** p<.001
^ At the time that outcome (in this case, special education) for focal child first occurred, or age 11 if never
^^ Average during ages 6-11 (or from age 6 through age of outcome if occurred)
^^^ At time of survey
1
Table 4. Cox Proportional Hazard IV Models of Elementary School Outcomes
Special Education
Std. Hazard
Covariates
Coeff. Error
Ratio
Race/Ethnicity of Child (omitted=African American Male)
Latina Female
-0.75
(0.65) 0.472
Latino Male
-0.69
(0.66) 0.502
African American Female
-0.93
(0.61) 0.395
Caregiver has Depressive Symptomatology^^^ 0.41
(0.34) 1.507
Caregiver is Immigrant
-0.76
(0.56) 0.468
Caregiver has HS Diploma or Better^
0.36
(0.39) 1.433
Proportion Residing with Two Parents^^
-0.12
(0.18) 0.887
Natural Log of Household Income^^
-0.42** (0.14) 0.657
Household Economic Stressor Scale^^
0.40
(0.21) 1.492
Cumulative # of Household Moves^
0.37** (0.12) 1.448
Neighborhood Characteristics (IVs)
Social Vulnerability Index
0.44
(0.33) 1.553
Occupational Prestige Score
0.41* (0.20) 1.507
Percent African American
-1.21* (0.47) 0.298
Percent Foreign Born
0.16
(0.17) 1.174
Percent Dwellings Built Pre-1940
-0.16
(0.26) 0.852
Violent Crime Rate
-0.41
(0.52) 0.664
Property Crime Rate
0.22
(0.39) 1.246
Observations
409
N_family clusters
245
Chi-sq [coeffs. of all vars.=0]
78.5***
Chi-sq [coeffs. of neigh'd. vars.=0]
16.3*
log pseudolikelihood
-116.7
Repeated a Grade
Suspended from School
Std. Hazard
Std. Hazard
Coeff. Error
Ratio Coeff. Error
Ratio
-0.27
(1.11)
-0.41
(1.13)
0.26
(0.60)
-0.039 (0.47)
-1.12
(0.67)
0.81
(0.44)
-0.040 (0.23)
-0.72*** (0.19)
-0.18
(0.27)
0.40* (0.16)
0.763
0.664
1.297
0.962
0.326
2.248
0.961
0.487
0.835
1.492
-3.83**
-1.57
-0.47
0.75
-0.35
0.90
0.48
-0.31
0.73**
0.82***
(1.44)
(1.09)
(0.42)
(0.41)
(0.76)
(0.60)
(0.32)
(0.20)
(0.27)
(0.16)
0.022
0.208
0.625
2.117
0.705
2.460
1.616
0.733
2.075
2.270
-0.65
0.057
-0.59
0.61*
-0.16
0.27
0.39
402
241
56.8***
12.9
-193.8
0.522
1.059
0.554
1.840
0.852
1.310
1.477
0.77
0.11
-0.19
1.01**
0.89
-1.07
0.85
410
246
80.4***
19.4**
-66.4
(0.41)
(0.58)
(0.49)
(0.34)
(0.53)
(0.62)
(0.57)
2.160
1.116
0.827
2.746
2.435
0.343
2.340
(0.51)
(0.38)
(0.55)
(0.27)
(0.27)
(0.40)
(0.40)
Clustered standard errors in parentheses; * p<.05; ** p<.01; *** p<.001
^ At the time that outcome (in this case, special education) for focal child 1st occurred, or age 11 if never
^^ Average during ages 6-11 (or from age 6 through age of outcome if occurred)
^^^ At time of survey
2
Appendix Table 1. Pearsonian Correlations among Indicators of Geographic
Context
1. Social Vulnerability Index
2. Percentage of Foreign Born Residents
3. Percentage of African American Residents
4. Percentage of Latino Residents
5. Percentage of Dwellings Built Pre-1940
6. Occupational Prestige Score
7. Property Crime Rate (per 1,000 population)
8. Violent Crime Rate (per 1,000 population)
1
2
3
1
-0.2500*
0.1629*
0.2104*
-0.0552
-0.5008*
0.7850*
0.8880*
1
-0.0331
0.7034*
0.1135
-0.3820*
-0.2670*
-0.0629
1
-0.2872*
-0.0549
-0.1455
0.0960
0.1637*
4
5
6
7
1
0.0193
1
-0.6593* 0.3587*
1
0.1037 0.0497 -0.2146*
1
0.3328* -0.0318 -0.5310* 0.7054*
* p <.05
^ Measured at year of dropping out, or at age 18 if never (or time of survey if < 18 then)
8
1
3
APPENDIX A. INVESTIGATING QUASI-RANDOM ASSIGNMENT IN OUR DHA
NATURAL EXPERIMENT
Introduction
Although often advocated (e.g., Oakes 2004), some question whether natural
experiments can be leveraged to draw convincing implications about causal
neighborhood effects. The main reservation from doing so has been the lack of
assurance that they in fact produce a quasi-random assignment of households, and thus
convincingly avoid geographic selection bias. This appendix uses our natural
experiment involving public housing in Denver and investigates whether it convincingly
produced an essentially random allocation of households across neighborhoods.
Methods of Analyzing Randomness of Initial Assignment
A few investigations of neighborhood effects employing natural experiments have
probed the degree to which quasi-random assignment was achieved. Three methods
have been employed. First, the allocation processes employed in the natural
experiments are described in detail in an effort to uncover points at which non-random
selections could occur (e.g., Oreopoulos, 2003; Edin, Fredricksson and Åslund, 2003;
Jacob, 2004; Lyle, 2007; Piil Damm, 2009; 2014). Second, the sample of individuals
analyzed is divided across two or more locations and their mean characteristics are
compared statistically. Third, regression is used to assess whether there are any nonzero realtionships between individual characteristics and neighborhood characteristics.
We employ all three strategies here and present a fourth, original approach involving
Monte Carlo simulation.
Possibilities for Tenant Self-Selections and Staff Selections in the Denver
HousingAuthority Allocation Process
First, we explore the possibility of selection arising because the tenant can potentially
choose between two DHA units that may be located in quite different neighborhoods.
Our independent evaluation of DHA records showed that: 69.5 percent accepted their
first offer from DHA, 18.8% accepted their second offer; 7.9 percent ended up rejecting
both offers and taking a third offer later (after returning to the bottom of the wait list); 3.8
percent rejected three or more offers before being placed.
Perhaps more revealing than acceptance rates is probing whether applicants ended up
in neighborhoods they would have selected on their own. Before their initial assignment
to a DHA dwelling, clients were asked by DHA whether they had any geographic location
preferences. DHA administrative data show that 42.5 percent of the clients in our
sample did not articulate any locational preference, approximately one-third expressed
general geographical areas (i.e., Southwest Denver) while the remaining 23.5 percent
provided responses that ranged from specific addresses to specific DHA developments
(i.e., North Lincoln Campus of Learners). In order to assess whether those who stated a
preference were assigned to a housing unit in their specified area, a number of different
approaches were taken. For those who specified a particular address, we checked to
see if that address was the DHA unit to which the client was initially assigned. For those
who specified a preference for a particular DHA development, we used the unit number
reported by DHA (which has an abbreviation of the development embedded in it) to
assess whether the initial DHA unit was located within that development. For those who
4
specified a preference for a particular neighborhood, we relied on our survey data to
determine whether the original DHA unit was in the specified neighborhood. Lastly,
initially assigned DHA units were mapped to identify where within the Denver
metropolitan area they were located for those who specified a preference for a particular
part of the metro area. Once these assessments were made, we were able to calculate
frequencies and percentages for those who specified a geographic preference and got it
(N=190; 25.8 percent) and those who specified a geographic preference but didn’t get a
housing unit that met that preference (N=233; 31.7 percent). If the vast majority of
households in our sample had a strong geographic preference and were granted this
preference by the DHA assignment process, one would challenge the process as one
producing a quasi-random assignment. However, our analyses indicate to the contrary
that the vast majority of the respondents to our survey (74.2 percent) were either
instances where there was no geographic preference articulated, or where the client’s
stated preference was not honored. Since we are unable to ascertain the geographic
location of all potential DHA unit vacancies that arose during the times that each client
was assigned to their initial unit, we are unable to perform any formal statistical tests to
determine whether the frequencies we obtained for those who were assigned their
expressed preference were any different than what would be expected by chance.
A second potential source of selection can arise from the actions of the DHA staff. If
staff have multiple vacancies to consider at one time they may make dwelling offers on
the basis of observable characteristics of the applicants at the top of the waiting list.
Though our interviews with DHA staff uncovered no suggestions that this occurred, we
nevertheless must acknowledge this possibility.
In sum, a close examination of the DHA dwelling allocation process leaves substantial
room for selection. A non-trivial share of DHA applicants did not accept their first offer
from DHA (30.5 percent) and ended up in a neighborhood they said they preferred (26
percent). It may also be possible that DHA staff practiced some selection in their
dwelling offers, though we have no direct evidence of this.
Comparisons of Individual Characteristics Across Space
A second way we test the randomness of the DHA assignment process is by
ascertaining the degree to which there are any systematic patterns of where individuals
with particular characteristics end up residing in their first DHA units. In other words, we
investigate whether certain types of households end up disproportionately allocated to
particular places, whether it be due to DHA practices and/or to choices made by
applicants regarding, for instance, refusing first options. We parse space in two ways:
across DHA housing developments and by census tracts. In both variants we examine a
wide range of individual characteristics—26 variables in all—measuring attributes that
are typically gathered in surveys used in neighborhood effects research and many
others that are not (but we have acquired through our aforementioned survey). These
individual characteristics are listed in columns of Exhibit A-1.31
Our method involves regressing each individual characteristic on a series of dummy
variables. In one variant these dummies signify different DHA developments; in the
31
Note that in our study we consider only Latino and African American residents of DHA, thus we
measure only African American ethnic status, with Latino ethnicity being the reference group.
5
other they signify census tracts.32 We stratify these regressions by family size (zero or
one child; two children; three or more children) because there is a distinct geographic
pattern in Denver of where public housing units of various bedroom configurations are
located. Our test of quasi-random assignment is whether the place-based dummy
variables denoting where DHA households were originally placed are significantly
different from zero.33 If they are, we reject the null hypothesis of random assignment of
applicants to DHA dwellings.
32
The scattered-site DHA developments are not identified in their allocation process by individual
address but rather by broader geographic area encompassing several census tracts (though we
are aware of the tract of each development). This produces the seemingly anomalous situation
shown in Exhibits A-1 and A-2 where apparently many more tracts are represented than
“developments.”
33 Here the number of children in the household refers to the number of eligible children for our
study and not the total number of all children in the household. So, it is possible for households
with 0-1 eligible children to have other siblings with the same father.
Exhibit A-1A. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 0-1 Child
P/C is single parent
(1=yes, 0=no)
P/C employment
P/C received TANF P/C receiving Food
P/C had checking
P/C had health
status at time of
P/C disability status at
P/C hourly wage at
at time of DHA
Stamps at time of
account at time of insurance at time of
DHA move-in
time of survey (1=yes;
time of DHA move-in
move-in (1=yes, DHA move-in (1=yes, DHA move-in (1=yes,
DHA move-in
(1=employed, 0=not
0=no)
0=no)
0=no)
0=no)
(1=yes, 0=no)
employed)
Coeff.
-0.111
-0.500
-0.100
-7.78e-16
-1.03e-15
-0.240
-0.278
-0.167
-0.111
-3.08e-16
-7.88e-16
-0.333
-0.227
-0.154
-9.83e-16
-1.000
-0.200
-0.355
-0.250
-0.316
Coeff.
-1.133
-0.300
-0.200
0.200
-0.133
-1.080
-0.106
-0.467
-0.522
-0.800
-0.800
-0.467
-0.300
-1.031
-0.300
-0.800
-0.400
-0.284
-0.250
-0.116
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Home
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
Quigg Newton Homes
Sun Valley Annex
Pacific Place
T Bean Tower (Elderly & Disabled)
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
Thomas Connole (Elderly & Dis.)
East Village
Combined Devel-Disp Housing S.
Combined Devel-Disp Housing E.
Combined Devel-Disp Housing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
261
0.898
0.590
0.0696
P value
0.638
0.052
0.667
1.000
1.000
0.248
0.170
0.416
0.604
1.000
1.000
0.282
0.280
0.490
1.000
0.032
0.456
0.083
0.239
0.139
261
0.917
0.566
0.0710
P value
0.113
0.698
0.775
0.886
0.886
0.086
0.863
0.450
0.419
0.568
0.568
0.617
0.636
0.126
0.779
0.568
0.621
0.645
0.696
0.857
Coeff.
-2.653
-4.701
-2.984
6.466
-0.251
-6.033
-0.594
-8.084
-10.08
-13.78
-13.78
-7.451
-5.520
-6.469
-6.284
-13.78
-7.034
-4.886
-2.848
-1.142
261
1.491
0.0850
0.111
P value
0.607
0.402
0.556
0.524
0.970
0.184
0.893
0.071
0.032
0.175
0.175
0.271
0.229
0.185
0.417
0.175
0.230
0.274
0.538
0.806
Coeff.
-0.0889
-0.200
-8.31e-15
-0.200
-0.200
-8.12e-15
-0.117
-0.133
-0.0889
-0.200
-0.200
-0.200
-0.0182
0.108
-0.200
0.800
0.600
-0.103
0.1000
-0.0421
261
1.842
0.0175
0.133
P value
0.652
0.351
1.000
0.606
0.439
1.000
0.490
0.435
0.619
0.606
0.606
0.439
0.917
0.563
0.499
0.040
0.008
0.545
0.572
0.813
Coeff.
-1.156
-0.267
0.1000
-0.600
-0.267
-1.000
-0.294
-0.333
-0.156
-0.600
-0.600
0.400
-0.191
-0.369
-0.100
-0.600
-0.600
-0.342
-0.350
-0.337
261
0.930
0.550
0.0719
P value
0.062
0.691
0.869
0.621
0.741
0.066
0.577
0.533
0.781
0.621
0.621
0.621
0.728
0.526
0.914
0.621
0.391
0.521
0.527
0.545
Coeff.
-1.044
-0.267
0.200
-0.600
0.0667
-0.760
-0.0167
0.0333
0.0111
-0.600
-0.600
0.400
0.0364
-0.754
0.400
0.400
-0.400
0.0129
-0.150
-0.0737
261
0.891
0.599
0.0691
P value
0.146
0.732
0.776
0.670
0.943
0.228
0.978
0.957
0.986
0.670
0.670
0.670
0.954
0.265
0.710
0.776
0.622
0.983
0.815
0.909
Coeff.
-0.956
0.1000
-7.11e-15
-0.400
0.600
-0.880
0.1000
-0.0667
-0.289
-0.400
-0.400
-0.400
-0.127
-0.862
-0.400
-0.400
0.200
-0.271
0.200
0.232
261
0.985
0.481
0.0759
P value
0.215
0.905
1.000
0.791
0.552
0.194
0.879
0.920
0.679
0.791
0.791
0.692
0.852
0.236
0.729
0.791
0.819
0.684
0.772
0.739
Coeff.
-0.911
-0.133
-9.46e-15
0.200
-0.467
-0.640
-0.0778
-0.133
-0.189
-0.800
-0.800
-0.133
0.0182
-0.646
0.200
0.200
-0.200
-0.0581
0.0500
-0.116
261
0.531
0.952
0.0424
P value
0.209
0.865
1.000
0.888
0.623
0.315
0.900
0.832
0.774
0.574
0.574
0.888
0.977
0.345
0.854
0.888
0.808
0.926
0.939
0.859
Exhibit A-1A. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 0-1 Child
(continued)
P/C had too little
P/C had difficulty
money for food at
paying all bills at
time of DHA move- time of DHA movein (1=yes, 0=no)
in (1=yes, 0=no)
Frequency that P/C
drank alcohol since
becoming a parent
Frequency that P/C
smoked marijuana
since becoming a
parent
Number of years
Number of years
P/C ever seen a
during childhood that during childhood that
psychiatrist (1=yes,
P/C lived in public
P/C lived in a home
0=no)
housing
owned by parents
P/C born in the
United States
(1=yes; 0=no)
Spanish language
interview (1=yes;
0=no)
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Home
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
Quigg Newton Homes
Sun Valley Annex
Pacific Place
T Bean Tower (Elderly & Disabled)
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
Thomas Connole (Elderly & Dis.)
East Village
Combined Devel-Disp Housing S.
Combined Devel-Disp Housing E.
Combined Devel-Disp Housing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
-0.756
0.133
0.400
0.800
0.467
-0.720
-0.0889
0.0333
0.0778
-0.200
-0.200
0.133
0.164
-0.508
0.300
-0.200
0.200
0.0903
-1.55e-15
0.274
261
0.713
0.812
0.0561
P value
0.327
0.873
0.597
0.597
0.644
0.288
0.893
0.960
0.911
0.895
0.895
0.895
0.811
0.485
0.795
0.895
0.819
0.892
1.000
0.693
Coeff.
-1.378
-0.267
-1.200
-0.600
-0.267
-0.960
-0.267
-0.133
-0.656
-0.600
-0.600
-0.267
-0.100
-0.908
-0.600
0.400
-0.600
-0.213
-0.250
-0.547
261
0.676
0.848
0.0534
P value
0.128
0.786
0.177
0.735
0.821
0.227
0.730
0.865
0.423
0.735
0.735
0.821
0.901
0.287
0.658
0.821
0.558
0.785
0.757
0.501
Coeff.
0.533
0.0333
-0.300
5.200
-3.467
-0.200
-0.189
-0.333
-0.356
1.200
0.200
1.867
-0.527
0.200
1.700
-0.800
0.400
0.0387
-0.350
0.0421
261
1.117
0.333
0.0851
P value
0.649
0.979
0.794
0.025
0.025
0.846
0.851
0.743
0.738
0.602
0.931
0.224
0.613
0.856
0.334
0.728
0.763
0.970
0.739
0.968
Coeff.
0.889
0.667
0.200
-7.48e-15
0.333
0.280
0.194
-0.300
0.111
1.000
1.000
-5.18e-15
-0.0455
0.769
-6.33e-15
-6.35e-15
0.600
0.194
1.000
0.421
261
0.573
0.929
0.0456
P value
0.356
0.524
0.832
1.000
0.791
0.741
0.813
0.719
0.899
0.597
0.597
1.000
0.958
0.397
1.000
1.000
0.583
0.816
0.247
0.627
Coeff.
0.356
0.300
0.400
-0.200
0.467
8.05e-15
0.0778
0.267
0.0222
-0.200
0.800
0.467
0.209
0.108
-0.200
-0.200
0.200
-0.00645
0.150
0.168
261
1.169
0.283
0.0888
P value
0.177
0.294
0.122
0.698
0.176
1.000
0.729
0.242
0.926
0.698
0.122
0.176
0.371
0.664
0.612
0.698
0.502
0.977
0.524
0.477
Coeff.
-0.689
10.37
-0.500
-5.800
-5.800
1.120
-1.244
-0.200
4.700
-5.800
-5.800
3.200
0.336
-2.492
-0.300
-5.800
-2.000
0.458
-3.200
-3.905
261
1.315
0.170
0.0987
P value
0.887
0.050
0.916
0.543
0.362
0.793
0.764
0.962
0.286
0.543
0.543
0.615
0.938
0.586
0.967
0.543
0.716
0.913
0.462
0.372
Coeff.
-6.889
-20.00
-9.300
7.000
-4.667
-14.96
-7.139
-9.900
-8.611
-20.00
-8.000
-7.000
-9.864
-6.077
4.500
7.000
-10.40
-5.968
-6.050
-7.053
261
1.436
0.106
0.107
P value
0.287
0.005
0.144
0.581
0.581
0.009
0.197
0.078
0.142
0.116
0.528
0.408
0.087
0.319
0.642
0.581
0.157
0.286
0.297
0.226
Coeff.
0.0889
0.200
0.100
0.200
-0.133
0.120
0.0889
0.133
0.144
0.200
0.200
0.200
0.109
0.0462
0.200
-0.800
-0.200
0.135
0.100
0.200
261
1.179
0.273
0.0895
P value
0.580
0.252
0.526
0.526
0.526
0.395
0.518
0.338
0.321
0.526
0.526
0.342
0.445
0.761
0.407
0.012
0.273
0.329
0.487
0.168
Coeff.
1.58e-15
1.84e-15
0.100
1.51e-15
1.75e-15
0.0800
0.0278
0.0667
1.56e-15
1.37e-15
1.51e-15
1.49e-15
0.0909
1.56e-15
1.47e-15
1.47e-15
0.200
1.25e-15
0.0500
0.0526
261
0.525
0.954
0.0419
P value
1.000
1.000
0.374
1.000
1.000
0.427
0.777
0.502
1.000
1.000
1.000
1.000
0.372
1.000
1.000
1.000
0.124
1.000
0.626
0.610
Exhibit A-1A. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 0-1 Child
(continued)
Biological father
always lived in
household with
child(ren) (1=yes;
0=no)
Parent have HS
Parent have any
Parent's age at time P/C African American diploma at time of higher education at
of DHA move-in
(1=yes; 0=no)
DHA move-in (1=yes; time of DHA move0=no)
in (1=yes; 0=no)
Kids share same
biological dad
(1=yes; 0=no)
Parent Depressive
Symptomatology
Scale at time of
interview
Parenting Efficacy
Scale at time of
interview
Parenting Beliefs
Scale at time of
interview
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Home
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
Quigg Newton Homes
Sun Valley Annex
Pacific Place
T Bean Tower (Elderly & Disabled)
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
Thomas Connole (Elderly & Dis.)
East Village
Combined Devel-Disp Housing S.
Combined Devel-Disp Housing E.
Combined Devel-Disp Housing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
-0.0889
0.133
-0.100
-0.200
-0.200
0.0800
-0.0611
-0.167
-0.0889
-0.200
-0.200
-0.200
-0.109
-0.0462
-0.200
-0.200
-3.54e-15
-0.0387
-3.49e-15
0.0105
261
0.619
0.897
0.0491
P value
0.657
0.540
0.611
0.611
0.446
0.649
0.721
0.337
0.624
0.611
0.611
0.446
0.540
0.807
0.506
0.611
1.000
0.823
1.000
0.953
Coeff.
5.756
6.200
0.900
11.20
11.87
5.680
4.561
4.300
0.311
30.20
17.20
12.20
3.609
8.123
5.200
29.20
15.00
1.910
7.850
9.095
261
1.677
0.0378
0.123
P value
0.338
0.342
0.879
0.343
0.132
0.282
0.375
0.408
0.954
0.011
0.145
0.122
0.499
0.152
0.564
0.014
0.028
0.713
0.146
0.094
Coeff.
0.289
-0.433
0.100
-0.600
0.0667
-0.240
-0.0722
-0.267
-0.156
0.400
0.400
0.400
-0.327
-0.138
0.400
-0.600
3.03e-15
-0.342
0.250
-0.337
261
2.800
0.000108
0.189
P value
0.269
0.127
0.697
0.243
0.845
0.296
0.747
0.239
0.511
0.436
0.436
0.243
0.159
0.574
0.308
0.243
1.000
0.131
0.286
0.153
Coeff.
0.133
-0.0333
0.100
-0.200
-0.200
0.360
0.217
0.233
0.0778
-0.200
0.800
-0.200
0.0727
-0.123
0.800
-0.200
6.26e-15
0.123
0.100
0.116
261
1.260
0.207
0.0950
P value
0.611
0.907
0.697
0.697
0.560
0.118
0.334
0.304
0.743
0.697
0.121
0.560
0.755
0.618
0.043
0.697
1.000
0.588
0.670
0.624
Coeff.
-0.200
-0.200
-0.100
-0.200
0.133
-0.120
-0.0889
-0.167
-0.144
-0.200
-0.200
0.133
-0.0182
0.185
-0.200
-0.200
-4.03e-15
-0.135
0.0500
-0.0947
261
1.165
0.286
0.0885
P value
0.267
0.306
0.571
0.571
0.571
0.448
0.564
0.285
0.376
0.571
0.571
0.571
0.909
0.277
0.459
0.571
1.000
0.384
0.756
0.559
Coeff.
-0.0444
0.0667
-0.300
0.400
0.0667
0.0800
0.178
0.133
-0.0444
0.400
0.400
0.0667
0.0364
0.0154
0.400
0.400
0.200
0.110
0.150
0.189
261
0.787
0.729
0.0615
P value
0.865
0.814
0.242
0.435
0.845
0.727
0.426
0.555
0.851
0.435
0.435
0.845
0.875
0.950
0.307
0.435
0.499
0.627
0.522
0.421
Coeff.
