Childcare expansion, Maternal employment and Child outcomes

There are two interventions to first-grade students attending Neulbom schools in the first semester of 2024. The first intervention is two programs free of charge from 1PM to 3PM, tailored to the developmental needs of younger students. These programs include curriculum-based educational activities (such as Korean, reading and discussion, mathematics, science, English, and computers), and play-centered physical education, arts, and music, and other activities (such as Lego, linking cubes, board games, cross blocks, I-ringos, and wooden blocks). The first-grade students uniquely benefit from the programs of Neulbom schools in the first semester of 2024. Students can decide to join the entire program or receive care services only.
The second intervention is to remove priority restrictions for applying for the childcare services, unlike the pre-existing after-school programs and childcare services which give priority to children from dual-income families, low-income households, single-parent families, and those recommended by teachers.

2024-03-02

2024-07-26

Employment(women's employment status, women's work hours), Education(amounts of money spent on child education in private sector, studying hours in private sector, students' cognitive and non-cognitive ability (e.g. number of close friends, reading the mind in the eyes test)), Household finance(expenditure),Life satisfaction (for both students and mothers)

Employment(number of workers employed at elementary schools (e.g. temporary workers for childcare management), women's work type), Education(satisfaction on childcare services), Family(desired number of children)

1. Empirical Analysis
To evaluate the effects of the Neulbom school public childcare expansion in Korea on maternal employment, child development, the substitution of private care services, the number of workers hired by schools, and the desired number of children, we exploit the quasi-experimental variations in the introduction of Neulbom schools across grades and schools in the first semester of 2024.
We employ a difference-in-difference framework by estimating equation (1) for students in grade1-4:

Y_ijg=α_0+τ_g+λ_j+δ_y1 TS×TG+X_ijg^' β_0+ε_ijg (1)

where TS is a dummy variable indicating whether schools are Nuelbom schools (treated school) and TG is also a dummy variable indicating whether the grade is first grade in Nuelbom schools (treated grade). The subscript i,j,g refer to students, schools and grades, respectively. Y includes several outcomes as explained in prior subsection. X includes individual and household characteristics such as household income, sex of students and whether the household is a multicultural family.
The parameter of interest of is δ_y1 which captures the differences in the aforementioned outcome variables of interest between the first- and other-grade students in Neulbom schools compared to those of non-Neulbom schools. Even in Neulbom schools, students in first grade and their parents can choose to utilize the programs offered by Neulbom schools. According to the MOE data, 74.3% of first-grade students are attending Neulbom schools. Therefore, δ_y1 captures the ITT (intention-to-treat) effect.

We examine the effect of Neulbom schools on several outcomes for compliers-those who utilize the programs and childcare services provided by Neulbom schools-by estimating equation (2) and (3) using an instrumental variable and a difference-in-difference model. We predict the proportion of students utilizing the services offered by Neulbom schools with equation (2) and use it in equation (3) to estimate the parameter of interest, δ_IV, which captures the effect of Neulbom schools on compliers.

P_ijg=α_1+τ_g1+λ_j1+δ_P TS×TG+X_ijg^' β_1+η_ijg, (2)
Y_ijg=α_2+τ_g2+λ_j2+δ_IV (hat of P_ijg)+X_ijg^' β_2+ψ_ijg, (3)

where we follow the same notations as in equation (1). P is a dummy variable indicating whether students attend Neulbom schools or not. We predict the proportion of students utilizing the services offered by Neulbom schools with equation (2). Then, to estimate the parameter of interest, δ_IV, which captures the effect of Neulbom schools on compliers, we estimate equation (3) with the predicted value of P, P ̂.
We acknowledge that δ_y1 or δ_IV may capture the effects of other events occurring in 2024 other than the introduction of Neulbom school on private tutoring expenses and maternal employment. To enhance the robustness of our analysis, we estimate the false effects of the Neulbom schools by using data from 2023 and compare those with the effects during 2024, using a difference-in-difference-in-difference (DDD) framework. In particular, we estimate the below equation (4):

Y_ijg=α_3+β_31 TT+β_32 TS+β_33 TG+γ_31 TT×TS+γ_32 TT×TG+γ_33 TS×TG+δ_y2 TT×TS×TG+φ_ijg, (4)

where we follow the same notations as in equation (1). TT is a dummy variable indicating the timing of the introduction of Neulbom schools. δ_y2 captures the differences between the effect of Neulbom schools in 2024 and 2023.

2. Data Collection and Key Outcomes
We are collecting data through surveys for sampled students and parents to use detailed individual level information on education and employment in the first half of 2024 and 2023. Our research team, in collaboration with MOE, developed a survey instrument, and we summarize our main outcome variables above. These are categorized into five main areas: 1) employment, 2) education, 3) household finance, 4) life satisfaction, and 5) family. Note that we are collecting data for both 2023 and 2024 based on participants’ recollections, enabling us to conduct robustness analysis and verify that the behaviors are consistent across students within the same school.

(To present the experimental design decently, attached is pdf file in Material section.)

[Sampling Procedure]
As we cannot randomly assign participants to treatment and control schools and grades, we rely on the exogeneity of the policy on beneficiaries, which allows us to exploit quasi-experimental setting. We explain how we "sample" schools for the treatment and control groups. Treatment schools are Neulbom schools and treatment grade is first grade. Control schools are non- Neulbom schools and control grades are second-to-fourth-grade students.
Neulbom schools and non-Neulbom schools were randomly selected based on random numbers generated by a computer to maintain a 3:2 ratio, proportionate to the total student population. Grades were then randomly assigned to each school to ensure an equal distribution of treatment grades (first grade) and control grades (second to fourth grades) in a 1:1 ratio. In each school, we survey one class from one grade.

School because the Neulbom school policy, from which this study exploits quasi-experimental setting, is implemented at school level

No

150 - 200 schools

2,000-2,500 students and their mothers.

100 – 133 schools, 1300 – 1600 students and their mothers.