HOUSING INSTABILITY AMONG HEAD START FAMILIES: THE ROLE OF PARENTING
PRACTICES, PARENTAL MENTAL HEALTH, AND CLASSROOM QUALITY ON
CHILDREN’S ACADEMIC AND SOCIAL-EMOTIONAL FUNCTIONING
By
Samanta Boddapati

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
School Psychology- Doctor of Philosophy
2017

ABSTRACT
HOUSING INSTABILITY AMONG HEAD START FAMILIES: THE ROLE OF PARENTING
PRACTICES, PARENTAL MENTAL HEALTH, AND CLASSROOM QUALITY ON CHILDREN’S
ACADEMIC AND SOCIAL-EMOTIONAL FUNCTIONING

By
Samanta Boddapati
National estimates indicate that young children account for a significant proportion of
people who are unstably housed (Bassuk, DeCandia, Beach & Berman, 2014). Housing
instability refers to a range of conditions that includes frequent residential mobility, living
doubled-up, and homelessness (Cunningham, Harwood & Hall, 2010). Emerging research has
demonstrated that instability during early childhood may affect long-term functioning, especially
for children living in poverty (Fowler, Henry, Schoeny, Taylor & Chavira, 2014; Ziol-Guest &
McKenna, 2014). Therefore, a primary goal of this secondary data analysis study is to compare
the pre-kindergarten outcomes of children who faced housing instability on important academic
and behavioral skills.
Beyond the effects of early housing instability on children, aspects of the social context,
such as parenting practices and quality of care at preschool, can have a beneficial influence on
children’s academic and social-emotional functioning (e.g., Herbers, Cutuli, Supkoff, Heistad,
Chan, Hinz, Masten, 2011). However, parental mental health challenges, particularly depression,
are higher among caregivers facing housing instability, which in turn can jeopardize parenting
practices that have well noted effects on child functioning (Suglia, Durarte & Sandel, 2011).
Despite these risks, positive environments beyond housing, such as high quality Head Start
classrooms, can serve as protective factors for children facing housing instability (Herbers et al.,
2011; Militois, Sesma, and Masten 1999; Pianta, Howes, Bryant, Clifford, Early & Barbarin,
2005). Therefore, another important aim of this study is to better understand the complex

relations between parental depression symptomology, parenting practices, classroom quality at
Head Start, and children’s functioning, among families who faced housing instability at Head
Start entry. Specifically, this study tested differences in a moderated-mediation model that aimed
to understand whether classroom quality at Head Start served as a moderator by interacting with
parenting in the noted relation between parental depression and child functioning through
parenting practices.
With the exception of mean level differences in parent engagement (.91 [z = - 2.98]),
multiple group structural equation modeling revealed no significant differences between children
and families who experienced housing instability during the Head Start years and those who
were stably housed. Although evidence that classroom quality mitigated risk was not present, the
findings did suggest a differential trend in the relation between classroom quality and children’s
social-emotional functioning between groups in the non-multiple group analyses. Overall, the
results also suggested the important role of parenting approaches during the pre-kindergarten
year for all Head Start children. Findings highlight potential unique differences between unstably
housed and stably housed children and future directions in research on the role of housing
instability during the pre-school years.

ACKNOWLEDGEMENTS

I am deeply grateful to a number of individuals who have contributed to this dissertation.
First, I would like to sincerely thank my advisor, Dr. Evelyn Oka, for her support, patience, and
mentorship. Not only do I appreciate her warm and thoughtful approach to mentorship, but I am
grateful to her for pushing me to think more critically about my ideas. She went above and
beyond as a mentor and chair, especially during personally trying times over the last six years. I
am truly grateful to her unwavering dedication and faith in me throughout the dissertation
process, even during times of self-doubt. Her guidance has undoubtedly been critical to my
development as a professional.
Second, I would like to thank the additional members of dissertation committee, Dr.
Dorinda Carter Andrews, Dr. Kristin Rispoli, and Dr. Cary Roseth for their patience,
constructive feedback, and time. I would like to thank Dr. Carter Andrews for challenging me to
think critically about issues of equity for which I am deeply passionate. I am grateful to Dr.
Rispoli for her thoughtful perspective, as well as her vast knowledge of the parent engagement
literature and statistics. I would like to thank Dr. Roseth for pushing me to think deeper about
child development throughout my years at MSU and for sharing his expertise in statistics.
I would also like to thank several family and friends. My graduate career would not have
been possible without my parents, Sam and Sobha Boddapati, who have always pushed me to do
my best and served as role models for strength and resilience throughout my life. I am thankful
to my sister, Neeta Boddapati, for the advice, laughter, weekend adventures, and listening ear. I
am grateful to my supportive in-laws, Cristina and Gerardo De Anda Sr., for their continuous
encouragement and praise. I would also like to thank my dear friend and colleague, Mohammed

iv

Palejwala, for his willingness to discuss statistical concepts over many lunches and dinners.
Finally, words cannot express my gratitude, admiration, and respect for my loving
husband, Jerry (Gerardo) De Anda, who has provided tireless support, encouragement, and an
endless supply of chocolate to make my dreams a reality. I am truly moved by and forever
grateful to his commitment to my career and the sacrifices he has made over the last six years to
share this journey with me.

v

TABLE OF CONTENTS

LIST OF TABLES ......................................................................................................................... ix
LIST OF FIGURES………………………………………………………………...…………......x
CHAPTER 1 ....................................................................................................................................1
INTRODUCTION ...........................................................................................................................1
Purpose of Study ..............................................................................................................................5
Research Questions .........................................................................................................................6
Theoretical Frameworks ..................................................................................................................6
Family Stress Model of Economic Hardship .............................................................................7
Risk and Resilience ....................................................................................................................8
CHAPTER 2 ..................................................................................................................................10
LITERATURE REVIEW ..............................................................................................................10
Importance of Housing Instability During Pre-School Years ........................................................10
Housing Instability: Definitions, Prevalence, and Historical Context ...........................................12
Prevalence ................................................................................................................................13
Historical Context and Trends .................................................................................................14
Caregiver and Family Factors Related to Housing Instability .......................................................15
Housing Instability and Poverty...............................................................................................15
Mental Health...........................................................................................................................16
Physical Health. .......................................................................................................................17
Psychosocial Stressors .............................................................................................................18
Housing Instability and Child Outcomes .......................................................................................18
Academic Functioning .............................................................................................................19
Social-Emotional Functioning .................................................................................................20
Related Health and Developmental Outcomes ........................................................................22
Parenting in The Context of Housing Instability ...........................................................................22
Parent Engagement ..................................................................................................................23
Parenting Approaches ..............................................................................................................26
Parental Mental Health as a Mechanism of Group Differences ..............................................30
Head Start and Housing Instability ................................................................................................32
Short and Long-Term Head Start Outcomes ...........................................................................33
Classroom Quality ...................................................................................................................35
Quality of care as protective process .......................................................................................38
Parenting and Classroom Quality ..................................................................................................39
Present Study .................................................................................................................................40
Research Questions and Hypotheses .......................................................................................42
CHAPTER 3 ..................................................................................................................................50
METHOD ......................................................................................................................................50
F.A.C.E.S Sampling Procedures ...................................................................................................51

vi

Attrition ..........................................................................................................................................52
Weighting and Design Effects .......................................................................................................53
Sample Characteristics ...................................................................................................................53
Full F.A.C.E.S Sample.............................................................................................................53
Pre-kindergarten Sample ..........................................................................................................54
Unstably Housed Sample .........................................................................................................56
Missing Data ..................................................................................................................................60
Variables and Measures .................................................................................................................61
Housing Status .........................................................................................................................62
Child Functioning ....................................................................................................................63
Pre-literacy and language skills .........................................................................................63
Social-emotional ................................................................................................................65
Teacher ratings ...................................................................................................................66
Parent ratings .....................................................................................................................67
Parenting Practices ...................................................................................................................67
Parent approaches ..............................................................................................................68
Warmth ..................................................................................................................68
Authoritative ..........................................................................................................68
Authoritarian ..........................................................................................................69
Energy ....................................................................................................................69
Parent engagement .................................................................................................................69
Parent/child activities ............................................................................................70
Parent Depression ...................................................................................................................70
Classroom Quality ..................................................................................................................71
Covariates ...............................................................................................................................72
Family economic risk............................................................................................72
Age ........................................................................................................................72
Maternal race ........................................................................................................73
Child’s gender .......................................................................................................73
Analytic Plan..................................................................................................................................73
Data Preparation and Cleaning ................................................................................................73
Preliminary Analyses ...............................................................................................................74
Confirmatory Factor Analysis..................................................................................................74
Structural Equation Modeling with Full Sample .....................................................................75
Subpopulation Analyses...........................................................................................................78
Multi-group Structural Equation Modeling .............................................................................79
Post-hoc Analyses ....................................................................................................................85
CHAPTER 4…………………………………………………………………………….……….86
RESULTS……………………………………………………..…………………………………86
Preliminary Analyses…………………………………………………………………………….86
Preliminary Differences by Housing Status ...................................................................................91
Correlations .............................................................................................................................92
Measurement Model……………………………………………………………………………..93
Structural Model…………………………………………………………………………………96
Base Model with Full Sample………………………………………………………….……..96

vii

Model with Covariates……………………………………………………………...………...97
Research Question 1: Structural Model………………………………………………….99
Structural Model with Subpopulations………………………………….…………...…104
Multiple Group Analyses……………………………………………………………………….107
Measurement Invariance……………………………………………………..................107
Research Questions 2 & 3: Mean Differences………………………………………….109
Research Question 4: Direct Paths……………………………………………………...110
Research Question 5: Indirect Paths for Mediation…………………………………….110
Research Question 6: Moderated-Mediation…………………………………………...111
Post-hoc Analyses……………………………………………………………………….……...114
CHAPTER 5……………………………………………………………………………………117
DISCUSSION……………………………………………………..……………………………117
Full Sample Pre-Kindergarten Year Findings…………………………………………………..118
Group Differences By Housing Risk…………………………………………………………...121
Differences in Levels of Parenting and Child Outcomes………………………………...…122
Differential Relations .............................................................................................................124
Race and Housing Status………………………………………………………………..….126
Clinical and Practical Implications………………..……………………………………………127
Limitations and Future Research……………………………………………………………….130
Conclusion……………………………………………………………………………………...135
APPENDIX…………………………………………………………………………..…………138
REFERENCES……………………………………….………………………………………...140

viii

LIST OF TABLES

Table 1: Methods Chart .................................................................................................................48
Table 2: Unweighted Attrition Rates of Full Sample ....................................................................52
Table 3: Unweighted Baseline Mean Scores of Available Predictor and Outcome Variables at
Head Start Entry.............................................................................................................................54
Table 4: Pre-Kindergarten Sample Characteristics with Stable and Unstable Groups ..................58
Table 5: Hypothesized Latent and Observed Variables For Inclusion in Study ............................61
Table 6: Unweighted and Weighted Full Pre-Kindergarten Year Sample Descriptive Statistics .............. 87
Table 7: Unweighted Group Descriptive Statistics and Preliminary Significance Testing ...........88
Table 8: Weighted Group Descriptive Statistics and Preliminary Significance Testing ...............89
Table 9: Correlation Matrix by Stable (upper) and Unstable (lower) Housing Group ..................90
Table 10: Five-Factor Confirmatory Factor Analysis Loadings and Communalities for Full
Sample............................................................................................................................................96
Table 11: Fit Statistics Across Structural Models..........................................................................99
Table 12: Parameter Estimates and Standard Error for Moderated-Mediation Model for Full
Sample..........................................................................................................................................102
Table 13: Construct-Level and Measurement Model Invariance Testing ...................................112
Table 14: Multi-Group Analysis Model Summary Table ............................................................113

ix

LIST OF FIGURES

Figure 1: Family Stress Model of Economic Hardship .................................................................41
Figure 2:Hypothesized Moderated-Mediation .............................................................................. 43
Figure 3: Full Hypothesized Model with Covariates, Error, and Disturbance Terms ...................47
Figure 4: Moderated-Mediation Model Full Sample ...................................................................103
Figure 5: Standardized Estimates for Sub-Groups for Full Model ..............................................106
Figure 6: Teacher-Rated Problem Behavior Scores by Maternal Race and Housing Status .......116

x

CHAPTER 1
INTRODUCTION
In 2013, statistics indicated that over 40% of families with children in the United States
(U.S.) reported a lack of adequate housing, a rate that has remained at or above 40% since 2005
(Federal Interagency Forum on Family and Child Statistics, 2015). These numbers included
children who lived in substandard or crowded housing conditions, as well as children from
families with housing costs that exceeded family earnings. During the late 2000s, the national
economic recession left a large segment of families in the U.S. unemployed, which exponentially
increased the number of families who faced stressors related to economic factors (Burgard,
Seefeldt & Zelner, 2012). The national landscape of that recession highlights the interrelatedness
of economic vulnerability, housing instability, and poverty in the United States. Housing
instability, coupled with risks of poverty, appears to pose unique challenges that extend beyond
the effects of poverty alone (Cohen & Wardrip, 2011; Sell et al., 2010). Yet, there is a scarcity of
literature focused on understanding early childhood and family functioning within the context of
housing instability.
Housing instability refers to a broad range of housing conditions experienced by a diverse
group of families and children. In the present study, housing instability is defined by a
continuum of housing risks that includes families who lack adequate housing (e.g., transitional
facilities, living with other families, or living in hotels), as well as those who have experienced
two or more residential moves in the prior year (e.g., Cutuli et al., 2013; Suglia, Chambers &
Sandel, 2015). This definition of housing instability is consistent with academic and legal
conceptualizations in the broader literature, which commonly include families who lack stable,

1

physical housing (e.g., homeless), live doubled up with relatives or acquaintances, or frequently
move due to economic hardship (Cunningham et al., 2010).
Studies with primarily school-aged children have consistently found that children who
face both poverty and housing instability are at a greater disadvantage than their stably housed
counterparts (Cutuli et al., 2013; Obradović et al., 2009). Unstable housing can endanger
children’s academic and social-emotional development due to disruptions in children’s daily
environment, resulting in problems related to school readiness, school transitions, attendance,
and mental health, including elevated risk for depression and conduct issues (Cutuli et al., 2013;
Edidin, Ganim, Hunter & Karnik, 2012; Rafferty, Shinn & Weiztman, 2004).
Despite the risk housing instability poses to children at any age, research suggests that
families with young children are most likely to face instability (Desmond & Perkins, 2016).
Furthermore, early childhood may be a particularly important developmental period for children
who face housing instability. Housing instability during early childhood has been linked to
academic and behavioral skills related to school readiness, defined as children’s capacity to
function in mainstream school settings (Desmond & Perkins, 2016; Fantuzzo, LeBoeuf, Chen,
Rouse & Culhane, 2012; Lewit & Baker, 1995; Ziol-Guest & McKenna, 2014). The prekindergarten year is especially critical because children’s functioning at entry into formal
schooling at kindergarten is predictive of long-term academic and social-emotional competence
(e.g., Duncan et al., 2007). Therefore, frequent disruptions in housing during the pre-school years
may place children at greater risk for decreased academic and social-emotional functioning in
comparison to stably housed children. In addition, recent literature has indicated that children
who face housing instability early in life, regardless of later housing status, are more likely to
experience persistent social-emotional challenges (Fowler et al., 2014; Rumbold et al., 2012).

2

However, few existing studies include between-group comparisons of unstably and stably housed
pre-school aged children.
The existing body of work on housing instability during early childhood is largely
focused on the ways in which housing instability places children at a disadvantage. Studies that
examine the role of home and school contexts for children facing housing instability are scarce.
An important feature of the home context that is critical to early development is parenting.
Research indicates that conditions of housing instability and poverty can negatively influence
parenting (Koblinsky, Morgan & Anderson, 1997; Gewirtz, DeGarmo, Plowman, August &
Realmuto, 2009). However, specific parenting practices (e.g., parental warmth and engagement
within children’s learning) have been noted to have a protective role in the relation between
economic risks and children’s outcomes (Herbers et al., 2011; Whittaker, Harden, See, Meisch &
Westbrook, 2011). Yet, these relations are still less well understood among populations faced
with housing risks (Whittaker et al., 2011).
Parental mental health appears to be an important factor in understanding the relation
between parenting and children’s early functioning for families who experience poverty
(Goodman et al., 2011). Specifically, the Family Stress Theory of Economic Hardship postulates
that poverty fosters a context in which mental health challenges are more pronounced, and the
effects of parental depression on child functioning are mediated through parenting practices
(Conger, Wallace, Yumei, Simons, McLoyd, & Brody, 2002). Studies of families who face
housing instability have found high rates of caregiver depression, a significant risk factor for
decreased parenting quality (Garg, Burrell, Tripodis, Goodman, Brooks-Gunn & Duggan, 2013;
Gonzales et al., 2011). Despite the importance of better understanding mental health risks as they
relate to parenting, there continues to be a lack of research that concurrently aims to understand

3

the pathways between housing status, multiple dimensions of parenting, parental depression, and
early childhood outcomes.
Another important aspect of the broader context of care are schools and early childhood
center-based programs, such as Head Start, that target families who face economic stressors.
These programs have a longstanding history of promoting school readiness outcomes for families
faced with economic hardship, as well as strive to incorporate comprehensive family-based
services (Bierman, Torres, Domitrovich, Welsh & Gest, 2009). Although research has found
mixed associations between classroom level factors and individual child-outcomes, other studies
have demonstrated that classroom-level indicators of quality may partially buffer contextual and
family stressors for certain groups of children (Burchinal, Peisner-Feinburg, Pianta & Howes,
2000; McWayne, Hahs-Vaughn, Cheung & Wright, 2012). This suggests that high classroom
quality offers the potential for consistent and quality care across settings for families faced with
housing instability despite the presence of stressors, but no studies to date have examined these
supports with Head Start populations who have experienced housing instability.
In summary, there are several gaps in the extant literature on children who face housing
instability during early childhood. First, there are few existing studies that attempt to disentangle
the effects of housing instability from poverty as a developmental context of risk for pre-school
aged children. Second, parenting practices offer a promising focus of research to understand how
housing problems can negatively affect children’s developmental outcomes. In particular,
parents facing significant housing stressors may be at increased risk for depression, which in turn
can affect children’s academic, language, and behavior through parenting practices. Third, there
is a gap in the research in examining aspects of the school context as they may relate to and
influence family-level stressors. Some evidence suggests that classroom quality within Head

4

Start may buffer risks for certain groups of children. Given the complex indirect relations
between parental depression, parenting, and children’s functioning, the role of classroom quality
in mitigating risk is currently unclear for children facing housing risks. Specifically, no known
studies have considered how these aspects of the social context are related to child outcomes
during early childhood for children who face housing instability in comparison to more stably
housed children. Lastly, research with families facing housing instability are often limited by
small sample sizes or represent families faced with a specific type of housing condition (e.g.,
families living in transitional facilities) due to difficulty with recruitment and retention of
participants.
Purpose of Study
The primary purpose of this study is to contribute to the literature on the distinct role of
housing instability as a developmental context of risk. This study focused on how the social
context, mainly parenting practices (parent engagement and parental approaches) and parental
depression symptomology, may be mechanisms by which housing instability affects family and
child functioning. Notably, this study examined the complex indirect and interactive relations of
the social context with development; specifically, the indirect role of parenting practices in the
relation between parental depressive symptoms and children’s outcomes (mediation), as well as
the interaction between parenting and classroom quality as a moderator in the same relations
(moderated-mediation). This study also explored whether these relations between developmental
outcomes and parenting and classroom quality are valid among a large national sample of prekindergarten children who attended Head Start and whether they were differentially related for
unstably and stably housed children. The results of this study can inform prevention efforts at the

5

child, family, and classroom levels for children facing housing instability and poverty during
early childhood.
Research Questions
1. Does classroom quality moderate the indirect relation between parent depression and
children’s outcomes through parenting practices for the full pre-kindergarten sample?
2.

Do children’s pre-kindergarten academic and social-emotional skills differ between
children who were unstably and stably housed?

3. Do caregivers’ parenting practices and levels of depressive symptoms differ between
parents of children who were unstably and stably housed?
4.

Do caregiver levels of depressive symptoms, parenting practices, and classroom quality
differentially predict children’s academic and social-emotional functioning for families of
children who were unstably and stably housed?

5. Do parenting practices differentially mediate the relation between caregiver depressive
symptoms and children’s academic and social-emotional functioning for children who
were unstably and stably housed during Head Start?
6. Does classroom quality differentially moderate the relation between parenting practices
and children’s academic and social-emotional functioning for children who were unstably
and stably housed during Head Start?
Theoretical Frameworks
This study is based on the integration of two theories: Family Stress Model of Economic
Hardship and Risk and Resilience. The Family Stress Model aims to understand the risk
pathways by which economic pressures may jeopardize child and family functioning. Although
the presence of risk factors, such as housing instability, increases the likelihood of less favorable

6

outcomes, not all individuals facing similar circumstances experience decreased functioning.
From a risk and resilience perspective, positive functioning in light of adversity is known as
resilience (Luthar, Cicchetti & Becker, 2000; Masten, 2001). Specific factors within the
environment or the individual that increase the likelihood of positive functioning in the presence
of risk are known as protective factors (Jenson & Fraser, 2006; Masten, 2001). This study will
build on research that examines how factors related to Head Start care may serve as protective
factors for academic and social-emotional outcomes for children facing housing instability (e.g.,
Masten, Cutuil, Herbers, Hinz, Obradović & Wenzel, 2014). Resiliency approaches suggest that
families, caregivers, and children may thrive in the context of adversity through the presence of
protective factors within the environment that are related to positive child and family outcomes.
Family Stress Model of Economic Hardship
One way that economic challenges in family functioning have been studied is through the
Family Stress Model of Economic Hardship (Conger et al., 2002). This theoretical model
postulates that economic factors, such as the low level of family income, create conditions that
elicit economic stress/pressures (Conger et al., 2002). Using Berkowitz’s (1989) FrustrationAggression Model, the Family Stress Model further theorizes that economic pressure (e.g.,
inability to meet financial needs) may increase affective problems in caregivers (e.g.,
depression), which in turn is related to a number of disturbances in family functioning, such as
use of inconsistent parenting practices and family conflict, that ultimately affect children’s
adjustment (Conger et al., 2002). Research examining how economic pressure is related to
parenting practices has suggested that caregivers are more likely to use harsher disciplinary
practices, display lower levels of sensitive/responsiveness, or inconsistences in parenting when
faced with greater economic pressures (Newland, Crnic, Cox & Mills-Koonce, 2013). Due to the

7

well-established nature of the relationship between poverty and economic pressure from previous
investigations of the Family Stress Model of Economic Hardship, this study assumes that factors
associated with poverty and housing instability are relevant hardship factors that create economic
pressures that can adversely affect parenting practices.
Risk and Resilience
This study will also draw upon a risk and resilience perspective to highlight the ways in
which factors within the home and school contexts (e.g., parenting, classroom quality) can
potentially serve as risk and protective factors (Masten, 2001). Risk factors are variables either
within the broader environment or inherent to an individual, which increase the likelihood of less
favorable outcomes (Jenson & Fraser, 2006; Masten, 2001). As previously noted, protective
factors are those that increase the chances of positive outcomes, despite the presence of risk
(Masten, 2001). The current study extends previous research that has established poverty and
socio-economic variables as significant risk factors to child development by examining the
influence of housing instability in conjunction with poverty, in addition to previously established
risks associated within these contexts (e.g., parental depression) on children’s outcomes (e.g.,
Curtis, Coorman, Noonan & Reichman, 2014).
The construct of resilience has been defined and studied in multifaceted fashion (Masten,
2001). For example, recent literature has begun to examine the construct of family resilience,
defined as adaptation of the family unit despite adverse circumstances. This type of approach
emphasizes the cognitive appraisals of stress at the level of the family. In contrast, the current
study is concerned with the presence or absence of specific environmental factors that may
buffer the cascading negative outcomes associated with economic risk (Patterson, 2002). Masten
(2001) described two important considerations to understanding resilience. First, positive

8

functioning is described as resilience only in the presence of an adversity that poses significant
risk for non-normative development. Factors associated with equal levels of positive functioning
across levels of risk are more accurately termed “promotive” factors (Jenson & Frazer, 2006;
Masten, 2001). Therefore, resilient functioning requires a threat or risk to development, such as
facing challenges associated with poverty or housing instability, in order for protective factors to
mitigate this risk. Second, Masten (2001) highlighted the ambiguity implied by the construct of
resilience, as the term is socially constructed to occur within the cultural, societal and historical
standards by which resilient outcomes are thought to represent normative development. This
suggests that resilient functioning may differ based upon factors within the broader context (e.g.,
cultural norms). In this study, resilient outcomes will be conceptualized through the
consideration of children’s school readiness skills. This conceptualization is consistent with
previous research studies that have used social skills and early academic outcomes as indicators
of positive functioning (e.g., Supkoff, Puig & Sroufe, 2012; Ziol-Guest & McKenna, 2014). Yet,
a limitation of this understanding of resilience is that the study may neglect to consider a broader
definition beyond certain outcomes valued by the mainstream Western perspective.

9

CHAPTER 2
LITERATURE REVIEW
This section reviews the literature in the areas of housing instability during early
childhood, parenting, and classroom quality, with an emphasis on Head Start populations. A
rationale for specifically focusing on housing instability during the pre-school years leads into a
discussion of the conceptualization of housing instability and the historical context surrounding
the current state of housing instability. Next, this review focuses on characteristics of families
and children who face housing instability. Specifically, this review emphasizes parent depression
and child development during early childhood, as well as ways in which these domains may
differ between unstably and stably housed families living in poverty. This review also highlights
two key aspects of the early childhood context that can promote children’s development:
parenting within the home and classroom quality at Head Start centers.
Importance of Housing Instability During Pre-School Years
The early childhood years, specifically the years prior to formal school entry, are an
important period of development with regard to early academic, social-emotional, and
developmental outcomes of children. It is well established that poverty is a developmental
context of risk that is associated with a myriad of lasting negative outcomes for children across
academic, social-emotional, and health domains (Duncan, Ziol-Guest & Kalil, 2010; Magnuson
& Waldfogel, 2005; Shonkoff et al., 2012). Emerging literature has noted that families with
younger children are also the most likely group to face housing instability (Desmond & Perkins,
2016). Therefore, recent literature in the area of early childhood has begun to focus on the
subgroup of children living in poverty who also face housing instability (e.g., Ziol-Guest &
McKenna, 2014).

10

Studies demonstrate that housing instability can negatively affect key academic and
social-emotional skills that develop during early childhood, beyond the effects of poverty alone
(Fantuzzo et al., 2012; Ziol-Guest & McKenna, 2014). Although housing instability at any age
may affect development, longitudinal research has suggested that children are particularly
vulnerable to housing instability from birth to five, especially with regard to internalizing and
externalizing behaviors (Fowler et al., 2014; Rumbold et al., 2012). Moreover, exposure to
housing instability at earlier ages is more likely to be linked to social-emotional functioning into
the childhood years (Fowler et al, 2014; Rumbold et al., 2012).
These findings in the area of early childhood and housing instability are particularly
concerning, as school readiness literature has clearly indicated that a range of skills at
kindergarten entry, including academic skills, cognitive functioning, and externalizing behaviors,
are predictive of long-term child functioning (Duncan et al., 2007; Pagani et al., 2010; Sabol &
Pianta, 2012). Specific to children facing housing instability, Herbers and colleagues (2012)
found that although early reading skills in first grade were predictive of oral reading fluency
throughout elementary and middle school for all children included in their district-wide study of
the Minneapolis Public Schools, this relation was stronger for children who faced housing
instability at any point in the six prior years prior. Similarly, Griffith, Arnold, Voegler-Lee and
Kupersmidt (2016) recently found that housing instability during pre-school, among other
family-level factors, in a sample of Head Start families was predictive of social-emotional
functioning in early kindergarten. These results suggest that children who experience housing
instability during the pre-kindergarten years may already be at risk of facing greater academic
and social-emotional disadvantages at kindergarten entry that can persist over time. These

11

findings highlight the importance of better understanding the pre-kindergarten functioning of
pre-school aged children.
Housing Instability: Definitions, Prevalence, and Historical Context
Research studies of housing instability often consider a broad range of families facing a
continuum of unstable housing conditions, such as families who are homeless or lack adequate
housing (e.g., transitional facilities, living with other families, or living in hotels), as well as
families who have experienced several residential moves (e.g., Cutuli et al., 2013; Suglia et al.,
2015). While some studies include families living in a specific housing condition (e.g., families
living in shelters), others combine families who face a range of unstable housing conditions into
one group (Cunningham et al., 2010; Suglia et al., 2015). Research has demonstrated that two to
three (or more) residential moves during the childhood years are negatively related to academic
achievement and social-emotional outcomes and may serve as an indicator of past or future
homelessness for the family, particularly among low-income families (Cutts et al., 2011;
Mantzicopoulous, & Knuston, 2000; Simpson & Fowler, 1994; Ziol-Guest & McKenna, 2014).
Federal agencies that govern programs for families facing housing instability also provide
further insight into the ways in which housing instability has been conceptualized in practice and
research. Cunningham et al. (2010) note that the Stewart B. McKinney Vento Act of 1987
(Public Law 100-77) was the first legislative action to protect the rights of homeless families.
The McKinney Vento Act mandates the provision of several programs through the U.S.
Department of Housing and Urban Development (HUD), as well as protects the educational
rights of children through the Education for Homeless Children and Youth Program enacted
through the U.S. Department of Education. Programs associated with both agencies include
services for families without fixed housing, such as families living in shelters, children in foster

12

care, and families residing in public spaces. Programs supported by the McKinney Vento
Education of Homeless Children and Youth Program further include children and families who
live doubled-up with other families due to economic hardship, children from migrant families,
and children or families temporarily residing in hotels/motels (Cunningham et al., 2010). In
2012, HUD expanded their definition through changes to the Homeless Emergency Assistance
and Rapid Transition to Housing Act (HEARTH), which were amendments to the Moving Ahead
for Progress for the 21st Century Act (MAP-21). These changes included the provision of
services to individuals who may be at-risk of losing their primary residence and cannot secure
new housing due to a lack of resources, as well as families who are unstably housed due to
barriers such as unemployment or disability (Zlotnick, 2009).
Together, the research literature and federal programs operationalize housing instability
in terms of a variety of conditions. Definitions include families who lack an immediate stable
residence (e.g., homeless, doubled-up, children in foster care) and those who are at-risk for
experiencing a loss of residence (e.g., frequent residential moves). Research that has examined
children’s outcomes often refers to both groups using the term homeless and highly mobile or
HHM (Cutuli et al., 2013; Obradović et al., 2009). Within this literature, terms such as
homelessness, residential instability, housing instability, and housing insecurity are also used
synonymously.
Prevalence
National statistics published by the National Center for Family Homelessness indicated
that in 2013, over 2.4 million or 1 in every 30 children experienced homelessness/housing
instability as defined by McKinney Vento Homeless Assistance Act (Bassuk et al., 2014). The
national report further noted that in 2013, the number of homeless children increased in 31states,

13

which was an 8% national increase from 2012. A study by Ma et al. (2008) found that 29.5% of
children from low-income households faced housing instability in a sample of 12,746 families
from the National Survey of America’s Families (NSAF). HUD (2014) most recently reported
that the number of homeless families with children represents 37% of the current homeless
population. In 2009, The National Center for Family Homelessness further reported that of 1.5
million children who experienced homelessness as defined by McKinney Vento Homeless
Assistance Act, 42% were under the age of 6. Furthermore, the U.S. Department of Education
estimated that over 1.2 million children attending public schools were homeless at some point
during the 2013 school year, a number that appears to be growing annually (National Center for
Homeless Education, 2013). These striking statistics demonstrate that families with young
children make up a large percentage of the current homeless population, with a steadily
increasing rate of young children that comprise the unstably housed population at the national
level.
Historical Context and Trends
Several scholars posit that the current state of housing and conditions of poverty are a
product of a longstanding history of residential segregation of persons of color in the U.S.
(Emerson, Chai & Yancey; Massey, 2001; Massey & Denton, 1993; Fong, 1996). Although a
thorough historical analysis is beyond the scope of the study, a consideration of the historical
context informs the current state of housing instability and whom it affects. Throughout the early
and mid-1900s, discriminatory housing practices, such as redlining and gerrymandering led to
impeding persons of color, predominantly black families, from home ownership (Massey &
Denton, 1993; Sugrue, 2005). In addition, high rates of unemployment due to discriminatory
hiring/work practices, as well as “White flight” mentalities confined families of color to specific