-0.0667
-6.233
5.100
-8.400
-3.400
1.600
-1.844
2.033
-1.067
8.600
8.600
3.933
-1.309
-0.785
-6.900
10.60
6.400
-1.787
-2.850
-1.453
261
0.955
0.518
0.0737
P value
0.990
0.295
0.343
0.435
0.635
0.739
0.694
0.668
0.830
0.424
0.424
0.583
0.788
0.879
0.401
0.325
0.303
0.706
0.562
0.768
Coeff.
-0.0889
0.467
-0.300
-4.200
0.467
-1.040
-1.256
-1.667
-1.811
-3.200
1.800
-2.200
-1.291
0.0308
1.800
-11.20
-3.600
-1.232
-2.150
-1.095
261
1.209
0.247
0.0915
P value
0.962
0.819
0.871
0.257
0.850
0.530
0.436
0.307
0.289
0.387
0.627
0.373
0.441
0.986
0.524
0.003
0.093
0.449
0.204
0.519
Coeff.
-1.911
0.367
-1.100
3.200
-3.133
-0.920
-0.828
-1.400
-0.578
-0.800
2.200
0.533
0.473
-0.415
0.700
-7.800
-2.600
-0.574
-1.750
-0.221
261
0.865
0.632
0.0672
P value
0.332
0.864
0.570
0.409
0.225
0.595
0.623
0.412
0.746
0.836
0.570
0.836
0.787
0.823
0.813
0.045
0.245
0.736
0.322
0.901
Exhibit A-1B. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 2 Children
P/C is single parent
(1=yes, 0=no)
P/C employment
P/C receiving
status at time of DHA
P/C disability status at P/C received TANF at
P/C hourly wage at
Food Stamps at
move-in
time of survey (1=yes; time of DHA move-in
time of DHA move-in
time of DHA move(1=employed, 0=not
0=no)
(1=yes, 0=no)
in (1=yes, 0=no)
employed)
P/C had checking
account at time of
DHA move-in
(1=yes, 0=no)
P/C had health
insurance at time
of DHA move-in
(1=yes, 0=no)
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Homes
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
220
Quigg Newton Homes
Sun Valley Annex
Pacific Place
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
East Village
Combined Devel-Disp Hsing S.
Combined Devel-Disp Hsing E.
Combined Devel-Disp Hsing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
0.333
0.333
0.0476
0.333
0.333
0.175
-0.0370
0.333
0.222
0.0833
0.333
-3.83e-15
0.0333
0.0333
0.333
0.333
0.0769
0.194
0.121
244
0.699
0.818
0.0560
P value
0.383
0.298
0.858
0.491
0.491
0.500
0.884
0.491
0.395
0.758
0.298
1.000
0.904
0.904
0.383
0.383
0.759
0.440
0.631
Coeff.
0.667
0.417
-0.476
0.667
-0.333
-0.386
0.444
-0.333
0.111
0.333
-1.833
1.16e-14
-0.133
0.167
-4.333
0.667
0.0256
0.500
0.212
244
2.087
0.00616
0.150
P value
0.600
0.695
0.591
0.678
0.836
0.655
0.600
0.836
0.898
0.710
0.085
1.000
0.884
0.856
0.001
0.600
0.975
0.550
0.800
Coeff.
9.167
4.542
0.792
17.17
-7.333
0.227
8.220
-7.333
0.146
3.164
5.042
-2.233
-3.933
0.552
-0.583
8.167
3.107
8.579
1.736
244
2.329
0.00184
0.165
P value
0.266
0.509
0.890
0.100
0.481
0.968
0.135
0.481
0.979
0.586
0.464
0.726
0.507
0.926
0.943
0.321
0.565
0.114
0.749
Coeff.
3.61e-15
0.500
0.143
4.17e-15
4.18e-15
0.0526
0.0741
4.58e-15
0.167
4.70e-15
4.33e-15
0.500
0.100
0.100
0.500
4.62e-15
0.128
0.139
0.0909
244
1.193
0.265
0.0919
P value
1.000
0.043
0.486
1.000
1.000
0.793
0.706
1.000
0.407
1.000
1.000
0.029
0.637
0.637
0.090
1.000
0.507
0.473
0.640
Coeff.
0.167
0.417
-0.476
-0.333
-0.333
0.140
0.0370
-0.333
0.167
0.250
0.167
-1.333
0.267
0.367
0.667
-0.333
-0.282
-2.55e-16
-0.0606
244
0.899
0.584
0.0709
P value
0.874
0.637
0.517
0.803
0.803
0.845
0.958
0.803
0.817
0.737
0.850
0.104
0.726
0.630
0.527
0.752
0.684
1.000
0.931
Coeff.
0.167
0.417
0.452
-0.333
-0.333
0.404
0.333
-0.333
0.333
0.417
0.417
0.500
0.267
0.467
0.167
-0.333
-0.128
0.0556
0.242
244
0.578
0.920
0.0467
P value
0.857
0.590
0.483
0.776
0.776
0.522
0.589
0.776
0.598
0.524
0.590
0.485
0.689
0.484
0.857
0.718
0.833
0.927
0.691
Coeff.
0.500
2.73e-14
-0.286
1.000
2.82e-14
-0.158
0.630
2.87e-14
0.222
0.333
-2.000
0.167
0.300
0.200
-4.500
0.500
0.154
0.556
0.333
244
2.352
0.00164
0.166
P value
0.685
1.000
0.739
0.521
1.000
0.851
0.443
1.000
0.792
0.702
0.053
0.861
0.736
0.822
0.000
0.685
0.849
0.493
0.682
Coeff.
2.46e-14
-0.250
-0.714
2.38e-14
2.38e-14
-0.842
-0.333
-1.000
-0.278
-0.333
-2.500
-0.500
-0.400
-0.1000
-5.000
-0.500
-0.410
-0.278
-0.242
244
1.840
0.0200
0.135
P value
1.000
0.816
0.425
1.000
1.000
0.335
0.697
0.538
0.751
0.713
0.021
0.615
0.666
0.914
0.000
0.697
0.626
0.742
0.775
Exhibit A-1B. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 2 Children
(continued)
P/C had too little
P/C had difficulty
Frequency that P/C
money for food at paying all bills at time
drank alcohol since
time of DHA move-in
of DHA move-in
becoming a parent
(1=yes, 0=no)
(1=yes, 0=no)
Frequency that P/C
smoked marijuana
since becoming a
parent
P/C ever seen a
psychiatrist (1=yes,
0=no)
Number of years
during childhood
that P/C lived in
public housing
Number of years
during childhood that
P/C lived in a home
owned by parents
P/C born in the
United States
(1=yes; 0=no)
Spanish language
interview (1=yes;
0=no)
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Homes
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
220
Quigg Newton Homes
Sun Valley Annex
Pacific Place
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
East Village
Combined Devel-Disp Hsing S.
Combined Devel-Disp Hsing E.
Combined Devel-Disp Hsing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
-0.167
-0.667
-1.024
0.333
-0.667
-0.772
-0.259
0.333
-0.556
-0.333
-2.417
-0.333
-0.0667
-0.567
-5.167
-0.167
-0.436
-0.194
-0.303
244
2.138
0.00481
0.153
P value
0.893
0.521
0.237
0.832
0.671
0.361
0.754
0.832
0.512
0.704
0.021
0.729
0.941
0.527
0.000
0.893
0.593
0.812
0.712
Coeff.
3.167
2.667
2.381
2.667
2.667
2.246
2.815
3.667
1.833
3.083
0.917
2.833
3.267
2.767
-1.833
2.667
2.974
3.167
2.879
244
1.514
0.0821
0.114
P value
0.082
0.080
0.061
0.246
0.246
0.070
0.021
0.111
0.140
0.017
0.546
0.045
0.013
0.035
0.312
0.142
0.013
0.008
0.017
Coeff.
-0.500
1.000
0.571
-1.000
-7.26e-14
-0.474
-0.481
-1.000
-0.722
-0.667
0.750
-0.833
-1.000
-1.500
0.500
-1.000
-1.051
-0.528
-1.061
244
0.447
0.979
0.0365
P value
0.846
0.642
0.749
0.758
1.000
0.786
0.778
0.758
0.680
0.713
0.727
0.675
0.589
0.418
0.846
0.697
0.533
0.755
0.532
Coeff.
-0.500
2.500
0.429
5.000
-1.82e-15
-0.842
-0.778
-1.000
-0.667
-0.667
-0.750
0.333
-1.100
-0.400
-0.500
-0.500
-0.564
-0.778
-0.394
244
1.044
0.411
0.0814
P value
0.804
0.138
0.760
0.050
1.000
0.539
0.562
0.694
0.628
0.639
0.656
0.831
0.448
0.783
0.804
0.804
0.669
0.557
0.767
Coeff.
0.167
0.667
0.0238
0.667
0.667
0.0351
0.222
-0.333
0.222
-0.167
-0.0833
-4.96e-15
0.0667
0.0667
0.167
-0.333
0.0513
0.222
-0.0303
244
1.226
0.238
0.0942
P value
0.710
0.077
0.939
0.241
0.241
0.909
0.458
0.557
0.469
0.600
0.824
1.000
0.837
0.837
0.710
0.458
0.862
0.452
0.919
Coeff.
1.34e-13
11.25
6.000
1.27e-13
16.00
6.579
2.630
1.22e-13
2.833
6.333
6.000
2.833
3.800
5.800
5.500
1.21e-13
5.077
1.944
4.394
244
0.952
0.520
0.0747
P value
1.000
0.057
0.222
1.000
0.073
0.171
0.575
1.000
0.556
0.204
0.309
0.604
0.455
0.254
0.435
1.000
0.273
0.675
0.345
Coeff.
17.83
-1.917
4.262
10.33
-8.667
-2.193
5.296
10.33
4.556
2.667
-1.917
-0.833
7.933
4.033
-8.667
3.833
3.513
3.361
0.758
244
0.879
0.609
0.0694
P value
0.106
0.835
0.578
0.458
0.533
0.769
0.470
0.458
0.544
0.732
0.835
0.922
0.317
0.611
0.431
0.727
0.626
0.642
0.917
Coeff.
7.02e-15
6.93e-15
-0.143
6.36e-15
6.37e-15
6.04e-15
-0.259
5.98e-15
-0.222
-0.167
-0.250
5.69e-15
-0.1000
-0.300
5.97e-15
-0.500
-0.0769
-0.139
-0.182
244
0.874
0.616
0.0690
P value
1.000
1.000
0.525
1.000
1.000
1.000
0.229
1.000
0.314
0.465
0.355
1.000
0.667
0.198
1.000
0.122
0.716
0.513
0.394
Coeff.
4.22e-15
4.24e-15
0.0714
4.46e-15
4.45e-15
4.67e-15
0.0370
4.67e-15
0.167
0.0833
0.250
4.77e-15
0.100
4.69e-15
4.68e-15
4.69e-15
0.0256
0.0556
0.152
244
0.744
0.771
0.0593
P value
1.000
1.000
0.655
1.000
1.000
1.000
0.808
1.000
0.287
0.607
0.193
1.000
0.545
1.000
1.000
1.000
0.865
0.713
0.317
11
Exhibit A-1B. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 2 Children
(continued)
Biological father
always lived in
household with
child(ren) (1=yes;
0=no)
Parent's age at time
of DHA move-in
P/C African
American (1=yes;
0=no)
Parent have HS
Parent have any
diploma at time of
higher education at
DHA move-in (1=yes; time of DHA move-in
0=no)
(1=yes; 0=no)
Kids share same
biological dad
(1=yes; 0=no)
Parent Depressive
Symptomatology
Scale at time of
interview
Parenting Efficacy Parenting Beliefs
Scale at time of
Scale at time of
interview
interview
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Homes
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
220
Quigg Newton Homes
Sun Valley Annex
Pacific Place
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
East Village
Combined Devel-Disp Hsing S.
Combined Devel-Disp Hsing E.
Combined Devel-Disp Hsing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
1.15e-14
1.14e-14
0.286
1.10e-14
1.10e-14
0.158
0.148
1.07e-14
1.01e-14
0.167
1.14e-14
0.167
0.100
0.200
1.06e-14
1.06e-14
0.205
0.167
0.182
244
0.522
0.951
0.0424
P value
1.000
1.000
0.222
1.000
1.000
0.489
0.507
1.000
1.000
0.482
1.000
0.521
0.679
0.408
1.000
1.000
0.351
0.450
0.411
Coeff.
-4.833
-7.583
-9.619
-6.333
-11.33
-10.07
-7.222
-3.333
-8.111
-10.42
-5.583
-4.833
-10.33
-7.733
-8.333
4.167
-4.333
-5.076
-4.606
243
1.044
0.411
0.0817
P value
0.554
0.267
0.092
0.540
0.273
0.071
0.185
0.747
0.147
0.072
0.414
0.445
0.080
0.190
0.308
0.610
0.419
0.346
0.393
Coeff.
-0.500
-0.500
-0.214
5.98e-15
5.99e-15
-0.526
-0.370
-1.000
-0.722
-0.750
-1.000
3.89e-15
-0.800
-0.400
5.07e-15
-0.500
-0.795
-0.111
-0.970
244
7.950
1.31e-16
0.403
P value
0.174
0.105
0.403
1.000
1.000
0.036
0.131
0.032
0.004
0.004
0.001
1.000
0.003
0.132
1.000
0.174
0.001
0.646
0.000
Coeff.
1.000
0.250
0.500
1.70e-14
1.70e-14
0.316
0.111
1.000
0.167
0.167
0.250
0.167
0.500
0.400
1.63e-14
1.61e-14
0.385
0.361
0.212
244
1.481
0.0938
0.112
P value
0.015
0.465
0.080
1.000
1.000
0.257
0.683
0.054
0.550
0.564
0.465
0.598
0.091
0.175
1.000
1.000
0.152
0.180
0.432
Coeff.
-0.333
-0.0833
-0.262
-0.333
-0.333
-0.281
-0.185
-0.333
-0.278
-0.250
-0.333
-0.167
-0.133
-0.333
-0.333
-0.333
-0.256
-0.167
-0.212
244
0.503
0.960
0.0409
P value
0.248
0.730
0.193
0.361
0.361
0.153
0.336
0.361
0.159
0.221
0.168
0.456
0.521
0.110
0.248
0.248
0.176
0.380
0.266
Coeff.
1.000
0.250
0.286
1.000
1.33e-14
0.316
0.407
1.000
0.333
0.417
0.250
0.167
0.400
0.200
1.33e-14
1.000
0.385
0.472
0.424
244
1.014
0.446
0.0792
P value
0.025
0.502
0.357
0.076
1.000
0.297
0.170
0.076
0.273
0.186
0.502
0.628
0.213
0.533
1.000
0.025
0.188
0.108
0.149
Coeff.
12.17
12.92
3.881
-4.333
13.67
2.140
6.815
-0.333
4.000
6.583
-3.083
2.500
9.267
3.167
2.667
-1.333
3.923
3.306
4.333
244
0.980
0.486
0.0767
P value
0.151
0.069
0.510
0.685
0.202
0.710
0.227
0.975
0.489
0.271
0.663
0.703
0.129
0.604
0.752
0.875
0.480
0.553
0.438
Coeff.
0.333
-5.917
-2.952
-0.667
0.333
-2.246
-3.481
-5.667
-1.833
-2.167
-1.167
-1.167
-4.067
-2.367
-5.667
-3.167
-3.077
-3.528
-2.061
244
0.954
0.517
0.0749
P value
0.921
0.037
0.209
0.876
0.938
0.328
0.122
0.184
0.426
0.363
0.679
0.655
0.095
0.330
0.093
0.347
0.165
0.112
0.355
Coeff.
-7.833
-1.333
-2.619
-4.333
-1.333
-1.386
-0.889
0.667
-1.333
-0.500
-2.583
-2.500
0.867
-1.433
-3.833
-0.833
-1.590
-3.111
-0.667
244
1.135
0.318
0.0878
P value
0.030
0.658
0.297
0.342
0.770
0.572
0.711
0.884
0.588
0.844
0.392
0.371
0.739
0.581
0.288
0.817
0.501
0.190
0.779
Exhibit A-1C. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 3+ Children
P/C is single
parent (1=yes,
0=no)
P/C employment
status at time of DHA
P/C disability status
P/C hourly wage at
move-in
at time of survey
time of DHA move-in
(1=employed, 0=not
(1=yes; 0=no)
employed)
P/C received
TANF at time of
DHA move-in
(1=yes, 0=no)
P/C receiving
P/C had checking
Food Stamps at account at time of
time of DHA moveDHA move-in
in (1=yes, 0=no)
(1=yes, 0=no)
P/C had health
insurance at time
of DHA move-in
(1=yes, 0=no)
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Homes
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
Quigg Newton Homes
Sun Valley Annex
Pacific Place
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
East Village
Combined Devel-Disp Hsing S.
Combined Devel-Disp Hsing E.
Combined Devel-Disp Hsing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
0.250
0.250
0.0682
0.250
0.250
5.54e-15
-0.150
-0.0658
-0.125
0.250
-0.250
-0.114
0.107
-0.0833
0.250
-0.132
0.00758
-0.00926
203
0.575
0.914
0.0533
P value
0.627
0.386
0.800
0.627
0.627
1.000
0.562
0.795
0.657
0.530
0.530
0.672
0.710
0.812
0.627
0.586
0.975
0.970
Coeff.
1.000
0.429
0.455
1.000
1.000
0.437
0.200
0.474
0.625
0.500
-4.500
0.727
-1.000
0.333
-2.31e-14
0.647
0.576
0.630
203
2.884
0.000169
0.220
P value
0.443
0.558
0.505
0.443
0.443
0.502
0.760
0.460
0.382
0.621
0.000
0.286
0.172
0.708
1.000
0.294
0.351
0.314
Coeff.
18.75
7.407
7.409
22.50
20.00
7.734
15.88
8.752
12.09
9.500
-1.30e-13
12.40
4.819
5.383
-4.42e-14
12.27
10.54
11.64
203
1.479
0.101
0.126
P value
0.077
0.211
0.180
0.034
0.059
0.144
0.003
0.093
0.038
0.246
1.000
0.025
0.416
0.455
1.000
0.015
0.036
0.022
Coeff.
-3.82e-15
-3.45e-15
0.0909
-3.79e-15
-3.79e-15
0.125
0.0667
0.158
-3.53e-15
-3.58e-15
-3.59e-15
0.182
0.143
0.333
-3.57e-15
0.0882
0.121
0.0370
203
0.452
0.974
0.0424
P value
1.000
1.000
0.604
1.000
1.000
0.456
0.693
0.339
1.000
1.000
1.000
0.300
0.447
0.147
1.000
0.578
0.445
0.818
Coeff.
-0.750
-0.179
-1.295
-0.750
-0.750
-0.125
-1.617
-0.329
-1.375
-0.750
-0.250
-0.205
-0.464
-0.417
-0.750
-0.279
-0.356
-0.231
203
1.202
0.263
0.105
P value
0.631
0.838
0.113
0.631
0.631
0.873
0.041
0.668
0.109
0.535
0.836
0.802
0.596
0.696
0.631
0.705
0.630
0.757
Coeff. P value
0.250
0.833
-0.0357 0.957
-0.205
0.741
0.250
0.833
0.250
0.833
0.188
0.752
-1.283
0.032
-0.171
0.769
-7.89e-16 1.000
-0.250
0.785
0.250
0.785
0.0682 0.912
-0.179
0.788
-0.0833 0.918
0.250
0.833
-0.103
0.854
-0.235
0.675
-0.0463 0.935
Coeff.
4.70e-15
0.143
-0.727
1.000
1.000
0.188
4.85e-15
0.105
0.625
0.500
4.49e-15
0.182
-1.286
0.333
1.000
0.676
0.455
0.370
203
1.169
0.291
0.103
203
1.525
0.0852
0.130
P value
1.000
0.850
0.303
0.459
0.459
0.781
1.000
0.874
0.398
0.632
1.000
0.796
0.090
0.718
0.459
0.290
0.477
0.567
Coeff.
0.750
0.464
0.386
-0.250
0.750
0.500
-0.250
0.539
0.375
-0.250
-0.250
0.477
-1.107
0.0833
0.750
0.544
0.417
0.528
203
1.359
0.157
0.117
P value
0.523
0.481
0.528
0.831
0.523
0.394
0.672
0.351
0.560
0.783
0.783
0.436
0.094
0.917
0.523
0.327
0.454
0.348
Exhibit A-1C. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 3+ Children
(continued)
P/C had too little
P/C had difficulty
Frequency that P/C
money for food at paying all bills at time
drank alcohol since
time of DHA moveof DHA move-in
becoming a parent
in (1=yes, 0=no)
(1=yes, 0=no)
Frequency that P/C
smoked marijuana
since becoming a
parent
P/C ever seen a
psychiatrist
(1=yes, 0=no)
Number of years
during childhood
that P/C lived in
public housing
Number of years
during childhood
that P/C lived in a
home owned by
parents
P/C born in the
United States
(1=yes; 0=no)
Spanish language
interview (1=yes;
0=no)
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Homes
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
Quigg Newton Homes
Sun Valley Annex
Pacific Place
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
East Village
Combined Devel-Disp Hsing S.
Combined Devel-Disp Hsing E.
Combined Devel-Disp Hsing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
-0.250
0.464
0.477
0.750
0.750
-0.0625
-0.583
0.118
0.625
0.250
0.250
0.295
-1.393
0.0833
-0.250
0.0735
0.0530
0.157
203
1.534
0.0822
0.131
P value
0.828
0.471
0.427
0.514
0.514
0.913
0.314
0.834
0.321
0.779
0.779
0.622
0.032
0.915
0.828
0.892
0.922
0.775
Coeff.
-0.250
0.179
-0.795
0.750
0.750
-0.562
-0.383
-0.461
0.375
0.250
-0.250
0.386
-1.393
-2.917
0.750
-0.515
-0.371
0.269
203
0.721
0.787
0.0659
P value
0.916
0.893
0.522
0.752
0.752
0.636
0.749
0.694
0.773
0.892
0.892
0.756
0.296
0.074
0.752
0.647
0.742
0.814
Coeff.
1.92e-14
1.286
0.0909
1.93e-14
1.000
-0.250
0.667
0.263
0.625
2.500
2.03e-14
-0.545
1.857
2.333
2.000
-0.382
-0.0909
0.963
203
0.852
0.637
0.0769
P value
1.000
0.420
0.951
1.000
0.725
0.860
0.641
0.851
0.688
0.257
1.000
0.713
0.244
0.230
0.482
0.776
0.946
0.480
Coeff.
-6.57e-15
-0.429
-0.364
-6.47e-15
1.000
-5.91e-15
-0.467
0.895
0.375
3.000
-6.14e-15
-0.727
1.286
-2.667
-6.07e-15
-0.941
0.455
0.630
203
1.086
0.369
0.0960
P value
1.000
0.789
0.808
1.000
0.727
1.000
0.746
0.525
0.811
0.177
1.000
0.626
0.423
0.173
1.000
0.487
0.737
0.646
Coeff.
-0.250
0.179
0.114
-0.250
0.750
0.0625
0.417
0.329
0.125
0.250
-0.250
-0.0682
0.179
0.417
-0.250
0.103
0.205
0.0463
203
0.990
0.473
0.0883
P value
0.649
0.563
0.692
0.649
0.174
0.820
0.133
0.225
0.678
0.557
0.557
0.812
0.563
0.268
0.649
0.692
0.432
0.861
Coeff.
-13.50
-6.071
-5.955
-13.50
-13.50
-9.375
-8.700
-8.342
-10.12
-6.000
-13.50
-8.682
-7.786
-7.167
12.50
-10.76
-11.50
-7.093
203
1.279
0.206
0.111
P value
0.139
0.235
0.211
0.139
0.139
0.040
0.059
0.064
0.043
0.395
0.057
0.069
0.128
0.250
0.170
0.013
0.008
0.105
Coeff.
8.000
1.286
6.545
22.00
4.000
7.750
5.867
5.684
7.250
8.500
4.000
12.73
8.143
16.67
22.00
10.18
10.58
5.407
203
0.879
0.604
0.0792
P value
0.549
0.863
0.348
0.100
0.764
0.246
0.382
0.387
0.321
0.411
0.699
0.069
0.277
0.068
0.100
0.108
0.095
0.398
Coeff.
