neighborhoods and selection 1 running head: …
TRANSCRIPT
NEIGHBORHOODS AND SELECTION 1
Running head: NEIGHBORHOODS AND SELECTION
Reciprocal Associations between Neighborhood Context and Parent Investments: Selection
Effects in Two Longitudinal Samples
September 2016
Thomas J. Schofield1 Melissa Merrick2
M. Brent Donnellan3 Chia-Feng Chen4
Fragile Families Working Paper WP16-08-FF
1 Department of Human Development and Family Studies, Iowa State University, 1360 Palmer, Iowa State
University, Ames, IA 50010. USA [email protected] phone:515-294-6914 fax:515-294-2502
2 Division of Violence Prevention, National Center for Injury Prevention and Control, Centers for Disease Control
and Prevention, Atlanta GA. USA [email protected]
3 Department of Psychology, Texas A&M University, College Station TX. USA [email protected]
4 Department of Human Development and Family Studies, Iowa State University. Ames, IA 50010. USA
NEIGHBORHOODS AND SELECTION 2
Abstract
The present study addresses the degree to which neighborhood disadvantage and parenting
investments are reciprocally linked over time, and the relative degree to which both show
indirect effects on child externalizing through each other. Data come from two studies: the first
followed families from the child’s birth to age 11 (N = 1,364), the second followed children from
birth to age 9 (N = 4,898). In both studies, material and emotional parenting
investments/resources predicted selection into neighborhoods over time, and neighborhood
disadvantage frequently predicted relative change in parenting investments. The prediction to
change in child externalizing was larger for parenting investments than it was for neighborhood
characteristics in most of the models tested.
Keywords: social environments; communities; neighborhoods
NEIGHBORHOODS AND SELECTION 3
Reciprocal Associations between Neighborhood Context and Parent Investments: Selection
Effects in Two Longitudinal Samples
The link between neighborhood attributes and child development has been replicated
across scores of samples, among families across a wide range of socioeconomic status (SES) and
ethnic backgrounds (Leventhal & Brooks Gunn, 2009). Based on this literature, policymakers
and other stakeholders frequently interpret these associations as causal. That is, neighborhood
context is now believed to create changes in child socioemotional outcomes. Even so, a
longstanding concern about neighborhood effects on child development is that families may
select nonrandomly into neighborhoods. Selection into neighborhoods would inflate associations
between neighborhood attributes and child outcomes, and call into question the underlying
directional assumption that correlations between neighborhoods and family processes reflect a causal
effect of neighborhoods. In the current study, we address this important issue, focusing on the outcome of
child externalizing behavior problems.
People select into environments (Jaffee & Price, 2007). Selection effects can refer to
constraints or preferences that are volitional (Scarr & McCartney, 1983) or nonconscious; they
result in people moving into environments that are in some way correlated with preexisting
personal characteristics (Cao, Mokhtarian, & Handy, 2009). In terms of the neighborhood
context, such selection effects are what urban sociologists such as Gans (1968) referred to when
he proposed that community life is primarily the aggregate of individual characteristics.
Selection is frequently studied in work on social contexts. For example, recent work on the
hypothesized influence of peers on adolescent development specifies both selection into peer
group, as well as peer influence (Dishion, 2013). Similar considerations are made when studying
the influence of one’s spouse/coparent on parenting behavior (Schofield et al., 2009; Schofield &
NEIGHBORHOODS AND SELECTION 4
Weaver, 2016). Failure to model selection into environments (e.g., niche picking) can spuriously
inflate associations between the environment and later development (Liberson, 1986), and can
create other errors of inference. For instance, one possibility is that poorer people tend to live in
poorer neighborhoods because they cannot leave the resource-providing social networks in
which they are embedded (such as childcare provided by friends, neighbors or family members).
Another possibility is that poorer families may have the income to move into a better
neighborhood yet lack the credit history to qualify for a good interest rate on a mortgage.
Consequently, even in instances where residents could potentially afford to live in a different
neighborhood, it would nonetheless be impractical to leave their current neighborhood. If the
poorest or least capable residents were to remain in a particular neighborhood while the more
resourceful or capable residents were to move to other neighborhoods, this would lead to an
overestimation of neighborhood effects. If selection is occurring with regard to neighborhoods, it
will require a reevaluation of how neighborhood models are to be specified. There is reason to
expect that selection occurs at the neighborhood level, and that the selection is related to the
same traits and processes that influence family dynamics. However, the selection-based
interpretation is less common partially due to a lack of evidence of selection into neighborhoods.
There have been very few tests of selection into neighborhoods (Bagley & Mokhtarian,
2002; Hedman, & van Ham, 2010). Qualitative research suggests multiple drivers of
neighborhood choice (Batty & Flint, 2010; Hickman, 2010; Mee, 2007), yet the few quantitative
studies that have tested for such selection effects include only demographic variables (Diez-Roux
et al., 1997; Ecob & Macintyre, 2000; McKinnish & White, 2011; Sampson & Sharkey, 2008).
However, people who live near each other are more likely to be similar with regard to their
personality (Jokela et al., 2014), their rate of extreme marital conflict (Hetling & Zhang, 2010),
NEIGHBORHOODS AND SELECTION 5
and child maltreatment experiences (Zhou, 2006). The typical interpretation of these correlations
is that the neighborhood context is exerting a causal influence on residents, making them friendly
or aggressive, responsible or impulsive. The selection-based interpretation is that residents with
similar personalities, levels of marital conflict, or child maltreatment are either attracted to or
pushed into particular neighborhoods. Therefore, an adequate accounting of preexisting
characteristics would include not only SES, but also parent attributes which are associated with
later child behavioral outcomes and have been shown to be distributed nonrandomly across
neighborhoods including the home environment (Bradley & Caldwell, 1977), personality
(McCabe, 2014), parenting behaviors (Cowan, Cowan, & Barry, 2011), support from a stable
nurturing romantic relationship (Erel & Burman, 1995), and so on. Exemplars of many of these
categories are distributed nonrandomly across neighborhoods, further implicating them in the
hypothesized selection process (Hetling & Zhang, 2010; Jokela et al., 2014; Zhou, 2006).
