Comparing public, private, and informal preschool programs in a national sample of low-income children

Early Childhood Research Quarterly 36 (2016) 91–105

Contents lists available at ScienceDirect

Early Childhood Research Quarterly

Comparing public, private, and informal preschool programs in a national sample of low-income children Rebekah Levine Coley a,∗ , Elizabeth Votruba-Drzal b , Melissa Collins a , Kyle DeMeo Cook a a b

Department of Counseling, Developmental, and Educational Psychology, Boston College, 140 Commonwealth Ave., Chestnut Hill, MA 02467, United States Department of Psychology, University of Pittsburgh, 4200 Fifth Ave., Pittsburgh, PA 15260, United States

a r t i c l e

i n f o

Article history: Received 10 November 2014 Received in revised form 22 October 2015 Accepted 8 November 2015 Keywords: Early childhood education Preschool Head Start Public pre-K School readiness

a b s t r a c t Recent research has found that center-based early education and care (EEC) programs promote gains in cognitive skills for low-income children, but knowledge is limited concerning diverse types of EEC arrangements. This paper contrasts the primary EEC arrangements (Head Start, public centers, private centers, and home care) attended by economically disadvantaged children in the US with data on 4250 low-income children from the nationally-representative ECLS-B cohort. Results found public centers and Head Start programs provided children with the most educated and highly trained teachers and with the most enriching learning activities and global quality, with private centers showing moderate levels and home EEC very low levels of quality. Nonetheless, after adjusting for differential selection into EEC through propensity score weighting, low-income children who attended private EEC centers showed the highest math, reading, and language skills at age 5, with children attending Head Start and public centers also showing heightened math and reading skills in comparison to children experiencing only parent care. No differences were found in children’s behavioral skills at age five in relation to EEC type. Results support enhanced access to all center preschool programs for low-income children, and suggest the need for greater understanding of the processes through which EEC affects children’s school readiness skills. © 2015 Elsevier Inc. All rights reserved.

1. Introduction The use of nonparental care for children prior to school entry has grown dramatically in recent decades, driven by increased needs of parents in the workforce as well as enhanced provision of publicly supported early education programs. As rates of maternal employment and single-parent families expanded, an increasing proportion of families required alternate care providers for their young children. At the same time, evidence grew regarding the potential for early education and care (EEC) programs, particularly high-quality center-based programs in the year or two prior to kindergarten, to improve the school readiness skills of children (Yoshikawa et al., 2013), that is, the nascent language, literacy, math, and behavioral skills that are essential for a positive transition to kindergarten and continued educational success (Snow, 2006). Much of this evidence focused on the efficacy of centerbased preschool programs to bolster the school readiness skills of economically disadvantaged children, offering a potential mechanism to reduce the expanding achievement gaps between poor and advantaged children (Magnuson, Waldfogel, & Washbrook,

2012; Reardon, 2011). As such, policy makers and scholars have increasingly turned to early education programs as a mechanism for supporting the nascent skills of economically disadvantaged children, helping them to prepare for future educational and economic success. And yet, with the plethora of EEC programs and funding models that have emerged in the US, there is a dearth of information regarding which EEC settings are most effective in supporting the school readiness skills of children from low-income families. This study seeks to provide a careful analysis of the EEC settings attended by low-income children in the US, using a nationally representative sample of children followed prospectively from early childhood through kindergarten entry. By comparing children attending home-based, private center, public center, and Head Start programs, this study seeks first to provide a detailed description of diverse EEC arrangements and second, using quasi-experimental analysis techniques, to assess how diverse EEC arrangements support the school readiness skills of economically disadvantaged children. 1.1. The early education and care landscape

∗ Corresponding author. Fax: +1 617 552 1981. E-mail address: [email protected] (R.L. Coley). http://dx.doi.org/10.1016/j.ecresq.2015.11.002 0885-2006/© 2015 Elsevier Inc. All rights reserved.

A great diversity of EEC programs are used by low-income families in the US, including Head Start programs, public preschool

92

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

centers, and private childcare or preschool centers, as well as homebased care. A recent report on a national sample of children in 2012 found that more than 76% of 3–5 year-old children used nonparental care at least one day a week, with 60% in center-based EEC programs and 36% in home-based care, with some children in more than one care type (Mamedova & Redford, 2015). Within the arena of center-based EEC programs, a growing proportion of children are attending publicly-funded programs. National estimates suggest that 29% of 4-year-olds in 2012–2013 were in state-funded preschool programs, with an additional 3% served by special education public programs and another 10% served by federally-funded Head Start (Barnett, Carolan, Squires, Clarke Browne, & Horowitz, 2015). The high use of publicly-supported EEC reflects the dramatic surge in such programs in recent years, although it is important to note that funding levels and access have vacillated due to budget shifts at both federal and state levels, with large waiting lists often reported for child care subsidies as well as public preschool and Head Start slots (Barnett & Carolan, 2013; Schulman & Blank, 2013). In short, a wide variety of funding streams and program models have emerged to provide EEC services to low-income children, with an equally broad range of regulatory mechanisms targeting structural and process quality features. Nearly all states in the US have implemented or are developing quality rating and improvement systems (QRIS) to track quality of EEC programs and provide information to parents, practitioners, and policy makers (Tout et al., 2010). And yet, regulations and quality indicators remain highly variable across EEC arrangements. Home-based EEC in particular has variable and limited regulations, and numerous studies have found low levels of both structural (e.g., teacher training and education) and process (e.g., the quality of the materials and teacher-child interactions) quality across home EEC programs (Coley, Li-Grining, & Chase-Lansdale, 2006; Fuller, Kagan, Loeb, & Chang, 2004). On the other end of the spectrum, Head Start is highly regulated. Head Start programs are required to serve primarily poor children and children with disabilities; to provide health services, family services, and family involvement programs in addition to early education; and to use research-based curricula to promote children’s learning and development. Teacher education requirements have been less rigorous, and although they increased in the past decade, data from the most recent cohort of the Head Start Family and Child Experiences Survey (FACES) found that less than half of Head Start teachers had a bachelor’s degree and slightly more than half had training in early childhood education (Hulsey et al., 2011). Public and private EEC centers do not have the uniformity of quality regulations that Head Start does, but public preschool programs in particular have been found to have numerous indicators of high quality. Using data from the National Center for Early Development and Learning’s (NCEDL) Multi-State Pre-Kindergarten Study, Clifford and colleagues reported that nearly 70% of teachers in public preschool programs had a bachelor’s degree or higher, more than three-quarters of the programs offered additional services for families and children, and essentially all used a learning curriculum (Clifford et al., 2005). Other work has directly compared public and private centers, finding that publicly-operated EEC programs had teachers with greater education and training, higher pay, and more stability than private centers (Bellm, Burton, Whitebook, Broatch, & Young, 2002). Research also has contrasted global program quality, assessed through measures such as the ECERS-R and FDCERS, across different EEC arrangements serving low-income preschool children. Such research has found that Head Start programs showed higher ratings of global quality than other centers, which in turn were higher than homes (Li-Grining & Coley, 2006; Fuller et al., 2004). This research did not, however, distinguish between public versus private center-based programs. It is also important to note that global quality measures such as the ECERS-R have come under

increased scrutiny, with recent research finding validity weaknesses and limited connections to students’ school readiness skills in large national samples (Gordon, Fujimoto, Kaestner, Korenman, & Abner, 2013; Sabol, Soliday Hong, Pianta, & Burchinal, 2013; Votruba-Drzal, Coley, Koury, & Miller, 2013; Weiland, Ulvestad, Sachs, & Yoshikawa, 2013). Indeed, a recent assessment of the type of quality indicators used in state QRIS systems found that most indicators showed no substantial association with children’s functioning (Sabol et al., 2013), raising additional concerns over the quality measures used in many policy and assessment systems. In contrast, a number of recent evaluations of curricular models in Head Start and public preschool programs found positive causal impacts on low-income children’s early reading and math skills (Assel, Landry, Swank, & Gunnweig, 2007; Clements & Sarama, 2007; Fantuzzo, Gadsden, & McDermott, 2011; Lonigan, Farver, Phillips, & Clancy-Menchetti, 2011), highlighting the importance of structured literacy and math learning activities in EEC programs. Together, past research suggests that Head Start and publiclyfunded EEC programs are likely to show higher quality than private centers and home-based EEC. Although research has highlighted some differences in quality indicators across EEC arrangements, much of this work has taken a piecemeal approach, often assessing only one type of EEC at a time or using local samples, and knowledge remains limited concerning how diverse regulations may translate into quality features of varied EEC programs across the country. One of the goals of this study is to expand this comparative view, using a nationally representative sample of children to compare quality characteristics across the four major types of EEC programs attended by low-income preschool-age children: Head Start, public centers, private centers, and home care. 1.2. EEC settings and children’s school readiness The second goal of this research is to test associations between EEC arrangements and children’s development. A host of research studies have found that center-based EEC programs in the year or two prior to kindergarten can help raise the school readiness skills of economically disadvantaged children (Yoshikawa et al., 2013). Center EEC is associated with heightened reading, math, and language scores in comparison to parent or home care, although results have been more mixed in relation to behavioral skills (Coley, Votruba-Drzal, Miller, & Koury, 2013; Gormley & Gayer, 2005; Gormley, Gayer, Phillips, & Dawson, 2005; Loeb, Bridges, Bassok, Fuller, & Rumberger, 2007; Magnuson, Meyers, Ruhm, & Waldfogel, 2004; Votruba-Drzal et al., 2013). A primary limitation in this research base is the restricted attention to differences across subtypes of center-based EEC. Given the diversity of teacher qualifications, classroom quality, and access to other services and supports across Head Start, public, and private EEC programs (Smith, Kleiner, Parsad, & Farris, 2003), we might expect that public EEC programs may be most effective at supporting the development of low-income children, and private programs least effective. Prior research has not adequately addressed these hypotheses, either because studies have combined diverse EEC program types into broader categories, have used reports of EEC setting with significant validity concerns, or simply have not conducted tests comparing the effectiveness between different EEC types. For example, a number of recent experimental and quasiexperimental studies have assessed impacts of public preschool or Head Start programs on children’s school readiness skills. In a set of rigorous studies, Gormley and colleagues (2005; Gormley, Phillips, Newmark, Welti, & Adelstein, 2011) assessed Oklahoma’s universal public preschool program, finding that the program led to increases in children’s language, literacy, and math skills, and to no meaningful changes in behavioral skills. Evaluation of the Boston public preschool program found similar results, with

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

moderate-sized benefits for cognitive skills and small positive effects on executive function outcomes (Weiland & Yoshikawa, 2014). The national Head Start Impact Study, similarly, found consistent short-term cognitive benefits, with more mixed behavioral results (U.S. Department of Health and Human Services, Administration for Children and Families, 2010). Although these studies used random assignment or quasi-experimental methods to assess program impacts, the ambiguity of the control groups, which combined children in diverse EEC arrangements, preclude a careful comparison of different EEC types. Re-analyses of the Head Start Impact Study have sought to address this limitation by comparing children who attended Head Start to peers who attended other types of EEC programs. Two recent papers, using principal score matching and principal stratification to adjust for selection bias, found cognitive benefits of Head Start in comparison to parent or home EEC, with less consistent evidence of behavioral benefits (Feller, Grindal, Miratrix, & Page, 2014; Zhai, Brooks-Gunn, & Waldfogel, 2014). These studies uncovered no differences in child functioning between those who attended Head Start versus other centers, a category which combined both public and private center programs. Work with the Fragile Families Study sample using propensity score matching techniques found similar results in terms of differences between Head Start and parent or home care; however, they also found that children who attended Head Start had improved parent-reported behavioral functioning compared to children in other center-based EEC (Zhai, Brooks-Gunn, & Waldfogel, 2011). In short, these studies have found benefits of Head Start in comparison to parent or home care, but generally few differences in the functioning of children who attended Head Start versus other center programs, without delineating between public and private centers. Other studies using nationally representative longitudinal surveys (the Early Childhood Longitudinal Study, Birth Cohort [ECLS-B] and the Early Childhood Longitudinal Study, Kindergarten Cohort [ECLS-K]) have compared outcomes between children attending different types of EEC arrangements. For example, Lee and colleagues separated pre-kindergarten (pre-k) from other types of centers, finding that children who attended pre-k programs had higher reading scores than their peers who attended Head Start, with the latter outperforming children in home EEC or parent care in cognitive skills, but underperforming in teacher-reported behavioral outcomes (Lee, Zhai, Brooks-Gunn, Han & Waldfogel, 2014). Research with the ECLS-K data, in contrast, found benefits of center EEC in comparison to parent care for children’s cognitive skills, but no benefits of Head Start (Magnuson et al., 2004). However, this research base remains limited with a lack of attention to differences between public and private EEC centers, and between these two options and Head Start or home programs. A second central limitation is the reliance on parent reports of EEC type, a reliance shared by essentially all of the studies listed above. Scholars have raised substantial validity concerns about parent reports of EEC (Magnuson, Ruhm, & Waldfogel, 2007), and indeed, study design and verification research has found that parents do not reliably and correctly distinguish between terms such as pre-k, preschool, and center programs (Datta, 2013). Thus, it is essential to assess whether differences in child functioning across distinct EEC settings replicate with more valid reports, such as teacher reports of EEC type. 1.3. Selection into EEC Within this research area, it is essential to highlight the importance of attending to differential selection into EEC. Extant research has delineated a broad range of factors associated with selection into EEC programs, ranging from entrance requirements regarding family income or welfare status, to local availability, to child