3.25e-15
-0.143
3.52e-15
3.21e-15
3.18e-15
-0.187
-0.133
-0.263
2.85e-15
2.86e-15
2.82e-15
-0.273
-0.143
2.92e-15
2.82e-15
-0.206
-0.152
-0.185
203
0.530
0.941
0.0493
P value
1.000
0.542
1.000
1.000
1.000
0.370
0.526
0.202
1.000
1.000
1.000
0.212
0.542
1.000
1.000
0.298
0.444
0.356
Coeff. P value
-2.92e-16 1.000
0.143
0.375
-5.00e-16 1.000
-3.69e-16 1.000
-3.81e-16 1.000
0.187
0.192
-6.40e-16 1.000
0.211
0.137
-6.93e-16 1.000
-7.06e-16 1.000
-7.14e-16 1.000
0.0909 0.544
-6.44e-16 1.000
-9.54e-16 1.000
-6.52e-16 1.000
0.0294 0.828
0.0606 0.655
0.0741 0.590
203
0.823
0.671
0.0745
ExhibitA-1C. Relationships Between DHA Resident Characteristics and DHA Developments: Households with 3+ Children
(continued)
Biological father
always lived in
Parent's age at time
household with
of DHA move-in
child(ren) (1=yes;
0=no)
Parent have HS
P/C African
diploma at time of
American (1=yes;
DHA move-in (1=yes;
0=no)
0=no)
Parent have any
Kids share same
higher education at
biological dad
time of DHA move(1=yes; 0=no)
in (1=yes; 0=no)
Parent Depressive
Parenting Efficacy Parenting Beliefs
Symptomatology
Scale at time of
Scale at time of
Scale at time of
interview
interview
interview
DHA Development
Arrowhead Townhouses
Columbine Homes
Curtis Park Homes
FHA Repossessed East
Goldsmith Village
South Lincoln
North Lincoln COL
Quigg Newton Homes
Sun Valley Annex
Pacific Place
Platte Valley Homes
Westridge Homes
Westwood Homes
Stapleton Homes
East Village
Combined Devel-Disp Hsing S.
Combined Devel-Disp Hsing E.
Combined Devel-Disp Hsing W.
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent or Caregiver;
reference group = Arapaho Cts.
bold = p <.05
Coeff.
-0.250
-0.107
-0.0682
-0.250
0.750
-0.188
0.150
-0.145
0.250
0.250
-0.250
-0.0682
-0.107
0.417
-0.250
0.162
-0.00758
0.120
203
1.349
0.162
0.117
P value
0.613
0.699
0.792
0.613
0.130
0.448
0.547
0.552
0.356
0.514
0.514
0.792
0.699
0.218
0.613
0.489
0.974
0.611
Coeff.
3.250
12.25
2.795
4.250
7.250
6.687
4.250
6.566
8.125
13.25
3.750
9.068
1.393
13.58
11.25
9.515
7.098
8.139
203
1.183
0.279
0.104
P value
0.729
0.021
0.568
0.650
0.439
0.154
0.368
0.155
0.114
0.069
0.605
0.065
0.791
0.035
0.230
0.033
0.111
0.071
Coeff.
-1.28e-14
-0.571
-9.76e-15
-1.16e-14
-1.15e-14
-0.500
-0.467
-0.895
-0.375
-1.000
-9.39e-15
-0.545
-0.857
-1.02e-14
-1.39e-14
-0.647
-0.303
-0.963
203
5.652
1.47e-10
0.356
P value
1.000
0.031
1.000
1.000
1.000
0.034
0.050
0.000
0.146
0.007
1.000
0.027
0.001
1.000
1.000
0.004
0.174
0.000
Coeff.
-2.63e-15
0.571
0.455
-2.46e-15
-2.54e-15
0.375
0.200
0.316
0.375
-1.50e-15
0.500
0.273
0.143
0.333
1.000
0.412
0.394
0.370
203
0.705
0.803
0.0645
P value
1.000
0.061
0.110
1.000
1.000
0.168
0.464
0.238
0.208
1.000
0.235
0.336
0.639
0.369
0.066
0.110
0.126
0.155
Coeff. P value
-5.40e-15 1.000
-5.08e-15 1.000
-5.63e-15 1.000
-5.38e-15 1.000
-5.34e-15 1.000
0.125
0.417
0.0667 0.667
0.0526 0.728
-5.19e-15 1.000
-5.19e-15 1.000
-5.19e-15 1.000
0.182
0.258
-5.13e-15 1.000
-4.91e-15 1.000
-5.22e-15 1.000
0.0294 0.840
0.152
0.299
0.148
0.315
Coeff. P value
-6.77e-15 1.000
0.143
0.625
0.273
0.317
-6.64e-15 1.000
1.000
0.056
0.375
0.151
0.400
0.129
0.316
0.219
0.125
0.662
-6.01e-15 1.000
-6.04e-15 1.000
0.273
0.317
0.429
0.144
-6.34e-15 1.000
-6.05e-15 1.000
0.529* 0.033
0.303
0.221
0.259
0.300
Coeff.
-3.250
8.321
7.659
13.75
20.75
2.687
-1.383
2.803
2.625
3.250
2.250
-0.886
-0.250
8.417
-3.250
1.956
-0.0682
0.194
203
0.627
0.876
0.0578
203
1.101
0.354
0.0972
203
1.111
0.344
0.0981
P value
0.763
0.170
0.175
0.204
0.056
0.619
0.799
0.598
0.657
0.698
0.788
0.875
0.967
0.255
0.763
0.702
0.989
0.970
Coeff.
3.500
-1.643
0.500
-4.500
-5.500
0.688
1.033
0.974
0.250
1.46e-14
0.500
0.318
1.786
-0.167
2.500
0.206
0.773
0.352
203
0.715
0.793
0.0654
P value
0.337
0.422
0.793
0.218
0.132
0.706
0.573
0.587
0.900
1.000
0.859
0.867
0.383
0.947
0.493
0.905
0.654
0.840
Coeff.
0.750
-1.679
-2.795
-9.250
-2.250
-1.438
-1.983
-1.197
-0.250
0.250
-4.750
-2.068
-3.679
-2.250
0.750
-0.779
-1.311
-2.028
203
0.740
0.767
0.0675
P value
0.865
0.498
0.227
0.037
0.611
0.515
0.373
0.582
0.918
0.942
0.166
0.371
0.139
0.456
0.865
0.709
0.531
0.339
15
Results of these tests using DHA development dummies are presented in Exhibit A-1(AC), those using census tract dummies in Exhibit A-2(A-C). Exhibit A-1shows that there
were few statistically significant differences in individual characteristics across the
various DHA developments: of 1,482 coefficients across all family size strata only 72
(5%) were so.34 A similar aggregate portrait emerges from Exhibit A-2: of 3,640
coefficients across all family size strata only 202 (5.5%) were significant.35 Examination
of individual characteristics reveals, however, that African Americans with two or more
children were not proportionally distributed across all DHA developments or census
tracts where such developments were located.36 We cannot be sure whether any
systematic actions by the DHA and/or African American applicants to DHA produced this
result, but the outcome was clearly inconsistent with quasi-random assignment across
developments or neighborhoods. The second notable revelation was that DHA residents
with disabilities (most of whom had two or fewer children) were also allocated nonrandomly to a relatively few developments, producing a distinct profile for their census
tract characteristics. This is not surprising, inasmuch as certain DHA developments are
designed especially for elderly and disabled residents and other, scattered-site
developments are rendered off-limits to the disabled because of expectations of tenant
contributions to dwelling and grounds maintenance. Conditioning on ethnic and
disability status, however, we think this evidence offers a compelling case that DHA
allocations were quasi-random across developments and neighborhoods because only
three (3) percent of the remaining coefficients proved statistically significant in both
Exhibits A-1and A-2 there was no pattern to these coefficients. This percentage could
have been generated by chance even if true random assignment had been undertaken.
34
The percentages across the 0-1 child, 2 children, 3+ children strata were 3%, 6%, 6%, respectively.
The percentages across the 0-1 child, 2 children, 3+ children strata were 4%, 6%, 6%, respectively.
36
Seventeen of the 37 DHA site coefficients were significantly different from zero for the African
American characteristic combining both family size strata; the corresponding figure for the 97 tract
coefficients was 44.
35
Exhibit A-2A. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 0-1 Child
P/C is single parent
(1=yes, 0=no)
Census Tract
2.0100
2.0200
3.0100
3.0200
5.0200
7.0100
7.0200
8.0000
9.0200
9.0300
9.0400
10.0000
11.0100
14.0100
15.0000
16.0000
18.0000
19.0000
21.0000
23.0000
24.0300
31.0200
35.0000
36.0200
37.0200
37.0300
40.0300
41.0100
41.0200
41.0300
41.0400
43.0100
44.0400
45.0100
45.0200
46.0100
46.0200
50.0100
51.0200
54.0000
55.0300
68.0900
83.0300
Coeff.
-2.46e-14
-0.267
-1.000
-0.667
-0.500
-2.46e-14
-0.190
-0.0645
-2.45e-14
-1.000
-2.45e-14
-0.500
-0.0625
-2.47e-14
-0.333
-0.0667
-2.44e-14
-0.305
-2.45e-14
-2.44e-14
-0.167
-1.000
-2.43e-14
-2.43e-14
-2.43e-14
-0.500
-2.44e-14
-2.44e-14
-2.43e-14
-0.333
-0.500
-0.500
-2.48e-14
-0.333
-0.200
-1.000
-1.000
-1.000
-1.000
-2.46e-14
-0.333
-2.49e-14
-1.000
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent
or Caregiver;
reference group is
Tract 1.0200
bold = p <.05
261
1.524
0.0276
0.232
P value
1.000
0.524
0.082
0.155
0.314
1.000
0.646
0.875
1.000
0.082
1.000
0.254
0.881
1.000
0.476
0.873
1.000
0.455
1.000
1.000
0.689
0.082
1.000
1.000
1.000
0.314
1.000
1.000
1.000
0.476
0.314
0.314
1.000
0.476
0.633
0.082
0.082
0.082
0.045
1.000
0.435
1.000
0.082
P/C employment
status at time of DHA
P/C hourly wage at
move-in
time of DHA move-in
(1=employed, 0=not
employed)
Coeff.
-1.000
-0.733
-1.000
-0.333
8.11e-15
-0.333
-0.476
-0.613
6.88e-15
6.67e-15
6.62e-15
-0.500
-0.625
-1.000
-0.667
-0.333
-1.000
-0.695
5.87e-15
-0.500
-1.000
-1.000
5.30e-15
-0.500
-1.000
-0.500
4.81e-15
4.92e-15
-1.000
-0.333
4.70e-15
-1.000
-0.333
-0.333
-1.200
8.40e-15
8.34e-15
8.66e-15
-0.500
-1.000
-0.556
-0.333
4.36e-15
261
0.242
1.000
0.0458
P value
0.604
0.602
0.604
0.832
1.000
0.832
0.733
0.658
1.000
1.000
1.000
0.734
0.656
0.604
0.672
0.813
0.549
0.613
1.000
0.764
0.475
0.604
1.000
0.764
0.604
0.764
1.000
1.000
0.604
0.832
1.000
0.549
0.832
0.832
0.394
1.000
1.000
1.000
0.764
0.604
0.699
0.832
1.000
Coeff.
-17.00
-12.11
-17.00
-10.67
2.350
-2.000
-8.343
-11.32
5.000
2.000
1.500
-7.917
-10.90
-17
-12.00
-5.205
-17.00
-5.769
2.000
-9.500
-8.504
-17.00
-1.000
-7.750
-17.00
-8.325
7.000
0.250
-17.00
-1.250
1.450
-17.00
-4.763
-3.000
-10.66
3.000
1.000
4.000
-7.775
-17
-9.111
-3.467
17.00
261
1.352
0.0853
0.211
P value
0.190
0.201
0.190
0.314
0.834
0.850
0.374
0.225
0.700
0.877
0.908
0.424
0.249
0.190
0.257
0.582
0.131
0.533
0.877
0.398
0.367
0.190
0.938
0.490
0.190
0.458
0.589
0.985
0.190
0.906
0.897
0.131
0.653
0.777
0.261
0.817
0.938
0.758
0.489
0.190
0.346
0.743
0.190
P/C disability status
at time of survey
(1=yes; 0=no)
Coeff.
-1.000
-0.933
-5.22e-14
-1.000
-1.000
-1.000
-0.810
-0.935
-1.000
-1.000
-5.37e-14
-1.000
-0.938
-1.000
-1.000
-0.800
-0.500
-0.881
-1.000
-0.500
-0.722
-5.51e-14
-1.000
-1.000
-5.52e-14
-0.500
-1.000
-1.000
-1.000
-0.667
-1.000
-5.60e-14
-0.667
-1.000
-0.733
-1.000
-1.000
-5.18e-14
-1.000
-1.000
-0.889
-1.000
-1.000
261
1.709
0.00711
0.253
P value
0.042
0.009
1.000
0.013
0.019
0.013
0.023
0.008
0.042
0.042
1.000
0.008
0.009
0.042
0.013
0.026
0.238
0.012
0.042
0.238
0.043
1.000
0.042
0.019
1.000
0.238
0.042
0.042
0.042
0.096
0.019
1.000
0.096
0.013
0.041
0.042
0.042
1.000
0.019
0.042
0.015
0.013
0.042
P/C receiving Food
P/C received TANF
Stamps at time of
at time of DHA moveDHA move-in (1=yes,
in (1=yes, 0=no)
0=no)
Coeff.
-1.000
-0.867
-7.97e-15
-1.000
-1.000
-0.667
-0.571
-0.581
-8.65e-15
-1.000
-1.000
-0.667
-0.688
-1.000
-0.333
-0.333
-0.500
-0.983
-9.41e-15
-9.48e-15
-1.111
-1.000
-9.87e-15
-1.000
-1.00e-14
-1.000
-1.000
-1.000
-1.03e-14
-1.000
-1.000
-1.000
-1.000
-1.000
-0.733
-1.000
-1.000
-1.000
-1.000
-1.000
-0.889
-0.667
-1.000
261
0.306
1.000
0.0572
P value
0.547
0.474
1.000
0.460
0.486
0.623
0.634
0.626
1.000
0.547
0.547
0.599
0.570
0.547
0.806
0.783
0.728
0.406
1.000
1.000
0.357
0.547
1.000
0.486
1.000
0.486
0.547
0.547
1.000
0.460
0.486
0.486
0.460
0.460
0.545
0.547
0.547
0.547
0.486
0.547
0.472
0.623
0.547
Coeff.
-1.000
-0.600
-3.06e-14
-0.667
-0.500
-0.333
-0.333
-0.419
-3.04e-14
-1.000
-3.03e-14
-0.667
-0.250
-1.000
-3.14e-14
-0.267
-0.500
-0.712
-3.01e-14
-0.500
-1.000
-3.00e-14
-2.98e-14
-0.500
-3.00e-14
-1.000
-3.01e-14
-3.01e-14
-3.02e-14
-1.000
-0.500
-0.500
-1.000
-0.667
-1.000
-3.05e-14
-1.000
-3.06e-14
-0.500
-1.000
-0.444
-0.333
-3.05e-14
261
0.260
1.000
0.0489
P value
0.605
0.671
1.000
0.672
0.765
0.833
0.811
0.762
1.000
0.605
1.000
0.651
0.859
0.605
1.000
0.850
0.765
0.605
1.000
0.765
0.476
1.000
1.000
0.765
1.000
0.550
1.000
1.000
1.000
0.526
0.765
0.765
0.526
0.672
0.478
1.000
0.605
1.000
0.765
0.605
0.757
0.833
1.000
P/C had checking
account at time of
DHA move-in
(1=yes, 0=no)
Coeff.
9.19e-14
0.333
1.000
0.333
1.000
1.000
0.238
0.0323
1.000
1.000
1.000
0.500
0.375
9.18e-14
9.19e-14
0.400
9.19e-14
0.102
1.000
9.24e-14
-0.111
9.26e-14
9.27e-14
0.500
9.27e-14
1.000
1.000
9.26e-14
1.000
0.667
1.000
0.500
0.667
0.333
-0.400
9.19e-14
9.18e-14
1.000
9.18e-14
9.18e-14
0.333
1.000
1.000
261
0.287
1.000
0.0538
P value
1.000
0.826
0.630
0.844
0.579
0.556
0.874
0.983
0.630
0.630
0.630
0.753
0.804
1.000
1.000
0.792
1.000
0.945
0.630
1.000
0.941
1.000
1.000
0.781
1.000
0.579
0.630
1.000
0.630
0.694
0.579
0.781
0.694
0.844
0.792
1.000
1.000
0.630
1.000
1.000
0.830
0.556
0.630
P/C had health
insurance at time of
DHA move-in
(1=yes, 0=no)
Coeff.
-1.000
-0.333
-1.84e-14
-0.333
-1.85e-14
-1.79e-14
-0.190
-0.355
-1.000
-1.90e-14
-1.90e-14
-0.333
-0.375
-1.000
-1.88e-14
-0.200
-1.85e-14
-0.508
-1.98e-14
-1.98e-14
-0.778
-2.01e-14
-2.00e-14
-2.02e-14
-2.02e-14
-2.02e-14
-2.03e-14
-2.03e-14
-2.03e-14
-1.95e-14
-2.05e-14
-0.500
-0.667
-0.667
-0.867
-1.83e-14
-1.000
-1.82e-14
-1.82e-14
-1.000
-0.111
-0.667
-2.06e-14
261
0.216
1.000
0.0411
P value
0.605
0.813
1.000
0.833
1.000
1.000
0.892
0.798
0.605
1.000
1.000
0.821
0.790
0.605
1.000
0.887
1.000
0.712
1.000
1.000
0.580
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.765
0.673
0.673
0.540
1.000
0.605
1.000
1.000
0.605
0.939
0.673
1.000
Exhibit A-2A. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 0-1 Child (cont.)
P/C had too little
P/C had difficulty
money for food at time paying all bills at time
of DHA move-in
of DHA move-in
(1=yes, 0=no)
(1=yes, 0=no)
Census Tract
2.0100
2.0200
3.0100
3.0200
5.0200
7.0100
7.0200
8.0000
9.0200
9.0300
9.0400
10.0000
11.0100
14.0100
15.0000
16.0000
18.0000
19.0000
21.0000
23.0000
24.0300
31.0200
35.0000
36.0200
37.0200
37.0300
40.0300
41.0100
41.0200
41.0300
41.0400
43.0100
44.0400
45.0100
45.0200
46.0100
46.0200
50.0100
51.0200
54.0000
55.0300
68.0900
83.0300
Coeff.
3.54e-14
0.333
1.000
0.667
0.500
0.333
0.333
0.290
3.43e-14
1.000
1.000
0.333
0.250
1.000
0.333
0.467
0.500
-0.186
3.63e-14
3.64e-14
-0.111
3.62e-14
3.62e-14
3.62e-14
3.62e-14
3.62e-14
3.61e-14
1.000
3.61e-14
0.667
0.500
3.61e-14
0.333
0.333
-0.267
3.55e-14
1.000
1.000
3.59e-14
3.60e-14
0.333
0.667
3.60e-14
Observations
F-Test
p value
Pseudo R ²
261
0.259
1.000
0.0488
Note: P/C = Parent
or Caregiver;
reference group is
Tract 1.0200
bold = p <.05
P value
1.000
0.825
0.628
0.692
0.780
0.843
0.823
0.845
1.000
0.628
0.628
0.832
0.868
0.628
0.843
0.757
0.780
0.899
1.000
1.000
0.941
1.000
1.000
1.000
1.000
1.000
1.000
0.628
1.000
0.692
0.780
1.000
0.843
0.843
0.859
1.000
0.628
0.628
1.000
1.000
0.828
0.692
1.000
Coeff.
9.63e-14
0.467
9.63e-14
0.667
1.000
0.667
0.524
0.0645
9.63e-14
1.000
9.62e-14
0.333
0.438
9.65e-14
0.333
-0.200
1.000
0.0508
9.65e-14
9.65e-14
-0.333
1.000
9.65e-14
9.64e-14
1.000
9.65e-14
9.64e-14
1.000
1.000
0.333
9.64e-14
0.500
0.333
-2.333
-0.267
1.000
9.64e-14
1.000
9.65e-14
9.66e-14
0.556
0.333
1.000
261
0.383
1.000
0.0706
P value
1.000
0.789
1.000
0.732
0.628
0.732
0.762
0.970
1.000
0.675
1.000
0.855
0.801
1.000
0.864
0.909
0.628
0.976
1.000
1.000
0.847
0.675
1.000
1.000
0.675
1.000
1.000
0.675
0.675
0.864
1.000
0.809
0.864
0.232
0.878
0.675
1.000
0.675
1.000
1.000
0.755
0.864
0.675
Frequency that P/C
drank alcohol since
becoming a parent
Coeff.
4.03e-14
0.133
4.06e-14
1.667
-4.000
1.333
0.238
0.548
4.20e-14
8.000
2.000
0.833
0.750
3.94e-14
2.000
0.600
0.500
0.678
4.32e-14
3.000
1.500
4.38e-14
4.38e-14
0.500
4.43e-14
4.45e-14
4.47e-14
1.000
4.46e-14
2.000
1.000
-3.500
0.333
1.667
1.000
4.00e-14
2.000
3.97e-14
3.96e-14
2.000
1.444
-2.667
1.000
261
1.299
0.117
0.205
P value
1.000
0.950
1.000
0.484
0.114
0.575
0.910
0.793
1.000
0.006
0.493
0.708
0.724
1.000
0.401
0.778
0.843
0.744
1.000
0.235
0.479
1.000
1.000
0.843
1.000
1.000
1.000
0.731
1.000
0.401
0.692
0.166
0.889
0.484
0.638
1.000
0.493
1.000
1.000
0.493
0.506
0.263
0.731
Frequency that P/C
smoked marijuana
since becoming a
parent
Coeff.
7.87e-14
-0.200
7.87e-14
7.88e-14
0.500
0.333
-0.0952
0.161
7.88e-14
4.000
7.88e-14
0.667
-0.312
7.81e-14
7.86e-14
0.133
0.500
0.203
7.66e-14
5.000
0.667
7.63e-14
7.63e-14
0.500
7.63e-14
7.64e-14
7.63e-14
2.000
7.64e-14
7.74e-14
7.63e-14
3.000
7.87e-14
0.667
0.667
7.86e-14
7.86e-14
7.85e-14
7.85e-14
1.000
0.444
0.333
1.000
261
0.777
0.837
0.133
P value
1.000
0.911
1.000
1.000
0.813
0.867
0.957
0.927
1.000
0.103
1.000
0.721
0.861
1.000
1.000
0.940
0.813
0.907
1.000
0.019
0.708
1.000
1.000
0.813
1.000
1.000
1.000
0.414
1.000
1.000
1.000
0.158
1.000
0.739
0.709
1.000
1.000
1.000
1.000
0.683
0.807
0.867
0.683
P/C ever seen a
psychiatrist (1=yes,
0=no)
Coeff.
2.98e-14
0.467
1.000
3.08e-14
2.99e-14
0.333
0.429
0.258
1.000
1.000
3.03e-14
0.500
0.375
2.99e-14
0.333
0.467
0.500
0.237
1.000
0.500
0.556
3.11e-14
3.11e-14
0.500
3.11e-14
3.12e-14
3.12e-14
3.12e-14
3.13e-14
0.667
3.13e-14
3.13e-14
0.667
0.667
0.267
1.000
2.99e-14
2.99e-14
2.99e-14
2.99e-14
0.222
0.667
3.15e-14
261
0.972
0.526
0.162
P value
1.000
0.342
0.138
1.000
1.000
0.544
0.379
0.593
0.138
0.138
1.000
0.331
0.444
1.000
0.544
0.342
0.391
0.621
0.138
0.391
0.256
1.000
1.000
0.391
1.000
1.000
1.000
1.000
1.000
0.225
1.000
1.000
0.225
0.225
0.587
0.138
1.000
1.000
1.000
1.000
0.657
0.225
1.000
Number of years
during childhood
that P/C lived in a
home owned by
parents
P value
Coeff.
P value
1.000
1.17e-12
1.000
0.565
7.533
0.536
0.345
1.17e-12
1.000
1.000
9.000
0.509
1.000
25.00
0.085
1.000
22.00
0.107
0.484
10.62
0.379
0.377
12.42
0.301
1.000
25.00
0.135
1.000
22.00
0.188
1.000
18.00
0.281
0.096
1.12e-12
1.000
0.680
10.19
0.402
1.000
1.17e-12
1.000
0.222
25.33
0.064
0.555
13.80
0.258
1.000
7.000
0.628
0.496
10.20
0.391
0.271
17.00
0.309
1.000
13.50
0.350
0.579
12.06
0.320
1.000
27.00
0.107
1.000
27.00
0.107
0.413
13.50
0.350
0.238
1.19e-12
1.000
1.000
8.500
0.556
1.000
27.00
0.107
1.000
10.00
0.549
1.000
1.20e-12
1.000
0.630
18.00
0.187
0.388
15.00
0.300
1.000
21.00
0.147
1.000
15.00
0.271
0.748
7.000
0.607
0.630
12.07
0.322
1.000
27.00
0.107
1.000
21.00
0.209
1.000
16.00
0.338
1.000
5.500
0.703
1.000
1.17e-12
1.000
0.432
14.00
0.261
1.000
15.33
0.261
1.000
27.00
0.107
Number of years
during childhood that
P/C lived in public
housing
Coeff.
