|
Field
Abstract
|
Before
This research builds on a pre-study that assessed the content knowledge in mathematics of a representative sample of primary school teachers in Morazan, El Salvador. The results of the pre-study show that the average math teacher scores less than 50% correct answers on an exam covering the official mathematics curriculum for second to sixth graders, i.e. the material they are supposed to teach. This randomized controlled trial aims at evaluating whether and to what extent the content knowledge of primary school teachers can be improved during a 5 months in-service teacher training. The teacher training program evaluated in this trial has two elements: 1) Self-studying using a computer-assisted learning software, and 2) participation in four workshops where the material covered in the self-study modules is recapped and discussed together with expert teachers. Importantly, both the software-based self-studying and the attendance in workshops are incentivized, meaning that the participants get compensated for each computer-assisted learning module they successfully complete as well as for their attendance in the monthly workshops.
|
After
Research on basic education in developing countries has recently been shifting its focus from accessibility to schooling towards the quality of schooling. Inevitably, this will put the performance of teachers at the center of the debate. While recent data from African countries, India and El Salvador document alarmingly low levels of teacher content knowledge, we lack rigorous evidence as to how this problem may be tackled. This field experiment aims at providing novel insights on the effectiveness of subject-based teacher training programs. The evaluated intervention targets primary school math teachers and consists of a five-month in-service training combining (1) computer-assisted self-studying and (2) monthly workshops. The design of the study allows to identify the causal effect of the implemented teacher training program on the content knowledge of teachers and the math skills of their students.
|
|
Field
Trial End Date
|
Before
September 06, 2019
|
After
October 16, 2020
|
|
Field
JEL Code(s)
|
Before
I20, I21, I28, O15
|
After
C93, I20, I21, I28, O15
|
|
Field
Last Published
|
Before
April 17, 2019 08:19 PM
|
After
June 28, 2019 01:27 AM
|
|
Field
Primary Outcomes (End Points)
|
Before
Math skills of primary school teachers measured via pencil and paper assessments covering the local curriculum of grades 2 -- 6.
|
After
Math skills of primary school teachers measured via pencil and paper assessments covering the local curriculum of grades 2 -- 6. We also assess their students' math ability at the end of the school year 2020 based on grade-specific assessments.
|
|
Field
Experimental Design (Public)
|
Before
Starting point are all primary schools in Morazan, excluding the 50 smallest schools with only one or two teachers covering grades 3--6. The NGO sent out invitations to all eligible teachers in these schools, who were asked to register via telephone, internet or filling out a registration form.
Based on these registrations the applicants were invited to an information meeting in March 2019, where an unannounced baseline math assessment was conducted. Some of the applicants took exactly the same assessment during a representative survey in September 2018 (about 6 months before), and therefore only completed the registration and received additional information about the program.
Based on the baseline assessments, the worst performing applicant of every school was selected; applicants were informed that participation in the March-meeting did not guarantee admission to the program, but the pre-selection criteria were not communicated to them. Following this procedure 175 teachers from 175 different schools across Morazan remained in the sample.
Stratifying on baseline scores and teacher gender, the 175 teachers were randomly assigned to the either the control or treatment group. This gives 87 teachers that participate in the program and 88 teachers constituting the control group.
|
After
Starting point are all primary schools in Morazan, excluding the 50 smallest schools with only one or two teachers covering grades 3--6. The NGO sent out invitations to all eligible teachers in these schools, who were asked to register via telephone, internet or filling out a registration form.
Based on these registrations the applicants were invited to an information meeting in March 2019, where an unannounced baseline math assessment was conducted. Some of the applicants took exactly the same assessment during a representative survey in September 2018 (about 6 months before), and therefore only completed the registration and received additional information about the program.
Based on the baseline assessments, the worst performing applicant of every school was selected; applicants were informed that participation in the March-meeting did not guarantee admission to the program, but the pre-selection criteria were not communicated to them. Following this procedure 175 teachers from 175 different schools across Morazan remained in the sample.
Stratifying on baseline scores and teacher gender, the 175 teachers were randomly assigned to the either the control or treatment group. This gives 87 teachers that participate in the program and 88 teachers constituting the control group.
To assess the impact of the intervention on the students' math abilities, an assessment with the students of the 175 teachers in the sample will be conducted in September / October 2020 (i.e. towards the end of the school year 2020). If a teacher teaches math in more than one class, we randomly sample one class per teacher to conduct the student assessments and collect the monitoring data.
|
|
Field
Randomization Unit
|
Before
We randomize on the teacher level stratifying on their gender and baseline score. However, the pre-selection based on the baseline score leaves only one teacher per school in the sample.
|
After
We randomize on the teacher level stratifying on their gender and baseline score. Note that the pre-selection based on the baseline score leaves only one teacher per school in the sample. Hence, for teachers the randomization of the treatment was not clustered, while for students it is clustered on the teacher level.
|
|
Field
Planned Number of Clusters
|
Before
175 teachers (and schools)
|
After
175 teachers (and schools) and 175 classes (one class per teacher).
|
|
Field
Planned Number of Observations
|
Before
88 teachers (and schools) in the control group, 87 teachers (and schools) in the treatment group.
|
After
88 teachers (and schools) in the control group, 87 teachers (and schools) in the treatment group. The students' of these teachers are also assessed in September / October 2020 (i.e. towards the end of the school year 2020); while the class sizes may change at the start of the new school year in January 2020, we expect on average about 18 students per teacher.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
88 teachers (and schools) in the control group, 87 teachers (and schools) in the treatment group.
|
After
88 teachers (and schools) in the control group, 87 teachers (and schools) in the treatment group. About 1580 students in the control and 1580 students in the treatment group (this is an estimate as class sizes may change at the start of the new school year in January 2020).
|
|
Field
Power calculation: Minimum Detectable Effect Size for Main Outcomes
|
Before
MDE = 0.2--0.27 standard deviations. Calculations based on formula by Bloom (2007) and the following parameter values: power=80%; alpha (level of significance)=0.1; R-squared : 0.5--0.7; P (share of control units)=0.5; n (total observations)=175
|
After
MDE_TEACHERS = 0.23--0.30 standard deviations. Calculations based on formula by Bloom (2007) and the following parameter values: power=80%; alpha (level of significance)=0.05; R-squared : 0.5--0.7; P (share of control units)=0.5; n (total observations)=175.
MDE_STUDENTS = 0.14--0.17 standard deviations. Calculations based on formula by Bloom (2007) and the following parameter values: power=80%; alpha (level of significance)=0.05; R-squared (between): 0.32--0.53; R-squared (within): 0.03--0.24; Rho (Interclass correlation)=0.13--0.20; P (share of control units)=0.5; J (total clusters)=175; n (observations per cluster)=18
|
|
Field
Secondary Outcomes (End Points)
|
Before
|
After
Class cancellation rates and student attendance rates based on at least four unannounced visits per class throughout the school year 2020. This yields about 700 data points, 350 for the control group and 350 for the treatment group.
|