Replication Files for Upping the Ante: The Equilibrium Effects of Unconditional Grants to Private Schools, 2012-2014
Other Household Survey [hh/oth]
Private schools that rely entirely on student fees for financing are increasingly popular in many low-income countries and parents often prefer these schools to government-run ones. In Pakistan, children in these schools tend to outperform students in government-run schools. But financial constraints can limit the growth of these private schools, whose fees are set low to attract poor students, especially if they cannot access formal credit markets. Researchers from Pomona College, Harvard University and the World Bank have designed an impact evaluation to study private financing models - grants and loans - to support private schools in Pakistan.
The intervention centered on two financing approaches: a grant model and microloans. The program included a pilot microloan intervention to allow researchers to better develop and target loan products. This randomized control trial covered about 2,000 schools in about 650 villages across two districts in Punjab, Pakistan's most populous province.
Baseline, midline and endline surveys were conducted at both school and individual level. Within each survey, there were specific sections aimed to collect information from different perspectives. Thus, the survey initially included sections to be answered by the school owner, head teacher, class teacher, children, and operational head of the school. However, during the implementation process some changes were made in consultation with the World Bank's Strategic Impact Evaluation Fund (SIEF) and other parties involved. As a result, the final evaluation (or the endline survey) for this project shifted its focus to more specific objectives, concentrating on certain sections of initial surveys but also including additional components that would serve to the development of other projects.
The replication files for the associated American Economic Review (AER) Journal publication - Upping the Ante: The Equilibrium Effects of Unconditional Grants to Private Schools ("https://www.aeaweb.org/articles?id=10.1257/aer.20180924") are documented here for public use.
Kind of Data
Sample survey data [ssd]
Unit of Analysis
Edited datasets for public distribution
The scope of the study includes:
- child roster
- math, civics, Urdu and English test scores
- school enrollment and attendance
- current posted fees
- collected revenues
Rural areas of the district of Faisalabad in Punjab province.
The target population is low-cost private schools in rural areas of Faisalabad district. Sampling is at the village level, so urban, and peri-urban villages are excluded. Furthermore, for the design of the intervention, villages without any private schools, or with only one private school, are excluded. Villages with population over 10,000 or high village aggregated revenue are also excluded.
Government schools, and schools where money is not contained within the school itself (i.e. some network schools share money across multiple schools in the network), are also excluded.
Producers and sponsors
Asim I Khwaja
Center for Economic Research in Pakistan
Managing surveys and data entry
Strategic Impact Evaluation Fund
CNIC NSF Budget
Tameer Microfinance Ltd
All eligible schools that consented to participate across the 266 villages are included in the final randomization sample for the study. This includes 822 private and 33 NGO schools, for a total of 855 schools; there were 25 eligible schools (about 3 percent) that refused to participate in either the ballot or the surveys. The reasons for refusals were: impending school closure, lack of trust, survey burden, etc. Appendix Figure A1 of the Online Appendix (https://www.aeaweb.org/content/file?id=13118) summarizes the number of villages and schools in each experimental group.
Weighting with unequal selection probabilities is described in section A.3 of the Online Appendix (https://www.aeaweb.org/content/file?id=13118)
Dates of Data Collection
Data Collection Mode
Village Listing: This survey collects identifying data such as school names and contact numbers for all public and private schools in our sampling frame.
School Survey Long: This survey is administered twice, once at baseline in summer 2012 and again after treatment in the first follow-up round in May 2013. It contains two modules: the first module collects detailed information on school characteristics, operations and priorities; and the second module collects household and financial information from school owners. The preferred respondent for the first module is the operational head of the school, i.e. the individual managing day-to-day operations; if this individual was absent the day of the survey, either the school owner, the principal or the head teacher could complete the survey. For the second module, the preferred respondent was either the legal owner or the financial decision-maker of the school. In practice, the positions of operational head or school owner are often filled by the same individual.
School Survey Short: This survey is administered quarterly between October 2013 and December 2014, for a total of four rounds of data. Unlike the long school survey, this survey focuses on our key outcome variables: enrollment, fees, revenues and costs. The preferred respondent is the operational head of the school, followed by the school owner or the head teacher. Please consult Appendix Figure A3 of the Online Appendix (https://www.aeaweb.org/content/file?id=13118) to see the availability of outcomes across rounds.
Child Tests and Questionnaire: We test and collect data from children in our sample schools twice, once at baseline and once after treatment in follow-up round 3. Tests in Urdu, English and Mathematics are administered in both rounds; these tests were previously used and validated for the LEAPS project (Andrabi et al., 2002). Baseline child tests are only administered to a randomly selected half of the sample (426 schools) in November 2012. Testing is completed in 408 schools for over 5000 children, primarily in grade 4. If a school had zero enrollment in grade 4 however, then the preference ordering of grades to test was grade 3, 5, and then 6. A follow-up round of testing was conducted for the full sample in January 2014. We tested two grades between 3 and 6 at each school to ensure that zero enrollment in any one grade still provided us with some test scores from every school. From a roster of 20,201 enrolled children in this round, we tested 18,376 children (the rest were absent). For children tested at baseline,
we test them again in whichever grade they are in as long as they were enrolled at the same school. We also test any new children that join the baseline test cohort. In the follow-up round, children also complete a short survey, which collects family and household information (assets, parental education, etc.), information on study habits, and self-reports on school enrollment.
Teacher Rosters: This survey collects teacher roster information from all teachers at a school. Data include variables such as teacher qualifications, salary, residence, tenure at school and in the profession. It was administered thrice during the project period, bundled with other surveys. The first collection was combined with baseline child testing in November 2012, and hence data was collected from only half of the sample. Two follow-up rounds with the full sample took place in May 2013 (round 1) and November 2014 (round 5).
Investment Plans: These data are collected once from the treatment schools as part of the disbursement activities during September-December 2012. The plans required school owners to write down their planned investments and the expected increase in revenues from these investments— whether through increases in enrollment or fees. School owners also submitted a desired disbursement schedule for the funds based on the timing of their investments.
Public Use Files
Use of the dataset must be acknowledged using a citation which would include:
- the Identification of the Primary Investigator
- the title of the survey (including country, acronym and year of implementation)
- the survey reference number
- the source and date of download
Jishnu Das, World Bank, Asim I Khwaja, Harvard Kennedy School and Tahir Andrabi, Pomona College. Replication Files for Upping the Ante: The Equilibrium Effects of Unconditional Grants to Private Schools, 2012-2014. Ref. PAK_2013_FPPSIE_v01_M_v01_A_AEA. Dataset downloaded from [URL] on [date].
Disclaimer and copyrights
The user of the data acknowledges that the original collector of the data, the authorized distributor of the data, and the relevant funding agency bear no responsibility for use of the data or for interpretations or inferences based upon such uses.
DDI Document ID
Development Data Group
Documentation of the Study
Date of Metadata Production
DDI Document version