A second perspective that supports selection into neighborhoods comes from economic
work on the material investments parents make in their children (Mayer, 1997). That is, where
children live reflects in part an investment parents make in their child, similar to having reading
materials in the home or health insurance for the child. Higher quality neighborhoods afford
children greater opportunities for safety, outdoor recreation, and positive peer interactions that
may not be available in lower quality neighborhoods. Economic forces are indeed a driver of
selection, as only families that can afford to move to better neighborhoods have the option to do
so. Once economic constraints are taken into account, however, the choice of neighborhoods
becomes an indicator of material investment in the child’s well-being. Therefore, selection into
neighborhoods ought to be subject to the same drivers that predict other parental investments
(either material or emotional see Schofield et al., 2011).
NEIGHBORHOODS AND SELECTION 6
In the current study, we hypothesize reciprocal associations between neighborhood SES
and parent investments (Figure 1). We expect neighborhood SES will predict relative change in
parent investments (Path A) and child externalizing problems (Path C). These paths are
consistent with the dominant perspective that neighborhoods causally influence family processes
and child development. However, we also expect that parent investments (and related resources)
will predict selection into or out of neighborhoods (Path B). This path is very rarely measured,
and has never been tested over time to our knowledge. Finally, we expect that parent investments
and related resources will predict relative changes in child externalizing problems (Path D). This
last path is consistent with the dominant perspective that parents causally influence child
development.
We focus on neighborhood SES because it is the canonical variable of interest for many
developmental studies of neighborhood effects (Leventhal & Brooks-Gunn, 2009), and we focus
on externalizing problems because most neighborhood effects relate to deviant or delinquent
behavior, of which externalizing behavior problems are the developmental precursors. We
include household income as a parenting investment in all models, because of the clear role
economic constraints would have on the ability to relocate into or out of a neighborhood. In
addition to these theoretical processes, the present study also addresses the issue of replication.
Concerns about the robustness of research results, and the capacity to replicate across
populations have long been a concern in the field of child development (Weisz, 1978). In order
to more comprehensively evaluate the current hypotheses, we test our theoretical model across
two longitudinal samples: families participating in the NICHD Study of Early Child Care and
Youth Development and families participating in the Fragile Families Study. Both samples
experience considerable variation in neighborhood SES, both include prospective longitudinal
NEIGHBORHOODS AND SELECTION 7
data which are necessary to test for selection (Jokela et al. 2014), and both include multiple
measures of parental investments which are necessary to assess the support for these hypotheses.
Method
Primary Sample
Participants
Participants were the families in the National Institute of Child Health and Human
Development (NICHD) Study of Early Child Care and Youth Development (SECCYD). These
1,364 families were recruited in 1991, shortly after their child’s birth, from hospitals at 10 sites
across the United States. Specific recruitment procedures are detailed elsewhere (NICHD Early
Child Care Research Network, 2005). This study was conducted with approval from the
affiliated Institutional Review Boards. This sample included a substantial proportion of low-
education parents (30% had a high school degree or less), ethnic minority families (13% were
African American compared with the national proportion of 12%), and the mean income level
was the same as the U.S. average ($37,000).
Procedures and Variables
Detailed measures of family demographics, maternal behaviors, and children’s
characteristics and adjustment were obtained from multiple informants beginning when children
were 1 month of age and continued annually until they were 15 years old.
Neighborhood socioeconomic status (SES). Family addresses were geocoded and used
at seven assessments to identify corresponding neighborhood attributes from the 1990 U.S.
census (at the blockgroup level). In the current study, census measures of neighborhood SES
included percent of female-headed households, residents who are unemployed, in poverty, and
lack a high school degree. These data were available at the 1 month, 15 month, 36 month, 54
NEIGHBORHOODS AND SELECTION 8
month, first grade, third grade, and fifth grade assessments (mean α = .75, range: .71-.78).
Descriptive statistics for neighborhood variables are presented in Table 1.
Income to needs ratio. Family income was reported by mothers at each assessment.
Support from romantic partner. Mothers completed the intimacy subscale of the Love
and Relationships Scale (created for this study) at the following assessments: 1 month, 36
months, 54 months, first grade, and third grade. Six items were answered on a range from
1(strongly disagree) to 5(strongly agree). Sample items included “My spouse/partner can really
understand my hurts and joys” and “My spouse/partner listens to me when I need someone to
talk to” (α = .87).
Parenting behavior. Mothers were recorded when interacting independently with their
children during play scenarios and problem-solving tasks. Observations were obtained eight
times between the child’s birth and grade 5, and sent to a single site for coding. In order to
maintain a developmentally appropriate measure of sensitivity, indicators changed somewhat
over time (NICHD Early Child Care Research Network, 1997, 1998). At 6, 15, and 36 months,
sensitivity was the sum of three 4-point ratings: sensitivity to the child’s non-distress signals
(e.g., acknowledging the child's affect, contingent vocalizations by the mother, facilitating the
manipulation of an object or child movement), positive regard (e.g., speaking in a warm tone of
voice, hugging or other expressions of physical affection, an expressive face), and reflected
intrusiveness (e.g., taking away objects or food while the child still appears interested, not
allowing the child to handle toys he/she reaches for, insisting that the child do something in
which he/she is not interested, not allowing the child to make choices).
At 54 months and in grades 1, 3 and 5, sensitivity was the sum of three 7-point ratings:
supportive presence (e.g., pay attention to the child when the child talks, be engaged in the
NEIGHBORHOODS AND SELECTION 9
interaction, appear to enjoy interacting with the child), respect for autonomy (e.g., ask the child’s
opinion, negotiate rule with the child, acknowledge the child’s perspective), and reflected
hostility (e.g., point out child's weaknesses, put the child down, use a negative or sarcastic tone
of voice). Inter-coder reliability was established by having two coders assess approximately 20%
of the tapes, randomly drawn from each assessment period (rICC > .70). Confirmatory factor
analyses supported a single factor solution at each timepoint, with good fit and standardized
loadings above .40.
