An Instrumental Variable Approach for Head Start Attendance on Low Income Children

Abstract

Purpose: This study examined (1) the causal impact of Head Start on children’s comprehensive outcomes and (2) why families did not comply to the original assignment. Method: Based on the Head Start Impact Study data, children between the ages of 3 and 4 years (N = 3780) were examined for assignment, attendance, and causal impacts on outcomes. Results: Effect of instrumental variable analysis was greater than that of assignment and attendance for measured outcomes. Assignment to the Head Start, older child age (age 4 over age 3), urban residence, and higher family income were associated with non-compliance. Inclusion of care quality into the model illustrated important nuances in predicting families’ compliance with Head Start versus control condition. Discussion: Head Start participation among low income children should be maintained for its positive causal effects on children. Quality of childcare as well as several child and family barriers to access to Head Start needs to be addressed to enhance Head Start enrollment.

Keywords

causal head start impact instrumental variable analysis compliance quality of care

The Head Start (HS) program provides federally funded early childhood education services aimed at ensuring school readiness of low-income children (Lamb-Parker et al., 2001). Head Start programs are centered on improving children’s early academic and social-emotional skills such as language, numeracy, and resilience (Improving HS for School Readiness Act, 2007). Numerous studies have examined short- and long-term HS effects at the local, regional, and national levels. Early, large-scale evaluations indicated positive results of HS participation, such as increases in academic skills (Barnett & Hustedt, 2005), with small to moderate effect size estimates (.09-.26; Puma et al., 2010). Other findings suggested HS children were more likely to complete high school and attend college, and less likely to be convicted of a crime. Long-term effects also include increased emotional understanding and social problem solving (Bierman, et al., 2008).

Most research on the effects of HS participation suggests children who attend the program may have advantages in school readiness over those who do not. For instance, among 173, 4-year-old children randomly assigned to Head Start or a waitlist condition, vocabulary and phoneme awareness gains across 1 year were greater for HS children compared to controls (Abbott-Shim, Lambert, & McCarty, 2003). In a large-scale study, HS children (N = 1,569) appeared to close gaps in reading and mathematics achievement between themselves and their more affluent peers from first through third grade (Kreisman, 2003). Nevertheless, these effects are often small to moderate in magnitude with some studies suggesting attenuation over time (Lee, Brooks-Gunn, Schnur, & Liaw, 1990). Furthermore, there are often differences between the magnitude of academic outcomes in comparison to emotional outcomes. Lee and colleagues (2014) found that children who attended Head Start had higher early reading and math scores than those that were placed in parental or non-parental care. However, children in HS exhibited more behavioral challenges compared to children in parental care. Alternatively, Hubbs-Tait et al. (2002) found that HS attendance predicted teacher ratings of higher social competence compared to children who did not attend HS regardless of the number of risk factors faced by children in the sample.

Head Start Impact Study and Outcomes

The HS Impact Study (HSIS) includes data from approximately 5,000 children and families randomly assigned to HS or a control condition in the 2002–2003 program year. Initial findings (Puma et al., 2005) indicated some positive effects on child outcomes, but results were mixed. While small to moderate positive effects were found for both 3- and 4-year-old children on measures of pre-reading, pre-writing, vocabulary, and literacy, no effects were apparent for oral comprehension, phonological awareness, early math, or social skills. Findings also suggested small to moderate positive effects for health care access and health status for children enrolled in HS who entered the study as 3-year-olds, with similar findings for health care access but not status among the 4-year-old group. Overall, the advantages children gained during HS yielded very few statistically significant differences in outcomes at the end of first grade (Puma et al., 2005), calling into question the long-term benefits of HS over alternative care options.

Following the mixed findings reported by Puma and colleagues (2005), researchers have examined HSIS data for nuances that may better explain the effects of HS attendance on children and families. Several of these analyses suggest greater gains for children who enter the program with considerable adversity. For instance, Bloom and Weiland (2015) examined variation in effects of HS on cognitive and social-emotional outcomes for children, which indicated that HS program effects served a compensatory function to mitigate disparities in cognitive outcomes for children who were eligible for the program. Furthermore, it appeared that HS was specifically beneficial for dual language learners and Spanish-speaking children. Likewise, Kline and Walters (2016) examined cognitive and vocabulary awareness outcomes for 3,571 HSIS participants and concluded that Head Start resulted in the largest effects for children deemed unlikely to otherwise attend preschool, and those who are unlikely to attend Head Start versus other programs. Lee and colleagues (2014) found that children who entered HS with lower cognitive ability, attended full-time, and had parents with less education, often appeared to benefit more than their HS peers. On the other hand, children who attended HS full-time and had parents with higher levels of education, appeared to exhibit more behavior problems.

Variability Within Head Start

Broadly, findings about HS’s long-term effects present a somewhat puzzling picture. Some research suggests initial gains are modest at best and fade by school entry. Still other research points to the relative benefits of HS over other programs, particularly for children who are unlikely to receive similar programming in other venues. Study limitations shed light on directions for clarifying these mixed conclusions. Many studies employed ad hoc control groups, were subject to attrition, and had small sample sizes and therefore limited statistical power (Garces, Thomas, & Currie, 2002). Studies such as the HSIS address these limitations by prospectively assigning participants to conditions and including a large representative sample.

Subsequently, several researchers have re-analyzed the HSIS data in an attempt to illuminate individual characteristics and contextual conditions associated with differential effects across specific HSIS subgroups. Support for this approach is provided by findings from Morris et al. (2018) showing that the HSIS results regarding average effects conceal the considerable variation in effectiveness. Comparing children in the HSIS assigned to HS with controls who attended center-based care (rather than the entire control group; N = 3485), Mackintosh and McCoy (2019) found that early math skills were not significantly different across the groups following 1 year of preschool, but that these skills were higher for children in both groups who also demonstrated higher levels of social competence. In another study using HSIS data, researchers found that children assigned to HS demonstrated greater vocabulary and math skills and had higher parent-reported social and behavioral skills than those in parental and relative or non-relative care. However, these differences were not evident when comparing children assigned to HS with those who attended alternative center-based care (Zhai, Brooks-Gunn, & Waldfogel, 2014).

Several factors may have contributed to the variable effects suggested by previous HSIS evaluations. Teacher qualifications and training varied widely across programs. According to Puma and colleagues (2010), about 30% of children in both cohorts had teachers with Bachelor’s degrees, another 30% had teachers with Associate’s degrees, and approximately 40% of the children were taught by teachers who did not have either degree. However, children who were provided access to HS were more likely than controls to have a teacher who held a bachelor’s degree, completed college-level early childhood education coursework, or obtained their child development associate’s degree compared to controls. Furthermore, teacher training and mentoring varied greatly. For instance, less than half the children had teachers who received 25 hours of training in the previous year or received mentoring at least one a month.

Another aspect in which HS program quality might be lacking is in the classroom experience. Although the majority of children attended classrooms with good ratings of quality, which emphasized language, literacy, and mathematics, and met ratio standards, there was vast variation in these experiences across cohorts. For example, about 40% of the children in both cohorts were in classrooms that did not emphasize language and literacy or math activities. However, children in the HS group were more likely to be in classrooms with higher quality than those in the control group as measured by standardized observation (Puma et al., 2010).

Head Start Assignment and Family Decision-Making

Importantly, random assignment to HS in the HSIS was a matter of eligibility. That is, families were provided access to HS programming if randomly assigned to the treatment group, but there was no guarantee of their enrollment in a HS program. A substantial portion of families (i.e., 15% and 20%, respectively, in the 3- and 4-year-old cohorts) who were assigned to the treatment group did not enroll in HS programs (Puma et al., 2005; 2010). Likewise, research in other samples indicates that low-income parents place their children in prekindergarten programs other than HS, and many working parents use typical child care, even if they must pay for it, or depend on other family members to care for their children (Besharov & Higney, 2007). Tang, Coley, & Votruba-Drzal (2012) found that among low-income families living in urban settings, maternal work schedule, geographic location, and, to some degree, ethnicity were influential in families’ care choices for their children. Both child and family characteristics related to a preference for HS versus other care, including difficult child temperament and higher maternal literacy skills. HS or other center-based care was chosen more often than White compared to Latina mothers.

Other research, however, calls into question ethnically based differences in care preferences and instead suggests nuanced social and policy influences underly differences in care choices. Furthermore, researchers found that family structure (i.e., the ages and number of children), education level of parents, employment history, and personal beliefs were strongly associated with choosing center care over home-based care (Huston et al., 2002). For some parents, choice in type of care is related to how much the family prioritizes a setting that will care for children when they are sick and how much they emphasize the importance of a trained care provider (Early & Burchinal, 2001). Still yet, in a sample of 355 employed mothers with children under 6 years old, participants rated caregiver warmth and education level, as well as the use of a play-based curriculum as the most critical factors in their decision regarding type of care (Kensinger Rose & Elicker, 2008). Income level also appears to play a role for families in choice of child care, as mothers in high-income families appear more likely to choose child care based on quality versus concerns like cost or location (Peyton, Jacobs, O’Brien, & Roy, 2001).

There is a clear need for more research elucidating factors underlying familial care choices. Some clarity may be gained from findings regarding issues related to families’ decisions to remain in HS for a second year taken from the HS Impact Study. The likelihood of children returning for a second year of HS was significantly related to many factors, including program services, availability of other care options, as well as satisfaction with the center’s sensitivity to cultural issues and facilitation of child development. Families were also more likely to return if they experienced recent financial hardship, were Hispanic, had Spanish as a home language, and recently immigrated to the US (Puma et al., 2005, 2010).

Quality of Head Start in Childcare Choice

Another potential determining factor for childcare choice is the quality of HS programs and classrooms, given demonstrated associations between classroom quality and child outcomes (e.g., Bryant, Burchinal, Lau, & Sparling, 1994; Gelber & Isen, 2013). Research using data from the Early Childhood Longitudinal Study-Birth Cohort suggested that family resources and needs appeared to have greater influence on care decisions in the first 2 years of children’s lives, but factors such as parent education and income played a greater role in later care decisions (Coley, Votruba-Drzal, Collins, & Miller, 2014). These findings suggest quality of care may become more salient for families as children age, as parental education and family income are often associated with preference for higher-quality care. Raver and colleagues (2008) found that HS classrooms vary substantially in their overall quality. While many classrooms and teachers scored high on components such as creating a positive emotional climate and behavioral management of classrooms, many classrooms demonstrated lower classroom quality via factors such as lack of positive climate and difficulty with child behavior management. Considering potential variability in HS classroom quality and subsequent outcomes, the role of perceived quality in family decisions to enroll in Head Start warrants further study.

