巴基斯坦的失学儿童在哪里？利用卫星数据和机器学习方法绘制学校可达性地图，并应用于2022年的洪水

Policy Research Working Paper Where Are Pakistan’s Out-of-School Children? Mapping Accessibility of Schooling Using Satellite Dataand Machine Learning Methods with an ApplicationPublic Disclosure Authorized Umar SaifAmer HasanAyesha TahirKoen GevenMohsen Ali Policy Research Working Paper11341 Abstract According to administrative data, as many as 25 millionchildren between the ages of 5 and 16 are out-of-schoolin Pakistan. However, little is known about where thesechildren are in relation to available public and privateschooling options. This paper brings together novel datafrom satellite imagery, the population census, governmentschool censuses of public schools, and publicly availablelocation data on private schools to understand the out-of-school phenomenon at the community level. The paper uses of the total) are enrolled in private schools, higher thanpreviously reported official estimates. The paper also pres-ents detailed data on how the out-of-school population,accessibility, and whitespace are distributed geo-spatiallyacross Pakistan. As an application of the strength of usingsatellite imagery, it analyzes the effect of a major flood eventin 2022 on accessibility, whitespace, and catchment areas.Doing so suggests that the 2022 floods affected the school-ing of approximately 3.5 million children in Pakistan. Lastly, This paper is a product of the Education and Skills Global Department. It is part of a larger effort by the World Bank toprovide open access to its research and make a contribution to development policy discussions around the world. PolicyResearch Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted atahasan1@worldbank.org. A verified reproducibility package for this paper is available athttp://reproducibility.worldbank. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about developmentissues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry thenames of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those Where Are Pakistan’s Out-of-School Children? Mapping Accessibility of Schooling UsingSatellite Data and Machine Learning Methods with an Application to the 2022 Floodsª Umar Saifa, Amer Hasanb, Ayesha Tahird, Koen Gevenb, Mohsen Alic, Aliza Masooda, Muhammad AbdulRahmanc, Sheena Fazilib Keywords:public schooling, private schooling, out-of-school children, machine learning, satellite imagery,floods, JEL classification:I21, I30 1.IntroductionRecent improvements in the precision of satellite data as well as in analytical approaches have allowed researchers to understand development even in data-poor settings. These techniquesallow researchersto movebeyond the limitationsoftraditional data gathering techniquessuch as surveys and administrativedata.But importantly, satellite data now provides a highspatial resolution that is potentially more preciseeven in areas where traditional techniques penetrate.In recent years, researchers have therefore usedsatellite data to map a variety of topics such as economic growth(Hendersonet al, 2012), local economicactivity (Gibson et al, 2021), poverty (Asher et al, 2021),school test scores(Runfolaet al, 2021),deforestation(Burgesset al, 2012), and groundwater management (Ryan & Sudarshan, 2022). This paper Pakistan’s schooling landscape features a large public system alongside a vast network of private schools,both in urban and rural areas. Early school surveys in Pakistan documented the rise and penetration oflow-cost private providers in rural areas (Andrabiet al, 2007). Since then, the literature has analyzed theeducation system in Pakistan as a joint public–private market rather than as two separate worlds (Andrabi,Das, and Khwaja 2008; Nguyen and Raju 2014). Research has focused on the interplay between the publicand private sectors in rural education markets (Andrabi et al; 2024) as well as how public-privatepartnership schools fill important gaps in service delivery in rural Pakistan (Barrera-Osorio et al. 2022; Official statistics haveimproved in recent years, and now provide a more stable picture of the public systemat the aggregate level, but it is less clear what publicly availableofficial statistics can tell us about the private sector.1The size of the public-school sector (primary through higher secondary) has remainedlargely static over time,increasingmarginally from 149,961 to 150,689schoolsbetween 2016 and2022.Data from theEconomic Census (released in August 2025) show a slightly lower estimate of143,599public schools across Pakistan. Due to a lack of effective regulationand the challenges of monitoring atscale, it has been harder for government to estimate the number of private schools. Until recently,thegovernment has relied on extrap