Examining the Mixed-Delivery System During Pre-K Expansion in North Carolina

Abstract

State-funded pre-K programs have proliferated over the past several decades. Many states have relied on a mixed-delivery approach, wherein pre-K is provided across diverse childcare settings, including public schools, community-based organizations, and private childcare centers. Leveraging data from a natural experiment (2001–10) of North Carolina’s Pre-Kindergarten Program, we assessed county-level variation in the allocation of funded slots to different pre-K settings and the dispersion of slots within or across centers. We found that heterogeneity in the implementation of the mixed-delivery system had minimal effects on children’s fifth grade achievement, suggesting fairly consistent effects across the system. When small differences emerged, they suggested that North Carolina’s Pre-Kindergarten Program funding was most beneficial for student achievement when allocated to private for-profit and nonprofit centers.

Keywords

achievement early childhood educational policy quasi-experimental analysis

Introduction

Over the past several decades, publicly funded pre-K programs have significantly expanded across the United States (e.g., Cascio, 2021). In addition to federally funded initiatives, such as Head Start, 44 states operate state-funded pre-K programs (Friedman-Krauss et al., 2022). These public investments attempt to capitalize on the foundational and malleable nature of early childhood to promote subsequent development, particularly among children from historically marginalized and underserved groups. Indeed, some evidence suggests that early intervention efforts may produce intermediate and long-term effects on achievement and social-emotion-related outcomes (e.g., Chetty et al., 2011; Deming, 2009; Gray-Lobe et al., 2023).

Despite this long-term potential, findings across the pre-K evaluation literature are quite variable (e.g., Lipsey et al., 2018; Phillips et al., 2017; Watts et al., 2023; Weiland et al., 2020). Integral to our understanding of mixed effectiveness is the acknowledgment that pre-K is not monolithic in nature; there is no single form of pre-K that exists across states or settings. Thus, examining varied implementation of pre-K may be key to understanding when and where programs will produce positive (e.g., Gray-Lobe et al., 2023) or negative (e.g., Durkin et al., 2022) effects.

Our study sought to understand whether variation in the delivery settings of a large state-funded pre-K program is related to the academic achievement of affected children in the state. Some states place pre-K programs primarily in elementary schools, essentially adding an additional year of education to primary school (e.g., Tennessee and Oklahoma). In contrast, other states invest within the existing infrastructure through a mixed-delivery approach in which pre-K is implemented in Head Start and community-based centers alongside public school classrooms (e.g., North Carolina). Across these different settings, some quality standards remain consistent and commonplace (see, e.g., benchmarks in National Institute for Early Education Research), but programs often diverge in their core programmatic components (e.g., curriculum requirements) and implementation features (e.g., student–teacher ratio and childcare setting). Moreover, program implementation also differs within a given state because counties, cities, and individual sites often determine how to implement pre-K programming.

An important feature of the mixed-delivery system is that it could provide ample childcare choices and flexibility to meet each family’s needs while also financially supporting the early educational system that has historically cared for young children. Although a sizable body of research has qualitatively explored differences across pre-K settings (e.g., Bellm et al., 2002; Desimone et al., 2004; Early et al., 2005; Little, 2020; Wilinski, 2017), policy researchers have relatively little empirical insight into whether this diverse provision of early education drives heterogeneity in program impacts in the short and long term.

Our study explored this practical variation in state approaches to pre-K scale-up to better understand pre-K program effects. We used county-level variation in the scale-up of North Carolina’s Pre-Kindergarten Program (NC Pre-K), a state-funded program in which counties disburse funding for pre-K slots across diverse childcare settings, paired with unusually rich historical administrative records derived over an 18-year period. We focused on differences in the types of centers that counties fund (e.g., Head Start, public schools, and private centers) and differences in the concentration of funded slots in the centers approved for NC Pre-K (e.g., Do counties allocate their funded slots within a few centers or do they disperse them widely across centers?). We investigated the population- and individual-level consequences of this diverse system by examining associations between these policy choices and the achievement of all children as well as the more direct effects of such decisions on the children who were funded to participate in NC Pre-K.

Overall, we found limited evidence indicating that children differentially benefited from the allocation of funds to any specific childcare setting. To the extent that differences were observed, they suggested that NC Pre-K funds were slightly more beneficial when provided to private for-profit and nonprofit centers. These settings also had a greater concentration of funded slots, which also was associated with somewhat stronger fifth grade achievement among pre-K attendees; however, this finding was not observed in the analyses for all age-eligible children. We view these findings as largely descriptive in nature, and we discuss how they may be understood in the context of future research and policy implementation regarding state pre-K expansion.

Literature Review

The growing demand for publicly funded pre-K programs has prompted a shift toward a mixed-delivery approach (e.g., Barnett et al., 2003). This diverse system mirrors other social service sectors (e.g., medical, healthcare, behavioral health, and Individuals with Disabilities Education Act [IDEA] Part C) that have encouraged collaboration among public, private, and community providers to maximize the delivery and accessibility of a particular service. Within the context of early childhood education (ECE) mixed-delivery systems, public investments are distributed to a myriad of established preschool providers within the community, including public schools, Head Start grantees, and private childcare settings (e.g., for-profit centers, nonprofit centers, and faith-based organizations; e.g., Garver et al., 2023; Weiland et al., 2024).

Expanding in such a heterogeneous way is beneficial for the scale-up of public efforts because states can capitalize on the use of existing resources (e.g., staff, funding, classroom space, and playgrounds) and reduce the time and financial costs associated with opening new, program-specific classrooms (e.g., Ackerman et al., 2009; Garver et al., 2023; Weiland et al., 2024). In addition to efficiencies for program scale-up, delivering pre-K across mixed settings allows programs to be flexible and responsive to the various needs of the community. As a result, parents are afforded greater autonomy over their childcare decisions, which further facilitates alignment between the childcare setting and the family’s interests and needs (e.g., proximity to home/work, childcare hours, cultural values, and wraparound services).

However, from the policymaker’s perspective, one may be interested in whether these differences may systematically influence the effectiveness of public preschool investments. A primary challenge is that families’ childcare choices are not random, and childcare opportunities are not equitably located across communities. In turn, correlational evidence suggests that child outcomes vary by pre-K program setting (e.g., Ansari et al., 2020; Peisner-Feinberg et al., 2019; Weiland et al., 2024), which may be due to underlying observed differences in structural and process quality (e.g., Bellm et al., 2002; Desimone et al., 2004; Early et al., 2005; Little, 2020; Wilinski, 2017). Indeed, qualitative research has highlighted the perception of higher-quality care in public school settings because teachers often have higher levels education, pay parity with K–12 teachers, and lower levels of teacher turnover (e.g., Bellm et al., 2002; Desimone et al., 2004; Early et al., 2005; Little, 2020; Wilinski, 2017). Such settings also offer greater alignment and stability between pre-K and kindergarten, which may help to confer additional benefits for children (e.g., Desimone et al., 2004; Kagan & Kauerz, 2007; Little, 2020).

Similarly, empirical work also has suggested that public pre-K offered in school-based settings may be associated with more beneficial child outcomes than public pre-K delivered in community-based centers (Ansari et al., 2017, 2020; Magnuson et al., 2007). For example, a recent meta-analysis of five mixed-delivery programs found that children in public school classrooms generally demonstrated more rapid growth in math skills during pre-K than children in community childcare centers, with positive, nonsignificant associations on end-of-pre-K language and inhibitory control (Weiland et al., 2024). These correlational findings were echoed in earlier work by Peisner-Feinberg et al. (2019), who reported faster rates of learning on language, math, and behavioral skills among children who attended Georgia’s universal pre-K program in public school settings compared with those who attended community-based settings (but see Henry et al. [2006] for an opposite pattern of effects).

Even still, patterns of fadeout and persistence as a function of preschool setting are unclear. Children who attended public school–based Georgia pre-K showed larger initial impacts at the end of pre-K but faster fadeout on achievement measures by the end of second grade relative to their peers in community-based settings. Such descriptive findings highlight the potential for impacts to persist longer among children who attended community-based settings despite the smaller program effects observed initially. However, it is unclear the extent to which these divergent persistence rates may be due to family selection into different pre-K settings and the quality of program provision. Thus, more work is needed to understand the long-term effects of pre-K attendance within the mixed-delivery system and the extent to which provider type is differentially predictive of subsequent achievement.