Attitudes about being a parent. Mothers reported on their beliefs about parenting three
times between the child’s birth and grade six using the Modernity scale (Schaefer & Edgerton,
1985). The 30 items on this scale provide an estimate of how progressive (democratic, child-
centered) versus traditional (authoritarian, strict, adult-centered) the parent’s attitudes are toward
child rearing and discipline. Sample items include “parents should teach their children that they
should be doing something useful at all times” (reflected), “parents should go along with the
game when their child is pretending something” and “children have a right to their own point of
view and should be allowed to express it.” Responses ranged from 1=strongly disagree to
5=strongly agree. Consistent with prior work using this scale (Dowsett, Huston, Imes, &
Gennetian, 2008), we combined the items into a single index, which had composite reliabilities
above .80 across all timepoints for both parents.
Mother personality. Mothers completed the 12-item scales for agreeableness (α = .74),
neuroticism (α = .84), and extraversion (α = .75) from the revised Neo Personality Inventory
(Costa & McCrae, 1992) when the child was 6 months old.
NEIGHBORHOODS AND SELECTION 10
Maternal mental health. Mothers’ depressive symptomatology was assessed using
questions from the Center for Epidemiologic Studies Depression Scale (Radloff, 1977) when
children were 1, 6 and 15 months, and in fifth grade (all α > .85).
Home environment. The 55-item HOME Inventory (Bradley & Caldwell, 1984) was
used to measure the quality and quantity of stimulation and support available at home when
children were 15 months old and again when children were in third grade (α > .80).
Child externalizing behavior problems. Mothers completed age-appropriate versions of
the Child Behavior Checklist (CBCL; Achenbach 1991) when children were 24 months, 36
months, 54 months, in Kindergarten, and in grades 1, 3, and 4. The T score standardization was
used in the current analyses.
Analyses
As shown in Figure 1, all hypothesis tests were conducted longitudinally. The seven
assessments of neighborhood disadvantage were fit into an autoregressive model, as were the
seven assessments of family income, and the measures of other investments (which varied in the
number of times they were measured). Separate models were run for each investment resource,
though they all included household income as a covariate, and externalizing problems as an
outcome. We used Mplus Version 6 (Muthén & Muthén, 2012) to estimate the models using full
information maximum likelihood estimation. Study hypotheses were evaluated using structural
equation models with model fit assessed using the standard chi-square index of statistical fit that
is routinely provided under maximum likelihood estimation of parameters, as well as practical fit
indices (Browne & Cudeck, 1993; Tucker & Lewis, 1973; Hu & Bentler, 1999). Models were
simplified by equating equivalent paths across time (when it did not result in a significant loss of
fit).
NEIGHBORHOODS AND SELECTION 11
Results
As shown in Table 1, the families in this study lived in neighborhoods reflecting a wide
range of income, employment, household structure, and education. Across the seven
assessments, 9.5% of the residents had an income-to-needs ratio lower than one, and that
estimate ranged across neighborhoods from 0% to 100%. Separate models were estimated for
each parent investment/investment resource. The fit indices from these six models are presented
on the left side of Table 2. For example, the model in which the parenting investment resource
was support from a romantic partner had an absolute fit of χ2(232) = 909.6, with acceptable
relative fit: RMSEA of .046, TLI = .957.
The standardized coefficients from these six analytic models appear in Table 3. For
example, the first model was the model in which the parenting investment resource was support
from a romantic partner. Information from this model appears in the first column of data. Path A
in Figure 1 reflects the hypothesis that neighborhood socioeconomic disadvantage will predict
parenting investment resources (both material and emotional). Consistent with this hypothesis, in
this first model the coefficients from neighborhood socioeconomic disadvantage to income were
all significant, and ranged in magnitude from -.12 to -.03. The coefficients from neighborhood
disadvantage to support from a romantic partner were not significant [range: .03 to .03]. Across
the remaining models neighborhood disadvantage frequently predicted lower household income,
and predicted emotional parenting investments two thirds of the time. The magnitudes of these
coefficients are small, because they represent change over a relatively brief period
(approximately 2 years). Over time, they would cumulate into a larger effect.
Path B in Figure 1 reflects the hypothesis that parenting investment resources (both
material and emotional) will predict neighborhood socioeconomic disadvantage. Consistent with
NEIGHBORHOODS AND SELECTION 12
this hypothesis, household income predicted changes in neighborhood socioeconomic
disadvantage consistently across all six models (100% of paths tested). Emotional investment
resources also predicted changes in neighborhood socioeconomic disadvantage (92% of paths
tested). The neighborhood data across these assessments all came from the same census year,
meaning that observed changes in neighborhood socioeconomic disadvantage only represent
families who relocated, not changes over time in neighborhood socioeconomic condition.
Path C1 and C2 in Figure 1 reflect the hypothesis that neighborhood disadvantage will
predict changes over time in child externalizing behavior problems, after accounting for parents’
emotional and material investment resources. Consistent with this hypothesis, neighborhood
disadvantage consistently predicted relative changes in externalizing behavior across all six
models (100% of paths tested).
Paths D1 and D2 in Figure 1 reflect the expectation that parenting investment resources
will predict changes over time in child externalizing behavior problems, net of neighborhood
socioeconomic disadvantage. Consistent with this expectation, relative changes in externalizing
problems were predicted by family income (100% of paths tested) and emotional investment
resources (100% of paths tested). To compare the magnitude of these coefficients, a model
constraint was applied which constrained the path from neighborhood disadvantage to
externalizing to be equal in magnitude (in absolute value) to the path from emotional investments
to externalizing. This model constraint significantly worsened model fit for all models except
observed sensitivity. Similar model constraints comparing neighborhood disadvantage to
household income significantly worsened model fit for all models except home environment.
Both emotional investment and material investment resources were more strongly linked than
neighborhood disadvantage to changes in child externalizing problems.
NEIGHBORHOODS AND SELECTION 13
Finally, inclusion of both income and emotional investments as longitudinal processes
allowed for an examination of reciprocal associations over time between those two constructs.
Household income was associated with increases in emotional investments (80% of paths tested),
and emotional investments predicted increases in household income (100% of paths tested).