93

and family characteristics (Chaudry, Henly, & Meyers, 2010; Coley, Votruba-Drzal, Collins, & Miller, 2014; Weber, 2011). It is beyond the scope of the current study to provide a detailed review of this extensive literature, but a recent analysis of the nationally representative ECLS-B data provides a strong example of patterns across much of the research base. This study (Coley et al., 2014) found that human capital factors such as maternal employment were associated with greater selection into all types of EEC over parent care, and greater selection of home EEC than center or Head Start programs. Parental education, in contrast, was associated with greater selection into centers over Head Start and home EEC or parent care. Low income and welfare receipt, not surprisingly, were associated with heightened selection of Head Start over other types of centers. Racial/ethnic differences emerged as well, with African American and immigrant children more likely to use Head Start and less likely to use parent or home care in comparison to white children. Rural children were more likely to attend Head Start programs, whereas children in large urban areas were in parent care more often. Competition for EEC was also important, with less home care availability associated with greater use of EEC centers. Little research has explicitly addressed how low-income families differentially select into the diverse array of EEC options which may be available to them, distinguishing between public and private centers and Head Start programs in comparison to more informal home EEC settings. Together, this research highlights the necessity of carefully attending to differential selection into EEC both in an effort to understand how children end up in different types of EEC programs and in seeking to delineate repercussions for children’s development. 1.4. Research goals and hypotheses Given the heightened public and policy attention to expanding preschool opportunities for low-income children and increasing preschool quality, it is essential to bolster our understanding of the diverse EEC arrangements experienced by low-income children today. This study assessed a nationally representative sample of children to examine the variety of EEC arrangements attended by 4-year-old children in low-income families. Improving on prior research, we use provider reports to delineate EEC type and distinguish Head Start, other publicly funded centers, private centers, and home-based EEC, providing a more valid and detailed breakdown of EEC type than prior research. Our first aim is to provide rich descriptive information on the quality characteristics of EEC programs. Our second aim is to delineate and adjust for differential selection into EEC type and, having done so, assess which type of EEC programs are most effective at promoting economically disadvantaged children’s cognitive and behavioral skills. Based upon prior research, we expected that Head Start and public EEC centers would show the highest levels of quality and would be most effective at promoting children’s school readiness skills, particularly in the realms of cognitive and language skills. 2. Method 2.1. Participants Data were drawn from the Early Childhood Longitudinal Study, Birth Cohort (ECLS-B), a longitudinal study following a nationally representative sample of approximately 10,700 children (the ECLSB requires that all Ns be rounded to the nearest 50) born in the U.S. in 2001 from infancy through kindergarten entry (Chernoff, Flanagan, McPhee, & Park, 2007). The ECLS-B sampling criteria excluded children who died or were adopted prior to 9 months of age and children born to mothers under 15 years of age. The

94

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

ECLS-B collected four waves of data from primary caregiver interviews (with the child’s mother in 98% of cases) and child assessments when children were, on average, 10 months, 2 years, 4 years, and 5 years of age, with response rates of 74%, 93%, 91%, and 92% (for children not yet in kindergarten at wave 4, the ECLS-B returned the following year for a final interview). For children in regular nonparental care at the age 4 interview, the child’s primary EEC provider was interviewed (response rate of 87% of providers) and a subset of EEC programs was assessed with structured observations (response rate of 57% of programs). At the age-5 wave the child’s kindergarten teacher (or primary EEC provider for children who had not yet entered kindergarten) was interviewed (response rates of 76% of kindergarten teachers and 92% of EEC providers). In the current study, we focus on children’s EEC experiences when children were 4 years of age (wave 3) and on their cognitive and behavioral skills when children were age 5 and most were transitioning into kindergarten (wave 4). Moreover, we focus on children from low-income families, defined as living in a family at age 4 with household income at or below 200% of the federal poverty line. We chose this income cut-off to concur with the most common definition of low income and to match the modal income eligibility limits for public preschool programs, which range from 100% to 230% of poverty (Barnett, Hustedt, Hawkinson, & Robin, 2006). This income criterion also allowed us to capture the majority of children in the ECLS-B who were attending the EEC arrangements of primary interest (90% of children in Head Start and 65% of children in public preschool programs). Of the 8900 children who remained in the sample at age 4, 48% (N ≈ 4250) met this lowincome criterion. Within this analytic sample, missing data at the variable level ranged from 0 to 59% (with the largest amount of missing data derived from purposeful data collection decisions, primarily the collection of EEC quality observations on only a subset of cases). Because missing data introduce biases into the sample, missing data were imputed using multiple imputation by chained equations implemented in Stata 12 to create 20 complete datasets (Little & Rubin, 2002; Royston, 2005). The missing data imputation included all of the variables used in this analysis (described below), as well as earlier measures of child skills and family characteristics to improve the accuracy of the imputation. Just over half (52%) of the analytic sample were boys, and children averaged over 4 years of age at the wave that EEC was assessed. Families were racially and ethnically diverse: 36% were native-born white, 19% native Black, 11% native Hispanic, 4% native-born from other racial/ethnic groups, 2% Asian (nearly all immigrants), 23% Hispanic immigrant, and 5% immigrants from other (non-Asian or Hispanic) backgrounds. Family income was low, averaging just over the poverty line. The majority of mothers were employed at least some of the time since the child’s infancy and nearly all received some type of welfare. In relation to EEC, the majority of children in the sample, 73%, attended an EEC program at age 4, with 22% in home EEC arrangements, 21% in Head Start programs, 19% in private centers, and the smallest proportion, 10%, in public centers; the remaining 27% were in parent care. 2.2. Measures 2.2.1. Children’s school readiness skills Children’s cognitive skills and behavioral functioning were assessed at age 5 (wave 4). Cognitive skills were measured using direct assessments developed specifically for the ECLS-B, comprised of items drawn from well-validated standardized instruments including the Peabody Picture Vocabulary Test Third Edition (PPVT-III; Dunn & Dunn, 1997), the PreLAS 2000 (Duncan & DeAvila, 1998), the Preschool Comprehensive Test of Phonological & Print Processing (Lonigan, Wagner, Torgeson, & Rashotte, 2002), and the Test of Early Mathematics Ability (3rd ed., Ginsburg & Baroody, 2003). The

early reading assessment (˛ = 0.92) consisted of 74 items that measured early reading and language skills, including letter knowledge, word recognition, print conventions, and phonological awareness. The math assessment (˛ = 0.92) consisted of 58 items focused on number sense, properties, operations, and probability. Analyses utilized the IRT scale scores calculated by the ECLS-B for these assessments. Children also completed the Let’s Tell Stories subscale of the PreLAS (Duncan & DeAvila, 1998), which assessed children’s expressive language skills through a storytelling task. Responses were coded on a 0–5 scale with higher scores indicating greater coherence, fluency, and complexity of language use. Behavioral functioning at age 5 was assessed via teacher reports (or provider reports for those not yet in kindergarten) on items drawn from the Preschool and Kindergarten Behavior Scales—Second Edition (PKBS-2; Merrell, 2003), the Social Skills Rating Scales (SSRS; Gresham & Elliott, 1990) and items created specifically for the ECLS-B and the Head Start Family and Child Experiences Study (FACES). Teachers/caregivers rated the frequency of the child’s engagement in behaviors on 5-point scales (ranging from “never” to “very often”). Factor analyses led to the construction of three measures. Externalizing problems assessed children’s impulsive, disruptive, and aggressive behaviors (7 items; ˛t = 0.86, ˛p = 0.97). Approaches to learning measured children’s attention, independence, task completion, and eagerness to learn (6 items, ˛t = 0.89, ˛p = 0.83). Finally, prosocial skills included behaviors such as making friends, sharing, and comforting others (6 items, ˛tp = 0.87). 2.2.2. EEC type and characteristics EEC provider reports were used to delineate EEC type for all children attending nonparental care for 5 or more hours per week at the age 4 interview. Children receiving 0–4 h per week of nonparental care (per parent report) were coded as parent care. Providers from each child’s primary nonparental care setting were interviewed. Primary nonparental care arrangement was defined as the arrangement where the child attended the most hours per week; 14% of children attended more than one type of EEC. EEC providers were asked first whether care was provided in a private home or in a center, with the former coded as home care. Center-based EEC providers were further asked to delineate the program type, choosing from options including public school prekindergarten (coded as public centers), Head Start program (coded as Head Start centers), or child care center, preschool/nursery school, or private school prekindergarten (all of which were coded as private centers). It is important to note that this designation of EEC type is somewhat imprecise, and reflects the broader national environment in which many EEC programs receive funding from multiple sources and must comply with diverse regulations and standards. Most notably, many public prekindergarten programs include slots for children in privately-run child care centers, and other children use public subsidies in private centers (Barnett et al., 2015); hence the private center group likely misclassifies some public prekindergarten children as being in private centers. Nonetheless, it is likely that provider reports of EEC settings are more valid than those of parents, and validation checks on these setting types (e.g., clarifying that public centers were primarily located in public schools whereas private centers were located in separate buildings, religious organizations, or other businesses) provided additional support for the type categorizations. Additional EEC characteristics were reported by parents, by providers, or through direct assessments. Parents reported on the hours per week the child attended the EEC program and the amount they paid for that EEC per month. EEC providers reported on their own characteristics, including their education, coded to indicate whether they had a bachelor’s degree or higher versus not, and whether or not they had a degree in early childhood education or