Another feature of mixed delivery systems is the extent to which state-funded slots are concentrated into a smaller set of providers that create classrooms with mostly income-eligible children or whether slots are distributed more equally across the market of pre-K providers. Because childcare providers blend ECE funding sources to support a full day of care or to provide a broader set of services (e.g., tuition, Title I, or Head Start; e.g., Barnett, 2018; Duer & Jenkins, 2022), the mix of peers within a pre-K center may vary widely. In the case of an income-targeted pre-K program, concentrating funds within a few centers or funding stand-alone classrooms may create more segregated learning environments wherein state-funded children are exposed to only other state-funded children. This not only would limit opportunities for socioeconomic peer effects but also would decrease the potential for spillover impacts from public investment in pre-K on all age-eligible children. In contrast, wider disbursements throughout the mixed-delivery system could promote integration across socioeconomic and racial/ethnic groups, which could generate opportunities for funded children to benefit from exposure to more advantaged peers (e.g., Botvin et al., 2024; Weiland & Yoshikawa, 2014). Further, these integrated learning environments also may confer benefits to nonfunded children because participating centers must adhere to program quality guidelines to receive public funding (e.g., Dodge et al., 2017).

Taken together, mixed-delivery systems aim to improve equitable access to publicly funded pre-K by providing ECE within diverse early childhood settings. At the same time, research has suggested that variation in the mixed-delivery approach can lead to significant differences in pre-K quality, program impacts, and the types of children receiving care. Given the continued demand for accessible state-funded programs, more research is needed to assess the longer-term effects of this policy variation.

NC Pre-K

North Carolina’s Pre-Kindergarten Program is a targeted, state-funded pre-K program that began in 2001. Originally titled, “More at Four,” the program was designed to promote school readiness among disadvantaged 4-year-old children in North Carolina (Peisner-Feinberg, 2003; Peisner-Feinberg & Schaaf, 2009). Children are considered eligible for the program if their gross family income is at or below 75% of the state median income, with eligibility extended to children who meet additional inclusion criteria (e.g., developmental disability/delay, physical disability, and military parent; Peisner-Feinberg, 2003).

NC Pre-K is implemented using a mixed-delivery approach in which eligible children receive a funded slot to attend preschool at an existing certified early childhood setting (e.g., Head Start, public school, or private for-profit or nonprofit centers that meet the state’s certification criteria) within their residing county (Barnett, 2018; Peisner-Feinberg & Schaaf, 2009). As a result, NC Pre-K is provided across diverse childcare settings (i.e., public schools, Head Start centers, and for-profit and nonprofit community centers) in which funded and nonfunded children attend. Because pre-K centers must comply with state guidelines to receive NC Pre-K funding (e.g., classroom curricula, teacher training and education levels, and class size; Peisner-Feinberg & Schaaf, 2009), these public investments have the potential to confer benefits to funded children as well as nonfunded children who attend certified centers with funded peers.

Research on the short- and long-term effectiveness of NC Pre-K has been well documented for over a decade. These studies have leveraged the state’s allocation of pre-K funding using a natural experiment design (e.g., Dodge et al., 2017; Ladd et al., 2014; Watts et al., 2023; see Appendix Figure A2 in the online version of the journal) and have consistently reported positive effects of NC Pre-K funding on academic achievement. Despite this overarching positive effect, recent work by Watts et al. (2023) found that NC Pre-K funding had greater effects on children who experienced more adversity, highlighting the compensatory nature of the program. We built on these recent findings by further exploring the heterogeneous impact of NC Pre-K by setting type and by the extent to which enrollment is concentrated in specific sites or disbursed across a community. To our knowledge, we are the first to examine broadly whether implementation differences across the state differentially affected longer-term achievement outcomes for age-eligible children.

This Study

We used data from the scale-up of NC Pre-K to better understand the local policy implementation decisions made within a highly varied mixed-delivery system. We examined the county- and child-level consequences related to county-level NC Pre-K implementation differences in the delivery of NC Pre-K. In line with previous literature on NC Pre-K (i.e., Watts et al., 2023), we exploited variation in county-level funding for the program to understand how program access impacts fifth grade achievement among children considered age eligible for the state-funded program. We focused primarily on academic achievement rather than behavioral or administrative outcomes because this was the only outcome that Watts et al. (2023) found to produce positive effects across subgroups (with null effects reported on special education and grade retention). Specifically, we examined how funding-induced center-type slot allocation differences relate to long-term achievement within counties over time.

We first examined how variation in a county’s allocation of slots across different types of centers affected the population-level achievement of all age-eligible children in a given county and year. Specifically, we asked (a) What is the association between allocating slots in different types of childcare settings and student achievement? and (b) Does the concentration of slots in a few centers versus the dispersion of slots across many centers predict later achievement?

We then examined the association between the type of center in which an individual child enrolled and that student’s later achievement by incorporating newly obtained administrative data on the group of children who were funded by NC Pre-K that included unusually detailed family and child characteristics as well as the characteristics of the classrooms and centers they attended to consider the extent to which implementation differences in the mixed-delivery approach differentially predicted their achievement. For these analyses of pre-K enrollees, we asked (a) Does the type of center that NC Pre-K children attended differentially predict their achievement? and (b) Among pre-K attenders, does the proportion of funded children in their NC Pre-K classroom (i.e., concentration of slots) affect their fifth grade achievement?

Of course, disentangling factors related to pre-K effectiveness from systemic and cultural factors, as well as family choice, presents a clear challenge. We do not claim that any observed effects related to implementation factors arose purely from exogenous variation. For example, community-based settings¹ within mixed-delivery systems have been shown to serve more children from Black and Hispanic families than public school settings, with the highest concentration of children of color typically observed in Head Start centers (Reid et al., 2019; Weiland et al., 2024). Children in these settings also frequently have greater educational support needs (e.g., dual language learner status or Individualized Education Program) and lower levels of achievement at the beginning of pre-K than children who attend public pre-K in school-based settings (Reid et al., 2019).

Given that parent choice is promoted within mixed-delivery systems, it is possible that these divergent child characteristics may be explained in part by parent selection of pre-K settings that are aligned with the family’s cultural interests and the child’s educational needs. However, the concentration of minoritized children within community centers also may reflect differences in pre-K delivery methods by neighborhood income. Thus, it is important to consider the potential influence of these factors when examining children’s enrollment across different pre-K settings, and it is possible that these factors also influence county funding allocations. Although we tested for such selection across counties and center types, we view our results as primarily descriptive in nature, providing early evidence in an area that sorely needs more work regarding the associations between differences in pre-K implementation and student achievement.

Methods

Sample

Participants included children born in North Carolina between October 17, 1987, and August 31, 2005. Birth records obtained from the NC Division of Public Health were matched with state administrative data to link each child’s county of birth to their fifth grade test scores (N = 1,207,576; see Watts et al. [2023] for more information). As can be seen in Table 1, most of the students in this sample identified as non-Hispanic White (58%), non-Hispanic Black (29%), and Hispanic (non-White or non-Black; ~7%). Gender was dichotomously categorized based on the information available, with half the students identified as male and half as female. Additionally, 8% of students were considered low birthweight, 65% had married parents, and 24% had a mother with <12 years of education.

Table 1.

Descriptive Statistics by Analytic Sample

Variable Name	County slot allocation analyses, M (SD)	Pre-K enrollee analyses, M (SD)
Key variables
County PS slot allocation (%)	0.22 (0.31)	0.49 (0.26)
County HS slot allocation (%)	0.05 (0.12)	0.13 (0.16)
County HS in PS slot allocation (%)	0.01 (0.08)	0.03 (0.12)
County for-profit slot allocation (%)	0.15 (0.23)	0.25 (0.21)
County nonprofit slot allocation (%)	0.06 (0.11)	0.09 (0.11)
County concentration of funded slots (%)	0.39 (0.42)	0.82 (0.15)
Child demographic information
Academic composite (std)	−0.00 (1.00)	−0.30 (0.90)
Male	0.50	0.51
Child is White	0.59	0.33
Child is Black	0.29	0.38
Child is Native American	0.02	0.03
Child is Asian	0.01	0.01
Child is Hispanic	0.07	0.20
Child is mixed race	0.03	0.04
Normal birth weight	0.82	0.83
Low birth weight	0.08	0.10
High birth weight	0.10	0.07
Family characteristics
Child is first born	0.43	0.38
Mother’s education <12 years	0.24	0.34
Mother’s marital status	0.65	0.46
Mother’s immigration status	0.09	0.21
Mother’s age in years	26.0 (5.92)	24.8 (5.70)
Mother is Black	0.29	0.39
Mother is Native American	0.02	0.03
Mother is Asian	0.01	0.02
Mother is Hispanic	0.07	0.18
Mother is other race	0.00	0.00
No father information	0.14	0.15
Observations	1,172,534	102,249

Note. PS = public school; HS = Head Start. Child-level descriptive information is provided for the population-level sample used in our county funding analyses (column 1) as well as the subsample of funded children for our pre-K attenders analyses (column 2).