Replication Sample
Participants
The Fragile Families and Child Well-being Study (Fragile Families), is a longitudinal
birth cohort study of 4,898 children born between 1998 and 2000 in 20 U.S. cities with
populations of 200,000 or more (Reichman et al., 2001). This study deliberately oversampled
births to unmarried parents, yielding a sample where one-quarter of births were to married
parents and three-quarters were to unmarried parents. The baseline interviews were conducted
shortly after the child’s birth with follow-up interviews one, three, five, and nine years later. This
study was conducted with approval from the affiliated Institutional Review Boards. Eighty-seven
percent of eligible mothers responded to the first wave of the survey, and subsequent response
rates have been above 70%. The sample is ethnically diverse: 47% Black, 21% White, 27%
Latino, and 5% other. Just over 80% of the mothers had been born in the U.S. Mother’s average
age was 25 years, and two thirds had at least graduated from high school.
Measures
Neighborhood socioeconomic status (SES). Family addresses were geocoded and used
at each assessment to identify corresponding neighborhood attributes from the 2000 U.S. census
(at the tract level). In the current study, census measures of neighborhood SES included percent
of female-headed households, residents who are unemployed, in poverty, and lack a high school
NEIGHBORHOODS AND SELECTION 14
degree (mean α = .88, range: .87-.89). Descriptive statistics for neighborhood variables are
presented in Table 1.
Income to needs ratio. Family income was reported by mothers at each assessment.
Support from romantic partner. Support from the mother’s romantic partner was
assessed at year one and year three using a 5-item scale created for this study. Items were
answered on a scale from 1(never) to 3(often). Sample items include “How often is your partner
fair/willing to compromise when you have a disagreement” and “How often does your current
partner encourage you or help with things that are important to you” (α = .60 at year 1, .60 at
year 3).
Parenting behavior. At years two, three, and five mothers reported on the degree to
which their child was spanked by the mother, father, or mother’s romantic partner. Two
questions were answered in reference to each caregiver: “Has [caregiver] spanked the child in the
past month” 0(no) and 1(yes) and “how often did [caregiver] spank the child in the past month”
1(once or twice), 2(a few times in the past month), 3(a few times each week) and 4(every day).
From these questions an overall scale was created for each parent ranging from 0(did not spank
child) to 5(spanked every day), then combined across caregivers (α = .66 at year 2, .68 at year 3,
.62 at year 5).
Attitudes about being a parent. At years two, three, and five mothers completed a four-
item scale created for this study which assessed the mother’s impression of being a parent. Items
were answered on a scale from 1(strongly disagree) to 4(strongly agree) and included items such
as “Taking care of children is more work than pleasure” and “would be doing better in life
without child” (mean α = .64, range: .61-.66).
NEIGHBORHOODS AND SELECTION 15
Mother personality. At year three mothers completed six items assessing impulsivity (α
= .84) from the Impulsivity scale (Dickman, 1990).
Maternal mental health. Maternal depression was reported by the mother at years two,
three, and five. The Fragile Families draws on the Composite International Diagnostic Interview
- Short Form (CIDI-SF), Section A (Kessler et al. 1998) to assess depression. The CIDI is a
standardized instrument for assessment of mental disorders and follows the criteria of the
Diagnostic and Statistical Manual of Mental Disorders – Fourth Edition (DSM-IV; APA 1994).
The CIDI-SF takes a portion of the full set of CIDI questions and estimates the probability that a
respondent would be positively diagnosed with depression if given the full CIDI interview.
Specifically, women were asked if they had experienced feelings of depression or anhedonia in
the past year, lasting two weeks or more. If so, they were asked about seven additional
symptoms: (1) losing interest (2) feeling tired (3) weight changes (4) trouble with sleep (5)
trouble concentrating (6) feeling worthless or (7) thinking about death. Symptoms were summed
into a scale of depressive symptomatology.
Home environment. At year three and year five mothers completed a 29-item version of
the HOME Inventory (Bradley & Caldwell, 1984) with items measuring parental warmth, verbal
skills, lack of hostility, and physical interior of home (α = .82 at year 3, .84 at year 5).
Child externalizing behavior problems. At year 5 mothers completed a 14-item version
of the Child Behavior Checklist (CBCL; Achenbach 1991), and a 38-item version at year 9.
Results
As shown on the right side of Table 1, the families in this study lived in neighborhoods
reflecting a wide range of income, employment, household structure, and education. Across the
five assessments, 17.4% of the families in these neighborhoods were living below the poverty
NEIGHBORHOODS AND SELECTION 16
line, and that estimate ranged across neighborhoods from 0% to 100%. Separate models were
estimated for each parent investment/investment resource. The fit indices from these six models
are presented on the right side of Table 2. For example, the model in which the parenting
investment resource was support from a romantic partner had an absolute fit of χ2(53) = 862.2,
with acceptable relative fit: RMSEA of .056, TLI = .907.
The standardized coefficients from these six analytic models appear in Table 4. Although
neighborhood disadvantage continued to predict relative changes in family income (83% of paths
tested), it was less frequently linked with changes in emotional investment resources across
models (43% of paths tested) than it was in the SECCYD. Income consistently predicted relative
changes in neighborhood disadvantage (96% of paths tested), as did emotional investment
resources (64% of paths tested). The neighborhood data across these assessments all came from
the same census year, meaning that observed changes in neighborhood socioeconomic
disadvantage only represent families who relocated, not changes over time in neighborhood
socioeconomic condition. Externalizing behavior problems was predicted by neighborhood
disadvantage (67% of paths tested), household income (83% of paths tested), and emotional
investment resources (100% of paths tested).
Similar to Study 1, model constraints were applied which constrained the path from
neighborhood disadvantage to externalizing to be equal in magnitude (in absolute value) to the
path from emotional investments to externalizing. This model constraint significantly worsened
model fit for all models except home environment. Similar model constraints comparing
neighborhood disadvantage to household income significantly worsened model fit for all models
except home environment. Both emotional investment and material investment resources (when
NEIGHBORHOODS AND SELECTION 17
significant) were more strongly linked than neighborhood disadvantage to changes in child
externalizing problems.