2.31e-13
5.333
12.00
2.02e-13
2.33e-13
1.98e-13
6.429
8.065
2.41e-13
2.41e-13
2.41e-13
16.17
3.813
2.28e-13
12.67
5.467
2.28e-13
6.169
14.00
2.43e-13
5.111
2.48e-13
2.48e-13
9.000
15.00
2.50e-13
2.51e-13
2.51e-13
2.51e-13
5.000
9.500
2.52e-13
2.14e-13
3.333
4.467
2.30e-13
2.31e-13
2.29e-13
2.30e-13
2.29e-13
7.444
2.18e-13
2.56e-13
261
0.773
0.843
0.133
261
0.981
0.511
0.163
P/C born in the
United States
(1=yes; 0=no)
Coeff.
-2.37e-14
-0.133
-2.37e-14
-2.35e-14
-2.37e-14
-2.35e-14
-0.0952
-0.0645
-2.38e-14
-2.38e-14
-2.39e-14
-2.31e-14
-2.36e-14
-2.36e-14
-2.36e-14
-0.133
-2.36e-14
-0.102
-2.37e-14
-2.37e-14
-0.167
-1.000
-2.37e-14
-2.38e-14
-2.37e-14
-2.37e-14
-1.000
-2.38e-14
-2.38e-14
-0.333
-2.38e-14
-0.500
-2.36e-14
-2.36e-14
-0.133
-2.36e-14
-2.36e-14
-2.36e-14
-2.36e-14
-2.36e-14
-2.37e-14
-0.333
-2.38e-14
261
0.865
0.709
0.146
P value
1.000
0.660
1.000
1.000
1.000
1.000
0.751
0.829
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.660
1.000
0.731
1.000
1.000
0.580
0.017
1.000
1.000
1.000
1.000
0.017
1.000
1.000
0.325
1.000
0.165
1.000
1.000
0.660
1.000
1.000
1.000
1.000
1.000
1.000
0.325
1.000
Spanish language
interview (1=yes;
0=no)
Coeff.
-1.36e-15
0.133
-1.36e-15
-1.22e-15
-1.36e-15
-1.22e-15
0.0952
-1.62e-15
1.000
-1.35e-15
-1.34e-15
-1.18e-15
-1.43e-15
-1.31e-15
-1.25e-15
0.0667
-1.34e-15
0.0508
-1.27e-15
-1.26e-15
0.0556
-1.25e-15
-1.24e-15
-1.23e-15
-1.25e-15
-1.23e-15
-1.23e-15
-1.23e-15
-1.23e-15
0.333
-1.22e-15
-1.22e-15
-1.27e-15
-1.31e-15
-1.24e-15
-1.35e-15
-1.34e-15
-1.34e-15
-1.33e-15
-1.34e-15
-1.31e-15
-8.30e-16
-1.21e-15
261
0.905
0.643
0.152
P value
1.000
0.525
1.000
1.000
1.000
1.000
0.647
1.000
0.001
1.000
1.000
1.000
1.000
1.000
1.000
0.751
1.000
0.804
1.000
1.000
0.790
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.156
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
Exhibit A-2A. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 0-1 Child (cont.)
Biological father
always lived in
household with
child(ren) (1=yes;
0=no)
Census Tract
2.0100
2.0200
3.0100
3.0200
5.0200
7.0100
7.0200
8.0000
9.0200
9.0300
9.0400
10.0000
11.0100
14.0100
15.0000
16.0000
18.0000
19.0000
21.0000
23.0000
24.0300
31.0200
35.0000
36.0200
37.0200
37.0300
40.0300
41.0100
41.0200
41.0300
41.0400
43.0100
44.0400
45.0100
45.0200
46.0100
46.0200
50.0100
51.0200
54.0000
55.0300
68.0900
83.0300
Coeff.
3.92e-14
0.0667
1.000
0.333
3.92e-14
3.93e-14
0.0952
0.0323
3.90e-14
3.90e-14
3.90e-14
0.333
3.94e-14
3.90e-14
3.92e-14
0.133
3.87e-14
0.220
3.86e-14
3.86e-14
0.111
3.84e-14
1.000
3.85e-14
1.000*
3.85e-14
3.85e-14
3.85e-14
3.85e-14
0.333
3.85e-14
0.500
3.91e-14
0.667
0.200
3.91e-14
1.000
3.90e-14
0.500
3.90e-14
0.111
4.00e-14
1.000
Observations
F-Test
p value
Pseudo R ²
261
1.513
0.0299
0.231
Note: P/C = Parent
or Caregiver;
reference group is
Tract 1.0200
bold = p <.05
P value
1.000
0.849
0.038
0.396
1.000
1.000
0.784
0.926
1.000
1.000
1.000
0.364
1.000
1.000
1.000
0.704
1.000
0.520
1.000
1.000
0.750
1.000
0.038
1.000
0.038
1.000
1.000
1.000
1.000
0.396
1.000
0.230
1.000
0.090
0.569
1.000
0.038
1.000
0.230
1.000
0.756
1.000
0.038
Parent's age at time
of DHA move-in
Coeff.
19.00
5.867
24.00
14.33
3.500
7.333
0.619
-2.194
-3.000
-7.000
5.03e-13
3.000
-1.750
16.00
0.667
-2.600
25.00
0.576
-7.000
-2.500
6.833
26.00
-3.000
27.00
-6.000
13.00
3.000
-6.000
-9.000
2.667
4.500
2.000
6.333
-5.000
3.733
-8.000
7.000
-1.000
-3.500
27.00
-0.111
8.667
12.00
261
1.829
0.00277
0.266
P value
0.195
0.583
0.102
0.231
0.783
0.540
0.953
0.835
0.838
0.633
1.000
0.789
0.870
0.275
0.956
0.808
0.050
0.956
0.633
0.844
0.521
0.077
0.838
0.034
0.682
0.306
0.838
0.682
0.539
0.823
0.723
0.875
0.596
0.676
0.727
0.585
0.633
0.946
0.783
0.066
0.992
0.469
0.413
Parent have HS
Parent have any
Kids share same
P/C African American
diploma at time of
higher education at
biological dad (1=yes;
(1=yes; 0=no)
DHA move-in (1=yes; time of DHA move-in
0=no)
0=no)
(1=yes; 0=no)
Coeff.
3.94e-14
0.200
3.92e-14
3.87e-14
0.500
1.000
0.286
0.387
3.86e-14
3.86e-14
3.86e-14
0.167
0.438
4.01e-14
0.667
0.667
4.05e-14
0.458
1.000
1.000
0.833
4.22e-14
1.000
1.000
1.000
0.500
1.000
1.000
1.000
0.333
1.000
1.000
1.000
3.97e-14
0.400
3.94e-14
3.95e-14
1.000
1.000
1.000
0.111
0.667
4.23e-14
261
1.938
0.00113
0.277
P value
1.000
0.677
1.000
1.000
0.380
0.063
0.548
0.413
1.000
1.000
1.000
0.740
0.362
1.000
0.215
0.166
1.000
0.329
0.129
0.080
0.082
1.000
0.129
0.080
0.129
0.380
0.129
0.129
0.129
0.535
0.080
0.080
0.063
1.000
0.405
1.000
1.000
0.129
0.080
0.129
0.821
0.215
1.000
Coeff.
-1.000
-0.533
-1.000
-1.000
-0.500
-0.667
-0.714
-0.742
-1.000
-1.000
-1.000
-0.833
-0.563
-9.49e-14
-0.333
-0.733
-0.500
-0.508
-1.000
-1.000
-0.722
-1.000
-1.000
-0.500
-9.88e-14
-9.90e-14
-1.000
-1.000
-1.000
-1.000
-0.500
-0.500
-1.000
-0.333
-0.933
-1.000
-9.52e-14
-1.000
-0.500
-1.000
-0.667
-1.000
-1.000
261
1.093
0.333
0.178
P value
0.134
0.273
0.134
0.067
0.386
0.221
0.139
0.122
0.134
0.134
0.134
0.102
0.247
1.000
0.540
0.132
0.386
0.285
0.134
0.084
0.136
0.134
0.134
0.386
1.000
1.000
0.134
0.134
0.134
0.067
0.386
0.386
0.067
0.540
0.056
0.134
1.000
0.134
0.386
0.134
0.180
0.067
0.134
Coeff.
2.20e-14
0.0667
2.20e-14
2.20e-14
2.19e-14
0.333
0.190
0.0323
2.19e-14
2.19e-14
2.19e-14
2.19e-14
2.21e-14
2.19e-14
2.21e-14
0.133
2.18e-14
0.102
2.18e-14
2.18e-14
0.111
2.18e-14
2.18e-14
0.500
2.18e-14
2.18e-14
2.19e-14
2.18e-14
2.18e-14
0.333
0.500
2.18e-14
0.667
0.333
0.333
2.20e-14
2.20e-14
1.000
2.20e-14
2.20e-14
0.111
0.333
2.18e-14
261
1.055
0.390
0.173
P value
1.000
0.842
1.000
1.000
1.000
0.372
0.565
0.922
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.689
1.000
0.755
1.000
1.000
0.738
1.000
1.000
0.207
1.000
1.000
1.000
1.000
1.000
0.372
0.207
1.000
0.075
0.372
0.318
1.000
1.000
0.029
1.000
1.000
0.744
0.372
1.000
Coeff.
-3.11e-14
-0.267
-1.000
-3.12e-14
-3.09e-14
-3.12e-14
-0.381
-0.419
-3.08e-14
-1.000
-1.000
-0.333
-0.188
-3.14e-14
-0.333
-0.600
-3.13e-14
-0.288
-3.15e-14
-0.500
-0.333
-3.15e-14
-1.000
-3.16e-14
-3.15e-14
-3.16e-14
-1.000
-1.000
-3.16e-14
-3.26e-14
-0.500
-3.16e-14
-3.14e-14
-0.333
-0.400
-3.12e-14
-3.12e-14
-3.13e-14
-3.13e-14
-3.13e-14
-0.222
-0.333
-3.17e-14
261
0.944
0.576
0.158
P value
1.000
0.580
0.130
1.000
1.000
1.000
0.425
0.376
1.000
0.130
0.130
0.508
0.696
1.000
0.536
0.214
1.000
0.540
1.000
0.382
0.487
1.000
0.130
1.000
1.000
1.000
0.130
0.130
1.000
1.000
0.382
1.000
1.000
0.536
0.407
1.000
1.000
1.000
1.000
1.000
0.651
0.536
1.000
Parent Depressive
Symptomatology
Scale at time of
interview
Coeff.
1.51e-13
11.87
31.00
5.000
5.000
15.00
8.571
9.516
5.000
18.00
1.57e-13
4.167
12.19
5.000
17.33
13.80
5.500
10.02
1.000
9.500
13.28
21.00
-1.000
2.500
-1.000
4.500
5.000
21.00
30.00
3.333
1.87e-13
11.00
7.667
6.667
8.800
10.00
7.000
8.000
4.500
19.00
11.89
7.000
4.000
261
0.928
0.603
0.155
P value
1.000
0.245
0.027
0.661
0.679
0.189
0.396
0.343
0.720
0.198
1.000
0.696
0.231
0.720
0.129
0.176
0.649
0.314
0.943
0.432
0.191
0.133
0.943
0.836
0.943
0.709
0.720
0.133
0.032
0.770
1.000
0.363
0.501
0.558
0.388
0.473
0.616
0.566
0.709
0.174
0.253
0.539
0.774
Parenting Efficacy
Scale at time of
interview
Coeff.
1.000
-0.667
-5.000
1.333
-2.000
-1.667
-0.238
-0.581
-2.000
-2.000
3.000
1.667
0.437
3.000
0.333
1.000
-2.500
-0.0169
3.000
0.500
-0.111
-10.00
3.000
3.000
3.000
-1.000
-2.000
2.000
-10.00
-0.333
2.500
-3.500
-4.000
2.000
0.733
2.000
-4.000
1.000
3.000
-2.000
1.111
1.667
-7.000
261
1.290
0.123
0.204
P value
0.832
0.846
0.288
0.728
0.623
0.664
0.944
0.864
0.671
0.671
0.524
0.643
0.898
0.524
0.931
0.771
0.539
0.996
0.524
0.902
0.974
0.034
0.524
0.462
0.524
0.806
0.671
0.671
0.034
0.931
0.539
0.391
0.298
0.603
0.831
0.671
0.395
0.832
0.462
0.671
0.751
0.664
0.138
Parenting Beliefs
Scale at time of
interview
Coeff.
7.000
3.400
7.000
3.667
4.500
1.667
4.429
2.548
2.000
-1.000
7.000
4.167
1.375
4.000
5.333
3.067
2.500
3.068
1.000
1.500
2.333
-4.000
7.000
-3.000
5.000
5.000
3.000
7.000
3.04e-13
1.333
3.05e-13
2.000
5.333
4.333
3.533
5.000
4.000
2.90e-13
7.000
3.000
3.222
0.667
2.000
261
0.964
0.540
0.160
P value
0.161
0.351
0.161
0.368
0.298
0.682
0.220
0.477
0.688
0.841
0.161
0.274
0.705
0.423
0.191
0.400
0.563
0.388
0.841
0.728
0.520
0.423
0.161
0.487
0.316
0.247
0.547
0.161
1.000
0.743
1.000
0.643
0.191
0.288
0.332
0.316
0.423
1.000
0.106
0.547
0.386
0.870
0.688
Exhibit A-2B. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 2 Children
P/C is single parent
(1=yes, 0=no)
Census Tract
2.0200
3.0100
4.0100
4.0200
5.0200
7.0100
7.0200
8.0000
9.0200
9.0300
9.0500
10.0000
11.0100
11.0200
13.0100
13.0200
14.0200
14.0300
15.0000
16.0000
18.0000
19.0000
21.0000
23.0000
24.0300
31.0100
35.0000
36.0100
36.0200
36.0300
37.0300
41.0100
41.0200
41.0300
41.0400
42.0100
42.0200
43.0100
43.0400
44.0300
44.0400
45.0100
45.0200
46.0200
47.0000
48.0200
50.0100
54.0000
55.0300
68.0900
69.0100
83.0300
85.3400
Coeff.
-0.167
3.29e-15
3.00e-15
2.96e-15
2.89e-15
-0.500
-0.267
-0.250
-0.500
-0.500
-1.000
3.04e-15
3.71e-15
-0.500
-1.000
-1.000
-0.250
-1.000
2.81e-15
-0.278
3.47e-15
-0.277
2.93e-15
2.90e-15
-0.111
2.96e-15
-0.143
2.96e-15
2.97e-15
4.07e-15
2.96e-15
3.94e-15
2.91e-15
4.14e-15
2.95e-15
2.94e-15
-1.000
-1.000
-0.500
-0.500
2.96e-15
3.38e-15
-0.308
2.97e-15
2.94e-15
2.97e-15
-0.500
-0.250
-0.200
2.91e-15
2.96e-15
2.93e-15
2.92e-15
Observations
F-Test
p value
Pseudo R ²
244
0.932
0.610
0.206
Note: P/C = Parent
or Caregiver;
reference group is
Tract 1.0200
bold = p <.05
P value
0.601
1.000
1.000
1.000
1.000
0.231
0.396
0.419
0.231
0.167
0.051
1.000
1.000
0.231
0.051
0.051
0.489
0.051
1.000
0.372
1.000
0.359
1.000
1.000
0.733
1.000
0.669
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.051
0.051
0.231
0.231
1.000
1.000
0.332
1.000
1.000
1.000
0.231
0.448
0.567
1.000
1.000
1.000
1.000
P/C employment
status at time of DHA
move-in (1=employed,
0=not employed)
Coeff.
-0.583
-0.667
-3.72e-14
-3.71e-14
-0.500
-0.500
-0.733
-0.400
-0.500
-0.500
-3.64e-14
-0.250
-0.429
-0.500
-1.000
-1.000
-0.250
-3.61e-14
-5.000
-1.056
-0.667
-0.574
-1.000
-3.71e-14
-0.333
-3.70e-14
-0.286
-0.500
-0.500
-3.57e-14
-3.72e-14
-3.57e-14
-3.68e-14
-3.56e-14
-3.69e-14
-3.71e-14
-3.70e-14
-3.72e-14
-0.500
-3.71e-14
-3.72e-14
-0.143
-0.462
-1.000
-3.62e-14
-1.000
-0.500
-2.750
-0.200
-1.000
-3.59e-14
-1.000
-3.61e-14
244
0.861
0.735
0.194
P value
0.604
0.620
1.000
1.000
0.734
0.734
0.508
0.714
0.734
0.695
1.000
0.845
0.717
0.734
0.579
0.579
0.845
1.000
0.001
0.337
0.620
0.589
0.497
1.000
0.772
1.000
0.809
0.734
0.734
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.734
1.000
1.000
0.904
0.680
0.579
1.000
0.579
0.734
0.019
0.871
0.579
1.000
0.579
1.000
P/C hourly wage at
time of DHA move-in
Coeff.
-12.13
-12.00
0.900
0.500
-7.750
-9.500
-13.97
-8.889
-18.50
-10.37
-1.250
-6.625
-8.190
-9.425
-18.50
-18.50
-4.398
8.500
-11.75
-10.96
-12.50
-6.530
-18.50
-5.075
-7.989
6.500
-2.757
-7.000
-8.000
0.333
1.100
-3.390
6.000
2.833
1.500
0.500
-0.480
-1.250
-8.250
2.375
-0.500
-3.214
-10.67
-18.50
-18.50
-18.50
-7.500
-9.937
-4.400
-18.50
9.500
-18.50
1.500
244
1.267
0.127
0.261
P value
0.086
0.154
0.936
0.965
0.400
0.303
0.045
0.194
0.046
0.194
0.912
0.406
0.268
0.306
0.102
0.102
0.581
0.451
0.203
0.111
0.138
0.326
0.046
0.524
0.268
0.564
0.709
0.447
0.385
0.968
0.922
0.687
0.595
0.736
0.894
0.965
0.966
0.912
0.370
0.796
0.965
0.663
0.128
0.102
0.102
0.102
0.415
0.173
0.568
0.102
0.400
0.102
0.894
P/C disability status at
time of survey (1=yes;
0=no)
Coeff.
0.0833
3.44e-15
3.01e-15
1.000
3.39e-15
0.500
0.133
0.100
3.40e-15
3.38e-15
3.41e-15
0.500
0.286
3.37e-15
3.13e-15
3.07e-15
3.12e-15
3.10e-15
0.500
0.111
3.38e-15
0.0638
3.35e-15
3.16e-15
0.333
3.11e-15
3.56e-15
3.10e-15
0.500
3.30e-15
3.07e-15
0.333
3.27e-15
3.26e-15
3.21e-15
3.14e-15
3.10e-15
1.000
0.500
3.08e-15
1.000
3.68e-15
0.154
3.15e-15
1.000
1.000
3.31e-15
3.24e-15
3.22e-15
3.29e-15
3.14e-15
3.13e-15
3.20e-15
244
1.555
0.0168
0.302
P value
0.722
1.000
1.000
0.008
1.000
0.104
0.564
0.660
1.000
1.000
1.000
0.061
0.246
1.000
1.000
1.000
1.000
1.000
0.104
0.627
1.000
0.773
1.000
1.000
0.165
1.000
1.000
1.000
0.104
1.000
1.000
0.235
1.000
1.000
1.000
1.000
1.000
0.008
0.104
1.000
0.008
1.000
0.509
1.000
0.008
0.008
1.000
1.000
1.000
1.000
1.000
1.000
1.000
P/C received TANF at
P/C receiving Food
time of DHA move-in Stamps at time of DHA
(1=yes, 0=no)
move-in (1=yes, 0=no)
Coeff.
0.583
-1.59e-15
-8.42e-16
1.000
-1.27e-15
-8.29e-16
0.533
0.600
0.500
0.500
-1.48e-15
0.750
0.286
-1.16e-15
-6.77e-16
1.000
0.250
-6.14e-16
1.000
-2.09e-15
0.333
0.383
0.500
0.500
-0.667
1.000
0.286
0.500
0.500
-2.52e-15
-9.69e-16
0.333
-1.07e-15
0.333
-1.04e-15
-1.03e-15
-9.44e-16
1.000
0.500
-9.93e-16
1.000
0.286
0.538
1.000
-7.03e-16
-7.14e-16
0.500
-0.750
0.200
-1.19e-15
-6.83e-16
-1.07e-15
-7.22e-16
244
0.396
1.000
0.0994
P value
0.536
1.000
1.000
0.509
1.000
1.000
0.566
0.513
0.686
0.640
1.000
0.483
0.773
1.000
1.000
0.509
0.815
1.000
0.418
1.000
0.767
0.668
0.686
0.640
0.490
0.509
0.773
0.686
0.686
1.000
1.000
0.767
1.000
0.767
1.000
1.000
1.000
0.509
0.686
1.000
0.509
0.773
0.566
0.509
1.000
1.000
0.686
0.443
0.846
1.000
1.000
1.000
1.000
Coeff.
0.167
-0.167
-0.500
0.500
-2.82e-14
-2.83e-14
0.167
0.200
0.500
-2.82e-14
0.500
0.250
0.0714
0.500
0.500
0.500
-0.250
-0.500
-2.82e-14
0.222
-0.167
0.160
-2.81e-14
0.250
0.0556
0.500
0.0714
0.500
-2.83e-14
-3.167
-0.500
0.167
-0.500
0.167
0.500
0.500
0.500
-0.500
0.500
-0.500
0.500
-0.0714
0.192
0.500
-0.500
0.500
-2.85e-14
-1.250
0.1000
-0.500
0.500
0.500
-0.500
244
1.068
0.365
0.230
P value
0.825
0.854
0.680
0.680
1.000
1.000
0.823
0.785
0.613
1.000
0.680
0.770
0.928
0.613
0.680
0.680
0.770
0.680
1.000
0.763
0.854
0.823
1.000
0.770
0.943
0.680
0.928
0.613
1.000
0.001
0.680
0.854
0.680
0.854
0.680
0.680
0.680
0.680
0.613
0.613
0.680
0.928
0.798
0.680
0.680
0.680
1.000
0.111
0.904
0.680
0.680
0.680
0.680
P/C had checking
account at time of
DHA move-in (1=yes,
0=no)
Coeff.
-0.417
0.167
0.500
0.500
-0.500
-0.500
-0.300
-0.100
-0.500
0.250
-0.500
-0.500
0.0714
0.500
-0.500
-0.500
0.250
0.500
-5.000
-0.722
-0.167
-0.223
-0.500
0.250
-0.167
0.500
0.0714
-0.500
0.500
-0.500
0.500
-0.167
0.500
0.167
0.500
0.500
-0.500
-0.500
0.500
-0.500
0.500
-0.357
-0.346
-0.500
-0.500
0.500
-0.500
-2.500
0.1000
-0.500
-0.500
0.500
0.500
244
1.022
0.445
0.222
P value
0.700
0.897
0.773
0.773
0.724
0.724
0.778
0.924
0.724
0.838
0.773
0.683
0.950
0.724
0.773
0.773
0.838
0.773
0.001
0.494
0.897
0.827
0.724
0.838
0.880
0.773
0.950
0.724
0.724
0.699
0.773
0.897
0.773
0.897
0.773
0.773
0.773
0.773
0.724
0.724
0.773
0.753
0.748
0.773
0.773
0.773
0.724
0.026
0.933
0.773
0.773
0.773
0.773
P/C had health
insurance at time of
DHA move-in (1=yes,
0=no)
Coeff.
-0.417
-0.333
-1
2.87e-15
-0.500
-0.500
-0.267
-0.250
3.14e-15
-0.250
-1.000
-0.250
-0.143
2.93e-15
2.98e-15
2.91e-15
-0.250
-1.000
-5.000
-0.556
-0.333
-0.532
-0.500
-0.250
-0.333
2.91e-15
-0.286
-0.500
-0.500
-0.333
2.89e-15
-0.333
2.94e-15
2.60e-15
2.93e-15
2.87e-15
-1
2.88e-15
2.85e-15
-1.000
2.91e-15
-0.143
-0.0769
3.06e-15
2.88e-15
3.24e-15
-1.000
-2.500
-0.200
2.87e-15
3.21e-15
3.17e-15
3.21e-15
244
0.763
0.876
0.175
P value
0.714
0.806
0.584
1.000
0.737
0.737
0.812
0.821
1.000
0.846
0.584
0.846
0.905
1.000
1.000
1.000
0.846
0.584
0.001
0.617
0.806
0.621
0.737
0.846
0.775
1.000
0.811
0.737
0.737
0.806
1.000
0.806
1.000
1.000
1.000
1.000
0.584
1.000
1.000
0.502
1.000
0.905
0.946
1.000
1.000
1.000
0.502
0.035
0.873
1.000
1.000
1.000
1.000
Exhibit A-2B. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 2 Children (cont.)