We also linked administrative data from NC Pre-K child enrollment records housed at the University of North Carolina. Although data were originally available for all funded children who attended pre-K between 2001 and 2018 (N = 439,959), 62% (n = 272,477) of these children were outside of our sampling years.² Among children within our sampling frame, we successfully matched 63% to fifth grade school records. Therefore, our matched subsample (n = 106,182) reflects the children who (a) received funding to attend NC Pre-K between 2001 and 2010, (b) had matched North Carolina birth records, and (c) had matched public school fifth grade academic records.³ As expected, this subsample was found to be more disadvantaged than our primary county-level sample. As demonstrated in Table 1, 38% of participants identified as non-Hispanic White, 42% as non-Hispanic Black, and 20% as Hispanic. Reflecting the targeted nature of the program, the vast majority of participants in this subsample were eligible in fifth grade for free or reduced-priced lunch (90%; see Table 2), 45% had married parents, and 35% had a mother with <12 years of education.

Table 2.

Descriptive Statistics of North Carolina Pre-K Program Enrollees by Center-Type Attendance, Pooled Over Time

Variable Name	All site types, M (SD)	Public school, M (SD)	Head Start, M (SD)	Head Start in public school, M (SD)	Private for-profit school, M (SD)	Private nonprofit school, M (SD)
Classroom concentration (%)	0.85 (0.18)	0.82 (0.19)	0.84 (0.19)	0.79 (0.23)	0.92 (0.11)	0.86 (0.18)
Child information
Male	0.51	0.51	0.50	0.52	0.50	0.51
White	0.33	0.38	0.18	0.35	0.32	0.33
Black/African American	0.42	0.35	0.58	0.39	0.45	0.46
Native Hawaiian/Pacific Islander	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01
Asian	0.02	0.02	0.01	0.01	0.02	0.01
Hispanic	0.20	0.21	0.19	0.25	0.19	0.17
American Indian/Alaska Native	0.03	0.04	0.04	<0.01	0.03	0.03
Disability status	0.05	0.05	0.04	0.04	0.03	0.05
LEP status	0.16	0.18	0.13	0.20	0.15	0.14
Normal birth weight	0.83	0.83	0.83	0.83	0.84	0.83
Low birth weight	0.10	0.10	0.11	0.10	0.09	0.10
High birth weight	0.07	0.07	0.06	0.07	0.07	0.07
Family characteristics
Child is first born	0.38	0.35	0.40	0.38	0.41	0.41
Child family size	3.98 (1.39)	4.12 (1.43)	3.79 (1.34)	3.90 (1.33)	3.86 (1.33)	3.83 (1.33)
Mother’s employment status	0.79	0.81	0.71	0.73	0.80	0.78
Mother’s education <12 years	0.34	0.35	0.39	0.42	0.30	0.31
Mother’s marital status	0.46	0.50	0.31	0.41	0.45	0.45
Mother’s immigration status	0.21	0.22	0.19	0.23	0.20	0.17
Mother’s age in years	24.8 (5.70)	25.0 (5.67)	24.0 (5.61)	24.5 (5.69)	24.9 (5.72)	24.8 (5.81)
Mother is Black	0.39	0.32	0.54	0.34	0.42	0.43
Mother is Native American	0.03	0.04	0.04	<0.01	0.02	0.02
Mother is Asian	0.02	0.02	0.01	0.01	0.01	0.01
Mother is Hispanic	0.18	0.19	0.17	0.23	0.17	0.15
Mother is other race	<0.01	<0.01	<0.01	<0.01	<0.01	<0.01
No father information	0.15	0.13	0.20	0.14	0.15	0.16
Family income, % FPL
>200%	0.06	0.07	0.01	0.03	0.07	0.07
Between 186 and 200%	0.03	0.03	<0.01	0.02	0.04	0.03
Between 131 and 185%	0.15	0.16	0.03	0.08	0.19	0.18
<130%	0.75	0.73	0.96	0.87	0.69	0.71
Missing (but >186%)	0.01	0.01	<0.01	0.01	0.01	0.01
Observations	102,249	51,404	14,369	3,194	24,382	8,980

Note. LEP = limited English proficiency; FPL = federal poverty level. An additional family income variable was created, “Missing (but >186%),” to include children who had a family income >186% of FPL but who could not be more specifically categorized (n = 1,027). Maternal employment status also was imputed using the county program year mean (n = 3,867). All variables were found to significantly differ across settings at the p < 0.01 level, except for normal birthweight (p = 0.053) and mother is other race (p = 0.317).

Measures

As we describe in more detail below, our analyses were split into two sets of models. The first set examined the association between the distribution of pre-K slots across center types at the county year level and student achievement for the entire population of NC public school children. The second set tested the relation between child-level enrollment in different center types for the subsample of children who enrolled in an NC Pre-K–funded slot and achievement. We begin by describing the achievement measure before describing the county- and child-level analysis measures, respectively.

Academic Achievement

Student fifth grade state mathematics and reading achievement test scores were obtained from the NC Department of Public Instruction through the NC Education Research Data Center at Duke University, and test scores were standardized by subject and academic year to account for differences in the assessment instruments over time. Student standardized math and reading scores then were averaged together and re-standardized to generate a standardized composite of fifth grade academic achievement.

Population-Level Analysis Measures

County Slot Allocation to Different Center Types

Center-type funding measures the proportion of funded slots allocated to each of the five center types in a county per year: public school, Head Start, private for-profit center, private nonprofit center, or a Head Start located in a public school. The denominator was the total number of funded slots that were available in that same county and year. We also explored supplemental analyses (see Appendix Table A4 in the online version of the journal) using a broader three-category classification of center type: public school, Head Start (either provided at a center or public school), and community center (including both private for-profit and private nonprofit). It should be noted that this variable ranged to zero because (a) our county-slot allocation analyses incorporated several nonprogram pretrend years and (b) program funding was not available in all North Carolina counties until the 2003–04 academic year. Thus, in any year where a county’s funding was equal to zero, center allocations are also zero.

Slot Concentration

Slot concentration captures the degree to which a county’s pre-K slots were concentrated within a select number of centers or disbursed across centers each year. To calculate this variable, we divided the number of funded slots in a given county and year by the total number of preschool-aged children attending NC Pre-K classrooms in that county and year, regardless of funding status.⁴ Because classrooms can include both funded and nonfunded peers, this number tells us whether greater or fewer non–NC Pre-K-exposed children relates to population benefits. As with our measure of center-type funding, this measure of slot concentration also ranged to zero.⁵

Pre-K Enrollee-Level Analysis Measures

Center Type Enrollment

For analyses examining children enrolled in and funded by NC Pre-K, center type attendance identifies the type of childcare setting in which each funded child attended NC Pre-K. This measure was derived from child-level administrative data. We generated five indicator variables to identify whether children were enrolled (yes = 1) or not enrolled (no = 0) in NC Pre-K at a public school, Head Start center, private for-profit center, private nonprofit center, or at a Head Start located in a public school. As with our population-level analyses, we also examined the impact of using less granular distinctions (see Appendix Table A5 in the online version of the journal). Descriptive statistics of NC Pre-K attenders across the five center types is provided in Table 2. Half the directly funded children attended NC Pre-K in a public school (50.3%), roughly one third attended a private community center (23.8% at a for-profit center and 8.8% at a nonprofit center), and more than one sixth attended NC Pre-K at a Head Start located either in a center (14.1%) or in a public school (3.1%). These numbers are consistent with reports on NC Pre-K over time (e.g., Peisner-Feinberg et al., 2019).

Classroom Slot Concentration

Classroom slot concentration was generated for each funded child to reflect their exposure to NC Pre-K–funded peers. For each funded child, we divided the number of funded classmates in their preschool classroom (including the funded child themselves) by the total number of funded and nonfunded classmates (i.e., total class size). This classroom slot concentration variable ranges from 36 to 100%. Shown in Table 2, the proportion of funded NC Pre-K classmates was the lowest for Head Start classrooms located in public schools (M = 0.78, SD = 0.23) and the highest for classrooms in private for-profit centers (M = 0.92, SD = 0.11).

NC Pre-K Funding

Program administrative records were used to measure variation in the annual amount of NC Pre-K funding provided to each North Carolina county divided by the number of 4-year-olds, scaled to thousands and converted to 2019 dollars. This was matched to children across each of the 18 birth cohorts based on their county of birth. This assumes that all children attended preschool within the county in which they were born.⁶ Appendix Figure A2 in the online version of the journal illustrates the growth and variation in the number of pre-K centers receiving NC Pre-K funds within counties during program scale-up.

Missingness on Key Variables

Missingness on a few variables reduced our analytic samples. First, our pre-K enrollee analyses relied on a subsample of 102,249, which was reduced from our original matched subsample (n = 106,182) due to missing fifth grade achievement (n = 3,241), class size and center type information (n = 335), or child/family demographic information (n = 357). Missing classroom and center characteristics also reduced our population analytic sample (n = 36,489). This, in combination with the missing child/family covariate data, reduced the county-level sample from 1,207,576 to 1,172,534.