14

inner city neighborhoods (Sugrue, 2005). Massey and Denton (1993) further postulate that this
confinement fueled economic conditions (e.g., movement of jobs away from black
neighborhoods, lack of affordable housing) that created conditions of poverty within
neighborhoods in which high concentrations of persons of color lived. Conditions of poverty
perpetuated a greater deterioration of these neighborhoods, often located within urban centers.
Overall, research has supported these proposed residential patterns, predominantly
through studies consistently finding that families of color are more likely to face poverty,
especially within inner city neighborhoods (Fong, 1996; Sugrue, 2005; Massey & Denton, 1993).
Furthermore, research suggests that families of color are less likely to face voluntary moves, as
well as more likely to move to neighborhoods with similar class compositions (Cohen &
Wardrip, 2011).
Caregiver and Family Factors Related to Housing Instability
Research on factors related to chronic and temporary housing instability for families has
found that a number of factors associated with poverty, mental health, physical health, and
psychosocial stressors are related to housing instability among caregivers. Although poverty and
parental depression are the main focus of this particular study due to the established nature of
these factors as they relate to housing instability, health conditions and psychosocial stressors are
also briefly reviewed as additional related factors that may affect caregivers facing housing
instability to fuel conditions of economic stress.
Housing Instability and Poverty
As noted earlier, housing instability may be more problematic for families living in
poverty because their mobility is often unplanned or involuntary and a result of hardship factors
or loss of assets (Ackerman, Kogos, Schoff, Izard, 1999; Clark, 2010; Cohen & Wardrip, 2011;

15

Ziol-Guest & McKenna, 2014). Related to these findings, families who move from low-income
or high poverty neighborhoods are often likely to move to similar neighborhoods, suggesting that
these moves are rarely the result of securing higher quality of housing or neighborhoods (South
& Crowder 1998). These findings suggest that housing instability may be more characteristic of
families living in poverty, due to factors out of the immediate control of the family (e.g., loss of
housing due to limited resources).
Existing research provides insight into the correlated demographic characteristics of
families facing housing instability. In a 5-year longitudinal study, Shinn et al. (1998) compared
over 400 families that applied for sheltered housing New York. Results indicated that 80% of the
families seeking sheltered housing at baseline had their own apartment close to five years later.
Subsidized housing was the largest predictor of stability at 5-year follow-up. In addition, other
demographic characteristics, including being African American, a younger mother, and recently
having a child were predictive of initial application for sheltered housing, but were not related to
longer term housing stability five years later. While parental education level was found to predict
poverty, it was not predictive of shelter requests. Similarly, across recent studies, caregivers
facing housing instability were more likely to identify as part of a racial minority group, as well
as report that they were a single parent, had their child at either a younger or older age than their
counterparts, received fewer years of education, and/or immigrated from another country
(Carrion et al., 2014; Cutts et al., 2011).
Mental Health
An additional noteworthy finding from Shinn et al. (1998) concerned the mental health
conditions of the caregivers. They found that rates of substance use and mental illness were
higher after experiences with homelessness, which suggested that mental health conditions were

16

less predictive of entry into sheltered housing, but may have developed as a result of experiences
with housing instability. Findings from other studies, however, suggest that pre-existing mental
health conditions, such as depression, may serve a precipitating risk factor for housing instability
(Corman, Curtis, Nonnan & Reichman, 2016; Curtis et al., 2014).
Consistent with Shinn et al. (1998), recent literature has also suggested that rates of
depression and other mental health conditions are more prevalent among mothers who have
faced housing instability for longer periods of time (Suglia et al., 2011; Zabikiewicz, Patterson &
Wright, 2014). For example, Zabikiewicz et al. (2014) found that women who were homeless for
two or more years were more likely to experience depression and post-traumatic stress, but that
the chances of these disorders were higher among women who were also caregivers to children
when compared to those who were not parents/caregivers. Similarly, chances of substance abuse
increased among women who were parents. Another recent study found that conditions of
housing disarray, defined as crowded, dark, noisy conditions as opposed to deterioration, defined
as aesthetics of the living conditions (e.g., holes in walls), were predictive of depression and
generalized anxiety among women experiencing housing instability (Suglia et al., 2011). Studies
have also continually noted that caregiver depression appears to be most predictive of housing
insecurity across both cross-sectional and longitudinal studies (Banyard & Graham-Bermann,
1998; Cutts et al., 2011). Yet, results from earlier research suggested that rates of depression
among caregivers facing housing instability were similar to rates among caregivers facing other
conditions of poverty (Bassuk, Buckner, Perloff & Bassuk, 1998).
Physical Health
With regard to health outcomes, research has found a higher incidence of specific health
conditions across adults facing housing instability. Conditions such as cardiovascular disease and

17

hypertension are common among individuals facing housing instability (Chong, et al., 2014;
Vijayaaghavan et al., 2013). Other conditions that are prevalent among homeless adults,
specifically women with children, are sexually transmitted infections, poor nutrition, as well as
lowered quality of prenatal care, which appears to be related to low-birth weight or premature
birth of children (Richards, Merrill & Baksh, 2011; Vijayaraghavan, Tochterman, Hsu, Johnson,
Marcus & Caton, 2012). Additionally, researchers have also found that previous or ongoing
sheltered status appears to be related to lack of primary care medical services and higher rates of
use of emergency departments among women (Duchon, Weitzman, and Shinn, 1999;
Vigayaraghavan et al., 2012).
Psychosocial Stressors
Experiences with specific psychosocial stressors appear to be more characteristic of
families facing housing instability. Rates of domestic violence, intimate partner violence, and
exposure to other types of trauma have been reported to be particularly high among families
experiencing housing instability (Shinn et al. 1998; Vigayaraghvan et al., 2012). These additional
stressors compounded with stressors related to housing instability have been noted to influence
caregiver/adult psychological wellbeing, mental health, interpersonal relationships/interactions,
and parenting practices across studies that have compared low-income stably housed families
with those facing housing instability (Bassuk, Perloff & Dawson, 2001; Ingram, Corning &
Schmit, 1996; Metraux & Culhane, 1999).
Housing Instability and Child Outcomes
Housing instability is a context that can affect children’s outcomes both directly and
indirectly through the additional challenges imposed by housing conditions on caregiver and
family functioning (Ackerman et al., 1999; Cohen & Wardrip, 2011; Suglia et al., 2011). With

18

the growing awareness of the importance of early academic and social-emotional functioning to
long-term child development, the research literature has attempted to better understand the
relation between housing instability and specific developmental outcomes of children during
early childhood (Fantuzzo, LeBoeuf, Chen, Rouse & Culhane, 2012; Masten et al., 2012; ZiolGuest & McKenna, 2014).
Academic Functioning
The majority of the existing research on housing instability and children’s development
has centered on academic achievement of school-aged children (Cutuli et al., 2013; Obradović et
al., 2009). Despite somewhat mixed results regarding the influence of housing instability on
academic skills, several studies with large, representative sample sizes have reported lowered
levels of reading, math, and cognitive skills that can persist throughout schooling when
compared to their stably housed peers from low-income households (Cutuli et al., 2013;
Obradović et al., 2009; Yu, North, LaVesser, Osborne & Spitznagel, 2008). Furthermore, studies
have noted that unstably housed children experience greater challenges with other aspects of the
school experience, even when few differences across achievement have been found (e.g.,
Rafferty, Shinn & Weiztman, 2004). For example, Raffetry et al. (2004) found that parents of
formerly homeless youth reported that their children had a less positive school experience and
were more likely to repeat a grade level.
More specific to early childhood, research has indicated that children who face housing
instability during early childhood may demonstrate below average language and literacy skills
(Schmitt, Finders & McClelland, 2015; Ziol-Guest & McKenna, 2014). For example, in a recent
study, Ziol-Guest & McKenna (2014) analyzed data from over 2,000 children that participated in
the Fragile Families and Child Wellbeing study. Results indicated that three or more moves by

19

the age of 5 predicted lowered receptive vocabulary and phonological awareness skills.
However, the effect of housing instability was significant only for children from low-income
families.
Fantuzzo et al. (2012) further analyzed data on over 10,000 children enrolled in the
School District of Philadelphia from birth through third grade. In order to determine the unique
contribution of the effects of mobility, researchers controlled for variance explained by several
potential confounding factors, including SES, race, and prior achievement. Results indicated that
children who experienced both homelessness and more frequent school transitions demonstrated
lowered standardized test scores and a higher rate of teacher reported classroom behavior
problems. These findings directly suggest that housing instability poses greater risk for academic
and behavioral problems across school and early childhood care settings.
Social-Emotional Functioning
Research in the area of mental health and social-emotional functioning for children faced
with housing instability has largely focused on adolescent samples. Research with predominately
adolescent homeless/runaway youth samples has indicated that children faced with housing
instability, specifically youth who lack immediate physical residence, may be at increased risk
for experiencing trauma, substance use, anxiety and depression (see review by Edidin et al.,
2012). While researchers speculate that mental health concerns, such as mood and anxiety
disorders, may pre-exist in this population at earlier ages due to history of trauma, abuse, or other
conflicts, there also appears to be an increase of mental health challenges as a result of exposure
to homelessness among adolescent samples (Castro, Gustafson, Ford, Edidin, Smith, Hunter &
Karnick, 2014; Kamieniecki, 2001). Although findings of this body of literature are largely

20

focused on a specific group, these results demonstrate the potential influence of housing
instability on mental health.
There are fewer studies that directly examine the mental health/social-emotional
outcomes of children facing housing instability with school-aged and early childhood samples.
Some of the existing literature suggests that children facing housing instability may not only be
at risk for externalizing, internalizing, and self-regulation challenges, but that the effects of
housing instability during specific periods of development may have a lasting influence on
behavior and social-emotional functioning (Fowler et al., 2014; McCoy & Raver, 2014; Yu et
al., 2008 Ziol-Guest & McKenna, 2014; ). Notably, Fowler et al. (2014) found that housing
instability during infancy, early childhood, and adolescence were more predictive of behavior
problems in a large sample of children reported to be at high risk for child maltreatment.
Preschool-aged children specifically between the ages of 4 and 6 who moved more frequently in
the 12 months prior to data collection were more likely to exhibit externalizing behaviors and
continued to demonstrate problem behaviors at the three-year follow-up.
Moreover, caregiver mental health, particularly maternal depression, is also predictive of
children’s development (e.g., Goodman et al., 2011; Koblinsky, Kuvalanka & Randolph, 2006).
In a meta-analysis of 193 studies conducted on the effects of maternal depression on child
functioning, Goodman et al. (2011) demonstrated that maternal depression was related to broad
measures of internalizing and externalizing functioning, as well as to negative emotions in
children. Although effect sizes were small, they were slightly larger across studies of families
living in poverty. In addition, specific conditions of poverty, such as being from a single-parent
home were associated with increases in effect sizes for certain clusters of externalizing
behaviors. Results of this study also found that children exposed to maternal depression earlier

21

in life may be at greater risk for developing future internalizing and externalizing challenges.
Given the previously noted rates of depression among caregivers who experience housing
instability (e.g., Suglia et al., 2011), these results provide further evidence of the importance of
considering parental depression and specific conditions of poverty as they relate to child
functioning to examine differences across contexts of risk during early childhood.
Related Health and Developmental Outcomes
Other lines of inquiry that may affect early academic social-emotional functioning
among children facing housing instability have indicated decreased health outcomes and higher
rates of developmental disabilities among children who face housing instability and poverty
more broadly. Generally, research has noted that poverty and housing instability may serve as
risk factors for poorer health outcomes, including higher rates of illness (e.g., flu, infections),
lead poisoning, lack of nutrition/food insecurity, and respiratory illnesses, particularly asthma
(Bashir, 2002; Bratt, 2002; Cutuli, Herbers, Lafavor, Ahumada, Masten & Oberg, 2014;
Weinreb, Goldberg, Bassuk, Perloff, 1998). Although developmental delays in several areas are
often thought to be characteristic of conditions imposed by poverty in general, recent research
suggests that delays in areas such as language and motor skills may be exacerbated for those who
also experience housing instability (Carrion et al., 2014; Chiu & DiMarco, 2010).
Parenting in the Context of Housing Instability
While a significant amount of research has noted the potential risks that housing
instability poses to child development, there has been less research examining how parenting, an
important determinant of early developmental skills for children, is affected within the context of
housing instability. The research on parenting during early childhood for families living in
poverty provides partial support for the importance of understanding specific parenting practices

22

in the context of economic risk more broadly. The existing literature on parenting in the context
of housing instability provides insight into the ways in which housing instability may create
unique challenges to parenting for caregivers with young children. The Family Stress Model
provides further explanation for the pathways by which parenting and child development may be
affected by economic hardship (Conger et al., 2002; McLoyd, Jayarantne, Ceballo & Borquez,
1994).
In this study, parenting practices are defined multi-dimensionally drawing on two areas
of parenting research in early childhood: parental approaches and parent engagement. With
roots in the parenting styles literature, parental approaches have traditionally been studied as
patterns of behaviors on two dimensions: warmth and demandingness. Parenting is viewed as
falling on a continuum ranging from warm, responsive behaviors (e.g. affectionate) to harsher
approaches (e.g., rigid, verbal threats), as well as a continuum from controlling (e.g., strict, high
expectations) to few limits (e.g., few or no boundaries) (Baumrind, 1967; Landry, Smith, Swank,
Assel & Vellet, 2001; Petril & Deater-Deckard, 2004). Recent research has expanded how
parenting practices are conceptualized to move beyond parenting approaches to recognize how
active and responsive parental engagement in learning/development, such as enriching activities
and involvement with early academics, relates to early outcomes (Sheridan, Knoche, Kupzyk,
Edwards & Marvin, 2011). Both areas of parenting are addressed in the present study to provide
a more complete, multi-dimensional understanding of high risk families by examining variation
across the types of practices, as discussed further below.
Parent Engagement
Parent engagement, which is an expansion of the term “parent involvement”, refers to a
variety of practices that parents engage in with children to foster development (Sheridan, et al.,

23

2011). Although there are many ways to conceptualizing parent engagement, Edwards, Sheridan
and Knoche (2008) define parent engagement as interactions that foster warmth, support for
autonomy, and involve an active participation in learning. Much of the literature on children
facing economic risks has focused on the benefits of parental involvement in learning (e.g.,
literacy support within the home) and the quality of the interactions during these activities in
which parents may develop children’s early learning competencies and readiness for school
(McWayne, Fantuzzo, Cohen & Sekino, 2004). Research has linked aspects of the home
environment and parental behaviors, including several forms of literacy and enrichment within
the home (e.g., indicators of shared book reading, interactive styles of book reading, parents’
own literacy practices, attending community events) to early literacy skills among children
facing economic risks (e.g., Bracken & Fishel, 2008; Burgess, Hecht & Lonigan, 2002; Hood,
Conlon & Andrews, 2008). In addition, parental involvement within center-based facilities (e.g.,
Head Start) and parental expectations for educational attainment also have documented relations
to school readiness skills (Foster, Lambert, Abbott-Shim, McCartyn & Franze, 2005; McWayne
et al., 2004; Suizzo & Stapleton, 2007).
Although parent engagement practices appear to be universally beneficial, families facing
economic risks are not a homogenous group. Research has demonstrated that there may be
potential barriers to accessing important resources for home-based stimulation and involvement
in children’s schooling, including access to resources and supports (Desimone, 1999). In
addition, research has highlighted the differential use of specific practices that relate to the socioeconomic context (Bradley & Corwyn, 2014; Vandermaas-Peeler, Nelson, Bumpass & Sassine,
2009). For example, Vandermaas-Peeler et al. (2009) found that parents with young children
from both middle and low-income homes engaged in similar amounts of guidance while reading

24

and playing with their children, but that the practices of middle-income families included a
greater emphasis on literacy. Yet, both groups made attempts to foster interest among their
children in literacy activities. Such findings indicate the relevance of understanding the ways in
which specific aspects of the economic context may result in different parenting practices, but
also how certain practices may be similar across families.
Despite the growing attention to context-specific differences in parent engagement
practices, there has been little work to understand the ways in which housing instability may
affect caregiver engagement independently of poverty. An earlier study by Koblinsky, Morgan,
and Anderson (1997) compared the parenting practices of children from low-income families
that included 31 homeless and 28 stably housed families with children in Head Start. The results
suggested that homeless children lived in less positive physical environments, as measured by
the Home Observation for Measurement of Environment (HOME), that included more crowded
and disorganized spaces. Results further suggested that children of homeless mothers were less
likely to receive cognitive and social stimulation within the home. Similar to low-income
families, researchers also speculate that families facing housing instability experience significant
barriers, including a lack of transportation, resources, and employment, to forms of school-based
parent involvement, such as attendance at school events or assisting in classroom services
(Jozefowicz-Simbeni & Isreal, 2006).
A recent qualitative examination of the types of literacy practices used by mothers and
school-aged children living in a homeless shelter in California found that many of the parents
reported barriers to engaging in literacy practices due to a lack of physical space or privacy, but,
many were, in fact, engaged in literacy practices. Mothers were both observed and reported to
engage in literacy practices through public libraries, churches, and schools (MacGillivray et al.,

25

2009). Although school was highlighted as a place for literacy learning, many of the mothers
described schools in terms of their evaluative components of school (e.g., homework, testing,
grades). Researchers believed that this focus might have reflected the challenges associated with
adjustment to the expectations of new schools with each transition. These findings also imply
that community-based supports, such as churches and public libraries, function as important
sources of literacy exposure and resources for parent engagement for families facing housing
instability.
In summary, the literature on parent engagement as related to housing instability has
demonstrated the importance of better understanding the unique literacy practices of unstably
housed families. To date, results of these studies suggest that traditional forms of engagement,
such as school-based involvement or home-based literacy practices, may be compromised in the
presence of housing stressors. Nevertheless, parents may rely on community-based and other
alternative forms of support to engage in practices that promote children’s learning and socialemotional growth (Koblinsky et al., 1997; MacGillivray et al., 2009).
Parenting Approaches
Approaches to parenting which foster positive child outcomes have been widely studied
in the child development literature (Baumrind, 1967; Landry et al., 2001; Riley et al., 2014).
Research has indicated that several behavioral dimensions of parenting lead to more favorable
cognitive and social-emotional outcomes for children, including a warm, supportive, and
responsive caregiving style that is also characterized by boundaries and rules, which is referred
to as authoritative parenting (Landry et al., 2001; Mistry, Benner, Biesanz, Clark & Howes,
2010; Petril & Deater-Deckard, 2004; Shaw, Winslow, Owens, Vondra, Cohn & Bell, 1998).

26

Research has indicated variation in the use of specific parenting approaches by context
(e.g., Lee, Lee & August, 2011). Some existing literature has noted that the stressors associated
with socio-economic risks, such as financial stress, may be linked to lower rates of
responsiveness and warmth (Jackson, Brooks-Gunn, Huang & Glassman, 2000; Lee et al., 2011).
Current literature has found that the relation between financial stress, parental approaches, and
child outcomes is much more complex and that the role of socio-economic risks on parenting
behaviors and child outcomes vary with factors specific to the developmental context, such as
child’s age, race, and type of SES indicator/risk measured (Bradley & Corwyn, 2014; Dotterer,
Iruka, & Pungello, 2012). These contextual nuances underscore the need to better understand
parenting approaches among specific developmental age groups that face similar types of risks.
There is a scarcity of recent literature, however, on the ways in which parenting
approaches may differ among those facing housing instability and socio-economic risks during
early childhood. Existing findings have suggested that homeless mothers may provide less
academic stimulation, as well as display lower levels of warmth and affection within the home
environment in comparison to stably housed, low-income mothers (Koblinsky et al., 1997).
Mothers from families who were homeless have also been found to be less responsive and
observed to display fewer instances of praise.
Moreover, research has also examined the use of physical punishment and firm
disciplinary practices in samples of homeless families. Koblinsky et al. (1997) found no
differences between homeless and low-income, stably housed mothers in the use of discipline,
particularly physical punishment, with half of the sample reporting use of spanking within the
previous week. More recent literature has suggested a higher rate of physical punishment and
overall risk for child maltreatment among those faced with housing instability (Park, Ostler &

27

Fertig, 2015; Perlman & Fantuzzo, 2013) For instance, a longitudinal secondary data analysis of
over 2,000 families found that housing instability was associated with higher rates of physical
(e.g., spanking) and psychological (e.g., shouting, swearing) aggression towards children in
comparison to other low-income mothers in the sample (Park et al., 2015). Results further
suggested a higher likelihood of these behaviors occurring during earlier waves of data collection
when children were younger (e.g., 3 or 5 years).
Yet, this literature-base has often been critiqued for perpetuation of deficit perspectives
of the parenting practices of low-income and homeless families (McWayne, Mattis, Green
Wright, Limlingan & Harris, 2016). An alternative way in which parenting differences have been
studied has been from a more strengths-based perspective. Recent studies with families of
children facing housing instability have focused on the benefits associated with positive, warm
interactions among unstably housed families. For example, Gewirtz, DeGarmo, Plowman,
August, and Realmuto (2009) videotaped and coded a series of parent-child tasks in a sample of
200 dyads with school-aged children (6 to 12 years) living in supportive housing facilities. In
addition to parenting behaviors, self-efficacy beliefs or parental beliefs of competence related to
sense of control were also assessed. Results indicated that both parenting behaviors, such as
positive and less coercive interactions, and self-efficacy were related to children’s behavioral
adjustment. However, parenting practices mediated the relationship between self-efficacy and
child behavior, indicating that effect of self-efficacy beliefs on children’s behavioral adjustment
may be explained through observed parenting approaches.
In a related study, Narayan, Herbers, Plowman, Gewirtz and Masten (2012) evaluated
expressed emotion of 39 caregivers with younger children, aged 4 to 7 years, living in
emergency shelters. Analysis of coded videotaped speech samples during interactive activities

28

with their children indicated that parental warmth was associated with positive involvement and
effective parenting (e.g., problem solving, skill encouragement). Criticisms used by parents were
correlated with coercive parenting (less positive involvement, problem solving, skill
encouragement). Among parents who spoke for longer periods of time in the speech sample,
negative affect was associated with fewer positive parenting behaviors and greater levels of
teacher reported externalizing behaviors.
This research has also attempted to study if parenting is promotive and/or protective to
the development of children faced with residential instability (e.g., Herbers, et al., 2011; Militois
et al. 1999; Riley et al., 2014). Although high quality parenting may be promotive or beneficial
across all levels of risk, a small evidence-base suggests that parenting practices may
differentially affect children facing housing instability, such that parenting may serve as a
possible protective factor (e.g., Militois et al., 1999). For example, in a recent study, Herbers et
al. (2011) evaluated the role of parenting quality in child academic achievement using a sample
of 58 parent-child dyads with children aged 4 to 7 years residing in a shelter. Parenting quality
was measured through ratings conducted by both the initial interviewer and a trained observer
across a series of questions rated on a five-point scale. Notably, results of the study demonstrated
that parenting quality moderated the relation between risk and academic functioning. This
indicated that children who were experiencing risk and higher quality of parenting had higher
academic functioning in comparison to children experiencing similar risk and lower quality
parenting. These findings are consistent with literature that has indicated that parenting
approaches that are characterized by sensitive, warm interactions, can be protective for children
facing other types of socio-economic risks more broadly (Riley et al., 2014; Whittaker, Harden,

29

See, Meisch & Westbrook, 2011). This suggests that positive parenting practices could have a
stronger relation to children’s outcomes for those facing greater risks.
To date, studies on approaches to parenting in the context of housing instability have
primarily covered three main areas. First, findings have focused on the variability in certain
parenting approaches, with findings largely concluding that housing instability may result in
lowered levels of warmth and increased use of harsher disciplinary practices (e.g., Koblinsky et
al., 1997; Park et al., 2015). Second, strengths-focused studies have demonstrated that warm
interactions have positive benefits across both early childhood and school-aged children who
face house instability, consistent with previous research that has examined parenting approaches
more broadly (Landry et al., 2012; Riley et al., 2014). Lastly, some of this research has further
suggested that positive parenting practices can moderate risk (Herbers et al., 2011; Militois et al.,
1999).
Parental Mental Health as a Mechanism of Group Differences
Research in the area of poverty offers some insight into potential explanations for the
ways in which housing instability may create a distinctive parenting context. As previously
noted, several studies have provided support for the Family Stress Model, which suggests that
indicators of economic hardship can influence child outcomes through the quality of parenting
(Conger, et al., 2002; McLoyd et al., 1994). From an economic stress perspective, housing
instability may serve as an additional risk related to economic hardship among families living in
poverty, which in turn may affect parenting practices to a greater extent.
Several studies have examined indicators of family stress that are associated with socioeconomic risks and parenting, such as parental depression and relationship strain among adult
caregivers (Conger et al., 2002; Newland et al., 2013). In addition, barriers to resources may also

30

place additional demands on parents (Haber & Toro, 2004). The Family Stress Model of
Economic Hardship theorizes that economic pressure (e.g., unable to meet financial needs) may
increase affective problems in caregivers (e.g., depression), which is associated with a number of
disturbances in family functioning, such as parenting practices (Conger et al., 2002). One of the
most widely researched indicators of the manifestation of economic pressure has been parental
mental health, specifically maternal psychopathology, as it relates to decreased quality of
parenting (Garg, Burrell, Tripodis, Goodman, Brooks-Gunn & Duggan, 2013; Gonzales et al.,
2011; Newland et al., 2013). These findings may have specific implications for families facing
housing instability, as previously noted research suggested a higher rates of caregiver
psychopathology, such as depression, among caregivers facing housing instability (Curtis et al.,
2016; Curtis et al., 2014; Gerwirtz et al., 2009; Suglia et al., 2011). This body of literature
clearly denotes that parental psychopathology may be a critical process that is related to
parenting quality within the context of housing instability.
The specific conditions of housing instability associated with quality of parental mental
health and adverse outcomes are unclear. Although families facing housing instability may be
exposed to a number of conditions that might affect parenting quality, extant research provides
some evidence that housing conditions may be linked to the quality of parenting (Howard,
Cartwright & Barajas, 2009; Koblinsky et al., 1997; Suglia et al., 2011; Lindsey, 1998). Across
studies that have examined families residing in shelters and transitional facilities, parents
reported that restrictions (e.g., rules of the facilities in which they reside), as well as lack of
privacy affect their parenting practices (Lindsey, 1998; Meadows-Oliver, 2003; Thrasher &
Mowbray, 1995). Parents faced with these conditions also noted difficulty with adhering to
family routines (Mayberry, Shinn, Benton & Wise, 2014). Additionally, chaotic physical

31

conditions, such as crowded, noisy, and disorganized spaces have been suggested to pose barriers
to parenting practices (Koblinsky et al., 1997; Suglia et al., 2011). These set of findings
demonstrate that factors related to housing instability are salient considerations with regards to
parenting practices. Yet, these studies provide limited insight into the processes and pathways
that affect parenting practices among unstably housed parents of children during early childhood.
Head Start and Housing Instability
Another context that children facing poverty and housing instability commonly
experience are early childhood centers, such as Head Start. Research has consistently
demonstrated that Head Start improves critical academic/cognitive and behavioral indicators of
school readiness among children living in poverty (e.g., Bierman et al., 2009). The literature base
on long-term effects of Head Start on children’s functioning is somewhat mixed. However, a
number of studies have reported gains during the Head Start years in academic and socialemotional domains of functioning (e.g., Lee, Brooks-Gunn, Schnur & Liaw, 1990; Puma et al.,
2005). In recent years, research has shown that classroom quality is predictive of short and
longer-term outcomes (e.g., Pianta et al., 2005)
The reauthorization of the Improving Head Start for School Readiness Act in 2007
included new provisions specific to families facing housing instability. These changes included
categorical eligibility for highly mobile children, greater emphasis on identifying and enrolling
children facing housing instability, and greater attention to services required in order to address
the needs of these families (Head Start for School Readiness, 2007; Institute for Children
Poverty and Homelessness, 2011). A 2011 policy report estimated that over 1 million children in
the U.S. were enrolled in Head Start and close to 40,000 were classified as homeless, with more
likely experiencing frequent mobility (Institute for Children Poverty and Homelessness, 2011).

32

These numbers represented a 50% increase of Head Start enrollment of families facing housing
instability from 2008 estimates. Although legislative changes have ensured greater enrollment of
families facing housing instability into Head Start, there has been little follow-up research that
has specifically examined the ways in which Head Start and quality of center-based care may
affect child outcomes for children from families facing housing instability.
Short and Long-Term Head Start Outcomes
A broad literature base provides evidence for the potential benefits of Head Start for
children facing socio-economic risk factors. Despite the variation of programs across the U.S.,
research has demonstrated positive short-term benefits for child and family functioning.
Research has indicated that by the end of Head Start, children show improvements in several
areas of functioning, including cognitive skills, such as working memory and attention, as well
as in skills related to literacy and numeracy (Bierman et al., 2009; Puma, Bell, Cook, Heid &
Lopez, 2005; Welsh, Nix, Blair, Bierman & Nelson, 2010; Zill & Resnick, 2006). In addition,
participation in Head Start programming has the potential to improve social-emotional skills,
such as prevention of externalizing problems and improvement in prosocial skills (Bierman et
al., 2009; Milfort & Greenfield, 2002; Puma et al., 2005). Furthermore, Head Start programs
have also been noted to improve health and family outcomes, particularly when programs strive
to engage parents through home and school communication, parent training efforts, and
interventions (McWayne et al., 2004; Puma et al., 2005; Webster-Stratton, Reid & Hammond,
2001).
Despite the demonstrated short-term success and potential for Head Start to improve
important school readiness skills among children facing socio-economic risks, research has
indicated mixed outcomes with regard to the sustainability of gains made during the Head Start

33

years. Key findings, such as those associated with the High Scope Perry Preschool Study, which
demonstrated long-term positive outcomes, including higher rates of graduation, higher earnings,
and improved social and emotional wellbeing between children living in poverty who attended
preschool and those who did not attend preschool, provided early scientific evidence for the
importance of high quality preschool for children living in poverty (e.g., Belfield, Nores, Barnett
& Schweinhart, 2006). The long term benefits of Head Start have recently been challenged,
however, with findings related to the Head Start Impact Study, a national evaluation conducted
by the U.S. Department of Health and Human Services (2005) of Head Start programming,
which included a comparison group of waitlist children who did not receive Head Start services.
Results of the evaluation indicated small effect sizes in academic, social-emotional, and health
domains for children who began the program at age 3 or 4 in academic and health domains
during the Head Start year. However, there were few lasting effects through third grade,
particularly for children who entered at age 4. There were only small, sustained benefits for
children who began Head Start at age 3 in the areas of parenting practices, social-emotional
skills, and literacy skills. Improvements maintained through third grade were limited to specific
groups of children. For example, children whose parents had symptoms representative of mild
depression showed greater gains in literacy and social-emotional outcomes throughout their third
grade year.
While these gains appear small, researchers have suggested that these findings are
consistent with previous longitudinal studies that have examined Head Start and other preschool
programs (e.g., Perry/High Scope). These previous studies have also found that the beneficial
effects for children who attended pre-school compared to those who received other forms of
care, may decrease over time and certain outcomes may not be apparent until school-aged years,

34

adolescence, or adulthood (Currie & Niedell, 2007; Schweinhart & Weikart, 1997). Furthermore,
alternative outcomes, such as long-term educational placements and grade promotion have been
used in other studies to demonstrate the outcomes associated with Head Start (Barnett & Hustedt,
2005; Ludwig & Phillips, 2008). In addition, the effects of preschool may vary based on child’s
abilities at entry and may improve from initial functioning, but children’s improvements also
depend upon other contextual factors, such as home environment (Lee et al., 1990).
Furthermore, a secondary analysis of Head Start Impact Study data indicated that Head Start
involvement increased parent involvement through time spent with children (Gelber & Isen,
2013). All of these findings suggest that while gains are most evident after the initial Head Start
year, there appear to be continued gains throughout the course of childhood into adulthood
associated with high quality preschool programming. Moreover, these research findings also
portray the need to understand the influence of early childhood programming on specific groups
of children faced with different risk factors.
Classroom Quality
Beyond the findings related to the short and long-term gains associated with Head Start,
it is difficult to make general claims about the effectiveness of Head Start due to the variation in
quality across centers (Raver, Jones, Li-Grining, Mtezger, Champion & Sardin, 2008). Research
in recent decades has paid greater attention to environmental factors specific to Head Start
centers that are related to positive outcomes. Although program and other classroom factors such
as higher spending per child and individual teacher characteristics, may result in better outcomes,
global indicators of quality at the classroom level have consistently been found to be predictive
of positive child outcomes (Curie & Niedell, 2007; Raver et al., 2008). Classroom quality has
been conceptualized as both the physical (structural) and interactive elements (e.g., student-

35

teacher interactions) of the classroom that optimize student success (Pianta et al., 2005). Pianta et
al. (2005) further note that interactive elements between students, teacher, and materials are often
referred to as process quality of the classroom.
With regard to process quality, research has identified several important components that
are related to children’s classroom success, including positive teacher-student relationships,
positive climate, and instructional quality as evidenced through practices such as support or
feedback during learning activities (Hamre & Pianta, 2001; 2005). One of the most extensively
used approaches to evaluating process quality is based on the Classroom Assessment and
Scoring System (CLASS) framework, which proposes that three aspects of process quality are
related to student outcomes: emotional support, classroom organization, and instructional support
(La Paro, Pianta, Hamre & Struhlman, 2002). The CLASS framework has been validated across
national samples of children from pre-school to fifth grade, as well as draws on a literature base
that has demonstrated that each of these broad indicators of quality is related to student success
(Hamre, Pianta, Mashburn and Downer, 2007; Hamre et al., 2007). The domain of emotional
support includes consideration of the general climate, teacher sensitivity, and level of
attentiveness to student perspectives (Hamre et al., 2007). Consideration of classroom
organizational elements, such as behavior management and opportunities for learning in different
formats, as well as instructional support, which includes the level of feedback provided to
students, modeling of language, and development of concepts are additional components of the
CLASS framework (Hamre et al., 2007; Pianta et al., 2005). Extant research has found relations
between these framework domains and student achievement, student-teacher relationships,
classroom behavior, language development, and social skills among samples of preschool-aged
children (e.g., Hamre & Pianta, 2005; Mashburn et al., 2008; Pianta et al., 2005).

