The rise of fine-grain geospatial data has the potential to change education policy planning and increase educational equity and efficiency across the globe. In our paper published in the academic journal Development Engineering, we make available a data-driven, open-source framework that uses freely-available population data from WorldPop to precisely identify populated areas in low- and middle-income countries (LMICs) that are not currently served by public primary schools. Identifying these “education deserts” serves as a first step towards making schools physically accessible to all learners, especially the most marginalized.
Challenges with universal enrollment remain
In spite of the rapid rise in primary school enrollment worldwide, 1 of every 6 school-age children in LMICs were out of school in 2018 (UNESCO). Having a school physically accessible is the first-order necessity for attending school, and long, sometimes dangerous distances to a school can play a determining role in a student’s decision not to attend. Systematically identifying localized areas where populations still lack physical access to schools can offer critical insights to inform targeted interventions that remove barriers to school enrollment, like school construction, remote education offerings, or improved access to school transportation options. This is particularly relevant since the typical aggregate enrollment rates that policymakers have historically relied on, even at a sub-regional level, may not tell the whole story about pertinent barriers to education at a hyper-local level, as we discuss in our paper and show in Figure 2 below.
What are “education deserts” and how can decision-makers locate them?
‘Education deserts” are populated areas that do not have reasonable physical access to a school (the definitions of ‘reasonable’ and ‘school’ may be defined differently for each context). We develop a highly scalable and open-source tool – which can be systematically applied to almost any country or region in the world – to precisely pinpoint education deserts, as we illustrate for Guatemala below. We calculate the shortest straight-line distance between each “population pocket” of 1km2 in the country and the nearest public primary school. Since we know how many people live in each pocket (from the geolocated population data), we can then calculate statistics that describe the dynamics of physical access in unprecedented detail across the country. An additional benefit of our approach is that the population data we use are freely available, so this framework can be applied to any circumstance as long as analysts are able to obtain location data for relevant schools. Importantly, we extend our methodology so it can incorporate geographical features, other approaches to measuring distances, and local policy rules about school placement and construction, making our approach a reliable and flexible starting point for the optimal planning of school placement in LMICs.
A 3-D geographic distribution of the Guatemalan population that is at least 3 km away from a public primary school
We also provide an algorithm to prioritize the construction of new schools in specific education deserts given their relative population sizes, set budget constraints, or minimum targets of population reached by each facility, to ensure a more efficient placement of schools. In other words, we can help answer the question: “If policymakers were to invest in school construction, what construction locations would most reduce the share of population living in education deserts?”
Using GIS tools to identify education deserts in Guatemala
Applying this methodology using data from Guatemala in 2017, we find that education deserts are somewhat rare, but highly localized, and that a relatively few but strategically placed schools could significantly universalize physical access to education. Only 5% of the population lives further than 3 km from a public primary school, and the median Guatemalan person lives 0.8 km from a public primary school. We also find that typical enrollment rates are only weakly correlated with physical access to schools at the department level, highlighting that our analyses do indeed provide novel information about local barriers to enrollment in a given area, as we show in the figure below.
The percentage of regional population living in education deserts compared to enrollment rates
If policymakers had placed half of all new schools opened between 2008 and 2017 using our school placement algorithm, there would not be a single person living in an education desert by 2017. In a more realistic scenario, we find that if only 1 in 20 of the new schools opened between 2008 and 2017 had been optimally-placed using our algorithm, Guatemala would have seen the same reduction in education deserts that it actually experienced with 20 times as many schools placed under their status quo policy. Given that the elimination of an education desert is not (and should not) be the sole goal for the opening of a new school, this result is especially hopeful and actionable for policymakers: it roughly indicates that – at least in the Guatemalan context – substantial strides in physical access can be made even if only a small fraction of new schools are constructed with physical access in mind.
Identifying education deserts can support last-mile enrollment
The type of disaggregated, fine-grain analyses that our methodology yields can be especially valuable as policymakers and investors attempt to guarantee universal access to education for every child around the world. The systematic identification of areas with no physical access to school is a first step towards more inclusive and equitable education systems. Unfortunately, the regions where it is most valuable to identify education deserts are often the same regions where traditional, aggregate administrative data is typically most lacking, and, in turn, where our methodology can be the most beneficial when coupled with local expertise.
As developing countries traverse the last mile on primary enrollment and the demand for secondary and tertiary education rises, policymakers need to consider how many new schools to build and where – to support goals such as universal primary enrollment or secondary education. Ultimately, having a school nearby is a necessary, though not in itself sufficient, first step towards supporting children to achieve their full potential through quality education.
How to access these tools and codebase to identify more education deserts globally
We deliver all of these analytic components in an extensively documented open-source codebase designed with a “plug-and-play” utility that can be used by policymakers, practitioners, and researchers. The entirety of our analysis can be replicated across contexts with minimal effort, low computational requirements, and zero cost, as all requisite software and packages used in our analysis are also free and open-source. This codebase is free and publicly available here where the code can be downloaded and adapted by other analysts. Individuals can submit a pull request on Github with any inquiries about the code. For the full analysis and more detailed methodology, please access our paper published in Development Engineering.
This blog was drafted on behalf of the GIS for Education Working Group, convened by the Education Commission and co-chaired by the EdTech Hub. Thanks to Jessica Bergmann (Education Commission) for her inputs and contributions.