Covariates

We included covariates to enhance the precision of our estimates and control for time-varying factors not accounted for by the county fixed effects but that still may have impacted NC Pre-K funding and/or elementary school achievement. At the child level, we included children’s race/ethnicity (non-Hispanic Black, non-Hispanic Native American, non-Hispanic Asian, non-Hispanic mixed race, and Hispanic), sex, birth weight (defined as low [<1,000 to <2,500 g], normal [2,500 to <4,500 g], or high [>4,500 g]), and quarter of birth. At the family level, we included characteristics of the mother’s race (Black, Native American, Asian, White, or other race), ethnicity (Hispanic or not Hispanic), marital status (married = 1), immigration status (yes = 1), years of education, age (in years) at child’s birth, and primipara (yes = 1), as well as father presence at birth (no = 1). Lastly, we incorporated time-varying county controls, including the proportion of births to non-Hispanic Black mothers, Hispanic mothers, and mothers with <12 years of education; total county population (log); number of births (log); the percent of the population receiving SNAP; the percent of the population receiving Medicaid; and the median family income (in 2019 dollars).

Additionally, we incorporated the rich child and family demographic information for NC Pre-K enrollees as controls in our analyses of pre-K enrollees (see Table 2): disability status, limited English proficiency status, family size, maternal employment (employed = 1), and three dichotomously coded family income categories (i.e., family income >200% of the federal poverty level [FPL], between 186 and 200% of the FPL, and <130% of the FPL). To retain our subsample size, an additional category was generated for students who were known to be >186% of the FPL but did not have the data needed to separate them into their more specifically defined categories (n = 1,027). Additionally, we imputed missing data on maternal employment using mean imputation, which we based on the county program year mean (n = 3,867).

Analytic Approach

Research Design

Our analytic approach for the county-level estimates mirrored that of Watts et al. (2023) such that fifth grade academic achievement scores were regressed on the level of NC Pre-K funding provided to the county during the fiscal year(s) that the child was age eligible for preschool. This natural experiment leverages both between- and within-county variation. First, between-county variation in the timing of NC Pre-K funding receipt can be exploited because funding was not equally distributed across all 100 counties each year. Because the amount of NC Pre-K funding that each county received was not always consistent from year to year, there exists within-county variation in funding over time. This means that children from the same birth cohort would have experienced different levels of NC Pre-K funding exposure had they also been born in different counties. Further, children who were born in the same county would have had different levels of exposure to NC Pre-K funding if they were also in different birth cohorts. Thus, this natural experiment enabled us to examine the impact of NC Pre-K funding by facilitating a comparison of children within the same county of birth, over time, before and after funding changes.

Population-Level Analyses

Our first set of analyses examined whether children’s academic achievement was associated with a county’s allocation of slots across center type in a given year. We used two-way fixed effects for the county and program year with ordinary least squares and included interaction terms to test for moderation of NC Pre-K funding by our county implementation moderators of interest (i.e., center-type funding and slot concentration):

\begin{array}{l} E_{i c y} = δ N C P K_{c y} + γ M O D_{c y} + λ N C P K_{c y} \times M O D_{c y} \\ + X_{c} α + Z_{i} β + θ_{y} + π_{c} + ε_{i c y} \end{array}

where $E_{i c t y}$ represents the fifth grade academic achievement outcome for child i born in county c who was age eligible to participate in NC Pre-K in year y. On the right side of the equation, continuous variable $N C P K_{c y}$ represents the amount of NC Pre-K funds (in $1,000 units adjusted for the population of 4-year-old children) provided to county c in program year y. Additionally, $M O D_{c y}$ denotes the main effect of a grand mean-centered moderator of interest (i.e., county slot allocation by center type or county slot concentration), whereas $λ$ demonstrates the interaction between these county/year moderators and NC Pre-K funding to determine whether each moderator influenced the effect of NC Pre-K funding on fifth grade achievement. Fixed effects for the academic year that a child was age eligible for NC Pre-K, program year, $θ_{y}$ , and county of birth, $π_{c}$ , control for unobserved year-specific factors and stable characteristics associated with each child’s county of birth. This accounts for any endogeneity in the provision of NC Pre-K funding during the program’s rollout and limited comparisons of a given child’s academic achievement with other children residing in the same county over time.

We also included a vector of time-varying county covariates (e.g., Smart Start funding, percent of population receiving SNAP, and proportion of births to non-Hispanic Black mothers), provided by $X_{c}$ , and a vector of individual child-level covariates (e.g., child race/ethnicity, child birthweight, maternal education level, and birth quarter), represented by $Z_{i}$ .

Slot Concentration

In our next set of models that exploit cross-county funding variation, we examined the influence of counties’ decisions to concentrate or disperse their NC Pre-K slots across providers. We first assessed the relation between NC Pre-K slot concentration and later academic achievement. Next, we stepped in the center-type allocation variables, with public school funding excluded as the reference group. Lastly, we examined whether the NC Pre-K funding effect was moderated by county slot concentration in a given year to examine whether allocating pre-K slots in various ways improved the efficacy of state pre-K dollars invested in student achievement.

Pre-K Enrollee Analyses

We also assessed whether funded children’s fifth grade achievement was associated with the type of center that they attended, or the proportion of funded peers in their NC Pre-K classroom. These analyses were designed to answer the question, “Among children who were directly funded by NC Pre-K, were differences in center type or classroom slot concentration associated with their achievement?”

To answer this question, we employed a series of ordinary least squares regressions with two-way fixed effects for the county and program year. These pre-K enrollee analyses no longer capitalize on differences across counties in their funding as an identification strategy, although we control for NC Pre-K funding. These analyses focus on our two pre-K attender variables of interest:

E_{i c y} = γ T y p e_{c y} + γ C o n c_{c y} + δ N C P K_{c y} + X_{c} α + Z_{i} β + θ_{y} + π_{c} + ε_{i c y}

where $T y p e_{c y}$ and $C o n c_{c y}$ denote center type attendance and classroom concentration, and all other terms are the same as earlier. However, our vector of individual child-level covariates (e.g., child race/ethnicity, child birth weight, maternal education level, and birth quarter), represented by $Z_{i}$ includes some additional child-level covariates derived from administrative records for funded children (e.g., maternal employment, family income, child’s English proficiency, and disability status).

We first examined the association between NC Pre-K center type attendance and fifth grade achievement by including all center type indicator variables in the model together to test for joint significance (with public school attendees as the reference group). Next, we assessed the relationship between NC Pre-K classroom slot concentration and pre-K enrollees’ fifth grade achievement. We then added center type attendance, followed by the corresponding interaction model. This allowed us to test whether the influence of classroom composition on fifth grade achievement was different for children who attended NC Pre-K at a Head Start, private for-profit center, private nonprofit center, or Head Start provided in a public school.

Results

Descriptive Results

Table 1 provided child-level descriptive information for each analytic sample, with our population sample presented in column (1) and our pre-K attender sample presented in column (2). Children in our pre-K attender sample were slightly more disadvantaged than those in our pre-K enrollee sample, reflecting the targeted nature of NC Pre-K eligibility.

Across all counties and years (column 1), 22% of funded slots were allocated to public school settings, 15% to for-profit centers, 6% to nonprofit centers, 5% to Head Start centers, and 1% to Head Start programs located in public schools. Note that these averages include pretrend years in which program funding was zero (i.e., funding was not yet active). Among the years in which NC Pre-K funding was active (column 2), approximately half of the funded slots were provided to public schools, over one third of slots were provided to community centers (25% to for-profit centers and 9% to nonprofit centers), and around one sixth of slots were to Head Start settings (13% to Head Start centers and 3% to Head Start programs offered in public schools).

Beyond the child-level information provided in Table 1, Figure 1 and Appendix Figure A3 in the online version of the journal illustrate the differences in county funding approaches over time. Figure 1 displays variation in center-type slot allocation, showing that the average county allocated over half of its slots to public school settings (see Appendix Table A1 in the online version of the journal for an exploration into selection of center type slot allocation within counties). Appendix Figure A3 in the online version of the journal portrays the within-county variation in classroom slot concentration, demonstrating that over time, counties developed classrooms that increasingly were exclusively composed of NC Pre-K participants.

Figure 1.

Variation in the average allocation of county North Carolina Pre-K Program slots across center types.

Main Findings

Results of our population analyses and NC Pre-K enrollee analyses are presented in Tables 3 and 4, respectively. Consistent with prior NC Pre-K work (e.g., Dodge et al., 2017; Watts et al., 2023), we began by estimating the effect of additional NC Pre-K funding for each child in a given county and year regardless of whether they enrolled in the program. Proportion variables are rescaled to 10 percentage point units to improve the interpretability of our results. Estimates were generated in Stata 17.0 using the reghdfe package (Correia, 2017).