P/C had too little
money for food at
time of DHA move-in
(1=yes, 0=no)
Census Tract
2.0200
3.0100
4.0100
4.0200
5.0200
7.0100
7.0200
8.0000
9.0200
9.0300
9.0500
10.0000
11.0100
11.0200
13.0100
13.0200
14.0200
14.0300
15.0000
16.0000
18.0000
19.0000
21.0000
23.0000
24.0300
31.0100
35.0000
36.0100
36.0200
36.0300
37.0300
41.0100
41.0200
41.0300
41.0400
42.0100
42.0200
43.0100
43.0400
44.0300
44.0400
45.0100
45.0200
46.0200
47.0000
48.0200
50.0100
54.0000
55.0300
68.0900
69.0100
83.0300
85.3400
Coeff.
0.0833
0.667
1.000
1.000
-4.02e-14
-4.11e-14
0.600
0.450
-4.04e-14
0.250
-4.02e-14
-4.13e-14
0.286
0.500
-4.06e-14
-4.07e-14
0.250
-4.06e-14
-4.500
-0.167
0.333
0.191
0.500
0.250
0.444
1.000
0.429
1.000
-4.06e-14
-4.11e-14
-4.11e-14
0.333
1.000
1.000
1.000
1.000
-4.09e-14
-4.10e-14
0.500
0.500
1.000
0.429
0.308
-4.05e-14
-4.07e-14
1.000
0.500
-1.875
0.800
-4.07e-14
-4.06e-14
1.000
-4.05e-14
Observations
F-Test
p value
Pseudo R ²
244
0.923
0.625
0.205
Note: P/C = Parent
or Caregiver;
reference group is
Tract 1.0200
bold = p <.05
P value
0.939
0.610
0.568
0.568
1.000
1.000
0.578
0.672
1.000
0.840
1.000
1.000
0.803
0.727
1.000
1.000
0.840
1.000
0.002
0.876
0.799
0.853
0.727
0.840
0.691
0.568
0.709
0.485
1.000
1.000
1.000
0.799
0.568
0.444
0.568
0.568
1.000
1.000
0.727
0.727
0.568
0.709
0.777
1.000
1.000
0.568
0.727
0.099
0.504
1.000
1.000
0.568
1.000
P/C had difficulty
paying all bills at time
of DHA move-in
(1=yes, 0=no)
Coeff.
-1.833
0.500
0.500
-0.500
-0.500
0.500
0.0333
-0.200
0.500
-0.250
-0.500
-0.500
-0.214
-1.49e-15
-0.500
0.500
0.250
0.500
-5.000
-1.167
-3.500
-0.372
-4.500
-0.250
-0.278
0.500
0.214
-1.46e-15
-0.500
-0.500
0.500
0.167
-0.500
0.167
0.500
-0.500
-0.500
0.500
-1.48e-15
-1.33e-15
0.500
0.0714
-0.192
-0.500
0.500
0.500
-1.05e-15
-2.375
0.1000
-0.500
0.500
-0.500
0.500
244
0.908
0.653
0.202
P value
0.242
0.789
0.842
0.842
0.807
0.807
0.983
0.895
0.807
0.888
0.842
0.778
0.896
1.000
0.842
0.842
0.888
0.842
0.015
0.445
0.062
0.801
0.029
0.888
0.862
0.842
0.896
1.000
0.807
0.789
0.842
0.929
0.842
0.929
0.842
0.842
0.842
0.842
1.000
1.000
0.842
0.965
0.902
0.842
0.842
0.842
1.000
0.143
0.953
0.842
0.842
0.842
0.842
Frequency that P/C
drank alcohol since
becoming a parent
Coeff.
-1.250
-0.833
-1.500
-10.50
1.05e-13
1.500
-1.433
-2.350
-5.500
-3.750
0.500
0.500
-0.929
-1.500
-0.500
-1.500
-0.250
-0.500
1.05e-13
-0.111
-1.167
-0.968
-0.500
-0.500
-1.278
-1.500
-2.071
1.000
2.000
-1.167
2.500
-1.500
-1.500
-1.500
-1.500
-10.50
-1.500
-1.500
0.500
0.500
-1.500
-0.929
-1.500
-1.500
-10.50
2.500
-1.000
1.05e-13
-1.300
-0.500
-1.500
-0.500
0.500
244
1.389
0.0571
0.279
P value
0.536
0.730
0.643
0.001
1.000
0.570
0.471
0.231
0.038
0.102
0.877
0.827
0.661
0.570
0.877
0.643
0.913
0.877
1.000
0.955
0.629
0.612
0.850
0.827
0.536
0.643
0.329
0.705
0.449
0.629
0.440
0.534
0.643
0.534
0.643
0.001
0.643
0.643
0.850
0.850
0.643
0.661
0.455
0.643
0.001
0.440
0.705
1.000
0.556
0.877
0.643
0.877
0.877
Frequency that P/C
smoked marijuana
since becoming a
parent
Coeff.
-0.333
0.500
-1.500
0.500
-1.500
-0.500
-1.567
-1.500
-1.000
-0.500
-1.500
2.000
-2.643
-1.500
-0.500
-1.500
0.250
-0.500
-1.000
-0.222
-1.500
-1.266
-1.500
-0.500
-0.389
-1.500
-2.214
-1.000
2.000
-1.167
-1.500
-1.500
4.500
-1.500
-1.500
-1.500
-1.500
-1.500
-5.500
-1.000
4.500
-1.214
-0.962
-1.500
0.500
-0.500
-1.000
-1.000
-1.300
-0.500
-1.500
-0.500
-0.500
244
0.952
0.572
0.210
P value
0.844
0.805
0.581
0.854
0.499
0.822
0.349
0.363
0.652
0.795
0.581
0.299
0.139
0.499
0.854
0.581
0.897
0.854
0.652
0.893
0.459
0.430
0.499
0.795
0.823
0.581
0.214
0.652
0.368
0.565
0.581
0.459
0.099
0.459
0.581
0.581
0.581
0.581
0.014
0.652
0.099
0.495
0.569
0.581
0.854
0.854
0.652
0.569
0.484
0.854
0.581
0.854
0.854
P/C ever seen a
psychiatrist (1=yes,
0=no)
Coeff.
0.500
0.667
1.000
1.000
-2.33e-14
0.500
0.400
0.300
0.500
-2.35e-14
1.000
1.000
0.571
-2.36e-14
-2.39e-14
1.000
0.250
-2.38e-14
0.500
0.333
0.667
0.468
-2.35e-14
0.250
0.333
1.000
0.429
1.000
1.000
0.333
-2.31e-14
0.667
1.000
0.667
1.000
1.000
-2.31e-14
1.000
1.000
-2.32e-14
-2.31e-14
0.143
0.385
1.000
-2.38e-14
-2.39e-14
-2.38e-14
0.500
0.400
1.000
-2.39e-14
1.000
-2.39e-14
244
1.127
0.277
0.239
P value
0.182
0.137
0.096
0.096
1.000
0.308
0.278
0.409
0.308
1.000
0.096
0.019
0.146
1.000
1.000
0.096
0.556
1.000
0.308
0.361
0.137
0.186
1.000
0.556
0.384
0.096
0.276
0.042
0.042
0.456
1.000
0.137
0.096
0.137
0.096
0.096
1.000
0.096
0.042
1.000
1.000
0.716
0.302
0.096
1.000
1.000
1.000
0.197
0.329
0.096
1.000
0.096
1.000
Number of years
during childhood that
P/C lived in public
housing
Coeff.
3.417
6.333
-2.52e-13
12.00
-2.66e-13
-2.52e-13
7.800
5.050
10.000
-2.66e-13
-2.70e-13
11.25
-2.93e-13
-2.64e-13
-2.52e-13
19.00
15.50
-2.52e-13
5.500
5.222
5.333
3.830
7.500
-2.57e-13
1.889
-2.55e-13
3.143
-2.56e-13
-2.62e-13
11.33
-2.53e-13
4.667
-2.62e-13
-2.88e-13
-2.58e-13
-2.56e-13
-2.55e-13
-2.54e-13
-2.54e-13
-2.54e-13
-2.53e-13
-2.90e-13
5.385
19.00*
-2.51e-13
-2.53e-13
-2.68e-13
7.000
3.600
16.00
-2.52e-13
-2.56e-13
11.00
244
0.977
0.525
0.214
P value
0.562
0.369
1.000
0.205
1.000
1.000
0.181
0.378
0.196
1.000
1.000
0.094
1.000
1.000
1.000
0.046
0.021
1.000
0.476
0.365
0.450
0.492
0.332
1.000
0.754
1.000
0.612
1.000
1.000
0.109
1.000
0.508
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.359
0.046
1.000
1.000
1.000
0.252
0.577
0.092
1.000
1.000
0.245
Number of years
during childhood that
P/C lived in a home
owned by parents
Coeff.
-0.333
-2.500
-9.500
12.50
-9.500
10.000
3.367
3.600
4.000
10.50
-9.500
-2.750
8.786
9.500
14.50
-9.500
-9.500
9.500
-9.500
2.000
-9.500
2.351
-9.500
9.000
4.389
15.50
5.214
4.000
2.000
-9.500
-9.500
3.833
9.500
1.167
4.500
7.500
-9.500
-9.500
9.500
-9.500
17.50
2.071
5.654
-9.500
-9.500
6.500
-9.500
3.500
0.300
-9.500
3.500
3.500
-9.500
244
0.847
0.758
0.191
P value
0.971
0.822
0.525
0.403
0.436
0.412
0.714
0.691
0.743
0.321
0.525
0.795
0.369
0.436
0.332
0.525
0.369
0.525
0.436
0.826
0.394
0.789
0.436
0.394
0.645
0.300
0.594
0.743
0.870
0.394
0.525
0.731
0.525
0.917
0.763
0.616
0.525
0.525
0.436
0.436
0.242
0.832
0.542
0.525
0.525
0.663
0.436
0.717
0.977
0.525
0.815
0.815
0.525
P/C born in the United
States (1=yes; 0=no)
Coeff.
0.333
0.167
0.500
0.500
0.500
-1.06e-14
0.433
0.300
0.500
-9.60e-15
0.500
0.500
0.214
-9.66e-15
0.500
0.500
0.500
0.500
0.500
0.389
0.500
0.351
0.500
0.500
0.389
0.500
0.0714
0.500
0.500
0.500
0.500
0.500
0.500
0.167
0.500
0.500
-0.500
0.500
0.500
0.500
0.500
0.500
0.269
0.500
0.500
0.500
0.500
0.375
0.300
0.500
0.500
0.500
0.500
244
0.649
0.967
0.153
P value
0.234
0.618
0.265
0.265
0.173
1.000
0.117
0.270
0.173
1.000
0.265
0.116
0.466
1.000
0.265
0.265
0.116
0.265
0.173
0.155
0.136
0.185
0.173
0.116
0.175
0.265
0.808
0.173
0.173
0.136
0.265
0.136
0.265
0.618
0.265
0.265
0.265
0.265
0.173
0.173
0.265
0.090
0.333
0.265
0.265
0.265
0.173
0.196
0.328
0.265
0.265
0.265
0.265
Spanish language
interview (1=yes;
0=no)
Coeff.
-0.417
-0.167
-0.500
-0.500
-0.500
2.22e-14
-0.433
-0.400
-0.500
-0.250
-0.500
-0.500
-0.214
2.32e-14
-0.500
-0.500
-0.500
-0.500
-0.500
-0.444
-0.500
-0.479
-0.500
-0.500
-0.500
-0.500
-0.357
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.500
-0.375
-0.500
-0.500
-0.500
-0.500
-0.500
244
1.016
0.456
0.221
P value
0.029
0.462
0.101
0.101
0.045
1.000
0.021
0.031
0.045
0.245
0.101
0.021
0.282
1.000
0.101
0.101
0.021
0.101
0.045
0.017
0.028
0.008
0.045
0.021
0.011
0.101
0.074
0.045
0.045
0.028
0.101
0.028
0.101
0.028
0.101
0.101
0.101
0.101
0.045
0.045
0.101
0.013
0.009
0.101
0.101
0.101
0.045
0.057
0.017
0.101
0.101
0.101
0.101
21
Exhibit A-2B. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 2 Children (cont.)
Census Tract
2.0200
3.0100
4.0100
4.0200
5.0200
7.0100
7.0200
8.0000
9.0200
9.0300
9.0500
10.0000
11.0100
11.0200
13.0100
13.0200
14.0200
14.0300
15.0000
16.0000
18.0000
19.0000
21.0000
23.0000
24.0300
31.0100
35.0000
36.0100
36.0200
36.0300
37.0300
41.0100
41.0200
41.0300
41.0400
42.0100
42.0200
43.0100
43.0400
44.0300
44.0400
45.0100
45.0200
46.0200
47.0000
48.0200
50.0100
54.0000
55.0300
68.0900
69.0100
83.0300
85.3400
Biological father
always lived in
household with
child(ren) (1=yes;
0=no)
Coeff.
P value
1.27e-14
1.000
1.30e-14
1.000
1.42e-14
1.000
1.42e-14
1.000
1.34e-14
1.000
1.43e-14
1.000
0.133
0.616
0.100
0.703
0.500
0.158
0.250
0.414
1.000
0.022
1.45e-14
1.000
1.27e-14
1.000
0.500
0.158
1.43e-14
1.000
1.44e-14
1.000
0.500
0.103
1.000
0.022
1.30e-14
1.000
0.222
0.399
1.30e-14
1.000
0.149
0.559
1.35e-14
1.000
1.39e-14
1.000
1.23e-14
1.000
1.41e-14
1.000
0.143
0.614
0.500
0.158
1.37e-14
1.000
1.27e-14
1.000
1.42e-14
1.000
1.26e-14
1.000
1.37e-14
1.000
0.333
0.302
1.39e-14
1.000
1.40e-14
1.000
1.000
0.022
1.41e-14
1.000
0.500
0.158
1.41e-14
1.000
1.42e-14
1.000
0.143
0.614
0.154
0.566
1.42e-14
1.000
1.43e-14
1.000
1.000
0.022
0.500
0.158
0.250
0.371
0.400
0.177
1.37e-14
1.000
1.43e-14
1.000
1.000
0.022
1.43e-14
1.000
Coeff.
-5.417
-2.167
-11.50
0.500
-6.000
4.000
-4.167
-8.650
-10.50
-1.000
3.500
-4.750
-4.357
3.500
11.50
-15.50
5.750
-2.500
-5.500
-5.611
-3.500
-5.585
3.000
-2.000
-0.389
-2.500
-4.000
2.500
-3.500
-2.500
12.50
-4.833
-3.500
-1.500
1.500
1.500
0.500
-9.500
-10.00
-0.500
4.500
1.214
-5.577
2.500
7.500
12.50
-1.500
-1.750
3.700
-8.500
-7.500
-5.500
-13.50
Observations
F-Test
p value
Pseudo R ²
244
1.178
0.213
0.247
243
0.915
0.641
0.204
Note: P/C = Parent
or Caregiver;
reference group is
Tract 1.0200
bold = p <.05
Parent's age at time of
DHA move-in
P value
0.433
0.793
0.300
0.964
0.507
0.658
0.541
0.198
0.246
0.898
0.752
0.544
0.548
0.699
0.300
0.163
0.463
0.821
0.543
0.405
0.672
0.393
0.740
0.798
0.956
0.821
0.588
0.782
0.699
0.762
0.260
0.558
0.752
0.856
0.892
0.892
0.964
0.391
0.269
0.956
0.685
0.867
0.417
0.821
0.498
0.260
0.868
0.807
0.625
0.443
0.498
0.620
0.224
P/C African
American (1=yes;
0=no)
Coeff.
0.250
3.63e-15
3.48e-15
1.000
4.86e-15
3.55e-15
0.133
0.250
4.95e-15
4.77e-15
4.46e-15
0.500
0.143
5.09e-15
6.17e-15
6.10e-15
0.250
1.000
1.000
0.833
2.69e-15
0.553
4.92e-15
1.000
0.778
1.000
0.714
1.000
1.000
0.667
1.000
1.000
1.000
1.000
1.000
1.000
3.37e-15
1.000
1.000
1.000
1.000
3.94e-15
0.538
5.87e-15
6.03e-15
5.90e-15
1.000
5.70e-15
0.200
1.000
1.000
1.000
5.95e-15
244
3.164
4.14e-09
0.469
P value
0.427
1.000
1.000
0.049
1.000
1.000
0.667
0.413
1.000
1.000
1.000
0.162
0.665
1.000
1.000
1.000
0.484
0.049
0.016
0.007
1.000
0.064
1.000
0.006
0.017
0.049
0.032
0.016
0.016
0.077
0.049
0.008
0.049
0.008
0.049
0.049
1.000
0.049
0.016
0.016
0.049
1.000
0.086
1.000
1.000
1.000
0.016
1.000
0.562
0.049
0.049
0.049
1.000
Parent have HS diploma
at time of DHA move-in
(1=yes; 0=no)
Coeff.
0.167
0.333
3.52e-15
1.000
1.97e-15
3.54e-15
0.400
0.250
1.52e-15
0.500
1.61e-15
0.250
0.143
0.500
1.000
1.000
0.750
1.000
1.23e-15
0.389
5.80e-16
0.213
0.500
0.750
0.222
3.21e-15
0.143
0.500
0.500
1.000*
3.30e-15
0.333
2.52e-15
0.667
2.87e-15
3.13e-15
3.23e-15
3.31e-15
3.36e-15
0.500
3.43e-15
0.143
0.308
3.50e-15
3.65e-15
1.000
0.500
0.375
0.200
2.57e-15
3.64e-15
3.14e-15
1.000
244
1.044
0.406
0.226
P value
0.631
0.421
1.000
0.073
1.000
1.000
0.242
0.458
1.000
0.204
1.000
0.525
0.695
0.271
0.073
0.073
0.057
0.073
1.000
0.251
1.000
0.516
0.271
0.057
0.531
1.000
0.695
0.271
0.271
0.017
1.000
0.421
1.000
0.109
1.000
1.000
1.000
1.000
1.000
0.271
1.000
0.695
0.372
1.000
1.000
0.073
0.271
0.296
0.598
1.000
1.000
1.000
0.073
Parent have any
higher education at
time of DHA move-in
(1=yes; 0=no)
Coeff.
-1.93e-14
0.333
-1.97e-14
-1.97e-14
-1.94e-14
-1.97e-14
0.133
0.1000
-1.94e-14
0.250
1.000
0.250
0.143
-1.94e-14
-1.98e-14
-1.98e-14
-1.98e-14
-1.98e-14
-1.92e-14
0.111
0.333
0.0851
-1.94e-14
-1.96e-14
0.111
1.000
0.143
0.500
0.500
-1.94e-14
-1.97e-14
0.333
-1.95e-14
-1.94e-14
1.000
-1.96e-14
-1.96e-14
-1.96e-14
-1.97e-14
-1.96e-14
-1.97e-14
0.286
-1.92e-14
-1.97e-14
-1.98e-14
-1.98e-14
-1.95e-14
-1.96e-14
-1.93e-14
-1.95e-14
1.000
-1.97e-14
-1.97e-14
244
1.173
0.219
0.247
P value
1.000
0.230
1.000
1.000
1.000
1.000
0.560
0.657
1.000
0.343
0.008
0.343
0.558
1.000
1.000
1.000
1.000
1.000
1.000
0.624
0.230
0.698
1.000
1.000
0.640
0.008
0.558
0.101
0.101
1.000
1.000
0.230
1.000
1.000
0.008
1.000
1.000
1.000
1.000
1.000
1.000
0.242
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
0.008
1.000
1.000
Kids share same
biological dad (1=yes;
0=no)
Coeff.
-0.333
0.167
-0.500
-0.500
0.500
-0.500
0.0333
-0.200
-1.09e-15
-1.29e-15
-0.500
-0.250
0.0714
-1.28e-15
-0.500
-0.500
-1.46e-15
-0.500
-0.500
-0.278
-0.167
-0.138
-0.500
0.500
-0.0556
-0.500
0.0714
-0.500
-0.500
-0.167
0.500
0.167
0.500
-0.167
0.500
0.500
0.500
-0.500
-0.500
-1.17e-15
-0.500
0.0714
-0.192
0.500
-0.500
-0.500
-1.06e-15
-0.125
-0.100
-0.500
0.500
-0.500
0.500
244
0.964
0.549
0.212
P value
0.373
0.709
0.404
0.404
0.307
0.307
0.928
0.582
1.000
1.000
0.404
0.555
0.856
1.000
0.404
0.404
1.000
0.404
0.307
0.447
0.709
0.695
0.307
0.239
0.885
0.404
0.856
0.307
0.307
0.709
0.404
0.709
0.404
0.709
0.404
0.404
0.404
0.404
0.307
1.000
0.404
0.856
0.605
0.404
0.404
0.404
1.000
0.747
0.807
0.404
0.404
0.404
0.404
Parent Depressive
Symptomatology
Scale at time of
interview
Coeff.
7.750
10.33
5.000
3.000
-3.000
14.50
11.27
6.150
4.500
-3.750
27.00
14.25
1.429
12.50
8.000
7.000
1.250
1.000
4.000
4.167
2.333
5.915
4.500
8.250
5.889
-3.000
4.143
7.500
14.00
3.000
14.00
3.333
-3.000
-1.667
25.00
8.000
-5.000
11.00
4.500
-2.000
-4.000
2.429
4.923
13.00
4.000
5.000
-3.500
5.375
12.40
15.00
-1.000
-3.000
-5.000
244
1.045
0.404
0.226
P value
0.271
0.220
0.658
0.790
0.745
0.117
0.105
0.368
0.625
0.638
0.017
0.075
0.847
0.176
0.478
0.535
0.875
0.929
0.664
0.544
0.781
0.374
0.625
0.302
0.414
0.790
0.575
0.416
0.130
0.721
0.215
0.692
0.790
0.843
0.028
0.478
0.658
0.330
0.625
0.828
0.723
0.742
0.482
0.250
0.723
0.658
0.704
0.461
0.109
0.185
0.929
0.790
0.658
Parenting Efficacy
Scale at time of
interview
Coeff.
-0.750
1.667
-12.00
-3.000
0.500
0.500
-1.800
-0.600
9.63e-14
1.000
-1.000
-4.250
0.857
1.500
1.000
1.000
0.750
2.000
-4.000
-0.611
-1.667
-1.489
-2.500
-0.250
0.222
2.000
-3.286
-6.000
-5.000
-0.667
-2.000
-2.667
1.000
-0.667
-3.000
2.000
2.000
-8.000
-2.000
2.000
-3.000
-1.000
-1.308
-6.000
9.45e-14
2.000
1.500
-0.250
-2.400
2.000
-8.000
1.000
-6.000
244
1.033
0.426
0.224
P value
0.789
0.619
0.008
0.505
0.892
0.892
0.515
0.825
1.000
0.753
0.824
0.182
0.771
0.683
0.824
0.824
0.813
0.656
0.276
0.823
0.619
0.574
0.496
0.937
0.938
0.656
0.265
0.103
0.174
0.842
0.656
0.426
0.824
0.842
0.505
0.656
0.656
0.076
0.586
0.586
0.505
0.734
0.639
0.183
1.000
0.656
0.683
0.931
0.435
0.656
0.076
0.824
0.183
Parenting Beliefs
Scale at time of
interview
Coeff.
-1.750
1.000
-6.000
-1.000
-0.500
-0.500
0.933
-1.050
-1.500
0.250
-4.000
-1.000
1.143
-4.500
-2.000
-1.000
-4.250
-3.000
-3.500
-1.778
-0.333
-0.851
3.000
-5.750
-3.333
1.000
-5.286
-3.500
-3.000
-3.000
-2.000
-1.667
-4.000
-1.000
-4.000
-1.000
-12.00
-5.000
3.000
2.000
1.15e-13
0.571
-0.923
-5.000
2.000
1.11e-13
1.000
-0.750
-2.000
-1.000
3.000
3.000
1.000
244
1.066
0.369
0.229
P value
0.561
0.781
0.214
0.836
0.899
0.899
0.753
0.719
0.703
0.942
0.407
0.769
0.717
0.254
0.678
0.836
0.214
0.534
0.375
0.545
0.926
0.765
0.446
0.093
0.280
0.836
0.095
0.375
0.446
0.404
0.678
0.643
0.407
0.781
0.407
0.836
0.014
0.301
0.446
0.612
1.000
0.856
0.758
0.301
0.678
1.000
0.800
0.810
0.544
0.836
0.534
0.534
0.836
Exhibit A-2C. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 3+ Children
P/C is single parent
(1=yes, 0=no)
Census Tract
3.0100
5.0200
6.0000
7.0200
8.0000
9.0200
9.0300
9.0400
9.0500
10.0000
11.0100
11.0200
13.0100
14.0200
14.0300
15.0000
16.0000
19.0000
21.0000
23.0000
24.0300
35.0000
36.0100
36.0200
41.0100
41.0300
41.0400
42.0200
43.0400
44.0300
44.0400
45.0100
45.0200
46.0200
47.0000
48.0200
50.0100
2.0000
54.0000
55.0300
68.0900
83.1100
83.1200
119.0200
Coeff.