Discussion
Overall, as hypothesized, neighborhood SES predicted change over time in parent
investment resources. Support for this hypothesis was more frequent for family income than for
emotional parent investments. It may be the case that neighborhood context more consistently
offers opportunities or resources for changing income than opportunities or resources for
changing emotional investments. Similar findings have been documented regarding parent SES
and parent investments. For instance, after accounting for selection effects, parent SES predicts
material investments but not emotional investments (Schofield et al., 2011). Of course, it may be
that different elements of the family environment not included in the current study would be
more affected by neighborhood socioeconomic disadvantage. A third possibility is that the
structural characteristics which influence these individual and family level attributes (Kling et
al., 2001) are not reflected in aggregate census data.
There was support for selection into neighborhoods based on both material and emotional
investment resources. In both samples, people moved into (or out of) neighborhoods based on the
six emotional investment resources, as well as the material investment resource (i.e., household
income). Although this is consistent with previous research on personal characteristics and
selection into neighborhoods (Jokela, 2009), it is important that our findings not be interpreted to
suggest that individuals choose neighborhoods based on individual characteristics alone. There
are many additional structural determinants that impact decision-making, and an extensive body
of literature in the field of social epidemiology (Berkman, Kawachi, & Glymour, 2014) around
the social construction of health and health inequities that goes beyond individual characteristics
NEIGHBORHOODS AND SELECTION 18
as critical drivers of relocation (Arroyo, 2006). For example, while education is one potential
consideration for choosing to live in a particular neighborhood, high quality education is not
equitably available to all. High poverty neighborhoods receive less funding per student than low
poverty neighborhoods, which can contribute to high teacher turnover rates, poorer educational
facilities, and other impacts (Arroyo, 2006; Condron & Roscigno, 2003). One study showed that
barriers to education may be even greater for African American children because of harsher,
more negative disciplinary responses due, in part, to racial stereotypes by teachers and
administrators (Okonofua & Eberhardt, 2015). Policies and structural inequities that that inhibit
moving out of impoverished neighborhoods are certainly not limited to education; additional
challenges for people living in poverty include, among others, reduced access to affordable
housing, banking services, and employment opportunities (Metzler et al., in press). As such,
future efforts in the exploration between neighborhood attributes and health should continue to
integrate a health equity perspective in drafting research aims, generating hypotheses, and
drawing conclusions. Nevertheless, selection effects based on individual or family-level
characteristics are real and merit continued study.
This nonexperimental data cannot provide strong causal inference. Although both
samples showed variation in most neighborhood characteristics, it remains possible that other
samples could provide a different pattern of results. We would note, however, that the current
two datasets offered us rich measurement of family and personal characteristics, which are
necessary to adequately test for selection effects. Third, although these analyses covered the
years from birth to late childhood, it is possible that outcomes measured in later periods of
development would provide less support for our hypotheses. Some family environment measures
had comparatively modest reliability; we would expect larger effects with more reliable
NEIGHBORHOODS AND SELECTION 19
measures. Finally, our focus was on socioeconomic disadvantage, and a different pattern of
results could emerge for indicators of neighborhood social organization, like collective efficacy.
The neighborhood environment likely affects resident behavior. At the same time,
residents are not randomly sorted into neighborhoods. Studies of neighborhood effects should
include longitudinal tests of both processes (i.e., selection and neighborhood effects) in order to
make justifiable claims about which process is operating. Inferring causal effects from cross-
sectional data on neighborhoods and family processes will likely conflate the causal effect of
neighborhood environment with previous selection into the neighborhood environment.
NEIGHBORHOODS AND SELECTION 20
References
Arroyo, C. G. (2006). The funding gap. The Education Trust Web site.
http://1k9gl1yevnfp2lpq1dhrqe17.wpengine.netdna-cdn.com/wp-
content/uploads/2013/10/FundingGap2006.pdf Accessed July 12, 2015.
Bagley, M. N. & Mokhtarian, P. L. (2002). The impact of residential neighborhood type on
travel behavior: A structural equations modeling approach. Annals of Regional Science,
36, 279-297.
Batty, E. and Flint, J. (2010) Self esteem, comparative poverty and neighbourhoods: A
working paper. Produced as part of Living Through Change in Challenging
Neighbourhoods Study. Joseph Rowntree Foundation: York. Available at:
http://research.shu.ac.uk/cresr/living-through-change/reports.html
Berkman, L.F., Kawachi, I., & Glymour, M. (2014). Social Epidemiology. (Ed.). New York,
NY: Oxford University Press.
Berry, T.R., Spence, J.C., Blanchard, C., Cutumisu, N., Edwards, J., & Nykiforuk, C. (2010).
Changes in BMI over 6 years: the role of demographic and neighborhood characteristics.
International Journal of Obesity, 34, 1275-83. doi: 10.1038/ijo.2010.36.
Berry, T.R., Spence, J.C., Blanchard, C.M., Cutumisu, N., Edwards, J., & Selfridge, G.
(2010). A longitudinal and cross-sectional examination of the relationship between
reasons for choosing a neighbourhood, physical activity and body mass index.
International Journal of Behavior Nutrition and Physical Activity, 5, 57. doi:
10.1186/1479-5868-7-57.
Bradley, R. H,. & Caldwell, B. M. (1977). Home observation for measurement of the
NEIGHBORHOODS AND SELECTION 21
environment: a validation study of screening efficiency. American Journal of Mental
Deficiency, 81, 417-420.
Bradley, R.H., & Caldwell, B.M. (1984). The relation of infants' home environments to
achievement test performance in first grade: a follow-up study. Child Development, 55,
803-809. PMID: 6734319
Bronfenbrenner, U., & Morris, P. A. (2006). The bioecological model of human
development. In R. M. Lerner & W. Damon (Eds.), Handbook of Child Psychology: Vol.
1. Theoretical models of human development (6th ed., pp. 793–828). Hoboken, NJ:
Wiley.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A.
Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury
Park, CA: Sage.
Cao, X., Mokhtarian, P. L., & Handy, S. L. (2009). Examining the impacts of residential self-
selection on travel behavior: A focus on empirical findings. Transport Reviews, 29, 359-
395.
Cohen, D. A., Mason, K., Bedimo, A., Scribner, R., Basolo, V., & Farley, T. A. (2003).
Neighborhood physical conditions and health. American Journal of Public Health, 93,
467-471.