Limitations in Previous Evaluations

The methodology employed in the HSIS (Puma et al., 2005, 2010) precluded researchers from guaranteeing that children and families randomly assigned to the non-HS group did not actually participate in the program. Though agencies involved in the research study agreed not to serve families in the control group, this did not preclude families from seeking services through other agencies in nearby communities. A number of families from the non-HS sample (17.6%) participated in at least some HS services for their children during the years the study was conducted. Likewise, noted that most studies examining HS effects compared children enrolled in HS with all other children. These controls included children with a diverse range of experiences such as parental, other center-based, and home-based care.

There is a need to clearly define the reference group in studies of HS effects to provide conclusive findings. This need has been partially addressed in some research, such as that by Zhai et al. (2011) which explicitly defined groups of children (i.e., parental care, prekindergarten, other center-based care, or other non-parental care) in reference to those participating in HS and found that HS effects varied depending on the reference group. Likewise, Lee et al. (2014) found that HS participants exhibited better academic outcomes compared to children placed in parental or non-parental care; however, outcomes were the same as or worse than those who had gone to pre-K or another type of center-based care. However, the analyses did not account for when children entered HS programs or number of years children received services. Relatedly, it was shown that approximately 29% of children in the Early Childhood Longitudinal Study-Kindergarten (ECLS-K) sample identified by parents or schools as attending HS programs did not actually attend those programs. This high rate of over-report is common (see also Lee et al., 2014) and warrants caution in the interpretation of findings.

Furthermore, while there may be promising overall treatment effects of HS compared to other programs, wide variability in the samples studied warrant caution regarding conclusions pertaining to overall average treatment effects. Subramanian, Kim, & Christakis (2018) suggest that while an average outcome can be somewhat helpful to discern overall change resulting from a given program, it does not ensure that this number is uniformly meaningful for populations or individuals. The researchers suggest that following evidence of an overall, positive average treatment effect, there is a need to evaluate what portion of the treatment group demonstrated the desired effect suggested by the overall findings. They propose solutions for incorporating separate mean and variance estimates in the treatment and control groups to determine the degree to which an overall effect in addition to subgroup improvements may be evident.

In addition to an over-reliance on average treatment effects, few studies have examined outcomes for children enrolled in HS comprehensively, especially in relation to social-emotional and health outcomes. This is a notable gap in the literature given HS’s emphasis on whole child functioning across the cognitive, social-emotional, and health domains (Zigler & Styfco, 2001; National HS Association, 2020). While the existing literature using HSIS data has conducted relatively comprehensive analyses of different outcomes, many studies of HS programs are focused on singular outcomes (e.g., Bonuck, Schwartz, & Schechter, 2016; Chor, 2018).

The Present Study

Given recent attention to the difficulties in extrapolating conclusions from the HSIS to real-world applications (Stuart & Rhodes, 2017), limited overall treatment effects of HS found in the HSIS evaluation (Puma et al., 2010), and likelihood of treatment effect variability (e.g., Morris et al., 2018) there is a need to better understand outcomes for children who participate in HS, and what factors may be related to families’ decisions to choose other center-based programs over Head Start. The present study serves as a unique contribution to the literature by examining the causal impact of HS attendance using compliance (actual enrollment) and by investigating the factors affecting families’ choice to enroll in HS versus non-HS programs. Specifically, the present study examined for the following hypothesis

1. To estimate the causal impact of HS enrollment on children’s comprehensive outcomes (cognitive, social-emotional, and health) with an instrumental variable (IV) analysis. We hypothesized that HS enrollment improved the outcomes.

2. To investigate why children who were assigned to HS did not enroll in HS but attended other care programs. We estimated the determinants of such a non-compliance to HS assignment. In particular, we hypothesized that the quality of care is associated with the prevalence of non-compliance.

Method

Head Start Impact Study (HSIS) Data

The current study used the Head Start Impact Study (HSIS) data. The IRB office notified that the activity described in the present study was considered not to be “research” as defined by the Common Rule as codified in the U.S. Department of Health and Human Services (DHHS) regulations for the protection of human research subjects since this secondary data analysis study using the Head Start Impact Study data does not contain any identifiable personal information as defined by the Department of Health and Human Services [45.CFR.46.102(f)].

Target Sample

This study included data from children entering HS at age 3 and 4 in Fall 2002 who were measured their outcomes in Spring 2003. This cohort was selected to allow for analyses examining outcomes following a single year of HS attendance. Of the original 4,442 children, outcome data were available for 3,780 children who remained enrolled in HSIS in Spring 2003. Attrition was as follows: 662 children left the study following assignment (287 HS, 375 control), leaving 2,359 HS-assigned children and 1,421 controls. Following assignment, there was a portion of children who did not comply with the original assignment, including “no-shows” (HS-assigned but did not attend HS) and “cross-overs” (assigned to control group but attended HS). Among children who comprised the final control group (n = 1,656), care arrangements included other center-based (n = 638), parental (n = 748), relative (n = 151), and non-relative (n =119) care. In sum, the current sample included 3,780 children: 2,124 in the Head Start group (actually enrolled) and 1,656 in the control group (enrolled other types of care).

Measures

Head Start Assignment and Enrollment

The current study used three types of Head Start variables. Head Start Assignment [Intent to Treat (ITT)] identified children assigned to Head Start (ITT HS) and children assigned to the control group (ITT NHS). Head Start Enrollment identified children who enrolled in Head Start (HS) and those who did not enroll in Head Start (NHS). Instrumental Variable (IV) of Head Start measured Head Start enrollment that accounted for the variance of Head Start assignment and enrollment, related to observed and non-observed conditional variables.

Child Cognition: Woodcock Johnson III Tests of Achievement (WJIII-A)

The WJ III Achievement (Schrank, McGrew, & Woodcock, 2001, Table 10, p.19) is measuring academic skills and abilities similar to those measured by other achievement tests. The concurrent criterion validity suggests that WJ III-A measure academic skills and abilities similar to those measured by the Wechsler Individual Achievement Test (WIAT) (0.31 to 0.82) and the Kaufman Test of Educational Achievement (KTEA) (0.29 to 0.81).

Oral Comprehension

The WJIII-A Oral Comprehension subscale was used to measure children’s listening, reasoning, and vocabulary skills. Children were read an analogy or passage with a word omitted and asked to verbally provide the missing word. Median Test Reliability- WJ III Tests of Achievement, Oral Comprehension is 0.85 (Schrank et al., 2001, Table 4, p.10; US DHHS, 2010, p.3-8).

Letter-Word Identification

The WJIII-A Letter-Word Identification subscale was used to measure children’s ability to identify letters and words. Median Test Reliability- WJ III Tests of Achievement, Letter-Word Identification is 0.94 (Schrank et al., 2001, Table 4, p.10; US DHHS, 2010, p.3-7).

Spelling

This subtest evaluates children’s ability to write letters and words presented verbally by an examiner. Items progress in difficulty, beginning with basic line drawing and ending with spelling full words. Median Test Reliability- WJ III Tests of Achievement, Spelling is 0.90 (Schrank et al., 2001, Table 4, p.10; US DHHS, 2010, p.3-8).

Applied Problems

The applied problems subtest requires children to complete practical math problems read verbally by an examiner, using basic math calculation skills. Median Test Reliability- WJ III Tests of Achievement, Applied Problem is 0.93 (Schrank et al., 2001, Table 4, p.10; US DHHS, 2010, p.3-8).

Child Receptive Language: Peabody Picture Vocabulary Test, Third Edition (PPVT-3)

The PPVT-3 measured children’s ability to identify common objects and action words represented in an array of pictures presented by an evaluator. Children were asked to examine a group of four pictures and identify the picture that best matched a word presented by the evaluator (Published reliability = 0.95). An adaptive, shorter version of the PPVT was developed for the HSIS using Item Response Theory (US DHHS, 2010, p.3-5).

Child Social-Emotional Functioning

Child–Parent Relationship

The quality of the child-parent relationship was measured using the 15-item Child–Parent Relationship Scale (Pianta, 1992). The 8-item conflict subscale measures the degree to which a parent feels that his or her relationship with a particular child is characterized by negativity. Cronbach alphas for paternal conflict were .78 to .80 at 54 months. The 7-item closeness scale assesses the extent to which a parent feels that the relationship is characterized by warmth, affection, and open communication. Cronbach alphas for paternal closeness were .72 to .74. The total positive relationship score was used in the current study, which can range from 15 to 75 with higher scores indicating a more positive child–parent relationship.

Child–Teacher Relationship

Teacher perceptions of the quality of their relationships with target children were measured using the Student–Teacher Relationship Scale. Scores ranged from 15 to 75, with higher scores indicating a more positive child–teacher relationship. The Cronbach Alpha for the internal consistency coefficients of the Child–Teacher Relationship Scale is 0.73 (Seven & Ogelman, 2014).

Social Skills and Positive Approaches to Learning (Achenbach, Edelbrock, & Howell, 1987)

Parents rated children’s social skills and learning behaviors using a 7-item scale, with ratings ranging from 0 (not true) to 2 (very true). Scores ranged from 0 to 14, with higher scores indicating greater positive social skills and positive approaches to learning. The test–retest reliability r was .87. The intraclass correlation between any two randomly chosen individuals measured in 2003 was 0.62 for age 3 and 0.63 for age 4 cohorts (US DHHS, 2010, 3-33).

Social Competencies

Parents were asked to provide information on social capabilities using a Social Competencies Checklist. Items addressed prosocial behaviors such as sharing and caring for one’s belongings, as well as interpersonal competence. The intraclass correlation between any two randomly chosen individuals measured in 2003 was 0.54 for age 3 and 0.56 for age 4 cohorts (US DHHS, 2010, 3-33).

Behavioral Problems (Achenbach et al., 1987)

Behavioral issues including aggression, defiance, inattention, hyperactivity, shyness, and withdrawal were rated by parents across 14 items which ranged from 0–28 with higher scores indicating more behavioral issues. The Cronbach Alpha for the internal consistency is from .89 to .90 (Zill, 1990).

Health-Related Outcomes

Parents responded to 4 questions asking about children’s health care services (US DHHS, 2010, p.3-28), including whether the child received since September of the program year (1) dental care (2) a hearing screening, and (3) a vision screening, (4) whether the child has been healthy, and (5) whether the child had a regular primary medical care provider. Responses were scored 1 for “yes” and 0 for “no” for each question.