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

a related field. Providers also reported the number of years they had been an EEC provider and the number of hours they attended professional training in the prior year. This variable was coded into three categories: no training, 1–14 h, and 15 or more hours. These categories were constructed based on NIEER quality benchmarks (Barnett & Carolan, 2013; Barnett et al., 2015) and prior studies showing that teachers who engage in more than 15 h of annual inservice training tend to have better teaching practices (Whitebook, Howes & Phillips, 1989). In relation to the EEC program, providers reported on the number of other children cared for along with the focal child, the number of non-English speaking children, and the number of providers, which we used to create a child:adult ratio. Providers also reported whether the EEC program was reimbursed for meals or snacks through the United States Department of Agriculture’s (USDA) Child and Adult Care Food Program (CACFP), which we view as a proxy for serving economically disadvantaged children. Providers reported on aspects of enrichment in the program as well, including the number of books and the frequency with which they engaged (0 = never to 5 = every day) in literacy activities (10 items such as learning letter names, learning about rhyming, discussing new words, practicing writing, listening to stories where children see print) and math activities (10 items including counting out loud, engaging in shape and pattern activities, playing math games, and working with counting manipulatives). The literacy activity items were standardized and averaged into a total literacy activities score (˛ = 0.86); similarly, the math items were standardized and averaged into a total math activities score (˛ = 0.87). These scales were correlated at 0.76 and hence were averaged into a total activities score. A final measure of global EEC quality was derived from direct observations conducted with a subset of the settings (n ≈ 950). Global EEC quality was assessed through structured observations using the Early Childhood Environment Rating Scale—Revised (ECERS-R) (Harms, Clifford, & Cryer, 1998) for centers and the Family Day Care Rating Scale (FDCRS) for homes (Harms & Clifford, 1989), both of which create total quality scores ranging from 1 (inadequate) to 7 (excellent) with strong internal reliability (˛ = 0.88 for centers and 0.89 for homes). These measures assess both structural and process quality features of EEC programs, such as having adequate and safe indoor and outdoor play and learning spaces, sanitary feeding and toileting practices, rich use of language, responsive and warm interactions and supervision, and developmentally appropriate and accessible learning materials and activities across a broad range of topic areas. 2.2.3. Child, family, and community characteristics A host of child, family, and community characteristics assessed in the ECLS-B were included due to their association in prior research with selection into EEC (Coley et al., 2014). Child characteristics reported by mothers included age in months at the age 4 interview, gender, whether the child was part of a multiple birth (i.e., twins, triplets, etc.), and birth weight, which was coded to indicate low birth weight of less than 2500 g. Mothers reported on children’s general health at ages 10 months, 2 years, and 4 years (1 = excellent to 5 = poor) and reports were recorded into an indicator reflecting whether the child was ever reported to be in fair/poor health. Additional indicators assessed whether the child was ever reported to have a physical or cognitive/ behavioral disability across the first three waves. Additional characteristics included child cognitive and behavioral functioning at age 2 (collected, on average, just prior to the time children were initiating care in their EEC setting). Cognitive functioning was directly assessed at age 2 using the Bayley Short Form-Research Edition (BSF-R; Chernoff et al., 2007; adapted from Bayley, 1993), which measures diverse domains of cognitive devel-

95

opment including exploration of objects, early problem-solving, and communication (˛2 = 0.88). Child temperament, a measure of children’s behavioral functioning tapping into adaptability, social engagement, and behavioral regulation, also was assessed at wave 2 directly and by maternal reports using selected items from the Infant/Toddler Symptom Checklist (ITSC; DeGangi, Poisson, Sickel, & Wiener, 1995) and the Behavior Rating Scale (BRS; Bayley, 1993), with higher scores representing a more regulated temperament (20 items, ˛2 = 0.86). We also included an indicator of whether the child had entered kindergarten at the age 5 interview and the number of months of kindergarten experienced at the time of assessment. A broad array of family characteristics was measured, with timevarying variables assessed at ages 10 months, 2 years, and 4 years and averaged for continuous variables or categorized for categorical variables to capture children’s family context from infancy through age 4. Race/ethnicity and immigrant status were combined into a set of mutually exclusive groups delineating native-born Whites, Blacks, Hispanics, and other races (including parents of different races, Native Americans, and other); Asians (94% of whom were immigrants); and immigrants of Hispanic and other origin. An additional indicator designated families whose primary language was not English. Family income-to-needs was calculated at each wave by dividing the annual household income by the federal poverty line for a family of its size, with scores averaged across the first three measurement periods. Parental education was represented as a series of dummy variables indicating whether the highest level of attainment across the first three waves of data collection was less than a high school diploma, high school diploma/GED, some college or vocational school, or a Bachelor’s degree or higher. Mothers’ employment status and welfare receipt (receipt of TANF, SNAP, or WIC) were measured at each interview and then delineated as stable across the three waves, some across the three waves, or none. An additional measure of maternal employment assessed hours per week employed and was averaged across the three waves. Family structure indicators delineated whether the mother was stably married, ever married, stably cohabitating, ever cohabitating but never married, or stably single across the first three measurement periods (child ages 10 months to 4 years). The mother’s age at the birth of her first child was included as a continuous variable. Additional indicators designated whether there was stably, ever, or never more than two adults living in the household. The number of children in the household was coded continuously and averaged across the three measurement points. Two additional measures assessed parental functioning. Maternal depression was assessed at child ages 10 months and 4 years using a modified version of the Center for Epidemiological Studies-Depression Scale (CES-D; Radloff, 1977; 12 items, ˛ = 0.88–0.89) and dichotomized to indicate whether the mother scored in the moderately to severely depressed range; these indicators were combined to delineate mothers who stably, sometimes, or never reported moderate to high levels of depressive symptoms. Finally, cognitive stimulation in the home was assessed using items from the Short Form of the Home Observation for Measurement of the Environment Inventory (HOME-SF; Caldwell & Bradley, 1979, 2001), the National Household Education Survey (NHES), and additional questions. Items assessed the frequency with which parents engage in a variety of learning activities with their children, such as reading, telling stories, or singing to their child. Reports from each wave were aggregated into composites which were then averaged to create a measure of cognitive stimulation in the home environment (46 items, ˛ = 0.82). The final set of variables assessed contextual factors. Two measures of the availability of publicly funded EEC were assessed at the state level using information from the National Institute for Early Education Research’s (NIEER) 2005 State Preschool Yearbook (Barnett, Hustedt, Robin, & Schulman, 2005). The first delineated

96

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

the state child care assistance subsidy eligibility limit for a family of three as a percent of the state median income (SMI) as of October 2003; the second indicated the percent of 4-year-olds in the state enrolled in state preschool programs in the 2004–2005 school year (reported at 0 for states with no state preschool programs). State preschool programs were defined as a state-funded or directed initiative serving 3–4 year-olds in a group setting for at least two days per week with a primary focus on education. Programs had to be distinct from (but could be coordinated with) the state’s child care subsidy system, and state supplements for Head Start that increased the number of children served in the state were considered state preschool programs. Additional measures of EEC availability were created using methods developed by Coley and colleagues (2014; based on work by Gordon & Chase-Lansdale, 2001). A special tabulation of the 2000 Decennial Census provided the number of children under age six living in each ZIP code in the U.S., as well as the number of childcare center workers and family daycare workers employed in each ZIP code. Data were aggregated across all ZIP codes falling within a 20mile radius of the centroid of each child’s ZIP code of residence using Geographic Information Systems software. Measures of availability of center care and home-based care were created by dividing the number of children under six by the number of care providers (center or home) and logging to correct for non-normality. Accordingly, this measure assessed competition for EEC, with larger numbers indicating lower availability of care. Finally, each family’s geographic location was characterized as falling within a large urban city, small urban area, suburb, or rural area using children’s ZIP codes and Rural-Urban Commuting Area Codes created by USDA. 2.3. Analytic plan The first research aim was to provide a rich description of the features of EEC programs used by low-income children. This analysis was done by estimating a series of no-constant regression models in which the independent variables were indicators of EEC type and the dependent variables were characteristics of EEC settings. Wald tests were used to assess significant differences in characteristics across the different EEC types. These analyses were weighted with 90 replicate weights (w3r1–w3r90) using jackknife replication methods to adjust for differential sampling and attrition, to properly estimate standard errors given the sampling plan, and to provide estimates drawn from low-income families in a nationally representative sample. To address the second research aim, which was to examine links between EEC type and children’s cognitive and behavioral skills, we conducted a series of OLS regression analyses with propensity score weighting (PSW) techniques to take into account the differential selection of children into EEC (Imbens, 2000; Rosenbaum & Rubin, 1984). A quasi-experimental analytic method, propensity score (PS) analyses restructure correlational data to mimic randomized experimental data where treatment and control groups are equated on observed, pre-existing characteristics (Rosenbaum & Rubin, 1983). Although adjusting for the propensity to be in the “treatment” group has been shown to remove a substantial portion of selection bias in nonexperimental research (for example, see Leon & Hedeker, 2005), it is important to note that PS techniques cannot control for unobserved factors, the influence of which may even be magnified by matching on observables (Pearl, 2009). PSW techniques followed the three-step procedure described by Imbens (2000). The first step used a multinomial logistic regression model to estimate the propensity of children to be in each EEC type

as a function of all observed child, family, and community covariates drawn from interviews at child ages 10 months, 2 years, and 4 years, as described above. This multinomial logistic regression model also provides information on child, family, and contextual characteristics associated with differential selection into EEC type. The second step in the PSW analysis involved creating propensity score weights by taking the inverse of the child’s conditional probability of experiencing the EEC type that the child actually received (Imbens, 2000). In developing and testing the PSW modeling strategy, we found ample evidence for the rigor and strength of this technique. First, we explored the distributions of the PS weights across the EEC types. To rein in the influence of the tail ends of the PS weights and strengthen the internal validity, we followed the technique of Smith and Todd (2005), employing trimming procedures and discarding cases in the region of non-overlap. Using these trimmed PS weights, we re-estimated the multinomial logistic regression model predicting EEC type to assess the rigor of the PS weights to adjust for differential pre-existing characteristics across the EEC types. Initial analyses with sampling weights found that 19% (83 of 430) of the pairwise comparisons between predictors showed significant differences across EEC type. After weighting with the PS weights, significant differences were dramatically reduced to 6% of all comparisons, although not completely eliminated (details discussed below in results section). Thus, as an added protection against bias, the full set of covariates, which are those listed above along with an indicator for whether the child was in kindergarten and the number of months they were in kindergarten at the age 5 interview, were included in the third and final step of the PSW modeling, OLS regression models predicting children’s cognitive and behavior skills at age 5. EEC treatment-specific propensity score weights were applied to generate the average treatment effect of EEC program types. We note that in comparison to PS matching techniques, PS weighting improves generalizability and power by maximizing sample size, but as occurred here, may result in groups that are not completely equal on all observed characteristics (Imbens, 2000). In developing this modeling strategy, we also considered alternate models. For example, we assessed OLS regression models using the ECLS-B sampling weights, which increase the external validity of the results but provide less rigor in adjusting for differential selection into EEC. We also considered including age 4 lagged measures of child functioning variables. However, descriptive analyses found that children had begun receiving care in their primary EEC setting at a mean of 29.48 (SD = 19) months, meaning that children had already experienced their delineated EEC program for approximately 23 months prior to the age 4 data collection, making the use of age 4 functioning an inappropriate choice as a lagged variable. Hence, we included the age-2 measures of children’s cognitive and behavioral skills, which were assessed, on average, just prior to initiation of care in the age-4 EEC program, to provide the most temporally-appropriate manner of adjusting for prior child functioning. Finally, we also considered the use of multilevel models with state fixed or random effects to further adjust for differential availability of EEC; however, the distribution of the ECLS-B sample did not provide adequate statistical power in a number of the smaller states in the U.S. to properly estimate multilevel models, and hence we decided that inclusion of measures of state subsidy generosity, state public pre-k slots, and competition for both formal and informal EEC slots coded as individual-level variables provided adequate adjustment for differential availability of EEC options. Results from additional sensitivity tests are described in the results section.