Conger, R. D., Patterson, G. R., & Ge, X. (1995). It takes two to replicate: A
mediational model for the impact of parents’ stress on adolescent adjustment. Child
Development, 66, 80-97.
Costa, P. T., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO-PI-R) and
NEIGHBORHOODS AND SELECTION 22
NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological
Assessment Resources.
Cowan, C. P., Cowan, P. A., & Barry, J. (2011). Couples' groups for parents of preschoolers:
Ten-year outcomes of a randomized trial. Journal of Family Psychology, 25, 240-250.
Derogatis, L.R. (1983). SCL–90–R: Administration, scoring, and procedures manual II.
Baltimore: Clinical Psychometric Research.
Dickman, S.J. (1990) Functional and Dysfunctional Impulsivity: Personality and Cognitive
Correlates. Journal of Personality and Social Psychology, 58, 95-102.
Digman, J.M. (1997). Higher-order factors of the Big Five. Journal of Personality and
Social Psychology, 73, 1246-1256. PMID: 9418278
Diez-Roux, A. V., Nieto, F. J., Muntaner, C., Tyroler, H. A., Comstock, G. W., Shahar, E.,
Cooper, L. S., Watson, R. L., & Szklo, M. (1997). Neighborhood environments and
coronary heart disease: A multilevel analysis. American Journal of Epidemiology, 146,
48-63.
Ecob, R., & Macintyre, S. (2000). Small area variations in health related behaviours; do
these depend on the behaviour itself, its measurement, or on personal characteristics?
Health Place, 6, 261-274. PMID: 11027952
Erel, O., & Burman, B. (1995). Interrelatedness of marital relations and parent-child relations: A
meta-analytic review. Psychological Bulletin, 118, 108-132. doi: 10.1037/0033-
2909.118.1.108
Frank, J. D., Saelens, B. E., Powell, K. E., & Chapman, J. E. (2007). Stepping towards causation:
Do built environments or neighborhood and travel preferences explain physical activity,
driving, and obesity? Social Science and Medicine, 65, 1898-1914.
NEIGHBORHOODS AND SELECTION 23
Gans, H. J. (1968). People and Plans: Essays on Urban Problems and Solutions. New
York: Basic Books.
Garner, DM, & Garfinkel, PE. (1979). The Eating Attitudes Test: an index of the symptoms
of anorexia nervosa. Psychological medicine, 9, 273-9. PMID: 472072
Hedman, L. & van Ham, M. (2010). Understanding Neighbourhood Effects: Selection Bias
and Residential Mobility. In Neighbourhood Effects Research: New Perspectives (pp. 79-
99). Springer.
Hetling, A., & Zhang, H. (2010). Domestic violence, poverty, and social services: does location
matter? Social Science Quarterly, 91, 1144-1163.
Hickman, P. (2010). Understanding residential mobility and immobility in challenging
neighborhoods: Research paper number 8. Centre for Regional Economic and Social
Research. Sheffield Hallam University. Available at:
http://research.shu.ac.uk/cresr/living-through-
change/documents/RP8_ResidentialMobility.pdf
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural equation modeling, 6, 1-55.
Janz, K.F., Witt, J. & Mahoney, L. T. (1995). The stability of children's physical activity as
measured by accelerometry and self report. Medical Science and Sports Exercise, 9,
1326-1332. PMID: 8531633
Jokela, M. (2009). Personality predicts migration within and between US states. Journal of
Research on Personality, 43, 79–83.
Jokela, M., Bleidorn, W., Lamb, M. E., Gosling, S. D., & Rentfrow, P. J. (2014). Geographically
NEIGHBORHOODS AND SELECTION 24
varying associations between personality and life satisfaction in the London metropolitan
area. Proceedings of the National Academy of Sciences, 112, 725-730.
Keegan, T.H., Hurley, S., Goldberg, D., Nelson, D.O., Reynolds, P., Bernstein, L., Horn-
Ross, P.L., & Gomez, S.L. (2012). The association between neighborhood characteristics
and body size and physical activity in the California teachers study cohort. American
Journal of Public Health, 102, 689-697. doi: 10.2105/AJPH.2011.300150
Kiecolt-Glaser, J. K.,& Newton, T. L. (2001). Marriage and health: His and hers. Psychological
Bulletin, 127, 472-503.
King, W. C., Brach, J. S., Belle, S., Killingsworth, R., Fenton, M., et al. (2003). The relationship
between convenience of destinations and walking levels in older women. American
Journal of Health Promotion, 18, 74-82.
King, A. C., Toobert, D., Ahn, D., Resnicow, K., Coday, M., et al. (2006). Perceived
environments as physical activity correlates and moderators of intervention in five
studies. American Journal of Health Promotion, 21, 24-35.
Kling, J. R., Liebman, J. B., & Katz, L. F. (2001). Bullets don’t got no name: Consequences of
fear in the ghetto. Joint Center for Poverty Research, Working Paper 255. Available at
http://www.jcpr.org/wpfiles/kling_liebman_latz.PDF?CFID=1802590&CFTOKEN=5318
0781
Kobasa, S. C. (1979). Stressful life events, personality, and health: An inquiry into hardiness.
Journal of Personality and Social Psychology, 37, 1-11.
Landgraf, J.M., Abetz, L., & Ware, J.E. (1996). The CHQ Users’ Manual (1st edition).
Boston: The Health Institute, New England Medical Center.
Leventhal, T., & Brooks-Gunn, J. (2000). The neighborhoods they live in: The effects of
NEIGHBORHOODS AND SELECTION 25
neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126,
309-337.
Leventhal, T., Fauth, R.C., & 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, 933-52.
Ludwig, J., Duncan, G. J., Gennetian, L. A., Katz, L. F., Kessler, R. C., Kling, J. R., &
Sanbonmatsu, L. (2013). Long-term neighborhood effects on low-income families:
Evidence from Moving to Opportunity. American Economic Review Papers and
Proceedings, 103, 226-231.
Macintyre, S., Ellaway, A., & Cummins, S. (2002). Place effects on health: How can we
conceptualize, operationalize and measure them? Social Science and Medicine, 55, 125-
139.