36

Specific to early childhood centers, research has emphasized the importance of emotional
and instructional supports (Burchinal et al., 2006; Mashburn, Hamre, Downer & Pianta, 2006;
Mashburn et al., 2008; Hamre & Pianta, 2001; 2005). Recent literature has indicated that
emotional and instructional process quality more consistently predict children’s outcomes across
studies (Pianta et al., 2005; Mashburn et al., 2008). For instance, using three different measures
of classroom quality, Mashburn et al. (2008) found that dimensions of instructional support (e.g.,
feedback, modeling of language) were predictive of academic and language skills (e.g.,
vocabulary, oral expression, achievement) and that emotional support was predictive of scores on
a teacher reported social skills measure across 671 pre-K classrooms.
In addition to the process aspects of classroom quality, structural elements also appear to
be related to child outcomes across early childhood settings. La Paro, Thomason, Lower, KinterDuffy & Cassidy (2012) note that the Early Childhood Environment Rating Scale (ECERS-R)
has been the most commonly used tool to evaluate structural elements of the classroom that are
related to child outcomes. Although there is some overlap with dimensions of process quality,
the ECERS-R assesses characteristics of the physical environment such as space, types activities,
and interactions of language/reasoning across seven subscales (Harms, Clifford & Cryer, 1998).
Research has broadly found that dimensions of the ECERS that capture stimulation within the
classroom environment and warm, supportive approaches are predictive of preschool and
kindergarten success (Burchinal et al., 2006). More specifically, research suggests associations
between ECERS-R scores and children’s language skills, mainly expressive language for
children in Head Start (Burchinal, Roberts, Nabors & Bryant, 1996; Love et al., 2003; Mashburn
et al., 2008).

37

Quality of care as protective process. To date, no existing studies have examined
classroom quality or Head Start in the relation between risk and child outcomes for families
facing housing instability. Some researchers have considered the protective role of Head Start
involvement on child outcomes for children facing poverty risks and family stressors, but the
findings appear to depend upon the risk factors under consideration and demographic
characteristics, such as race or gender (Burchinal et al., 2006; Burchinal, Peisner-Feinburg,
Bryant & Clifford, 2000; Caughy, DiPietro & Strobino, 1994; Vallotton, Harewood, Ayoub, Pan,
Mastergeorge & Brophy-Herb, 2012).
The quality of center-based care may be more likely to buffer risks for specific groups
(Burchinal et al., 2006; Vallotton et al., 2012). For example, Burchinal et al. (2000) used
secondary data analysis with three large-scale studies to examine the potential role of the quality
of center-based care. Results of the study indicated that classroom quality served as a promotive
factor, such that there were no differences across the examined risk factors in the role of
classroom quality on most academic or behavioral outcomes, with the exception of language
skills. These results demonstrated that quality of care appeared to have a stronger or protective
effect on language skills for children of color who also faced socio-economic risks.
In a more recent longitudinal study, Burchinal et al. (2006) further found that the quality
of center –based care, along with parenting and children’s language skills, all moderated the
relation between cumulative risk and children’s academic skills and behavior problems in a
sample of 75 African American children. These results along with similar findings reported
earlier by Peisner-Feinburg et al. (2001), suggest that there are likely differential effects of the
quality of care for children and families with different background characteristics. Furthermore,

38

such findings also highlight the relevance of more clearly understanding the role of center-based
care across other specific conditions of risk, such as housing instability.
Parenting and Classroom Quality
Few studies have examined indicators of quality across social contexts together.
Furthermore, no studies have examined indicators of parenting and early childhood care contexts
for children facing housing instability in one model. Several researchers, however, have used
person-centered analyses to examine the combinations of quality indicators that lead to the most
favorable outcomes (Bulotsky-Shearer, Wen, Faria, Hahs-Vaughn & Korfmacher, 2012;
McWayne et al., 2012). Using the 1997 Head Start Family and Child Experiences Survey,
Bulotsky-Shearer et al. (2012) demonstrated that a profile of high parental school and home
involvement, coupled with high classroom quality scores was associated with the most favorable
academic and social-emotional outcomes.
McWyane et al. (2012) further used the Head Start Family and Child Family Experiences
Survey (2006) and found that children’s patterns of skills remained relatively stable throughout
their initial Head Start year, but combinations of several contextual variables, including child,
parent, and classroom characteristics influenced changes in children’s academic and social skills.
While there was clear evidence that parenting styles, family-level demographic characteristics
(e.g., parental education), and family involvement in Head Start were positively associated with
change over time, there was less clear support for the role of classroom quality using the
ECERS-R, as it did not significantly predict change in children’s outcomes over time.
Researchers argued, however, that the greater consideration of process elements of classroom
quality, as well as the possible longitudinal effect of classroom quality on child outcomes, may
be more appropriate considerations in future studies. Studies examining the role of quality in

39

relation to parenting and classroom practices have demonstrated the beneficial and possible
protective role of high levels of both parenting and classroom quality. Although the role of
classroom quality is somewhat less clear, recent evidence that includes more sensitive measures
of process classroom quality have demonstrated positive child outcomes (Pianta et al., 2005).
In summary, the literature on parenting and classroom quality suggests that housing
status may interact with aspects of the social context to produce differential outcomes for
children during early childhood. Evidence in support of the Family Stress Theory suggests that
parental mental health may affect parenting practices, a strong predictor of child functioning
(Howard, Cartwright & Barajas, 2009; Suglia et al., 2011). However, the role of additional
contexts in which children participate (e.g., Head Start) in the established relationships between
parent depression, parenting practices, and child outcomes is understudied for children facing
housing instability. Classroom quality has been found as one such factor in the Head Start setting
that may mitigate risks for certain groups of children facing greater risks (Burchinal et al., 2000).
Yet, it is currently unclear how classroom quality and parenting interact to influence child
functioning for children who may experience housing risks.
Present Study
The present study used secondary data analysis to examine whether housing instability
during the Head Start years is related to child functioning during the pre-kindergarten Head Start
year. Due to the large body of research evidence demonstrating the importance of prekindergarten skills at formal school entry for kindergarten and childhood functioning, outcomes
at the end of the pre-kindergarten year were specifically examined to gain greater insight into the
processes that might relate to child functioning at this time point (e.g., Duncan et al., 2007).
Furthermore, the study aimed to address gaps in understanding how parenting practices,

40

specifically parental approaches, such as warmth during interactions, endorsement of the use of
discipline, and encouragement of independence, as well as parental engagement to promote
learning and development, are related to child outcomes in the context of housing instability.
Given the likelihood of higher rates of depression-related concerns among caregivers facing
housing instability reported by previous research, the present study examined whether parental
depression could have a role in the mediational pathway (see Figure 1) that directly affected
parenting practices and indirectly related to child outcomes (e.g., Sugila et al., 2011). Yet,
children and families interact with and participate in different social contexts, including
educational settings, which can buffer the risks at the level of the family. Therefore, the present
study also examined if Head Start classroom quality moderated the hypothesized mediational
pathway (moderated-mediation) as high quality classroom practices at early childhood centers
have been identified as contextual influences that can promote positive child functioning (Pianta,
2005; Riley et al., 2014).
Figure 1: Family Stress Model of Economic Hardship

Results of this study directly expand upon extant literature in two ways. First, this study
adds to literature by including an examination of whether the interactive role of parenting and
classroom quality affect children’s functioning during the pre-kindergarten year. Second, the

41

primary relevance of the study was to understand the differential relations in the associations
between parental depression, parenting, classroom quality, and children’s outcomes for children
who experienced housing instability using structural equational modeling techniques.
Research Questions and Hypotheses
1. Does classroom quality moderate the indirect relations between parent depression and
children’s outcomes through parenting practices for the full pre-kindergarten sample?
I hypothesized that in addition to the latent constructs of parenting practices (engagement
and approaches) serving as mediators in the relation between parental depression and children’s
functioning, classroom quality will further moderate these indirect pathways among the full prekindergarten sample. It is important to note that the hypothesized moderation would occur
between potential mediators (parenting practices) and outcomes (child functioning), signifying a
moderated- mediation as the full effect (Baron & Kenny, 1986; Hayes & Preacher, 2004) as
depicted in Figure 2. This would signify that the interactive effect is dependent upon the levels of
the parenting and classroom quality latent constructs.

42

Figure 2: Hypothesized Moderated-Mediation

2. Do children’s pre-kindergarten academic and social-emotional skills differ between
children who were unstably and stably housed?
I predicted differences between children who were unstably and those stably housed, such
that the unstably housed group would demonstrate lower mean scores in academic skills at prekindergarten assessment. However, I expected these differences to be less pronounced due to the
influence of Head Start, as consistent with previous research that has demonstrated that Head
Start enrollment is related several areas of academic (e.g., language and literacy) and socialemotional (e.g., social skills and problem behaviors) functioning (Bierman et al., 2009; Puma et
al., 2005; Zill & Resnick et al., 2006).
3. Do caregivers’ parenting practices and levels of depressive symptoms differ between
parents of children who were unstably and stably housed?
Consistent with previous research in the area of parenting and housing instability (e.g.,
Koblinksy et al., 1997), I expected that that caregivers of children who faced housing instability
during Head Start may have endorsed lower mean levels of parental engagement, parental
approaches (e.g., warmth), and higher levels of parental depression.
43

4. Do caregiver depressive symptoms, parenting practices, and classroom quality
differentially predict children’s academic and social-emotional functioning for families of
children who were unstably and stably housed?
I expected that parent depression would negatively predict children’s outcomes at Head Start
completion across both groups due to the strong association noted in the literature between
caregiver depression and children’s functioning (e.g., Goodman et al., 2011). I further
hypothesized that parenting practices and classroom quality would positively predict children’s
academic and social-emotional outcomes. I hypothesized that the predictive relations would
hold across both the unstably housed and stably housed groups and would likely function in a
similar fashion across both groups. However, I discuss the potential for the paths between
parenting practices and child outcomes to differ in RQ 5 based on the notion of resilience as part
of the hypothesized indirect pathways in RQ 5
5. Do parenting practices differentially mediate the relation between caregiver depressive
symptoms and children’s academic and social-emotional functioning for children who
were unstably and stably housed during Head Start?
Based on the Family Stress Theory of Economic Hardship, I predicted the direct and
indirect relationships depicted in Figure 1 for both stably housed and unstably housed families in
terms of the nature of the meditational relationship between the two groups. I expected a partial
meditational path within the full sample and both groups whereby parenting practices mediated
the relation between parental depression and child outcomes, which has been a robust finding
with regards to the Family Stress Theory.
Although I predicted that parenting practices would mediate the relation between parental
depression and child outcomes, the resilience/strengths-based literature, which has not typically

44

taken the mental health (e.g., depression) of caregivers into account, has studied parenting
practices as a moderator. Thus, although I predicted that the more likely pathway would be the
meditational pathway, especially when also accounting for parental depression it was possible
that a differential association could emerge. Given that the indirect pathways are a product of
two direct pathways, I predicted that if a difference did emerge in the full mediational relation, it
would occur between the direct paths from parenting (approaches, engagement) to child
outcomes (social-emotional, language and literacy). Although I hypothesized in earlier questions
that there may be mean level differences in depression among caregivers who faced housing
instability, I did not expect that differences in the associations between depression and parenting
for the two groups. Research has demonstrated that poverty and certain specific conditions of
poverty may moderate the relationship between depression and parenting (e.g., Goodman et al.,
2011), but I did not predict significant differences in this pathway between groups included in
the study because both groups will be drawn from a sample of families facing several risk factors
associated with poverty.
6. Does classroom quality differentially moderate the relation between parenting practices
and children’s academic and social-emotional functioning for children who were unstably
and stably housed during Head Start?
I predicted that classroom quality will have a differential effect for children facing
housing instability. This supports research on resilience that has indicated that environmental
factors may interact and be protective for children faced with housing instability (e.g., Militois et
al., 1999; Herbers et al., 2011). Although there is not literature on children facing housing
instability specifically, this prediction is rooted in research that indicates that classroom quality

45

can have a differential effect for children from specific groups facing higher risk with regards to
child and family-level indicators (e.g., Burchinal et al., 2006).
Similar to RQ 1, the hypothesized moderation would occur between potential mediators
(parenting practices) and outcomes (child functioning), signifying a moderated- mediation as the
full effect (Baron & Kenny, 1986; Hayes & Preacher, 2013) as depicted in Figure 2. If a
differential relation of classroom quality is present between groups, the influence of parenting
practices in the indirect effect of parental depression on child outcomes may be dependent upon
both classroom quality and housing status

46

Figure 3: Full Hypothesized Model with Covariates, Error, and Disturbance Terms
Covariate paths depicted as dashed lines; * = dummy coded in analyses

47

Table 1: Methods Chart
Research Question
1.Does classroom quality moderate the indirect relations between
parent depression and children’s outcomes through parenting
practices for the full pre-kindergarten sample?

Variables*

Covariates

Analysis

Parent/Caregiver
Depression
Parental Approaches
Parent Engagement
Language and
Literacy
Social-Emotional
Skills
Classroom Quality
Language and
Literacy
Social-Emotional
Skills

Child’s Age
Child’s Gender
Maternal Race
Family Economic Risk

Structural Equational Modeling
(SEM)

Child’s Age
Child’s Gender
Maternal Race
Family Economic Risk

Multi-group SEM -latent means

3. Do caregivers’ parenting practices and level of depressive
symptoms differ between parents of children who were unstably and
stably housed?

Parent/Caregiver
Depression
Parental Approaches
Parent Engagement

Child’s Age
Child’s Gender
Maternal Race
Family Economic Risk

Multi-group SEM -latent and
observed means

4. Do caregiver depressive symptoms, parenting practices, and
classroom quality differentially predict children’s academic and
social-emotional functioning for families of children who were
unstably and stably housed?

Parent/Caregiver
Depression
Parental Approaches
Parent Engagement
Language and
Literacy
Social-Emotional
Skills

Child’s Age
Child’s Gender
Maternal Race
Family Economic Risk

Multi-group SEM

2.Do children’s pre-kindergarten academic and social-emotional
skills differ between children who were unstably and stably housed?

48

Table 1 (cont’d)
5. Do parenting practices differentially mediate the relation between
caregiver depressive symptoms and children’s academic and socialemotional functioning for children who were unstably and stably
housed during Head Start?

Parent/Caregiver
Depression
Parental Approaches
Parent Engagement
Language and
Literacy
Social-Emotional
Skills

Child’s Age
Child’s Gender
Maternal Race
Family Economic Risk

6. Does classroom quality differentially moderate the relation
between parenting practices and children’s academic and socialemotional functioning for children who were unstably and stably
housed during Head Start?

Parent/Caregiver
Depression
Parental Approaches
Parent Engagement
Language and
Literacy
Social-Emotional
Skills
Classroom Quality

Child’s Age
Child’s Gender
Maternal Race
Family Economic Risk

*Bold= Latent Variable; Italics = Observed Variable (see Figure 5 for indicators for each latent variable)

49

Multi-group SEM

Multi-group SEM

CHAPTER 3
METHOD
The proposed study is a secondary analysis of data from the 2009 Head Start Family and
Child Experiences Survey (F.A.C.E.S.), funded by the U.S. Department of Health and Human
Services, Administration for Children and Families. The purpose of F.A.C.E.S. data collection
was to gather a wide range of longitudinal quantitative information on child, family, and centerbased characteristics to evaluate the outcomes of Head Start children, as well as inform Head
Start and early childcare initiatives more broadly (Malone et al., 2013). For the 2009 sample,
data was collected for over 3,000 children attending 60 Head Start programs located across the
U.S. To date, there have been four rounds of F.A.C.E.S data collected, beginning in 1997. The
present study extracted data from a subset of the 2009 F.A.C.E.S sample consisting of children
who were in their pre-kindergarten year. The study closely examined pre-kindergarten
differences between children who were unstably and stably housed during Head Start. Table 1
summarizes the corresponding variables and statistical approaches employed for each research
question.
The most recent round of F.A.C.E.S (2009) is a longitudinal study that spanned three
years of data collection (2009-2012) from a range of informants and sources. F.A.C.E.S data
collection included direct child assessments, evaluations of child functioning by parents and
teachers, observations of classrooms, and interviews with center directors. The F.A.C.E.S study
recruited families of three and four-year-old children at the start of their initial Head Start year
and additional data were collected on a yearly basis through the end of the child’s kindergarten
year. Children in the three-year-old-cohort were assessed in Fall 2009 (baseline), Spring 2010
(end of first Head Start year), Spring 2011 (end of second Head Start year), and Spring 2012

50

(end of kindergarten year). Children in the four-year-old-cohort were evaluated in Fall 2009
(baseline), Spring 2010 (end of Head Start year), and Spring 2011 (end of kindergarten year).
The present study was a cross-sectional examination of the pre-kindergarten year, with the
majority of data sampled at Head Start completion (Spring 2010 for four-year olds and Spring
2011 for three-year-olds). In addition, housing status of these children throughout Head Start was
also used for the present study. Relevant measures of interest are described below.
F.A.C.E.S Sampling Procedures
F.A.C.E.S data collection consisted of a multistage sampling procedure, during which
Head Start programs, centers, classrooms, and children were identified for sampling during each
of the stages (Malone at al., 2013). Although a full description of each stage of sampling is
detailed in the F.A.C.E.S user manual (see Malone et al., 2013), a summary of these procedures
is provided below.
Sampling for F.A.C.E.S employed a probability proportional to size (PPS) method across
the first three stages of sampling, which was based on the procedures described by Chromy
(1979). This procedure includes random sampling of units within randomly selected clusters. The
use of PPS helped ensure the least biased probability of inclusion in the study for each program,
center, and classroom. Families and children were randomly sampled using a computer algorithm
of approximately 36 newly enrolled children within each of the selected centers in the previous
stage of sampling. Sampling procedures further included stratification techniques at each stage
of sampling in order to ensure representative selection of the final sample. As described in the
F.A.C.E.S user manual, both explicit procedures that involved creation of strata based on a set of
characteristics (e.g., census region, minority enrollment, urbanicity), as well as implicit
procedures that involved pre-specification of specific characteristics (e.g., percentage of children

51

with disabilities, dual language learners, language within the home) prior to sampling were used
to decrease bias in sample selection. F.A.C.E.S did not oversample any subgroup of children or
classrooms for this particular cohort.
The sample of the F.A.C.E.S study resulted in the inclusion of a large sample of children,
families, classrooms, and centers. The study included 3,349 children across 486 classrooms
located in 60 programs at baseline in Fall of 2009. Approximately 10 children per classroom
with parent consent were targeted for inclusion in the study at baseline data collection. Children
who left Head Start during the years in which they were eligible for Head Start during the study
period were excluded from the original study. The F.A.C.E.S user manual indicates that these
children accounted for roughly 10% of sample loss from Fall of 2009 to Spring of 2010 and a
22% overall sample loss from Spring of 2010 to Spring of 2011.
Attrition
Unweighted estimates of the attrition for the total sample are presented in Table 2
below. Although the estimates below reflect the percentage of complete data across both three
and four-year-olds, the focus of the present study is on data from Head Start exit or the prekindergarten year (Spring 2010 for four-year-olds; Spring 2011 for three-year-olds).
Table 2: Unweighted Attrition Rates of Full Sample
Child Assessments

Parent Interview

Teacher Data

Fall 2009

94% (3,149)

93%(3,119)

97% (3,259)

Spring 2010

95%(2,879)

86% (2,601)

96% (2,906)

Spring 2011

89% (2,141)

79% (1,916)

81% (1,956)

Spring 2012

84% (935)

81%(896)

80%(892)

52

Weighting and Design Effects
Sampling weights are used in complex survey designs to account for unequal probability
of sampling and clustering that results from the employment of a stratified multistage sampling
procedure, as opposed to a simple random sampling procedure (Bell, Onwuegbuzie, Ferron, Jiao,
Hibbard & Kromey, 2012; Osborne, 2011). Weighting allows for generalizability of the results to
the full population of interest and adjusts for non-responders. Research studies that have
compared unweighted and weighted estimates of nationally representative data, including
previous rounds of Head Start F.A.C.E.S, have generally demonstrated that weighted analyses
tend to provide more accurate estimates of parameters and standard errors (e.g., Hahs-Vaughn,
McWayne, Bulotsky-Shearear, Wen and Faria, 2011; Thomas & Heck, 2001).
F.A.C.E.S documentation includes over 40 sampling weights to adjust for both the
probability of representation within the sample and nonresponse at the desired time point(s). The
sampling weight that was applied in the present study was PRAO5WT. This weight is intended
for cross-sectional analyses that incorporate data from the pre-kindergarten year, specifically
from the parent interview, child assessment, teacher interview, and classroom observation
measures.
Sample Characteristics
Full F.A.C.E.S Sample
The full F.A.C.E.S sample is intended to be a representative sample of children who
attended Head Start programming between the years of 2009-2012. The full unweighted sample
comprised 3,349 children who were enrolled in Head Start during the fall of 2009.
Approximately 48.3% (n= 1,619) of the sample were male and 48% (n= 1,609) were female. The
average age at the time of enrollment was 46 months (sd= 6.5) or 3.8 years of age.

53

Approximately 38% of children were identified by parents as Hispanic/Latino, 30% as African
American, 20% as White, and 5% as Multi-racial, 2% as Asian American and less than 1% as
American Indian or Alaskan Native or Other. At the time of Head Start enrollment,
approximately 58% (n= 1,945) of the sample lived in poverty and 61% (n= 2,004) reported that
they received multiple forms of public assistance.
Pre-kindergarten Sample
Due to the emphasis on the pre-kindergarten year functioning, the sample was weighted
to adjust for non-response and reflect the broader population during the pre-kindergarten year.
Weighting adjusted the sample to include those with data present for the pre-kindergarten
measures on the majority of indicators of interest. Although weighting excludes a sizeable
portion of the initial sample, Table 3 indicates that few significant mean differences at baseline
(Head Start entry) emerged between the included and excluded samples on variables used within
SEM modeling in the present study that were available at baseline (Fall 2009). Independent
samples t-tests indicated that in comparison to those without pre-kindergarten sampling weights,
children with pre-kindergarten weights had higher mean parent-rated problem behavior scores,
t(3044)= -2.96, p<.005 and lower standardized receptive vocabulary scores, t(29808)= 3.03,
p<.005. Yet mean scores indicated only an approximate three point difference on standardized
testing of receptive vocabulary.
Table 3: Unweighted Baseline Mean Scores of Available Predictor and Outcome Variables
at Head Start Entry

Parent
Depression
Parent Rated

Included –
Pre-K
Weight
Mean (SD)
4.76(5.80)

Excluded –
No Pre-K
Weight
Mean (SD)
5.10(6.05)

5.72 (3.72)

5.33(3.50)

54

Table 3 (cont’d)
Problem
Behaviors*
Parent Rated 12.03(2.55) 12.00(2.56)
Social Skills
Weekly
11.30(2.14) 11.35(2.00)
Parent-Child
Activities
Score
Monthly
4.82(2.21)
4.68(2.14)
Outside
Activities
Score
Woodcock
95.54
96.25(19.30)
Johnson
(17.29)
Letter Word
Identification
Peabody
79.88
82.09(19.05)
Picture
(20.83)
Vocabulary
Test*
Expressive 78.47(15.50) 78.40(15.61)
One Word
Vocabulary
Test
* Indicates significant mean differences at p<.005
The final pre-kindergarten year sample used within the study was comprised of 1,554
children (weighted n= 356,547). As indicated in Table 4, the sample was composed of equal
numbers of male and female children. Similarly, approximately half of the sample entered Head
Start at the age of three and the other half at the age of four. Children’s primary racial identity
was reported as follows: 40% Hispanic/Latino, 32% Black/African American, 21%
White/Caucasian, 5% Multi-Racial, 2% Asian/Pacific Islander, and less than 1% identifying as
other.
Approximately 87% (n=1,313) of survey respondents identified as a biological or
adoptive mother. Therefore, maternal characteristics were of primary interest in the study.

55

Maternal race followed a similar pattern of representation as child’s race: 38% Hispanic/Latino,
32% Black/African American, 26% White/Caucasian, 2% Asian/Pacific Islander, 2% MultiRacial, and 1% identifying as other. Family-level indicators of economic risk indicated that 63%
(n= 962) of the sample reported meeting criteria for federal definitions of poverty status at the
time of pre-kindergarten data collection. Furthermore, 45% (n= 708) of the children in the
sample were from a single-parent home with the mother as the primary caregiver and 43%(n=
664) were from two-parent homes with both mother and father as caregivers. Respondents
reported that approximately 37% (n= 537) of mothers attained less than a high school diploma,
33% (n= 484) earned a diploma, 24% (n= 326) obtained vocational training or some college, and
6% (n= 89) earned a Bachelor’s degree. According to native language information collected only
at baseline, 70% (n= 1,035) of the pre-kindergarten sample reported that English was the primary
language spoken to the child within the home at the time of Head Start entry.
Unstably Housed Sample
A primary sample of interest in the present study is families who faced housing
instability during Head Start. The primary inclusion criteria for housing instability in the present
study was based on the housing status of families at each data collection time point (e.g., Head
Start entry, pre-kindergarten year) who were also assigned a pre-kindergarten weight used for
data analysis. Consistent with previous conceptualizations of housing instability within the
research literature, the present study defined housing instability as either a lack of stable housing
or as families who faced frequent moves in a twelve-month period of time. In other words,
children with the relevant pre-kindergarten sampling weight who lacked stable housing (e.g.,
doubled-up, shelters) at any data collection time point before kindergarten entry and/or those
who faced at least two moves during the twelve-month period prior to a data collection time

56

point were included in the unstably housed group (Cutts, 2011). Although 25% (n=375) children
with a pre-kindergarten sampling weight present experienced housing instability at least once
during Head Start, only children for whom housing information was available for a minimum of
two data collection time points were included in the creation of the housing status variable in
order to most accurately summarize the status of families in the sample as either having
experienced instability or stability, resulting in a final sample of 368 children.
As detailed in Table 4, approximately 24% (n= 368, weighted n= 84,072) of children in
the pre-kindergarten sample were unstably housed at least once during Head Start.
Approximately 23% (n= 88) of the unstably housed families reported experiencing housing
instability at two or more time points during Head Start. Child characteristics were comparable
to the larger pre-kindergarten sample. Also consistent with the larger sample, approximately 90%
(n= 328) of survey respondents identified as a biological or adoptive mother. Family level
characteristics, including maternal race and economic stressors were also explored among the
unstably housed group. Mother’s identified their primary race as follows: 39% Hispanic/Latino,
26% Black/African American, 32% White/Caucasian, 1% Asian/Pacific Islander, 1% MultiRacial, and 1% identifying as another race. Family-level indicators of economic risk indicated
that 70% (n= 254) of the unstably housed sample reported meeting criteria for federal definitions
of poverty status at the time of pre-kindergarten data collection. Furthermore, 51% (n= 187) of
the children in the unstably housed group were from a female led single-parent home and
36%(n= 138) were from two-parent homes. Among the respondents who were unstably housed
during Head Start, 41% (n= 143) of mothers attained less than a high school diploma, 36% (n=
126) earned a diploma, 20% (n= 69) obtained vocational training or some college, and 3% (n=
12) earned a Bachelor’s degree.

57

Table 4: Pre-Kindergarten Sample Characteristics with Stable and Unstable Groups
Full Pre-Kindergarten Sample
%
Total Sample

100
%

Unweighte
dN
1,554

Weighted
N
365,547

Survey Respondent
Mother

87%

1,313

304,865

Father
Grandparent
Other

7%
4%
2%

102
56
20

3,074
2,676
1,720

Child’s Gender
Male

50%

775

181,890

Female

50%

771

182,022

Mother’s Race
White/Caucasian

26%

353

92,838

African American

32%

485

115,558

Hispanic/Latino

38%

640

136,659

Asian
Multi-Racial
Other
Child’s Race
White/Caucasian

2%
2%
1%

18
30
12

6,143
8,810
2,417

21%

296

75,639

African American

32%

485

117,020

Hispanic/Latino

40%

668

143,652

Stably Housed
%

Unstably Housed

Unweighted
Percent (N)
1179

Weighted
Percent (N)
280,097

24%

Unweighted
Percent(N)
375

Weighted
Percent(N)
85,450

86
%
8%
5%
2%

985

230,012

90%

328

74,853

77
47
17

20,423
12,810
4,953

7%
3%
<.5
%

25
9
3

5,693
2,528
442

50
%
50
%

588

139,739

49%

187

43,299

583

183,773

51%

188

42,152

24
%
34
%
37
%
2%
3%
1%

249

65,798

32%

104

27,040

389

93,520

26%

96

22,038

477

103,074

39%

163

33,585

13
27
9

4,764
7,975
1,988

2%
1%
1%

5
3
3

1,379
835
429

20
%
34
%
39
%

211

54,725

25%

85

20,914

391

95,109

26%

94

21,911

500

108,477

41%

168

35,175

75
%

58

%

Table 4 (cont’d)
Asian
Multi-Racial
Other

2%
5%
<1%

19
64
5

5,828
18,149
1,162

2%
5%
<1
%

16
44
3

4,916
12,549
728

1%
7%
1%

3
20
3

912
5,601
435

Child’s Cohort
Three-Year-Old

50%

739

183,047

566

141,137

50%

173

41,910

Four-Year-Old

50%

815

183,000

50
%
50
%

613

138,960

51%

202

43,540

Poverty Status
No

37%

577

135,496

463

110,260

30%

114

25,209

Yes

63%

962

227,000

40
%
60
%

704

167,340

70%

258

59,399

Family Structure
Bio/Adoptive Mother and Father

43%

664

155,713

526

125,707

35%

138

30,006

Bio/Adoptive Mother Only

45%

708

165,331

514

120,927

52%

194

4,404

Bio/Adoptive Father Only
One Bio Parent/One non-bio Parent
Biological Grandparents
Other Non-Relative
Maternal Education
Less than Diploma

2%
5%
1%
4%

32
81
15
54

7,394
184,741
4,302
14,066

45
%
43
%
2%
5%
1%
4%

26
60
12
41

5,573
13,864
3,154
10,872

2%
6%
1%
4%

6
21
3
13

1,821
4,877
1,148
3,194

37%

537

124,080

390

90,130

42%

147

33,951

Diploma

33%

484

111,298

356

82,754

35%

128

28,544

Vocational/Some College

24%

326

7,8913

257

62,833

20%

69

16,080

Bachelor Degree or Higher
Home Language at Entry
English

6%

89

20,900

36
%
33
%
25
%
7%

76

18,495

3%

13

2,338

70%

1035

254,539

792

196,527

68%

243

58,012

Non-English

30%

511

109,373

71
%
29
%

379

81,944

32%

132

27,438

59

Missing Data
Malone et al. (2013) indicated that F.A.C.E.S sampling weights adjust for nonresponse of
full instruments and minimize missing data, however that missing data can still be present at the
item or survey level. Upon considering sampling weights, missing information on demographic
variables of interest (e.g., race, age, gender) ranged from 0-3% on all study variables of interest.
Missing data across variables of interest used within modeling ranged from 0 – 6%, with the
exception of expressive language scores, for which 20% of the data was not present. However, as
described in the F.A.C.E.S User Guide, data for language assessments may be unavailable to a
greater extent than other forms of information due to specific routing procedures used to
determine the tests that should be given at each wave (Malone et al., 2013).
The present study employed a robust maximum likelihood (ML) estimation method,
which uses the Full Information Maximum Likelihood (FIML) as the default method to address
missing data (Enders & Bandalos, 2001; Kline, 2011). This approach uses all available or
observed information to estimate the model, as opposed to deleting or imputing missing values
(Enders & Bandalos, 2001). Although no value is substituted, this approach functions similarly
to multiple imputation due to the computation of a model based on observed information.
However, an assumption of FIML according to Enders (2010) is that data is either
missing completely at random (MCAR) or missing at random (MAR). Under the MCAR
assumption, data that is missing is not thought to be dependent upon the value of a specific
variable. For example, children who scored low on reading tasks were no more likely to be
missing on the letter-word identification task than those who scored high. The MAR assumption
is less restrictive and assumes that data that is missing may be dependent upon other variables,
but that these variables are accounted within the model. For example, lower literacy and

60

language functioning may be dependent upon economic risk indicators, and therefore, the model
accounted for these relations. It is important to highlight there is no true method to determine
missing data mechanisms, particularly given the complex nature of secondary data analysis of
national data. Although weighting partially adjusts for nonresponse and ensures that a given
score represents the appropriate number of persons in the population, the most salient
theoretically driven relations were accounted for in the model under the assumption that data
may be MAR.
Variables and Measures
Data of relevance to the proposed study included select information from direct child
assessments, observations of the classroom, rating scales (parent and teacher), and parent
interview. The present study included five latent variables: parental approaches, parent
engagement, social-emotional skills, language/literacy, and classroom quality. Indicators for the
hypothesized latent variables are described separately below and depicted in Table 5.
Table 5: Hypothesized Latent and Observed Variables For Inclusion in Study
Construct
Parent Approaches