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

3. Results

97

Table 1 presents descriptive data on characteristics of EEC programs (omitting children in parent care) with significant differences between types indicated by matched superscripts. These descriptive results show numerous significant differences across the four EEC types. First, considering family experiences, results show that home EEC programs had children attending for the greatest number of hours per week (over 31 h), about 1/2 a standard deviation (SD) higher than all other EEC types, which each averaged in the low 20 s. Private centers had by far the highest costs for parents, averaging over $1900 per month, about 1 SD higher than family costs for Head Start and public EEC programs, which averaged just over $250 a month with home EEC programs in between. Several significant differences across EEC types emerged in relation to indicators of program quality. Teachers in public centers were the most highly educated, with 79% of teachers holding a bachelor’s degree, a rate nearly double that of teachers in Head Start programs and private centers, and many times higher than home providers. Teachers with degrees in early childhood education or a related field followed a similar pattern. Home EEC providers had particularly low levels of education, and also reported minimal hours of professional training, with 82% of home providers reporting no training in the past year, a rate that was closer to 20% for the other EEC types. Providers in Head Start programs were the most likely to acquire the recommended 15 or more hours of training per year, followed by teachers at public and private EEC centers. Head Start programs were distinct in a number of other ways as well, reporting the highest group size, the largest number of nonEnglish speakers, and the largest proportion of government meal reimbursement, a sign of serving substantial numbers of poor children. Public centers were not far behind on all of these indicators, and they also had the highest child:adult ratios. Together, these results suggest that public preschool settings generally had teachers with the highest education and training, but that, along with Head Start programs, they also had the largest classrooms, the highest child:adult ratios, and the greatest numbers of poor children and those learning English as a second language. EEC programs also varied on indicators of global quality and learning opportunities and materials. On observed measures of global program quality, Head Start and public centers outscored private centers by almost 1/2 a SD and outscored home settings by nearly 1 SD, indicating moderate to large differences. Similarly, Head Start and public centers reported more frequent literacy and math activities than private centers, with differences of nearly 1/3 of a SD, and outscored home EEC programs by more than 1 SD. Further, public centers reported significantly more books than all other types, with home EECs again scoring lowest, with differences ranging from about 1/5–1/2 of a SD. In short, Head Start and public center programs showed the highest levels of global quality and materials and activities for learning, with home EEC showing the lowest levels, and private programs evincing middling quality.

we highlight the overarching patterns of results. In relation to child characteristics, results indicate that higher age-2 cognitive scores were associated with a higher likelihood of attending Head Start in comparison to home or private center settings, whereas age-2 adaptive temperament was associated with a higher likelihood of attending home versus Head Start settings. Children from multiple births were more likely to be in Head Start versus home programs. Child age was a very strong predictor of EEC type: older children were more likely to be in public centers than in parent, home, Head Start, or private center settings, and more likely to be in private centers or Head Start than in parent care or home arrangements. A broad array of family characteristics was associated with EEC type. Native born African American children were more than twice as likely to be in Head Start programs and just under twice as likely to be in public centers as in parent, home, or private center arrangements in comparison to native white children. Children in Hispanic immigrant families were least likely to be in private centers in comparison to parent, home or Head Start programs, whereas children from other immigrant families were three times as likely to be in Head Start versus parent care in comparison to native white children. Non-English status was also associated with a heightened likelihood of Head Start use in comparison to home care. In relation to socioeconomic characteristics, higher family income and parental education were predictive of greater use of private centers than Head Start settings. Use of welfare programs was associated with about a 4-fold increase in Head Start use in comparison to other EEC types. Maternal employment, in contrast, was associated with a much stronger likelihood of home EEC care in comparison to all other types, and with a greater likelihood of private centers than parent care. Family structure was also important. Mothers with partners, whether spouses or cohabiting partners, were notably more likely to have children in parent care than all other types of EEC, and the stable presence of more than two adults in the household predicted higher likelihood of the child being in home-based care compared to all other EEC types. In contrast, higher maternal age and having fewer children were associated with a greater use of private centers in comparison to many other types of EEC. In short, these results suggest that even within this low-income sample, greater socioeconomic disadvantage was associated with a higher likelihood of children being in Head Start programs and a lower likelihood of private center programs. Finally, in relation to contextual characteristics, children in states with higher subsidy eligibility for public preschool were 3 to 4 times more likely to be in public preschool programs than in parent care or other EEC arrangements, and the percent of children in state preschool was an even stronger predictor. Children in communities with greater competition for informal EEC were also more likely to be in public preschool in comparison to home EEC arrangements. Urbanicity was also important: children in small urban areas were less likely to be in home versus parent care, whereas rural children were about twice as likely to be in Head Start as all other EEC types, in comparison to children in suburban communities. These results suggest the importance of EEC policies and availability in driving family EEC choices.

3.2. Selection into EEC type

3.3. Propensity score weighting models limiting selection bias

The next set of analyses addressed selection into EEC type through a multinomial logistic regression model predicting EEC type from child, family, and contextual characteristics. Table 2 presents results, with relative risk ratios and standard errors presented for each type of EEC in comparison to parent care, and matched superscripts designating significant differences between EEC arrangements. Although it is beyond the scope of the current paper to discuss analysis of selection into EEC in great detail (e.g., through delineation of effect sizes for every paired difference), here

Table 3 presents results of the multinomial logistic regression models re-estimated using the trimmed PS weights. As noted above, the use of the PS weights removed a substantial proportion, but not all, of the significant differences between EEC types. Few significant differences remained in relation to child characteristics or contextual characteristics. Most of the remaining differences were associated with a handful of family variables. Some race/ethnicity/ nativity differences remained, with native Blacks, native others, and Asians showing higher rates of Head

3.1. Characteristics of EEC programs

98

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

Table 1 Preschool characteristics across EEC settings. Home (N ≈ 850) %/M (SD)

Head Start (N ≈ 1000) %/M (SD)

Public center (N ≈ 450) %/M (SD)

Private center (N ≈ 750) %/M (SD)

Significant differences between EEC types

Individual characteristics Number of hours in EEC EEC cost/month

31.68 (18.63)abc 1161.67 (1810.16)abc

24.25 (12.79)a 259.41 (911.67)ad

24.36 (13.39)b 266.05 (901.99)be

22.88 (13.65)c 1921.26 (2069.65)cde

Home > HS Pub Priv Priv > Home > HS Pub

Structural quality BA or greater Early childhood degree 0 Training hours last year 1–14 Training hours last year 15+ Training hours last year Years as an EEC provider Group size Child: adult ratio Num. non-english Reimbursed for meals

11%abc 16%abc 82%abc 7%abc 11%abc 10.44 (9.79)abc 3.15 (4.64)abc 2.79 (2.95)abc 0.92 (3.10)abc 26%abc

43%ad 62%ade 16%ad 22%a 62%ad 13.42 (9.08)a 14.33 (5.63)ad 5.49 (2.84)ad 4.66 (6.08)ad 92%ade

79%bde 78%bdf 18%b 26%b 57%b 12.19 (8.87)b 14.08 (5.74)be 6.06 (3.77)bde 3.74 (5.42)b 79%bdf

41%ce 53%cef 22%cd 27%c 51%cd 13.27 (8.96)c 12.63 (6.17)cde 5.33 (3.19)ce 2.85 (4.94)cd 44%cef

Pub > Priv HS > Home Pub > HS > Priv > Home Home > Priv > HS; Home > Pub HS Pub Priv > Home HS > Priv > Home; Priv>Home HS Pub Priv > Home HS Pub > Priv > Home Pub > HS Priv > Home HS > Priv > Home; Pub > Home HS > Pub > Priv > Home

Process quality FDCRS/ECERS total score Literacy & math activities Num. of Books in Setting

3.53 (1.19)abc −0.48 (0.81)abc 55.14 (130.05)abc

4.71 (1.07)ad 0.28 (0.45)ad 89.58 (129.71)aef

4.57 (1.09)be 0.23 (0.43)be 146.12 (188.61)beg

4.17 (1.20)cde 0.08 (0.49)cde 114.62 (193.14)dfg

HS Pub > Priv > Home HS Pub > Priv > Home Pub > Priv > HS > Home

Note: matched superscripts within rows indicate significant differences at the p < 0.05 level.

Start use than native whites. Cohabitation and unstable marriages were associated with lower rates of parent care versus other types of EEC. The strongest remaining results were related to maternal employment, which continued to be significantly associated with heightened home care over all other EEC arrangements, and private centers over parent care. 3.4. Predicting children’s cognitive and behavioral skills at age 5 The next set of analyses considered how EEC type at age 4 was associated with children’s cognitive and behavioral skills at age 5, when most children had just entered kindergarten. The prior results indicate that PS weighting removed most, but not all, of the significant patterns of differential selection into EEC type; hence, in predicting children’s functioning, we incorporated the trimmed PS weights and included all measured covariates assessed from infancy through age 4, in addition to children’s exposure to kindergarten by the age 5 assessments, thus adjusting for all measured differences in children, families, and community contexts. Table 4 provides results from the multivariate PSW models predicting children’s functioning at age 5, comparing each EEC type to parent care. Significant differences between EEC types are presented using matched superscripts. In these models measures of children’s cognitive and behavioral skills at age 5 were standardized; thus coefficients represent the standard deviation unit difference in skills for children in each EEC type versus parent care. Results indicate that home EEC, Head Start centers, public centers, and private centers at age 4 were all associated with significantly higher math scores at age 5 than parent care, with small to moderate effect sizes ranging from 0.11 SD units (for home versus parent care) to 0.29 SD unit differences (for private centers versus parent care). Moreover, significant differences in math scores emerged between children in private centers versus Head Start, public centers, and home EEC settings, with effect sizes ranging from 0.11 to 0.18 SDs. A similar pattern emerged for reading skills: attending Head Start, public centers, and private centers was associated with higher reading skills at age 5 in comparison to parent care, with effect sizes ranging from 0.14 to 0.28 SDs. Children who attended public and private centers outscored their peers in home care (0.15 and 0.23 SDs, respectively) in their reading skills, and children in private centers also outscored those who attended

Head Start programs (0.14 SD). Fewer differences emerged in relation to children’s expressive language skills. Children attending private centers had significantly higher language scores than children in parent care, home EEC, or public centers, with effect sizes of 0.12–0.16 SDs. The final three columns of Table 4 present results of models predicting children’s behavioral skills at age 5. Results were consistent across children’s externalizing behaviors, approaches to learning, and prosocial skills, with no significant differences emerging across EEC program type. A number of additional model specifications were estimated to check the robustness of the results. Models were estimated clustering at the state level to adjust standard errors; including only the covariates not used to create the PS weights; and incorporating both clustering and the more limited covariates. Results did not change substantially from those presented in Table 4. In a final set of models, we incorporated the set of EEC characteristics from Table 1 into the models predicting child outcomes to assess whether these variables attenuated significant associations between EEC type and children’s cognitive skills at age 5. Because EEC characteristics were, by definition, only assessed for EEC programs, children in parent care were dropped from these analyses. Results, presented in Table 5, first show the re-estimated associations between EEC type and child outcomes, weighted with PSWs, estimated only among children in EEC (with children in home care serving as the omitted group): these models simply replicate results presented in Table 4. The next panel shows results from models incorporating the EEC characteristics. Two patterns are notable. First, associations between EEC type and children’s cognitive skills diminished in size and statistical significance, with the only remaining significant differences showing that children in private centers scored higher than peers in public centers in math skills (0.16 SD units) and higher than peers in Head Start in reading skills (0.11 SD units). Differences between children in private or public centers versus home care all diminished to non-significance. Secondly, the results show very limited connections between EEC characteristics and children’s skills. The only significant pattern indicated that larger group sizes were associated with very small positive increments in math and reading skills, with a 0.01 SD increase for each additional child in the classroom. (We confirmed that multicollinearity between EEC characteristics was not a concern in these models.)