Table 3.

County Slot Allocation Analyses: Association Between Center Type Funding, Slot Concentration, and Fifth Grade Achievement

	NC Pre-K funding	Center type allocation		Funded slot concentration
Predictors	(1)	(2)	(3)	(4)	(5)	(6)
NC Pre-K funding ($000s)	0.031* (0.011)	0.031* (0.010)	0.027* (0.008)	0.029* (0.010)	0.030* (0.010)	0.039** (0.015)
Center type allocation (PS reference)
HS slots (10 pp)		0.001 (0.004)	−0.001 (0.006)		0.002 (0.003)	0.002 (0.003)
HS in PS slots (10 pp)		−0.000 (0.003)	0.001 (0.004)		0.000 (0.003)	0.000 (0.003)
For-profit slots (10 pp)		0.004 (0.002)	0.004*** (0.002)		0.003 (0.002)	0.003 (0.002)
Nonprofit slots (10 pp)		0.005 (0.005)	−0.005 (0.004)		0.006 (0.004)	0.006 (0.004)
Center type allocation × funding interactions
% HS slots × NC Pre-K funding			0.002 (0.005)
% HS in PS slots × NC Pre-K funding			−0.000 (0.003)
% For-profit slots × NC Pre-K funding			−0.001 (0.003)
% Nonprofit slots × NC Pre-K funding			0.015* (0.005)
County concentration
Concentration of funded slots (10 pp)				0.003 (0.002)	0.003 (0.002)	0.004 (0.002)
Concentration of funded slots × NC Pre-K funding						−0.002 (0.003)
Observations	1,172,534	1,172,534	1,172,534	1,172,534	1,172,534	1,172,534
Joint F-test p value			0.018**

Note. NC Pre-K = North Carolina Pre-K Program; PS = public school; HS = Head Start; pp = percentage point. .Standard errors (in parentheses) are clustered at the county level. All models include Smart Start funding (in $000s), child and family covariates, county covariates, and fixed effects for the county, program year, and quarter of birth. Child covariates include child’s sex (male), race and ethnicity (Black, Native American, Asian, Hispanic, or mixed race), and birth weight (low, high). Family covariates include mother’s education (<high school), marital status (married), immigration status, age (in years), race and ethnicity (Black, Native American, Asian, Hispanic, or other race), whether a father was present at birth, and first-born status. County covariates include the percent of births to Black mothers, Hispanic mothers, and low-education mothers; the number of births (log); total population (log); median family income; percent of population receiving food stamps; and percent of population enrolled in Medicaid. The proportion of slots allocated to each center type and the concentration of funded slots were generated for a given county and year. All proportion variables were rescaled to 10% units.

p < 0.01; **p < 0.05; ***p < 0.10.

Table 4.

Pre-K Enrollee Analyses: Association Between Center Type Attendance, Classroom Concentration, and Fifth Grade Achievement Among Children Enrolled in the North Carolina Pre-K program

	Center type attendance	Class concentration		All
Predictors	(1)	(2)	(3)	(4)
Center-type attendance (PS reference)
HS	−0.005(0.013)		−0.006(0.013)	−0.006(0.013)
HS in PS	−0.007(0.025)		−0.005(0.024)	0.009(0.025)
For profit	0.086*** (0.016)		0.081*** (0.016)	0.082*** (0.015)
Nonprofit	0.026(0.017)		0.023(0.017)	0.024(0.017)
Classroom characteristics
Class concentration (10 pp)		0.010*** (0.002)	0.007** (0.002)	0.006**** (0.003)
Interactions
Class concentration × HS				0.004(0.006)
Class concentration × HS in PS				0.018* (0.008)
Class concentration × For profit				−0.000(0.006)
Class concentration × Nonprofit				−0.006(0.007)
Observations	102,249	102,249	102,249	102,249
Joint F-test p value	<0.001***			0.128

Note. PS = public school; HS = Head Start; pp = percentage point. Standard errors (in parentheses) are clustered at the county level. All models include North Carolina Pre-K Program funding (in $000s), Smart Start funding (in $000s), child and family covariates, county covariates, and fixed effects for the county, program year, and quarter of birth. In addition to the covariates described in Table 3, we include pre-K class size, child disability status, and limited English proficiency status, as well as family income (family income that is >200% of the federal poverty level [FPL], between 186 and 201% of the FPL, <130% of the FPL, with an additional category for those >186% of FPL who could not be more specifically categorized; n = 1,027). Maternal employment status at enrollment was imputed using the county program year mean (n = 3,867). Center type attendance reflects the childcare setting that each child attended. Classroom concentration reflects the proportion of funded slots in each North Carolina Pre-K Program classroom. All proportion variables were rescaled to 10% units.

p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.10.

Population Analyses

Center Type Slot Allocation

Column (1) of Table 3 begins with the main effect of NC Pre-K funding for children in our population analytic sample (n = 1,172,534). Recall that these estimates come from the entire 4-year-old population regardless of whether a given child attended NC Pre-K. We found that a $1,000 increase in county-level NC Pre-K funds per 4-year-old child was related to a significant increase in fifth grade achievement of 0.031 SD (p < 0.01). This effect is similar in both magnitude and significance to that in Watts et al. (2023). Next, we added the center type funding measures across the five center types (column 2) and their respective interactions with NC Pre-K funding (column 3; public school funding as the reference group). Main effect estimates for these variables were generally small and not statistically significant (see Appendix Table A2 in the online version of the journal for models examining the main effect of each center type), and the main effect of NC Pre-K funding remained unchanged. This suggests that the funding effect is not mediated by the allocation of funds to the five center types.

Including the interaction between center type allocation and NC Pre-K funding resulted in a marginally significant but small coefficient related to the proportion of children enrolled in private for-profit centers ( $β$ = 0.004; p < 0.10) and a significant interaction between NC Pre-K funding and the proportion of children enrolled in private nonprofit centers ( $β$ = 0.015; p < 0.01). The set of interaction terms was jointly statistically significant (p = 0.018). This pattern suggests that pre-K may have been more effective when it occurred in private nonprofit community centers relative to the other center types. However, we caution the interpretation of this interaction for a few reasons. First, the interaction effect was diminished when the broad, three-pronged center type categorizations (see Appendix Table A4 in the online version of the journal) were examined instead of the specific, five-pronged categorizations displayed here. Second, the interaction itself is difficult to conceptualize considering that the cost per NC Pre-K slot varies depending on the setting type, with private-center slots costing more than the other center types (Barnett, 2018).

Slot Concentration

Shown in columns (4)–(6) of Table 3, county slot concentration across centers was not significantly predictive of later county year-level achievement. The interaction between NC Pre-K funding and county slot concentration in a given year was negative but also nonsignificant.

Pre-K Enrollee Analyses

Before examining the effects of center attendance on child outcomes, we first investigated student selection into various types of centers. Table 2 provides descriptive information on our expanded set of covariates for NC Pre-K enrollees by center type attendance. Overall, child and family characteristics were relatively consistent across the five center type categories. The most notable difference that emerged was with respect to Head Start, indicating that children who attended NC Pre-K in a Head Start center were more economically disadvantaged (e.g., 96% with family incomes <130% of the federal poverty level) and more likely to identify as Black (58%) than children who attended the other center types. Beyond this distinction, we observed few notable differences between children in the types of centers they attended. Although families have substantial choice in where to enroll their child in NC Pre-K, their choices do not seem systematically different from one another, at least on observable characteristics.

Center Type Attendance

Table 4 presents findings using the subsample of children (n = 102,249) who were enrolled in NC Pre-K. We begin with our joint attendance model, which includes all center type variables, with public school as the reference group. Shown in column (1), children who attended NC Pre-K at a for-profit center scored ~0.09 SD higher on their fifth grade assessments (p < 0.001) than their public-school-attending peers. For models with main effects of each of the five center types compared with a pooled reference group, see Appendix Table A3 in the online version of the journal. This table reflects the other center types compared similarly with one another, with private for-profit centers consistently holding an advantage over all other center types.

Classroom Concentration

We also evaluated the effect of children’s exposure to funded peers. These models, presented in Table 4, begin by assessing the main effect of classroom concentration in column (2), followed by a model stepping in the five center type categories in column (3), with the interaction between classroom concentration and center type attendance in column (4). NC Pre-K classroom concentration predicted fifth grade achievement such that a 10 percentage point increase in share of funded preschool classmates resulted in a 0.010 SD increase in later achievement (p < 0.001). The coefficient decreased when we included center type attendance ( $β$ = 0.007; p < 0.01) in columns (3) and (4) and interactions between classroom concentration and center type attendance ( $β$ = 0.006; p < 0.10). Of the five interactions examined, one was significant, indicating that having a greater concentration of funded classmates was more beneficial for children who attended NC Pre-K at a Head Start in a public school relative to the other center types ( $β$ = 0.019; p < 0.05). However, the joint significance test for these models was just above the 0.10 threshold, indicating that these measures were not meaningfully contributing to the model and thus should be interpreted with caution.