Quality of Care

Quality of care was measured for all children except those in the control group who were cared for by their parents (n = 748). This composite variable included items from 12 standardized and non-standardized measures: (1) The Early Childhood Environment Rating Scale-Revised Edition for children enrolled in center-based care or the Family Day Care Rating Scale for children enrolled in non-relative home-based care. Both the ECERS-R and FDCRS assess overall quality of the care environment including caregiver–child interactions; (2) the Arnett Scale of Lead Teacher Behavior a 30-item measure of observed lead teacher behavior across the dimensions of sensitivity, harshness, detachment, permissiveness, independence; (3) Literacy activities, including how often teachers/care providers reported using each of 12 reading and language activities with children in their classroom or child care home (e.g., identifying letter names, practice writing or spelling their name, practicing letter sounds, making up stories); (4) Math activities, including how often teachers/care providers reported using each of 8 math activities with children in their classroom or child care home (e.g., counting out loud, playing with shape blocks, working with rulers or measuring cups); (5) Other activities, including how often teachers and care providers reported using non-academic instructional activities (arts and crafts, games, sports, and chores) with the children in their classroom or child care home; (6) The staff:child ratio; (7) Education of teacher/care providers; (8) Child development associates degree/early childhood education coursework completed by teacher/care providers; (9) Training completed by teachers in the past year (25+ hours vs. <25 hours); (10) Parent involvement, which included questions such as how often parents volunteered, attended meetings or parent–teacher conferences, and assisted with field trips; (11) Home visits; and (12) Program services to children and families. In the current sample, the mean quality of care composite score was 0.615 (SD = 0.230, range = 0 to 1). Higher scores indicated higher quality of care.

Analytic Approach

We make a distinction between “assignment” and “enrollment” of the Head Start program—the former referring to a random allocation to the Head Start program, and the latter is about whether the (non) assigned children attended the program. As shown later, the data show that about 18.2% of children (430 children out of 2359) who were assigned to the program did not attend it. Similarly, 13.7% of children (195 children out of 1421) who were not assigned to the Head Start program ended up with attending it. In order to explore the importance of the Head Start program from a social services delivery perspective, it is crucial to investigate the impact of attending the program, rather than that of being assigned to the program, especially when there are numbers of non-compliances.

A: Effects of Assignment

Let us consider the following linear regression model as equation (1)

Y_{j} = α + τ X_{j}^{assign} + Z_{j}^{'} β + ε_{j}

(1)

Y_{j}

is the outcome variable for child j (see Table 1 for the list of outcome variables), and

X_{j}^{assign}

indicates whether she or he is assigned to the Head Start program. The vector

Z_{j}

represents control variables in the regression model, including child characteristics (age, gender, special needs, race, baseline cognitive test scores, baseline social-emotional scores, and baseline health status) and family characteristics (mother [age at the child birth, education, marital status] and family [status of both biological parents, language speaking at home, immigration status, residential location, income, household risk factors, primary caregiver age], see Table 1). Finally,

ε_{j}

is idiosyncratic error. The Intention-to-treat (ITT) effect of the Head Start program is represented by the coefficient

τ

, and given that the assignment was random,

τ

represents the causal impact of being assigned to the program.

Table 1.

Descriptive Statistics for the Study Variables by Head Start Attendance.

Attendance	Not attended=0			Attended=1			p-value	Difference
	Mean	S.D	N	Mean	S.D	N
Head Start assignment	0.260	0.439	1656	0.908	0.289	2124	0.000	−0.649
Cognitive
PPVT	91.73	9.54	1582	92.28	9.30	2093	0.079	−0.550
WJIII Applied problem solving	86.57	19.08	1572	89.05	17.43	2084	0.000	−2.482
WJIII Spelling	91.12	12.82	1594	93.32	11.63	2096	0.000	−2.199
WJIII Word identification	90.27	17.22	1584	94.16	17.67	2098	0.000	−3.890
WJIII Oral comprehension	91.95	13.91	1545	90.96	13.72	2073	0.034	0.985
Social
Child–parent relationship	63.43	7.75	1569	63.83	7.55	1944	0.123	−0.400
Behavioral problems	6.09	3.61	1638	5.78	3.65	2053	0.010	0.308
Social competencies	10.99	1.34	1639	11.00	1.40	2053	0.795	−0.400
Social Skills	12.49	1.66	1638	12.43	1.75	2053	0.278	−0.012
Child–teacher relationship	64.39	8.00	526	64.26	8.63	1954	0.742	0.138
Health
Dental check-up	0.49	0.50	1639	0.73	0.45	2051	0.000	−0.234
Health status	0.80	0.40	1640	0.81	0.39	2053	0.538	−0.008
Hearing check-up	0.49	0.50	1577	0.75	0.43	1942	0.000	−0.257
Vision check-up	0.49	0.50	1577	0.75	0.43	1943	0.000	−0.264
Place for routine care	0.97	0.16	1639	0.98	0.14	2053	0.240	−0.006
Other variables
Age entered in head Start	3.47	0.50	1656	3.42	0.49	2124	0.001	0.056
Gender (1=male; 0=female)	0.50	0.50	1656	0.51	0.50	2124	0.640	−0.008
Special needs	0.12	0.32	1656	0.14	0.34	2124	0.072	−0.020
Academic skills prior to head Start	0.26	0.44	1656	0.23	0.42	2124	0.069	0.026
Social skills prior to head Start	12.23	1.83	1656	12.27	1.78	2124	0.502	−0.040
Health status prior to head Start	0.79	0.41	1656	0.79	0.41	2124	0.914	−0.001
Black	0.28	0.45	1656	0.31	0.46	2124	0.038	−0.031
Hispanic	0.37	0.48	1656	0.38	0.49	2124	0.530	−0.010
White	0.34	0.48	1656	0.30	0.46	2124	0.007	0.041
Age of caregiver	29.01	7.27	1656	29.32	7.57	2124	0.208	−0.307
Teenage mother	0.17	0.38	1656	0.15	0.36	2124	0.202	0.015
Less than high school education	0.38	0.49	1656	0.38	0.48	2124	0.620	0.008
High school diploma or GED	0.33	0.47	1656	0.33	0.47	2124	0.852	−0.003
Beyond high school education	0.29	0.45	1656	0.29	0.45	2124	0.737	−0.005
Married	0.55	0.50	1656	0.55	0.50	2124	0.568	−0.009
Both biological parents live with a child	0.50	0.50	1656	0.50	0.50	2124	0.646	−0.008
Home language (English=1)	0.71	0.45	1656	0.69	0.46	2124	0.129	0.023
Biological mother recent immigrant	0.19	0.39	1656	0.20	0.40	2124	0.526	−0.008
Living in urban	0.84	0.37	1656	0.84	0.37	2124	0.974	0.000
Family income/month	1.85	1.09	1656	1.88	1.11	2124	0.368	−0.032
No/low household risk	0.74	0.44	1656	0.73	0.44	2124	0.592	0.008
Moderate household risk	0.18	0.39	1656	0.19	0.39	2124	0.549	−0.008
High household risk	0.08	0.27	1656	0.08	0.27	2124	0.989	0.000

B: Effects of Enrollment

The next model in equation (2) instead analyzes attendance rather than assignment

Y_{j} = γ + θ X_{j}^{attend} + Z_{j}^{'} δ + ε_{j}

(2)

in this model, we use the same outcome variable and the same set of control variables.

X_{j}^{attend}

represents whether one attends the Head Start program or not, and

ε_{j}

is idiosyncratic error. Although the assignment was based on lottery, whether to attend the program assigned or not would be determined by non-random factors such as travel distance, parental preference and need etc., that are not observed in the data. In order to address the potential bias stemming from the omitted variables, we employ an instrumental variable (IV) approach (Wooldridge, 2010). The IV approach is an approach to correct such a bias and to draw causal inferences, which has been widely used in empirical economics and other social sciences, and more recently, in medical sciences and epidemiology. In this approach, the estimation is conducted in two steps (but simultaneously). First, the treatment variable (

X^{attend}

) is regressed on the IV and the control variables with the same set of control variables as shown in equation (3)

X_{j}^{attend} = ζ + {φIV}_{j} + Z_{j}^{'} η + e_{j}

(3)

an IV has to be an exogenous determinant of the treatment variable (attendance) and should not be correlated with other factors. In this study, we use the random assignment to the Head Start program as our IV. This IV is random and hence is uncorrelated with other factors, and it is well correlated with the enrollment. Second, the outcome variable (

Y_{j}

in equation (2)) is regressed on the predicted values of the treatment variable. This gives an unbiased causal impact of the treatment variable on the outcome variable. More precisely, in this estimation, the variations of the treatment variable via the IV only are used to explain the variations of the outcome variable. Therefore we estimate a local average treatment effect for the “compliers,” whose decision was potentially affected by assignment (Angrist, Imbens & Rubin, 1996). However, the estimated impact does not represent the effect for those who attend or did not attend Head Start irrespective of the assignment (“non-compliers”).

Further, to determine which baseline characteristics affecting children’s compliances with the original assignment, several sets of logistic regression analysis were conducted: (1) for all children, (2) for Head Start assigned children, and (3) for Non-Head Start assigned children. We also added quality of care of each service provider to determine the association between quality and non-compliances. In the statistical analyses, all cognitive and socio-emotional outcomes were standardized to Z-score. For these standardized outcomes, the regression coefficient shows the effect of a change in Head Start status on standard deviation point changes in the outcome variables. Health outcome variables were used as original dichotomized values. The current study used STATA 14/SE.

Results

Descriptive Analysis

Table 1 describes study variables by Head Start enrollment status. Child’s age and child race differed depending on actual Head Start attendance status. Younger children (3.42 vs. 3.47, p < .001) were more likely to attend Head Start. Black (31.5% vs. 28.4%, p < .05) children were more likely to attend Head Start, whereas white (30.4% vs. 34.5%, p < .01) children were less likely to attend Head Start programs. Descriptive statistics for the study variables by Head Start assignment is available in Table 2. Most baseline variables did not differ between children who were assigned to Head Start and did not assign to Heed Start except special needs status, teenage mother, and household risk status. As shown in Table 2, children with special needs were more likely to assign to Head Start than control group (14% vs. 12%, p < .05). Children born to teenage mothers were less likely assigned to Head Start (16% vs. 18%, p < .05). Children living in a No/Low household risk were less assigned to Head Start (75% vs. 78%, p < .05).

Table 2.

Descriptive Statistics for the Study Variables by Head Start Assignment.