McCabe, J. E. (2014). Maternal personality and psychopathology as determinants of parenting
behavior: A quantitative integration of two parenting literatures. Psychological Bulletin,
140, 722-750. doi: 10.1037/a0034835
Mee, K. (2007) “I ain’t been to heaven yet? Living here, this is heaven to me”: Public
housing and the making of home in inner Newcastle. Housing, Theory and Society, 24,
207-228.
McKinnish, T. & White, T. K. (2011). Who Moves to Mixed-Income Neighborhoods? Regional
Science and Urban Economics, 41, 187–195. doi: 10.1016/j.regsciurbeco.2011.01.007
Metzler, M., Merrick, M. T., Klevens, J. Ford, D., & Ports, K. A. (In press). Adverse Childhood
Experiences and Life Opportunities: Time to Shift the Narrative? Children and Youth
Services Review.
NEIGHBORHOODS AND SELECTION 26
Murray, A. L. & Booth, T. (2015). Personality and physical health. Current opinion in
psychology, 5, 50-55.
NICHD Early Child Care Research Network. (2005). Child care and child development:
Results from the NICHD Study of Early Child Care and Youth Development. New York,
NY: Guilford Press.
Norman, G. J., Carlson, J. A., O’Mara, S., Sallis, J. F., Patrick, K., Frank, L. D., & Godbole, S.
V. (2013). Neighborhood preference, walkability, and walking in overweight/obese men.
American Journal of Health Behavior, 37, 277-282.
Østbye, T., Malhotra, R., Stroo, M., Lovelady, C., Brouwer, R., Zucker, N., & Fuemmeler, B.
(2013). The effect of the home environment on physical activity and dietary intake in
preschool children. International Journal of Obesity, 37, 1314-1321.
Pabayo, R., Belsky, J., Gauvin, L., & Curtis, S. (2011). Do area characteristics predict
change in moderate-to-vigorous physical activity from ages 11 to 15 years? Social
Science and Medicine, 72, 430-438. doi: 10.1016/j.socscimed.2010.09.039.
Petersen, A.C., Crockett, L., Richards, M., & Boxer, A. (1988). A self-report measure of
pubertal status: Reliability, validity, and initial norms. Journal of Youth and Adolescence,
17, 117-133. doi: 10.1007/BF01537962.
Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the
general population. Applied Psychological Measurement, 1, 385-401.
Ross, C.E., & Mirowsky, J. (2008). Age and the balance of emotions. Social Science and
Medicine, 66, 2391-400. doi: 10.1016/j.socscimed.2008.01.048.
Sampson, R. J., & Sharkey, P. (2008). Neighborhood selection and the social reproduction
of concentrated racial inequality. Demography, 45, 1-29.
NEIGHBORHOODS AND SELECTION 27
Sanbonmatsu, L., Kling, J. R., Duncan, G. J., & Brooks-Gunn J. (2006). Neighborhoods and
academic achievement: Results from the Moving to Opportunity experiment. Journal of
Human Resources, 41, 649-691.
Scarr, S. & McCartney, K. (1983). How people make their own environments: A theory of
genotype environment effects. Child Development, 54, 424-435. PMID: 6683622
Schaefer, E. S., & Edgerton, M. (1985). Parent and child correlates of parental modernity. In I. E.
Siegel (Ed.), Parental belief systems (pp. 287–318). Hillsdale, NJ: Erlbaum.
Sciandra, M., Sanbonmatsu, L., Duncan, G.J., Gennetian, L.A., Katz, L.F., Kessler, R.C.,
Kling, J.R., & Ludwig, J. (2013). Long-term effects of the Moving to Opportunity
residential mobility experiment on crime and delinquency. Journal of Experimental
Criminology, 9. doi: 10.1007/s11292-013-9189-9.
Smith, K. R., Zick, C. D., Kowaleski-Jones, L., Brown, B. B., Fan, J. X., & Yamada, I. (2011).
Effects of neighborhood walkability on healthy weight: Assessing selection and causal
influences. Social Science Research, 40, 1445-1455.
Stevens, R. B., & Brown, B. B. (2011). Walkable new urban LEED_neighborhood-development
(LEED_ND) community design and children’s physical activity: Selection,
environmental, or catalyst effects? Physical Activity, 8.
Thornberry, T., Huzinga, D., & Loeber, R. (1989). Measures developed for multisite study
on causes and correlates of delinquency. Unpublished manuscript.
Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum likelihood factor
analysis. Psychometrika, 38, 1–10.
Veit, C.T., & Ware, J.E. (1983). The structure of psychological distress and well-being in
general populations. Journal of Consulting and Clinical Psychology, 51, 730-42.
NEIGHBORHOODS AND SELECTION 28
Weisz, J. R. (1978). Transcontextual validity in developmental research. Child
Development, 49, 1-12.
Wells, N. M., Evans, G. W., & Yang, Y. (2010). Environments and health: Planning decisions as
public-health decisions. Journal of Architectural and Planning Research, 27, 124–143.
Yen, I.H., Michael, Y.L., & Perdue, L. (2009). Neighborhood environment in studies of
health of older adults: a systematic review. American Journal of Preventive Medicine, 37,
455-463. doi: 10.1016/j.amepre.2009.06.022.
Yu, C., & Zhu, X. (2013). Impacts of residential self-selection and built environments on
children’s walking to school behaviors. Environment and Behavior, 47, 268-287.