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

99

Table 2 Multinomial logistic regressions predicting selection into types of EEC, compared to parent care. Home (N ≈ 850) RRR (SE)

Head Start (N ≈ 1000) RRR (SE)

Public center (N ≈ 450) RRR (SE)

Private center (N ≈ 750) RRR (SE)

Significant differences between types

Child characteristics Bayley’s mental score Adaptive temperament Low birth weight Multiple birth Poor health Physical disability Cog/behav disability Boy Age wave 3

0.99 (0.01)a 1.06 (0.16)a 1.29 (0.19) 0.81 (0.15)a 0.74 (0.19) 1.11 (0.32) 1.15 (0.54) 1.16 (0.15) 1.01 (0.02)ab

1.01 (0.01)ab 0.77 (0.11)a 1.02 (0.15) 1.32 (0.24)a 0.76 (0.16) 0.8 (0.21) 1.68 (0.70) 1.19 (0.15) 1.04 (0.02)*c

1.00 (0.01) 0.73 (0.14) 1.19 (0.22) 0.86 (0.19) 0.89 (0.24) 0.98 (0.35) 1.78 (0.88) 1.11 (0.19) 1.11 (0.02)***acd

0.99 (0.01)b 1.04 (0.17) 1.10 (0.17) 1.18 (0.22) 0.64 (0.17) 0.94 (0.29) 1.58 (0.76) 1.00 (0.14) 1.04 (0.02)**bd

HS > Home Priv Home > HS

Family characteristics Native black Native hispanic Native other Asian Hispanic immigrant Other immigrant Non-english household Income-to-needs Less than HS degree Some college BA/BS or greater Some mat. employment Stable mat. employment Average hours emp. Sometimes welfare Stable welfare Sometimes cohabitating Stably cohabitating Sometimes married Stably married Mat age at first child Average # kids in HH Sometimes >2 adults Stably >2 adults Cognitive stimulation Sometimes mat. depress Stable mat. depression

0.99 (0.22)ab 1.35 (0.33) 1.62 (0.44) 1.26 (0.49) 0.95 (0.30)a 1.66 (0.62) 0.80 (0.22)a 1.06 (0.13) 0.92 (0.19) 0.91 (0.15) 0.77 (0.19) 3.64 (0.82)***abc 4.47 (1.71)***ab 1.03 (0.01)***a 0.96 (0.27)a 0.74 (0.21)a 0.88 (0.21) 0.32 (0.09)*** 0.49 (0.12)*** 0.35 (0.08)*** a 0.99 (0.02)ab 0.91 (0.05) 1.11 (0.16) 1.85 (0.43)**abc 0.88 (0.18) 1.04 (0.17) 1.12 (0.26)

2.17 (0.47)***ac 1.21 (0.31) 1.67 (0.45) 0.82 (0.30) 0.97 (0.32)b 3.07 (1.20)*** 1.73 (0.50)b 0.82 (0.12)a 0.62 (0.12)** 1.02 (0.15) 0.54 (0.14)*a 1.28 (0.26)a 1.76 (0.66)a 1.02 (0.01) 4.11 (1.75)***abc 4.37 (1.85)***abc 0.96 (0.22) 0.50 (0.13)** 0.60 (0.14)*a 0.52 (0.11)*** 1.00 (0.02)c 0.91 (0.05) 0.88 (0.12) 0.81 (0.18)a 1.08 (0.20) 1.02 (0.16) 1.01 (0.21)

1.81 (0.50)*bd 0.89 (0.31) 1.58 (0.57) 1.03 (0.42) 0.54 (0.22) 2.35 (1.17) 0.94 (0.32) 1.04 (0.18) 0.75 (0.21) 1.04 (0.22) 0.73 (0.24) 1.26 (0.34)b 1.69 (0.84)b 1.01 (0.01)a 1.44 (0.59)b 1.35 (0.55)b 1.09 (0.33) 0.54 (0.20) 0.78 (0.24)b 0.70 (0.20)ab 1.04 (0.02)a 0.97 (0.07)a 0.94 (0.18) 0.67 (0.23)b 0.84 (0.21) 1.10 (0.22) 0.62 (0.18)

0.90 (0.21)cd 0.85 (0.23) 1.34 (0.39) 0.85 (0.30) 0.40 (0.15)**ab 1.50 (0.58) 1.02 (0.31) 1.19 (0.14)a 0.78 (0.21) 1.09 (0.18) 1.20 (0.28)a 1.76 (0.40)**c 2.77 (1.09)** 1.02 (0.01)* 1.30 (0.35)c 0.85 (0.23)c 0.78 (0.20) 0.31 (0.10)*** 0.33 (0.08)***ab 0.35 (0.08)***b 1.04 (0.02)*bc 0.81 (0.06)***a 0.82 (0.13) 0.72 (0.19)d 0.99 (0.20) 0.86 (0.15) 0.80 (0.20)

HS Pub > Home Priv

Contextual characteristics Subsidy Elig/State Med Inc. % 4yo in State Pre-K Formal EEC competition Informal EEC competition Large urban Small urban Rural

0.83 (0.40)a 0.76 (0.32)a 1.31 (0.29) 0.83 (0.12)a 0.72 (0.13) 0.64 (0.13)* 0.95 (0.19)a

1.32 (0.59)b 1.54 (0.61)bc 1.09 (0.22) 1.07 (0.14) 0.97 (0.17) 0.88 (0.17) 1.96 (0.36)***abc

3.93 (2.16)**abc 14.21 (6.43)***abd 1.01 (0.26) 1.21 (0.20)a 0.70 (0.17) 0.74 (0.18) 1.19 (0.29)b

1.22 (0.61)c 0.51 (0.23)cd 0.94 (0.24) 1.13 (0.17) 0.71 (0.14) 0.91 (0.18) 0.85 (0.17)c

Pub > Home HS Priv Pub > HS > Priv; Pub > Home

HS > Home

Pub > HS Priv; Priv > Home

Home HS > Priv HS > Home Priv > HS

Priv > HS Home > HS Pub Priv Home > HS Pub Home > Pub HS > Home Pub Priv HS > Home Pub Priv

HS, Pub > Priv Pub > Home Priv Pub > Home; Priv > Home HS Pub > Priv Home > HS Pub Priv

Pub > Home

HS > Home Pub Priv

Note: sampling weight w3r0 used. Matched superscripts within rows indicate significant differences between selection into different child care types at the p < 0.05 level. Omitted groups include White, high school, never employed, never welfare, stably single, never additional adults, never depression, and suburban.

4. Discussion Recent years have seen notable efforts by federal, state, and local governments to expand early childhood education opportunities as a mechanism for stemming the intergenerational transmission of inequality and preparing children for formal schooling (Barnett & Carolan, 2013). To help inform such efforts, it is essential to understand whether all types of early childhood arrangements function similarly, or whether different types of EEC programs are more or less effective in promoting the skills of low-income children. The goals of this study were to paint a rich descriptive portrait comparing numerous types of EEC arrangements accessed by preschool-aged children in low-income families, and, using quasiexperimental methods, to delineate whether distinct EEC types were most effective at promoting children’s cognitive and behavioral school readiness skills. As the first study to use provider reports of EEC type to delineate home care settings from Head Start, public centers, and private centers, this provides unique infor-

mation comparing and contrasting diverse EEC arrangements in a national sample of low-income children. When considering how to increase the use of EEC programs that best support children’s school readiness skills, it is necessary to keep in mind selection processes by which children access different types of programs, which range from parental resources and preferences to contextual opportunities and constraints. An important literature is emerging which seeks to conceptualize and empirically map such selection processes (Chaudry et al., 2010; Coley et al., 2014; Weber, 2011). The current study expanded this work by focusing exclusively on low-income children and distinguishing between various types of centers. Replicating prior work (Coley et al., 2014; Fuller et al., 2004; Weber, 2011), we found that children in African American and immigrant families, rural families, and families with lower parental income, education, maternal age, and higher welfare receipt were more likely to be in Head Start programs than private centers or parent care, whereas greater maternal employment was associated with a heightened likeli-

100

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

Table 3 Propensity score weighted multinomial logistic regressions predicting selection into types of EEC, compared to parent care. Home (N ≈ 850) RRR (SE)

Head Start (N ≈ 1000) RRR (SE)

Public center (N ≈ 450) RRR (SE)

Private center (N ≈ 750) RRR (SE)

Child characteristics Bayley’s mental score Adaptive temperament Low birth weight Multiple birth Poor health Physical disability Cog/behav disability Boy Age wave 3

1.00 (0.01) 0.85 (0.11) 0.88 (0.12) 1.13 (0.20) 1.25 (0.24) 0.90 (0.21) 0.83 (0.28) 0.93 (0.10) 0.97 (0.01)**a

1.01 (0.01) 0.85 (0.10) 0.96 (0.12) 1.11 (0.18) 1.23 (0.22) 0.84 (0.17) 0.92 (0.28) 0.87 (0.09) 0.99 (0.01)a

0.99 (0.01) 0.93 (0.14) 1.05 (0.18) 1.13 (0.24) 1.30 (0.29) 0.85 (0.25) 1.09 (0.43) 0.99 (0.14) 0.98 (0.02)

1.01 (0.01) 0.76 (0.10)* 0.97 (0.14) 1.34 (0.24) 1.38 (0.29) 0.92 (0.22) 0.89 (0.31) 0.97 (0.12) 0.97 (0.01)

Family characteristics Native black Native hispanic Native other Asian Hispanic immigrant Other immigrant Non-english household Income-to-needs Less than HS degree Some college BA/BS or greater Some mat. employment Stable mat. employment Average hours emp. Sometimes welfare Stable welfare Sometimes cohabitating Stably cohabitating Sometimes married Stably married Mat age at first child Average # kids in HH Sometimes >2 adults Stably >2 adults Cognitive stimulation Sometimes mat. depression Stable mat. depression

0.75 (0.14)a 0.85 (0.17) 0.69 (0.13)a 0.60 (0.18)a 0.78 (0.21) 0.96 (0.30) 1.03 (0.24) 0.85 (0.09) 1.13 (0.19) 1.26 (0.17) 0.92 (0.19) 2.24 (0.42)***ab 3.57 (1.13)***abc 1.01 (0.01) 0.91 (0.21) 1.08 (0.25) 1.43 (0.27) 0.91 (0.20) 1.73 (0.33)*** 1.29 (0.22)a 1.00 (0.02) 0.95 (0.05) 0.84 (0.10) 1.21 (0.22) 0.91 (0.15) 0.93 (0.13) 0.93 (0.16)

1.10 (0.20)a 1.09 (0.23) 1.11 (0.21)abc 1.30 (0.38)a 1.03 (0.26) 0.78 (0.24) 0.86 (0.19) 0.99 (0.12) 1.29 (0.20) 1.17 (0.15) 1.11 (0.23) 1.00 (0.17)a 1.77 (0.58)a 1.00 (0.01) 1.49 (0.51) 1.87 (0.64) 1.03 (0.19) 1.30 (0.27) 1.60 (0.32)* 1.32 (0.23) 1.01 (0.01) 0.94 (0.04) 0.91 (0.11) 0.91 (0.18) 0.72 (0.12) 0.85 (0.11) 0.81 (0.13)

0.86 (0.19) 0.83 (0.23) 0.52 (0.15)b 1.10 (0.41) 0.87 (0.31) 0.62 (0.25)* 1.04 (0.31) 0.80 (0.12) 1.06 (0.24) 1.04 (0.18) 0.87 (0.23) 0.88 (0.20)bc 1.17 (0.47)b 1.01 (0.01) 0.98 (0.31) 0.88 (0.29) 1.34 (0.34) 1.60 (0.48)*** 2.46 (0.64)* 1.66 (0.39) 1.00 (0.02) 0.92 (0.06) 0.84 (0.13) 0.97 (0.24) 0.88 (0.18) 1.09 (0.18) 0.77 (0.20)

1.03 (0.21) 1.13 (0.25) 0.71 (0.15)c 0.83 (0.24) 0.86 (0.25) 0.76 (0.24) 1.19 (0.28) 0.94 (0.09) 1.29 (0.28) 1.36 (0.19)* 1.22 (0.24) 1.46 (0.28)*c 2.52 (0.86)**c 1.00 (0.01) 1.07 (0.23) 1.14 (0.25) 1.00 (0.22) 1.38 (0.35) 1.73 (0.39)* 1.32 (0.25)b 1.00 (0.01) 0.95 (0.05) 0.84 (0.11) 0.77 (0.17) 1.00 (0.18) 0.88 (0.12) 0.83 (0.17)

Contextual characteristics Subsidy Elig/State Median Inc. % 4yo in State Pre-K Formal EEC competition Informal EEC competition Large urban Small urban Rural

1.30 (0.50) 1.27 (0.46) 0.81 (0.14) 1.04 (0.13) 1.26 (0.19)a 1.12 (0.18) 0.97 (0.15)

1.20 (0.43) 1.00 (0.32) 0.91 (0.15) 1.05 (0.12) 0.91 (0.13)b 1.03 (0.16) 0.92 (0.14)

1.30 (0.63) 1.31 (0.49) 0.73 (0.16) 1.20 (0.16) 1.22 (0.25) 1.29 (0.28) 1.23 (0.25)

1.47 (0.62) 2.04 (0.76) 0.95 (0.20) 0.96 (0.13) 1.15 (0.19) 0.97 (0.16) 0.98 (0.17)

Significant differences between types

HS > Home HS > Home HS > Home Pub Priv HS > Home

Home > HS Pub; Priv>Pub Home > HS Pub Priv

Priv > Home

Home > HS

Note: propensity score weight used. Matched superscripts within rows indicate significant differences at the p < 0.05 level. Omitted groups include white, high school, never employed, never welfare, stably single, never additional adults, never depression, and suburban.