Head Start Assignment	Control group = 0			Head Start=1			p-value	Difference
	Mean	S.D	N	Mean	S.D	N
Head Start attendance	0.137	0.344	1421	0.818	0.386	2359	0.000	−0.680
Cognitive
PPVT	91.36	9.64	1368	92.44	9.25	2312	0.001	−1.078
WJIII applied problem solving	86.48	19.19	1357	88.87	17.51	2304	0.000	−2.395
WJIII Spelling	91.15	12.45	1378	93.08	12.00	2317	0.000	−1.931
WJIII Word identification	89.89	17.05	1370	94.02	17.72	2317	0.000	−4.137
WJIII oral comprehension	91.52	14.05	1333	91.29	13.68	2290	0.619	0.236
Socio-emotional
Child–parent relationship	63.41	7.82	1346	63.80	7.53	2167	0.141	−0.390
Behavioral problems	6.09	3.58	1407	5.81	3.67	2284	0.022	0.282
Social competencies	11.00	1.30	1409	10.99	1.41	2283	0.915	0.003
Social Skills	12.46	1.67	1408	12.45	1.74	2283	0.957	0.005
Child–teacher relationship	63.89	8.37	543	64.38	8.56	1983	0.238	−0.487
Health
Dental check-up	0.53	0.50	1410	0.68	0.47	2280	0.000	−0.155
Health status	0.80	0.40	1410	0.81	0.39	2283	0.292	−0.014
Hearing check-up	0.50	0.50	1358	0.72	0.45	2161	0.000	−0.216
Vision check-up	0.50	0.50	1358	0.72	0.45	2162	0.000	−0.215
Place for routine care	0.97	0.16	1410	0.98	0.15	2282	0.400	−0.004
Other variables
Age entered in head Start	3.45	0.50	1796	3.45	0.50	2646	0.750	0.005
Gender (1=male; 0=female)	0.49	0.50	1796	0.50	0.50	2646	0.681	−0.006
Special needs	0.11	0.32	1796	0.14	0.35	2646	0.016	−0.025
Academic skills prior to head Start	0.25	0.43	1796	0.24	0.43	2646	0.459	0.010
Social skills prior to head Start	12.25	1.77	1796	12.25	1.80	2646	0.912	−0.006
Health status prior to head Start	0.79	0.41	1796	0.80	0.40	2646	0.507	−0.008
Black	0.30	0.46	1796	0.31	0.46	2646	0.668	−0.006
Hispanic	0.37	0.48	1796	0.38	0.49	2646	0.519	−0.010
White	0.33	0.47	1796	0.31	0.46	2646	0.274	0.016
Age of caregiver	28.65	7.06	1796	29.08	7.52	2646	0.057	−0.427
Teenage mother	0.18	0.39	1796	0.16	0.37	2646	0.034	0.024
Less than high school education	0.39	0.49	1796	0.37	0.48	2646	0.287	0.016
High school diploma or GED	0.33	0.47	1796	0.34	0.47	2646	0.659	−0.006
Beyond high school education	0.28	0.45	1796	0.29	0.45	2646	0.495	−0.009
Married	0.55	0.50	1796	0.56	0.50	2646	0.620	−0.008
Both biological parents live with a child	0.49	0.50	1796	0.50	0.50	2646	0.916	−0.002
Home language (English=1)	0.71	0.45	1796	0.70	0.46	2646	0.339	0.013
Biological mother recent immigrant	0.19	0.39	1796	0.20	0.40	2646	0.500	−0.008
Living in urban	0.84	0.36	1796	0.84	0.36	2646	0.894	−0.001
Family income/month	1.85	1.02	1796	1.87	1.05	2646	0.675	−0.013
No/low household risk	0.78	0.42	1796	0.75	0.43	2646	0.025	0.029
Moderate household risk	0.15	0.36	1796	0.18	0.38	2646	0.064	−0.021
High household risk	0.07	0.25	1796	0.07	0.26	2646	0.309	−0.008

Research Question 1

Hypothesis 1

Does the causal impact of HS enrollment on children’s comprehensive outcomes (cognitive, social-emotional, and health) with an instrumental variable (IV) analysis higher than the impact of Head Start assignment and Head Start attendance?

Analysis

Using a series of linear regressions, we estimated the intention-to-treat (ITT) effect of Head Start assignment on children’s cognitive, social-emotional, and health outcomes. Using a similar regression model, association between Head Start attendance and children’s outcomes was estimated. Finally, using a simultaneous equation model with an instrumental variable, a causal effect of Head Start attendance on children’s outcomes was estimated (see Method section for more details).

Findings

Cognitive Outcomes

Table 3 showed the results of Head Start assignment (ITT), Head Start attendance, and IV analysis for children’s cognitive outcomes. Our hypothesis—Head Start improved children’s cognitive outcomes—were supported for Peabody picture vocabulary test score (PPVT), applied problem solving, spelling, letter-word identification, and were not supported for oral comprehension. Effect of attending the head start program is greater than that of being assigned to it. For Peabody picture vocabulary test score (PPVT), for example, getting assigned to the program is associated with 0.125 standard deviation point increase in the test score (SE = .03, p < .001). Actual attendance of Head Start increased 0.097 standard deviation points of the test score (SE = .03, p < .001). The size of the effect becomes bigger in the IV regression analysis, showing that attending the head start program increases the test score by 0.181 standard deviation points (SE = .04, p < .001). Effect of Head Start attendance for Applied problem solving skills were also higher (b = 0.154, SE = .03, p < .001) than effect of Head Start assignment (b = 0.137, SE = 0.03, p < .001). The size of the effect becomes bigger in the IV regression analysis, showing that attending the head start program increases the test score by 0.202 standard deviation points (SE = .05, p < .001). For Letter-Word Identification scores, the size of the effect becomes bigger in the IV regression (b = 0.329, SE = .05, p < .001) than that of attendance (b = 0.212, SE = .03, p < .001) and assignment (b = 0.223, SE = .03, p < .001). Attending the Head Start program has more positive effect on Spelling (b = 0.160, SE = .03, p < .001) than assigned the Head Start program (b = 0.153, SE = .03, p < .001) but lower than the effects of IV analysis (b = 0.224, SE = .05, p < .001).

Table 3.

Effects of Head Start Assignment, Head Start Attendance, and Instrumental Variable Analysis on Children’s Cognitive, Social, and Health Outcomes.

	Cognitive outcomes
(SE)	PPVT	Applied problem Solving	Spelling	Word identification	Oral comprehension
Head Start	0.125***	0.137***	0.153***	0.223***	0.001
Assignment	(0.027)	(0.031)	(0.032)	(0.031)	(0.027)
N	3680	3661	3695	3687	3623
Head Start	0.097***	0.154***	0.160***	0.212***	−0.030
Attendance	(0.027)	(0.030)	(0.032)	(0.031)	(0.026)
N	3675	3656	3690	3682	3618
Head Start	0.181***	0.202***	0.224***	0.329***	0.001
(IV)	(0.040)	(0.045)	(0.047)	(0.046)	(0.039)
N	3675	3656	3690	3682	3618
Robust F	3072.19	3035.87	3081.69	3066.24	2912
p-value	0.000	0.000	0.000	0.000	0.000

	Social outcomes
	Child–parent	Behavioral	Social	Social	Child–teacher
	Relationship	Problems	Competencies	Skills	Relationship
Head Start	0.045	−0.075*	−0.008	−0.008	0.051
Assignment	(0.034)	(0.031)	(0.032)	(0.032)	(0.046)
N	3513	3691	3692	3691	2526
Head Start	0.045	−0.087**	0.006	−0.036	0.003
Attendance	(0.033)	(0.031)	(0.031)	(0.031)	(0.046)
N	3513	3691	3692	3691	2480
Head Start	0.067	−0.111*	−0.012	−0.011	0.077
(IV)	(0.050)	(0.046)	(0.047)	(0.046)	(0.073)
N	3513	3691	3692	3691	2480
Robust F	2850.7	3078.09	3068.26	3086.27	983.09
p-value	0.000	0.000	0.000	0.000	0.000

	Health related outcomes
	Dental	General	Hearing	Vision	Primary
	Treatment	Health status	Check-ups	Check-ups	Care provider
Head Start	0.152***	0.015	0.212***	0.212***	0.004
Assignment	(0.016)	(0.013)	(0.017)	(0.017)	(0.005)
N	3690	3693	3519	3520	3692
Head Start	0.235***	0.012	0.267***	0.265***	0.006
Attendance	(0.016)	(0.012)	(0.016)	(0.016)	(0.005)
N	3690	3693	3519	3520	3692
Head Start	0.225***	0.022	0.312***	0.313***	0.005
(IV)	(0.024)	(0.019)	(0.024)	(0.024)	(0.008)
N	3690	3693	3519	3520	3692
Robust F	3051.92	3052.3	2917.08	2902.63	3056.04
p-value	0.000	0.000	0.000	0.000	0.000

* p<0.05, ** p<0.01, *** p<0.001.

Social-Emotional Outcomes

Table 3 showed the results of Head Start assignment (ITT), Head Start attendance, and IV analysis for children’s social-emotional outcomes. As we hypothesized, Health Start improved children’s behavioral problem scores. However, our hypothesis was not supported for child–parent relationship, social competencies, social skills, and child–teacher relationship. Effect of attending the head start program is greater than that of being assigned to it for children’s total behavioral problems scores. Getting assigned to the Head Start program is associated with 0.075 standard deviation point decrease (SE = .03, p < .001) in the behavioral problem scores. This decrease is more significant for the effects of Head Start enrollment (b = −0.087, SE = .03, p < .001). The size of the effect becomes bigger in the IV regression analysis, showing that attending the head start program decreases the behavioral problem scores by 0.111 standard deviation points (SE = .05, p < .001). Although child–parent relationship (.045, .045, .67) and child–teacher relationship (.051, .003, .077) scores increased from Head Start assignment (ITT), Head Start enrollment, to IV analysis, they were not statistically significant. Further, social competencies and social skills were not significantly affected by Head Start assignment, Head Start attendance, and IV analysis.

Health-Related Outcomes

As shown in Table 3, effect of attending the head start program is greater than that of being assigned to it. As we hypothesized, IV analysis indicated that Head Start attendance has the significant causal impacts on children’s dental, vision, and hearing treatment status. Our hypothesis however was not supported for general health status and attending primary care. For preventative dental health treatment, for example, by getting assigned to the Head Start program, the probability of dental treatment was increased by 15.2% points (b = .152, SE= .016, p< .001). With Head Start attendance, the probability of having dental treatment was increased by 23.5% points (b = .235, SE = .016, p < .001) and that of IV analysis was increased by 22.5% points (b = .225, SE= .024, p < .001). The probability of having hearing check-up was 21.2% points (b = .212, SE= .017, p < .001) with Head Start assignment and 26.7% points (b = .267, SE= .016, p < .001) with Head Start attendance. The probability of having hearing check-up was increased by 31.2% points (b = .312, SE= .024, p < .001) with the IV analysis. The similar result was obtained for vision check-up. The probability of having vision check-ups with IV analysis was increased by 31.3% points (b = 0.313, SE= .024, p < .001) which was also higher than effect of Head Start assignment and Head Start attendance.

