The Equity Equation Project
Evidence from the DARE-RC Programme
Affiliated Institutions: Lahore University of Management Sciences (LUMS), DARE-RC
Geographical scope: Lahore Division
What is the context?
Pakistan currently grapples with a severe national education crisis, with approximately 26.2 million children, nearly one-third of the school-age population, remaining outside the formal schooling system. While the government has declared an Education Emergency, the National Education Policy Development Framework (2024) notes that the system’s primary failure is not merely a lack of data, but the inability to use that data to create targeted policies. Currently, decision-making often relies on broad guesses that fail to capture the local, real-life reasons why a family keeps a child at home.
The Equity Equation Project (EEP) addresses this gap by moving away from simply reporting past failures to predicting future solutions. By placing the family unit at the center of analysis, the project’s simulation model recognizes that enrollment is not driven by education policies alone; rather, it is shaped by a mix of poverty, gendered expectations, and safety concerns that define the lived reality of marginalized children.
The overarching objective of this project is to strengthen evidence-based planning. By integrating government records with local household and geospatial (GIS) data, the EEP allows policymakers to simulate how different families respond to specific policies, such as building a new school or providing transport, before any funds are committed. This ensures that the limited government resources are spent where they will have the greatest impact on Pakistan’s most vulnerable children.
What did we learn?
● With 26.2 million children out of school, the primary obstacle behind Pakistan’s education crisis is not only a lack of data, but the inability to use existing data to make targeted, evidence-based policy decisions.
● The Equity Education Project (EEP) has developed a model that allows policymakers to test potential interventions by predicting how families will respond to a particular policy, before they commit public funds. For this, the government does not need new data. The predictive model adapts existing government datasets (PSLM, MICS), and uses ‘Explainable AI’ so officials can explore the thinking behind each recommendation.
● Children are left out of school for a mix of reasons that vary from one neighborhood to the next so generic ‘one-size-fits-all’ policies will not work. In rural areas, girls are especially affected and infrastructure gaps like missing toilets and boundary walls matter, whereas in urban areas, poverty and low access to digital devices are the bigger obstacles. Given these nuances, interventions must be tailored to local needs and can first be tested out in the predictive model.
● The model’s simulation data found that distance cancels out quality: once a school is more than 1.5km away, its quality no longer matters to parents. The model also shed light on the impact of disability on enrolment and highlighted several behavioral patterns e.g., child labor as a coping strategy that follows dropout, and the hurdles for households lacking documentation like B-Forms.
● Pakistan’s substantial data is underutilised in decision making. This can be improved using predictive models, bringing together data sets, and using innovations in data governance. This will help the government shift from reactive reporting to precision governance of Pakistan’s education emergency.
How did we do it?
The study utilizes a rigorous mixed-methods design that integrates quantitative, qualitative, and geospatial data into a cohesive modeling framework. The research was conducted in two primary phases to ensure technical accuracy and policy relevance:
● Phase 1: Conceptualization and Development: The team synthesized secondary data from national surveys: Pakistan Social and Living Standards Measurement Survey (PSLM) and Multiple Indicator Cluster Survey (MICS) to identify macro-level predictors of enrollment. This was augmented by a primary survey of 800 households in the Lahore division and qualitative Educational Journey Mapping to capture micro-level behavioral insights, which were then used to train a simulation model predicting probability of enrollment for a family with given demographic, access, and school-level factors.
● Phase 2: Calibration and Engagement: The project leveraged Explainable AI (XAI) techniques that unpack how different factors interact to influence school enrollment, allowing non-technical officials to see the logic behind every prediction.
● Co-Design Process: Through a collaborative approach involving the Pakistan Institute of Education (PIE) and provincial departments, the model was calibrated to administrative realities to facilitate institutional adoption.
Limitations:
● Geographic Scope: While this conceptualization study provides high-fidelity insights, primary data collection was concentrated in the Lahore Division, which may limit immediate generalizability to vastly different provincial terrains.
● Implementation Assumptions: Successful adoption assumes existing functional planning units and data interoperability capacity, which currently varies across provinces; uptake planning in regions like KP and Balochistan must account for these varying baselines.
Crucially, the methodology demonstrated that existing national data pipelines can be enhanced with targeted behavioral questions to support robust simulations without requiring extensive additional investment.