Variable Type
Latent

Parent Engagement

Latent

Social-Emotional Skills

Latent

Language & Literacy

Latent

Classroom Quality

Latent

61

Indicators
Warmth
Authoritative
Authoritarian
Energy
Head Start Involvement
Parent-Child Weekly Activities
Parent-Child Monthly Out of Home
Activities
Parent Rated Problem Behaviors
Teacher Rated Problem Behaviors
Parent Rated Social Skills
Teacher Rated Social Skills
Woodcock Johnson Letter Word
Identification
Peabody Picture Vocabulary Test
Expressive One Word Picture Vocabulary
Test
Early Childhood Environmental Rating
Scale, Short Form

Table 5 (cont’d)

Parent/Caregiver Depression
Family Economic Risk Index
Maternal Race
Child’s Gender
Child’s Age at Entry
Housing Status

ObservedContinuous
ObservedCategorical
ObservedCategorical
ObservedDichotomous
ObservedContinuous
ObservedDichotomous

Classroom Assessment Scoring SystemEmotional Support, Instructional Support,
and Classroom Organization Subscales
-------

Housing Status
Housing instability was represented by a dichotomous variable (HINST 0= No, 1= Yes)
that was created from two existing variables at multiple time points in the dataset: type of
housing (PnM07) and the number of moves experienced the twelve months prior to the data
collection time period (PnM08). Consistent with research and legal definitions of housing
instability, families that reported two or more moves during the twelve months prior to data
collection time points or those that identified as living doubled up or in transitional housing at
any time point during Head Start were coded as having faced housing instability. For children
who were in the three-year-old cohort, housing status information was extracted from Head Start
entry, spring of their first Head Start year, and the pre-kindergarten year. For children who were
in the four-year-old cohort, housing status information was extracted from Head Start entry and
the pre-kindergarten year. As noted in the description of the unstably housed sample, only
children with the relevant pre-kindergarten sampling weight for whom housing information was
available for a minimum of two data collection time points were included in the creation of the

62

housing instability variable in order to most appropriately summarize the housing status of
families in the sample.
Child Functioning
Child functioning during the pre-kindergarten year was assessed in terms of children’s
academic and social-emotional skills. The proposed study conceptualized academic skills using
standardized literacy and language measures. Social-emotional functioning was comprised of
parent and teacher-rated social skills and problem behaviors.
Pre-literacy and language skills. Although academic functioning comprises a range of
skills, research has demonstrated that language and literacy skills are important predictors of
future academic achievement (e.g., Foster et al., 2005), and therefore, were used to assess
academic functioning of children in the present study. The Letter-Word Identification subtest
from the Woodcock Johnson-III Tests of Achievement or The Batería III Woodcock-Munoz for
Spanish-speaking bilingual children measures letter-sound correspondence, an early indicator of
phonological awareness. In this task, children were asked to identify pictures, letters, and words.
Trained data collectors administered this direct child assessment at each data collection time
point in the study. The standard score was used in data analyses (M= 100, SD= 15). Notably,
F.A.C.E.S. discontinued the test after three incorrect responses, which is inconsistent with
standardized directions of the subtest that dictate that test items are discontinued after six
incorrect responses. These changes to the standardized directions may have underestimated the
literacy skills of some children.
The Letter-Word Identification sub-test has been used in previous published work using
F.A.C.E.S (1997; 2006) datasets to conceptualize early literacy skills (Bulotsky-Shearer et al.,
2012; McWyane et al., 2012). Cronbach’s alpha estimates for the current sample ranged from

63

.85-.93 for the English version and .67-.97 for the Spanish version across all time points (Malone
et al., 2013). Developers of the sub-test reported split-half reliability coefficients that ranged
between .97 and .99 for English speaking children and .84-.98 for Spanish speaking children
(Woodcock, McGrew & Mather 2001; Woodcock, Munoz-Sandoval, McGrew, & Mather, 2007).
The sub-test has also been noted to demonstrate high theoretical construct validity (Woodcock, et
al., 2001).
Two domains, receptive and expressive vocabulary, were used to measure language
functioning. Consistent with previous published work using F.A.C.E.S. data (e.g., BulotskyShearer et al., 2012), two widely used reliable and valid instruments, the Peabody Picture
Vocabulary Test, 4th Edition (PPVT-4) and Expressive One Word Vocabulary Test (EOWVT-3),
were used in the present study to assess language. Data tables from the current sample indicated
reliability estimates on the PPVT-4 that ranged from .91 to .97 for the English version (Malone
et al., 2013). The test has noted concurrent validity estimates ranging from .70 to .80 with similar
measures of language (Dunn & Dunn, 2007). Developers of the test reported split- half reliability
estimates of .95 to .93 for children 3 to 5 years of age. Developers of the EOWVT-3, reported
split-half estimates for all ages ranged from .96 to .99 and average test retest reliability is
reported to be approximately .90 (Brownell, 2000). Cronbach’s alpha estimates for the current
sample ranged from .86 to .90 for the EOWVT-3 (Malone et al., 2013). Standard scores from the
tests were included in data analyses (M=100, SD= 15).
Although bilingual versions of the language tests (Test de Vocabulario de Imagnes
Peabody and Expressive One-Word Picture Vocabulary Test, Bilingual Edition) were also
administered in select cases, the English norms were used in the present study for two primary
reasons. First, scores across the English (e.g., PPVT-4) and bilingual versions (e.g., TVIP) are

64

not interchangeable due to the use of different norming samples (Brassard & Boehm, 2007;
Malone et al., 2013). Since all of the children in the sample were administered the PPVT-4, the
standard scores derived using the English norms were included in analyses. Second, scores using
both sets of norms (English and bilingual) on the EOWVT-3 were reported for select children,
resulting in overlapped scores and a lack of clarity on the most appropriate score for each
particular child. In order to remain consistent, the standard scores derived from English norms
for the EOWVT-3 and PPVT-4 scores were included in the present study. Lastly, the use of the
scores from the English versions of the tests is consistent with previous research that has targeted
analyses to the broader group of Head Start children (e.g., Choi, Elicker, Christ & Dobbs-Oates,
2016), however it should be noted that use of English scores may underestimate language
functioning of some dual-language students within the sample.
Social-emotional. Children’s social skills and problem behaviors were assessed to
examine social-emotional functioning. Four sets of measures comprised these two domains in the
present study. Social-emotional functioning was measured through rating measures that
represented parent-reported social skills (PnSSPAL), parent-reported problem behaviors
(PnPBERPB), as well as teacher-reported social skills (RnSSRS) and teacher-reported problem
behaviors (RnBPROB2). Parent interviews were completed both in-person and via telephone by
F.A.C.E.S staff at Head Start centers and teachers were provided rating scales. Social skills and
problem behaviors were measured across 21 parent items and 26 teacher items using a
combination of items adapted from valid and reliable instruments, including the Social Skills
Rating System (SSRS; Gresham & Elliot, 1990), Behavior Problem Index (BPI; Peterson & Zill,
1986), and the Personal Maturity Scale (Alexander, Entwilse & Thompson, 1987). Due to

65

copyrights on select measures (e.g., SSRS), item-level information was not provided in the
dataset.
F.A.C.E.S included two composite scales produced from these items, social skills and
problem behaviors, which were available in the dataset and were used to serve as an overall
index of these constructs in the present study. Parent-reported social skills scores also included
approaches to learning questions, whereas teacher-reported social skills score did not. All socialemotional measures were used in previous rounds of F.A.C.E.S data collection (Malone et al.,
2013; West, Tarullo, Aikens, Sprachman, Ross & Carlson, 2007).
Teacher ratings. Teachers rated children on a 3-point Likert scale with items ranging
from never (1) to very often (3). Social skills included 12 teacher items related to the child’s
cooperative classroom behaviors (e.g., assisting with cleanup, following rules). Items were
drawn from the Personal Maturity Scale and SSRS. The Personal Maturity Scale was originally
used in the National Survey of Children and later in a large longitudinal study of children in the
Baltimore schools (e.g., Alexander et al., 1987). The full scale consists of 13 items, measuring
attention, cooperation, compliance, and perceived interest. Earlier studies reported moderate to
high reliability (α= .77 and .86) of the items with young children in first and second grade
(Alexander et al., 1987; Alexander, Entwisle & Dauber, 1993). Similarly, SSRS test developers
reported high reliability (α= .94) for preschool –aged children. Cronbach’s alpha estimates of the
combined social skills items for the current F.A.C.E.S sample ranged from .88 to .90 for the full
sample (Malone et al., 2013).
Teachers rated the frequency of 14 specific problem behaviors that included aggressive
behaviors, hyperactivity, and withdrawn behaviors. The items were drawn from the Personal
Maturity Scale and BPI. The BPI, based on the Child Behavioral Checklist, was used in two

66

previous national studies, including National Health Interview Survey and the National
Longitudinal Study of Youth, and demonstrated moderate to high reliability, with estimates that
ranged from .80 and .91 in studies that have used items from these surveys (Brand & Brinich,
1999; Zlotnick, Johnson & Khon, 2006). Reliability estimates for the current F.A.C.E.S sample
ranged from .86 to .88 (Malone et al., 2013).
Parent ratings. Parents rated their children also on a 3-point Likert scale with items
ranging from 1 (not true) to 3 (very true). The 21 parent items were drawn from the similar
measures described above for teachers. Parent items included measurement of problem behaviors
(e.g., aggression) and ratings of their perceptions of children’s social skills (e.g., ability to make
friends). Entwisle and Alexander (1987) demonstrated adequate reliability (α= .87) of items from
the Personal Maturity Scale with parents. Test developers of the SSRS reported internal
consistency estimates ranging from .73 to .87 for problem behavior subscales from which items
were drawn for the present study (Gresham & Elliot, 1990). Lastly, the BPI also demonstrated
reliability estimates close to .90 for parents in previous national studies (e.g., Brand & Brinich,
1999). Cronbach’s alpha estimates for the current sample ranged from .72 to .79 for problem
behaviors (Malone et al., 2013). Reliability coefficients for the social skills items in the present
F.A.C.E.S were slightly lower and ranged from .68 to .72 across the data collection time points
(Malone et al, 2013).
Parenting Practices
Consistent with a comprehensive conceptualization of parenting that includes
consideration of both engagement and approaches to parenting, parenting practices were
represented by seven available composite scores for parental warmth (PnWarm), authoritative
parenting (PnAuthv), authoritarian parenting (PnAuthrn), parental energy (PnEnergy), parent

67

involvement in Head Start (PnPInvHS), home-based parent-child activities (PnPWKAC2), and
out of home parent-child activities (PnMoAct). These items were administered as part of the
parent interview at the completion of each Head Start year.
Parent approaches. Items that comprised the warmth, authoritative, authoritarian, and
energy subscales were derived from 13 items from the Child Rearing Practices Report (CPR;
Block, 1965). Parents rated the extent to which they believed statements reflected their practices
on a scale of 1 (exactly) to 5 (not at all). However, F.A.C.E.S developers did not compute the
original subscales of the CPR that were validated as part of the test development, but rather new
subscales that grouped similar items of the measure have consistently been used across
F.A.C.E.S cohorts (West et al., 2007). Items from the measure have also been used in other
national longitudinal studies (e.g., Early Childhood Longitudinal Study -Birth Cohort) and
secondary analyses of F.A.C.E.S data (e.g., Bulotsky-Shearer et al.,2012). Previous research has
demonstrated construct and content validity of the larger CPR scale items (e.g., Kochasnka,
Kuczysnki & Radke-Yarrow, 1989). In the present study, the internal consistency of the 13 items
that comprised parental approaches was moderate (α= .70). Each of the F.A.C.E.S derived
subscales are described below. All composites were readily available within the F.A.C.E.S
dataset.
Warmth. The 4 items from the measure that comprised the warmth score reflected use of
parent perceptions of their interactions with their child. Example items include “my child and I
have warm intimate moments together” and “I make sure my child knows that I appreciate what
(he/she) tries to accomplish.”
Authoritative. The authoritative parenting score was derived using 3 items that were
intended to screen for parenting behaviors that reflected parental beliefs about setting boundaries

68

and encouraging independence. Example items include “I encourage my child to be independent
of me” and “I teach my child that misbehavior or breaking the rules will always be punished one
way or the other.”
Authoritarian. The authoritarian parenting score was derived using 3 items that assessed
parental attitudes about physical punishment and setting rules. Example items include “I believe
physical punishment to be the best way of disciplining” and “I believe a child should be seen and
not heard.”
Energy. The energy score was comprised of 3 items that were intended to reflect the
parents’ consistency in adherence to rules and consequences for misbehavior. Example items
include “Once I decide how to deal with a misbehavior, I follow-through on it” and “there are
times, I just don’t have the energy to make my child behave as they should.”
Parent engagement. Parental engagement items included involvement in Head Start
activities and enrichment activities both within and out of the home. With regards to involvement
in Head Start, parents were administered 14 items related to the frequency of their involvement
in Head Start during the years their child was enrolled at Head Start (e.g., volunteered in
classroom, attended conferences) that comprised one score (PnINVHS). Items were scored as 1
(not yet) to 5 (once a week). Items demonstrated face validity with previous parent engagement
research related to activities that foster children’s schooling experiences and skills during early
childhood (Fantuzzo, Tighe, & Childs, 2000; McWayne et al., 2004). Similar to published
research with a previous cohort of F.A.C.E.S (2006) data, which indicated an adequate internal
consistency (α= .80) estimate of parent involvement items, the present study found support for
strong internal consistency (α= .87) across parent involvement items (McWayne et al., 2012).

69

Parent/child activities. Parents were administered 14-items that gauged whether they
engaged in home enriching parent-child activities within the previous week of the interview,
including telling stories, teaching letters-numbers, and playing games. In addition, 11-items that
assessed whether parents engaged in out of home parent-child activities on a monthly basis was
also administered, which included activities such as, visiting the zoo, library, and community
events. These items comprised two scores: one for weekly activities (PnWCK2) and monthly
out of home activities (PnMOACT). All items were drawn from use in previous national
longitudinal studies (e.g., National Household Education Survey, Early Childhood Longitudinal
Study -Birth Cohort), including earlier rounds of F.A.C.E.S data collection (Malone et al., 2013;
West et al., 2007). Consistent with previous studies that demonstrated variable internal
consistency (α= .60-.66) of enrichment items across studies (Bulotsky-Shearer et al., 2012;
Foster et al., 2005), Cronbach’s alpha estimates in the present sample with pre-kindergarten
items indicated borderline acceptable estimates for weekly activities (α= .67) and monthly out of
home activities items (α= .64). However, these items demonstrated face validity with broader
measures of cognitive stimulation within the home environment (e.g., Home Observation for
Measurement of the Environment [Bradley & Caldwell, 1984]).
Parent Depression
The short form of the Center for Epidemiological Studies-Depression Scale (CES-D;
Radloff, 1977), a 12 item screener for depression, was administered at each wave to the
parent/caregiver who responded to the survey items. The scale is designed to screen for current
(state-specific), on-going depression symptomology. Respondents rated their level of
symptomology on a 4-point scale during the parent interview. Scores range from 0-36, with a
cutoff score of 15 indicating severe depression, 10-14 indicating moderate levels of depression,

70

and 5-9 indicating mild levels of depression. Internal consistency of the measure is reported at
roughly .85 and ranged from .86 to .89 in the current sample across the data collection time
points (Malone et al., 2013; Radloff, 1977). The full CES-D demonstrates concurrent validity
with similar measures, including the Beck Depression Inventory (Radloff, 1977).
Classroom Quality
Trained observers rated classroom quality across both structural (OnECERSS) and
process quality (OnCLSSIS, OnCLSSES, OnCLSSO) indicators at the end of each Head Start
year. Structural aspects of classroom quality were measured using 21 items from the Early
Childhood Environmental Rating Scale, Short Form (ECERS-R; Harms, et al., 2005). Items
represented in the dataset from this measure were most closely aligned with the Teaching and
Interactions, as well as Provisions for Learning subscales. A factor analysis study indicated that
items of the ECERS-R loaded on one factor, suggesting that that using a smaller sub-set of the
items may be as reliable as the full measure (see Perlman, Zellman & Le, 2004). Total ECERS-R
scores are comprised of a mean score and range from inadequate (1) to 7 (excellent). Reliability
estimates in the F.A.C.E.S sample ranged from .85 to .87 for the ECERS-R (Malone et al.,
2013). The short form also demonstrated high inter-rater reliability, with estimates ranging from
.83 to .93 for the present sample.
Process quality indicators from the Classroom Assessment Scoring System (CLASS;
Pianta et al., 2008) were also available in the dataset. Scores for the three broad domains that
comprised the eleven subscales, Emotional Support, Instructional Support, and Classroom
Organization were used to represent the process quality components of classroom quality.
Trained observers rated classrooms on each of the subscales that measure student-teacher
interactions and aspects of the classroom that promote student learning, assigning a mean score

71

from 1 (minimally characteristic) to 7 (highly characteristic) for each of the domains rated. The
C.L.A.S.S is a widely used tool in research and has been included in several large-scale,
nationally representative studies (e.g., Hamre et al., 2007; La Paro, Pianta & Stuhlman, 2004). A
large validation study confirmed the theorized three-factor structure (Hamre et al., 2007). La
Paro et al. (2004) reported moderate associations of the C.L.A.S.S subscales with the ECERS-R
total score (r= .52 and .40). Cronbach’s alpha estimates for the F.A.C.E.S sample ranged from
.82 to .90 (Malone et al., 2013). Malone et al. (2013) further reported that inter-rater reliability
estimates demonstrated high correspondence between raters, with estimates ranging from .81 to
.86 across subscales.
Covariates
Four covariates of theoretical relevance to constructs in the model were selected for
inclusion in the study as control variables. Each is described below.
Family economic risk. The family economic risk (PnECRISK) risk variable is a
F.A.C.E.S derived index score that was created by the developers using three economic risk
variables: family-level poverty status (PnPovtry), maternal education (PR1MOMED), and family
structure (PnFMSTRC). The family economic risk variable represents the number of risk factors
faced by the family at each time point, with a max score of three. A score was assigned based on
the number of risks that included meeting criteria for poverty status, a maternal education below
a high school diploma, and being from a single-parent home. The economic risk score included
in the present study was dummy coded for inclusion in the model.
Age. The child’s age (P1RCAGE) in months at Head Start entry was included to control
for differential years of exposure to Head Start across the three and four-year-old cohorts by the
time they reached the pre-kindergarten year.

72

Maternal race. Maternal race (MRACE) was selected for inclusion due to the large
percentage of respondents identifying as the child’s mother (~90%). Due to the low frequency of
certain groups, the MRACE variable was re-coded to represent a collapsed variable that included
identification as White, Black/African American, Latino/Hispanic, or within a category termed
“other” to represent the additional sample. Maternal race was dummy coded for inclusion in the
model.
Child’s gender. Child’s gender (CHGENDER) was represented by binary variable that
identified the gender of the child as male or female.
Analytic Plan
Table 1 depicts the analyses with their corresponding research questions. Analyses were
conducted using the Statistical Package for the Social Sciences (SPSS Version 23) with the
Complex Samples Module and Mplus Version 7.31 (Muthén & Muthén, 2015). Both packages
were selected for their ability to handle complex sampling designs. The primary research
questions were answered through Structural Equation Modeling (SEM) and Multi-Group SEM in
Mplus Version 7.31. However, several steps were conducted to execute these analyses that
included: data cleaning/preparation, preliminary analyses, Confirmatory Factor Analysis (CFA),
testing the full conceptual model through SEM with latent variables, subpopulation analyses, and
Multi-Group SEM.
Data Preparation and Cleaning
Initial stages of data analysis consisted of data preparation and cleaning. The raw format
of the data consists of child and family data by year (e.g., Fall, 2009, Spring, 2010) and is not
separated by cohort. Although all children in the study entered Head Start in Fall of 2009,
approximately half the sample was three years of age at Head Start entry and the other half was

73

four years of age, resulting in two potential pre-kindergarten years, Spring 2010 for the fouryear-old cohort and Spring 2011 for the three-year-old cohort. As consistent with
recommendations from the F.A.C.E.S user manual, pre-kindergarten variables were created
using the PKYEAR grouping variable to guide selection as to the appropriate wave from which
to draw data for each child in the study. All pre-kindergarten variables were created using the
syntax feature of SPSS Version 23.
Preliminary Analyses
Preliminary analyses, including descriptive statistics, correlations, and t-tests were
conducted on observed variables prior to considering the full model. These analyses were
conducted with and without sampling weights. Preliminary analyses served two purposes in the
current study. First, descriptive statistics and correlations of the predictor and outcome variables
provided greater insight into the sample and relations between the study variables to rule-out
issues that may hinder statistical analyses (e.g., multicollinearity). In addition, preliminary group
difference testing without latent variables, such as t-tests between stably housed and unstably
housed groups on parenting practices, caregiver depression, pre-academic (language and literacy)
scores, and social-emotional skills provided a baseline from which to understand potential
sources of difference between groups. However, differences using the full latent variable model
were of primary interest.
Confirmatory Factor Analysis
CFA was used to test the latent factor structure of the model (Figure 3). Several observed
variables or indicators, further described in the variables and measures section, were
hypothesized to load on each construct based on theoretical and practical considerations. For
example, scores on a receptive language test, expressive language test, and a letter-word

74

identification task were selected to load together on a construct that comprised language and preliteracy skills. CFA was conducted in a two-step fashion as proposed by the broader literature
(Brown, 2015; Anderson & Gerbing, 1988). These steps included initially testing the factor
structure of each construct separately, followed by testing the entire measurement model that
included all relevant latent constructs and their indicators. Additionally, the latent constructs
were first tested on the full sample, followed by the sub-populations of interest (unstably housed
and stably housed) prior to the multi-group analyses. Five latent variables were used in the study:
parenting approaches, parent engagement, classroom quality, language and literacy, and socialemotional skills.
Structural Equation Modeling with Full Sample
The next step consisted of testing the overall structural model (Figures 2 and 3) with the
full sample (RQ 1). All hypothesized paths of interest, including the moderated-mediation, were
specified in the full model. A latent variable SEM approach was selected in the proposed study
to parsimoniously and simultaneously test a number of observed variables as latent constructs
through the specification of both direct and indirect paths. The hypothesized base model was respecified using modification indices and theory as detailed further in the results section.
Although several scholars discuss the relevance of resampling procedures (bootstrapping)
in testing mediation, this approach is not compatible in complex survey designs due to the lack
of independence across observations (Preacher & Hayes, 2004; Muthén & Muthén, 2015).
Therefore, the present study included the default settings in Mplus to test mediation via the
MODEL INDIRECT command for the base model and an identical procedure that computes new
parameters using MODEL CONSTRAINT for the moderated-mediation. The default approach to
testing mediation in Mplus is the delta method, a method that estimates if the standard error

75

distribution of the product of the two direct effects (SEβa·SEβb) that comprise the mediation are
significantly different from zero (Bollen & Stine, 1990). Similar to other related tests of
mediation (e.g., Sobel test), this approach assumes that distributions are normally distributed and
may produce more biased estimates of the standard error when sample sizes are small. Several
scholars have demonstrated that testing mediational effects with sufficiently large samples (N >
150) produces similar estimates across both the delta and bootstrapped methods (Bollen & Stine,
1990; Tofighi & MacKinnon, 2016).
Recent literature has described a two-step process for estimating moderation models with
latent variables interactions (Maslowsky, Jager & Hemken, 2014; Muthén & Asparouhov, 2015).
Maslowsky et al. (2014) indicated that a model without the latent variable interaction is first
fitted and adjusted, particularly due to the underdeveloped methods for evaluating model fit with
latent variable interactions. In Mplus, latent variable interaction models require alternate
commands (e.g, Type = RANDOM) that produce limited fit statistics. Since it is assumed that
latent variable interactions add no additional variance, means, or covariance that are not already
included in the base model, overall fit is expected to be comparable to the base model (Muthén &
Asparouhov, 2015). The model without the interaction serves as the base model to which the
latent variable model is compared using a more limited set of indices that emphasize parsimony
(e.g., Alkaline Information Criterion). In the present study, the moderation paths of interest, the
differential influence of classroom quality on the hypothesized mediational paths, caregiver
depression on children outcomes through parenting, were computed by creating latent variable
interaction terms as outlined by Hayes and Preacher (2013) in their discussion of conditional
process analysis. Specifically, these terms were computed between classroom quality and

76

parenting (parent engagement · classroom quality and parent approaches · classroom quality)
using the XWITH command in Mplus.
The present study used several absolute and relative SEM model fit statistics. As per
recommendations found within the literature, decisions of model fit were based on the Root
Mean Square Error of Approximation, Comparative Fit Index, Standardized Root Mean Square
Residual and Chi-Square Test (Kline, 2011). Consistent with Kline (2011), these fit indices,
with cut-off criteria stated below, were used to make decisions regarding model fit. It was
expected, however, that the Chi-Square Test would be significant and not an appropriate index
due to the large weighted sample size, and thus, the other indices were weighted more heavily in
determining overall model fit (Kline, 2011; Hu & Bentler, 1999). The Alkaline Information
Criterion was further used for model comparisons.
Chi-Square ( χ 2 ). The chi-square statistic tests model fit or misfit with the overall data.
A non-significant chi-square value indicates better fit, however, this statistic often performs
poorly with large samples and tends to accept the null model, despite true differences (Kline,
2011; Hu & Bentler, 1999).
Comparative Fit Index (CFI): This relative fit statistic is an indication of how much the
model differs from the null hypothesis (H0) or no relations between elements of the model.
Typically, values that range between .9 and 1.0 suggest better fit, with .95 and above often
preferred (Hu & Bentler, 1999).
Root Mean Square Error of Approximation (RMSEA): This statistic is referred to as an
overall (absolute) index of model fit that estimates the variance explained by the model. Lower
values of approximately .06 or below suggest better fitting models (Hu & Bentler, 1999).

77

Standardized Root Mean Square Residual (SRMR): The SRMR is a standardized pooled
estimation of the square root of the differences in residuals within the covariance matrix. Values
between 0 and .08 are considered appropriate, but a value of less than .05 is preferred (Hu &
Bentler, 1999; Kline, 2011). A limitation, however, of the SRMR is decreased sensitivity within
models with many parameters.
Alkaline Information Criterion (AIC). AIC is a measures of the parsimony and quality of
a specific model, and is used to compare multiple models (Burnham and Anderson, 2004; Kline,
2011). In general, models with lowest AIC values are preferred over those with larger statistics,
which may indicate a number of parameters that do not contribute meaningfully to the model.
However, these fit statistics are only meaningful in the context of model comparisons and are not
interpreted on their own (Burnham & Anderson, 2004). In comparing models, Burnham and
Anderson (2004) suggested that a model that differs by 10 or more on the AIC from another
model is a very large difference, such that the model with the larger AIC is not likely to explain
the relations within the data well.
Subpopulation Analyses
Upon testing the model with the full sample, the model was tested with both housing subgroups via the SUBPOPULATION command in Mplus. Testing the model with both groups
allowed for testing the measurement and structural model with the two nested sub-populations of
interest, the unstably housed and stably housed groups. As suggested within the literature
(Brown, 2015; Kline, 2011; Reynolds & Keith, 2013), testing the model fit separately among the
sub-populations of interest prior to multi-group analyses is a necessary step. In order for multigroup analyses to proceed, the model must demonstrate adequate fit to the sub-groups of interest
in order to make meaningful comparisons of difference between competing models. A model that

78

shows poor fit to one group may require re-specification, as further detailed within the results
section.
Although weighting ensures representativeness of estimates to the larger sample and
accounts for design structure, the focus of certain analyses on a sub-population (domain) within
the larger data warrants adjustments to reflect the specific strata from which sampling occurred,
therefore discarding cases not included in sample of the proposed study would likely produce
biased estimates of standard error due to the sampling of the sub-population from several strata
(Graubard & Korn, 1996). In order to obtain the most precise estimates of standard error, all
observations were retained in the analyses and subpopulation commands were applied using
housing status as the grouping variable. In Mplus and other similar software, this approach sets
the weights of cases not within the subpopulation of interest to zero, while still retaining all
design elements of the complex structure (Lee & Forthofer, 2006).
Multi-group Structural Equation Modeling
Once an adequate fitting model for both sub-populations was established, tests of
measurement invariance were conducted. These series of tests included restricting factor
loadings, intercepts and variances to establish that constructs were measured similarly across the
two nested groups (Brown, 2015). Testing for measurement invariance consists of a series of
steps that constrains components of the measurement model to test if underlying constructs
function similarly across groups (Brown, 2015; Keith, 2014).
Typically, the first step of invariance testing is establishing configural invariance in order
to demonstrate that a similar model holds for both groups. Configural invariance includes
specifying a model that is allowed to freely vary across groups, with the exception of latent
means. Second, metric invariance is established by holding the factor loadings constant across

79

both groups. Metric invariance, also known as weak invariance, establishes that indicators of
latent constructs function similarly for both groups (Keith, 2014; Meredith & Teresi, 2006).
Next, scalar invariance, or strong invariance is established by constraining intercepts across
groups. Establishing this form of invariance allows for mean level comparisons on latent
variables with the assurance that mean differences are a result of underlying differences and not
differences in the scales of the indicators (Keith, 2014). Although the last form of measurement
invariance, strict invariance or constraints on residuals and covariances, is not considered
necessary, this form of invariance was also tested with the full model. These forms of invariance
ensure that error and covariances in the model do not affect cross-group comparisons (Brown,
2015). Measurement invariance was initially tested separately for each latent construct using the
Mplus defaults (MODEL = CONFIGURAL METRIC SCALAR), followed by imposing
constraints on the full model manually to also establish strict invariance and build a model for
which to use for subsequent analysis.
Upon establishing measurement invariance, the multi-group SEM (MG-SEM) with
housing status (unstable vs. stable) as the grouping variable was applied to answer all research
questions. MG- SEM includes the application of a series of constraints to test the fit of
competing models to provide evidence of differential (moderation) group relations between
components of the model. The questions were primarily concerned with examining structural
differences between stably and unstably housed children and families across latent and observed
means (RQ 2 & 3), direct paths (RQ 4), indirect paths (RQ 5), and moderated-mediation (RQ 6).
Notably, only those paths with hypothesized differences in RQ 2- 6 were constrained and relaxed
in an attempt to obtain a better fitting model that was comparable to baseline. Furthermore, to
avoid multiple comparisons that could increase the potential for Type I errors, constraints were

80

applied in an omnibus fashion for each research question, as opposed to individually testing each
parameter. The three primary steps of this approach for each set of paths of interest are
described below.
1. Free Model: First, a model with all parameters allowed to vary freely was tested across both
groups. This step establishes a baseline model for comparison of models tested in subsequent
steps for this analysis. For example, to test the direct paths between parenting and children’s
social-emotional skills, a model that first allowed these paths to freely vary in both groups
was specified.
2. Constrained Model: The second step involved constraining parameters of interest for each
research question (e.g., means, direct paths, indirect paths). This means these parameters
were set as equal to one another. In the example highlighting a direct path of interest between
parenting and social-emotional skills, this path was constrained for this phase of testing. The
purpose of applying this constraint was to see if model fit changes. A comparable or
improved fit would suggest two invariant models, which signifies little difference in the
underlying structure of the model between the stably housed and unstably housed groups
across paths. Conversely, if fit degrades, this would signify differential patterns of relations
between the two groups.
3. Selective release of constraints: The third step involved selectively releasing constraints
where differences were implicated by modification indices and difference testing for
hypothesized paths in order to test if model fit changes. Using the example discussed above,
if evidence of variance was found in Step 2, then the select paths that demonstrated variance
would be released for comparison to the more restrictive model.