Findings for Research Question 2: Factors Affecting Compliances

Hypothesis 2

Are child and family characteristics variables associated with the non-compliances of Head Start assignment? Does quality of care affect this association?

Analysis

To investigate the determinants of non-compliances to Head Start assignment, we regressed attrition and non-compliances on children’s baseline characteristics including socioeconomic and demographic factors. We did in two ways: without quality and with quality in the model.

Findings

No Quality of Care Variable in the Model

As shown in Table 4, out of original 4,442 children, 662 children did not complete Head Start Impact Study (attrition rate = 14.9%). Among 3780 children who remained in the HSIS in spring 2003, 3155 children complied with the original assignments and 625 children did not comply with the original assignments). Table 5, column 1 showed higher attrition among children in control group, older children (age 4), boys, children cared by younger providers, children in teenage mothers, children of less educated mothers, children living in urban area, and children living in low household risks. Table 5, column 2 indicated for the association between baseline characteristics and the status of non-compliances. Head Start assigned children were more likely to non-comply with the original assignment by 4.5% points (p < .001) than those who were not assigned. Children’s age (older children, 2.7% point, p < .05), those living in urban area (4.9% point, p < .01), and living in a higher family income (1.7% point, p < .01) were positively associated with non-compliance. As shown in Table 5, column 3, among children who were assigned to Head Start, older children (6.5% point, p < .001), children with lower pre-academic skills (4.3% point, p < .05), children speaking English at home (6.2% point, p < .05), and those living in urban area (5.1% point, p < .05) tended not to comply the assignment. For non-Head Start (control group) children, younger children (3.8% point, p < .05), boys (−3.7% point, p < .05), black children (5.6% point, p < .05), children living in urban area (4.5% point, p < .05), and children living a higher family income (3.2% point, p < .001) were associated with non-compliances.

Table 4.

Attrition and Noncompliance from Head Start Impact Study.

Variable	Obs	Mean	Std. Dev.	Min	Max
Attrition	4442	14.90%	0.356	0	1
Nocomply	3780	16.53%	0.372	0	1
		Enroll
		Treatment	Control	Attrition	Sum
Assign	Treatment	1929	430	287	2646
Assign	Control	195	1226	375	1796
	Sum	2124	1656	662	4442

Table 5.

Effects of Baseline Variables Predicting Attrition and Noncompliances With and Without Quality.

b(SE)	Attrition	Nocomply	Nocomply	Nocomply	Nocomply	Nocomply	Nocomply
b(SE)		Total	Head Start Assigned	Control Group	Total (Quality)	Head Start Assigned (Quality)	Control Group (Quality)
Head Start	−0.091***	0.045***			−0.081***
Assignment	(0.011)	(0.012)			(0.025)
Age	0.024*	0.027*	0.065***	−0.038*	0.034*	0.093***	−0.097***
	(0.011)	(0.013)	(0.017)	(0.019)	(0.014)	(0.012)	(0.028)
Gender	−0.022*	−0.024	−0.015	−0.037*	−0.014	−0.005	−0.040
	(0.011)	(0.012)	(0.016)	(0.019)	(0.013)	(0.012)	(0.029)
Special needs	0.019	0.023	0.003	0.060	0.044*	0.053**	0.022
	(0.016)	(0.020)	(0.025)	(0.033)	(0.020)	(0.018)	(0.043)
Academic skills	0.001	0.024	0.043*	−0.005	0.009	−0.011	0.059
Prior Head Start	(0.013)	(0.015)	(0.020)	(0.022)	(0.016)	(0.013)	(0.036)
Social skills	0.000	−0.010	−0.012	−0.007	−0.008	0.003	−0.017
Prior Head Start	(0.005)	(0.006)	(0.009)	(0.010)	(0.007)	(0.006)	(0.015)
Health status	0.001	−0.001	−0.003	0.001	0.000	0.002	0.011
Prior Head Start	(0.013)	(0.015)	(0.020)	(0.023)	(0.016)	(0.014)	(0.037)
White	ref	ref	ref	ref	ref	ref	ref
Black	0.017	0.002	−0.028	0.056*	0.031	−0.007	0.036
	(0.014)	(0.016)	(0.021)	(0.023)	(0.017)	(0.016)	(0.035)
Hispanic	0.005	0.011	0.003	0.027	0.008	−0.014	0.062
	(0.016)	(0.020)	(0.027)	(0.028)	(0.021)	(0.019)	(0.042)
Age of caregiver	−0.004***	0.000	−0.001	0.001	0.001	0.001	0.000
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)
Teenage mother	0.075***	0.011	0.009	0.003	0.034	0.017	0.013
	(0.018)	(0.020)	(0.027)	(0.029)	(0.022)	(0.020)	(0.046)
< High school	ref	ref	ref	ref	ref	ref	ref
High school	−0.034*	−0.017	−0.006	−0.030	0.002	0.009	−0.066
	(0.014)	(0.016)	(0.020)	(0.024)	(0.017)	(0.014)	(0.038)
> High school	−0.041**	−0.005	0.009	−0.028	0.016	0.049**	−0.082*
	(0.015)	(0.017)	(0.022)	(0.027)	(0.018)	(0.016)	(0.040)
Married	0.056***	−0.013	−0.020	−0.005	−0.001	−0.003	0.004
	(0.016)	(0.017)	(0.023)	(0.025)	(0.019)	(0.017)	(0.040)
Biological parents	−0.013	−0.023	−0.031	−0.013	−0.003	−0.003	−0.022
	(0.015)	(0.017)	(0.022)	(0.025)	(0.018)	(0.016)	(0.040)
English speaking	0.016	0.036	0.062*	−0.003	−0.004	0.012	0.017
	(0.019)	(0.023)	(0.031)	(0.034)	(0.024)	(0.023)	(0.049)
Immigrant	0.009	0.024	0.020	0.030	0.026	0.014	0.036
	(0.018)	(0.021)	(0.027)	(0.033)	(0.022)	(0.020)	(0.047)
Living in urban	0.028*	0.049**	0.051*	0.045*	0.021	0.000	0.038
	(0.014)	(0.015)	(0.021)	(0.022)	(0.017)	(0.014)	(0.037)
Family income	−0.007	0.017**	0.009	0.032***	0.019**	0.008	0.030*
	(0.004)	(0.006)	(0.007)	(0.009)	(0.006)	(0.005)	(0.013)
Low risk	ref	ref	ref	ref	ref	ref	ref
Moderate risk	−0.187***	0.007	−0.001	0.022	0.020	−0.005	0.024
	(0.014)	(0.018)	(0.024)	(0.028)	(0.020)	(0.017)	(0.046)
High risk	−0.231***	−0.012	−0.011	−0.002	−0.014	−0.018	−0.016
	(0.022)	(0.027)	(0.036)	(0.040)	(0.029)	(0.025)	(0.066)
Quality					−0.048***	−0.225***	0.150***
					(0.011)	(0.011)	(0.012)
Constant	0.190***	0.077*	0.096*	0.121*	0.144***	0.110**	0.364***
	(0.031)	(0.036)	(0.048)	(0.049)	(0.042)	(0.036)	(0.077)
R-squared	0.069	0.016	0.019	0.030	0.057	0.313	0.207
N	4442	3780	2359	1421	2963	2166	797

* p<0.05, ** p<0.01, *** p<0.001

Quality of Care Variable in the Model

We hypothesized that quality of care of different programs are associated with non-compliances. Figure 1 provides the distribution of the index of quality of care for different types of care program (Head Start, home-based care, and center-based care). Quality is a composite variable included items from 12 standardized and non-standardized measures. Overall, the quality of care at Head Start is the best; center-based care comes next, followed by home-based care, though some good home- and center-based care are better than poor Head Start care. Table 5, column 5 shows the results of non-compliance after adding quality of care in the model (excluding 817 children that were consisted of children cared by parents [n =748], children in home-based care [n = 68], and children in center-based care [n = 1]). Quality was negatively associated with non-compliance (b = −.048, SE = .04, p < .001), implying that children who were assigned to high quality service providers were more likely to comply with original assignment. Head Start assigned children were less likely to non-compliance to the original assignment (b = −0.081, SE = 0.02, p < .001). Children with special needs (b = .044, SE = .02, p < .05) and those living in a higher family income household (b = .019, SE = .01, p < .01) tended for more non-compliance. Within Head Start assigned children (Table 5, column 6), non-compliance children tended to be older children, children with special needs, and those whose mothers had more years of education. Within non-Head Start children (Table 5, column 7) shows quality (b = .150, SE = 0.01, p < .001) was positively associated with non-compliance. Higher non-compliance was found for younger children, children of mothers with less years of education, and children in higher income household.

Figure 1.

Density of qualify of care.

Table 6 showed what child and family characteristics were associated with quality of care. As shown in Table 6, column 1, Head Start assigned children (b = 1.019, SE = 0.04, p < .001), older children (b = 0.10, SE = 0.03, p < .01), children with special needs (b = 0.11, SE = 0.05, p < .05) attended higher quality care. Black children attended lower quality care (b = −0.13, SE = 0.04, p < .01) than white children. This was also found within Head Start assigned children (Table 6, column 2). Black children attended significantly lower quality Head Start (b = −.15, SE = 0.04, p < .001) than white Head Start children. Older age children (b = .193, SE = 0.028, p < .001) and children with special needs (b = .120, SE = 0.040, p < .01) attended higher quality Head Start than their counterparts. Children who did not assigned but attended Head Start (non-compliance) attended lower quality Head Start (b = −1.305, SE = 0.044, p < .001). The opposite effect was found for non-Head Start children (control group). As shown in Table 6, column 3, children who did not comply for the assignment attended higher quality care (b = 1.091, SE = 0.089, p < .001) than those who comply to the original control group assignment.

Table 6.