0.389
-0.611
0.389
0.0812
-0.0111
0.389
-0.111
0.139
0.389
0.264
0.189
0.389
0.389
0.103
-0.611
-0.111
0.158
0.124
0.389
0.389
0.189
0.189
0.389
0.0253
0.389
-0.111
0.389
0.389
0.389
0.389
0.389
-0.278
0.0556
0.389
0.389
0.389
-0.611
-0.611
0.0556
0.389
0.389
-0.111
0.389
0.389
Observations
F-Test
p value
Pseudo R ²
Note: P/C = Parent
or Caregiver;
reference group is
Tract 2.0200
bold = p <.05
203
0.647
0.954
0.153
P value
0.268
0.206
0.421
0.635
0.946
0.421
0.751
0.593
0.268
0.187
0.427
0.268
0.421
0.622
0.206
0.669
0.356
0.365
0.421
0.268
0.427
0.427
0.421
0.888
0.136
0.751
0.268
0.421
0.421
0.421
0.421
0.344
0.772
0.421
0.421
0.421
0.206
0.206
0.772
0.421
0.421
0.751
0.421
0.421
P/C employment
status at time of
DHA move-in
(1=employed, 0=not
employed)
Coeff.
0.556
0.556
-0.444
0.325
0.222
0.556
0.0556
0.306
0.556
0.0556
-0.0444
0.0556
0.556
0.413
-0.444
-0.194
-0.0598
-0.150
0.556
0.556
-2.044
0.356
0.556
-0.0808
0.306
0.556
0.556
-0.444
-0.444
-0.444
-0.444
-0.444
-1.111
-0.444
-0.444
0.556
-0.444
0.556
0.333
0.556
0.556
0.0556
0.556
0.556
203
0.628
0.964
0.149
P value
0.571
0.681
0.742
0.498
0.629
0.681
0.955
0.674
0.571
0.921
0.947
0.955
0.681
0.481
0.742
0.789
0.900
0.695
0.681
0.571
0.002
0.593
0.681
0.872
0.674
0.571
0.571
0.742
0.742
0.742
0.742
0.588
0.040
0.742
0.742
0.681
0.742
0.681
0.535
0.681
0.681
0.955
0.681
0.681
P/C hourly wage at
time of DHA move-in
Coeff.
8.686
6.401
-7.849
5.418
4.327
11.34
2.401
5.526
15.15
0.601
0.351
1.651
14.15
9.794
-7.849
-3.811
-1.580
2.824
16.15
10.15
-4.099
4.001
18.15
-0.548
2.929
17.15
12.90
-7.849
-7.849
-7.849
-7.849
-7.849
-1.768
-7.849
-7.849
10.25
-7.849
15.15
4.568
15.15
12.15
1.151
8.151
16.15
203
1.476
0.0433
0.291
P value
0.205
0.497
0.406
0.106
0.179
0.230
0.726
0.277
0.028
0.877
0.940
0.809
0.135
0.018
0.406
0.453
0.636
0.292
0.088
0.139
0.377
0.389
0.056
0.876
0.564
0.013
0.061
0.406
0.406
0.406
0.406
0.171
0.637
0.406
0.406
0.278
0.406
0.109
0.224
0.109
0.199
0.866
0.388
0.088
P/C disability status
at time of survey
(1=yes; 0=no)
Coeff.
-0.167
-0.167
-0.167
-0.0128
-0.167
-0.167
-0.167
-0.167
0.333
-0.167
-0.167
-0.167
0.833
-0.0238
-0.167
0.333
-0.0897
-0.0784
-0.167
-0.167
-0.167
0.0333
-0.167
0.0152
-0.167
-0.167
-0.167
-0.167
-0.167
-0.167
-0.167
-0.167
-0.0556
-0.167
-0.167
-0.167
-0.167
-0.167
-0.0556
-0.167
-0.167
-0.167
-0.167
-0.167
203
0.697
0.919
0.163
P value
0.460
0.592
0.592
0.907
0.117
0.592
0.460
0.320
0.141
0.196
0.277
0.460
0.008
0.860
0.592
0.048
0.416
0.375
0.592
0.460
0.277
0.828
0.592
0.896
0.320
0.460
0.460
0.592
0.592
0.592
0.592
0.378
0.653
0.592
0.592
0.592
0.592
0.592
0.653
0.592
0.592
0.460
0.592
0.592
P/C received TANF
at time of DHA movein (1=yes, 0=no)
Coeff.
0.111
0.611
0.611
0.150
-1.122
-0.389
-0.389
-0.389
0.611
0.111
0.211
0.611
0.611
0.183
0.611
-0.139
-0.697
-0.0948
0.611
-0.389
-0.189
0.211
-0.389
0.0657
0.111
-0.389
-0.389
0.611
0.611
0.611
-0.389
-0.389
-0.167
0.611
0.611
0.611
-0.389
-0.389
-0.167
0.611
-0.389
0.111
-0.389
-0.389
203
0.332
1.000
0.0847
P value
0.922
0.696
0.696
0.787
0.036
0.804
0.732
0.644
0.591
0.864
0.784
0.591
0.696
0.788
0.696
0.869
0.210
0.831
0.696
0.732
0.806
0.784
0.804
0.910
0.895
0.732
0.732
0.696
0.696
0.696
0.804
0.682
0.789
0.696
0.696
0.696
0.804
0.804
0.789
0.696
0.804
0.922
0.804
0.804
P/C receiving Food
Stamps at time of
DHA move-in
(1=yes, 0=no)
Coeff.
0.444
0.444
0.444
0.291
0.178
0.444
-0.556
-0.306
0.444
0.0694
0.244
0.444
0.444
0.159
0.444
-0.0556
0.0598
-0.232
-0.556
-0.0556
0.244
0.244
0.444
0.0808
-0.0556
-0.556
-0.0556
0.444
0.444
-0.556
-0.556
0.111
-0.111
0.444
-0.556
0.444
0.444
-0.556
7.63e-16
0.444
0.444
-0.0556
-0.556
-0.556
203
0.202
1.000
0.0532
P value
0.612
0.713
0.713
0.497
0.665
0.713
0.526
0.638
0.612
0.889
0.681
0.612
0.713
0.762
0.713
0.932
0.889
0.498
0.645
0.949
0.681
0.681
0.713
0.857
0.932
0.526
0.949
0.713
0.713
0.645
0.645
0.879
0.817
0.713
0.645
0.713
0.713
0.645
1.000
0.713
0.713
0.949
0.645
0.645
P/C had checking
account at time of
DHA move-in
(1=yes, 0=no)
Coeff.
-0.167
-0.167
-0.167
0.141
0.433
-0.167
0.333
0.583
0.333
0.0833
-0.167
-0.167
0.833
0.405
-0.167
0.0833
-0.782
-0.0784
0.833
0.833
0.0333
0.0333
-0.167
0.379
0.0833
0.833
0.833
-0.167
0.833
-0.167
0.833
-0.167
-1.056
-0.167
0.833
0.833
0.833
0.833
0.611
0.833
0.833
0.333
-0.167
0.833
203
0.549
0.989
0.133
P value
0.864
0.901
0.901
0.766
0.341
0.901
0.731
0.418
0.731
0.880
0.800
0.864
0.533
0.485
0.901
0.908
0.100
0.836
0.533
0.391
0.960
0.960
0.901
0.447
0.908
0.391
0.391
0.901
0.533
0.901
0.533
0.837
0.048
0.901
0.533
0.533
0.533
0.533
0.251
0.533
0.533
0.731
0.901
0.533
P/C had health
insurance at time of
DHA move-in (1=yes,
0=no)
Coeff.
0.278
0.278
0.278
0.0470
-0.122
0.278
-0.722
0.278
-0.722
0.0278
0.278
0.278
0.278
0.135
0.278
-0.472
-0.261
-0.310
0.278
-0.222
-0.122
0.0778
0.278
0.00505
-0.222
0.278
-0.222
0.278
0.278
-0.722
0.278
-0.0556
-1.278
0.278
0.278
0.278
0.278
-0.722
0.0556
0.278
0.278
-0.722
0.278
0.278
203
0.424
0.999
0.106
P value
0.744
0.813
0.813
0.910
0.759
0.813
0.396
0.659
0.396
0.954
0.630
0.744
0.813
0.791
0.813
0.454
0.530
0.351
0.813
0.794
0.832
0.893
0.813
0.991
0.724
0.744
0.794
0.813
0.813
0.538
0.813
0.938
0.007
0.813
0.813
0.813
0.813
0.538
0.905
0.813
0.813
0.396
0.813
0.813
Exhibit A-2C. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 3+ Children (cont.)
P/C had too little
money for food at
time of DHA move-in
(1=yes, 0=no)
Census Tract
3.0100
5.0200
6.0000
7.0200
8.0000
9.0200
9.0300
9.0400
9.0500
10.0000
11.0100
11.0200
13.0100
14.0200
14.0300
15.0000
16.0000
19.0000
21.0000
23.0000
24.0300
35.0000
36.0100
36.0200
41.0100
41.0300
41.0400
42.0200
43.0400
44.0300
44.0400
45.0100
45.0200
46.0200
47.0000
48.0200
50.0100
2.0000
54.0000
55.0300
68.0900
83.1100
83.1200
119.0200
Coeff.
0.722
-0.278
-0.278
0.261
0.389
0.722
0.222
-0.278
0.222
0.472
0.522
-0.278
0.722
-0.135
-0.278
-0.0278
0.261
-0.278
0.722
0.222
0.122
0.122
-0.278
-0.00505
0.222
-0.278
0.222
-0.278
-0.278
-0.278
-0.278
-0.278
-1.056
0.722
-0.278
-0.278
0.722
0.722
0.0556
-0.278
0.722
0.222
0.722
-0.278
Observations
F-Test
p value
Pseudo R ²
203
0.511
0.995
0.124
Note: P/C = Parent
or Caregiver;
reference group is
Tract 2.0200
bold = p <.05
P value
0.384
0.808
0.808
0.520
0.318
0.528
0.789
0.652
0.789
0.319
0.354
0.738
0.528
0.785
0.808
0.964
0.520
0.392
0.528
0.789
0.828
0.828
0.808
0.991
0.718
0.738
0.789
0.808
0.808
0.808
0.808
0.689
0.021
0.528
0.808
0.808
0.528
0.528
0.903
0.808
0.528
0.789
0.528
0.808
P/C had difficulty
paying all bills at
time of DHA movein (1=yes, 0=no)
Coeff.
1.111
0.111
1.111
0.803
0.511
1.111
0.611
0.111
0.611
0.611
0.311
0.611
1.111
0.111
1.111
-4.139
-0.274
-0.0654
0.111
0.611
0.311
0.711
1.111
0.657
0.611
0.111
0.611
0.111
0.111
-8.889
0.111
-2.889
-1.667
1.111
0.111
0.111
1.111
0.111
0.556
0.111
1.111
0.111
1.111
0.111
203
1.359
0.0887
0.275
P value
0.461
0.957
0.593
0.276
0.470
0.593
0.685
0.921
0.685
0.477
0.761
0.685
0.593
0.902
0.593
0.000
0.710
0.912
0.957
0.685
0.761
0.487
0.593
0.397
0.585
0.941
0.685
0.957
0.957
0.000
0.957
0.023
0.045
0.593
0.957
0.957
0.593
0.957
0.501
0.957
0.593
0.941
0.593
0.957
Frequency that P/C
drank alcohol since
becoming a parent
Coeff.
0.889
-0.111
-0.111
-0.419
0.222
1.889
0.389
2.639
-0.111
1.139
0.489
2.889
1.889
0.0317
-0.111
1.639
-0.188
0.0948
-0.111
-0.111
0.689
0.289
-0.111
-0.202
-2.111
-0.111
0.389
-0.111
-0.111
0.889
-0.111
-3.111
1.889
-0.111
-9.111
0.889
1.889
-0.111
-0.444
-0.111
0.889
0.389
0.889
1.889
203
0.876
0.690
0.196
P value
0.641
0.966
0.966
0.653
0.804
0.473
0.838
0.064
0.954
0.296
0.706
0.131
0.473
0.978
0.966
0.248
0.840
0.899
0.966
0.954
0.595
0.823
0.966
0.837
0.137
0.954
0.838
0.966
0.966
0.735
0.966
0.053
0.072
0.966
0.001
0.735
0.473
0.966
0.671
0.966
0.735
0.838
0.735
0.473
Frequency that P/C
smoked marijuana
since becoming a
parent
Coeff.
-0.833
-0.833
-0.833
-1.372
-0.633
-0.833
-0.833
0.417
-0.833
-1.208
0.167
2.667
-9.833
-0.833
-0.833
-2.833
-1.218
-1.010
-0.833
-0.833
-0.633
0.567
-0.833
-1.015
0.167
-0.833
-0.333
-0.833
-0.833
2.167
-0.833
-3.833
0.389
-0.833
-0.833
-0.833
-9.833
-0.833
-1.056
-0.833
0.167
-0.333
0.167
0.167
203
1.061
0.386
0.228
P value
0.661
0.751
0.751
0.141
0.478
0.751
0.661
0.768
0.661
0.266
0.897
0.162
0.000
0.464
0.751
0.046
0.191
0.176
0.751
0.661
0.623
0.660
0.751
0.299
0.906
0.661
0.861
0.751
0.751
0.409
0.751
0.017
0.709
0.751
0.751
0.751
0.000
0.751
0.312
0.751
0.949
0.861
0.949
0.949
P/C ever seen a
psychiatrist (1=yes,
0=no)
Coeff.
-1.77e-15
-0.500
-0.500
-0.269
-0.100
-0.500
-1.85e-15
-0.500
-0.500
-1.79e-15
-0.100
-1.59e-15
0.500
-0.0714
0.500
0.250
-0.269
-1.50e-15
0.500
-0.500
-0.300
0.1000
0.500
-0.136
-1.59e-15
-1.63e-15
-0.500
-0.500
0.500
-0.500
-0.500
-0.167
-0.167
-0.500
-0.500
-0.500
-0.500
-0.500
-0.0556
-0.500
0.500
0.500
0.500
-0.500
203
0.855
0.723
0.192
P value
1.000
0.331
0.331
0.140
0.567
0.331
1.000
0.072
0.181
1.000
0.692
1.000
0.331
0.748
0.331
0.366
0.140
1.000
0.331
0.181
0.236
0.692
0.331
0.476
1.000
1.000
0.181
0.331
0.331
0.331
0.331
0.593
0.414
0.331
0.331
0.331
0.331
0.331
0.785
0.331
0.331
0.181
0.331
0.331
Number of years
Number of years
during childhood that during childhood that P/C born in the United
P/C lived in public
P/C lived in a home States (1=yes; 0=no)
housing
owned by parents
Coeff.
1.056
21.56
-5.444
-0.291
-2.911
-5.444
1.056
1.056
5.556
1.056
-5.444
2.056
-5.444
-5.444
15.56
4.056
2.940
-0.121
-5.444
-5.444
3.156
-3.044
-5.444
-2.263
-5.444
-5.444
-5.444
-5.444
-5.444
-5.444
-5.444
-5.444
-1.000
-5.444
18.56
6.556
-5.444
-5.444
0.222
-5.444
-5.444
-5.444
-5.444
-5.444
203
0.920
0.617
0.204
P value
0.865
0.012
0.524
0.923
0.317
0.524
0.865
0.818
0.370
0.765
0.196
0.740
0.524
0.143
0.070
0.378
0.332
0.960
0.524
0.380
0.453
0.469
0.524
0.477
0.237
0.380
0.380
0.524
0.524
0.524
0.524
0.294
0.768
0.524
0.031
0.443
0.524
0.524
0.948
0.524
0.524
0.380
0.524
0.524
Coeff.
0.333
-12.67
12.33
2.333
-1.733
14.33
8.333
-5.917
-7.667
-7.167
-3.467
-12.67
14.33
3.048
-12.67
6.333
-2.359
-1.196
-12.67
14.33
-1.067
-3.467
6.333
1.061
-5.917
14.33
0.833
14.33
-12.67
6.333
14.33
5.333
1.778
-12.67
-12.67
-12.67
14.33
14.33
7.667
14.33
-3.667
13.33
14.33
14.33
203
1.153
0.261
0.243
P value
0.969
0.292
0.304
0.583
0.671
0.233
0.339
0.360
0.379
0.150
0.557
0.147
0.233
0.558
0.292
0.327
0.579
0.725
0.292
0.101
0.857
0.557
0.597
0.812
0.360
0.101
0.924
0.233
0.292
0.597
0.233
0.464
0.709
0.292
0.292
0.292
0.233
0.233
0.109
0.233
0.760
0.127
0.233
0.233
Coeff.
-0.222
0.278
-0.722
0.0470
0.144
-0.722
0.278
0.278
0.278
0.153
0.278
0.278
0.278
-0.151
0.278
0.278
0.278
0.131
-0.722
-0.222
0.278
0.278
0.278
0.0960
0.278
-0.222
0.278
0.278
0.278
0.278
0.278
-0.389
0.167
0.278
0.278
0.278
0.278
-0.722
0.278
0.278
0.278
-0.222
0.278
0.278
203
1.323
0.109
0.269
P value
0.400
0.445
0.048
0.715
0.244
0.048
0.293
0.157
0.293
0.310
0.122
0.293
0.445
0.339
0.445
0.157
0.032
0.206
0.048
0.400
0.122
0.122
0.445
0.479
0.157
0.400
0.293
0.445
0.445
0.445
0.445
0.079
0.249
0.445
0.445
0.445
0.445
0.048
0.056
0.445
0.445
0.400
0.445
0.445
Spanish language
interview (1=yes;
0=no)
Coeff.
0.278
-0.222
-0.222
-0.145
-0.222
-0.222
-0.222
-0.222
-0.222
-0.0972
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.134
-0.222
-0.222
-0.222
-0.222
-0.222
-0.131
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
0.444
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
-0.222
0.278
-0.222
-0.222
203
0.989
0.500
0.216
P value
0.145
0.396
0.396
0.119
0.013
0.396
0.243
0.116
0.243
0.370
0.086
0.243
0.396
0.052
0.396
0.116
0.018
0.073
0.396
0.243
0.086
0.086
0.396
0.179
0.116
0.243
0.243
0.396
0.396
0.396
0.396
0.006
0.034
0.396
0.396
0.396
0.396
0.396
0.034
0.396
0.396
0.145
0.396
0.396
Exhibit A-2C. Relationships Between DHA Resident Characteristics and Census Tracts: Households with 3+ Children (cont.)
Biological father
always lived in
household with
child(ren) (1=yes;
0=no)
Census Tract
3.0100
5.0200
6.0000
7.0200
8.0000
9.0200
9.0300
9.0400
9.0500
10.0000
11.0100
11.0200
13.0100
14.0200
14.0300
15.0000
16.0000
19.0000
21.0000
23.0000
24.0300
35.0000
36.0100
36.0200
41.0100
41.0300
41.0400
42.0200
43.0400
44.0300
44.0400
45.0100
45.0200
46.0200
47.0000
48.0200
50.0100
2.0000
54.0000
55.0300
68.0900
83.1100
83.1200
119.0200
Coeff.
0.833
-0.167
-0.167
0.0641
0.233
-0.167
0.333
0.333
-0.167
0.0833
0.0333
-0.167
-0.167
0.262
0.833
0.333
0.0641
0.0392
-0.167
-0.167
-0.167
0.0333
-0.167
0.0152
0.0833
0.333
-0.167
0.833
-0.167
-0.167
-0.167
0.500
0.167
0.833
-0.167
0.833
-0.167
0.833
-0.0556
0.833
0.833
0.333
0.833
0.833
Observations
F-Test
p value
Pseudo R ²
203
1.179
0.231
0.247
Note: P/C = Parent
or Caregiver;
reference group is
Tract 2.0200
bold = p <.05
P value
0.012
0.713
0.713
0.689
0.131
0.713
0.311
0.172
0.612
0.656
0.881
0.612
0.713
0.183
0.067
0.172
0.689
0.760
0.713
0.612
0.454
0.881
0.713
0.928
0.732
0.311
0.612
0.067
0.713
0.713
0.713
0.070
0.354
0.067
0.713
0.067
0.713
0.067
0.757
0.067
0.067
0.311
0.067
0.067
Parent have HS
Parent have any
Parent's age at time P/C African American
diploma at time of
higher education at
of DHA move-in
(1=yes; 0=no)
DHA move-in (1=yes; time of DHA move-in
0=no)
(1=yes; 0=no)
Coeff.
0.833
7.333
1.333
2.872
2.200
7.333
5.833
-2.167
9.333
6.083
-0.667
1.833
1.333
4.905
7.333
4.583
-3.205
-0.225
11.33
-0.167
-2.067
-2.667
6.333
3.788
5.333
8.333
-1.167
-2.667
9.333
1.333
7.333
5.000
-4.667
-1.667
-0.667
2.333
2.333
8.333
4.667
11.33
1.333
-18.17
2.333
2.333
203
0.988
0.502
0.216
P value
0.895
0.399
0.878
0.351
0.457
0.399
0.355
0.643
0.140
0.092
0.876
0.771
0.878
0.194
0.399
0.328
0.299
0.927
0.193
0.979
0.629
0.533
0.466
0.243
0.255
0.187
0.853
0.759
0.284
0.878
0.399
0.344
0.178
0.848
0.939
0.788
0.788
0.338
0.178
0.193
0.878
0.004
0.788
0.788
Coeff.
-0.111
-0.111
-0.111
0.274
0.489
-0.111
-0.111
-0.111
-0.111
0.264
-0.111
-0.111
-0.111
0.175
-0.111
0.889
0.889
0.418
-0.111
-0.111
0.889
0.489
0.889
0.434
0.889
0.889
0.889
0.889
0.889
0.889
0.889
-0.111
0.111
-0.111
-0.111
0.889
0.889
0.889
-4.56e-15
-0.111
0.889
0.389
0.889
-0.111
203
3.075
0.000000153
0.461
P value
0.719
0.794
0.794
0.072
0.001
0.794
0.719
0.628
0.719
0.136
0.596
0.719
0.794
0.345
0.794
0.000
0.000
0.001
0.794
0.719
0.000
0.021
0.038
0.007
0.000
0.005
0.005
0.038
0.038
0.038
0.038
0.668
0.512
0.794
0.794
0.038
0.038
0.038
1.000
0.794
0.038
0.210
0.038
0.794
Coeff.
0.167
-0.333
-0.333
-0.0256
4.61e-16
0.667
0.167
0.167
-0.333
0.167
0.0667
0.167
-0.333
0.0952
-0.333
0.167
0.0513
-0.0392
-0.333
-0.333
0.0667
-0.133
0.667
0.0303
0.167
0.167
0.167
0.667
0.667
-0.333
-0.333
-0.333
4.64e-16
-0.333
0.667
-0.333
-0.333
0.667
0.111
-0.333
-0.333
0.167
-0.333
-0.333
203
0.562
0.986
0.135
P value
0.657
0.520
0.520
0.889
1.000
0.199
0.657
0.549
0.375
0.436
0.793
0.657
0.520
0.671
0.520
0.549
0.780
0.789
0.520
0.375
0.793
0.601
0.199
0.875
0.549
0.657
0.657
0.199
0.199
0.520
0.520
0.289
1.000
0.520
0.199
0.520
0.520
0.199
0.589
0.520
0.520
0.657
0.520
0.520
Coeff.
-0.0556
-0.0556
-0.0556
0.0983
-0.0556
-0.0556
-0.0556
0.444
-0.0556
-0.0556
-0.0556
0.444
-0.0556
-0.0556
-0.0556
-0.0556
-0.0556
0.0327
-0.0556
0.444
-0.0556
0.144
-0.0556
-0.0556
-0.0556
0.444
-0.0556
-0.0556
-0.0556
-0.0556
0.944
0.278
-0.0556
-0.0556
-0.0556
-0.0556
-0.0556
-0.0556
-0.0556
0.944
-0.0556
-0.0556
0.944
-0.0556
203
2.076
0.000555
0.366
P value
0.760
0.824
0.824
0.268
0.514
0.824
0.760
0.001
0.760
0.591
0.652
0.015
0.824
0.609
0.824
0.680
0.531
0.645
0.824
0.015
0.652
0.242
0.824
0.551
0.680
0.015
0.760
0.824
0.824
0.824
0.000
0.069
0.576
0.824
0.824
0.824
0.824
0.824
0.576
0.000
0.824
0.760
0.000
0.824
Kids share same
biological dad
(1=yes; 0=no)
Coeff.