81

For the majority of the models (RQ 2-4), differences in the fit of competing models were
tested using a likelihood ratio test for equality (chi-square difference test = Δχ 2 ) and changes in
the CFI statistic (ΔCFI). For situations that required computing new parameters (moderatedmediation RQ 5-6), a similar test, the Wald Test (Wald X2), was used.
Chi-square difference test = Δχ 2 : The chi-square difference test is a likelihood ratio test
for equality that is conducted to examine if differences in model fit are statistically significant
between competing models in multi-group analysis (Kline, 2011). This statistic was used to
compare if the differences in chi-square between more constrained and relaxed models are
statistically significant. Significance indicates group differences across the parameter or paths of
interest. However, the most precise option for studies that employ Maximum Likelihood
estimation methods is the Satorra-Bentler Chi-Square Test (S-B Δχ 2 , Satorra, 2000). Publicly
available formulas through the developers of Mplus (https://www.statmodel.com/chidiff.shtml)
were used to compute this statistic in Excel, as Mplus version 7.31 does not readily perform this
test for structural invariance testing (Muthén & Muthén, 2015). These formulas are included in
the Appendix.
Difference in CFI (ΔCFI): Cheung and Rensvold (2002) proposed the use of additional fit
criteria, such as the CFI, between competing models as an alternative to the chi-square statistic.
CFI is less sensitive to sample size and has been frequently used in conjunction with the chisquare difference test in recent literature (e.g., Milfont & Fischer, 2015). Although the present
study relied upon S-B Δχ 2 , differences in the CFI between models were also used as a
supplementary metric by which to compare results. Using the cut-off criteria proposed by
Cheung and Rensvold (2002), a difference in a CFI of .01 or higher between models was

82

considered significant. Similar to the S-B Δχ 2 test, significant differences in the CFI indicate
evidence of variance between the paths/parameters of interest.
Wald Test: In models with limited fit statistics and/or those that required the computation
of new/additional parameters, the Wald Test (via the MODEL TEST command) was employed
to detect if parameters were statistically different from one another or from the null hypothesis
(β1 = β2, with β1 and β 2 as the same parameter across groups). Similar to other likelihood ratio
tests, a non-significant Wald X2 indicates that the parameters may not statistically differ from one
another. In the case of constraints, the Wald Test determines if additional parameters/constraints
are significantly different from one another.
Following the procedures above, targeted mean differences of interest for RQ 2 and RQ 3
in child outcomes, parent depression, and parenting practices between the unstably housed and
stably housed groups were constrained upon establishing measurement invariance. Once means
were constrained, the fit of this model was compared to a model where means were freely
allowed to vary. Difference testing (S-B ΔX2 and ΔCFI), along with modification indices, were
used to identify specific means that could be released in subsequent models. Standardized
intercept differences were explored to consider the magnitude of mean differences on latent
constructs between the two groups.
RQ 4 was aimed at specific differences of the direct paths of environmental quality
indicators to child outcomes. All six direct paths of interest (parent engagement à socialemotional functioning; parent engagement à language and literacy; parent approaches à socialemotional functioning; parent approaches à language and literacy; parent depression à socialemotional functioning; parent depression à language and literacy) were constrained to compare
model fit to the best fitting model from the previous research question. Difference testing (S-B

83

ΔX2 and ΔCFI) and modification indices were examined to consider differential predictive
relations between the two groups.
RQ 5 tested differences in the four hypothesized indirect relations of parenting
approaches and engagement on the relationship between caregiver depression and children’s
outcomes. Differences in the mediational paths between the groups were considered in two
ways. First, all direct paths that comprised the mediational paths were constrained and compared
to the best fitting model from the previous research question. Since mediation is the product of
two direct effects, differences in the indirect path would be seen on one or both of the direct
paths that comprise the indirect effect. Second, differences in mediational paths across groups
was also tested by the Wald X2. For this approach, new parameters that reflected the indirect
effect were created through the MODEL CONSTRAINT command for each group (Muthén &
Muthén, 2015). Then, the indirect effects in both groups were set as equal to one another to test
if the indirect effects differentially contributed to the model for the stable and unstably housed
groups.
Lastly, RQ 6 targeted differences in the moderated-mediation. Due to the addition of the
latent variable interaction, an alternative method to multiple group analysis that used latent class
mixture model within two known classes (or observed variable) was required to answer this final
research question. Traditional multi-group analyses in Mplus lacks the capability to handle latent
variable interactions, however mixture modeling that uses an observed class variable (e.g.,
KNOWNCLASS command) is identical to the multi-group approach when an observed variable
is used to classify the groups, but allows for latent variable interactions (Muthén & Muthén,
2015). Differences in model fit were compared using parsimony fit statistics (e.g., AIC) and the
Wald Test.

84

Post-hoc Analyses
Additionally, the analytic plan included exploratory post-hoc analyses, such as Analysis
of Variance (ANOVA), to better explain the challenges that arose during data analysis. More
specifically, the post-hoc analysis sought to gain insight into the relations between maternal race
and housing due to degraded model fit as a result of including race as a covariate for the unstably
housed group, as detailed in the results section.

85

CHAPTER 4
RESULTS
Preliminary Analyses
Preliminary descriptive statistics and analyses are presented in Tables 6- 9. Table 6
presents the unweighted and weighted descriptive statistics of the full pre-kindergarten sample.
Tables 7 and 8 include descriptive statistics by housing status and preliminary univariate
significance testing (e.g., t-test) for mean differences. Lastly, Table 9 presents the correlation
matrix for variables in the model. Correlations are presented for both the stable and unstably
housed groups in one table.

86

Table 6: Unweighted and Weighted Full Pre-Kindergarten Year Sample Descriptive Statistics
Unweighted N
Unweighted Mean
Unweighted
Weighted
Standard Deviation
N
Parent/Caregiver
Depression
Parenting
Warmth
Energy
Authoritative
Authoritarian*
Weekly Activities
Monthly Activities
Head Start Involvement
Social-Emotional Functioning
Parent-Reported Problem
Behaviors*
Parent-Reported Social
Skills
Teacher -Reported
Problem Behaviors*
Teacher-Reported Social
Skills
Language and Literacy
Letter Word Identification
Receptive Language
Expressive Language
Classroom Quality
Structural Quality
Emotional Support
Classroom Organization
Instructional Support
*Scales reverse coded

Weighted
Mean

Weighted
Standard Error

1,540

3.93

5.28

362,661

4.06

.183

1,551
1,537
1,550
1,510
1,554
1,554
1,545

4.29
3.97
3.50
3.43
11.77
5.30
6.21

.505
.740
.530
.736
1.94
2.31
3.04

364,482
361,907
363,960
355,551
365,547
365,547
362,978

4.30
3.99
3.49
3.44
11.80
5.22
6.16

.014
.029
.016
.024
.071
.083
.160

1,554

18.77

3.69

365,547

18.77

.135

1,554

12.66

2.41

365,547

12.57

.080

1,554

25.29

4.15

365,547

25.22

.170

1,552

18.40

4.23

365,333

18.18

.170

1,455
1,527
1,245

99.63
86.66
84.25

14.01
16.79
13.94

341,888
359,968
393,898

100.030
87.34
84.36

.574
1.05
.712

1,554
1,554
1,554
1,554

4.32
5.33
4.77
2.20

.745
.512
.639
.640

365,547
365,547
365,547
365,547

4.31
5.32
4.76
2.21

.082
.135
.041
.043

87

Table 7: Unweighted Group Descriptive Statistics and Preliminary Significance Testing
Stably
Housed
N= 1140*
N
Parent Depression
Parenting
Warmth
Energy
Authoritative
Authoritarian
Weekly Activities
Monthly Activities
Head Start Involvement
Social-Emotional Functioning
Parent-Reported Problem
Behaviors
Parent-Reported Social Skills
Teacher-Reported Problem
Behaviors
Teacher-Reported Social Skills
Language and Literacy
Letter Word Identification
Receptive Language
Expressive Language
Classroom Quality
Structural Quality
Emotional Support
Classroom Organization
Instructional Support

Mean(SD)

N

Unstably
Housed
N = 368*
Mean (SD)

1,130

3.45(5.03)

366

4.33(5.67)

Independent Samples TTest
-2.00(562.33)*

1,138
1,128
1,137
1,104
1,140
1,140
1,134

4.30(.501)
3.97(.746)
3.52(.515)
3.45(.747)
11.83(1.87)
5.34(2.29)
6.29(3.01)

367
366
367
360
368
368
365

4.27(.501)
3.97(.750)
3.49(.564)
3.41(.717)
11.68(2.10)
5.16(2.39)
6.02(2.93)

.729(1503)
.023(1492)
.720(1502)
.971(1462)
1.21(586.1)
1.31(1506)
1.51 (1497)

1,140

18.99(3.57)

368

18.29(4.02)

2.76(565.7)**

1,140
1,140

12.67(2.38)
25.35(4.05)

368
368

12.54(2.52)
25.13(4.48)

.921(1506)
.885(1506)

1,138

18.43(4.12)

368

18.16(4.53)

1.09(1504)

1,072
1,122
921

99.96(13.89)
87.04(16.78)
84.43(13.67)

339
361
290

98.51(14.30)
85.72(16.80)
83.95(14.08)

1.66(1409)
1.29(1481)
.511 (1206)

1,140
1,140
1,140
1,140

4.31(.773)
5.33(.517)
4.77(.640)
2.19(.640)

368
368
368
368

4.37(.686)
5.35(.508)
4.75(.634)
2.21(.644)

-1.34(1481)
-.591(1506)
.290(1506)
-.663(1506)

*= p<.05; **= p <.005

88

Table 8: Weighted Group Descriptive Statistics and Preliminary Significance Testing
Stably Housed
N
Mean(SE)

Unstably Housed
N
Mean (SE)

Parent Depression
Parenting
Warmth
Energy
Authoritative
Authoritarian
Weekly Activities
Monthly Activities
Head Start Involvement
Social-Emotional Functioning
Parent Reported Problem
Behaviors
Parent Reported Social Skills
Teacher Reported Problem
Behaviors
Teacher Reported Social Skills
Language and Literacy
Letter Word Identification

269,169

3.94(.228)

83,730

4.36(.340)

T-Test (General Linear
Model)
-.999(48)

270,225
269,703
269,703
262,984
269,703
271,009
269,140

4.30(.016)
3.98(.032)
3.51(.020)
3.46(.031)
11.85(.079)
5.26(.092)
6.25(.177)

83,792
83,661
83,792
82,102
84,073
84,073
83,372

4.28(.030)
4.01(.056)
3.45(.029)
3.44(.058)
11.65(.129)
5.08(.144)
5.90(.267)

.528(48)
-.381(48)
2.01(48)*
.191(48)
1.37(48)
1.20(48)
1.27(48)

271,009

18.88(.132)

84,073

18.34(.269)

2.05(48)*

271,009
271,009

12.54(.088)
25.28(.189)

84,073
84,073

12.54(.155)
24.98(.300)

.007(48)
.895 (48)

270,795

18.21(.215)

84,073

18.07(.256)

.400(48)

253,903

100.42(.549)

77,937

98.76(.990)

1.86(48)

Receptive Language
Expressive Language
Classroom Quality
Structural Quality
Emotional Support
Classroom Organization
Instructional Support

267,369
227,001

87.64(1.14)
84.41(.85)

82,313
68,320

86.93(1.31)
84.54(.906)

.604 (48)
-.122 (48)

271,009
271,009
271,009
271,009

4.29(.091)
5.31(.045)
4.76(.041)
2.20(.044)

84,073
84,073
84,073
84,073

4.32(.073)
5.36(.046)
4.74(.056)
2.22(.055)

-.497(48)
-.434(48)
.401(48)
-.418 (48)

*= p<.05

89

Table 9: Correlation Matrix by Stable (upper) and Unstable (lower) Housing Groups
4
-.01
.21*
.17*
.01
.05
.11*
-.08

5
-.09*
.37*
.07
.01
.08
-.10*
-.02

6
.02
.11*
.21*
.04
-.01
.45*
.19*

7
.04
.04
.15*
.05
-.08
.44*
.34*

8
.01
.07*
.16*
.01
.02
.25*
.35*
-

9
-.23*
.15*
.29*
-.06
.15*
.14*
.10*
.13*

10
-.07
.07
.03
-.01
-.02
.03
.05
.07

11
-.12*
.20*
.19*
-.06
.05
.20*
.17*
.15*

12
-.10*
.08*
.01
.02
.02
.04
.02
.03

13
-.02
.05
.09*
.02
.01
.02
.02
-.01

14
-.03
.07*
.21*
-.03
.09*
.16*
.06
.05

15
.03
.06
.09*
-.06
.10*
.01
.01
.01

16
-.03
.03
.00
-.01
.00
-.03
-.02
-.05

17
-.01
.01
-.02
.00
.03
-.04
-.03
-.01

18
.01
.02
.00
.03
.01
-.01
-.02
.03

19
-.06
.01
-.03
.00
.01
-.07
-.05
-.01

9 PBEH
-.20* .23*
.28*
-.06
10 TBEH
-.06
.04
-.07
-.10
11 PSS
-.08
.22*
.28* .10*
12 TSS
-.04
.07
.01
-.05
13 WJLW
-.01
.04
.04
-.02
14 PPVT
.07
.17*
.22*
-.06
15 EOWVT
-.06
.14
.04
-.09
16 ECERSS
-.03
.04
.04
.03
17 CLASSE
-.07
-.06
-.03
-.02
18 CLASSI
.01
-.06
-.04
-.01
19 CLASSO
-.07
-.07
-.05
.04
*Bolded values significant at p value < .001

.20*
-.01
.04
.05
.02
.14*
.11
-.02
-.05
-.01
-.02

.13
-.09
.22*
-.06
.01
.20*
.05
.12
.03
.05
.08

.03
-.04
.17*
.03
.08
-.01
-.04
.02
-.04
-.03
-.04

.03
.06
.13*
.12
.14
-.10
-.03
.01
-.06
-.08
-.15*

.31*
.37*
.22*
.19*
.24*
.23*
.01
.06
-.04
.10

.17*
.21*
.63*
.19*
.05
.13
.03
.09
.01
.07

.27*
.15*
.14*
.26*
.16*
.19*
.07
.01
-.01
.02

.12*
.61*
.11*
.25*
.13
.19*
.13*
.13
-.05
.08

.14*
.14*
.05
.16*
.36*
.39*
.02
-.03
.11
.08

.22*
.15*
.04
.17*
.43*
.72*
.10
.05
.11
.07

.12*
.17*
.02
.16*
.44*
.71*
.11
.07
.14
.14

-.01
.01
.00
.11*
.03
.09*
.10*
.59*
.40*
.47*

-.05
.09*
-.03
.12*
.04
.05
.06
.61*
.52*
.72*

-.01
.04
-.02
.04
.12*
.11*
.12*
.34*
.54*
.50*

-.04
.12*
-.01
.10*
.05
-.01
.05
.47*
.71*
.54*
-

1 DEP
2 WARM
3 ENERGY
4 AUTHV
5 AUTHRN
6 WACT
7 MACT
8 INV

1
-.09
-.10
-.01
-.11
.01
.09
-.01

2
-.08*
.37*
.11
.33*
.21*
.14*
.01

3
-.12*
.32*
.20*
.16*
.29*
.13*
-.05

90

Preliminary Differences by Housing Status
As indicated in Tables 8 and 9, few significant mean differences between the stably and
unstably housed groups emerged in preliminary testing. However, this testing revealed some
differences between the unweighted and weighted results. The unweighted results (Table 8)
indicated significant differences in mean levels of parental depression between the stably housed
(M= 3.45 SD = 5.03) and unstably housed (M= 4.33 SD= 5.67) groups; t(562.33)= -2.00, p<.05.
Parents who were unstably housed during Head Start reported higher levels of depression
symptoms. In addition, significant differences in parent reported problem behaviors emerged
between the unstably housed (M= 18.29 SD = 4.02) and stably housed (M= 18.99 SD= 3.57)
children; t(565.7),= 2.76 p <.01, such that parents of children who were stably housed during
Head Start endorsed higher mean levels of problem behaviors.
Consistent with the unweighted results, the weighted (Table 9) results suggested mean
level differences in parent reported problem behaviors between the two groups; t(48)= 2.05 p <
.05. However the weighted results indicated differences in mean scores of the authoritative
parenting sub-scale between the unstably housed (M= 3.45 SE = .03) and stably housed (M=
3.51, SE = .02 ) parents; t(48)= 2.01 p<.05. Although the difference between the two groups was
not significant when the data were weighted, parental depression scores continued to suggest that
unstably housed parents were more likely to endorse slightly higher levels of depressive
symptomology. These preliminary results should be interpreted with caution, as multiple
comparisons increase the likelihood of Type I error.
Differences in the covariates, the economic risk index, maternal race, child’s gender, and
child’s age at entry, were also explored. Unweighted chi-square testing indicated significant
differences in economic risk between the two groups, χ2 (3, N = 1440) = 16.98, p <.001.

91

Weighted percentages signified that unstably housed children were more likely than stably
housed children to face one (35% versus 29%) or two (48% versus 37%) indicators of economic
risks at the time of pre-kindergarten data collection. Similarly, chi-square testing suggested
differences in the distribution of families across the two groups by maternal race, χ2 (3, N =
1508) = 13.10 p <.005. An examination of weighted percentages indicated that mothers who
experienced housing instability were more likely than stably housed counterparts to be White
(33% versus 24%) or Latino (40% versus 36%). There were no significant differences between
groups with respect to child’s gender or age at entry into Head Start.
Correlations. Correlations (Table 9) were examined for exploratory group differences in
associations between the study variables. Although similar associations were found among both
groups, several noteworthy differences emerged. One difference that emerged was the greater
number of associations between parental depression symptoms and parenting, as well as
depression and child behavior for the stably housed group. More specifically, depression
symptomology was significantly correlated with warmth (r= -.08, p<.001), energy (r= .-.12, p
<.001), authoritarian parenting (r= -.09, p<.001), parent-reported behavior (r= -.23, p<.001),
parent-reported social skills (r= -.12, p<.001), and teacher-reported social skills (r= -.10, p<.001)
scores for the stably housed group. For the unstably housed group, depression was significantly
correlated with only parent-reported behavior (r= -.20, p<.001). In addition, parental
involvement in Head Start was significantly correlated with the classroom organization sub-scale
of the Classroom Assessment Scoring System (r=-.15, p<.001) for the unstably housed group,
whereas this difference was close to zero and non-significant among the stably housed group (r=
-.01, p<.001). Furthermore, teacher-reported social skills were significantly related to both
receptive (r= .16, p <.001) and expressive language (r= .19, p<.001) in the unstably housed

92

sample, whereas these differences were non-significant among the stably housed group (r= .04
and .02, p<.001).
Measurement Model
Confirmatory factor analysis (CFA) was first conducted on the full sample to assess the
proposed five-factor structure of the model. The measurement portion of the model specified five
latent constructs: parent approaches, parent engagement, classroom quality, social-emotional
functioning, and language and literacy. Parent approaches was specified by four sub-scale scores
from the parenting questionnaire: warmth, authoritative, authoritative, and energy. Parent
engagement was specified with three indicator scores: monthly out-of-home activities, weekly
parent-child activities, and involvement in Head Start. Classroom quality was specified with the
three sub-scales process quality (emotional support, instructional support, and classroom
organization) and the structural quality score. Social-emotional functioning was specified by the
two broader behavioral composites for the parent-reported items (problem behaviors and social
skills) and the two behavioral composite scores for the teacher-rated items (problem behaviors
and social skills). Lastly, the language and literacy construct was indicated by receptive
language, expressive language, and letter-word identification.
The fit of the initial five-factor structure fit the data reasonably well, RMSEA = 0.048,
90% CI = 0.044 -.052, CFI = 0.872, SRMR = .060, χ2 = 572.82, p = 0.0, df = 125. As previously
noted, it was expected that the chi-square fit statistic would be significant due to the large
weighted sample (Kline, 2011). Although this value is reported throughout, additional fit indices
(e.g., CFI, RMSEA, SRMR) were primarily used to make statistical decisions. Changes to the
initial five-factor structure were identified using modification indices, communalities (R2) and fit
statistics. Salient theoretical considerations were used in conjunction with statistical evidence.

93

However, modifications to theoretical models are considered exploratory and results should be
interpreted with caution (Brown, 2015). In the present study, modifications were allowed in
order to include a preliminary exploration of a more complex model with an understudied group
of unstably housed children.
Two main modifications were made to the measurement model. First, the authoritative
approaches score was removed from the model. Decisions to remove factors occur in light of
theoretical and statistical significance. In general, indicators with standardized loadings below
.30, low communalities, and/or non-significant contributions are typically removed (Brown,
2015). Authoritative parenting demonstrated a low loading (.20) and very low communality
(R2= .05), indicating that this sub-scale contributed little to the construct and model, despite
demonstrating a significant loading.
Consistent with the broader literature, modification indices were used sparingly and only
when they were conceptually meaningful to avoid contributing to atheoretical and/or over-fitted
models (Brown, 2015; Joreskog, 1993). Modification indices suggested three conceptually
justifiable changes. Although a primary assumption of CFA is that the observed variables are
measured without error, three residuals between indicators from the same measure were included
in the model. These modifications were deemed as reasonable due to the scores originating from
the same broader measure, thus likely increasing the chance that specific indicators have more
similarity with other indicators of the construct. The three additional terms allowed to co-vary
included teacher-rated social skills with teacher-rated problem behaviors, classroom instructional
support with classroom organization, and authoritarian parenting with energy.
Two additional noteworthy changes were explored, but ultimately not included in the
model. First, a cross-loading of parental energy with both the parent approaches and parent

94

engagement latent variables. This potential cross-loading was identified via modification indices
and appeared theoretically plausible because of select items (e.g., adhering to plans) that may be
pertinent to both parental approaches and engagement. Furthermore, correlations (Table 9)
indicated that associations between energy and indicators of parent engagement (e.g., weekly
activities) were moderate. When cross-loaded, energy loaded significantly on both constructs and
model fit improved slightly, however, energy was retained as an indicator of parent approaches
due to a low cross-loading on parent engagement (<.10).
Second, models in which social-emotional indicators were considered as both separate
constructs by rater and as one unidimensional construct were explored. Although there was an
improvement in the measurement model in which social-emotional constructs were separated by
rater (RMSEA = 0.034, 90% CI = 0.028 -.037, CFI = 0.940, SRMR = .040, χ2 = 267.68, p = 0.0,
df = 102), the unidimensional social-emotional model was ultimately selected (RMSEA = 0.035,
90% CI = 0.031 -.040, CFI = 0.938, SRMR = .047, χ2 = 312.48, p = 0.0, df = 106) in favor of
parsimony with regards to the structural paths of the model, given the complexity of the full
model. Although teacher-rated social skills demonstrated a marginally low loading (.26) with
social-emotional functioning as a unidimensional construct, this indicator was retained in the
model due to the marginal contribution to the model and in order to maintain a cohesive
construct balanced by both parent and teacher-report.
The final five-factor measurement model fit the data well, RMSEA = 0.035, 90% CI =
0.031 -.040, CFI = 0.938, SRMR = .047, χ2 = 312.48, p = 0.0, df = 106. All loadings were
significant and fit statistics indicated improvement in model fit from the baseline measurement
model. Information on the specific loadings of each indicator is summarized in Table 10.
Although tests of measurement invariance are detailed in subsequent sections, preliminary

95

subpopulation analyses indicated that the measurement model demonstrated adequate fit with
both the stably housed (RMSEA = 0.039, 90% CI = 0.034-.044, CFI = 0.930, SRMR = .050, χ2 =
289.62, p = 0.0, df = 106) and unstably housed (RMSEA = 0.049, 90% CI = 0.039-.050, CFI =
0.912, SRMR = .063, χ2 = 200.56, p = 0.0, df = 106) sub-groups.
Table 10: Five-Factor Confirmatory Factor Analysis Loadings and Communalities for Full
Sample
Parent Approaches
Energy
Authoritative
Authoritarian
Warmth
Parent Engagement
Monthly Activities
Weekly Activities
Head Start Involvement
Classroom Quality
Classroom Emotional Support
Classroom Organization
Classroom Instructional Support
Structural Quality
Language and Literacy
Expressive Language
Receptive Language
Pre-literacy Skills
Social-Emotional Functioning
Parent Reported Problem Behaviors
Parent Reported Social Skills
Teacher Reported Problem Behaviors
Teacher Reported Social Skills

Standardized

Unstandardized

R2
Contribution

0.68
-0.53
0.56

1.00
-0.76
0.56

0.46
-0.28
0.32

0.72
0.67
0.47

1.00
0.74
0.84

0.52
0.41
0.22

0.94
0.77
0.56
0.63

1.00
0.96
0.72
0.98

0.88
0.59
0.31
0.40

0.91
0.86
0.50

1.00
1.07
0.51

0.83
0.73
0.25

0.67
0.50
0.31
0.26

1.00
0.49
0.52
0.42

0.45
0.24
0.10
0.08

Structural Model
Base Model with Full Sample
The first step in testing the structural model included evaluating model fit of the base
mediational model without covariates. Model fit was used to determine the appropriateness of
proceeding with multi-group analyses, as each model was tested with the full sample and the two
housing sub-groups separately to make decisions about subsequent steps. Due to a lack of

96

availability of traditional model fit indices, latent variable moderation was included only in the
final model and compared using fit indices of parsimony (AIC). Model fit indices throughout the
development of the structural model phase are summarized in Table 11.
The base mediational model fit the full sample well, RMSEA = 0.035, 90% CI = 0.030 .039, CFI = 0.937, SRMR = .046, AIC = 114295.61, χ2 = 342.77, p = 0.0, df = 120. Similarly, the
base model fit the stably housed group well, RMSEA = 0.038, 90% CI = 0.033 -.043, CFI =
0.929, SRMR = .050, AIC = 83635.29, χ2 = 314.13, p = 0.0, df = 120. The model also
demonstrated adequate fit with the unstably housed sample, RMSEA = 0.046, 90% CI = 0.036.052, CFI = 0.915, SRMR = .061, AIC = 2725.39, χ2 = 214.27, p = 0.0, df = 120.
Model with Covariates
Next, a model that included hypothesized covariates was specified. Categorical covariates
were dummy coded for inclusion in the model. Based on theoretical considerations, paths from
the economic risk index categories to depression, parent engagement, parent approaches, and
language and literacy were included. Paths from maternal race to depression, parent approaches,
and parent engagement were included. Paths from gender to behavior, parent approaches, and
parent engagement were included. Lastly, paths from child’s age at Head Start entry to parent
approaches, parent engagement, behavior, and language and literacy were included. Ultimately,
age was trimmed from the model due to not significantly contributing to any of the specified
variables. This change appeared reasonable given that the sample was heterogeneous with
regards to age due to the focus of the study on the pre-kindergarten year and the sample was
relatively balanced by cohort.
All other pre-specified paths were initially retained. With the addition of the covariates,
the model fit dropped considerably. The fit of the model for the full sample was marginal, but

97

adequate, RMSEA = 0.039, 90% CI = 0.036-.042, CFI = 0.886, SRMR = .046, AIC =
122608.54, χ2 =719.56, p = 0.0, df = 214. Fit statistics continued to indicate similar fit for the
stably housed group, RMSEA = 0.039, 90% CI = 0.036-.043, CFI = 0.889, SRMR = .046, AIC =
89808.34, χ2 = 592.97, p = 0.0, df = 214. However, model fit was poorer, with a noted drop in
several fit indices (CFI, RMSEA and SRMR) for the unstably housed group, RMSEA = 0.052,
90% CI = 0.045-.059, CFI = 0.846, SRMR = .064, AIC = 29003.16, χ2= 427.01, p = 0.0, df =
214.
Due to the considerable degradation of model fit with the addition of the covariates,
especially for the unstably housed group, model re-specification was targeted at the covariates.
Covariates were tested separately and results indicated that maternal race was responsible for the
greatest degradation in model fit, particularly for the unstably housed group. Although the
inclusion of this variable was both theoretically and statistically significant with regards to
depression and parenting, the model was re-specified without maternal race in order to proceed
with multi-group analyses. However, proceeding without this covariate warranted caution in
interpretation. Possible causes for the degradation in model fit as a result of this variable in
relation to the measurement components of the model are explored further in the post-hoc
analysis section to support interpretation of the final results.
The re-specified model fit the data well, RMSEA = 0.034, 90% CI = 0.031-.038, CFI =
0.93, SRMR = .043, AIC = 119930.38, χ2 = 491.69, p = 0.0, df = 175. Fit indices also indicated
reasonably well-fitting model for the sub-groups. The model fit the stably housed group
adequately, RMSEA = 0.035, 90% CI = 0.030-.039, CFI = 0.922, SRMR = .045, AIC =
87793.62, χ2 = 415.63, p = 0.0, df = 175. Model fit was also adequate for the unstably housed
group, RMSEA = 0.046, 90% CI = 0.038-.050, CFI = 0.886, SRMR = .058, AIC = 28558.72, χ2

98

= 312.99, p = 0.0, df = 175. Despite some borderline fit indices for the unstably housed group
(e.g., CFI below .90), this model was used for subsequent analysis to retain the majority of
constructs of interest, as a primary aim of the study was to examine the role of housing instability
beyond economic risks in the proposed relations. Therefore, analysis was centered around
preserving as many paths and covariates of theoretical interest as feasible.
Table 11: Fit Statistics Across Structural Models

Base Mediational
Model

Base Model
Covariates
Maternal Race,
Risk, and Gender

Final Base
Model
Covariates Risk
and Gender

χ2

df

P

CFI

RMSEA

Full
Sample
Stable

342.77

120

0.00

.937

.035

314.13

120

0.00

.929

.036

Unstable

214.27

120

0.00

.915

.046

Full
Sample
Stable

714.56

214

0.00

.887

.039

592.97

214

0.00

.889

.039

Unstable

427.01

214

0.00

.846

.052

Full
Sample
Stable

491.70

175

0.00

.930

.034

415.63

175

0.00

.922

.035

Unstable

312.98

175

0.00

.886

.046

RMSEA
90%
C.I.
.030.039
.033.043
.036.052
.036.042
.036.043
.045.059
.031.038
.030.039
.038.050

SRMR

AIC

.046

114295.61

.050

83635.29

.061

2725.39

.046

122608.54

.046

89808.34

.064

29003.16

.043

119930.38

.045

87793.62

.058

28558.72

Research Question 1: Structural Model
The final structural model for the full sample included all theorized relations, with
economic risk and gender as covariates. Standard error estimates were derived using Taylor
Series Linearization. Figure 4 summarizes the standardized estimates for the full model. See
Table 12 for a summary of standardized and unstandardized parameter estimates, as well as
standard errors for each parameter in the full moderation model.

99

Two covariates, the family economic risk index score and child’s gender, were included
in the final model. Using the highest risk group as the reference, economic risk index scores were
significantly related to parental depression and children’s language and literacy scores.
Specifically, parents of children with the lowest levels of economic risk were less likely to
endorse depressive symptoms (βrisk0 = -.11, p<.001; βrisk1 = -.07, p<.05) and their children scored
higher on standardized academic measures (βrisk0 = .27, p<.001; βrisk1 = .25, p<.001; βrisk2 = .14,
p<.005) when compared to those facing the greatest risk. Furthermore, child gender was
significantly related to behavior ratings, such that parents and teachers rated males lower on the
latent construct of social-emotional functioning (β = -.24, p <.001).
The majority of the direct effects specified in the model were significant. Parental
depression significantly predicted parental approaches (β = -.17, p <.001), which suggested that
higher rates of depression negatively influenced the construct of parental approaches. Contrary to
expectations, however, parental depression was positively related to parent engagement (β = .08,
p <.05), such that higher parental depression scores predicted slightly higher levels of
engagement. However, while significant, this relation had a relatively lower significance level (p
<.05) compared to parent depression and parental approaches. Depression was also directly
related to children’s skills in the areas of social-emotional functioning as expected, such that
parental endorsement of depressive symptoms was negatively related to the latent construct of
social-emotional functioning (β = -.25, p <.001). Unexpectedly, endorsement of depressive
symptoms was predictive of higher levels of the latent construct of language and literacy (β =
.08, p < .005). As expected, parental approaches and parent engagement were positively related
to social-emotional functioning (βapproaches = .35 , p <.001 ; βengagement = .27, p <.001), as well as

100

language and literacy (βapproaches = .25, p <.001 ; βengagement = .11 , p <.05). Other theorized direct
associations between classroom quality and child functioning were non-significant.
Overall, evidence of moderated-mediation was not supported with the full prekindergarten sample. Results suggested that the moderated-mediation model was a slightly worse
fitting model (AIC = 119931.64) than the base mediational model (AIC = 114295.61). Partial
mediation, specifically the influence of parental depression on children’s social emotional
functioning through parent approaches (βindirect = -.06, p < .001) and language and literacy
through parent approaches (βindirect = -.04, p < .001) were supported by the results. These
relations indicated small, but significant relations such that that higher depressive symptoms
were related to lower quality parenting practices, which in turn negatively predicted children’s
social-emotional and academic skills. Yet, parent approaches partially mediated the association
between depression and children’s outcomes, as the direct relationships between depression and
child functioning remained significant (β = -.25, p <.001; β = .08, p < .005). As noted above,
classroom quality did not further moderate the indirect paths, suggesting that the level of
classroom quality did not impact the existing indirect influence of parental approaches in the
relation between parental depression and child functioning. Parent engagement positively
partially mediated the relation between depression and children’s social-emotional functioning
(βindirect = .02, p < .05). This relation was unexpected, however, as the results suggested that
higher levels of depressive symptoms positively predicted engagement, which in turn was related
to positive social-emotional functioning. Similar to parental approaches, the standardized indirect
effect was small in magnitude. The relation between parental depression and latent construct of
language and literacy via parental engagement was non-significant.