Sensitivity Analyses

Spillover Effects to Children Not Enrolled in NC Pre-K

One question raised by our slot concentration analyses is whether counties may benefit when students are indirectly exposed to NC Pre-K funding through quality improvements to childcare centers. For these exploratory analyses, we relied on data from the county population of 4-year-olds to determine the percent of the population that was directly exposed to funding (i.e., received a funded NC Pre-K slot), indirectly exposed to funding (i.e., attended an NC Pre-K center as a 4-year-old but did not themselves receive a funded slot), or were not at all exposed to NC Pre-K funding (i.e., did not attend NC Pre-K—used as the reference group). These three categories are visualized in Appendix Figure A1 in the online version of the journal, which highlights the increasing number of children directly and indirectly exposed to NC Pre-K funding as the pre-K program matured, coupled with the decreasing number of children who did not attend NC Pre-K. Shown in Appendix Table A6 in the online version of the journal, counties generally performed better when the number of directly funded children increased. These models do not indicate evidence of spillover effects because indirect exposure to NC Pre-K funding was not associated with benefits for nonfunded children.

Third and Fourth Grade Achievement

To ensure that our mixed-delivery findings were not spurious, we assessed whether our concentration and center type variables were associated with students’ third and fourth grade achievement in a manner similar to fifth grade achievement. Overall, our main findings showed that fifth grade achievement was consistent with that observed for the earlier grade levels across our county funding (see Appendix Tables A8 and A10 in the online version of the journal) and pre-K enrollee (see Appendix Tables A9 and A11 in the online version of the journal) analyses. Notably, for the third and fourth grade models, we did not observe statistically significant main effects of funding on achievement for the cohorts included in this paper. However, it should be noted that Ladd et al. (2014) reported statistically significant funding effects for NCs Pre-K on third grade achievement when analyses were restricted to earlier cohorts of students and models included birth year fixed effects (as in Watts et al., 2023).

Early Versus Later Scale-up

To explore whether the differential influence of NC Pre-K center type attendance varied as the program matured, we examined whether our pre-K attendance findings were consistent across cohorts of NC Pre-K enrollees (see Appendix Table A12 in the online version of the journal). Apart from children in the third cohort, attending NC Pre-K at a for-profit center remained significantly predictive of fifth grade achievement, with effects ranging in magnitude from 0.038 (nonsignificant) to 0.123.

Threshold Effects of Classroom Concentration

Given the nature of our concentration variables, it is possible that certain concentration levels are critical for achievement. The distribution of classroom slot concentration (shown in Appendix Figure A3 in the online version of the journal) limited our options for threshold magnitudes. As a result, we created a dichotomous variable of attending fully concentrated classrooms (i.e., 100% concentrated) relative to classrooms that were not fully concentrated (i.e., <100%). Appendix Table A13 in the online version of the journal presents the effect of full classroom concentration both with and without controls for center type attendance. Shown in column (1), the effect of attending a classroom with 100% concentration versus all other classrooms was nearly zero.

Discussion

The United States has experienced a proliferation of publicly funded pre-K efforts over the past several decades. Although these programs are motivated by the promise of lasting developmental benefits, existing early childhood research has underscored the variable nature of pre-K effectiveness, particularly in the long term (Burchinal et al., 2024). Although a growing body of research has attempted to explain these heterogeneous effects, few studies have investigated the influence of how and where early childhood programs are implemented. Such consideration is particularly important given the increased reliance on the mixed-delivery approach, which enables states to capitalize on existing resources while also prioritizing the varied needs of children and families. However, pre-K mixed-delivery systems also introduce further complexity and nuance that must be considered as we attempt to solve the pre-K puzzle (Phillips et al., 2017).

Our study explored heterogeneity in the scale-up of NC Pre-K to test how differences in early childhood investments affected children’s subsequent development. We relied on county-level data collected from 18 cohorts of participants who were age eligible to attend pre-K from 1992–93 to 2009–10. Using these data, we first evaluated whether the fifth grade achievement of all age-eligible children was related to differences in annual county funding decisions surrounding the types of centers that slots were allocated toward and the degree to which counties concentrated or dispersed their funded slots within and across, NC Pre-K centers. These county funding analyses then were mirrored using a subsample of NC Pre-K enrollees to assess the extent to which the fifth grade achievement of funded children was differentially affected by the type of center that these children attended or by their level of exposure to similarly funded peers.

Overall, we found little indication that children substantially benefited from the provision of NC Pre-K funds to any specific childcare setting. We instead found evidence that the benefits of pre-K investment were not strongly tied to center type, suggesting that program investments were likely to be effective across the mixed-delivery system. When differences occurred, they suggested that public pre-K funding was slightly more effective in promoting later achievement when allocated to private for-profit and nonprofit centers, although these differences typically were modest in magnitude. These private centers also were found to have a higher concentration of funded slots, which further promoted the achievement of NC Pre-K enrollees. However, slot concentration, measured at the county level, was not similarly beneficial for all age-eligible children.

Impact of Heterogeneity and Program Implementation

Our results suggest that variation in pre-K effectiveness may be attributed I part to the diverse implementation practices instituted within state mixed-delivery systems, including the location of program delivery and the concentration of public funds. Interestingly, we do not find any indication that the main mechanism of pre-K program scale-up (i.e., in public schools) is more beneficial than other modes of delivery in the longer term. In fact, our findings suggest that NC Pre-K funds were less effective when provided to public schools and more effective when allocated to private for-profit and nonprofit centers. This finding is surprising given the emphasis states often place on public schools during pre-K implementation and expansion because such settings are perceived to have higher levels of structural and process quality (Bellm et al., 2002; Desimone et al., 2004; Early et al., 2005; Little, 2020; McCormick et al., 2022; Reid et al., 2019; Wilinski, 2017) and greater alignment between pre-K and kindergarten (e.g., Desimone et al., 2004; Kagan & Kauerz, 2007; Little, 2020). However, qualitative research has highlighted potential difficulties surrounding implementing pre-K in public school settings. For example, pre-K teachers have reported that they are often more physically and socially isolated from the rest of the elementary school (e.g., Wilinski, 2017). Moreover, despite the abundance of resources available to pre-K classrooms (e.g., access to shared art rooms, playgrounds, and cafeterias; Desimone et al., 2004), such resources may not always be appropriate or usable for 4-year-old children (Little, 2020).

Furthermore, our findings contradict recent work by Weiland et al. (2024), who reported lower levels of skill growth among children who attended state-funded pre-K in community-based settings relative to public school settings. However, these authors only examined the proximal influence of public pre-K settings on children’s post-test achievement, which we were unable to explore in our dataset. Thus, it is unclear whether these post-test differences would have extended into later elementary school years and whether they would be consistent with the patterns observed in this study.

Descriptive studies also have highlighted the potential for selection into pre-K settings, because public school settings reportedly had higher concentrations of White students and teachers, whereas community-based settings had greater concentrations of minoritized children. As a result, it remains possible that our observed effects may be driven, at least in part, by unobserved processes related to parent/family decision making. However, we found limited evidence of selection into centers in our sample, as indicated by the available demographic information of NC Pre-K attendees. In fact, the only notable difference that emerged was that children who attended NC Pre-K in a Head Start center were more likely to identify as Black and be from more economically disadvantaged families than children who attended NC Pre-K at the other center types. However, the comparative disadvantage experienced by children who attended NC Pre-K in a Head Start center is understandable considering the vast array of services that Head Start provides. Notwithstanding the assignment of the “neediest” children to Head Start service receipt, we did not see further evidence to indicate systemic selection across NC Pre-K setting types.

Beyond the influence of program settings, we found that concentrating funded NC Pre-K slots within classrooms was more beneficial than dispersing slots across many classrooms and centers. In other words, children—specifically funded children—scored higher on fifth grade achievement tests when they attended pre-K classrooms with a greater number of funded peers. It is possible that because NC Pre-K is income targeted, classroom funding concentration may equate to reduced socioeconomic diversity and thus more economically segregated early learning environments. This not only would reduce the potential for socioeconomic peer effects but also would decrease the potential for public investments to spill over to nonfunded children. However, we did not find a significant association between county slot concentration and the fifth grade achievement of nonfunded age-eligible children.