Baseline Variables and Noncompliance of Assignments Predicting Quality.

b (SE)	Quality (n=2,963)	Quality (Head Start =2,166)	Quality (Control=797)
Nocomply	−0.305***
	(0.046)
Head Start	1.019***
Assignment	(0.037)
Age	0.101**	0.193***	0.159*
	(0.033)	(0.028)	(0.076)
Gender	−0.021	−0.023	0.039
	(0.033)	(0.028)	(0.076)
Special needs	0.112*	0.120**	0.136
	(0.048)	(0.040)	(0.117)
Academic skills prior Head	−0.083*	−0.067*	−0.140
Start	(0.039)	(0.033)	(0.093)
Social skills prior Head	0.044**	0.042**	0.047
Start	(0.016)	(0.014)	(0.039)
Health status prior Head	−0.030	−0.008	−0.091
Start	(0.040)	(0.034)	(0.096)
White	ref	ref	ref
Black	−0.129**	−0.151***	−0.028
	(0.043)	(0.036)	(0.100)
Hispanic	−0.049	−0.054	−0.083
	(0.050)	(0.044)	(0.114)
Age of caregiver	0.004	0.002	0.007
	(0.002)	(0.002)	(0.005)
Teenage mother	−0.054	−0.045	−0.021
	(0.053)	(0.045)	(0.124)
< High school	ref	ref	ref
High school	−0.096*	−0.068	0.019
	(0.041)	(0.035)	(0.098)
>High school	−0.072	0.017	0.017
Married	−0.086	−0.051	−0.118
	(0.046)	(0.039)	(0.108)
Biological parents	0.068	0.011	0.188
	(0.044)	(0.037)	(0.104)
English speaking	0.033	0.084	−0.111
	(0.058)	(0.050)	(0.130)
Immigrant	−0.030	−0.005	−0.096
	(0.054)	(0.045)	(0.127)
Living in urban	−0.050	−0.057	−0.027
	(0.045)	(0.038)	(0.112)
Family income	0.009	−0.001	0.002
	(0.015)	(0.013)	(0.033)
Low risk	ref	ref	ref
Moderate risk	0.115*	−0.002	0.308**
	(0.048)	(0.041)	(0.117)
High risk	0.113	0.004	0.289
	(0.074)	(0.062)	(0.181)
Nocomply	−1.305***
(Head Start)	(0.044)
Nocomply		1.091***
(Control)		(0.089)
Constant	−0.576***	0.467***	−0.994***
	(0.098)	(0.080)	(0.217)
R-squared	0.256	0.322	0.197

Discussion

This study provides unique insight into childcare choice using nationally representative data from the Head Start Impact Study. Results identify possible contributors to families’ decisions to enroll in Head Start and begin to unpack the role of childcare quality in what factors may contribute to enrollment decisions. Findings also extend the literature examining the broad effects of Head Start participation on children’s outcomes, particularly by comprehensively examining whole-child outcomes across the cognitive, social-emotional, and health domains. All child outcomes were measured immediately after 1 year of Head Start, thereby equalizing the duration of children’s enrollment across participants. An examination of non-compliances suggested parents act in the childcare market using quality of programs as a significant choice index. Several child and family covariates enhance or deter the likelihood of Head Start enrollment, revealing barriers to connecting families to particular childcare programs.

Quality and Childcare Choice Among Head Start-Eligible Families

The study identified factors related to parents’ decision to enroll their children in Head Start, or chose another form of childcare, following assignment to the Head Start group in the HSIS study. First, it is important to note that between the Head Start-assigned treatment and non-Head Start control groups, far more participants did not comply with the assignment to Head Start and enrolled in other programs than participants who were assigned as controls yet chose to enroll in a Head Start program. Accordingly, assignment to the Head Start group was a significant predictor of non-compliance to the original care assignment in the HSIS study.

Quality was included in our analyses to better understand potential reasons for compliance and non-compliance and provided additional nuance to contextualize family decision-making. Head Start assignment was a significant predictor of non-compliance to the original assignment after accounting for program quality. Importantly, this comparison was limited to children enrolled in Head Start or another center-based care program, as quality ratings were not available for children in parental or other relative or non-relative home-based care. Within this subgroup, this finding suggests that quality, rather than care type, is particularly influential in families’ decisions about childcare enrollment for their children. Children who were assigned to higher quality Head Start were six times more likely to remain in Head Start, whereas those assigned to lower quality Head Start often moved to the control group. Despite the intention for overall higher quality of Head Start programs resulting from federal program mandates and strict oversight, the current study showed there existed a range of quality of Head Start programs, consistent with previous findings (Raver et al., 2008). This variation likely influenced family decision making. Families’ inclination to make decisions based on quality is beneficial, given associations between quality and child outcomes (Gelber & Isen, 2013).

Family focus on quality and its relation to outcomes suggests a need for universal implementation and evaluation of childcare quality standards across Head Start and similar programs available for low-income children. Fiscal and labor supports are also needed to ensure programs can engage in continuous improvement efforts such as ongoing professional development and technical assistance. The study suggests that though increasing availability of free early childhood education services is intended to increase equity for low-income children, quality is a variable that must be supported. Other research using nationally representative samples suggested that most children enrolled in Head Start were not situated in high-quality classrooms, and that children of ethnic minority families and parents with lower levels of education were more likely to be enrolled in lower-quality classrooms (Bulotsky-Shearer et al., 2012). Head Start must work to increase quality uniformly across programs, and efforts should particularly target programs serving families with significant and/or multiple risk factors.

Across all care settings, several other factors were related to non-compliance with the original care assignment, including older child age, urban residence, and higher family income level. In general, family decision-making around childcare may be more heavily weighted toward these factors than care type when deciding on the best fit for themselves and their children. However, it is necessary to contextualize these results by considering family compliance within Head Start- and non-Head Start-assigned groups.

Child Age

Child age was a discriminating factor in families’ decision-making, as non-compliance with Head Start was associated with older child age, and non-compliance with non-Head Start was associated with younger child age. Relatively greater options for childcare for 4-year-olds, such as publicly funded pre-kindergarten, may have swayed families to eschew their original Head Start assignment. It is possible that connection to the child’s public elementary school where they would attend kindergarten was appealing for families if the alternative program was housed or otherwise connected with this school. In 2012–2013, 28% of US children attended publicly funded preschool programs, while only 10% attended Head Start programs, and an additional 3% attended special education preschool programs. The remaining 59% did not attend any publicly funded early learning programs. For 3-year-old children, Head Start may be more appealing given its broad focus on whole child learning, health, parent involvement, and social services that meet families’ unique needs (National Head Start Association, 2020). Pragmatically, Head Start is among few publicly funded, national preschool programs available to three-year-old children.

Urban Residence

Urban residency appeared to influence families’ decisions to both attend and not attend Head Start, which suggests a complex and somewhat puzzling picture with respect to familial region and childcare choice. It is possible that proximity to childcare centers, whether Head Start or not, or availability of parental or non-parental home-based care, played a greater role in urban families’ choice to not comply with their original assignment more so than preference for one type of care over another. Families living in urban areas may have fewer options available for affordable childcare, leading more parents to enroll their children in freely available programs or make use of parental/relative care over fee-based options. Results from a national survey indicated that 47% of parents living in urban areas reported there were either no good choices for early child care programs or they did not know about early care options, compared to 39% of parents in suburban and town settings (National Center for Education Statistics, 2018).

Family Income

Analyses indicated that non-compliance with a non-Head Start assignment, that is, moving from the control group to a Head Start program, was associated with higher family income. This finding aligns with other research citing associations between childcare subsidies and use of center-based care over multiple care arrangements (Ros Pilarz, 2018). With increased purchasing power, families may be equipped with the option to enroll in a center-based program that will better meet their child and family needs, rather than making use of care arrangements that are responsive to limited time and monetary resources. Parents with lower incomes may require longer hours of childcare due to longer shifts and inflexible work schedules, which may influence the decision to choose another type of care over Head Start. In 2016, less than half of Head Start programs provided services 5 days per week, and for longer than 6 hours per day. Availability of these services contrasts considerable need; in 2014, 3- and 4-year-old US children attended center-based childcare an average of 163 hours per month, approximately 2,000 hours per year.

Likewise, given established associations between family income, parental education and children’s skills (e.g., Conway, Waldfogel, & Wang, 2018), Head Start may be a preferred option among parents who are informed about its benefits in increasing school readiness (Coley et al., 2014). National data suggests children of parents with higher educational attainment are more likely to be enrolled in part- and full-day preschool and prekindergarten programs (Child Trends, 2019). What these findings do not explain, however, is why parents may have enrolled their children in Head Start versus other full-day center-based programs. This decision likely harkens back to the earlier discussion of the relevance of childcare quality versus program type, with Head Start potentially being perceived as having higher quality than other programs.

Minoritized Children and Families

Findings regarding compliance also suggest Head Start may be favored over other programs for minoritized families. Among those originally assigned to Head Start, families who did not speak English as a home language were more likely to non-comply. For control group children, those identified as Black were more likely to non-comply with the control assignment and enroll in Head Start. Together, these findings suggest a mixed picture regarding the choice for Head Start over other available care arrangements for children from minoritized families. In a sample of African American families, positive regard for the schools’ efforts to listen to and provide activities for parents predicted parent involvement in elementary school (Overstreet, Devine, Bevans, & Efreom, 2005). Head Start’s orientation toward explicitly engaging parents and providing opportunities for families to participate in culturally responsive educational activities (National Center on Cultural and Linguistic Responsiveness, 2012) may play a role in drawing Black families to this service, given the need for flexible parent engagement practices that address barriers to involvement among minority families (Mendez, 2010).

Given the well-established achievement gap between Black and white children resulting from systemic oppression that begins in early childhood (e.g., Burchinal et al., 2011), parents of Black children may have also been particularly attuned to program quality characteristics, and chose not to enroll their child if program quality was perceived to be low. It should be noted, however, that research examining racial and ethnic differences in preferences of childcare type have been mixed. For instance, Tang and colleagues (2012) found that white mothers were more likely to use Head Start and other center-based care for their children than Latina but not African American mothers. Other findings suggest that use of center-based versus other forms of care may be a result of social and political influence rather than familial racial/ethnic heritage. Together, this work suggests racial and ethnically based differences in care preference are complex and influenced by several individual, cultural, and societal factors.

Non-English speaking families may be more likely to make use of familial or non-familial care arrangements, given established disparities in preschool enrollment among Latino immigrant versus US-born Latino parents, in which the latter are more likely to place their children in the care of a relative or parent. Similar findings are reported for families with lower English language proficiency, with these families favoring parental care over preschool enrollment (Ansari, 2018). Though our sample included diverse families representing many cultural and ethnic groups, these findings suggest that sociocultural values and practices play a large role in preschool selection processes, reflected in current findings.