-0.389
-0.389
-0.389
-0.0812
-0.122
-0.389
-0.389
-0.139
-0.389
-0.264
-0.189
0.111
0.611
0.468
0.611
-0.139
-0.158
-0.00654
-0.389
-0.389
-0.389
-0.389
0.611
-0.0253
0.111
0.111
-0.389
-0.389
-0.389
-0.389
0.611
0.278
0.0556
-0.389
0.611
-0.389
-0.389
0.611
-0.167
0.611
0.611
-0.389
-0.389
-0.389
203
1.168
0.242
0.245
P value
0.258
0.411
0.411
0.628
0.448
0.411
0.258
0.585
0.258
0.178
0.417
0.746
0.197
0.023
0.197
0.585
0.346
0.961
0.411
0.258
0.096
0.096
0.197
0.886
0.662
0.746
0.258
0.411
0.411
0.411
0.197
0.334
0.767
0.411
0.197
0.411
0.411
0.197
0.375
0.197
0.197
0.258
0.411
0.411
Parent Depressive
Symptomatology
Scale at time of
interview
Coeff.
-2.222
-3.222
18.78
-3.607
1.844
-8.222
-2.222
-5.222
-2.222
4.778
-0.622
17.28
13.78
-3.937
1.778
9.028
0.932
-0.399
10.78
-6.222
-0.222
3.778
-4.222
0.141
-6.222
-9.722
4.278
-3.222
3.778
-7.222
-7.222
-5.222
-2.556
0.778
-7.222
-10.22
-4.222
-11.22
3.222
-11.22
18.78
-7.222
-8.222
7.778
203
1.035
0.425
0.224
P value
0.758
0.746
0.060
0.306
0.586
0.408
0.758
0.329
0.758
0.246
0.899
0.018
0.167
0.361
0.858
0.093
0.791
0.888
0.279
0.389
0.964
0.440
0.671
0.970
0.245
0.179
0.553
0.746
0.704
0.468
0.468
0.387
0.518
0.938
0.468
0.304
0.671
0.260
0.415
0.260
0.060
0.317
0.408
0.434
Parenting Efficacy
Scale at time of
interview
Coeff.
1.333
3.333
-7.667
0.0256
-0.333
-0.667
2.333
0.0833
1.333
-1.667
2.133
0.333
-4.667
1.190
-0.667
-0.667
0.872
0.627
-2.667
1.833
0.933
0.733
-2.667
0.333
1.083
3.333
-0.667
3.333
3.333
-2.667
1.333
2.667
0.444
-9.667
1.333
2.333
1.333
3.333
-0.444
3.333
-5.667
-0.167
-2.667
3.333
203
1.021
0.448
0.221
P value
0.578
0.313
0.021
0.982
0.767
0.840
0.330
0.963
0.578
0.223
0.190
0.889
0.159
0.406
0.840
0.707
0.456
0.503
0.419
0.444
0.566
0.652
0.419
0.786
0.542
0.165
0.781
0.313
0.313
0.419
0.686
0.184
0.735
0.004
0.686
0.480
0.686
0.313
0.735
0.313
0.087
0.944
0.419
0.313
Parenting Beliefs
Scale at time of
interview
Coeff.
1.444
-0.0556
-1.056
0.406
0.678
2.944
-4.056
2.194
-0.0556
0.319
2.544
1.944
2.944
1.087
-0.0556
-0.0556
0.0214
0.621
-1.056
1.444
0.544
3.144
1.944
2.308
2.194
-7.556
-5.056
-5.056
4.944
0.944
-5.056
-0.0556
-0.389
0.944
-1.056
2.944
-5.056
-5.056
3.389
4.944
-0.0556
-0.0556
1.944
1.944
203
0.954
0.559
0.210
P value
0.622
0.989
0.794
0.776
0.622
0.466
0.167
0.313
0.985
0.848
0.201
0.507
0.466
0.534
0.989
0.980
0.988
0.588
0.794
0.622
0.784
0.115
0.630
0.126
0.313
0.011
0.085
0.211
0.221
0.815
0.211
0.982
0.808
0.815
0.794
0.466
0.211
0.211
0.036
0.221
0.989
0.985
0.630
0.630
Relationships between Individual Characteristics and Neighborhood Characteristics
Even if (as we have found) there were non-random assignments to DHA developments
or neighborhoods on the basis of ethnic or disability status, it would not necessarily
follow that there would be a strong relationship between these statuses and a wide
variety of neighborhood characteristics. Thus, our third investigative strategy involves
the use of continuously measured neighborhood characteristics instead of dummy
variables to probe their potential systematic co-variation with characteristics of individual
DHA families. Specifically, we employed the same individual characteristics as above
and 12 characteristics of census tracts’ population and housing (percentages of: femaleheaded households, poor families and individuals, unemployed adults, those with only
elementary school education, those with college degrees, employees in
professional/technical occupations, non-Latino African American population, Latino
population, foreign-born population, housing vacancy rate, homes built prior to 1940,
homes that are owner-occupied) conventionally used in neighborhood effect studies.
We employed multivariate regression (again stratified by family size) to estimate the
statistical associations between 27 individual and 12 neighborhood characteristics. As
before, a quasi-random assignment would be reflected in coefficients approximating zero
and an insignificant F-test for the regression as a whole.
Results are shown in Exhibit A-3A-C. Overall, of the 36 regressions, 26 exhibited
insignificant F-tests. More convincingly, of the 972 regression estimates, 900 (92
percent) yielded coefficients that were statistically insignificant. Across the three familysize strata the percentages of insignificant coefficients were 91, 93 and 92, respectively,
suggesting that generally the outcomes corresponded to quasi-random assignment.
Further examination is required, however, to ascertain if there was any systematic
sorting by a particular household characteristic as revealed by that characteristic
garnering the bulk of the statistically significant coefficients.
Exhibit A-3A. Relationships Between DHA Resident and Neighborhood Characteristics: Households with 0-1 Child
Characteristics of DHA Resident
Percent female Percent of residents Percent of residents
Percent foreign born
headed households 25+ with LT 9th 25+ with Bachelor's
in neighborhood
in neighborhood
grade education
degree or more
Neighborhood
poverty rate
Percent Black
residents in
neighborhood
Percent Hispanic
residents in
neighborhood
Neighborhood
unemployment rate
Neighborhood
vacancy rate
Neighborhood Percent homes built
homeownership
before 1940 in
rate
neighborhood
Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value
P/C is single parent (1=yes, 0=no)
P/C employment status at time of DHA move-in (1=employed, 0=not employed)
P/C hourly wage at time of DHA move-in
P/C disability status at time of survey (1=yes; 0=no)
P/C received TANF at time of DHA move-in (1=yes, 0=no)
P/C receiving Food Stamps at time of DHA move-in (1=yes, 0=no)
P/C had checking account at time of DHA move-in (1=yes, 0=no)
P/C had health insurance at time of DHA move-in (1=yes, 0=no)
P/C had too little money for food at time of DHA move-in (1=yes, 0=no)
P/C had difficulty paying all bills at time of DHA move-in (1=yes, 0=no)
Frequency that P/C drank alcohol since becoming a parent
Frequency that P/C smoked marijuana since becoming a parent
P/C ever seen a psychiatrist (1=yes, 0=no)
Number of years during childhood that P/C lived in public housing
Number of years during childhood that P/C lived in a home owned by parents
P/C born in the United States (1=yes; 0=no)
Spanish language interview (1=yes; 0=no)
Biological father always lived in household with child(ren) (1=yes; 0=no)
Parent's age at time of DHA move-in
Year that P/C moved into first randomly assigned unit
P/C African American (1=yes; 0=no)
Parent have HS diploma at time of DHA move-in (1=yes; 0=no)
Parent have any higher education at time of DHA move-in (1=yes; 0=no)
Kids share same biological dad (1=yes; 0=no)
Parent Depressive Symptomatology Scale at time of interview
Parenting Efficacy Scale at time of interview
Parenting Beliefs Scale at time of interview
Note: P/C = Parent or Caregiver; OR = odds ratios for dichotomous outcomes;
all neighborhood characteristics measured at time of first DHA move-in
N=265
bold = p<.05
1.011
0.991
0.100
0.992
1.016
1.002
0.992
0.999
0.996
0.990
-0.036
0.018
0.997
-0.072
-0.025
1.003
0.996
0.986
0.084
0.100
1.004
1.005
0.975
0.986
0.029
-0.130
-0.267
0.160
0.156
0.177
0.396
0.016
0.743
0.222
0.846
0.526
0.119
0.493
0.026
0.650
0.401
0.856
0.791
0.786
0.140
0.678
0.594
0.548
0.419
0.025
0.035
0.847
0.094
0.023
1.045
0.974
-0.111
0.964
1.049
1.012
0.960
0.975
0.993
0.990
0.113
0.011
1.008
-0.188
0.120
1.035
0.977
0.969
0.450
-0.407
0.979
1.007
0.971
0.953
0.355
-0.169
-0.002
0.007
0.067
0.356
0.050
0.003
0.405
0.005
0.139
0.654
0.477
0.158
0.425
0.610
0.149
0.581
0.145
0.485
0.106
0.137
0.149
0.137
0.661
0.160
0.004
0.116
0.179
0.991
0.975
1.005
0.049
1.033
0.969
0.986
1.028
1.012
1.003
1.009
-0.111
-0.002
0.998
0.276
-0.307
0.967
1.005
1.017
-0.332
0.098
1.016
0.993
1.016
1.040
-0.223
0.070
-0.124
0.023
0.652
0.664
0.005
0.017
0.159
0.010
0.333
0.760
0.363
0.131
0.871
0.883
0.010
0.107
0.015
0.827
0.186
0.247
0.722
0.126
0.507
0.260
0.008
0.303
0.561
0.522
1.028
0.993
0.033
1.008
0.970
0.997
1.005
0.978
1.001
1.029
0.033
-0.015
1.008
0.024
0.061
1.032
0.988
0.995
-0.132
0.341
1.005
1.002
1.062
1.033
0.238
0.027
0.154
0.098
0.575
0.733
0.631
0.043
0.807
0.715
0.142
0.946
0.038
0.615
0.163
0.569
0.824
0.721
0.227
0.730
0.774
0.595
0.118
0.692
0.858
0.001
0.038
0.183
0.794
0.340
1.015
0.993
-0.053
0.990
1.026
1.007
0.988
0.996
0.994
0.992
0.003
0.014
0.998
-0.187
0.142
1.001
0.994
0.991
-0.004
-0.149
1.000
1.006
0.977
0.979
0.068
-0.096
-0.028
0.064
0.249
0.486
0.280
0.000
0.311
0.064
0.633
0.377
0.223
0.949
0.095
0.774
0.011
0.290
0.928
0.717
0.353
0.985
0.419
0.990
0.406
0.030
0.003
0.652
0.229
0.834
1.012
1.005
0.014
1.013
1.009
0.998
1.011
1.016
1.005
0.984
-0.015
0.007
1.009
0.058
-0.108
0.998
1.008
0.990
0.057
0.014
1.056
0.992
1.011
0.998
-0.035
-0.043
-0.109
0.276
0.512
0.743
0.195
0.275
0.833
0.190
0.144
0.511
0.080
0.605
0.129
0.250
0.197
0.129
0.876
0.656
0.448
0.603
0.891
0.000
0.365
0.323
0.782
0.669
0.330
0.106
1.012
0.991
0.001
0.982
1.007
1.002
0.986
0.987
0.994
1.003
0.031
-0.005
1.002
-0.061
0.075
1.017
1.000
0.992
0.041
0.001
0.978
1.010
0.990
0.990
0.101
0.008
0.065
0.082
0.162
0.973
0.019
0.293
0.799
0.024
0.092
0.363
0.587
0.272
0.298
0.755
0.183
0.317
0.077
0.985
0.339
0.714
0.989
0.001
0.160
0.264
0.158
0.211
0.860
0.367
1.010
1.013
-0.155
0.953
1.069
1.030
0.992
1.007
0.975
0.989
0.012
0.008
0.993
-0.238
0.329
0.975
1.023
1.007
-0.224
0.157
0.989
1.034
0.938
0.946
0.126
-0.070
0.283
0.663
0.482
0.304
0.066
0.001
0.115
0.682
0.731
0.208
0.551
0.910
0.668
0.717
0.147
0.217
0.431
0.614
0.795
0.577
0.676
0.559
0.093
0.034
0.007
0.678
0.672
0.275
1.022
1.015
-0.230
1.020
1.175
1.051
0.947
1.035
1.009
0.934
0.045
0.028
1.052
-0.147
-0.155
0.972
1.007
0.985
0.220
-0.906
1.033
0.999
0.859
0.893
-0.086
-0.118
-0.132
0.465
0.561
0.057
0.562
0.000
0.060
0.042
0.251
0.747
0.013
0.616
0.049
0.051
0.309
0.507
0.489
0.910
0.673
0.521
0.000
0.203
0.956
0.003
0.000
0.741
0.398
0.561
0.985
1.006
-0.006
0.997
0.984
0.998
1.005
1.005
1.004
1.002
0.021
-0.008
1.002
0.082
-0.064
1.004
1.009
1.009
0.079
-0.023
0.994
0.995
1.010
1.008
-0.115
0.075
0.040
0.022
0.277
0.897
0.699
0.009
0.671
0.339
0.448
0.450
0.733
0.510
0.148
0.781
0.095
0.437
0.699
0.486
0.263
0.517
0.838
0.273
0.377
0.251
0.178
0.194
0.113
0.622
0.997
1.007
-0.016
1.015
1.016
1.005
1.005
1.011
0.997
0.991
0.039
0.003
1.012
-0.081
-0.029
0.993
1.022
1.006
0.030
-0.109
1.008
1.010
0.990
1.001
0.002
-0.001
-0.023
0.695
0.280
0.724
0.107
0.029
0.505
0.492
0.146
0.626
0.215
0.206
0.585
0.092
0.097
0.730
0.538
0.215
0.545
0.804
0.327
0.215
0.155
0.337
0.855
0.984
0.984
0.778
Exhibit A-3B. Relationships Between DHA Resident and Neighborhood Characteristics: Households with 2 Children
Characteristics of DHA Resident
P/C is single parent (1=yes, 0=no)
P/C employment status at time of DHA move-in (1=employed, 0=not employed)
P/C hourly wage at time of DHA move-in
P/C disability status at time of survey (1=yes; 0=no)
P/C received TANF at time of DHA move-in (1=yes, 0=no)
P/C receiving Food Stamps at time of DHA move-in (1=yes, 0=no)
P/C had checking account at time of DHA move-in (1=yes, 0=no)
P/C had health insurance at time of DHA move-in (1=yes, 0=no)
P/C had too little money for food at time of DHA move-in (1=yes, 0=no)
P/C had difficulty paying all bills at time of DHA move-in (1=yes, 0=no)
Frequency that P/C drank alcohol since becoming a parent
Frequency that P/C smoked marijuana since becoming a parent
P/C ever seen a psychiatrist (1=yes, 0=no)
Number of years during childhood that P/C lived in public housing
Number of years during childhood that P/C lived in a home owned by parents
P/C born in the United States (1=yes; 0=no)
Spanish language interview (1=yes; 0=no)
Biological father always lived in household with child(ren) (1=yes; 0=no)
Parent's age at time of DHA move-in
Year that P/C moved into first randomly assigned unit
P/C African American (1=yes; 0=no)
Parent have HS diploma at time of DHA move-in (1=yes; 0=no)
Parent have any higher education at time of DHA move-in (1=yes; 0=no)
Kids share same biological dad (1=yes; 0=no)
Parent Depressive Symptomatology Scale at time of interview
Parenting Efficacy Scale at time of interview
Parenting Beliefs Scale at time of interview
Note: P/C = Parent or Caregiver; OR = odds ratios for dichotomous outcomes;
all neighborhood characteristics measured at time of first DHA move-in
N=216
bold = p<.05
Percent female Percent of residents Percent of residents
Percent foreign born Neighborhood poverty
headed households 25+ with LT 9th grade 25+ with Bachelor's
in neighborhood
rate
in neighborhood
education
degree or more
Percent Black
residents in
neighborhood
Percent Hispanic
residents in
neighborhood
Neighborhood
unemployment rate
Neighborhood
vacancy rate
Neighborhood
homeownership rate
Percent homes built
before 1940 in
neighborhood
Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value
1.003 0.754
1.001
0.960
1.005 0.721
1.011 0.508
0.997 0.713
1.019 0.090
0.997
0.746
0.982 0.467
0.984 0.586
1.003 0.728
1.003 0.692
1.007 0.316
0.965
0.023
1.018 0.157
0.970 0.018
0.997 0.629
1.032 0.001
0.979
0.003
0.987 0.547
0.993 0.792
1.007 0.231
1.001 0.906
-0.067 0.630
0.120
0.542
-0.558 0.454
0.027 0.841
0.004 0.981
-0.124 0.007
0.103
0.204
0.177 0.576
0.709 0.234
0.070 0.447
0.120 0.271
1.004 0.671
1.007
0.747
1.008 0.616
0.992 0.698
1.007 0.524
1.007 0.454
0.996
0.687
0.988 0.688
1.042 0.240
0.986 0.134
0.991 0.370
1.016 0.023
1.048
0.003
0.960 0.005
0.996 0.738
1.025 0.001
0.997 0.653
1.013
0.045
1.064 0.004
1.081 0.002
0.985 0.018
1.000 0.962
1.009 0.231
1.005
0.726
0.992 0.503
1.002 0.857
1.010 0.162
1.010 0.199
1.002
0.815
1.034 0.121
1.011 0.662
0.994 0.299
1.003 0.619
1.002 0.793
0.958
0.007
1.024 0.040
0.970 0.029
0.996 0.546
1.013 0.076
0.980
0.002
0.978 0.313
0.988 0.640
1.003 0.671
1.004 0.519
0.992 0.304
0.997
0.854
0.997 0.787
0.992 0.559
0.996 0.585
1.001 0.898
0.995
0.487
0.990 0.665
1.057 0.083
1.004 0.573
0.992 0.268
1.008 0.269
0.984
0.294
1.014 0.226
0.998 0.851
1.007 0.344
1.016 0.028
0.988
0.056
1.022 0.311
1.010 0.702
0.992 0.172
1.001 0.906
0.992 0.234
0.961
0.010
1.013 0.252
1.002 0.894
0.989 0.141
0.995 0.531
0.988
0.054
0.985 0.487
0.945 0.036
1.003 0.579
0.994 0.371
-0.063 0.374
0.122
0.192
-0.565 0.079
0.057 0.408
-0.041 0.619
-0.026 0.534
0.050
0.267
-0.070 0.686
-0.002 0.995
0.058 0.199
0.021 0.753
-0.044 0.565
0.148
0.094
-0.642 0.028
0.077 0.251
-0.015 0.868
-0.019 0.657
0.058
0.198
-0.029 0.874
-0.107 0.777
0.042 0.404
0.035 0.602
0.998 0.806
0.991
0.544
1.006 0.616
0.998 0.851
0.998 0.755
1.006 0.403
0.997
0.602
1.004 0.865
0.992 0.743
1.005 0.445
0.999 0.932
-0.301 0.287
-0.468
0.238
1.648 0.293
-0.327 0.212
-0.453 0.092
-0.048 0.787
-0.173
0.373
-0.761 0.225
1.291 0.353
0.217 0.257
-0.298 0.203
-0.415 0.038
0.123
0.764
0.612 0.697
0.109 0.691
-0.349 0.201
-0.126 0.406
0.080
0.672
-0.260 0.689
-0.051 0.971
0.146 0.434
-0.103 0.676
1.007 0.494
0.993
0.754
1.000 0.987
1.012 0.528
1.005 0.633
1.018 0.176
0.991
0.351
1.010 0.727
1.011 0.763
0.993 0.420
0.989 0.221
0.973 0.095
1.026
0.408
1.001 0.958
1.007 0.776
0.984 0.280
0.942 0.065
1.019
0.197
0.961 0.346
0.996 0.935
1.009 0.438
1.022 0.103
0.981 0.070
0.971
0.155
1.011 0.447
0.984 0.392
0.988 0.220
0.993 0.523
0.989
0.185
0.987 0.654
1.017 0.617
1.014 0.090
0.991 0.341
0.157 0.424
-0.345
0.169
0.570 0.613
-0.273 0.069
0.081 0.722
-0.021 0.855
-0.102
0.439
0.083 0.863
1.304 0.093
-0.007 0.959
0.018 0.921
0.163 0.472
0.296
0.347
-0.591 0.649
0.262 0.176
0.244 0.307
0.100 0.433
0.073
0.641
0.375 0.471
-1.595 0.055
-0.185 0.184
0.101 0.617
1.016 0.025
0.968
0.032
1.024 0.042
0.983 0.172
1.009 0.211
1.097 0.000
0.969
0.000
1.041 0.052
1.066 0.011
0.998 0.678
1.025 0.000
0.994 0.403
0.992
0.625
0.999 0.926
1.007 0.601
0.993 0.375
1.016 0.028
0.990
0.154
0.980 0.365
1.001 0.981
1.009 0.190
1.001 0.931
1.000 0.986
1.022
0.375
0.997 0.881
1.052 0.005
0.995 0.680
1.012 0.239
1.009
0.384
0.980 0.539
0.914 0.066
0.999 0.895
1.002 0.863
1.000 0.979
0.997
0.839
1.008 0.465
1.009 0.473
0.995 0.465
1.007 0.337
0.999
0.932
0.977 0.283
0.989 0.657
1.002 0.793
1.006 0.346
0.025 0.951
-0.219
0.704
2.283 0.250
-0.055 0.889
0.130 0.770
-0.135 0.543
0.012
0.964
0.782 0.363
0.520 0.789
-0.193 0.467
0.104 0.766
-0.029 0.656
0.096
0.258
-0.533 0.038
0.042 0.498
-0.018 0.808
-0.010 0.791
0.035
0.398
-0.062 0.682
-0.020 0.950
0.043 0.307
0.018 0.748
-0.014 0.913
-0.019
0.919
-0.277 0.699
-0.068 0.580
0.007 0.961
-0.095 0.104
0.041
0.634
0.102 0.732
0.928 0.025
0.064 0.455
0.081 0.450
Exhibit A-3C. Relationships Between DHA Resident and Neighborhood Characteristics: Households with 3+ Children
Characteristics of DHA Resident
P/C is single parent (1=yes, 0=no)
P/C employment status at time of DHA move-in (1=employed, 0=not employed)
P/C hourly wage at time of DHA move-in
P/C disability status at time of survey (1=yes; 0=no)
P/C received TANF at time of DHA move-in (1=yes, 0=no)
P/C receiving Food Stamps at time of DHA move-in (1=yes, 0=no)
P/C had checking account at time of DHA move-in (1=yes, 0=no)
P/C had health insurance at time of DHA move-in (1=yes, 0=no)
P/C had too little money for food at time of DHA move-in (1=yes, 0=no)
P/C had difficulty paying all bills at time of DHA move-in (1=yes, 0=no)
Frequency that P/C drank alcohol since becoming a parent
Frequency that P/C smoked marijuana since becoming a parent
P/C ever seen a psychiatrist (1=yes, 0=no)
Number of years during childhood that P/C lived in public housing
Number of years during childhood that P/C lived in a home owned by parents
P/C born in the United States (1=yes; 0=no)
Spanish language interview (1=yes; 0=no)
Biological father always lived in household with child(ren) (1=yes; 0=no)
Parent's age at time of DHA move-in
Year that P/C moved into first randomly assigned unit
P/C African American (1=yes; 0=no)
Parent have HS diploma at time of DHA move-in (1=yes; 0=no)
Parent have any higher education at time of DHA move-in (1=yes; 0=no)
Kids share same biological dad (1=yes; 0=no)
Parent Depressive Symptomatology Scale at time of interview
Parenting Efficacy Scale at time of interview
Parenting Beliefs Scale at time of interview
Note: P/C = Parent or Caregiver; OR = odds ratios for dichotomous outcomes;
all neighborhood characteristics measured at time of first DHA move-in
N=265
bold = p<.05
Percent female Percent of residents Percent of residents
Percent foreign born Neighborhood poverty
headed households 25+ with LT 9th grade 25+ with Bachelor's
in neighborhood
rate
in neighborhood
education
degree or more
Percent Black
residents in
neighborhood
Percent Hispanic
residents in
neighborhood
Neighborhood
unemployment rate
Neighborhood
vacancy rate
Neighborhood
homeownership rate
Percent homes built
before 1940 in
neighborhood
Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value Coeff./OR P value
1.002 0.807
0.997
0.864
0.997
0.872
1.007
0.644
0.999 0.936
1.010
0.240
0.999
0.886
0.996 0.884
1.013
0.679
0.999
0.917
1.002 0.767
0.996 0.621
1.006
0.757
0.997
0.816
1.028
0.056
0.994 0.418
1.002
0.803
1.001
0.875
0.970 0.177
0.928
0.013
0.999
0.852
0.995 0.466
-0.054 0.562
0.127
0.579
-0.315
0.428
0.125
0.215
-0.060 0.469
-0.098
0.519
0.115
0.113
-0.110 0.633
-0.322
0.148
0.059
0.368
0.023 0.765
0.992 0.530
1.045
0.174
0.952
0.215
1.001
0.960
0.999 0.911
0.990
0.456
1.016
0.245
1.035 0.334
1.033
0.466
1.010
0.364
1.014 0.235
1.004 0.557
1.010
0.578
1.004
0.773
0.984
0.244
1.012 0.130
0.991
0.218
1.004
0.585
1.035 0.128
1.030
0.298
0.991
0.204
1.003 0.724
1.010 0.238
1.025
0.191
0.995
0.732
1.008
0.587
1.016 0.070
0.992
0.267
1.010
0.198
1.047 0.060
0.969
0.292
0.985
0.036
1.011 0.154
0.999 0.888
0.952
0.010
1.034
0.037
0.993
0.613
0.991 0.279
1.012
0.100
0.977
0.002
0.960 0.088
0.928
0.020
1.003
0.647
0.982 0.018
0.992 0.336
0.994
0.742
1.028
0.155
1.034
0.033
0.991 0.261
0.990
0.160
1.008
0.274
0.973 0.249
0.911
0.003
1.001
0.862
1.001 0.943
1.004 0.600
1.011
0.554
1.000
0.985
0.990
0.480
1.005 0.535
0.999
0.858
0.997
0.720
1.002 0.930
1.000
0.995
0.995
0.481
0.997 0.722
1.000 0.990
1.015
0.431
0.986
0.388
1.022
0.118
0.998 0.782
1.003
0.698
1.003
0.673
1.007 0.758
0.954
0.125
0.996
0.593
1.001 0.846
-0.003 0.905
-0.004
0.936
-0.028
0.761
-0.009
0.709
-0.001 0.979
-0.015
0.663
0.009
0.592
0.031 0.556
-0.007
0.889
0.005
0.717
0.001 0.946
0.008 0.225
-0.019
0.243
0.003
0.926
-0.004
0.644
0.005 0.371
0.008
0.485
-0.002
0.744
0.023 0.149
-0.004
0.828
-0.004
0.380
0.002 0.677
1.000 0.990
1.006
0.729
0.975
0.137
0.974
0.061
1.004 0.637
0.997
0.649
1.003
0.639
1.028 0.223
1.028
0.324
1.003
0.692
0.999 0.925
-0.017 0.867
-0.316
0.175
0.657
0.099
-0.245
0.009
-0.011 0.906
0.014
0.931
-0.143
0.056
-0.017 0.946
0.236
0.332
0.037
0.594
0.033 0.684
0.058 0.500
-0.213
0.310
-0.070
0.854
0.075
0.437
0.033 0.673
-0.076
0.595
-0.001
0.989
0.063 0.770
-0.283
0.177
-0.084
0.158
-0.129 0.049
1.008 0.478
1.016
0.520
0.981
0.249
0.988
0.491
1.011 0.352
1.003
0.784
0.998
0.860
1.018 0.564
1.081
0.086
0.986
0.142
1.003 0.753
0.995 0.770
0.983
0.623
0.996
0.890
0.987
0.642
0.988 0.467
0.988
0.498
1.014
0.388
1.015 0.734
0.953
0.441
1.021
0.155
1.004 0.776
0.985 0.088
0.961
0.047
1.020
0.189
0.988
0.444
0.983 0.065
0.991
0.284
0.986
0.068
0.944 0.027
0.963
0.251
1.010
0.210
0.980 0.014
-0.523 0.024
0.577
0.364
-0.622
0.582
0.516
0.055
-0.417 0.052
-0.933
0.011
0.301
0.145
-1.267 0.029
-1.362
0.018
0.212
0.241
-0.288 0.164
-0.171 0.193
0.118
0.726
-0.611
0.289
0.422
0.000
-0.154 0.194
-0.306
0.152
0.236
0.017
-0.296 0.376
-0.721
0.016
0.063
0.517
-0.077 0.494
1.026 0.001
0.975
0.153
1.020
0.193
0.966
0.016
1.023 0.004
1.053
0.000
0.967
0.000
1.064 0.007
1.097
0.002
0.986
0.041
1.013 0.076
0.990 0.192
0.973
0.146
1.040
0.015
1.002
0.897
0.987 0.112
0.999
0.855
0.995
0.541
0.979 0.363
0.938
0.042
1.012
0.102
1.004 0.539
0.974 0.123
0.994
0.859
1.017
0.444
1.038
0.113
0.979 0.195
1.011
0.367
0.996
0.780
0.957 0.324
0.957
0.445
1.008
0.553
1.004 0.782
0.989 0.167
0.976
0.205
1.003
0.821
1.019
0.183
0.991 0.263
0.995
0.508
1.000
0.990
0.989 0.626
0.929
0.024
1.008
0.279
0.992 0.273
-0.164 0.245
-0.228
0.522
-0.044
0.945
0.127
0.432
-0.162 0.197
-0.208
0.376
-0.043
0.719
-0.600 0.072
-0.491
0.159
0.063
0.544
-0.121 0.305
0.000 0.995
0.017
0.909
-0.050
0.846
-0.011
0.866
-0.009 0.866
0.122
0.192
-0.010
0.839
0.030 0.838
0.144
0.324
0.034
0.421
0.101 0.018
-0.121 0.119
-0.314
0.099
0.646
0.045
0.028
0.762
-0.123 0.073
-0.188
0.143
-0.018
0.796
-0.144 0.478
-0.249
0.213
0.069
0.228
0.034 0.616
29
Only two individual characteristics had a frequency of statistically significant coefficients that
were greater than average: African-American DHA tenant (15 percent) and household wages
(14 percent). It is noteworthy that although disability status generated a non-random
assignment to particular developments due to DHA rules (as shown in Exhibit A-1) this
apparently did not produce a strong association with particular neighborhood characteristics
because the locations where the disabled were assigned evinced considerable variation.
