This impact evaluation uses random assignment at the village level to estimate impacts of the NECS and IMAGINE programs on enrollment, attendance, learning and other education outcomes. Wave 1 data were collected in fall 2013. Follow-up data will be collected in spring 2015.
The EGRA descriptive study will measure reading skill of students enrolled in CI or CP and attending school on the day of data collection in a sample of NECS intervention schools. Round 1 data were collected in spring 2014.
Kind of data
Sample survey data [ssd]
Anonymized dataset for public distribution
basic skills education
compulsory and preschool education
Niger Threshold Program
Early education assessments
Early grade reading
The program is being implemented in rural villages throughout Niger.
Unit of analysis
Individuals, households, and community
Households and primary school aged children (ages 5 - 14) living in rural Niger.
Producers and sponsors
Mathematica Policy Research
Millennium Challenge Corporation
NECS: The sample contains 204 villages, from which 40 households with school age children were to be selected per village. We encountered 22 villages that had fewer than 40 eligible households, leading to 267 fewer households in the sample than initially anticipated. As a result, our sample contains information from 7,893 households, 17,266 children between the ages of 5-14. Village level information was also gathered, as was information that was directly observable for the main school in each village. Information for 202 schools is in the sample. The response rate is 99.9% for households, 94.6% for individual children, and 99.0% for school infrastructure.
EGRA: The target sample size was 1,000 students, from across 21 schools, with 50 students per school (25 from first (CI) grade and 25 from second (CP) grade). As the data collection was underway, some schools had less then the anticipated number of students, so additional schools were added until we reached the desired sample size. In total, the final sample consisted of 1,010 students from 27 schools - 600 in CI and 410 in CP.
Deviations from sample design
There are no deviations from the sample design.
NECS: The response rate for households is 99.9%, individual response rate for the children is 94.6%, and school response rate is 99.0%.
EGRA: The response rate for individuals is 99.7%.
There are no weights included in the public use data.
Dates of collection
NECS Wave 1
EGRA Round 1
Data collection supervision
NECS: The data collection team hired and trained 56 interviewers to collect household and school data. They were organized by linguistic groups into 14 teams, with each team comprised of 3 interviewers led by an experienced field supervisor. The teams were then assigned a region and surveys were conducted simultaneously throughout the country.
EGRA: The data collection team hired and trained 18 interviewers to collect data. They were organized by linguistic groups into 4 teams, with each team led by an experienced field supervisor. The teams were assigned schools corresponding to their linguistic group and surveys were conducted simultaneously throughout the country.
NECS: Village, household, individual
EGRA: Individuals, school
Centre International d’Etudes et de Recherches Sur Les Populations Africaines
NECS: The local data collection firm entered and cleaned the data using The Census and Survey Processing System (CSPro). We implemented double data entry of 10% of the sample and checked the accuracy of the double-entered data with the data that was originally received. The data entry error rate was very small—2.05 percent when checking at the household level and 0.65 percent at the individual level—and therefore no additional double entry was conducted. We also conducted checks on the data set for out-of-scope responses, item nonresponse, and inconsistent patterns. In addition, we tested and confirmed the ability to merge the household and school data. Inconsistencies and potential errors were identified and corrected when they were the result of data entry error. Mathematica staff devised a data cleaning protocol designed to resolve inconsistencies and out-of-range data that were not a result of data entry error.
EGRA: Following completion of data collection activities, the local data collection firm entered and cleaned the data using The Census and Survey Processing System (CSPro). We implemented double data entry of 50% of the sample and checked the accuracy of the double-entered data with the data that was originally received. The data entry error rate was very small at 0.34%.