Data-Driven Analysis for Improving Educational Policies

Data analytics can support evidence-based decision-making in public policies by enabling the identification of patterns, forecasting needs, and prioritizing actions. Consequently, data-driven analysis can aid policymakers in redesigning and enhancing educational policies, such as textbook distribution for public schools. However, there is no consensus on a structured approach for descriptive analytics in this context. This study presents a descriptive approach to textual data analysis aimed at improving policy implementation and monitoring, with a focus on effort, productivity, and quality of reviews produced by textbook evaluators. Through a case study, we apply natural language processing techniques to analyze thousands of answers to rubrics during the pedagogical evaluation of a public call for literary works under the Brazilian textbook program (Programa Nacional do Livro e do Material Didático - PNLD). The PNLD is one of the most extensive textbook policies, impacting millions of students. Our findings shed light on challenges related to the effort involved and the quality of written reports in the pedagogical evaluation process. Analyzing reports, which reflect some desired and undesired behaviors of evaluators, can offer policymakers insights for making informed decisions and improving textbook programs worldwide. Our descriptive approach to textual data analysis leverages insights to enhance transparency, inform improvements, and guide policy implementation through real-time monitoring.