Of course, geographic selection bias arises to the extent that individual household
characteristics that are not observed (or controlled statistically) are correlated with both
neighborhood characteristics and child outcomes. In this regard it is revealing to separate the
individual characteristics listed in Exhibit A-3 into the first 15 (which were not observable to DHA
officials because they were gleaned from our household survey) and the last 12 (which likely
were). Ninety-five percent of the former set’s coefficients were not statistically significant,
whereas only 88 percent of the later set’s were. This is consistent with the notion that, although
DHA’s assignment process may not have produced a completely random assignment across
neighborhood characteristics based on household characteristics that DHA staff could observe,
it nevertheless likely produced such based on household characteristics that they could not
observe.
We therefore conclude that this third piece of evidence suggests the DHA allocation process
produced a quasi-random assignment across geography, with the possible exception of two
individual characteristics observable by the DHA—African-American ethnicity and household
wages—that are easily controlled in our analyses. Even more importantly, we conclude that the
DHA allocation process produced a quasi-random assignment across geography in terms of
individual characteristics not observable by the DHA (but observable to us from our survey).
This gives us some confidence that any additional household characteristics we do not observe
in our study are similarly quasi-randomly allocated across neighborhood characteristics.
Relationships between Typically Unobserved Individual Characteristics and Neighborhood
Characteristics using Monte Carlo Simulation
Recall that the key issue at hand is whether DHA’s assignment of public housing tenants to
neighborhoods effectively removes the correlation between unobservable (i.e., cannot be
controlled statistically) parental characteristics that might affect both characteristics of location
chosen and individual outcomes being investigated. We investigated this by examining the
degree to which a variety of characteristics of parents/caregivers in our sample that typically are
not observed in neighborhood effect studies were correlated with multiple characteristics of their
neighborhoods at the time of initial assignment by DHA. The intuition guiding our analysis is as
follows. An actual random assignment of DHA applicants to DHA dwellings will likely produce
by chance a few non-zero pairwise correlations between DHA household characteristics and
neighborhood characteristics. A Monte Carlo simulation repeating such random assignments
will generate bootstrapped standard errors and distributions of such correlations for each pair.
This provides the benchmark against which we will compare the actual pairwise correlations
between DHA household characteristics and neighborhood characteristics. If the pattern of the
actual correlations does not differ significantly from that produced by the simulation, we will fail
to reject the null hypothesis that the DHA assignment process yielded a quasi-random
geographic assignment of households according to their unobserved characteristics.
In particular, we implemented this strategy as follows. We considered here the unobserved (by
DHA and typically in other studies) characteristics of parents (listed in Exhibit A-4) and the
characteristics of census tracts considered above. For each of the three aforementioned family
30
sizes of DHA tenants we calculated the Pearsonian correlation between each pairwise
combination of parental characteristics and neighborhood characteristics observed when the
DHA first assigned our sample households to their DHA units.
As a comparative benchmark for these correlations we conducted Monte Carlo simulations in
which each sample household was, indeed, randomly assigned to one of the DHA units (for the
appropriate family size) with its associated bundle of neighborhood characteristics that we
observed whenever the initial assignment of household in our study actually occurred.37 In each
iteration after all households were randomly assigned we calculated correlations for all pairwise
combinations of parental characteristics and neighborhood characteristics. We used 10,000
repetitions of these simulations to produce distributions for all pairwise combinations of parental
characteristics and neighborhood characteristics and associated bootstrapped standard errors.
This allowed us to estimate: (1) for each correlation a 95 percent confidence interval, and (2)
across all pairwise correlations how many significantly different from zero would be expected by
chance when produced by a random assignment process.
The results are reported in Exhibit A-4. The parental characteristics are listed in the rows and
the three family-size strata in the columns. The cells show for how many of the possible
neighborhood characteristics the initial DHA assignment produced an actual correlation with the
given parental characteristic that was significantly different from zero at the 5 percent level (twotailed test); the actual correlation coefficient and the neighborhood characteristic involved are
reported in these cases. The exhibit shows that for families with no or one child and families
with two children, only eight (5 percent of possible correlations) were statistically different from
zero; the corresponding figure for families with three or more children was 12 (8 percent of
possible correlations). Our simulations showed that in over 98 percent and 95 percent of the
cases, respectively, a larger number of statistically significant correlations were produced by a
random assignment. This strongly indicated that the relatively rare non-zero correlations we
observed from initial DHA allocations of tenants to neighborhoods (shown in Exhibit A-4) were
consistent with those that would have been generated by a pure process of random assignment.
These results suggest that the DHA natural experiment likely removes the correlation between
parental characteristics (which we do not observe and cannot control in our Denver study) that
may potentially affect both initial DHA neighborhood characteristics and subsequent individual
outcomes.
37
The programming and execution of these simulations was conducted by Dr. Albert Anderson of PDQ
Inc., whose contribution we gratefully acknowledge.
31
Exhibit A-4. Simulation Results: Number of Statistically Significant Correlations between
Typically Unobserved Household Characteristics and Neighborhood Characteristics
Household Characteristic
Families with 0-1 Families with 2
Families with 3+
Child
Children
Children
Ever not enough food for family
0
1 (%black=.14)
0
while reside in this location
Ever unable to pay all bills while
2 (%foreign-born 2 (%elem. school 1 (%vacant = -.12)
reside in this location
= .13; %vacant = ed. = -.17;
-.16)
%vacant = -.14)
Frequency of alcohol use since
2 (%unemployed 0
1 (%black = -.09)
becoming parent
= -.16; %owner
=.13)
Frequency of marijuana use since 1 (%black = .17)
0
0
becoming parent
Frequency of drug use since
1 (%black = .13)
0
0
becoming parent
Ever seen psychologist,
0
0
0
psychiatrist or counselor
Did your parents ever live in
1 (%female
0
1 (%foreign born = public housing when you were
heads = .22)
.18)
growing up
Did your parents ever own their
0
3 (%elem. school 0
home when you were growing up
= .26; %college =
-.26; %own =
.20)
Born in U.S.
1 (%college = 0
0
.16;)
Primary language is Spanish
0
0
0
Father of child always lived in
0
0
5 (%female heads =
home while child growing up
-.11; %elementary
school = -.10; %poor
= -.10; %own = .09;
%pre-1940 homes=
-.12)
Parental depression (CESD)
0
1 (%Latino = .13) 2 (%elem. school =
scale
.13;
%Latino=.13)
Parental self-efficacy scale
0
0
0
Parental beliefs & practices scale 0
1 (%Latino = 2 (%college = -.09;
.21)
%black = -.12)
Source: authors’ calculations based on Monte Carlo simulations of Denver Child Study survey
data; statistically significant household-neighborhood characteristic correlations shown
parenthetically
32
Conclusion
Natural experiments involving residential placements under the auspices of some public
program offer potentially powerful vehicles for measuring neighborhood effects because they
can rupture the association between unobserved characteristics of the individuals being studied
and characteristics of their neighborhood. In this appendix we have investigated the extent to
which a natural experiment involving public housing in Denver offers such potential.
Our analysis of the Denver Housing Authority’s dwelling allocation procedures revealed
considerable room for tenant self-selection and/or DHA staff selection to enter. Nevertheless,
we found that the initial occupancy mimicked a quasi-random assignment process to DHA
dwellings or neighborhoods, with the exception of ethnicity and disability status. Only African
American ethnicity (and to a lesser degree, household wages) exhibited above-average
frequencies of associations with neighborhood conditions, however. This suggests that,
conditioned on ethnicity and wages, the DHA allocation process produced a quasi-random initial
assignment across neighborhood characteristics. The empirical implication is that models
estimating neighborhood effects using the current data must control for ethnicity and wages to
avoid geographic selection bias. We, in fact, do so in all analyses conducted in this report.
Even more importantly, two sorts of analyses indicate that the DHA allocation process produced
a quasi-random assignment across neighborhood conditions in terms of individual
characteristics not observable by the DHA (but observable to us from our survey). This gives us
some confidence that any additional household characteristics we do not observe in our study
are similarly quasi-randomly allocated across neighborhood characteristics.
References Cited in Appendix A but not in Main Text
Jacob, Brian. 2004. “Public Housing, Housing Vouchers, and Student Achievement: Evidence
from Public Housing Demolitions in Chicago,” American Economic Review 94: 233-58.
Lyle, David. 2007. Estimating and Interpreting Peer and Role Model Effects from Randomly
Assigned Social Groups at West Point,” The Review of Economics and Statistics 89(2): 289–
299.
Oakes, J. Michael. 2004. “The (Mis)estimation of Neighborhood Effects: Causal Inference for a
Practicable Social Epidemiology,” Social Science and Medicine 58: 1929–1952.
33
Appendix B:
Details of Denver Child Study Survey, Other Data Sources and Construction of Variables
Denver Child Study Household Survey
We developed and fielded during 2006-2008 the Denver Child Study telephone survey
(conducted in person for about 20 percent of sample who had no landline phones) that collected
retrospective and current information about the household, adults and children. Detailed
information related to multiple domains of outcomes was gathered for all eligible children
associated with each household. Each household’s residential mobility history was obtained so
it could be associated with neighborhood developmental context for children. Study eligibility
criteria were: (1) presence of children in the home between ages 0 and 18 years when they
moved into DHA; (2) family remained in DHA housing for at least two years; (3) family first
entered DHA in 1987 or later (when DHA’s current quasi-random assignment process came into
operation); and (4) were of Latino or African American ethnicity.
Attempts to recruit participants for the study were made by mail, phone and in person, in both
English and Spanish when appropriate. Compensation for participation took the form of either
cash or gift card. We estimate an overall participation rate of 56.5 percent (85 percent for those
still residing in DHA), with most non-participation due to our inability to locate the household;
less than six (6) percent refused to participate once contacted. Our team successfully
completed 710 interviews with the primary caregivers of eligible households whose surveys
subsequently passed our rigorous data verification and reliability checks. Children and youth
analyzed in our study were current or past members of these 710 households who spent two or
more years residing in DHA housing before reaching age 19.
Variables Measuring Characteristics of Caregivers and Households
Our Denver Child Study survey collected information on a wide variety of parental/caregiver
(“caregiver” hereafter) and household characteristics. This included time-invariant information
about caregiver gender, year of birth and national origin. Less conventionally, our survey asked
respondents whether they had used alcohol, marijuana and/or other illegal drugs since
becoming a parent, a situation we coded with a dummy variable indicator.
The survey also permitted us to estimate caregiver depressive symptomatology at time-ofsurvey and served as a control for caregiver affect when responding to survey questions, not
necessarily caregiver emotional state during onset of a particular child outcome. Interviewers
administered the Center for Epidemiologic Studies Depression (CES-D) Scale developed by
Radloff (1977). This scale is a self-evaluative questionnaire designed to assess whether an
individual possesses emotional traits or characteristics that commonly indicate the presence of
clinical depression. The scale is based on 20 questions about the emotions a person has felt
over the past week, with responses of 0 indicating less than 1 day, 1 indicating 1-2 days, 2
indicating 3-4 days, and 3, indicating 5-7 days that a particular emotion was felt during the
week. Scoring for items indicating positive feelings is reversed, and the points are summed.
Overall scores range from 0 to 60; with scores less than 16 indicating no depressive symptoms,
16 to 26 indicating sub-clinical depressive symptomatology and scores of 27 or higher indicating
34
clinical depressive symptomatology. We used a dummy variable indicating whether the
caregiver exhibited sub-clinical or clinical depressive symptomatology (i.e., scored at least 16 on
the CES-D scale). The Cronbach’s alpha for the CES-D has been found to be 0.80 or above
within the general population as well as various subpopulations. In this study, the Cronbach’s
alpha for the CES-D was 0.869. The scale is available at
http://www.chcr.brown.edu/pcoc/cesdscale.pdf).
We also were able to measure in the Denver Child Study many aspects of the caregiver and
household that varied over the life of the child, such as educational attainment, economic status,
disability status, marital status, fertility and employment histories, and access to health
insurance. The survey also included a Household Stressors index comprised of five items
which inquire about difficulties faced by the household in terms of finances, employment,
insurance, utilities, and housing. For each year in a child’s life, this index measures the
magnitude of stressors facing the household at the corresponding residential location.
Caregivers were asked whether they experienced any of the following events:
a. Unemployed a month or more?
b. Have a major illness or injury?
c. Have too little money to buy enough food for your family?
d. Have your electricity, gas, or phone service cut off?
e. Get evicted from your home?
Possible responses to each item (followed by their associated score) were no (0) and yes (1),
indicating either the presence or absence of a certain stressor. Scores could vary from 0 to 5,
with higher values indicating a greater degree of household stress. Cronbach’s alpha for this
index was .50.
Variables Measuring Characteristics of Children and Youth
Our Denver Child Study survey asked caregivers to supply information about all their children
with whom they had lived in DHA public housing for at least one year. In this manner we
collected detailed information about the children’s gender, ethnicity, birth order, residential
histories, health, exposure to violence, behaviors and activities, education and (for older
children), marital/fertility histories and labor market outcomes during early adolescence and
young adulthood. We also recorded focal child’s number of siblings during the time period
under analysis, and other behaviors of older siblings (e.g. dropping out of high school). Finally,
residential history information permitted us to compute the number of moves the household had
undertaken during the childhood of the observed youth.
Variables Measuring Characteristics of Neighborhoods Experienced by Children and Youth
It is generally accepted that “neighborhood” has both objective “space” dimensions (i.e.,
economic, demographic, social indicators associated with geographies) and subjective “place”
dimensions (i.e., the human experience of territory) (Fitzpatrick and LaGory, 2000). We
obtained a wide variety of neighborhood data about both dimensions from four sources.
The first source was the decennial U.S. Census, where we used census tract geographic scales
from 1980, 1990 and 2000 censuses. We employed the Neighborhood Change Data Base (a
Geolytics proprietary product) for this information because it adjusts data to account for changes
in tract boundaries between decennial censuses. For estimates of non-census year data, we
used linear interpolation or extrapolation. We gathered indicators that have been widely
35
employed in prior research on neighborhood effects, including percentages of: households
moving in during the prior year, female-headed households, families below the poverty line,
unemployed adults, Latino population, non-Latino African American population,38 foreign-born
population, homes that are renter-occupied, homes that were built during various periods, and
mean occupational prestige based on the General Social Survey prestige score weighted by the
observed proportional distribution of occupations of employees in the tract. Given high
correlations among several of these variables, we conducted four principal components
analyses, one for a comparable set of variables for each of the 1980-2000 censuses.39 For
each census year, the analysis produced a single component (with an eigenvalue greater than
unity) that consistently was comprised of the roughly equally weighted sum of census tract
percentages of: poor, unemployed, renters, and female household heads. We call this our
neighborhood social vulnerability score (ranging from 0-400).
We note that our neighborhood social vulnerability score is not strictly comparable to similar
indices of “neighborhood disadvantage” used in many prior scholarly works for two reasons.
First, our index sums neighborhood percentages of: unemployment, poverty, female-headed
households and renters; it does not include ethnic, racial, or nativity measures, as do most
others. Second, our models control for other neighborhood characteristics that are often
associated with “disadvantaged neighborhoods” but for which other studies have no direct
measures: notably crime and occupational prestige. Thus, other studies’ “neighborhood
disadvantage” variables serve as ambiguous proxies for a diverse bundle of conditions and
quantitative results using such measures should not be used as precedents for results using our
social vulnerability measure.
The second source of neighborhood information was the Denver-based Piton Foundation’s
Neighborhood Facts Database, which provided small area-based, annually measured
information culled from Denver administrative databases on characteristics that are not provided
by the Census. These included violent crimes reported to police per 1,000 population and
property crimes reported to police per 1,000 population. The Piton Foundation data are
aggregated to 77 named community areas consisting of two census tracts, on average, and
thus are measured at a larger spatial scale than our census-based data. Moreover, Piton series
are available only for the City and County of Denver, which produced shrinkage in our analysis
sample because some former DHA households interviewed had moved out of the county.
Quality Controls and Creation of Analytical Databases
We spent considerable effort cleaning, reconciling and augmenting the survey data. When our
audits revealed inconsistencies or omissions in the responses, we attempted to contact
respondents again and seek clarifications. Information provided by respondents on their
residential histories was cross-checked with residential location information contained in the
DHA administrative databases, U.S. Postal Service, Lexis-Nexis and Intelius address files as
well as several additional online search engines.
Once residential history information obtained on the survey was verified for accuracy, we geocoded each address, using the U.S. Bureau of the Census’ American FactFinder website utility.
In cases where respondents could not recall specific addresses but only proximate crossstreets, we verified these locations using MapQuest and then identified the corresponding
census tract using the aforementioned Census website showing tract boundaries. This
38
The ethnic makeup of Denver in 2000 was 52 percent non-Latino whites, 11 percent non-Latino African
Americans, and 32 percent Latinos.
39 The creation of our linked database occurred prior to the release of the 2010 Census and the five-year
average American Community Survey data.
36
procedure provided the census tract corresponding to each location in respondents’ residential
histories, which, in turn, permitted us to match each location to the aforementioned battery of
neighborhood indicators for census tract neighborhoods. We were able to successfully link 92
percent of the residential locations identified by respondents.
We then transformed these data for households and neighborhoods into the format of a childyear unit of observation. For each child-year there are variables associated with: (1) fixed child
characteristics, (2) fixed caregiver characteristics; (3) temporally varying child characteristics;
(4) temporally varying caregiver-household characteristics; (5) temporally varying neighborhood
characteristics; and (6) temporally varying outcomes.
References Cited in Appendix B
Fitzpatrick, Kevin, and Mark LaGory. 2000. Unhealthy Places: The Ecology of Risk in the Urban
Landscape. New York: Routledge.
Furtado, Bernardo. 2011. “Neighbourhoods in Urban Economics: Incorporating Cognitively
Perceived Urban Space in Economic Models,” Urban Studies 48(13): 2827-2848.
Muhajarine, Nazeem, Ronald Labonte, Allison Williams, and James Randall. 2008. “Person,
Perception, and Place: What Matters to Health and Quality of Life,” Social Indicators
Research 85(1): 53-80.
Radloff, L. 1977. The CES-D Scale—A Self-Report Depression Scale for Research in the
General Population, Applied Psychological Measurement 1(3), 385-401.
37
APPENDIX C:
MAPS OF DHA HOUSING AND RESIDENTIAL PATTERNS BY ETHNICITY AND
POVERTY, DENVER, CO.
Map 1: Locations of Sample Household’s First Denver Housing Authority Dwelling
Sources: Denver Housing Authority, Piton Foundation Neighborhood Facts database, map by
authors;
Note: year varies by date of first offer but range from 1987-2008; each point can represent
multiple DHA dwellings
38
Map 2: Percentages of African American Population by Census Tract, 2010
39
Map 3: Percentages of Latino Population by Census Tract, 2010
40
Map 4: African American and Latino Populations by Census Tract Poverty Rate, 2010