Facilitating innovative growth of low cost private schools: experimental evidence from Pakistan 2016-2019


The data contain information on 837 low-cost for-profit private schools (LCPS) from three districts in Punjab, Pakistan: Faisalabad, Gujranwala, and Sialkot. The past few decades have seen an exponential increase in the growth of these LCPS globally, and in countries like Pakistan and India, the private sector now commands a large and quickly increasing share of the market. Over forty percent of primary school enrolment in Pakistan is now in LCPS, and students in private schools in Pakistan far out-perform those in public schools. Yet, firm innovation and expansion is constrained for private schools, likely due to a range of supply-side and market level failures. The main research questions this study and the uploaded dataset seek to answer are: (1) To what extent are schools constrained by finance, and does the type of financing vehicle (loan vs equity) matter? (2) Is LCPS quality improvement constrained by a lack of access to appropriate quality-enhancing products and services, i.e. educational support services (ESS)? (3) Is there a positive interaction between access to finance and the provision of appropriate innovative investment opportunities? The dataset includes topics such as school administration, facilities, fees, enrolment, student population, finances, and financial expectations and literacy. Schools are uniquely identified using the variables mauza (administrative district) code and school code. While most of the variables are school-level, there are a few individual-level data pieces that were collected from the school owner. For each school we interviewed only one owner, therefore both schools and school owners are identified using the same mauza code and school code ID.Most interventions to improve education in developing countries require spending significant amounts of money on improving the quality of the inputs to the education system. While this is often a useful approach, in countries with weak governments and low tax collection, little resources are available to invest in schools. In these settings, such as in Pakistan, private schools have provided an alternative to the low quality public schools, and parents are willing to pay for the improved quality, and so even in many remote rural areas, parents can pick from sending their child to the public school or one of several private schools in the village. This variety of schools has prompted us to study education markets instead of the inputs to the production of learning, applying theories from studying Small and Medium Enterprises (SMEs) to private schools. Instead of going to schools and telling them which inputs they should focus on, we tend to ask them what prevents them from expanding in quality and quantity. Over the past decade, our research team led by Tahir Andrabi (Pomona College), Jishnu Das (World Bank), and Asim Ijaz Khwaja (Harvard University) has studied the education markets in Pakistan. Despite the tremendous growth in the low cost private school (LCPS) sector (rising from 3,300 schools in 1983 to over 70,000 in 2011) and relatively better quality than the public sector (LCPS are 1.5 years ahead in learning outcomes relative to government schools), there is also evidence of substantial untapped innovation potential in this sector. The team has gathered both primary data and implemented randomized controlled trials (RCTs) that reveal constraints to growth and quality improvement for LCPS. Two factors that contribute to this innovation constraint are the lack of financing (a financial market failure) and access to affordable educational support services (ESS) (an input market failure), which together create a very challenging context for school owners. The current project is a RCT that seeks to explain how alleviating these constraints one at a time or simultaneously would affect learning outcomes, enrolment and school profitability. The randomized component means that schools are randomly allocated to either receiving offers of a loan product or an equity product to alleviate financial constraints, and/or receive access to buying ESS such as teacher training, improved curricula or student testing services. The controlled component of the trial means that some - randomly chosen - schools do not get any of these treatments, which allows us to compare the treatment outcomes with the counterfactual. The two financial products were developed together with one of Pakistan's largest microfinance banks. The equity product represents an innovation in the type of financial product offered to SMEs, and it is particularly relevant to LCPS since its revenue-contingent interest rate (if the school earns more, it will pay a higher interest rate) effectively shifts some of the risk of an investment over to the bank, and LCPS tend to have to make more lumpy investments than other SMEs. Our theory is that a less risky financial product would allow schools to take on more risky investments, such as investments in quality-improving ESS services. Our earlier studies have shown that schools are, under most circumstances, more prone to undertake capacity-improving investments such as buying more chairs when they get access to finance, perhaps because the school owner can more accurately predict and advertise the result from buying new chairs than from buying teacher training services. This results in a stagnant quality development of learning outcomes, which are already very poor in rural Pakistan. In conclusion, the present study will broaden the knowledge about what policy makers and providers of financial and ESS products can do to facilitate the improvement of learning outcomes in challenging rural contexts where LCPS are present.

The majority of data collection took place on pen and paper, however some sections were collected on tablets using the program Open Data Kit (ODK). We arrived at the sample of schools included in this report through a multi-stage process, beginning with a complete listing of private schools from our previous work in the area. First, we limited this list to private schools in rural areas of Faisalabad, Gujranwala, and Sialkot districts. These districts were chosen for logistical reasons since they are (or border) districts in which we have previously worked (randomly drawn from districts in the Punjab), and we restricted to rural areas primarily because the financial needs of urban schools are significantly larger than our partner can prudently offer. We then arrived at our final sample by further restricting to schools that had expressed interest in receiving financing, which was done to decrease our required sample size since we expect take-up rates to be close to 60% for this screened-in population (as opposed to 20% for the general population). Treatment/Control randomization was done after the non-interested schools had been screened out, so both groups are completely comprised of schools that expressed a need for financing. Thus, our treatment effect will be unbiased with respect to any school/school owner characteristics that interest in financial services may signal. Using this procedure, we have surveyed 999 schools (total enrolment of 176,030 students) over the course of 4 rounds, of which 908 remain in the final sample. Each round lasted roughly two months, with the exception of Round 2, which took significantly less time due to the smaller number of schools being surveyed. Great care was taken in collecting this data to ensure that it is accurate. For example, enumerators took revenue, enrolment, and posted fee figures directly from the registers whenever possible (registers were available in 94% of schools) rather than rely on school owner memory. Furthermore, enumerators were given manuals with detailed instructions on how to record data under a variety of likely register structures (e.g. how to record enrolment data by grade if the register does not group children by grade) to maintain consistency across schools. Throughout surveying, supervisors (1:3 supervisor to enumerator ratio) made random checks to verify that these procedures were being followed, and all registers were photographed at the time of surveying to ensure that data was accurately recorded. After data had been collected a number of backchecks were conducted to ensure internal consistency.

DOI https://doi.org/10.5255/UKDA-SN-853776
Metadata Access https://datacatalogue.cessda.eu/oai-pmh/v0/oai?verb=GetRecord&metadataPrefix=oai_ddi25&identifier=5b8aba268d76ea81c46e561db09c04c8c1535c77b7813e3b2407484cf32e4e12
Creator Khwaja, A, Harvard Kennedy School; Andrabi, T, Pomona College; Das, J, World Bank
Publisher UK Data Service
Publication Year 2020
Funding Reference Economic and Social Research Council
Rights Asim Khwaja, Harvard Kennedy School. Tahir Andrabi, Pomona College. Jishnu Das, World Bank; The Data Collection is available to any user without the requirement for registration for download/access. Commercial use of data is not permitted.
OpenAccess true
Resource Type Numeric
Discipline Economics; Social and Behavioural Sciences
Spatial Coverage Faisalabad, Gujranwala, and Sialkot; Pakistan