101

Table 12: Parameter Estimates and Standard Error for Moderated-Mediation Model for Full Sample
Unstandardized

Standardized

Standard Error
(SE)

1.00
0.55**
0.77**

0.69**
0.55**
0.55**

0.05
0.04
0.04

1.00
0.75**
0.85**

0.71**
0.64**
0.47**

0.04
0.04
0.04

1.00**
0.96**
0.71**
0.98**

0.94**
0.76**
0.55**
0.63**

0.04
0.05
0.06
0.04

1.00
1.09**
0.51**

0.91**
0.86**
0.50**

0.02
0.02
0.02

1.00
0.52**
0.60**
0.52**

0.64**
0.49**
0.33**
0.29**

0.05
0.06
0.05
0.05

-0.02**
0.02*
1.65**
6.72**
0.40**
0.92*
-0.11**
0.21*
0.12
1.05

-0.17**
0.08*
0.35**
0.25**
0.28**
0.11**
-0.25**
0.08*
0.02
0.04

0.03
0.03
0.06
0.03
0.05
0.05
0.05
0.03
0.05
0.04

-0.11**

-0.04**

0.01

Factor Loadings
Parent Approaches
Energy
Warmth
Authoritarian
Parent Engagement
Monthly Activities
Weekly Activities
Head Start Involvement
Classroom Quality
Emotional Support
Organization
Instructional Support
Structural Quality
Language & Literacy
Expressive Language
Receptive Language
Pre-literacy Skills
Social-Emotional Functioning
Parent-Reported Problem Behaviors
Parent-Reported Social Skills
Teacher-Reported Problem Behaviors
Teacher-Reported Social Skills
Direct Effects
Depression à Approaches
Depression à Engagement
Approaches à Social-Emotional
Approaches à Language & Literacy
Engagement à Social-Emotional
Engagement à Language & Literacy
Depression à Social-Emotional
Depression à Language & Literacy
Classroom Quality à Social-Emotional
Classroom Quality à Language & Literacy
Indirect Effects (Mediation)
Depression à Approaches à Language & Literacy

102

Table 12 (cont’d)
Depression à Approaches à Social Emotional
Depression à Engagement à Social-Emotional
Depression à Engagementà Language & Literacy
Interactions (Moderated-Mediation)
Approaches X Classroom Quality à Language & Literacy
Approaches X Classroom Quality à Social-Emotional
Engagement X Classroom Quality à Language & Literacy
Engagement X Classroom Quality à Social-Emotional

** p value <.001 * p value <.05
Figure 4: Moderated-Mediation Model Full Sample

103

-0.03**
0.02*
0.01

-0.06**
0.01*
0.02

0.02
0.01
0.01

-3.08
0.56
0.56
-0.01

-0.06
0.06
0.03
0.00

0.05
0.06
0.04
0.04

Structural Model with Subpopulations
The same full structural model was tested with both sub-groups. The stable and unstable
groups (Figure 5) demonstrated a similar pattern of results for significant paths as the larger prekindergarten sample and evidence of moderated-mediation did not emerge among either group.
However, some differences emerged between the two groups when models were tested
separately. With regards to covariates, parents of children who faced lower risk were more likely
to display greater levels of engagement on the latent construct of engagement (βrisk0 = 1.39,
p<.05; βrisk1 = 1.42, p<.005; βrisk2 = 1.20, p<.05), whereas this covariate was not significant for
the stably housed group. On the contrary, risk was not predictive of parental approaches among
the unstably housed group, whereas among the stably housed parents facing less risk (one
economic risk factor) reported higher scores on the latent construct of parental approaches
compared to those facing the greatest risk (βrisk1 = .14, p<.05).
Within the hypothesized model for the unstably housed group, the indirect paths between
depression and children’s outcomes via parent engagement (βindirectsocial-emotional = .01, p < .33;
βindirectlanglit = -.01, p = .42) were non-significant for the unstably housed group, whereas the
social-emotional path remained only marginally significant for the stably housed group
(βindirectsocial-emotional = .02, p = .053). Furthermore, classroom quality was directly related to
children’s social-emotional functioning for children who experienced housing instability during
Head Start (β = .27, p < .05) such that higher classroom quality scores were predictive of
increases in social-emotional functioning. In contrast, this relation did not emerge for the stably
housed group or among the full sample. All other paths and factor loadings were comparable
between the two groups and to the full pre-kindergarten sample. Multi-group analysis was

104

conducted in subsequent steps to indicate if differential relations across groups in model
parameters were significant.

105

Figure 5: Standardized Estimates for Sub-Groups for Full Model

106

Multiple Group Analyses
Multiple group analyses were utilized to test for significant differential relations in the
hypothesized model between stably housed and unstably housed children. The initial step of
multiple group analyses included testing for measurement invariance. This step is necessary prior
to testing for structural differences between groups, including latent means and paths. Upon
establishing measurement invariance, structural paths were constrained and relaxed to evaluate
changes in model fit. The following section details measurement invariance testing, followed by
constraining and relaxing paths of theoretical interest as aligned with research questions.
Measurement Invariance
Each of the latent constructs were initially tested for configural, metric, and scalar
invariance separately to determine possible sources of variance across the general structure of the
constructs, factor loadings, or intercepts between the unstably and stably housed groups.
Individual tests of invariance are summarized in Table 13. Each of these tests included testing
subsequently more restrictive models, ultimately holding factor loadings and intercepts
constrained (scalar invariance). Constructs that are invariant in their loadings and interprets are
understood to demonstrate strong invariance (Meredith, 1993; Kline, 2011). Results supported
evidence of strong invariance for individual tests, such that difference testing using the S-B Δχ2
and ΔCFI were non-significant across more restrictive models for each of the latent constructs,
demonstrating that each of the five latent constructs were invariant across the stably housed and
unstably housed groups. This suggested that the underlying constructs had similar factor
structures and underlying meanings for both groups.
Measurement invariance testing was also conducted with the full model in order to
develop a baseline model from which to compare model fit for subsequent analyses. All MG-

107

SEM analyses are summarized in Table 14. In order to establish support for strict invariance, the
optimal form of measurement invariance that allows for comparisons of the structural
components of the model between groups, establishing measurement invariance with the full
model consisted of testing five models. Following recommendations from the literature (e.g.,
Brown, 2015; Keith, 2014), these included the following: 1. an equal form model with all
parameters (with the exception of means) allowed to vary freely between the groups (configural
invariance) 2. a model with factor loadings held equal (metric invariance) 3. a model with factor
loadings and intercepts held equal (scalar invariance) 4. a model with residual variances held
equal and 5. a model with residual covariances held equal.
Results supported strict invariance. The initial model (configural invariance model)
demonstrated adequate fit to the data (χ2 = 754.92, p = 0.00 ,df = 350, RMSEA = .039, 90% CI =
0.05-.043, CFI = 0.911, SRMR = .049, AIC = 116367.48). When factor loadings were
constrained to be equal across groups (Model 1), the difference in model fit between Model 0
and Model 1 was non-significant, S-B ΔX2 (12) = 1.02, p= .99; ΔCFI = .004. Similarly,
differences between Model 1 and Model 2, in which intercepts were also constrained to establish
scalar invariance was also non-significant, S-B ΔX2 (12) = 7.44, p = .83; ΔCFI = .002. Criteria for
strict invariance, residual variances (Model 3) and covariances (Model 4), were set to be equal
across groups. The difference in model fit between the scalar model (Model 2) and the model
with restricted residual variances (Model 3) was non-significant, S-B ΔX2 (17) = 16.94, p = .46;
ΔCFI = .000. Similarly, when residual covariances were also constrained (Model 4), the

difference in fit between Model 3 and 4 was also non-significant, S-B ΔX2 (4) = 4.52, p = .34;
ΔCFI = .000. Together, results established that the underlying measurement components of the

108

model are invariant between the stably housed and unstably housed groups, allowing for
meaningful comparisons of the structural components of the model between the two groups.
Research Questions 2 & 3: Mean Differences
Due to the model meeting strict invariance, latent mean differences were examined in the
multi-group framework. Using the residual covariance model (Model 5), the means of interest
for research questions 2-3 were constrained in an omnibus fashion in Model 6. Means of
relevance included constructs relevant to parenting (RQ 2) and child outcomes (RQ 3), including
parental approaches, parental engagement social-emotional functioning, language and literacy, as
well as the observed mean of parental depression. The Satorra-Bentler chi square test indicated a
significant degradation in model fit between the residual covariance model (Model 4) and the
constrained mean model (Model 5), S-B ΔX2 (6) = 13.52, p = .04. However, the difference in CFI
(ΔCFI = .001) indicated non-significant differences between the two models.
Modification indices were used to identify potentially significant improvements in model
fit. These indices suggested the release of constraints on the parent engagement intercept or
mean of the latent construct of parent engagement. Model 6 included relaxed constraints on
parent engagement and was compared again to Model 4. Relaxing constraints on the latent
construct of parent engagement (Model 6) resulted in comparable and non-significant changes in
model fit, ΔX2 (5) = 4.35, p = .50; ΔCFI = .002, suggesting Model 5 (latent mean of parent
engagement free) was the better fitting model. Using the stably housed group as the reference
group, the standardized mean difference was -.91 (z = -2.98), suggesting that the unstably housed
parents were significantly less likely than their stably housed counterparts to participate in forms
of parent engagement as captured by the latent construct.

109

Research Question 4: Direct Paths
For RQ 4, all relevant paths from parenting constructs to child outcomes, as well as
classroom quality to child outcomes were constrained to evaluate for differential influence of
housing instability across direct paths. Although sub-population analyses that examined the
model separately in each group suggested potential differences in direct paths from parent
engagement to child outcomes and from classroom quality to behavior, the model with these
direct paths constrained (Model 7) demonstrated non-significant differences with the model in
which the intercept of parent engagement was freed (Model 6), ΔX2 (8) = 11.04, p = .20; ΔCFI =
.001, suggesting no significant differences in the predictive relations between parenting and child
outcomes or classroom quality and child outcomes between the two groups.
Research Question 5: Indirect Paths for Mediation
First, to answer RQ 5, all direct paths that comprised the indirect mediational paths
relevant to the mediation were constrained. This involved constraining an additional two paths
from the previous model (parent depression à parent engagement and parent depression à
parent approaches). The model with all mediational paths constrained (Model 8) was compared
to the previous model (Model 7). Model fit was comparable and changes in fit were nonsignificant, ΔX2 (2) = 1.00, p = .61; ΔCFI = .000, suggesting that the paths that comprised the
mediational effect were invariant across two groups.
Additionally, differences in the indirect paths between the two groups were also tested
using the Wald Test statistic, such that equality constraints posed upon the new parameters that
contributed to the hypothesized mediation were set as equal to one another (βindirectstable =
βindirectunstable). The significance in difference between the constraints were then tested using the
Wald χ2, which confirmed the previous results that there were non-significant differences in the

110

mediational paths between the stably housed and unstably housed groups, Wald χ2(4)= 3.24, p =
.52.
Research Question 6: Moderated-Mediation
Despite a lack of significance of the moderation in the full sample or in the subpopulation analyses, differences in moderated-mediation were tested as part of multi-group
analyses to explore for potential group differences in the magnitude of the effect. As detailed in
the methods, moderated-mediation within multi-group analyses is tested using known class
mixture modeling. Difference testing between the two groups was conducted using the limited
available fit statistics (e.g., AIC) under the numerical integration algorithm and the Wald χ2 was
further used to compare the equality of the constraints imposed upon on the parameters of
interest.
Specifically, a model with constraints on the paths relevant to the moderation (Model 9)
was compared to a model in which the moderation was allowed to vary freely (Model 10). Model
9 was suggestive of a slightly lower AIC (118015.20) in comparison to Model 10 (118021.96),
or a ΔAIC of 6.76, indicating that the moderated-mediation model with constraints was a better
fitting model. In other words, the moderation did not have appear to exert a differential effect
across the groups. Second, the Wald Test statistic was used to compare the equality of the
constraints imposed upon the moderation paths. The results re-iterated the previous findings,
such that the differences across moderation paths were non-significant across the two groups,
Wald X2(2)= .99, p = .61.

111

Table 13: Construct-Level and Measurement Model Invariance Testing

Parenting Approaches
Configural
Metric
Scalar
Parent Engagement
Configural
Metric
Scalar
Classroom Quality
Configural
Metric
Scalar
Language & Literacy
Configural
Metric
Scalar
Social-Emotional Functioning
Configural
Metric
Scalar
Measurement Model
Configural
Metric
Scalar

X2

CFI

ΔCFI

SB ΔX2

Δdf

P
value

Models Compared

0.00
1.79
1.67

1.00
1.00
1.00

-.000
.000

-1.76
0.19

-2
2

-.42
.91

-Configural vs. Metric
Metric vs. Scalar

0.00
1.10
2.41

1.00
1.00
1.00

-.000
.000

-1.10
2.41

-2
2

-.57
.52

-Configural vs. Metric
Metric vs. Scalar

14.25
21.67
25.99

.981
.972
.970

-.009
.002

-3.81
1.59

-3
3

-.28
.67

-Configural vs. Metric
Metric vs. Scalar

0.00
1.08
4.51

1.00
1.00
.990

-.000
.001

-1.08
3.58

-2
2

-.58
.17

-Configural vs. Metric
Metric vs. Scalar

0.30
2.03
3.96

1.00
1.00
1.00

.000
.000
.000

-2.10
1.81

-3
3

-.55
.61

-Configural vs. Metric
Metric vs. Scalar

495.10
506.38
511.95

.926
.927
.928

-.001
.001

-11.41
7.65

-12
12

-.49
.81

-Configural vs. Metric
Metric vs. Scalar

112

Table 14: Multi-Group Analysis Model Summary Table

Measurement Invariance
Model 0:
Configural
(all free)
Model 1:
Metric
(factor loadings)
Model 2:
Scalar
(intercepts)
Model 3:
Residual Variances
Model 4 :
Residual Covariances
Means (RQ 2 & 3)
Model 5:
Means Constrained
Model 6:
Parent Engagement Free
Direct Paths (RQ 4)
Model 7:
Direct Paths Constrained
Indirect Paths (RQ 5)
Model 8:
All Paths of Mediation
Constrained
Moderated-Mediation (RQ 6)
Model 9:
Moderated-Mediation
Constrained
Model 10:
Moderated-Mediation Free

X2

CFI

ΔCFI

S-BΔX2

Δdf

P value

Wald χ2

P value

Models
Compared

745.92

.911

--

--

--

--

--

--

--

740.82

.915

.004

1.10

12

.99

--

--

Model 0 vs.
Model 1

745.19

.917

.002

7.43

12

.83

--

--

Model 1 vs.
Model 2

760.62

.917

.000

16.95

17

.46

--

--

764.36

.917

.000

4.52

4

.34

--

--

Model 2 vs.
Model 3
Model 3 vs
Model 4

778.41

.916

.001

13.52

6

.04*

--

--

768.13

.918

.002

4.35

5

.50

--

--

778.95

.917

.001

11.04

8

.20

--

779.46

.917

.000

1.00

2

.61

3.24

.52

--

--

--

--

--

--

--

--

--

--

--

--

--

--

.99

.61

113

--

Model 4 vs.
Model 5
Model 4 vs.
Model 6
Model 6 vs.
Model 7
Model 7 vs.
Model 8

Model 9 vs.
Model 10

Post-hoc Analyses
Post-hoc analyses were focused on identifying sources of variation related to
degradations in fit for the model with maternal race and univariate relations between race and
housing status. Due to the degradation in overall model fit, but significant contributions of race
in the structural model, the possibility of differences in the measurement model across latent
constructs with race as the grouping variable (White, African American, and Hispanic/Latino)
was explored. These exploratory results indicated that that although a similar model and factor
loadings held across groups (configural vs. metric S-B ΔX2 (24) = 35.95, p = .06; ΔCFI = .001),
intercepts were variant across groups. Specifically, there were significant differences between
models that held factor loadings (RMSEA = 0.049, 90% CI = 0.044-.053, CFI = 0.894, SRMR =
.073, χ2 = 758.31, p = 0.0, df =351) and intercepts (RMSEA = 0.062, 90% CI = 0.057-.066, CFI
= 0.827, SRMR = .091, χ2 = 1074.67, p = 0.0, df = 375) equal to one another, S-BΔX2 scalar vs.
metric (24) = 258.13, p = .00; ΔCFI = .07. Intercept variance across measurement components of
the model suggests differential meaning of latent constructs by race. Thus, group comparisons
that include race with latent constructs described in the study may be uninterpretable or must be
interpreted with caution. Further testing to examine partial invariance extended beyond the scope
of the current study and was not conducted. However, the lack of invariance by maternal race
may be one reason for the degradation in model fit when race was included as a covariate.
In addition, a series of univariate Analysis of Variance (4 X 2 ANOVA) using SPSS
Complex Samples (General Linear Model) were conducted to consider race (1= White, 2= Black,
3= Hispanic/Latino, 4 = Other) by housing (0= Stably Housed, 1 = Unstably Housed)
interactions across the indicators of the model. A significant interaction emerged between race
and housing status with regards to teacher-rated problem behaviors, F(3, 46) = 3.94, p <.05.

114

Using non-reverse coded scores for which lower scores indicated fewer problem behaviors,
results suggested differential patterns across racial groups by housing status (Figure 6). Among
both groups, children of mothers who identified as Hispanic/Latino were rated to have the fewest
behavioral problems by teachers, but this difference was more pronounced for children who were
also unstably housed at any point in Head Start (Mstable = 3.28, Munstable= 2.54). Furthermore,
children of African American mothers among the unstably housed group were rated to exhibit
higher problem behaviors in the classroom by teachers (Mstable= 3.67, Munstable = 5.81). No
additional interactions (race x housing status) emerged across parenting practices, depression
scores, child behavioral ratings, or child assessment measurements.
However, main effects of race emerged for several indicators included in the model. With
regards to parenting variables, significant main effects were found across scores for weekly
activities (F(3, 46) = 29.02 , p <.001), monthly activities (F(3, 46) = 16.84, p <.001), and parental
depression symptomology (F(3, 46) = 15.16, p <.001). Specifically, mothers who identified as
Hispanic/Latino reported the lowest mean levels of depression (M= 2.77), whereas those who
identified as African American reported the highest depression scores (M= 5.45). Furthermore,
mothers who identified as Hispanic/Latino reported fewer weekly activities (M= 11.01), whereas
all other groups reported comparable mean levels of weekly activities (~12). Mothers who
identified as Black/African American or Other reported the highest number of out of home
monthly activities (Mblack= 6.09, Mother= 5.78), whereas those who identified as White or
Hispanic/Latino reported similar levels of monthly engagement activities (Mwhite= 4.51, Mlatino=
4.80). In addition, main effects for race were present among receptive (F(3, 46) = 43.70, p
<.001) and expressive language (F(3, 46) = 16.46, p <.001). Children of mothers who identified
as White scored highest on tests of expressive (Mwhite= 89.12, Mblack = 83.21, Mlatino = 80.31,

115

Mother= 80.60) and receptive language (Mwhite= 97.74, Mblack = 89.49, Mlatino = 77.15, Mother=
89.06) in comparison to all other groups. Lastly, main effects were present by race (F(3, 46) =
5.03, p <.05) and housing (F(1, 48) = 1.39, p <.05) for parent-rated problem behaviors. Children
of mothers who identified as Hispanic/Latino (Mwhite= 5.03, Mblack = 4.77, Mlatino = 6.14, Mother=
5.97) or were unstably housed (Mstable = 4.96, Munstable= 6.00) reported more problem behaviors.
Due to large number of univariate tests and the increase in the Type I error rate, these differences
should be interpreted as exploratory.
Figure 6: Teacher-Rated Problem Behavior Scores by Maternal Race and Housing Status

116

CHAPTER 5
DISCUSSION
Early childhood appears to be a particularly sensitive period for the effects of housing
instability (Fowler et al., 2014). In particular, the pre-school years are especially salient, in light
of the wide literature-base on school readiness that suggests that skills at formal school entry are
predictive of longer-term functioning (e.g., Duncan et al., 2007). The Family Stress Model and
Risk/Resilience highlight the important role of social context in hindering and fostering school
readiness and early development for children experiencing poverty and housing instability.
The Family Stress Model illustrates the importance of parenting practices to children’s
functioning and suggests that poverty fosters a context that affects parenting through the
increased risk for parental depression (Conger et al., 2002). Although research has found that
parents who experience housing instability may be at higher risk for depression and lowered
levels of specific parenting practices, little is known about the differential relationships between
children from families who experience different levels of housing risks (e.g., Suglia et al., 2011).
In addition, early childhood care centers, such as Head Start, also have been linked to children’s
as pre-literacy skills and social-emotional functioning. Factors within Head Start, such as
classroom quality, serve to buffer risk associated with poverty for certain sub-sets of the
population (e.g., Burchinal et al., 2000). Yet, no studies have sought to expand the Family Stress
Model to include factors within Head Start that might intervene or buffer the pathways related to
parenting for families who experience housing stressors for children who face housing risks.
The present study examined the relations between parental depression symptomology,
parenting practices, classroom quality, and children’s functioning during the pre-kindergarten
Head Start year in a national sample of children who attended Head Start between 2009- 2012.

117

One goal of the present study was to consider whether classroom quality can buffer parenting
practices in the indirect relation between parental depression and children’s functioning through
parenting practices for the full pre-kindergarten sample. However, the primary aim of the study
was to consider the differential relations in the full model during the pre-kindergarten year
between children who experienced housing instability during Head Start in comparison to those
who were stably housed throughout Head Start.
This study is the first known empirical examination to use multi-group structural equation
modeling techniques to consider differential relations between unstably housed and stably
housed children. Specifically, the study contributes to the literature on two widely studied
theories: Family Stress Model and Risk/Resilience with a novel sample. In addition, the study is
a unique contribution to the broader literature because of the focus on understanding the complex
relations between multiple social context variables in the consideration of children’s prekindergarten functioning more broadly.
Full Sample Pre-Kindergarten Year Findings
The hypothesized model on the full sample was partially supported. However, evidence
of moderated-mediation was not supported. Classroom quality did not moderate the indirect
relation between parental depression and children’s functioning through parenting practices.
Furthermore, contrary to studies that have shown that preschool classroom quality is related to
early academic functioning, the latent construct of classroom quality was not significantly related
to latent constructs of children’s social-emotional or academic functioning in the present study at
the end of the pre-kindergarten year (Mashburn et al., 2008; Pianta et al., 2005). A plausible
explanation for these results may be that the cross-sectional design of the study may not have
captured the underlying role of classroom quality, which may be masked until later years.

118

Although research suggests gains associated with Head Start and pre-school are immediate, some
literature highlights long-term benefits into childhood and adolescence (Barnett & Hustedt, 2005;
Puma et al., 2005). This literature would suggest that it could be plausible that among all
children who attended Head Start, the role of various types of classroom environments were
similar for most children immediately following the pre-kindergarten year, however children
who were placed in higher quality classrooms during Head Start may exhibit more pronounced
differences throughout childhood and later in their schooling. Therefore, although research has
shown that outcomes associated with Head Start are typically present immediately following
Head Start, some of the benefits associated with quality early childhood care may be masked
until later in children’s developmental trajectory (Barnett & Hustedt, 2005; Ludwig & Miller,
1997).
As predicted, based on the Family Stress Theory, parental approaches partially mediated
the relation between parental depression and children’s pre-kindergarten social-emotional
functioning and language and literacy functioning. Depression negatively predicted parental
approaches, which in turn was positively related to latent constructs of children’s prekindergarten social-emotional functioning and language and literacy functioning. These results
suggested that the influence of parental depression is partially observed through parenting
practices, specifically those reflective of approaches to parenting (e.g., warm interactions, harsh
parenting). This relationship was in the expected direction and supports previous research that
has examined this theory in populations faced with economic stressors (Conger et al., 2002;
Riley et al., 2014). These results underscore the importance of high quality, salient mental health
and parenting supports for families facing sources of economic stress.

119

A few relations that emerged were, however, in the unexpected direction. First, in the
partial mediation of depression on children’s functioning through parenting, the direct
association between parental depression and the latent construct of language and literacy was in
the positive direction, suggesting that increases in endorsement of depression symptomology
were positively related to increase in children’s language and literacy skills. Similarly, parental
depression demonstrated a small, but significant positive association with parental engagement,
leading to a marginally significant, small mediation on social-emotional outcomes. Although the
standardized effects were small (.08), these relationships are contrary to previous research that
has suggested negative relations between parent depression and parenting behaviors, as well as
children’s language skills (e.g., Goodman et al., 2011).
A few explanations may possibly explain these unexpected relations. First, the depression
measure was skewed in the positive direction, with the average score among the full sample
reflecting mild symptomology (M=4.06). This indicated that depression symptomology in the
full sample was relatively low and a sample that exhibited greater variation may more accurately
detect expected relations. The lack of expected relations should be interpreted cautiously.
However, it is unclear as to why then the CES-D scores were sensitive enough to show expected
patterns between depression and parental approaches. Second, the measurement of the construct
of engagement may offer some insight into the positive association between depression and
parental engagement. Though from the broader literature, indicators that comprised this construct
which included parental behaviors such as taking a child on an outing or reading to a child, may
have been less reflective of the types of parent-child interactions that occur during such activities
(e.g., open-ended questioning). Therefore, parents who did endorse depression symptomology
may have still participated in parent-child activities, as well as literacy practices within the home

120

that could have been related to standardized assessment scores; perhaps even more readily in this
sample, given that Head Start aims to increase such opportunities for families. Yet, the types of
interactions with the child during the activity were not measured. Third, parental engagement
items demonstrated borderline levels of reliability in the present study, despite being drawn from
previous national studies and validated measures (e.g., HOME), which may suggest such items
were not a full representation of the construct of parental engagement.
Group Differences by Housing Risk
The primary aim of the study was to consider differences in the hypothesized model
between children who experienced housing instability during Head Start and those who were
stably housed throughout. Both mean level differences in parent depression, parenting, and
children’s pre-kindergarten functioning, as well as differential relations between parental
depression, parenting, classroom quality, and children’s functioning were examined through
MG-SEM, while controlling for economic risk and child gender. MG-SEM techniques provided
a sense of whether differences across nested models were statistically significant, such that a
grouping variable (e.g., housing status) moderated different pathways within the model.
Taken together, few significant differences emerged between stably housed and unstably
housed children in MG-SEM models when accounting for economic risk levels and child’s
gender. With the exception of mean level differences on the latent construct of parent
engagement, no additional differences were found. However, despite a lack of true moderation
indicated by MG-SEM models, initial testing of the model revealed interesting potential sources
of difference across groups that are discussed in further detail below.
All findings should be interpreted in light of the lack of heterogeneity across the sample.
The sample was comprised of a national sample of Head Start children, of which a large

121

percentage of the children faced economic risk factors (e.g., low maternal education, low family
income) indicative of poverty. For this reason, fewer differences may have emerged between
stably housed and unstably housed children and families. Nonetheless, the findings do suggest
that although few differences emerged, there may be unique reasons to consider housing risks as
a similar, but a differential context than other economic risks that are characteristic of poverty.
The absence of differences between groups may also be related to the contributions of
Head Start in mitigating risk for children who faced with housing risks. Specifically, the Head
Start effectiveness research (e.g., Head Start Impact Study) provides evidence for short-term and
immediate gains in areas of pre-literacy and social-emotional functioning (Puma et al., 2005;
Welsch et al., 2010). Given the well documented shorter-term gains associated with Head Start,
it is reasonable to suspect that enrollment in Head Start alone may increase parental and child
skills to a degree. Therefore, null results across groups within a sample of Head Start children
may, in fact, suggest that exposure to Head Start during early childhood may buffer the risks
associated with housing instability. This is an area that warrants future research with a more
rigorous design that includes a comparison group of children facing housing risk and not enrolled
in Head Start programming.
Differences in Levels of Parenting and Child Outcomes
MG-SEM revealed a mean difference in the latent construct of parent engagement.
Specifically, this difference translated to a large standardized difference (.91), which suggested
that parents of children who were unstably housed engaged in lower levels of engagement. This
finding is consistent with previous studies that have found that economic risks and housing
instability, specifically, may pose barriers for families to engage in children’s schooling and
learning-related activities (Jozefowicz-Simbeni & Isreal, 2006; Koblinsky et al., 1997). Yet, this

122

finding may be not fully capture the nuances of parent engagement, as other studies, particularly
qualitative in nature, have indicated that parents who face housing instability may utilize more
context-specific sources to parenting (MacGillvery et al., 2009). The present findings indicate
that parents of children attending in Head Start may face barriers to accessing involvement in
learning and schooling in forms historically characteristic of Head Start and other pre-school
programs. No additional mean level differences emerged in the MG-SEM between unstably
housed and stably housed families or children with regards to parent depression symptoms,
parental approaches, social-emotional functioning or academic skills. These findings suggest
similar pre-kindergarten levels of child and family functioning among families who experienced
housing instability and stably housed families.
The results that compared mean levels of functioning between unstably housed and stably
housed families are a particularly salient contribution to the literature due to the scarcity of
studies that allow for group comparisons due to sample size limitations. Furthermore, this is the
first known study to consider pre-kindergarten levels of functioning among children who
experienced housing instability during Head Start, despite the relevance of this developmental
period to long-term academic and social-emotional functioning and the potential for elevated risk
associated with housing instability during early childhood (Fowler et al., 2014; Rumbold et al.,
2012). The findings implied that among Head Start children, children who experienced housing
instability at least one time point in pre-school may enter kindergarten with comparable socialemotional and academic skills to those who were stably housed throughout Head Start.
Furthermore, when accounting for economic risk and gender, parents of children who were
unstably housed reported comparable levels of depression and employed similar approaches to
parenting, but were less engaged in schooling and learning-related activities. Although these

123

mean differences were not seen across univariate tests on indicators of parent engagement
individually, the use of latent variables allowed for the detection of these differences by
representing a common underlying broader construct.
Differential Relations
MG-SEM revealed no differential pathways between children who experienced housing
instability at least once during Head Start and those who were stably housed throughout. The
lack of differential relations held across direct, indirect, and moderated-mediation pathways,
such that there were non-significant differences in the relations that involved the mediators
(parent engagement and parent approaches) and moderated-mediation (classroom quality and
parenting) interactions. In other words, results suggested a similar pattern of relations, such that
none of the relations included in the model were significantly different based upon housing
status. These finding imply that housing status may not moderate the hypothesized model tested
in the present study.
Contrary to hypothesized relations, the results imply that housing status functions
similarly to other economic risks, evidenced by the lack of differences found across Head Start
children. This further indicates that similar outcomes may be expected across children who face
housing instability during pre-school and those who face economic risk, but are stably housed.
Although it is truly possible that housing instability is not a differential context of risk, apart
from poverty risks, other explanations may also be likely. Despite the larger sample size for a
high-risk population serving as a primary strength of the study, detection of mediation and
moderated-mediation relations require large samples. Therefore, it is also possible that the
sample size for unstably housed group did not provide ample power to detect complex, nuanced
differences of interest to the present study. In addition, the study utilized a broad

124

conceptualization of housing instability that included both families living in various types of
housing conditions and those who experienced frequent moves in a short period of time.
Although this definition is consistent with the approach used by many studies, other studies have
considered differences at the sub-group level (e.g., sheltered families, doubled-up) among those
who do experience housing risks (e.g., Gewirtz et al., 2009; Narayan et al., 2012). Furthermore,
if the sample and available measures had allowed for a more concurrent examination of housing
status, parenting, and child outcomes, the study may have produced differing results.
Despite not approaching significance as true moderation by housing status in multi-group
models, an initial examination of the model for each group separately prior to MG-SEM analyses
revealed one noteworthy difference. The direct path between classroom quality and children’s
pre-kindergarten behavior was significant for children who were unstably housed (β = .27,
p<.05), whereas this path was non-significant for children who were stably housed. Although
this differential relation may have been marginal and not large enough to be captured by MGSEM models, the findings may suggest that it could be possible that classroom quality may have
a slightly more pronounced influence for children who face housing risks. Such a finding is in
line with research has found that children facing certain risks may benefit to a greater extent
from high quality classrooms (Burchinal et al., 2006; Vallotton et al., 2013).
However, there are several reasons this relation may not have emerged in MG-SEM
models. It is possible that factors within the social context, such as classroom quality have more
pronounced effects for smaller sub-groups within the broader sample of unstably housed
children. Due to the inclusion of a broad range of housing conditions in the present study, this
relation might not have been detected clearly in the MG-SEM models. Furthermore, another
explanation may include the consideration of background characteristics that may interact with

125

the social context in mitigating risk. In the studies by Burchinal and colleagues (2000;2006),
classroom quality was most pronounced for African American children facing economic risks on
indicators of language functioning, indicating that the consideration of multiple background
characteristics, such as both race and housing status or gender and housing status could reveal
differential relations. This type of approach is also in line with literature that emphasizes the
concept of intersectionality, or the multiple aspects of identity that shape individual experiences
and is further discussed regarding the post-hoc analyses (Cole, 2009).
Race and Housing Status
Although not a primary aim of the study, exploratory post-hoc analyses considered
measurement invariance by maternal race due to the significant degradations in model fit with
race in the model, as well as and the relations between housing status and maternal race. The
inclusion of these exploratory analyses was important given the relationship between economic
stress and the historical context of housing instability in the United States (Cohen & Wardrip,
2011; Massey & Denton, 1993). Results of these analyses appeared to suggest variance by
maternal race with regards the measurement model, indicating that although maternal race may
by structurally relevant to the model, these relations may not have been measured similarly
across racial groups. Although out of scope within the present study and not explored further, the
exploratory results suggested intercept invariance. As discussed by Keith (2014), intercept
invariance typically signifies differential scaling of the indicators (e.g., starting points), relevance
of specific factors (differential item functioning), and/or bias related to constructs across groups.
Specific reason for these differences were not further explored in the present study.
Exploratory follow-up analyses that focused on univariate relations between race,
housing, and indicators of latent constructs were conducted. A particularly noteworthy area for