Notwithstanding the benefit associated with concentrating pre-K slots, our results should not be interpreted as support for the creation of more segregated pre-K classrooms. Note that we do not know the characteristics of nonfunded children. The hypothesized benefit of dispersing pre-K slots across classrooms and centers assumes that the nonfunded children in these pre-K centers are more advantaged than their funded peers. In this study, we were only able to observe the characteristics of children who received NC Pre-K funding and were unable to conclude whether the integration of nonfunded classmates contributed more demographic diversity. It is possible that children are more likely to attend pre-K centers that are nearest to their households, making the demographic characteristics of funded and nonfunded children potentially more similar than they are different. Assuming that these children were in fact demographically similar, then dispersing NC Pre-K–funded slots across classrooms and centers would not afford any additional benefit to funded children because they would be exposed to similarly homogeneous learning environments. If this is true, it remains unclear what mechanism could be driving the positive association between slot concentration and children’s achievement because such concentration is likely to generate less diverse program environments.

Limitations

Despite the strengths of our study, several limitations are worth mentioning. First, the variation in program implementation is not randomized and therefore is vulnerable to omitted variables bias from factors associated with NC Pre-K funding changes within counties over time. Related to this limitation is the inability to ascertain how counties are making their specific funding decisions. Despite the consistency in NC Pre-K slot costs across settings, the proportion of the slots covered varies from one setting to the next (i.e., 46% in Head Start, 56% in public school, and 72% in private centers; North Carolina Department of Health and Human Services, 2017). Thus, it is likely that funding decisions do not occur in a vacuum but rather through an interplay between supply (e.g., capacity and availability of alternate funds) and demand (e.g., parent choice and educational support needs) processes.

Second, our data were unable to elucidate the mechanism through which differences in fifth grade achievement may occur. As mentioned earlier, we could not determine whether there are observable or measurable differences in classroom and instructional quality that may be contributing to differences in later achievement. However, Weiland et al. (2024) recently examined variation in pre-K quality across several mixed-delivery systems, with lower levels of process quality observed within community-based settings. Future work should evaluate how these differences in quality may translate to longer-term effects.

Because the underlying mechanisms driving the effects are uncertain, it is unclear the extent to which our findings will generalize to other mixed-delivery systems. In this particular state (North Carolina), public investments appear to be more efficacious when provided to private pre-K centers and when funds are concentrated within centers. However, we do not fully understand why these specific policy decisions make the public pre-K program more effective. Therefore, we cannot confidently say whether these findings would transfer to other settings, whether they be other states or other mixed-delivery systems. Differences across these settings with respect to program auspice—for example, differences between public pre-K provided in a Head Start and public school—may be distinct and unique to the setting of interest. More work on the mixed-delivery system is needed to generate a more cohesive understanding of how variation policy and program implementation may contribute to long-term achievement effects.

Finally, our analyses solely focused on the impact of the mixed-delivery system on children’s elementary school achievement. We relied on fifth grade achievement because it is the sole outcome that Watts et al. (2023) found to have positive effects across subgroups, with null effects reported on grade retention and special education designation. However, variation in the mixed-delivery system also may have implications for other areas of children’s subsequent functioning. For example, using data from the Early Childhood Longitudinal Study–Kindergarten, Magnuson et al. (2007) observed a negative association between pre-K and behavioral problems for children attending pre-K in non-school-based settings. Thus, it is possible that our narrow focus on academic outcomes may be limiting our understanding of the ways in which the mixed-delivery system impacts children.

Conclusion

In sum, our findings provide further understanding regarding how mixed-delivery pre-K systems may produce distal outcomes for the entire population of eligible children as well as the children for which the program was designed to serve. We attempted to examine how policy decisions surrounding center type and slot concentration may translate to observable differences in achievement, but there remain many unknown factors that may be contributing to both the funding decisions themselves and the mechanisms driving the impacts on achievement. Given this uncertainty and the descriptive rather than causal nature of our analyses, we caution the application of these findings to early childhood policy. Future work should examine how different policy implementation decisions are being made by key stakeholders. Moreover, additional work is needed to explore the more micro-level processes that may contribute to differential achievement effects, including parent decision making strategies and variation in process quality across centers. Exploring such processes and the impacts they have on children’s subsequent outcomes is critical given the continued expansion of public pre-K across the United States and the practicality of the mixed-delivery approach.

Supplemental Material

sj-docx-1-ero-10.1177_23328584261441289 – Supplemental material for Examining the Mixed-Delivery System During Pre-K Expansion in North Carolina

Supplemental material, sj-docx-1-ero-10.1177_23328584261441289 for Examining the Mixed-Delivery System During Pre-K Expansion in North Carolina by Caroline M. Botvin, Jade M. Jenkins, Maria Sauval, Tiffany Wu, Kenneth Dodge, Ellen Peisner-Feinberg and Tyler W. Watts in AERA Open

Supplemental Material

sj-docx-2-ero-10.1177_23328584261441289 – Supplemental material for Examining the Mixed-Delivery System During Pre-K Expansion in North Carolina

Supplemental material, sj-docx-2-ero-10.1177_23328584261441289 for Examining the Mixed-Delivery System During Pre-K Expansion in North Carolina by Caroline M. Botvin, Jade M. Jenkins, Maria Sauval, Tiffany Wu, Kenneth Dodge, Ellen Peisner-Feinberg and Tyler W. Watts in AERA Open

Footnotes

Acknowledgements

The authors thank Ariel Ford, Justine Rogoff, Candace Witherspoon, Lorena Gonzalez, and Janessa Nieves in the North Carolina Division of Child Development and Early Education for discussing early versions of this work and organizing center visits to inform the writing of this paper.

Declaration of Conflicting Interests

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

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding was provided by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (Grant No. R01HD095930) and the Research Council of Finland (EDUCA Flagship, Grant Nos. 358924 and 358946). The State of North Carolina generally supports research on early childhood education but is not responsible for the conduct of this study. All views and opinions expressed in this paper are those of the authors alone.

ORCID iDs

Caroline M. Botvin

Jade M. Jenkins

Maria Sauval

Tiffany Wu

Ellen Peisner-Feinberg

Notes

Authors

CAROLINE M. BOTVIN is a doctoral candidate at Teachers College, Columbia University. She researches the short- and long-term effects of early childhood interventions.

JADE M. JENKINS is an associate professor at the University of California, Irvine, School of Education and co-director of the Center for Population, Inequality and Policy. She studies the impacts of early childhood policies on children, families, and providers.

MARIA SAUVAL is a postdoctoral researcher at Aalto University. Her research focuses on evaluating social and educational policies aimed at reducing inequalities and supporting child and youth development.

TIFFANY WU is a postdoctoral scholar at the University of Southern California. Her work focuses on how pre-K–12 policies can enhance access to and improve the quality of educational opportunities for children from underserved communities.

KENNETH DODGE is the William McDougall Distinguished Professor of Public Policy at the Sanford School of Public Policy, Duke University. He studies the development, prevention, and public policy of problem behaviors in children.

ELLEN PEISNER-FEINBERG is a research professor emerita and senior research scientist emerita at the University of North Carolina at Chapel Hill. She researches the short- and long-term effects of early educational interventions.

TYLER W. WATTS is an associate professor of psychology and education at Teachers College, Columbia University. He researches the long-term effects of educational interventions.

References

Ackerman

D. J.

Barnett

W. S.

Hawkinson

L. E.

Brown

McGonigle

E. A.

(2009). Providing preschool education for all 4-year-olds: Lessons from six state journeys (NIEER Policy Brief, Issue 18). National Institute for Early Education Research. https://nieer.org/research-library/providing-preschool-education-all-4-year-olds

Ansari

Pianta

R. C.

Whittaker

J. V.

Vitiello

V. E.

Ruzek

E. A.

(2020). Persistence and convergence: The end of kindergarten outcomes of pre-K graduates and their nonattending peers. Developmental Psychology, 56(11), 2027–2039. https://doi.org/10.1037/dev0001115

Ansari

López

Manfra

Bleiker

Dinehart

L. H. B.

Hartman

S. C.

Winsler

(2017). Differential third grade outcomes associated with attending publicly funded preschool programs for low-income, Latino children. Child Development, 88(5), 1743–1756. https://doi.org/10.1111/cdev.12663

Barnett

W. S.

(2018). Barriers to expansion of NC Pre-K: Problems and potential solutions. National Institute for Early Education Research.

Barnett

W. S.

Hustedt

J. T.

Robin

K. B.

Schulman

K. L.

(2003). The state of preschool. National Institute for Early Education Research.

Bellm

Burton

Whitebook

Broatch

Young

M. P.

(2002). Inside the pre-K classroom: A study of staffing and stability in state-funded prekindergarten programs. Center for the Child Care Workforce.

