Understanding the targeting and impact of a parental involvement campaign in Peru

Intervention consists of 12 weeks of SMS (text messages) sent by the Ministry of Education to parents or guardians of fourth grade students enrolled in treated classrooms in the sample of public elementary schools. Two messages are sent each week. The first text message of each week provides information about a best practice for parents to follow in supporting the academic progress of their student. This first message has identical content for all treated students. The second text message of the week varies by treatment status. Those in the "planning" treatment group receive messages prompting them to make a specific plan for implementing this advice. Those in the "action" treatment group receive messages prompting them to take immediate action to implement the advice.

2019-09-27

2019-11-25

1. National standardized exam subscores (math and communication)
2. Grade point average from national student registry

1. Disaggregated student grades by subject
2. Teacher attendance rate
3. Student attendance rate (if available from national student registry)

The sample consists of 117 schools with three sections (classrooms) of fourth grade students, according to the 2019 school census. Randomization is at the classroom level and stratified by school, such that a school with exactly three sections has one section in the control group (C), one section in the planning treatment group (T1), and one section in the action treatment group (T2). After drawing the sample and contacting schools, it was determined that 1 school had only 1 section, 8 schools had 2 sections, and 10 schools had 4 sections. Schools with 1 or 2 sections had no more than 1 section assigned to C, T1, or T2, while schools with 4 sections had at least 1 but no more than 2 sections assigned to C, T1, or T2. The stratified randomization was performed in such a way that enforced global balance in C/T1/T2 assignment.

All students within a classroom were assigned to the same treatment status.

Analysis will use regression adjustment to control for prior achievement (lagged grades), accounting for clustering at the appropriate level as determined by MacKinnon, Nielsen, and Webb (2020). Benjamini, Krieger, and Yekutieli (2006) will be used to adjust for multiple testing.

Prior to randomization, teachers are asked to indicate which students/families they believe would benefit most from a parental engagement intervention. This provides the key variable used in the study's main heterogeneity analysis, which will interact treatment with a dummy variable for whether the student is listed by the teacher.

At baseline, parents answer questions related to 1) their level of involvement in their child's education and 2) their propensity to plan. These will be used to create parental involvement and planning indices, following Anderson (2008). These indices will be interacted with treatment to estimate heterogeneous treatment effects.

Randomization done in office by a computer

Classroom

Yes

349 classrooms

Approximately 6,000 students

117 classrooms: Control
116 classrooms: Information + planning prompt
116 classrooms: Information + action prompt

Unadjusted for multiple testing: 0.10 standard deviations (unit of measurement is normalized score) Adjusted for multiple testing: MDE depends on correlation structure of p-values. Benjamini, Krieger, and Yekutieli (2006) will be used to adjust for multiple testing.