126

future exploration may be the interaction between race and housing status with regards to teacher
rated problem behavior, in which these results indicated a trend towards teachers rating children
of African American/Black mothers who also experienced housing instability as exhibiting
higher problem behaviors. Again, despite reaching beyond the scope of the present study, these
findings partially align with the literature that has suggested heightened perceptions of
externalizing behaviors of Black students in the classroom, even as young as pre-school (e.g.,
Graves & Howes, 2011; Mashburn et al., 2006). Scholars have interpreted such findings as a
possible result of mismatch of expectations between families and schools or racial bias, but have
also noted the importance of other context variables in producing differential experiences, such
as racial identities of the broader school population and teacher-student relationships (Graves &
Howes, 2011; Skiba & Williams, 2014). Nonetheless, the central relevance of these results to
the present study is the salience of considering more complex, differential patterns based on the
both housing and race. Although exploratory, these results highlight the importance of the
inclusion of intersectional analyses that consider context at the level of housing status and
multiple background factors that are related to child outcomes and family functioning as a future
direction.
Clinical and Practical Implications
The present study has several strengths and implications for the areas of early childhood
education, prevention, and intervention. A primary implication of the present study is the
importance of continual screening for specific types of risk throughout pre-school and childhood.
Although programs such as Head Start already target high-risk children, the present study
provides some insight into ways children and families who face housing risks in particular may
differ from the general Head Start population. The importance of continual screening by early

127

childhood care centers and schools would allow programs to link families with appropriate
resources and interventions for given developmental periods. Implications of this particular study
are such that families who face housing risks when their children are in pre-school may also face
greater barriers to typical forms of school and home-based engagement at kindergarten entry, and
thus, working with families to consider either alternative methods of engagement (e.g.,
identifying community resources, such as libraries) as identified by existing and future research
or reducing barriers for families (e.g., transportation) may increase the frequency of positive
parent-child activities (Koblinsky et al., 1997; MacGillivery et al., 2009).
The findings also have implications for improving the broader social context of children
who face poverty and housing risks. First, the findings highlight that relevance of parenting
interventions for children who live in poverty and face related risk factors. Although the
magnitude of the relation did not differ by housing status, approaches to parenting served as a
mediator in the relation between parental depression and children’s outcomes, particularly socialemotional functioning across both groups. This is a finding that has consistently held across both
children faced with poverty risks during early childhood more broadly and indicates the
importance of support for positive parenting practices during sensitive developmental periods
(Conger et al., 2002; Riley et al., 2014). This study specifically suggests that the prekindergarten year is a developmental period in which parenting approaches play an important
role.
Clinically, this may translate as accessible, coordinated behavioral health interventions,
such as parent training programs and parental psychotherapy related to the context of poverty.
Given the nature of the Head Start sample, drawing from work on the adaptations of evidencebased mental health interventions (e.g., Incredible Years), the findings also imply that the

128

incorporation and acknowledgement of psycho-social impact of contextual stressors of poverty
within evidence-based behavioral health initiatives with low-income families to deliver more
contextually salient interventions. Working with families to ensure that intervention
recommendations are relevant and feasible for families based on their contextual situations (e.g.,
housing instability) may be most effective.
In addition to parental supports, a noteworthy implication of the present study is the
relevance of early childhood social-emotional functioning. In addition to pre-latent variable
differences, the strongest overall relations emerged with respect to child social-emotional
functioning, particularly with regards to parental depression and parenting approaches both in the
full sample and sub-samples based on housing risks. This is consistent with literature that
suggests that the pre-school years appear to be an important time point for early social-emotional
development (Fowler et al., 2014; Rumbold et al., 2012). Therefore, the results largely lend
support to the literature on universal social-emotional interventions integrated into Head Start
programming. Integration of key skills measured by latent constructs, such as pro-social skills,
coping, and emotion recognition/regulation may be important pre-kindergarten skills to target for
the broader Head Start population, particularly for those facing greater threats in the form of
contextual risks, such as poverty and parental depression that can jeopardize parenting practices
within the home.
Additionally, the findings of the present study also implicate the possibility of differential
relations of quality within the classroom. Although classroom quality played a minimal role in
the overall model and was not a moderator in the mediated pathways that involved parenting, the
analysis of separate models by group membership indicated that classroom quality may have a
significant influence on social-emotional functioning for children who face housing risks. Yet,

129

this difference was not indicated in the full multi-group model. These findings may suggest
although parenting may be the more salient with regards to social context influences during the
pre-kindergarten year, classroom quality may play a greater role for a sub-population of children
who may face housing risks. Although further research has yet to confirm the role of classroom
quality and the present study suggests a minimal, largely non-significant role of classroom
quality, these preliminary findings may indicate that program-level investment in improving
Head Start classroom quality could benefit children facing certain housing risks to a greater
degree.
The present study also suggests the importance of considering housing instability in
conjunction with other background characteristics of the child and family. In the present study,
race, along with housing, emerged as a salient factor in post-hoc analyses, particularly with
respect to child social-emotional functioning in the classroom. In particular, the finding that
teacher rated behaviors were dependent upon and different for children in unstably and stably
housed groups by race implies that child behavior may be perceived differently in the classroom
as a product of multiple characteristics and background factors interact with housing status. In
lieu of these findings, early childhood centers that emphasize cultural sensitivity training for staff
around building school-family relations with families of varied backgrounds, coupled with
supports for marginalized families of color noted earlier, may not only reduce the potential for
bias in teacher perception of child behavior, but could also improve social functioning of the
larger student population of the Head Start center.
Limitations and Future Research
Despite the many strengths of the present study, including a large sample and rigorous
analytic procedures, there are several limitations and areas in which future research could focus.

130

First, future research may consider a longitudinal examination of the model included in the
present study. Although a cross-sectional design was selected to focus on the end of the prekindergarten year, it is possible that the influences of classroom quality and parenting may exert
themselves on academic and social-emotional functioning at later time points during schooling.
Therefore, a longitudinal design that may more clearly demonstrate the trajectories expected for
children who face housing instability during Head Start. More specifically, the use of other Head
Start datasets, including the Head Start Impact study may allow for meaningful comparison of
the developmental trajectories across unstable and stably housed children who did and did not
attend Head Start. Furthermore, comparisons of the dosage of housing instability at multiple time
points across the early childhood and early school-aged years may also be feasible with such
datasets. For example, questions related to whether those who were unstably housed throughout
Head Start fared worse into the early school-aged years compared to those who experienced
instability at fewer time points may be addressed. Such an approach was not feasible in the
present study due to the scope of the study and the lack of available measures of interest across
all time points.
Additionally, the use of secondary data imposes restrictions upon measurement and
definition of constructs. The nature of secondary analysis not only constrains the measures and
approaches used, but also removes the researcher from experiencing the challenges of data
collection and possible sources of error associated with sampling when interpreting results.
Although the dataset selected for use in the present study included accepted indicators of the
constructs of interest, there were both issues related to measurement (e.g., reliability) and/or the
potential for alternative forms of measurements of the constructs.

131

First, there were several measurement issues introduced with available measures. Many
of the measures used in the present study were the combination of several standardized
measures, which may have restricted the constructs intended to be measured by the original
scales. Moreover, F.A.C.E.S developers noted slight alterations to the standardized language and
literacy measures to balance large scale data collection with standardization (Malone et al., 2013)
Additionally, the use of secondary data limited the measurement of the constructs to some
measures with borderline low or moderate levels of reliability (e.g., parenting measures). Issues
related to measurement of indicators may have introduced more error in latent variables and may
not have fully captured the construct of interest. Future research that includes standardized
measures with a wider range of items to capture desired parenting and social-emotional concepts
would contribute further to the literature.
In addition, alternative forms of measurement may be useful to some study constructs.
Specific to parenting practices, while one of the intended goals of the study was to better
understand parenting practices within the context of house instability, the measures involved
approaches to parenting from a largely middle class notion of parenting. It may have also been
likely that the modality, phone interviews, used to collect sensitive information related to
depression symptoms may have resulted in response biases on depression or parenting measures.
Therefore, alternative measures or methodologies, such as qualitative or observational
approaches, may better capture the rich variation that likely exists across families facing housing
instability.
Also related to measurement, there are some challenges with the conceptualization of
housing instability. Due to the available items to represent housing instability and to maximize
the sample size, the present study used a widely accepted definition of housing instability from

132

the broader literature that included those who moved frequently in a short period of time and
those who lived in uncertain housing situations at any point during Head Start (e.g., homeless,
doubled-up). However, there may be variation among the sub-groups that wasn’t captured as a
result of grouping several forms of housing instability together. Although the present study did
not allow for analysis by housing sub-groups, future research could include one or more of the
groups to consider more differences at the housing sub-group levels.
Furthermore, families were included in the unstably housed group if they met criteria at
any point in time during Head Start. The rationale for this approach was to maximize sample
size, capture transiency throughout Head Start, and consider the pre-kindergarten functioning of
those who experienced instability throughout Head Start. Yet, it is likely that some families were
not represented due to losing housing for a shorter period of time between data collection time
points. Although a strength of the approach used in the study is that it included a broad group of
families who are likely to experience housing instability, it could also have masked immediate
effects of instability on parenting or children’s outcomes not explored in the study. Additionally,
this conceptualization of housing instability did not capture the potential role of the length of
exposure, nor whether more cumulative housing risk may differentially relate to outcomes,
Future research that continues to consider ways to conceptualize housing differences among
families would be a valuable contribution to the literature.
Another consideration is the limitations in the analytic approaches used in the present
study. Although SEM and use of latent variables allowed for complex modeling and a novel
approach to consider outcomes between children who experienced differing housing conditions,
there were limitations that were associated with the approaches used. One of the greatest
statistical challenges that arose throughout the study was accounting for latent variable

133

interactions in SEM models, which posed difficulties with assessing model fit. Furthermore,
several changes to the model were initiated as a result of statistical challenges, rendering the
findings exploratory. Some of these changes included alterations to the measurement model and
dropping race as a covariate from the structural base model due to poorer model fit with the
unstably housed group, calling into question the stability of the parameter estimates. Such
changes were made to proceed with analyses and answer the central research questions, however
do require replication by future studies to confirm the findings and thus should be considered
exploratory. Moreover, despite the large sample size for an understudied population, a
simulation study (e.g., Monte Carlo simulation) would be required to truly determine the
necessary sub-samples to determine adequate power. While such challenges are inherent within
SEM studies, it is important for future research to continue to consider alternative analytic
approaches to similar questions to reduce statistical error and consider novel methods to examine
outcomes associated with housing instability during early childhood, including both quantitative
and qualitative methodologies.
Despite the inclusion of a wide range of constructs, another area of measurement that
served as a limitation is the missing data. Even after weighting to adjust for sampling and
attrition in the broader dataset, language measures were incomplete due to F.A.C.E.S sampling
and language routing procedures. This level of missing information can bias conclusions drawn
about the language and literacy construct. Furthermore, general attrition in large scale surveys is
problematic and may be a less valid national representation of the Head Start sample.
A final area of development for future consideration is theoretical, particularly in
expanding the Family Stress Model. Findings from the present study provided support for this
model, yet also delineated the relevance of considering the ways in which other environmental

134

factors, such as classroom or Head Start level factors, interact with parenting. Additionally,
background characteristics of the child or family (e.g., race) may have a differential influence
based on housing conditions. Future research that expands and integrates the Family Stress
Model with models that account for context, such as Risk and Resilience or Ecological Systems
Theory would provide greater insight into the ways in which specific conditions of poverty (e.g.,
housing instability) collectively function.
Conclusion
The present study examined the differential relations between parental depression,
parenting, classroom quality, and children’s pre-kindergarten functioning across Head Start
children who were stably housed and those who experienced housing instability at least one-time
point during Head Start. The study specifically attempted to expand a more established indirect
relation that parenting practices partially explain the relation between parental depression and
child functioning by considering if classroom quality may mitigate risk differentially by housing
status during the pre-kindergarten year of Head Start.
The results somewhat supported the previously established indirect relations in which
approaches to parenting, but not parent engagement, partially explained the relation between
parental depression and children’s functioning across children who were stably housed and
unstably housed. However, classroom quality did not further exert a differential influence on
parenting. Furthermore, few significant group differences emerged between the two groups, with
the exception of differential mean levels of parent engagement practices. Although classroom
quality did significantly predict children’s behavior for the unstably housed group, but not the
stably housed group, this finding did not emerge as significant in the final model. The study
suggests that the influence of housing instability may function similarly to other forms of

135

poverty, but also indicates some areas of difference between the two groups. Therefore, future
directions of this work that include continued exploration of additional factors within Head Start
or the classroom that mitigate housing related risks would continue to expand this avenue of
research, as well as distinguish potential sources of prevention and intervention to buffer housing
related risks.
Although the study was largely focused on housing instability as a developmental context
of risk, the post-hoc analyses further suggested reason for future work to explore the potential
interactive influences of both housing and other background characteristics (e.g., race) in better
understanding the influences of housing instability on child and family functioning. These results
suggested a potentially dependent influence of race and housing for teacher behavior ratings,
such that African American children who experienced housing instability were also rated as
exhibiting the highest level of classroom behaviors. This exploratory finding warrants future
consideration and highlights the importance of a more intersectional approaches that attempt to
integrate multiple aspects of the family background in understanding child and family
functioning to ultimately provide more culturally salient services.
Overall, the present study highlights the importance of considering the potential for
housing instability as a context of risk, in conjunction with other poverty related risks, during the
pre-kindergarten year or prior to kindergarten entry. The findings specifically suggest that levels
of parental engagement may differ for children who experience housing instability as opposed to
those who remain stably housed. Furthermore, there may be differences between children who
experience housing risks and those who are stably housed as a function of race or culture.
Despite the modest findings, the present study is a valuable contribution to the broader literature
because it is the first known examination that employed SEM techniques and considered

136

interactive influences of the social context for children who experienced housing instability
during early childhood. Furthermore, the study expanded our understanding of the connection
between home-based factors, including both parental depression and parenting practices, and
child functioning. Findings from the present study also provide important future directions for
better understanding the role of housing instability prior to kindergarten entry to promote
functioning of an understudied group of children.

137

APPENDIX

138

APPENDIX
Satorra-Bentler Chi Square Formulas
Formulas publicly available at: https://www.statmodel.com/chidiff.shtml
1. cd= (d0 * c0 – d1*c1)/(d0-d1)
cd= difference testing scaling correction factor
d0= degrees of freedom in nested (or more restrictive) model
c0= scaling correction factor of nested model
d1= degrees of freedom in model to be compared
c1= scaling correction factor of model to be compared
2. S-Bχ2 = (T0*c0 – T*c1)/cd
T0= chi-square value of nested (or more restrictive) model
T1= chi-square value of comparison model

139

REFERENCES

140

REFERENCES

Ackerman, B. P., Izard, C. E., Schoff, K., Youngstrom, E. A., & Kogos, J. (1999). Contextual
risk, caregiver emotionality, and the problem behaviors of six and seven year-old children
from economically disadvantaged families. Child Development, 70(6), 1415-1427.
Alexander, K.L., Entwisle, D.R., & Dauber, S.L. (1993). First-grade classroom behavior: Its
short-and long-term consequences for school performance. Child Development, 64(3),
801-814.
Alexander, K. L., Entwisle, D. R., & Thompson, M. S. (1987). School performance, status
relations, and the structure of sentiment: Bringing the teacher back in. American
Sociological Review, 665-682.
Allison, P. D. (2002). Missing data: Quantitative applications in the social sciences. British
Journal of Mathematical and Statistical Psychology, 55(1), 193-196. doi:
10.1348/000711002159653
Anderson, D. R., & Burnham, K. P. (2002). Avoiding pitfalls when using information-theoretic
methods. The Journal of Wildlife Management, 912-918.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review
and recommended two-step approach. Psychological bulletin, 103(3), 411.
Baraldi, A. N., & Enders, C. K. (2010). An introduction to modern missing data analyses.
Journal of School Psychology, 48(1), 5-37. doi:10.1016/j.jsp.2009.10.001
Barnett, W. S., & Hustedt, J. T. (2005). Head Start's lasting benefits. Infants & Young
Children, 18(1), 16-24.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social
psychological research: Conceptual, strategic, and statistical considerations. Journal of
Personality and Social Psychology, 51(6), 1173-1182. doi: 10.1037/0022-3514.51.6.1173
Bashir, S. A. (2002). Home is where the harm is: inadequate housing as a public health crisis.
American Journal of Public Health, 92(5), 733-738. doi: 10.2105/AJPH.92.5.733
Bassuk, E. L., Buckner, J. C., Perloff, J. N., & Bassuk, S. S. (1998). Prevalence of mental health
and substance use disorders among homeless and low-income housed mothers. American
Journal of Psychiatry, 155(11), 1561-1564. doi: 10.1176/ajp.155.11.1561
Bassuk, E. L., DeCandia, C.J., Beach, C. & Berman, F. (2014). America’s Youngest Outcasts: A
report card on child homelessness. Washington DC: The National Center on Family
Homelessness.

141

Bassuk, E. L., Perloff, J. N., & Dawson, R. (2001). Multiply homeless families: The insidious
impact of violence. Housing Policy Debate, 12(2), 299-320. doi:
10.1080/10511482.2001.9521407
Banyard, V. L., & Graham-Bermann, S. A. (1998). Surviving poverty: stress and coping in the
lives of housed and homeless mothers. American Journal of Orthopsychiatry, 68(3), 479489. doi: 10.1037/h0080357
Baumrind, D. (1967). Child care practices anteceding three patterns of preschool behavior.
Genetic psychology monographs, 75(1), 43-88.
Belfield, C. R., Nores, M., Barnett, S., & Schweinhart, L. (2006). The High/Scope Perry
Preschool Program cost–benefit analysis using data from the age-40 followup. Journal of
Human Resources, 41(1), 162-190. doi: 10.3368/jhr.XLI.1.162
Bell, B. A., Onwuegbuzie, A. J., Ferron, J. M., Jiao, Q. G., Hibbard, S. T., & Kromrey, J. D.
(2012). Use of design effects and sample weights in complex health survey data: a review
of published articles using data from 3 commonly used adolescent health surveys.
American Journal of Public Health, 102(7), 1399-1405. doi: 10.2105/AJPH.2011.300398
Berger, L. M., Paxson, C., & Waldfogel, J. (2009). Income and child development. Children and
Youth Services Review, 31(9), 978-989. doi:10.1016/j.childyouth.2009.04.013
Berkowitz, L. (1989). Frustration-aggression hypothesis: examination and reformulation.
Psychological bulletin, 106(1), 59-73. doi: 10.1037/0033-2909.106.1.59
Bierman, K. L., Torres, M. M., Domitrovich, C. E., Welsh, J. A., & Gest, S. D. (2009).
Behavioral and cognitive readiness for school: Cross-domain associations for children
attending Head Start. Social Development, 18(2), 305-323. doi: 10.1111/j.14679507.2008.00490.x
Block, J. H. (1965). Child-rearing practices report. Educational Testing Service.
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley.
Bollen, K. A., & Stine, R. (1990). Direct and indirect effects: Classical and bootstrap estimates
of variability. Sociological Methodology, 20(1), 15-140.
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(3), 803-809.
Brand, A. E., & Brinich, P. M. (1999). Behavior problems and mental health contacts in adopted,
foster, and nonadopted children. Journal of Child Psychology and Psychiatry, 40(8),
1221-1229.

142

Brassard, M. R., & Boehm, A. E. (2007). Preschool assessment: Principles and practices. New
York: Guilford Press.
Bratt, R. G. (2002). Housing and family well-being. Housing Studies, 17(1), 13-26. doi:
10.1080/02673030120105857
Bracken, S. S., & Fischel, J. E. (2008). Family reading behavior and early literacy skills in
preschool children from low-income backgrounds. Early Education and Development,
19(1), 45-67. doi: 10.1080/10409280701838835
Bradley, R.H. & Corwyn, R.F. (2014). Age and ethnic variations in family process mediators of
SES. In M.H. Bornstein & R.H. Bradley (Eds). Socioeconomic status, parenting, and
child development (pp. 161-187). New York, NY: Routledge.
Brown, T. A. (2015). Confirmatory factor analysis for applied research. New York: Guilford
Press.
Brownell, R. (Ed.). (2000). Expressive one-word picture vocabulary test: Manual. Academic
Therapy Publications.
Bulotsky-Shearer, R. J., Wen, X., Faria, A. M., Hahs-Vaughn, D. L., & Korfmacher, J. (2012).
National profiles of classroom quality and family involvement: A multilevel examination
of proximal influences on Head Start children's school readiness. Early Childhood
Research Quarterly, 27(4), 627-639. doi: 10.1016/j.ecresq.2012.02.001
Burchinal, M. R., Peisner-Feinberg, E., Bryant, D. M., & Clifford, R. (2000). Children's social
and cognitive development and child-care quality: Testing for differential associations
related to poverty, gender, or ethnicity. Applied Developmental Science, 4(3), 149-165.
doi: 10.1207/S1532480XADS0403_4
Burchinal, M. R., Roberts, J. E., Nabors, L. A., & Bryant, D. M. (1996). Quality of center child
care and infant cognitive and language development. Child Development, 67(2), 606-620.
doi: 10.1111/j.1467-8624.1996.tb01754.x
Burchinal, M., Roberts, J. E., Zeisel, S. A., Hennon, E. A., & Hooper, S. (2006). Social risk and
protective child, parenting, and child care factors in early elementary school years.
Parenting: Science and Practice, 6(1), 79-113. doi: 10.1207/s15327922par0601_4
Burgard, S. A., Seefeldt, K. S., & Zelner, S. (2012). Housing instability and health: Findings
from the Michigan Recession and Recovery Study. Social Science & Medicine, 75(12),
2215-2224. doi: 10.1016/j.socscimed.2012.08.020
Burgess, S. R., Hecht, S. A., & Lonigan, C. J. (2002). Relations of the home literacy
environment (HLE) to the development of reading-related abilities: A one-year

143

longitudinal study. Reading Research Quarterly, 37(4), 408-426. doi:
10.1598/RRQ.37.4.4
Burnham, K. P., & Anderson, D. R. (2004). Multimodel inference understanding AIC and BIC in
model selection. Sociological Methods & Research, 33(2), 261-304. doi:
10.1177/0049124104268644
Carrion, B. V., Earnshaw, V. A., Kershaw, T., Lewis, J. B., Stasko, E. C., Tobin, J. N., &
Ickovics, J. R. (2014). Housing instability and birth weight among young urban mothers.
Journal of Urban Health, 92(1), 1-9. doi: 10.1007/s11524-014-9913-4
Castro, A. L., Gustafson, E. L., Ford, A. E., Edidin, J. P., Smith, D. L., Hunter, S. J., & Karnik,
N. S. (2014). Psychiatric disorders, high-risk behaviors, and chronicity of episodes
among predominantly African American homeless Chicago youth. Journal of health care
for the poor and underserved, 25(3), 1201-1216. doi: 10.1353/hpu.2014.0124
Caughy, M. O. B., DiPietro, J. A., & Strobino, D. M. (1994). Daycare participation as a
protective factor in the cognitive development of low-income children. Child
Development, 65(2), 457-471. doi: 10.1111/j.1467-8624.1994.tb00763.x
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing
measurement invariance. Structural equation modeling,9(2), 233-255. doi:
10.1207/S15328007SEM0902_5
Choi, J. Y., Elicker, J., Christ, S. L., & Dobbs-Oates, J. (2016). Predicting growth trajectories in
early academic learning: Evidence from growth curve modeling with Head Start
children. Early Childhood Research Quarterly, 36, 244-258. doi:
10.1016/j.ecresq.2015.12.017
Chong, M. T., Yamaki, J., Harwood, M., d'Assalenaux, R., Rosenberg, E., Aruoma, O., &
Bishayee, A. (2014). Assessing health conditions and medication use among the
homeless community in Long Beach, California. Journal of research in pharmacy
practice, 3(2), 56-61. doi: 10.4103/2279-042X.137073
Chromy, J. R. (1979). Sequential sample selection methods. In Proceedings of the Survey
Research Methods Section of the American Statistical Association (pp. 401-406).
Chiu, S. H., & DiMarco, M. A. (2010). A pilot study comparing two developmental screening
tools for use with homeless children. Journal of Pediatric Health Care, 24(2), 73-80. doi:
10.1016/j.pedhc.2009.01.003
Clark, S. L. (2010). Housing instability: Toward a better understanding of frequent residential
mobility among America’s urban poor. Center for Housing Policy. Retrieved from
http://www.partnering-for-change.org/wp-content/uploads/2014/11/Hoursing-InstabilityToward-a-Better-Understanding-of-Frequent-Residential-Mobility-Among-America’sUrban.pdf

144

Clifford, R. M., Barbarin, O., Chang, F., Early, D., Bryant, D., Howes, C., ... & Pianta, R.
(2005). What is pre-kindergarten? Characteristics of public pre-kindergarten programs.
Applied Developmental Science, 9(3), 126-143. doi: 10.1207/s1532480xads0903_1
Cohen, R., & Wardrip, K. (2011). Should I stay or should I go? Exploring the effects of housing
instability and mobility on children. Washington, DC: Center for Housing Policy.
Retrieved from http://mcstudy.norc.org/publications/files/CohenandWardrip_2009.pdf
Cole, E. R. (2009). Intersectionality and research in psychology. American psychologist, 64(3),
170-180. doi: 10.1037/a0014564
Conger, R. D., Wallace, L. E., Sun, Y., Simons, R. L., McLoyd, V. C., & Brody, G. H. (2002).
Economic pressure in African American families: a replication and extension of the
family stress model. Developmental psychology, 38(2), 179-193. doi: 10.1037/00121649.38.2.179
Corman, H., Curtis, M. A., Noonan, K., & Reichman, N. E. (2016). Maternal depression as a risk
factor for children's inadequate housing conditions. Social Science & Medicine, 149, 7683. doi: 10.1016/j.socscimed.2015.11.054
Cunningham, M., Harwood, R., & Hall, S. (2010). Residential instability and the McKinneyVento Homeless Children and Education Program: What we know, plus gaps in research.
Washington, DC: Urban Institute (NJ1).
Currie, J., & Neidell, M. (2007). Getting inside the “black box” of Head Start quality: What
matters and what doesn’t. Economics of Education review, 26(1), 83-99. doi:
10.1016/j.econedurev.2005.03.004
Curtis, M. A., Corman, H., Noonan, K., & Reichman, N. E. (2014). Maternal depression as a risk
factor for family homelessness. American Journal of Public Health, 104(9), 1664-1670.
doi: 10.2105/AJPH.2014.301941
Cutts, D. B., Meyers, A. F., Black, M. M., Casey, P. H., Chilton, M., Cook, J. T., ... & Frank, D.
A. (2011). US housing insecurity and the health of very young children. American
Journal of Public Health, 101(8), 1508-1514.
Cutuli, J. J., Desjardins, C. D., Herbers, J. E., Long, J. D., Heistad, D., Chan, C. K., ... & Masten,
A. S. (2013). Academic achievement trajectories of homeless and highly mobile students:
Resilience in the context of chronic and acute risk. Child Development, 84(3), 841-857.
doi: 10.1111/cdev.12013
Cutuli, J. J., Herbers, J. E., Lafavor, T. L., Ahumada, S. M., Masten, A. S., & Oberg, C. N.
(2014). Asthma and adaptive functioning among homeless kindergarten-aged children in
emergency housing. Journal of health care for the poor and underserved, 25(2), 717-730.
doi: 10.1353/hpu.2014.0099

145

Desimone, L. (1999). Linking parent involvement with student achievement: Do race and
income matter? The Journal of Educational Research, 93(1), 11-30. doi:
10.1080/00220679909597625
Desmond, M., & Perkins, K. L. (2016). Housing and Household Instability. Urban Affairs
Review, 52(3), 421-436. doi: 10.1177/1078087415589192
Dotterer, A. M., Iruka, I. U., & Pungello, E. (2012). Parenting, race, and socioeconomic status:
Links to school readiness. Family Relations, 61(4), 657-670. doi: 10.1111/j.17413729.2012.00716.x
Dowdy, S., Wearden, S., & Chilko, D. (2011). Statistics for research (Vol. 512). John Wiley &
Sons.
Duncan, G. J., Dowsett, C. J., Claessens, A., Magnuson, K., Huston, A. C., Klebanov, P., ... &
Japel, C. (2007). School readiness and later achievement. Developmental Psychology,
43(6), 1428. doi: 10.1037/0012-1649.43.6.1428
Duncan, G. J., Ziol-Guest, K. M., & Kalil, A. (2010). Early childhood poverty and adult
attainment, behavior, and health. Child Development, 81(1), 306-325. doi:
10.1111/j.1467-8624.2009.01396.x
Duchon, L. M., Weitzman, B. C., & Shinn, M. (1999). The relationship of residential instability
to medical care utilization among poor mothers in New York City. Medical Care, 37(12),
1282-1293.
Dunn, D. M., & Dunn, L. M. (2007). Peabody picture vocabulary test manual. Pearson.
Edidin, J. P., Ganim, Z., Hunter, S. J., & Karnik, N. S. (2012). The mental and physical health of
homeless youth: a literature review. Child Psychiatry & Human Development, 43(3),
354-375. doi: 10.1007/s10578-011-0270-1
Edwards, C. P., Sheridan, S. M. D., & Knoche, L. (2008). Parent engagement and school
readiness: Parent-child relationships in early learning. Lincoln, NE: Faculty
Publications, Department of Child, Youth, and Family Studies Retrieved from
http://digitalcommons.unl.edu/cgi/ iewcontent.cgi?article=1060&context
Emerson, M. O., Chai, K. J., & Yancey, G. (2001). Does race matter in residential segregation?
Exploring the preferences of white Americans. American Sociological Review, 66(6),
922-935.
Enders, C. K., & Bandalos, D. L. (2001). The relative performance of full information maximum
likelihood estimation for missing data in structural equation models. Structural Equation
Modeling, 8(3), 430-457. doi: 10.1207/S15328007SEM0803_5

146

Enders, C. K. (2010). Applied missing data analysis. New York, NY: Guilford Publications.
Entwisle, D. R., & Alexander, K. L. (1987). Long-term effects of Cesarean delivery on parents'
beliefs and children's schooling. Developmental Psychology, 23(5), 676-682. doi:
10.1037/0012-1649.23.5.676
Fantuzzo, J. W., LeBoeuf, W. A., Chen, C. C., Rouse, H. L., & Culhane, D. P. (2012). The
unique and combined effects of homelessness and school mobility on the educational
outcomes of young children. Educational Researcher, 41(9), 393-402. doi:
10.3102/0013189X12468210
Fantuzzo, J., Tighe, E., & Childs, S. (2000). Family Involvement Questionnaire: A multivariate
assessment of family participation in early childhood education. Journal of Educational
Psychology, 92(2), 367-376. doi: 10.1037/0022-0663.92.2.367
Fong, E. (1996). A comparative perspective on racial residential segregation. The Sociological
Quarterly, 37(2), 199-226. doi: 10.1111/j.1533-8525.1996.tb01746.x
Foster, M. A., Lambert, R., Abbott-Shim, M., McCarty, F., & Franze, S. (2005). A model of
home learning environment and social risk factors in relation to children's emergent
literacy and social outcomes. Early Childhood Research Quarterly, 20(1), 13-36. doi:
10.1016/j.ecresq.2005.01.006
Fowler, P. J., Henry, D. B., Schoeny, M., Taylor, J., & Chavira, D. (2014). Developmental
timing of housing mobility: Longitudinal effects on externalizing behaviors among at-risk
youth. Journal of the American Academy of Child & Adolescent Psychiatry, 53(2), 199208. doi: 10.1016/j.jaac.2013.12.003
Garg, A., Burrell, L., Tripodis, Y., Goodman, E., Brooks-Gunn, J., & Duggan, A. K. (2013).
Maternal mental health during children's first year of life: Association with receipt of
Section 8 rental assistance. Housing Policy Debate, 23(2), 281-297. doi:
10.1080/10511482.2012.762033
Gelber, A., & Isen, A. (2013). Children's schooling and parents' behavior: Evidence from the
Head Start Impact Study. Journal of Public Economics, 101, 25-38. doi:
10.1016/j.jpubeco.2013.02.005
Gewirtz, A. H., DeGarmo, D. S., Plowman, E. J., August, G., & Realmuto, G. (2009). Parenting,
parental mental health, and child functioning in families residing in supportive
housing. American Journal of Orthopsychiatry, 79(3), 336-347. doi: 10.1037/a0016732
Gonzales, N. A., Coxe, S., Roosa, M. W., White, R. M., Knight, G. P., Zeiders, K. H., & Saenz,
D. (2011). Economic hardship, neighborhood context, and parenting: Prospective effects
on Mexican–American adolescent’s mental health. American Journal of Community
Psychology, 47(1-2), 98-113. doi: 10.1007/s10464-010-9366-1

147

Goodman, S. H., Rouse, M. H., Connell, A. M., Broth, M. R., Hall, C. M., & Heyward, D.
(2011). Maternal depression and child psychopathology: a meta-analytic review. Clinical
Child and Family Psychology Review, 14(1), 1-27. doi: 10.1007/s10567-010-0080-1
Graubard, B. I., & Korn, E. L. (1996). Modeling the sampling design in the analysis of health
surveys. Statistical Methods in Medical Research, 5(3), 263-281. doi:
10.1177/096228029600500304
Graves Jr, S. L., & Howes, C. (2011). Ethnic differences in social-emotional development in
preschool: The impact of teacher child relationships and classroom quality. School
Psychology Quarterly, 26(3), 202-214. doi: 10.1037/a0024117
Gresham, F. M., & Elliot, S. N. (1990). Manual for the social skills rating system. Circle Pines,
MN: American Guidance Service.
Griffith, S., Arnold, D., Voegler-Lee, M. E., & Kupersmidt, J. (2016). Individual characteristics,
family factors, and classroom experiences as predictors of low-income kindergarteners’
social skills. Journal of Educational and Developmental Psychology, 6(1) doi:
10.5539/jedp.v6n1p59
Haber, M. G., & Toro, P. A. (2004). Homelessness among families, children, and adolescents:
An ecological–developmental perspective. Clinical Child and Family Psychology
Review, 7(3), 123-164. doi: 10.1023/B:CCFP.0000045124.09503.f1
Hahs-Vaughn, D. L., McWayne, C. M., Bulotsky-Shearer, R. J., Wen, X., & Faria, A. M. (2011).
Methodological Considerations in Using Complex Survey Data An Applied Example
With the Head Start Family and Child Experiences Survey. Evaluation Review, 35(3),
269-303. doi: 10.1177/0193841X11412071
Hamre, B. K., & Pianta, R. C. (2001). Early teacher-child relationships and the trajectory of
children's school outcomes through eighth grade. Child Development, 72(2), 625-638.
doi: 10.1111/1467-8624.00301
Hamre, B. K., & Pianta, R. C. (2005). Can instructional and emotional support in the first grade
classroom make a difference for children at risk of school failure? Child Development,
76(5), 949-967. doi: 10.1111/j.1467-8624.2005.00889.x
Hamre, B. K., Pianta, R. C., Mashburn, A. J., & Downer, J. T. (2007). Building a science of
classrooms: Application of the CLASS framework in over 4,000 early childhood and
elementary classrooms. Retrieved from
http://fcdus.org/sites/default/files/BuildingAScienceOfClassroomsPiantaHamre. pdf
Harms, T., Clifford, R. M., & Cryer, D. (1998). Early childhood environmental scale.
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A
regression-based approach. New York, NY: Guilford Press.