Zhou, Y. (2006). Spatial Analysis of Substantiated Child Maltreatment in Metro Atlanta,
Georgia. Thesis, Georgia State University.
http://scholarworks.gsu.edu/geosciences_theses/7
NEIGHBORHOODS AND SELECTION 29
Table 1
Neighborhood Attributes Across Both Studies
Early Child Care and
Youth Development
Fragile Families Project
Percent of: M SD Min Max M SD Min Max
Residents with income-to-needs ratio less than 1.00 9.5 10.2 0 100
- - - -
Families below poverty level - - - -
17.4 13.0 0 100
Residents who are unemployed 4.6 4.7 0 51
10.00 7.2 0 66
Households headed by single mothers with children under 18 9.1 9.2 0 76
20.2 13.0 0 100
Residents over 25 without a high school degree 17.6 13.6 0 91
71.3 15.1 28 100
NEIGHBORHOODS AND SELECTION 30
Table 2
Fit Indices From Final Models Across Both Studies
SECCYD
Fragile Families Study
Emotional investment/resource χ2 df RMSEA TLI
χ2 df RMSEA TLI
Support from romantic partner 909.6 232 0.046 0.957
862.2 53 0.056 0.907
Parenting behavior 1069.8 250 0.049 0.949
509.6 57 0.04 0.949
Attitude about parenting 779.9 194 0.047 0.964
927.1 54 0.057 0.904
Parent personality 672.2 162 0.048 0.965
358.8 38 0.042 0.955
Parent depression 1060.6 256 0.048 0.952
796.7 69 0.046 0.93
Home environment 1023.3 230 0.05 0.952
404.1 56 0.037 0.961
Note. SECCYD= Study of Early Child Care and Youth Development, RMSEA= Root mean square error
of approximation, TLI = Tucker Lewis index
NEIGHBORHOODS AND SELECTION 31
Table 3
Standardized Coefficients From Models of Investments, Neighborhoods, and Externalizing (Study of Early Child Care and Youth Development)
Support Observed Parenting Agreeable- Maternal Home
Path from RP sensitivity attitudes ness depression environment
A. Neighborhood→income [-.12,-.03]100 [-.12,-.02]100 [-.04,-.01]17 [-.12,-.02]100 [-.12,-.02]100 [-.02,-.03]100
A. Neighborhood→emotional investment [.00,.00]0 [.00,-.20]80 [.02,-.09]50 - [.04,.05]100 [-.16,-.12]100
B. Income→neighborhood [-.08,-.05]100 [-.07,-.05]100 [-.06,-.04]100 [-.05,-.08]100 [-.05,-.08]100 [-.07,-.04]100
B. Emotional investment→neighborhood [-.02,.01]67 [-.07,.02]83 [-.13,-.04]100 [-.03,-.03]100 [-.04,-.05]100 [-.06,-.03]100
C. Neighborhood→externalizing [.03,.04]100 [.03,.04]100 [.02,.03]100 [.02,.03]100 [.02,.03]100 [.02,.03]100
D. Income→externalizing [-.11,-.03]100 [-.14,-.03]100 [-.09,-.03]100 [-.13,-.03]100 [-.10,-.02]100 [-.03,-.03]100
D. Emotional investments→externalizing [-.07,-.05]100 [-.03,-.02]100 [-.14,-.03]100 [-.10,-.09]100 [.08,.23]100 [-.07,-.05]100
Stability in neighborhood [.72,.87]100 [.73,.86]100 [.72,.87]100 [.72,.86]100 [.73,.87]100 [.72,.86]100
Stability in income [.64,.94]100 [.63,.92]100 [.64,.93]100 [.64,.93]100 [.63,.94]100 [.63,.93]100
Stability in emotional investment [.53,.63]100 [.30,.48]100 [.72,.81]100 - [.41,.51]100 [.46,.59]100
Income→emotional investment [.02,.02]0 [.08,.23]100 [.04,.04]100 - [-.04,-.05]100 [.14,.18]100
Emotional investment→income [.02,.03]100 [.03,.04]100 [.03,.04]100 [.02,.02]100 [-.02,-.02]100 [.03,.04]100
Note. [min, max]# = percent of paths over time measuring this relation that were statistically significant. RP=romantic partner. Boldface = p < .05
NEIGHBORHOODS AND SELECTION 32
Table 4
Standardized Coefficients From Models of Investments, Neighborhoods, and Externalizing (Fragile Families Study)
Support Parent Parenting
Maternal Home
Path from RP harshness attitudes Impulsivity depression environment
A. Neighborhood→income [-.26,-.01]75 [-.21,-.01]75 [-.11,-.01]75 [-.23,-.04]100 [-.12,-.02]100 [-.22,-.00]75
A. Neighborhood→emotional investment [.03,.03]0 [-.01,.03]67 [.02,-.09]50 - [.01,.01]0 [-.07,-.07]100
B. Income→neighborhood [-.27,-.04]100 [-.24,-.04]100 [-.23,-.03]100 [-.23,-.04]100 [-.05,-.08]100 [-.20,-.00]75
B. Emotional investment→neighborhood [-.04,.00]50 [-.02,-.01]0 [-.13,-.04]100 [.02,.02]100 [.01,.04]33 [-.16,-.16]100
C. Neighborhood→externalizing [.03,.03]100 [.03,.03]100 [.02,.02]00 [.00,.00]00 [.04,.04]100 [.04,.04]100
D. Income→externalizing [-.10,-.03]67 [-.07,-.02]67 [-.08,-.02]100 [-.08,.01]67 [-.07,-.07]100 [-.06,-.06]100
D. Emotional investments→externalizing [-.07,-.05]100 [.14,.13]100 [-.20,-.12]100 [.25,.17]100 [.09,.09]100 [-.05,-.05]100
Stability in neighborhood [.01,.80]75 [.72- .90] [.72- .90] [.72- .90] [.72- .90] [.72- .90]
Stability in income [.31,.64]100 [.80-1.13] [.81-1.13] [.81-1.14] [.81-1.13] [.81-1.13]
Stability in emotional investment [.17,.17]100 [.38,.48]100 [.51- .58]100 - [.33- .38]100 [.32- .32]100
Income→emotional investment [-.01,.00]0 [-.08,.02]67 [.04,.04]100 - [-.07,-.07]100 [.21,.21]100
Emotional investment→income [.03,.07]100 [.02,.02]100 [-.02,.05]33 [.03,.03]100 [-.03,-.03]100 [.06,.06]100
Note. [min, max]# = percent of paths over time measuring this relation that were statistically significant. RP=romantic partner. Boldface = p < .05
NEIGHBORHOODS AND SELECTION 33
Neighborhood
SES
Neighborhood
SES
Parent
investments
Parent
investments
βB
βA
Child
externalizing
Child
externalizing
βC1 βC2
βD2βD1
Figure 1. Theoretical Model of Reciprocal Associations between Neighborhood Context and Parent Investment