Child Skills and Special Needs

Children with lower pre-academic skills at the beginning of the preschool year were from families more likely to not comply with a Head Start assignment. This finding may be reflective of families’ desire for specialized care, such as enrollment in an early childhood special education classroom, or parental/relative care. This inference is supported by our analysis that accounted for program quality, which suggested similar results in that Head Start non-compliance was associated with children who were identified as having special needs. It is possible that families of children with these characteristics perceived a more intense need for higher-quality services for their children, thus chose not to enroll when perceptions suggested available Head Start programs were not of high quality. For instance, parents of children with disabilities site efforts among early childhood programs to establish relationships as central to their child’s experience in typical education settings (Warren, 2017). It may be that Head Start programs of lower quality had reputations for poorly engaging parents, which then influenced parents’ decisions to not enroll their children with special needs.

Positive Effects of Head Start for Attendees

Another contribution of the current study is its analysis of comprehensive child outcomes, including children’s health in addition to the traditional analysis of cognitive and social-emotional outcomes. Results indicated positive effects of Head Start participation on children’s health, including regular care for dental health, vision, and hearing. Puma and colleagues (2010), in their primary analysis of the HSIS data, found a positive effect of Head Start access on three and four year-old children’s receipt of dental services over the course of the year, though effects on other health-related outcomes (e.g., health insurance coverage) were not evident after the first year of Head Start access. Similarly, low-income children enrolled in Head Start were found to be three times more likely to receive a dental check than their non-Head Start peers (Lee et al., 2014). This study adds to the literature by identifying positive effects of Head Start attendance on other, non-dental health outcomes, which were not evident in previous studies.

Consistent with other research (Abbott-Shim et al., 2003; Puma et al., 2010; Wen, Leow, Hahs-Vaughn, Korfmacher, & Marcus, 2012), results of the current analysis indicated positive effects of Head Start enrollment on children’s early mathematics, reading, and language skills. These findings align with that of the primary evaluation of the HSIS conducted by Puma and colleagues (2010). However, results of the current analysis identified a significant increase in the effect of Head Start attendance over that of Head Start assignment for vocabulary, math problem solving, and spelling skills. Research with other Head Start samples demonstrated a positive growth trajectory for cognitive skills in children attending Head Start (Wen et al., 2012), faster growth for children attending Head Start versus those placed on a wait list (Abbott-Shim et al., 2003), and reductions in achievement gaps for Head Start attendees that were not evident for children who did not attend preschool (Kreisman, 2003).

The current findings also demonstrated positive effects for children’s social-emotional outcomes, including reduced behavioral concerns. Again, positive effects of Head Start attendance outweighed that of Head Start assignment for children’s behavioral problems, hyperactivity, and child–parent closeness. These results help to clarify mixed findings regarding the effects of Head Start on children’s social-emotional competencies. Similar to the current study, Hubbs-Tait and colleagues (2002) found that attendance in Head Start positively predicted children’s teacher-rated sociability regardless of family risk factors. However, results from the primary HSIS evaluation were less conclusive particularly for 4-year-olds (Puma et al., 2010) and other research suggests children who attended Head Start had greater behavioral issues than their peers who were primarily cared for by their parents (Lee et al., 2014).

There is a need for future research clarifying the ambiguity in Head Start effects on children’s social-emotional and behavioral competencies. Varied findings across the current and previous studies (e.g., Lee et al., 2014; Puma et al., 2010) might be the result of differing reference groups and control variables. Lee et al. (2014) used data from the Early Childhood Longitudinal Study – Birth Cohort (ECLS–B), which was not a randomized design. Income among control groups varied widely between the two samples (e.g., among children cared for by both parents, ≤ $35,000 = 90% for HSIS vs. 65% for ECLS-B data), while the Head Start groups from both datasets had similar income distributions. Several studies document negative social-behavioral outcomes associated with children’s time in center-based care (Dearing et al., 2015; NICHD ECCRN, 2003). In contrast, others demonstrated null or positive associations (Atteberry, Bassok, & Wong, 2019; Zachrisson, Dearing, Lekhal, & Toppelberg, 2013). Various explanations exist for the apparent effects on children’s social-emotional outcomes, including reporting bias, selection bias, peer effects in classrooms, and social learning processes (Huston, Bobbitt, & Bentley, 2015; Pingault et al., 2015). Possibly, Head Start curriculum has focused more on cognitive functioning so children spend more time in academic activities and less in socio-emotional learning which could lead to poorer socio-emotional outcomes. Clearly more focused study on children’s social-emotional outcomes that captures children’s experiences, Head Start, and potential mechanisms linking these, should be addressed in a separate future study.

Study Limitations

There was no direct parent report of the reason for non-compliance to the original assignment available in the HSIS data. Thus, this study used several available covariates to identify predictors associated with non-compliance to the Head Start or control groups. Moreover, there was no measure of childcare quality for children cared by parents, relatives or non-relatives in home-based settings. Thus, a sizeable number of cases could not be included in analyses addressing childcare quality. Despite the benefits of including a measure of quality to identify nuances in the current study results, there is a need for future research that incorporates quality of home-based care in comparative analyses.

Discussion and Applications to Practice

Findings shed light on the nuance in what factors may influence families’ decisions to enroll their children in Head Start versus other center- or home-base care programs. Childcare programs, including Head Start, should consider these findings in addressing competition within the childcare market and reaching families at risk. First, our findings align with the body of evidence suggesting that childcare quality, more so than program type, is a primary factor in swaying families’ decisions to enroll children in a particular care program. Quality varies substantially both between center-based care settings including Head Start, and across centers within Head Start. Teacher professional development and quality improvement initiatives should focus on measuring and improving program quality for all care settings, both to improve the richness of children’s experiences and to increase enrollment and financial viability for struggling centers. There is also a vital need for research investigating the quality of home-based care and its relation to child outcomes, as this construct is rarely studied in the extant literature.

The decision to enroll in Head Start over other programs was influenced by higher family income and younger child age in this sample. Families with higher incomes may have greater purchasing power, therefore providing the opportunity to enroll in a program that is perceived as being more beneficial to their children’s development, rather than a program that offers the most convenient location or hours, for instance. Childcare stipends for lower-income families may increase purchasing power and therefore lead to more families enrolling in higher-quality care programs. Findings also suggest that families of younger children were more likely to choose Head Start over other programs, while the opposite was the case for families of older children. Families of older children may favor other publicly funded programs that broadcast clear objectives such as increasing reading readiness. Head Start centers may benefit from more targeted messaging and outreach demonstrating its potential for positive child outcomes.

Results also indicated lower likelihood of enrollment in Head Start among families whose used a non-English home language, while families of Black children were more likely to comply with a Head Start assignment. These findings suggest a need for Head Start programs to increase outreach and cultural responsivity to the needs of non-native English speaking families. Given increasingly diversity of US families, as well as ongoing concerns with racial trauma (Comas-Diaz et al., 2019) all childcare programs should engage in active, ongoing efforts to increase cultural responsivity and support the needs of minoritized families.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Kyunghee Lee

References

Abbott-Shim

Lambert

McCarty

(2003). A comparison of school readiness outcomes for children randomly assigned to a Head Start program and the program’s wait list. Journal of Education for Students Placed at Risk, 8(2), 191–214. https://doi.org/10.1207/S15327671ESPR0802_2

Achenbach

T. M.

Edelbrock

Howell

C. T.

(1987). Empirically based assessment of the behavioral/emotional problems of 2-3-year old children. Journal of Abnormal Child Psychology, 15, 629–650. https://doi.org/10.1007/BF00917246

Angrist

Imbens

G. W.

Rubin

D. B.

(1996). Identification of causal effects using instrumental variables. Journal of the American Statistical Association, 91(434), 444–455. https://doi.org/10.2307/2291629

Ansari

(2018). The selection of preschool for immigrant and native-born Latino families in the United States. Early Child Research Quarterly, 41, 149–160. https://doi.org/10.1016/j.ecresq.2017.07.002

Atteberry

Bassok

Wong

V. C.

(2019). The effects of full-day prekindergarten: Experimental evidence of impacts on children’s school readiness. Educational Evaluation and Policy Analysis, 41, 537–562.

Barnett

W. S.

Hustedt

J. T.

(2005). Head Start’s lasting benefits. Infants & Young Children, 18, 16–24. https://doi.org/10.1097/00001163-200501000-00003

Besharov

D. J.

Higney

C. A.

(2007). Head Start: Mend it, don’t expand it (yet). Journal of Policy Analysis and Management, 26, 678–681. https://doi.org/10.1002/pam.20272

Bierman

K. L.

Domitrovich

C. E.

Nix

R. L.

Gest

S. D.

Welsh

J. A.

Greenberg

M. T.

Gill

(2008). Promoting academic and social-emotional school readiness: The HS REDI program. Child Development, 79, 1802–1817.

Bloom

H. S.

Weiland

(2015, March) Quantifying variation in head start effects on young children’s cognitive and socio-emotional skills using data from the national head start impact study. New York, NY: MDRC.

10.

Bonuck

K. A.

Schwartz

Schechter

(2016). Sleep health literacy in HS families and staff: Exploratory study of knowledge, motivation, and competencies to promote healthy sleep. Sleep Health, 2, 19–24. https://doi.org/10.1016/j.sleh.2015.12.002

11.

Bryant

D. M.

Burchinal

Lau

L. B.

Sparling

J. J.

(1994). Family and classroom correlates of head start children’s developmental outcomes. Early Childhood Research Quarterly, 9, 289–309. https://doi.org/10.1016/0885-2006(94)90011-6

12.

Bulotsky-Shearer

R. J.

Wen

Faria

A.-M.

Hahs-Vaughn

D. L.

Korfmacher

(2012). National profiles of classroom quality and family involvement: A multilevel examination of proximal influences on Head Start children’s school readiness. Early Childhood Research Quarterly, 27, 627–639. https://doi.org/10.1016/j.ecresq.2012.02.001

13.

Burchinal

McCartney

Steinberg

Crosnoe

Friedman

S. L.

McLoyd

NICHD Early Child Care Research Network . (2011). Examining the black–white achievement gap among low-income children using the NICHD study of early child care and youth development. Child Development, 82, 1404–1420. https://doi.org/10.1111/j.1467-8624.2011.01620.x

14.

Child Trends Databank . (2019). Preschool and prekindergarten. Available at: https://www.childtrends.org/?indicators=preschool-and-prekindergarten

15.

Chor

(2018). Multigenerational head start participation: An unexpected marker of progress. Child Development, 89, 264–279. https://doi.org/10.1111/cdev.12673

16.

