View in EDS

Need Not Be a Surprise: Early-Warning Systems for Chronic Absenteeism

Saved in:
Bibliographic Details
Title: Need Not Be a Surprise: Early-Warning Systems for Chronic Absenteeism
Language: English
Authors: Nat Malkus, Sam Hollon, American Enterprise Institute (AEI)
Source: American Enterprise Institute. 2025.
Availability: American Enterprise Institute. 1150 Seventeenth Street NW, Washington, DC 20036. Tel: 202-862-5800; Fax: 202-862-7177; Web site: http://www.aei.org
Peer Reviewed: N
Page Count: 19
Publication Date: 2025
Document Type: Reports - Descriptive
Education Level: Elementary Secondary Education
Descriptors: Attendance, Prevention, Elementary Secondary Education, Models, Prediction, Accuracy
Geographic Terms: Rhode Island, Indiana
Abstract: This report shows that districts can use data they already routinely collect to predict which students will become chronically absent. Existing work to predict absenteeism in advance either is academic and too challenging for districts to use themselves or uses proprietary systems that are not publicly accessible. Accordingly, in this report, the authors approach the problem of predicting chronic absenteeism from a district leader's perspective. The authors explore and demystify the logic, trade-offs, and potential of early-warning systems for targeting attendance interventions and show how districts can predict which students will be absent using only the data that they "already" routinely collect. More specifically, the authors use data from Rhode Island and Indiana to train machine learning models to predict whether students will be chronically absent, how many total absences students will accrue, and how many more or fewer absences students will accrue compared with the previous year. The models in this report show that although absenteeism became harder to predict during the COVID-19 pandemic, it has since become more predictable again, even though the overall rate of chronic absenteeism remains high.
Abstractor: ERIC
Entry Date: 2026
Accession Number: ED677962
Database: ERIC
Description
Abstract:This report shows that districts can use data they already routinely collect to predict which students will become chronically absent. Existing work to predict absenteeism in advance either is academic and too challenging for districts to use themselves or uses proprietary systems that are not publicly accessible. Accordingly, in this report, the authors approach the problem of predicting chronic absenteeism from a district leader's perspective. The authors explore and demystify the logic, trade-offs, and potential of early-warning systems for targeting attendance interventions and show how districts can predict which students will be absent using only the data that they "already" routinely collect. More specifically, the authors use data from Rhode Island and Indiana to train machine learning models to predict whether students will be chronically absent, how many total absences students will accrue, and how many more or fewer absences students will accrue compared with the previous year. The models in this report show that although absenteeism became harder to predict during the COVID-19 pandemic, it has since become more predictable again, even though the overall rate of chronic absenteeism remains high.