Botvin

C. M.

Jenkins

J. M.

Carr

R. C.

Dodge

K. A.

Clements

D. H.

Sarama

Watts

T. W.

(2024). Can peers help sustain the positive effects of an early childhood mathematics intervention? Early Childhood Research Quarterly, 67, 159–169. https://doi.org/10.1016/j.ecresq.2023.11.001

Burchinal

Whitaker

Jenkins

Bailey

Watts

Duncan

Hart

(2024). Unsettled science on longer-run effects of early education. Science, 384(6695), 506–508.

Cascio

E. U.

(2021). Early childhood education in the United States: What, when, where, who, how, and why. In McCall

(Ed.), The Routledge handbook of the economics of education (pp. 30–72). Routledge.

10.

Chetty

Friedman

J. N.

Hilger

Saez

Schanzenbach

D. W.

Yagan

(2011). How does your kindergarten classroom affect your earnings? Evidence from project star. Quarterly Journal of Economics, 126(4), 1593–1660. https://doi.org/10.1093/qje/qjr041

11.

Correia

(2017). Linear models with high-dimensional fixed effects: An efficient and feasible estimator. Duke University Press.

12.

Deming

(2009). Early childhood intervention and life-cycle skill development: Evidence from head start. American Economic Journal: Applied Economics, 1(3), 111–134. https://doi.org/10.1257/app.1.3.111

13.

Desimone

Payne

Fedoravicius

Henrich

C. C.

Finn-Stevenson

(2004). Comprehensive school reform: An implementation study of preschool programs in elementary schools. Elementary School Journal, 104(5), 369–389. https://doi.org/10.1086/499758

14.

Dodge

K. A.

Bai

Ladd

H. F.

Muschkin

C. G.

(2017). Impact of North Carolina’s early childhood programs and policies on educational outcomes in elementary school. Child Development, 88(3), 996–1014. https://doi.org/10.1111/cdev.12645

15.

Duer

J. K.

Jenkins

(2022). Paying for preschool: Who blends funding in early childhood education? Educational Policy, 37(7), 1857–1885. https://doi.org/10.1177/08959048221103804

16.

Durkin

Lipsey

M. W.

Farran

D. C.

Wiesen

S. E.

(2022). Effects of a statewide pre-kindergarten program on children’s achievement and behavior through sixth grade. Developmental Psychology, 58(3), 470.

17.

Early

Barbarin

Bryant

Burchinal

Chang

Clifford

Kraft-Sayre

(2005). Pre-kindergarten in eleven states: NCEDL’s multi-state study of pre-kindergarten and study of state-wide early education programs (SWEEP). Preliminary descriptive report (NCEDL Working Paper). University of North Carolina, Chapel Hill.

18.

Friedman-Krauss

A. H.

Barnett

W. S.

Garver

K. A.

Hodges

K. S.

Weisenfeld

G. G.

Gardiner

B. A.

Jost

T. M.

(2022). The state of preschool 2021: State preschool yearbook. National Institute for Early Education Research.

19.

Garver

Weisenfeld

G. G.

Connors-Tadros

Hodges

Melnick

Plasencia

(2023). State preschool in a mixed delivery system: Lessons from five states. Learning Policy Institute.

20.

Gray-Lobe

Pathak

P. A.

Walters

C. R.

(2023). The long-term effects of universal preschool in Boston. Quarterly Journal of Economics, 138(1), 363–411. https://doi.org/10.1093/qje/qjac036

21.

Henry

G. T.

Gordon

C. S.

Rickman

D. K.

(2006). Early education policy alternatives: Comparing quality and outcomes of head start and state prekindergarten. Educational Evaluation and Policy Analysis, 28(1), 77–99. https://doi.org/10.3102/01623737028001077

22.

Kagan

S. L.

Kauerz

(2007). Reaching for the whole: Integration and alignment in early education policy. In Pianta

R. C.

Cox

M. J.

Snow

K. L.

(Eds.), School readiness and the transition to kindergarten in the era of accountability (pp. 11–30). Paul H Brookes Publishing.

23.

Ladd

H. F.

Muschkin

C. G.

Dodge

K. A.

(2014). From birth to school: Early childhood initiatives and third-grade outcomes in North Carolina. Journal of Policy Analysis and Management, 33(1), 162–187. https://doi.org/10.1002/pam.21734

24.

Lipsey

M. W.

Farran

D. C.

Durkin

(2018). Effects of the Tennessee Prekindergarten Program on children’s achievement and behavior through third grade. Early Childhood Research Quarterly, 45, 155–176. https://doi.org/10.1016/j.ecresq.2018.03.005

25.

Little

(2020). Educators’ views on the location of pre-K programs and its relation to features of P–3 alignment: An exploratory study. Children and Youth Services Review, 118, 105455. https://doi.org/10.1016/j.childyouth.2020.105455

26.

Magnuson

K. A.

Ruhm

Waldfogel

(2007). Does prekindergarten improve school preparation and performance? Economics of Education Review, 26(1), 33–51. https://doi.org/10.1016/j.econedurev.2005.09.008

27.

McCormick

M. P.

Mattera

S. K.

Maier

M. F.

Xia

Jacob

Morris

P. A.

(2022). Different settings, different patterns of impacts: Effects of a pre-K math intervention in a mixed-delivery system. Early Childhood Research Quarterly, 58, 136–154. https://doi.org/10.1016/j.ecresq.2021.08.005

28.

North Carolina Department of Health and Human Services. (2017, February 1). Study costs and effectiveness associated with NC Pre-K slots. Fiscal Research Division. https://ncchildcare.ncdhhs.gov/Portals/0/documents/pdf/S/SL_2016-94_Sec_12B_4a-Study_Costs_and_Effectiveness_Associated_with_NC_PreK.pdf?ver=2018-10-22-094114-607

29.

Peisner-Feinberg

E. S.

(2003). Child and program characteristics of the North Carolina more at four pre-kindergarten program: Year 1 (January–June 2002). FPG Child Development Institute. https://fpg.unc.edu/publications/child-and-program-characteristics-north-carolina-more-four-pre-kindergarten-program-year-1

30.

Peisner-Feinberg

E. S.

Schaaf

J. M.

(2009). Evaluation of the North Carolina more at four pre-kindergarten program: A look across time at children’s outcomes and classroom quality from pre-K through kindergarten (2003–2009). FPG Child Development Institute. https://fpg.unc.edu/sites/fpg.unc.edu/files/resource-files/MAF_Yr8_full_report.pdf

31.

Peisner-Feinberg

Van Manen

Mokrova

Burchinal

(2019). Children’s outcomes through second grade: Findings from year 4 of Georgia’s pre-K longitudinal study. FPG Child Development Institute. https://files.eric.ed.gov/fulltext/ED598162.pdf

32.

Phillips

D. A.

Lipsey

M. W.

Dodge

K. A.

Haskins

Bassok

Burchinal

M. R.

Weiland

(2017). Puzzling it out: The current state of scientific knowledge on pre-kindergarten effects. Brookings Institution.

33.

Reid

J. L.

Ready

D. D.

(2013). High-quality preschool: The socioeconomic composition of preschool classrooms and children’s learning. Early Education and Development, 24(8), 1082–1111. https://doi.org/10.1080/10409289.2012.757519

34.

Reid

J. L.

Melvin

S. A.

Kagan

S. L.

Brooks-Gunn

(2019). Building a unified system for universal pre-K: The case of New York City. Children and Youth Services Review, 100, 191–205. https://doi.org/10.1016/j.childyouth.2019.02.030

35.

Watts

T. W.

Jenkins

J. M.

Dodge

K. A.

Carr

R. C.

Sauval

Bai

Escueta

Duer

Ladd

Muschkin

Peisner-Feinberg

Ananat

(2023). Understanding heterogeneity in the impact of public preschool programs. Monographs of the Society for Research in Child Development, 88(1), 7–182. https://doi.org/10.1111/mono.12463

36.

Weiland

Yoshikawa

(2014). Does higher peer socio-economic status predict children’s language and executive function skills gains in prekindergarten? Journal of Applied Developmental Psychology, 35(5), 422–432. https://doi.org/10.1016/j.appdev.2014.07.001

37.

Weiland

Unterman

Shapiro

Staszak

Rochester

Martin

(2020). The effects of enrolling in oversubscribed Prekindergarten programs through third grade. Child Development, 91(5), 1401–1422. https://doi.org/10.1111/cdev.13308

38.

Weiland

McCormick

Duer

Friedman-Krauss

Pralica

Xia

Nores

Mattera

(2024). The mixed-delivery pre-K opportunity gap? Differences in demographics, quality, and children’s gains in community-based versus public school programs across five large-scale systems. Early Childhood Research Quarterly, 68, 247–259. https://doi.org/10.1016/j.ecresq.2024.05.004

39.

Wilinski

(2017). When pre-K comes to school: Policy, partnerships, and the early childhood education workforce. Teachers College Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.74 MB

0.00 MB

0.74 MB