Coley

R. L.

Votruba-Drzal

Collins

M. A.

Miller

(2014). Selection into early education and care settings: Differences by the developmental period. Early Childhood Research Quarterly, 29, 319–332. https://doi.org/10.1016/j.ecresq.2014.03.006

17.

Comas-Díaz

Nagayama Hall

Neville

H. A.

(2019). Racial trauma: Theory, research, and healing: Introduction to the special issue. American Psychologist, 74, 1–5. https://doi.org/10.1037/amp0000442

18.

Conway

Waldfogel

Wang

(2018). Parent education and income gradients in children’s executive functions at kindergarten entry. Children and Youth Services Review, 91, 329–337. https://doi.org/10.1016/j.childyouth.2018.06.009

19.

Dearing

Zachrisson

H. D.

Nærde

(2015). Age of entry into early childhood education and care as a predictor of aggression: faint and fading associations for young Norwegian children. Psychological Science, 26(10), 1595–1607. https://doi.org/10.1177/0956797615595011

20.

Early

D. M.

Burchinal

M. R.

(2001). Early childhood care: Relations with family characteristics and preferred care characteristics. Early Childhood Research Quarterly, 16(4), 475–497. https://doi.org/10.1016/S0885-2006(01)00120-X

21.

Garces

Thomas

Currie

(2002). Longer-term effects of Head Start. American Economic Review, 92, 999–1012. https://doi.org/10.1257/00028280260344560

22.

Gelber

Isen

(2013). Children’s schooling and parents’ behavior: Evidence from the HS Impact Study. Journal of Public Economics, 101, 25–38. https://doi.org/10.1016/j.jpubeco.2013.02.005

23.

Hubbs-Tait

Culp

A. M.

Huey

Culp

Starost

H.-J.

Hare

(2002). Relation of Head Start attendance to children’s cognitive and social outcomes: Moderation by family risk. Early Childhood Research Quarterly, 17, 539–558. https://doi.org/10.1016/S0885-2006(02)00189-8

24.

Huston

A. C.

Bobbitt

K. C.

Bentley

(2015). Time spent in child care: How and why does it affect social development? Developmental Psychology, 51, 621–634. https://doi.org/10.1037/a0038951

25.

Improving Head Start for School Readiness Act of 2007, Pub.L.No.110-134, 42 U.S.C. § 121, (2007). https://www.congress.gov/110/plaws/publ134/PLAW-110publ134.pdf

26.

Kensinger Rose

Elicker

(2008). Parental decision making about child care. Journal of Family Issues, 29, 1161–1184. https://doi.org/10.1177/0192513X07312103

27.

Kline

Walters

C. R.

(2016). Evaluating public programs with close substitutes: The case of head start, The Quarterly Journal of Economics, 131, 1795–1848, https://doi.org/10.1093/qje/qjw027

28.

Kreisman

M. B.

(2003). Evaluating academic outcomes of Head Start: An application of general growth mixture modeling. Early Childhood Research Quarterly 18, 238–254. https://doi.org/10.1016/S0885-2006(03)00027-9

29.

Lamb-Parker

Piotrkowski

C. S.

Baker

A. J. L.

Kessler-Sklar

Clark

Paey

(2001). Understanding barriers to parent involvement in Head Start: A research-community partnership. Early Childhood Research Quarterly, 16, 35–51. https://doi.org/10.1016/S0885-2006(01)00084-9

30.

Lee

V. E.

Brooks-Gunn

Schnur

Liaw

F-R.

(1990). Are head start effects sustained? A longitudinal follow-up comparison of disadvantaged children attending Head Start, no preschool, and other preschool programs. Child Development, 61, 495–507. https://doi.org/10.1111/j.1467-8624.1990.tb02795.x

31.

Lee

Zhai

Brooks-Gunn

Han

W.-J.

Waldfogel

(2014). Head start participation and school readiness: Evidence from the early childhood Longitudinal study-birth cohort. Developmental Psychology, 50, 202–215. https://doi.org/10.1037/a0032280

32.

Mackintosh

B. B.

McCoy

D. C.

(2019). Exploring social competence as a mediator of Head Start’s impact on children’s early math skills: Evidence from the HS Impact Study. Early Education and Development, 30, 655–677. https://doi.org/10.1080/10409289.2019.1576156

33.

Mendez

J. L.

(2010). How can parents get involved in preschool? Barriers and engagement in education by ethnic minority parents of children attending Head Start. Cultural Diversity and Ethnic Minority Psychology, 16, 26–36. https://doi.org/10.1037/a0016258

34.

Morris

P. A.

Connors

Friedman-Krauss

McCoy

D. C.

Weiland

Feller

Page

(2018). New findings on impact variation from the Head Start Impact Study: Informing the scale-up of early childhood programs. AERA Open, 4, 1–16. https://doi.org/10.1177/2332858418769287

35.

National Center for Education Statistics . (2018, May). Early childhood care arrangements: Choices and costs. https://nces.ed.gov/programs/coe/indicator_tca.asp

36.

National Center on Cultural and Linguistic Responsiveness (2012). Head Start cultural and linguistic responsiveness resource catalogue. Vol 3. Bank Street College of Education. Retrieved from: https://eclkc.ohs.acf.hhs.gov/sites/default/files/pdf/resource-catalogue-cultural-linguistic-responsiveness.pdf

37.

National HS Association (2020). The HS model. Retrieved from: https://www.nhsa.org/why-head-start/head-start-model

38.

National Institute of Child Health and Human Development, Early Child Care Research Network . (2003). Does amount of time spent in child care predict socioemotional adjustment during the transition to kindergarten? Child Development, 74, 976–1005. https://doi.org/10.1111/1467-8624.00582

39.

Overstreet

Devine

Bevans

Efreom

(2005). Predicting parental involvement in children’s schooling within an economically disadvantaged African American sample. Psychology in the Schools, 42, 101–111. https://doi.org//10.1002/pits.20028

40.

Peyton

Jacobs

O’Brien

Roy

(2001). Reasons for choosing child care: Associations with family factors, quality, and satisfaction. Early Childhood Research Quarterly, 16, 191–208. https://doi.org/10.1016/S0885-2006(01)00098-9

41.

Pianta

R. C.

(1992). Child-parent relationship scale. Charlottesville, VA: University of Virginia

42.

Pingault

J.B.

Tremblay

R.E.

Vitaro

Japel

Boivin

Côté

S.M.

(2015). Early Nonparental care and social behavior in elementary school: Support for a social group adaptation hypothesis. Child Development, 86, 1469–1488.

43.

Puma

Bell

Cook

Head

Lopez

Zill

Shapiro

(2005). Head Start impact study: First year findings. Washington, DC: U.S. Department of Health and Human Services, Administration for Children and Families.

44.

Puma

Bell

Cook

Heid

Shapiro

Broene

Jenkins

(2010). Head start impact study: Final report. Washington, DC: U.S. Department of Health and Human Services, Administration for Children and Families.

45.

Raver

C. C.

Jones

S. M.

Li-Grining

C. P.

Metzger

M. W.

Smallwood

Sardin

(2008). Improving preschool classroom processes: Preliminary findings from a randomized trial implemented in Head Start settings. Early Childhood Research Quarterly, 23, 10–26. https://doi.org/10.1016/j.ecresq.2007.09.001

46.

Ros Pilarz

(2018). Childcare subsidy programs and child care choices: Effects on the number and type of arrangements. Children and Youth Services Review, 95, 160–173. https://doi.org/10.1016/j.childyouth.2018.10.013

47.

Schrank

F. A.

McGrew

K. S.

Woodcock

R. W.

(2001). Technical Abstract (Woodcock-Johnson III assessment service bulletin no. 2). Itasca, IL: Riverside Publishing.

48.

Seven

Ogelman

(2014). The Reliability-validity studies for the Student-Teacher Relationship Scale (STRS). European Journal of Research on Education, 2(2), 179–179. https://doi.org/10.15527/ejre.201426262

49.

Stuart

E. A.

Rhodes

(2017). Generalizing treatment effect estimates from sample to population: A case study in the difficulties of finding sufficient data. Evaluation Review, 41, 357–388. https://doi.org/10.1177/0193841X16660663

50.

Subramanian

S. V.

Kim

Christakis

N. A.

(2018). The “average” treatment effect: A Construct ripe for retirement. A commentary on Deaton and Cartwright. Social Sciences & Medicine, 210, 77–82. https://doi.org/10.1016/j.socscimed.2018.04.027

51.

Tang

Coley

R. L.

Votruba-Drzal

(2012). Low-income families’ selection of child care for their young children. Children and Youth Services Review 34, 2002–2011. https://doi.org/10.1016/j.childyouth.2012.06.012

52.

U.S. Department of Health and Human Services, Administration for Children and Families (2010). Head start impact study. Technical report. Washington, DC. www.acf.hhs.gov/sites/default/files/documents/opre/hs_impact_study_tech_rpt.pdf

53.

Warren

S. R.

(2017). Hidden voices: Parents’ perspectives on the barriers to and facilitators of inclusion on their preschool children with disabilities. In K.

Scorgie

Sobsey

(Eds.), Working with families for inclusive education (International perspectives on Inclusive education, volume, 10) (pp. 151–174). Bingley: Emerald Publishing Limited.

54.

Wen

Leow

Hahs-Vaughn

D. L.

Korfmacher

Marcus

S. M.

(2012). Are two years better than one year? A Propensity score analysis of the impact of Head Start program duration on children’s school performance in kindergarten. Early Childhood Research Quarterly, 27, 684–694. https://doi.org/10.1016/j.ecresq.2011.07.006

55.

Wooldridge

(2010). Econometric analysis of cross section and panel data (2nd ed.). MIT Press

56.

Zachrisson

H. D.

Dearing

Lekhal

Toppelberg

C. O.

(2013). Little evidence that time in child care causes externalizing problems during early childhood in Norway. Child Development, 84, 1152–1170. https://doi.org/10.1111/cdev.12040

57.

Zhai

Brooks-Gunn

Waldfogel

(2014). Head Start’s impact is contingent on alternative type of care in comparison group. Developmental Psychology, 50, 2572–2586. https://doi.org/10.1037/a0038205

58.

Zigler

Styfco

S. J.

(2001). Extended childhood intervention prepares children for school and beyond. Journal of the American Medical Association, 285, 2378–2380. https://doi.org/10.1001/jama.285.18.2378

59.

Zill

(1990). Behavior problems based on parent report. Washington, DC: Child Trends. Retrieved from: http://www.childtrends.org/Files/Child_Trends-1990_11_01_ES_BehaviorProb.pdf