Table 4 Propensity Score Weighted Multivariate Predicting Low-Income Children’s Skills at Age 5. Mathematics

EEC type Home Head Start Public center Private center Significant differences

Intercept R-squared range F-score range

Reading

Expressive language

Externalizing behavior

Approaches to learning

Prosocial skills

B

SE

B

SE

B

SE

B

SE

B

SE

B

SE

0.11*a 0.17**b 0.12*c 0.29**abc Priv > Home HS Priv

(0.04) (0.04) (0.05) (0.05)

0.05ab 0.14**c 0.20**a 0.28***bc Priv > Home HS; Pub>Home −0.07 0.40–0.43 47.52–63.38

(0.05) (0.04) (0.06) (0.05)

0.02a 0.04 −0.02b 0.14**ab Priv > Home Pub

(0.05) (0.05) (0.07) (0.05)

0.07 0.09 0.01 0.01

(0.06) (0.07) (0.08) (0.07)

-0.01 0.03 -0.03 0.06

(0.06) (0.05) (0.08) (0.07)

0.05 0.03 0.01 0.04

(0.06) (0.06) (0.08) (0.06)

(0.11)

0.35* (0.14) 0.24–0.27 17.44–21.01

−0.07 (0.11) 0.41–0.45 40.85–56.94

−0.34 (0.18) 0.12–0.15 8.30–10.41

0.70*** (0.16) 0.16–0.19 12.35–17.21

0.40* (0.17) 0.11–0.14 6.98–9.15

Notes: N ≈ 4250 ***p < 0.001. **p < 0.01. *p < 0.05. t < 0.10. EEC types were compared to the omitted group of parent care. Matched superscripts within columns indicate significant differences at the p < 0.05 level between EEC types.

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

101

Table 5 Propensity regression characteristics. Mathematics

B

Reading

SE

Expressive language

Externalizing behavior

Approaches to learning

Prosocial skills

B

SE

B

SE

B

SE

B

SE

B

SE

Model 1: without EEC characteristics Head Start 0.06a (0.05) Public center 0.00b (0.06) ab 0.18*** (0.05) Private center Priv > HS Pub Significant differences

0.09a 0.16** 0.24***a Priv > HS

(0.05) (0.06) (0.05)

0.02 −0.03a 0.12*a Priv > Pub

(0.06) (0.07) (0.06)

0.02 −0.06 −0.06

(0.07) (0.08) (0.06)

0.05 −0.01 0.07

(0.05) (0.08) (0.06)

−0.02 −0.04 −0.01

(0.06) (0.08) (0.06)

Model 2: with EEC characteristics Head Start 0.00 Public center −0.05a Private center 0.11a

(0.07) (0.07) (0.07)

−0.00a −0.05 0.11a

(0.07) (0.07) (0.07)

−0.02 −0.07 0.05

(0.09) (0.09) (0.08)

0.06 −0.04 −0.04

(0.09) (0.11) (0.08)

0.00 −0.03 0.05

(0.08) (0.10) (0.08)

−0.09 −0.06 −0.03

(0.09) (0.11) (0.08)

Individual characteristics Hours/week in EEC(100’s) EEC cost/month (1000s)

0.00 0.01

(0.01) (0.01)

0.00 0.01

(0.01) (0.01)

−0.01 0.02

(0.02) (0.01)

0. 04 0.02

(0.02) (0.02)

−0.04 −0.01

(0.02) (0.02)

−0.02 −0.02

(0.02) (0.02)

Structural & process quality BA or greater Early childhood degree 1–14 Training hours 15+ Training hours Years as an EEC provider (10s) Group size Child: adult ratio Num. non-english Reimbursed for meals FDCRS/ECERS total score Literacy & math activities Num. of books (100s) Significant differences Model 2 intercept Model 2 R-squared range Model 2 F-score range

−0.01 −0.02 0.02 −0.06 0.02 0.01* −0.01 0.01 −0.01 −0.02 0.02 0.00 Priv > Pub 0.08 0.41–0.45 26.21–35.93

(0.05) (0.06) (0.06) (0.06) (0.02) (0.00) (0.01) (0.00) (0.05) (0.03) (0.03) (0.01)

−0.01 (0.05) −0.02 (0.06) 0.02 (0.06) −0.06 (0.06) 0.02 (0.02) 0.01* (0.00) −0.01 (0.01) 0.01 (0.00) −0.01 (0.05) −0.02 (0.03) 0.02 (0.03) 0.00 (0.01) Priv > Home 0.08 (0.17) 0.39–0.44 31.11–39.07

−0.01 −0.05 0.01 0.05 0.00 0.01 0.00 0.00 0.00 −0.02 0.01 0. 00

(0.06) (0.06) (0.07) (0.07) (0.02) (0.01) (0.01) (0.01) (0.06) (0.03) (0.04) (0.02)

−0.01 0.00 −0.03 0.02 −0.02 0.00 0.02 0.01 0.04 −0.03 −0.02 0.01

(0.07) (0.06) (0.09) (0.08) (0.03) (0.01) (0.01) (0.01) (0.06) (0.04) (0.05) (0.02)

−0.03 −0.01 0.04 −0.05 0.01 0.00 −0.01 −0.01 −0.01 0.04 0.06 −0. 02

(0.07) (0.06) (0.09) (0.09) (0.03) (0.01) (0.01) (0.01) (0.06) (0.04) (0.04) (0.02)

−0.12 0.00 0.00 −0.09 −0.01 0.00 0.01 0.00 0.09 0.03 0.03 0. 00

(0.08) (0.07) (0.08) (0.08) (0.03) (0.01) (0.01) (0.01) (0.07) (0.03) (0.04) (0.02)

(0.17)

0.40* (0.20) 0.26–0.29 11.16–13.26

−0.32 (0.23) 0.13–0.16 6.03–8.07

0.69 (0.21) 0.17–0.21 8.81–12.50

0.38 (0.22) 0.12–0.15 5.12–6.88

Notes: N ≈ 3100. ***p < 0.001. **p < 0.01. *p < .05. t < 0.10. EEC types were compared to the omitted group of home care. Matched superscripts within columns indicate significant differences at the p < 0.05 level between EEC types.

hood of using home EEC arrangements, and maternal marriage and cohabitation were associated with more parent care versus home or center use. Factors associated with a heightened likelihood of public centers included family welfare receipt, higher child age, and the availability of more generous state subsidies and public preschool programs; indeed, these were the only factors distinguishing use of public centers versus Head Start settings. Although a deep analysis and interpretation of selection results are deserving of richer attention than can be provided here, it is important to reiterate that even within this sample of children from low-income families (those with incomes less than two-times the federal poverty line, or approximately $38,700 for a family of four in 2005 dollars), families whose children attended private center EEC programs were generally more socioeconomically advantaged and families whose children attended Head Start were generally the most disadvantaged. These patterns likely to some degree reflect the eligibility requirements low-income families must meet to access diverse EEC options. For example, with the exception of small set-asides, families must have income below the poverty level to qualify for Head Start. On the other hand, in order for families to access state child care subsidies (often used for private centers), parents must be low-income and frequently must provide proof of employment. These patterns might also reflect trends in declining federal funds for Head Start, Temporary Assistance to Needy Families (TANF), and the Child Care and Development Block Grant (CCDBG), which are limiting supply and may be forcing state and community organizations to prioritize the most disadvantaged children for publicly-funded programs (Schulman & Blank, 2013). Federal funds through CCDBG and TANF funding, two of the primary mechanisms providing EEC subsidies for low-income families, peaked in 2001 and then declined (Schulman & Blank, 2013). Head Start funding

similarly has not kept up with demand, with Head Start programs currently serving only 42% of eligible three- and four-year-olds, down from 60% in 2000 (Currie & Neidell, 2007; Magnuson & Shager, 2010; Schmit, Matthews, Smith, & Robbins, 2013). Perhaps seeking to fill the gap between demand and supply, state and city early education programs have grown dramatically in the past decade (Barnett et al., 2015; National Institute for Early Education Research (NIEER), 2006; Samuels, 2014). Indeed, our results highlight the central role of state and local EEC supply, as children living in states with higher subsidy limits and public preschool slots were notably more likely than their peers to attend public centers. Together, these patterns suggest two central lessons: First, the importance for policy makers and educational leaders to carefully consider the needs and constraints of families in the targeting of public and private resources supporting access to EEC; and second, the importance of adjusting for differences in children, families, and contexts when seeking to delineate associations between different types of EEC arrangements and children’s development. Turning to EEC programs themselves, our results highlight notable discrepancies in quality characteristics across EEC type. Interestingly, patterns in EEC program characteristics showed opposing patterns to those for child and family characteristics, with public center teachers reporting the highest educational credentials, and Head Start and public centers showing the highest levels of global quality, learning activities, and ongoing professional training. The high levels of education and training of public EEC teachers is consistent with other research (Bellm et al., 2002; Clifford et al., 2005; Pianta et al., 2005; Smith et al., 2003) and may reflect the more rigorous educational credentials of these programs which often provide pay scales and supports similar to what elementary school teachers receive. Additional research is needed to assess

102

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

whether these characteristics translate into less turnover, better morale, and better working and learning conditions for teachers and children in public preschool programs. It is important to note that other measures of structural quality showed opposing patterns, with Head Start and public centers also having the largest group size and ratios, and serving the most disadvantaged and ELL students. Replicating prior research (Li-Grining & Coley, 2006), providers in home care settings reported the lowest levels of global program quality, learning activities, and provider education and training. Private centers showed moderate levels of both teacher education and training as well as global quality and learning activities but the highest cost for families. In short, our descriptive results paint a picture of notable disparities in the EEC programs attended by low-income children. Within this national sample of children from low-income families, the most advantaged children attended private EEC centers. And yet these programs provided middling levels of quality, with many measures indicating significantly lower quality than public and Head Start programs. The latter programs, in contrast, served the most disadvantaged children, providing these children with highly educated and trained teachers and with higher global quality and more consistent learning opportunities. What do these differences mean for children’s preparation for formal schooling? In prospective models adjusting for a broad range of child, family, and community characteristics, and incorporating propensity score weighting and child lags to help further control for selection biases, our results found, overall, that all three types of EEC centers predicted enhanced math and reading skills in children in comparison to not attending EEC of any type. Despite major differences in features of EEC settings, a diverse array of center-based preschool programs predicted increased academic skills for low-income children. With one exception, these benefits did not emerge from the use of home-based EEC programs. This pattern of results supports a broad array of policy initiatives that buttress low-income children’s access to center-based programs, including publicly supported programs, federal Head Start programs, and public childcare subsidies that can be used in private programs. Yet beyond this overarching pattern, our results further highlight that private EEC centers, which showed middling levels of quality but served more advantaged children, showed the most consistent and largest effect sizes for increased reading, math, and language skills among children, predicting significantly higher math scores than home, public center, or Head Start programs, significantly higher reading scores than home and Head Start programs, and significantly higher expressive language skills than home and public EEC programs. What might explain these results? Although we incorporated quasi-experimental analytic methods, our data were nonetheless correlational, and it is possible that models did not completely adjust for private preschool attendees’ more advantaged home contexts or for other unmeasured differences between the groups. Our selection models and covariates did not adjust for parental attitudes regarding early education, family educational engagement, or other important dimensions of parenting, such as warmth and responsivity, for example. Moreover, the models including quality features of EEC settings suggest that the measures assessed in this work, which parallel those used in state QRIS systems and a host of prior research, have very limited predictive validity in terms of children’s core cognitive and behavioral skills. This result is not new: analyses of these measures in other studies with both the ECLS-B and a host of other datasets have reported similar null results (although we note that some analyses of differential coding of ECERS-R scores, as well as work with other measures such as the CLASS, show somewhat more promising results in terms of predicting child outcomes, Burchinal, Vandergrift, Pianta, & Mashburn, 2009; Sabol et al., 2013), suggesting the need for greatly enhanced attention to mea-