148

Hayes, A. F., & Preacher, K. J. (2013). Conditional process modeling: Using structural equation
modeling to examine contingent causal processes. In G. R. Hancock & R. O. Mueller
(Eds.). Structural equation modeling: A second course (2nd Ed). Greenwich, CT:
Information Age Publishing.
Herbers, J. E., Cutuli, J. J., Lafavor, T. L., Vrieze, D., Leibel, C., Obradović, J., & Masten, A. S.
(2011). Direct and indirect effects of parenting on the academic functioning of young
homeless children. Early Education & Development, 22(1), 77-104. doi:
10.1080/10409280903507261
Herbers, J. E., Cutuli, J. J., Supkoff, L. M., Heistad, D., Chan, C. K., Hinz, E., & Masten, A. S.
(2012). Early reading skills and academic achievement trajectories of students facing
poverty, homelessness, and high residential mobility. Educational Researcher, 41(9),
366-374. doi: 10.3102/0013189X12445320
Hood, M., Conlon, E., & Andrews, G. (2008). Preschool home literacy practices and children's
literacy development: A longitudinal analysis. Journal of Educational Psychology,
100(2), 252. doi: 10.1037/0022-0663.100.2.252
Howard, K. S., Cartwright, S., & Barajas, R. G. (2009). Examining the impact of parental risk on
family functioning among homeless and housed families. American Journal of
Orthopsychiatry, 79(3), 326-335. doi: 10.1037/a0016599
Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis:
Conventional criteria versus new alternatives. Structural Equation Modeling: A
Multidisciplinary Journal, 6(1), 1-55. doi: 10.1080/10705519909540118
Ingram, K. M., Corning, A. F., & Schmidt, L. D. (1996). The relationship of victimization
experiences to psychological well-being among homeless women and low-income
housed women. Journal of Counseling Psychology, 43(2), 218. doi: 10.1037/00220167.43.2.218
Jackson, A. P., Brooks-Gunn, J., Huang, C., & Glassman, M. (2000). Single mothers in low
wage jobs: Financial strain, parenting, and preschoolers' outcomes. Child Development,
71(5), 1409-1423. doi: 10.1111/1467-8624.00236
Jenson, J. M., & Fraser, M. W. (2006). Social policy for children and families: A risk and
resilience perspective. Thousand Oaks, CA: Sage.
Jozefowicz-Simbeni, D. M. H., & Israel, N. (2006). Services to homeless students and families:
The McKinney-Vento Act and its implications for school social work practice. Children
& Schools, 28(1), 37-44. doi: 10.1093/cs/28.1.3

149

Kamieniecki, G. W. (2001). Prevalence of psychological distress and psychiatric disorders
among homeless youth in Australia: a comparative review. Australian and New Zealand
Journal of Psychiatry, 35(3), 352-358.
Keith, T. Z. (2014). Multiple regression and beyond: An introduction to multiple regression and
structural equation modeling. New York: Routledge.
Kline, R. (2011). Principles and practices of structural equation modeling (3rd edition). New
York: Guilford Press.
Koblinsky, S. A., Kuvalanka, K. A., & Randolph, S. M. (2006). Social skills and behavior
problems of urban, African American preschoolers: role of parenting practices, family
conflict, and maternal depression. American Journal of Orthopsychiatry, 76(4), 554-563.
doi: 10.1037/0002-9432.76.4.554
Koblinsky, S. A., Morgan, K. M., & Anderson, E. A. (1997). African American homeless and
low income housed mothers: Comparison of Parenting Practices. American Journal of
Orthopsychiatry, 67(1), 37-47. doi: 10.1037/h0080209
Kovacs, M., Devlin, B., Pollock, M., Richards, C., & Mukerji, P. (1997). A controlled family
history study of childhood-onset depressive disorder. Archives of General Psychiatry,
54(7), 613-623. doi: 10.1001/archpsyc.1997.01830190033004.
Kochanska, G., Kuczynski, L., & Radke-Yarrow, M. (1989). Correspondence between mothers'
self-reported and observed child-rearing practices. Child Development, 60(1) 56-63. doi:
10.2307/1131070
Landry, S. H., Smith, K. E., Swank, P. R., Assel, M. A., & Vellet, S. (2001). Does early
responsive parenting have a special importance for children's development or is
consistency across early childhood necessary?. Developmental Psychology, 37(3), 387.
doi: 10.1037/0012-1649.37.3.387
Langkamp, D. L., Lehman, A., & Lemeshow, S. (2010). Techniques for handling missing data in
secondary analyses of large surveys. Academic pediatrics, 10(3), 205-210.
doi:10.1016/j.acap.2010.01.005
La Paro, K. M., Pianta, R. C., & Stuhlman, M. (2004). The classroom assessment scoring
system: Findings from the prekindergarten year. The Elementary School Journal, 104(5),
409-426.
La Paro, K. M., Thomason, A. C., Lower, J. K., Kintner-Duffy, V. L., & Cassidy, D. J. (2012).
Examining the definition and measurement of quality in early childhood education: A
review of studies using the ECERS-R from 2003 to 2010. Early Childhood Research &
Practice, 14(1).

150

Lee, V. E., Brooks-Gunn, J., Schnur, E., & Liaw, F. R. (1990). Are Head Start effects sustained?
A longitudinal follow-up comparison of disadvantaged children attending Head Start, no
preschool, and other preschool programs. Child Development, 61(2), 495-507. doi:
10.1111/j.1467-8624.1990.tb02795.x
Lee, E. S., & Forthofer, R. N. (2006). Analyzing complex survey data. New York: Sage
Publications.
Lee, C. Y. S., Lee, J., & August, G. J. (2011). Financial stress, parental depressive symptoms,
parenting practices, and children's externalizing problem behaviors: Underlying
processes. Family Relations, 60(4), 476-490. doi: 10.1111/j.1741-3729.2011.00656.x
Lewit, E. M., & Baker, L. S. (1995). School readiness. The future of children, 128-139.doi:
10.2307/1602361
Lindsey, E. W. (1998). The impact of homelessness and shelter life on family relationships.
Family Relations, 47(4), 243-252. doi: 10.2307/584973
LoCasale-Crouch, J., Konold, T., Pianta, R., Howes, C., Burchinal, M., Bryant, D., ... &
Barbarin, O. (2007). Observed classroom quality profiles in state-funded pre-kindergarten
programs and associations with teacher, program, and classroom characteristics. Early
Childhood Research Quarterly, 22(1), 3-17. doi: 10.1016/j.ecresq.2006.05.001
Love, J. M., Harrison, L., Sagi‐Schwartz, A., Van IJzendoorn, M. H., Ross, C., Ungerer, J. A., ...
& Chazan‐Cohen, R. (2003). Child care quality matters: How conclusions may vary with
context. Child Development, 74(4), 1021-1033. doi: 10.1111/1467-8624.00584
Ludwig, J., & Phillips, D. A. (2008). Long term effects of Head Start on low income
children. Annals of the New York Academy of Sciences, 1136(1), 257-268. doi:
10.1196/annals.1425.005
Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical
evaluation and guidelines for future work. Child development, 71(3), 543-562.
Ma, C. T., Gee, L., & Kushel, M. B. (2008). Associations between housing instability and food
insecurity with health care access in low-income children. Ambulatory Pediatrics, 8(1),
50-57. doi: 10.1016/j.ambp.2007.08.004
MacGillivray, L., Ardell, A. L., & Curwen, M. S. (2009). Libraries, churches, and schools: The
literate lives of mothers and children in a homeless shelter. Urban Education 45(2), 221245 doi: 10.1177/0042085908322721
Mantzicopoulos, P., & Knutson, D. J. (2000). Head Start children: School mobility and
achievement in the early grades. The Journal of Educational Research, 93(5), 305-311.
doi: 10.1080/00220670009598722

151

Meadows-Oliver, M. (2003). Mothering in public: A Meta synthesis of homeless women with
children living in shelters. Journal for Specialists in Pediatric Nursing, 8(4), 130-136.
doi: 10.1111/j.1088-145X.2003.00130.x
Malone, L., Carlson, B.L., Aikens, N., Moiduddin, E., Klein, A.K., West, J…Rall, C. (2013).
Head Start Family and Children Experiences Survey: 2009 User’s Manual. Washington,
DC: Mathematica Policy Research
Mashburn, A. J., Hamre, B. K., Downer, J. T., & Pianta, R. C. (2006). Teacher and classroom
characteristics associated with teachers’ ratings of prekindergartners’ relationships and
behaviors. Journal of Psychoeducational Assessment, 24(4), 367-380. doi:
10.1177/0734282906290594
Mashburn, A. J., Pianta, R. C., Hamre, B. K., Downer, J. T., Barbarin, O. A., Bryant, D., ... &
Howes, C. (2008). Measures of classroom quality in prekindergarten and children’s
development of academic, language, and social skills. Child Development, 79(3), 732749. doi: 10.1111/j.1467-8624.2008.01154.x
Maslowsky, J., Jager, J., & Hemken, D. (2014). Estimating and interpreting latent variable
interactions A tutorial for applying the latent moderated structural equations
method. International journal of behavioral development. doi: 0165025414552301.
Massey, D. S. & Denton, N.A.(1993). American apartheid: Segregation and the making of the
underclass. American Journal of Sociology, 96(2), 329-357.
Masten, A. S. (2001). Ordinary magic: Resilience processes in development. American
Psychologist, 56(3), 227-238. doi: 10.1037/0003-066X.56.3.227
Masten, A. S., Cutuli, J. J., Herbers, J. E., Hinz, E., Obradović, J., & Wenzel, A. J. (2014).
Academic risk and resilience in the context of homelessness. Child Development
Perspectives, 8(4), 201-206. doi: 10.1111/cdep.12088
Mayberry, L. S., Shinn, M., Benton, J. G., & Wise, J. (2014). Families experiencing housing
instability: the effects of housing programs on family routines and rituals. American
Journal of Orthopsychiatry, 84(1) 95- 109. doi: 10.1037/h0098946
Meredith, W. (1993). Measurement invariance, factor analysis and factorial
invariance. Psychometrika, 58(4), 525-543. doi: 10.1007/BF02294825
Meredith, W., & Teresi, J. A. (2006). An essay on measurement and factorial
invariance. Medical Care, 44(11), S69-S77.
Metraux, S., & Culhane, D. P. (1999). Family dynamics, housing, and recurring homelessness
among women in New York City homeless shelters. Journal of Family Issues, 20(3),
371-396. doi: 10.1177/019251399020003004

152

McCoy, D. C., & Raver, C. C. (2014). Household instability and self-regulation among poor
children. Journal of Children and Poverty 20(2), 1-22. doi:
10.1080/10796126.2014.976185
McKinney-Vento Homeless Assistance Act of 1987, Pub. L. No. 100- 77, 101 Stat. 482, 42
U.S.C. § 11301 et seq. (1987).
McLoyd, V. C., Jayaratne, T. E., Ceballo, R., & Borquez, J. (1994). Unemployment and work
interruption among African American single mothers: Effects on parenting and
adolescent socioemotional functioning. Child Development, 65(2), 562-589. doi:
10.1111/j.1467-8624.1994.tb00769
McWayne, C., Fantuzzo, J., Cohen, H. L., & Sekino, Y. (2004). A multivariate examination of
parent involvement and the social and academic competencies of urban kindergarten
children. Psychology in the Schools, 41(3), 363-377. doi: 10.1002/pits.10163
McWayne, C. M., Hahs-Vaughn, D. L., Cheung, K., & Wright, L. E. G. (2012). National profiles
of school readiness skills for Head Start children: An investigation of stability and
change. Early Childhood Research Quarterly, 27(4), 668-683. doi:
doi:10.1016/j.ecresq.2011.10.002
McWayne, C. M., Mattis, J. S., Green Wright, L. E., Limlingan, M. C., & Harris, E. (2016). An
Emic, Mixed-Methods Approach to Defining and Measuring Positive Parenting Among
Low-Income Black Families. Early Education and Development, 1-25. doi:
10.1080/10409289.2016.1208601
Milfont, T. L., & Fischer, R. (2015). Testing measurement invariance across groups:
Applications in cross-cultural research. International Journal of psychological
research, 3(1), 111-130. doi: 10.21500/20112084.857
Milfort, R., & Greenfield, D. B. (2002). Teacher and observer ratings of head start children’s
social skills. Early Childhood Research Quarterly, 17(4), 581-595. doi:
doi:10.1016/S0885-2006(02)00190-4
Miliotis, D., Sesma, Jr, A., & Masten, A. S. (1999). Parenting as a protective process for school
success in children from homeless families. Early Education and Development, 10(2),
111-133. doi: 10.1207/s15566935eed1002_2
Mistry, R. S., Benner, A. D., Biesanz, J. C., Clark, S. L., & Howes, C. (2010). Family and social
risk, and parental investments during the early childhood years as predictors of lowincome children's school readiness outcomes. Early Childhood Research Quarterly,
25(4), 432-449. doi: 10.1016/j.ecresq.2010.01.002
Muthén, B. & Asporuhov, T. (2015). Latent variable interactions. Mplus Web Note November
2, 2015.

153

Muthén, L. K., & Muthén, B. O. (1998-2015). Mplus User's Guide. Seventh Edition. Los
Angeles, CA: Muthén & Muthén.
Myung, I. J. (2003). Tutorial on maximum likelihood estimation. Journal of Mathematical
Psychology, 47(1), 90-100.
Narayan, A. J., Herbers, J. E., Plowman, E. J., Gewirtz, A. H., & Masten, A. S. (2012).
Expressed emotion in homeless families: A methodological study of the five-minute
speech sample. Journal of Family Psychology, 26(4), 648-653. doi: 10.1037/a0028968
Nathanson, L., Lang, N., Than, V., Ketchie, B., Brush, L., and Kirshstein, R. (2003).
Recommended Cognitive Assessment Instrument for the ECLS-B Preschool Battery:
Results of the 2003 Pilot Test. Prepared for the National Center for Education Statistics,
U.S. Department of Education. Washington, DC: American Institutes for Research,.
National Center for Homeless Education (NCHE; 2014). Education for homeless children and
youth consolidated state performance report data. Greensboro, NCHE. Retrieved from
http://center.serve.org/ nche/downloads/data-comp-1011-1213.pdf
Newland, R. P., Crnic, K. A., Cox, M. J., & Mills-Koonce, W. R. (2013). The family model
stress and maternal psychological symptoms: mediated pathways from economic
hardship to parenting. Journal of Family Psychology, 27(1), 96-105. doi:
10.1037/a0031112
Obradović, J., Long, J. D., Cutuli, J. J., Chan, C. K., Hinz, E., Heistad, D., & Masten, A. S.
(2009). Academic achievement of homeless and highly mobile children in an urban
school district: Longitudinal evidence on risk, growth, and resilience. Development and
Psychopathology, 21(02), 493-518. doi: 10.1017/S0954579409000273
Osborne, J. W. (2011). Best practices in using large, complex samples: the importance of using
appropriate weights and design effect compensation. Practical Assessment, Research &
Evaluation, 16, 1531-7714.
Pagani, L. S., Fitzpatrick, C., Archambault, I., & Janosz, M. (2010). School readiness and later
achievement: a French Canadian replication and extension. Developmental Psychology,
46(5), 984-994. doi: 10.1037/a0018881
Park, J. M., Ostler, T., & Fertig, A. (2015). Physical and psychological aggression toward a child
among homeless, doubled-up, and other low-income families. Journal of Social Service
Research, 41(3), 413-423 doi: 10.1080/01488376.2015.1018660.
Patterson, J. M. (2002). Integrating family resilience and family stress theory. Journal of
Marriage and Family, 64(2), 349-360. doi: 10.1111/j.1741-3737.2002.00349.x
Peisner-Feinberg, E. S., Burchinal, M. R., Clifford, R. M., Culkin, M. L., Howes, C., Kagan, S.
L., & Yazejian, N. (2001). The relation of preschool child-care quality to children's

154

cognitive and social developmental trajectories through second grade. Child
Development, 72(5), 1534-1553. doi: 10.1111/1467-8624.00364
Perlman, S., & Fantuzzo, J. W. (2013). Predicting risk of placement: A population-based study
of out-of-home placement, child maltreatment, and emergency housing. Journal of the
Society for Social Work and Research, 4(2), 99-113. doi: 10.5243/jsswr.2013.7
Perlman, M., Zellman, G. L., & Le, V. N. (2004). Examining the psychometric properties of the
early childhood environment rating scale-revised (ECERS-R). Early Childhood Research
Quarterly, 19(3), 398-412. doi: 10.1016/j.ecresq.2004.07.006
Petrill, S. A., & Deater‐Deckard, K. (2004). Task orientation, parental warmth and SES account
for a significant proportion of the shared environmental variance in general cognitive
ability in early childhood: Evidence from a twin study. Developmental Science, 7(1), 2532. doi: 10.1111/j.1467-7687.2004.00319.x
Peterson, J. L., & Zill, N. (1986). Marital disruption, parent-child relationships, and behavior
problems in children. Journal of Marriage and the Family, 48(2), 295-307. doi:
10.2307/352397
Pianta, R., Howes, C., Burchinal, M., Bryant, D., Clifford, R., Early, D., & Barbarin, O. (2005).
Features of pre-kindergarten programs, classrooms, and teachers: Do they predict
observed classroom quality and child-teacher interactions?. Applied Developmental
Science, 9(3), 144-159. doi: 10.1207/s1532480xads0903_2
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects
in simple mediation models. Behavior Research Methods, Instruments, & Computers,
36(4), 717-731.
Puma, M., Bell, S., Cook, R., Heid, C., Broene, P., Jenkins, F., … & Downer, J. (2012). Third
Grade Follow-up to the Head Start Impact Study Final Report, OPRE Report # 2012-45,
Washington, DC: Office of Planning, Research and Evaluation, Administration for
Children and Families, U.S. Department of Health and Human Services.
Radloff, L. S. (1977). The CES-D scale a self-report depression scale for research in the general
population. Applied Psychological Measurement, 1(3), 385-401.
Rafferty, Y., Shinn, M., & Weitzman, B. C. (2004). Academic achievement among formerly
homeless adolescents and their continuously housed peers. Journal of School Psychology,
42(3), 179-199. doi: 10.1016/j.jsp.2004.02.002
Raver, C. C., Jones, S. M., Li-Grining, C. P., Metzger, M., Champion, K. M., & Sardin, L.
(2008). Improving preschool classroom processes: Preliminary findings from a
randomized trial implemented in Head Start settings. Early Childhood Research
Quarterly, 23(1), 10-26. doi: 10.1016/j.ecresq.2007.09.001

155

Rauh, V. A., Chew, G. R., & Garfinkel, R. S. (2002). Deteriorated housing contributes to high
cockroach allergen levels in inner-city households. Environmental Health Perspectives,
110(2), 323-327.
Reynolds, M. R., & Keith, T. Z. (2013). Measurement and statistical issues in child assessment
research. In D. H. Soklofske, C. R. Reynolds, & V. Schwean (Eds.), Oxford Handbook of
Child and Adolescent Assessment (pp. 48-83). New York: Oxford University Press.
Rozario, P. A., & Menon, N. (2010). An examination of the measurement adequacy of the CESD among African American women family caregivers. Psychiatry Research, 179(1), 107112. doi: 10.1016/j.psychres.2010.06.022
Richards, R., Merrill, R. M., Baksh, L., & McGarry, J. (2011). Maternal health behaviors and
infant health outcomes among homeless mothers: US Special Supplemental Nutrition
Program for Women, Infants, and Children (WIC) 2000–2007. Preventive Medicine,
52(1), 87-94. doi: 10.1016/j.ypmed.2010.10.010
Riley, M. R., Scaramella, L. V., & McGoron, L. (2014). Disentangling the associations between
contextual stress, sensitive parenting, and children's social development. Family
Relations, 63(2), 287-299. doi: 10.1111/fare.12063
Rumbold, A. R., Giles, L. C., Whitrow, M. J., Steele, E. J., Davies, C. E., Davies, M. J., &
Moore, V. M. (2012). The effects of house moves during early childhood on child mental
health at age 9 years. BMC Public Health, 12(1), 583. doi: 10.1186/1471-2458-12-583
Sabol, T. J., & Pianta, R. C. (2012). Patterns of school readiness forecast achievement and
socioemotional development at the end of elementary school. Child Development, 83(1),
282-299. doi: 10.1111/j.1467-8624.2011.01678.x
Satorra, A. (2000). Scaled and adjusted restricted tests in multi-sample analysis of moment
structures. In Heijmans, R.D.H., Pollock, D.S.G. & Satorra, A. (eds.), Innovations in
multivariate statistical analysis. (pp.233-247). London: Kluwer Academic Publishers.
Schafer, J. L., & Kang, J. (2008). Average causal effects from nonrandomized studies: a practical
guide and simulated example. Psychological Methods, 13(4), 279-313. doi:
10.1037/a0014268
Schmitt, S. A., Finders, J. K., & McClelland, M. M. (2015). Residential mobility, inhibitory
control, and academic achievement in preschool. Early Education and
Development, 26(2), 189-208. doi: 10409289.2015.975033
Schweinhart, L. J., & Weikart, D. P. (1997). The High/Scope preschool curriculum comparison
study through age 23. Early Childhood Research Quarterly, 12(2), 117-143. doi:
10.1016/S0885-2006(97)90009-0

156

Sell, K., Zlotnik, S., Noonan, K., & Rubin, D. (2010). The effect of recession on child well-being.
Philadelphia, PA: Children’s Hospital of Philadelphia.
Shaw, D. S., Winslow, E. B., Owens, E. B., Vondra, J. I., Cohn, J. F., & Bell, R. Q. (1998). The
development of early externalizing problems among children from low-income families:
A transformational perspective. Journal of Abnormal Child Psychology, 26(2), 95-107.
doi: 10.1023/A:1022665704584
Sheridan, S. M., Knoche, L. L., Kupzyk, K. A., Edwards, C. P., & Marvin, C. A. (2011). A
randomized trial examining the effects of parent engagement on early language and
literacy: The Getting Ready intervention. Journal of School Psychology, 49(3), 361-383.
doi: 10.1016/j.jsp.2011.03.001
Shinn, M., Weitzman, B. C., Stojanovic, D., Knickman, J. R., Jimenez, L., Duchon, L., ... &
Krantz, D. H. (1998). Predictors of homelessness among families in New York City:
from shelter request to housing stability. American Journal of Public Health, 88(11),
1651-1657. doi: 10.2105/AJPH.88.11.16
Shonkoff, J. P., Garner, A. S., Siegel, B. S., Dobbins, M. I., Earls, M. F., McGuinn, L., ... &
Wood, D. L. (2012). The lifelong effects of early childhood adversity and toxic stress.
Pediatrics, 129(1), 232-246. doi: 10.1542/peds.2011-2663
Simpson, G. A., & Fowler, M. G. (1994). Geographic mobility and children's
emotional/behavioral adjustment and school functioning. Pediatrics, 93(2), 303-309.
Sirin, S. R. (2005). Socioeconomic status and academic achievement: A meta-analytic review of
research. Review of Educational Research, 75(3), 417-453. doi:
10.3102/00346543075003417
Skiba, R. J., & Williams, N. T. (2014). Are Black kids worse? Myths and facts about racial
differences in behavior. The Equity Project at Indiana University.
South, S. J., & Crowder, K. D. (1998). Leaving the 'hood: Residential mobility between black,
white, and integrated neighborhoods. American sociological review, 17-26.
Suglia, S.F., Chambers, E., & Sandel, M.P. (2015). Poor housing quality and housing instability.
In M.T. Compton & R.S. Shim (Ed.), The Social Determinants of Mental Health (pp.
171-192). Arlington, VA: American Psychiatric Publishing.
Suglia, S. F., Duarte, C. S., & Sandel, M. T. (2011). Housing quality, housing instability, and
maternal mental health. Journal of Urban Health, 88(6), 1105-1116. doi:
10.1007/s11524-011-9587-0
Sugrue, T. J. (2005). The Origin of the Urban Crisis: Race and Inequality in Postwar Detroit.
Princeton: Princeton University Press.

157

Supkoff, L. M., Puig, J., & Sroufe, L. A. (2012). Situating resilience in developmental context.
In Unger, M. (Eds.) The social ecology of resilience: A handbook of theory and practice
(pp. 127-142). New York, NY: Springer Publications.
Suizzo, M.-A., & Stapleton, L. (2007). Home-based parental involvement in young children’s
education: Examining the effects of maternal education across U.S. ethnic groups.
Educational Psychology, 27(4), 533-556. doi: 10.1080/01443410601159936
Thrasher, S. P., & Mowbray, C. T. (1995). A strengths perspective: An ethnographic study of
homeless women with children. Health & Social Work, 20(2), 93-101. doi:
10.1093/hsw/20.2.93
Tischler, V., Rademeyer, A., & Vostanis, P. (2007). Mothers experiencing homelessness: Mental
health, support and social care needs. Health & Social Care in the Community, 15(3),
246-253. doi: 10.1111/j.1365-2524.2006.00678.x
Thompson, M. S., & Green, S. B. (2006). Evaluating between-group differences in latent
variable means. In Hancock, G.R. & Muller, R.O. (Eds.) Structural equation modeling:
A second course (pp. 119-169). Charlotte, NC: Information Age Publishing, Inc.
Thomas, S. L., & Heck, R. H. (2001). Analysis of large-scale secondary data in higher education
research: Potential perils associated with complex sampling designs. Research in Higher
Education, 42(5), 517-540. doi: 10.1023/A:1011098109834
Tofighi, D., & MacKinnon, D. P. (2016). Monte Carlo confidence intervals for complex
functions of indirect effects. Structural Equation Modeling: A Multidisciplinary
Journal, 23(2), 194-205. doi: 10.1080/10705511.2015.1057284
U.S. Department of Housing and Urban Development (HUD) (2014). The 2014 point-in-time
estimates of homelessness. The 2014 Annual Homeless Assessment Report to Congress,
(Vol 1.). Washington, DC:.HUD.
U.S. Department of Education (2014). Homeless program (McKinney- Vento), Homeless
Students Enrolled in LEAs with or without McKinney- Vento Subgrants, 2012-13.
Washington, DC: ED Data Express. Retrieved from: http://eddataexpress.ed.gov/dataelements.cfm
Vallotton, C. D., Harewood, T., Ayoub, C. A., Pan, B., Mastergeorge, A. M., & Brophy-Herb, H.
(2012). Buffering boys and boosting girls: The protective and promotive effects of Early
Head Start for children's expressive language in the context of parenting stress. Early
Childhood Research Quarterly, 27(4), 695-707. doi: 10.1016/j.ecresq.2011.03.001
Vandermaas-Peeler, M., Nelson, J., Bumpass, C., & Sassine, B. (2009). Social contexts of
development: Parent-child interactions during reading and play. Journal of Early
Childhood Literacy, 9(3), 295-317. doi: 10.1177/1468798409345112

158

Vijayaraghavan, M., Kushel, M. B., Vittinghoff, E., Kertesz, S., Jacobs, D., Lewis, C. E., ... &
Bibbins-Domingo, K. (2013). Housing instability and incident hypertension in the
CARDIA cohort. Journal of Urban Health, 90(3), 427-441.doi: 0.1007/s11524-0129729-z
Vijayaraghavan, M., Tochterman, A., Hsu, E., Johnson, K., Marcus, S., & Caton, C. L. (2012).
Health, access to health care, and health care use among homeless women with a history
of intimate partner violence. Journal of Community Health, 37(5), 1032-1039. doi:
10.1007/s10900-011-9527-7
Webster-Stratton, C., Reid, M. J., & Hammond, M. (2001). Preventing conduct problems,
promoting social competence: A parent and teacher training partnership in Head Start.
Journal of Clinical Child Psychology, 30(3), 283-302. doi:
10.1207/S15374424JCCP3003_2
Weinreb, L., Goldberg, R., Bassuk, E., & Perloff, J. (1998). Determinants of health and service
use patterns in homeless and low-income housed children. Pediatrics, 102(3), 554-562.
doi: 10.1542/peds.102.3.554
Welsh, J. A., Nix, R. L., Blair, C., Bierman, K. L., & Nelson, K. E. (2010). The development of
cognitive skills and gains in academic school readiness for children from low-income
families. Journal of educational psychology, 102(1), 43 - 53. doi: 10.1037/a0016738
West, J., Tarullo, L., Aikens, N., Sprachman, S., Ross, C., & Carlson, B. L. (2007). FACES 2006
study design. Washington DC: US Department of Health and Human Services,
Administration for Children and Families, Mathematica Policy Research.
Whittaker, J. E. V., Harden, B. J., See, H. M., Meisch, A. D., & Westbrook, T. P. R. (2011).
Family risks and protective factors: Pathways to Early Head Start toddlers’ social–
emotional functioning. Early Childhood Research Quarterly, 26(1), 74-86. doi:
10.1016/j.ecresq.2010.04.007
Woodcock, R. W., Muñoz-Sandoval, A. F., McGrew, K. S., & Mather, N. (2007). Batería III
Woodcock-Muñoz. Rolling Meadows, IL: Riverside.
Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). Woodcock-Johnson® III NU Tests of
Achievement.
Yu, M., North, C. S., LaVesser, P. D., Osborne, V. A., & Spitznagel, E. L. (2008). A comparison
study of psychiatric and behavior disorders and cognitive ability among homeless and
housed children. Community Mental Health Journal, 44(1), 1-10. doi: 10.1007/s10597007-9100-0
Zabkiewicz, D. M., Patterson, M., Wright, A., Barker, J., Keller, C., & Goering, P. (2014). A
cross-sectional examination of the mental health of homeless mothers: does the

159

relationship between mothering and mental health vary by duration of homelessness?
BMJ Open, 4(12) doi:10.1136/bmjopen-2014-006174
Zill, N., & Resnick, G. (2006). Emergent literacy of low-income children in Head Start:
Relationships with child and family characteristics, program factors, and classroom
quality. Handbook of Early Literacy Research, 2, 347-371.
Ziol‐Guest, K. M., & McKenna, C. C. (2014). Early childhood housing instability and school
readiness. Child Development, 85(1), 103-113. doi: 10.1111/cdev.12105
Zlotnick, C. (2009). What research tells us about the intersecting streams of homelessness and
foster care. American Journal of Orthopsychiatry, 79(3), 319-325. doi:
10.1037/a0017218
Zlotnick, C., Johnson, D. M., & Kohn, R. (2006). Intimate partner violence and long-term
psychosocial functioning in a national sample of American women. Journal of
Interpersonal Violence, 21(2), 262-275

160