Abstract

Based on the process of integration of diverse data sources, such as geospatial information, and administrative records it is intended to develop an analysis of home-to-school distances of students, and to increase the availability of data to measure strategic indicators identified
by the education sector that are aligned with the Sustainable Development Goals. With the support of the Data4Now Initiative, a cross-cutting training on geospatial analysis (cross-cutting) was delivered focused on strengthening knowledge and practical capacities among DANE staff members to process, classify and use satellite imagery/remote sensor data in statistical analysis, covering a range of relevant techniques such as artificial intelligence methods / neural networks, as well as to analyze and use integrated statistical-geospatial datasets for the production of statistical indicators related to poverty and education. Using different distance metrics, the exercise focused on urban areas and the sector to which the educational center belongs (official and nonofficial). It was observed that the distances for students enrolled in official educational
centers are shorter than the distances for students in non-official schools. Results point out that there is no strong relation between dropout rate and home-school distance: in 2018, the dropout rate increased from 1.7% to 3.1% between distance quintiles 1 and 5 for official educational centers, while for nonofficial educational centers this rate decreased by 0.1 percentage points (p.p.) between distance quintile 1 and 5. In 2020 there was significant growth in dropout rates for non-official schools (0.6 p.p. above the figure for 2019) despite the distance quintile 5 (average meters = 14.457,1) showing a dropout rate of 2,3% for non-official centers and 2,1% for official ones while distance quintile 1 shows dropout rates of 1,4% and 3,2%, accordingly. The joint work between DANE, the Ministry of Education, and Data4Now also shows a better understanding of the school dropout phenomenon based on the administrative, and geospatial data, using alternative analysis and processing methods.