surement development and work delineating the most important practices and characteristics of EEC supporting children’s development. Interestingly, only larger group size significantly predicted children’s enhanced cognitive skills, with small group size perhaps serving as a proxy for small, informal home care arrangements with limited cognitive enrichment in comparison to more structured larger center programs using validated curricula. Another possible explanation for our patterns of results is that children’s functioning in these programs is affected not only by structural and process quality features of the EEC programs, but also by the functioning of other children. If private preschool programs are filled with more advantaged and higher functioning children than are Head Start and public preschool programs, then the dynamics of the classroom, peer influences, and teachers’ ability to focus more time on learning activities or individual needs may enhance children’s learning in such programs. Unfortunately, the ECLS-B data did not allow us to more richly assess the composition of children within EEC classrooms or how this composition may affect classroom dynamics and processes. However, a small emerging body of research suggests that the average level of language or cognitive skills across children in EEC classrooms is predictive of differential growth in individual children’s language or cognitive skills through the preschool year (Coley, Kull, & Cook, under review; Henry & Rickman, 2007; Justice, Logan, Tzu-Jung, & Kaderavek, 2014; Mashburn, Justice, Downer, & Pianta, 2009). Other studies have found that children in classrooms with greater numbers of poor children experience less cognitive growth through the year (Reid & Ready, 2013; Weiland & Yoshikawa, 2014). Further research is needed in this arena to explore the potential role of classroom composition and peer functioning in enhancing or limiting how children benefit from EEC. In contrast to the results unearthed in this study related to children’s cognitive skills, it is important to highlight the different pattern in relation to behavioral skills. Across diverse measures of behavioral functioning reported by teachers including externalizing problems, learning behaviors, and prosocial skills, no significant differences emerged across children attending different types of EEC programs or parent care. This is in contrast to some prior research which has found negative effects of center-based EEC on children’s externalizing and learning-related behavioral skills (Coley et al., 2013; Loeb et al., 2007). Such work has argued that center EEC beginning in infancy or at higher intensity may be more predictive of negative behavioral outcomes (Coley et al., 2013), which may help to explain our null results, as we only assessed EEC at age 4. Another possibility is that center-based EEC settings are not predictive of negative behavioral outcomes for children from economically disadvantaged homes—an argument which concurs with results from experimental and quasi-experimental research on public EEC programs (Gormley & Gayer, 2005; U.S. Department of Health and Human Services, Administration for Children and Families, 2010) and with results from observational studies focused on low-income samples (Votruba-Drzal, Coley, & ˜ Chase-Lansdale, 2004; Votruba-Drzal, Coley, Maldonado-Carreno, Li-Grining, & Chase-Lansdale, 2010). Although it is encouraging that negative effects of center EEC programs on children’s behavioral skills did not emerge, these results highlight the need to explore practices and curricula that support improvements in young children’s emotional and behavioral development. In closing, it is important to consider the limitations of this research. As noted above, these include the ever-present concerns related to correlational data and the risks of selection bias affecting results; although our propensity score weighting technique lowered these risks, it did not eliminate them, and the data remain correlational. Other limitations include the following: although having EEC provider rather than parent reports of EEC type presents a substantial improvement over past literature,

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

we must nonetheless remain somewhat cautious concerning the EEC categories. There are a broad diversity of ECE arrangements with publicly-supported programs run at the local and state level, and ever-more-complex EEC funding and program models, with many programs incorporating mixed-funding streams into single centers or classrooms (New America Foundation, 2014), making a clear delineation of program type challenging. In particular, it is likely that children enrolled in public prekindergarten programs or using state child care subsidies were misclassified into the private center category in our EEC classification. It is also important to highlight the broad diversity of standards and quality both between but also within program type (Sabol et al., 2013), leading to caution in clumping diverse programs into monlithic type categories. Other measures in this study were also limited, particularly those assessing program quality. As noted above, the ECERS-R and FDCERS assessments have come under increased scrutiny as measures of global program quality (Gordon et al., 2013; Sabol et al., 2013; Votruba-Drzal et al., 2013; Weiland et al., 2013), and the measures of learning activities used in this study addressed only the frequency rather than the quality of such activities, and may suffer from reporter bias as they derived from teacher reports. Finally, it is important to note that the data related to EEC were collected predominantly in 2005, and the landscape of EEC in the US has changed notably since this time. Beyond these limitations, this paper adds to the literature by contrasting four different types of EEC and parent care using more nuanced categories and valid reporting than in past research. Overall, we found that compared to parent only care, all center-based EEC program types appear to promote the academic skills of lowincome children, with the highest functioning seen among children attending private centers. This finding has important policy implications as federal, state and local governments make decisions about resource allocation and eligibility requirements for families to access EEC. Although there were notable differences in quality across different types of EEC programs, these quality features were not significantly associated with children’s skills, replicating a growing base of research indicating that quality features typically regulated and supported through state QRIS systems are not strongly predictive of enhanced learning and development among children (Sabol et al., 2013). This discrepancy contributes both to ongoing concerns about the validity of program quality indicators and to the need for additional research delineating the central features of EEC that best promote children’s learning and development. While these issues remain essential questions in the literature, results from this study reiterate the importance of increasing access to center-based EEC for low-income children, and weaken the arguments that EEC should only be delivered in certain settings. As federal, state, and local policy makers continue to alter the funding and regulatory contexts for early education, considerations for ensuring low-income children have continued access to developmentally-promotive EEC programs is particularly important. Acknowledgements We gratefully acknowledge the support of the Spencer Foundation (# 201300115). The content is solely the responsibility of the authors and does not necessarily represent the official views of the grantors. A special thank you is also extended to the children and families that participated in the Early Childhood Longitudinal Study, Birth Cohort. References Assel, M. A., Landry, S. H., Swank, P. R., & Gunnewig, S. (2007). An evaluation of curriculum, setting, and mentoring on the performance of children enrolled in

103

pre-kindergarten. Reading and Writing, 20, 463–494. http://dx.doi.org/10.1007/ s11145-006-9039-5 Barnett, W. S., & Carolan, M. E. (2013). Trends in state funded preschool programs: survey findings from 2001–2002 to 2011–2012. New Brunswick, NJ: National Institute for Early Education Research and the Center for Enhancing Early Learning Outcomes. http://nieer.org/publications/policy-reports/trends-statefunded-preschool-programs-survey-findings-2001-2002-2011 Barnett, W. S., Carolan, M. E., Squires, J. H., Clarke Brown, K., & Horowitz, M. (2015). The state of preschool 2014: state preschool yearbook. New Brunswick, NJ: National Institute for Early Education Research. Barnett, W. S., Hustedt, J. T., Hawkinson, L. E., & Robin, K. B. (2006). The state of preschool 2006: state preschool yearbook. New Brunswick, NJ: National Institute for Early Education Research. Barnett, W. S., Hustedt, J. T., Robin, K. B., & Schulman, K. L. (2005). The state of preschool 2005: state preschool yearbook. New Brunswick, NJ: National Institute for Early Education Research. Bayley, N. (1993). Bayley scales of infant development (Second ed.). San Antonio, TX: The Psychological Corporation. Bellm, D., Burton, A., Whitebook, M., Broatch, L., & Young, M. (2002). Inside the pre-K classroom: a study of staffing and stability in state-funded prekindergarten programs. Washington, DC: Center for the Child Care Workforce. Burchinal, M., Vandergrift, N., Pianta, R., & Mashburn, A. (2009). Threshold analysis of association between child care quality and child outcomes for low-income children in pre-kindergarten programs. Early Childhood Research Quarterly, 25, 166–176. http://dx.doi.org/10.1016/j.ecresq.2009.10.004 Caldwell, B. M., & Bradley, R. H. (1979). Home observation for measurement of the environment. Little Rock, AR: Center for Child Development and Education, University of Arkansas. Caldwell, B. M., & Bradley, R. H. (2001). HOME inventory and administration manual (3rd ed.). University of Arkansas for Medical Sciences and University of Arkansas at Little Rock. Chaudry, A., Henly, A., & Meyers, M. (2010). Conceptual frameworks for child care decision-making. Washington, D.C: Office of Planning, Research, and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services. Chernoff, J. J., Flanagan, K. D., McPhee, C., & Park, J. (2007). Preschool: first findings from the third follow-up of the early childhood longitudinal study, birth cohort (ECLS-B). (NCES 2008-025). Washington, D.C: National Center for Education Statistics. Clements, D. H., & Sarama, J. (2007). Early childhood mathematics learning. In F. K. Lester (Ed.), Second handbook of research on mathematics teaching and learning (pp. 461–555). Charlotte, NC: Information Age. Clifford, R. M., Barbarin, O., Chang, F., Early, D. M., Bryant, D., Howes, C., et al. (2005). What is pre-kindergarten? Characteristics of public pre-kindergarten programs. Applied Developmental Science, 9(3), 126–143. http://dx.doi.org/10. 1207/s1532480xads0903 1 Coley, R. L., Kull, M., & Cook, K. D. (2015). Peer effects on children’s cognitive skills and teachers’ instructional quality in state-funded preschool programs. Boston College. Unpublished manuscript, under review. Coley, R. L., Li-Grining, C., & Chase-Lansdale, P. L. (2006). Low-income families’ child care experiences: meeting the needs of children and families. In N. Cabrera, R. Hutchins, & E. Peters (Eds.), From welfare to child care: what happens to children when mothers exchange welfare for work (pp. 149–170). Mahway, NJ: Lawrence Erlbaum Associates. Coley, R. L., Votruba-Drzal, E., Collins, M. A., & Miller, P. (2014). Selection into early education and care settings: differences by developmental period. Early Childhood Research Quarterly, 29(3), 319–332. http://dx.doi.org/10.1016/j. ecresq.2014.03.006 Coley, R. L., Votruba-Drzal, E., Miller, P. L., & Koury, A. (2013). Timing, extent, and type of child care and children’s behavioral functioning in kindergarten. Developmental Psychology, 49, 1859–1873. http://dx.doi.org/10.1037/a0031251 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, 83–99. http://dx.doi.org/10.1016/j.econedurev.2005.03.004 Datta, A. R. (2013). Say what you mean and mean what you say: Terminology agnosticism in child care questionnaires. Methodological brief OPRE 2013-33. Washington, D.C: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services. DeGangi, G. A., Poisson, S., Sickel, R. Z., & Wiener, A. S. (1995). The infant–toddler symptom checklist. Tucson, AZ: Therapy Skill Builders. Duncan, S. E., & DeAvila, E. A. (1998). PreLAS 2000, A test of language development for four-, five-, and six-year olds, Form C. Monterey, CA: CTB/McGraw-Hill. Dunn, L. M., & Dunn, L. M. (1997). Examiner’s manual for the PPVT-III: peabody picture vocabulary test (3rd ed.). Circle Pines, MN: American Guidance Service. Fantuzzo, J. W., Gadsden, V. L., & McDermott, P. A. (2011). An integrated curriculum to improve mathematics, language, and literacy for Head Start children. American Educational Research Journal, 48(3), 763–793. http://dx.doi.org/10. 3102/0002831210385446 Feller, A., Grindal, T., Miratrix, L. W., & Page, L. (2014). Compared to what? Variation in the impact of early childhood education by alternative care-type settings. Social Science Research Network. http://dx.doi.org/10.2139/ssrn.2534811 Fuller, B., Kagan, S. L., Loeb, S., & Chang, Y. W. (2004). Child care quality: centers and home settings that serve poor families. Early Childhood Research Quarterly, 19(4), 505–527. http://dx.doi.org/10.1016/j.ecresq.2004.10.006 Ginsburg, H. P., & Baroody, A. J. (2003). TEMA-3: test of early mathematics ability (3rd ed.). Austin, TX: Pro-Ed.

104

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105

Gordon, R. A., & Chase-Lansdale, P. L. (2001). Availability of child care in the United States: a description and analysis of data sources. Demography, 38, 299–316. http://dx.doi.org/10.1353/dem.2001.0016 Gordon, R. A., Fujimoto, K., Kaestner, R., Korenman, S., & Abner, K. (2013). An assessment of the validity of the ECERS-R with implications for measures of child care quality and relations to child development. Developmental Psychology, 49(1), 146–160. http://dx.doi.org/10.1037/a0027899 Gormley, W. T., & Gayer, T. (2005). Promoting school readiness in Oklahoma: an evaluation of Tulsa’s pre-K program. Journal of Human Resources, 40, 533–558. http://dx.doi.org/10.1037/0012-1649.41.6.872 Gormley, W. T., Gayer, T., Phillips, D., & Dawson, B. (2005). The effects of universal pre-K on cognitive development. Developmental Psychology, 41, 872–884. http://dx.doi.org/10.1037/0012-1649.41.6.872 Gormley, W., Phillips, D., Newmark, K., Welti, K., & Adelstein, S. (2011). Social-emotional effects of early childhood education programs in Tulsa. Child Development, 82, 2095–2109. http://dx.doi.org/10.1111/j.1467-8624.2011. 01648 Gresham, F. M., & Elliott, S. N. (1990). The social skills rating system. Circle Pines, MN: American Guidance Service. Harms, T., Clifford, R., & Cryer, D. (1998). Early childhood environment rating scale (Rev. ed.). New York: Teachers College. Harms, T., & Clifford, R. M. (1989). The family day care rating scale. New York: Teachers College. Henry, G. T., & Rickman, D. K. (2007). Do peers influence children’s skill development in preschool? Economics of Education Review, 26, 100–112. http:// dx.doi.org/10.1016/j.econedurev.2005.09.006 Hulsey, L. K., Aikens, N., Kopack, A., West, J., Moiduddin, E., & Tarullo, L. (2011). Head Start children, families, and programs: Present and past data from FACES. OPRE Report 2011-33a. Washington, D.C: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services. Imbens, G. W. (2000). The role of the propensity score in estimating dose–response functions. Biometrika, 87(3), 706–710. http://dx.doi.org/10.1093/biomet/87.3. 706 Justice, L. M., Logan, J. A., Lin, T. J., & Kaderavek, J. N. (2014). Peer effects in early childhood education testing the assumptions of special-education inclusion. Psychological Science, 25(9), 1722–1729. http://dx.doi.org/10.1177/ 0956797614538978 Lee, R., Zhai, F., Brooks-Gunn, J., Han, W. J., & Waldfogel, J. (2014). Head Start participation and school readiness: evidence from the early childhood longitudinal study-birth cohort. Developmental Psychology, 50(1), 202–215. http://dx.doi.org/10.1037/a0032280 Leon, A. C., & Hedeker, D. (2005). A mixed-effects propensity adjustment for effectiveness analyses of ordered categorical doses. Statistics in Medicine, 24, 647–658. Li-Grining, C. P., & Coley, R. L. (2006). Child care experiences in low-income communities: developmental quality and maternal views. Early Childhood Research Quarterly, 21, 125–141. Little, R. J. A., & Rubin, D. B. (2002). Statistical analysis with missing data (Second ed.). Hoboken, New Jersey: Whiley-InterScience. Loeb, S., Bridges, M., Bassok, D., Fuller, B., & Rumberger, R. W. (2007). How much is too much? The influence of preschool centers on children’s social and cognitive development. Economics of Education Review, 26, 52–66. http://dx. doi.org/10.1016/j.econedurev.2005.11.005 Lonigan, C., Farver, J., Phillips, B., & Clancy-Menchetti, J. (2011). Promoting the development of preschool children’s emergent literacy skills: a randomized evaluation of a literacy-focused curriculum and two professional development settings. Reading and Writing, 24(3), 305–337. http://dx.doi.org/10.1007/ s11145-009-9214-6 Lonigan, C., Wagner, R., Torgeson, J., & Rashotte, C. (2002). Preschool comprehensive test of phonological & print processing. Department of Psychology, Florida State University. Magnuson, K. A., Meyers, M., Ruhm, C., & Waldfogel, J. (2004). Inequality in preschool education and school readiness. American Educational Research Journal, 41, 115–157. http://dx.doi.org/10.3102/00028312041001115 Magnuson, K., Ruhm, C., & Waldfogel, J. (2007). Does prekindergarten improve school preparation and performance? Economics of Education Review, 26, 33–51. http://dx.doi.org/10.3386/w10452 Magnuson, K., & Shager, H. (2010). Early education: progress and promise for children from low-income families. Children and Youth Services Review, 32, 1189–1198. http://dx.doi.org/10.1016/j.childyouth.2010.03.006 Magnuson, K., Waldfogel, J., & Washbrook, E. (2012). The development of SES gradients in skills during the school years: evidence from the United States and England. In J. Ermisch, M. Jantti, & T. Smeeding (Eds.), From Parents to Children: The Intergenerational Transmission of Advantage. New York: Russell Sage. Mamedova, S., & Redford, J. (2015). Early childhood program participation, From the National Household Education Surveys Program of 2012 (NCES 2013-029 REV). Washington, D.C: National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Retrieved from http://nces.ed.gov/pubsearch accessed June 5, 2015. Mashburn, A. J., Justice, L. M., Downer, J. T., & Pianta, R. C. (2009). Peer effects on children’s language achievement during pre-kindergarten. Child Development, 80(3), 686–702. http://dx.doi.org/10.1111/j.1467-8624.2009.01291.x Merrell, K. W. (2003). Preschool and kindergarten behavior scales (Second ed.). Austin, TX: PRO-ED.

National Institute for Early Education Research (NIEER). (2006). Why cities are making preschool education available to all children. Preschool Matters, 4(6) http://nieer.org/publications/why-cities-are-making-preschool-educationavailable-all-children New America Foundation. (2014). Pre-k funding from state and federal sources. New America Foundation. Downloaded from http://febp.newamerica.net/ background-analysis/pre-k-funding Accessed 23.06.15 Pearl, J. (2009). Letter to the editor: remarks on the method of propensity score. Statistics in Medicine, 28, 1415–1416. Pianta, R., Howes, C., Burchinal, M., Bryant, D., Clifford, R., Early, D., et al. (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. http://dx.doi.org/10.1207/ s1532480xads0903 2 Radloff, L. S. (1977). The CES-D scale: a self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. Reardon, S. F. (2011). The widening academic achievement gap between the rich and the poor: new evidence and possible explanations. In R. Murnane, & G. Duncan (Eds.), Whither opportunity? Rising inequality and the uncertain life chances of low-income children (pp. 91–116). New York, NY: Russell Sage Foundation. Reid, J. L., & Ready, D. D. (2013). High-quality preschool: the socioeconomic composition of preschool classrooms and children’s learning. Early Education and Development, 24(8), 1082–1111. Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70(1), 41–55. Rosenbaum, P. R., & Rubin, D. B. (1984). Reducing bias in observational studies using subclassification on the propensity score. Journal of the American Statistical Association, 79(387), 516–524. http://dx.doi.org/10.1080/01621459. 1984.10478078 Royston, P. (2005). Multiple imputation of missing values: update of ice. Stata Journal, 5, 527–536. Sabol, T. J., Soliday Hong, S. L., Pianta, R. C., & Burchinal, M. R. (2013). Can rating pre-k programs predict children’s learning? Science, 341, 845–846. http://dx. doi.org/10.1126/science.1233517 Samuels, C.A. (2014, January). Cities seize momentum on pre-k expansion. Education Week. Retrieved from http://www.edweek.org.proxy.bc.edu/ew/ articles/2014/01/15/17preschool ep.h33.html?qs=cities+seize+momentum. Schmit, S., Matthews, H., Smith, S., & Robbins, T. (2013). Investing in young children: a fact sheet on early care and education participation, access, and quality. New York, NY, Washington, D.C: National Center for Children in Poverty, Center for Law and Social Policy. Schulman, K., & Blank, H. (2013). Pivot point: state child care assistance policies 2013. National Women’s Law Center. Retrieved from http://www.nwlc.org/sites/ default/files/pdfs/final nwlc 2013statechildcareassistancereport.pdf February Smith, T., Kleiner, A., Parsad, B., & Farris, E. (2003). Prekindergarten in U.S. public schools: 2000–2001 (NCES 2003-019). Washington, D.C: U.S. Department of Education, National Center for Education Statistics. Smith, J., & Todd, P. (2005). Does matching overcome LaLonde’s critique of nonexperimental estimators? Journal of Econometrics, 125(1–2), 305–353. http://dx.doi.org/10.1016/j.jeconom.2004.04.011 Snow, K. L. (2006). Measuring school readiness: conceptual and practical considerations. Early Educ. Dev., 17(1), 7–41. Tout, K., Starr, R., Soli, M., Moodie, S., Kirby, G., & Boller, K. (2010). The child care quality rating system assessment: compendium of quality rating systems and evaluations. Washington, D.C: Office of Planning, Research & Evaluation, Administration for Children & Families, U.S. Department of Health and Human Services. U.S. Department of Health and Human Services, Administration for Children and Families. (2010). Head Start impact study. Final report. Washington, D.C: U.S. Department of Health and Human Services, Administration for Children and Families. Votruba-Drzal, E., Coley, R. L., & Chase-Lansdale, P. L. (2004). Child care and low-income children’s development: direct and moderated effects. Child Development, 75, 296–312. http://dx.doi.org/10.1111/j.1467-8624.2004.00670. x Votruba-Drzal, E., Coley, R. L., Koury, A., & Miller, P. (2013). Center-based child care and academic skills development: importance of timing and household resources. Journal of Educational Psychology, 105, 821–838. http://dx.doi.org/10. 1037/a0032951 ˜ C., Li-Grining, C., & Votruba-Drzal, E., Coley, R. L., Maldonado-Carreno, Chase-Lansdale, P. L. (2010). Child care and the development of behavior problems among economically disadvantaged children in middle childhood. Child Development, 81, 1460–1474. http://dx.doi.org/10.1111/j.1467-8624. 2010.01485.x Weber, R. (2011). Understanding parents’ child care decision-making: a foundation for policy making (OPRE 2011-12). Washington, D.C: Office of Planning, Research and Evaluation, Administration for Children and Families, U.S. Department of Health and Human Services. Weiland, C., Ulvestad, K., Sachs, J., & Yoshikawa, H. (2013). Associations between classroom quality and children’s vocabulary and executive function skills in an urban public prekindergarten program. Early Childhood Research Quarterly, 28, 199–209. Weiland, C., & Yoshikawa, H. (2014). Does higher peer socio-economic status predict children’s language and executive function skills gains in prekindergarten? Journal of Applied Developmental Psychology, 35(5), 422–432.

R.L. Coley et al. / Early Childhood Research Quarterly 36 (2016) 91–105 Whitebook, M., Howes, C., & Phillips, D. (1989). Who cares? Child care teachers and the quality of care in America: National Child Care Staffing Study. Oakland, CA: Child Care Employee Project. Yoshikawa, H., Weiland, C., Brooks-Gunn, J., Burchinal, M. R., Espinosa, L. M., Gormley, W. T., & Zaslow, M. J. (2013). Investing in our future: the evidence base on preschool education. Society for Research in Child Development and Foundation for Child Development.

105

Zhai, F., Brooks-Gunn, J., & Waldfogel, J. (2011). Head Start and urban children’s school readiness: a birth cohort study in 18 cities. Developmental Psychology, 47(1), 134–152. Zhai, F., Brooks-Gunn, J., & Waldfogel, J. (2014). Head Start’s impact is contingent on alternative type of care in comparison group. Developmental Psychology, 50(12), 2572–2586.