TITLE:
Estimating Heterogeneous Treatment Effects of Early Childhood Interventions on Fifth-Grade Math Achievement: A Machine Learning-Augmented Causal Analysis of ECLS-K:2011
AUTHORS:
Xingtian Si
KEYWORDS:
Causal Inference, Machine Learning, Targeted Maximum Likelihood Estimation, Causal Random Forests, Special Education Program, Heterogeneity, Non-Randomized Data
JOURNAL NAME:
International Journal of Intelligence Science,
Vol.15 No.4,
September
3,
2025
ABSTRACT: The study focuses on identifying and distinguishing whether there are differences between those students receiving special education services later compared to their general-education peers entering kindergarten, in terms of early-childhood characteristics, and examines the long-term implications for fifth-grade mathematics achievement. Drawing on the nationally representative Early Childhood Longitudinal Study, Kindergarten Class of 2010-11 (ECLS-K:2011), we first quantify baseline differences across domains—demographic, academic, family, school composition, health, and behavioral by using standardized mean differences. Next, we apply approaches of machine learning to augment causal inference methods, including Bayesian Additive Regression Trees (BART), Targeted Maximum Likelihood Estimation (TMLE), and Causal Random Forests (CRF), to estimate heterogeneous treatment effects of early supports and interventions. Our analysis reveals that early math and reading proficiency, self-regulation skills, and socioeconomic indicators are among the strongest predictors of special-education placement. We demonstrate that CRF, in particular, excels at uncovering complex, nonlinear relationships and subgroup-specific impacts, enabling precise estimation of how different combinations of behavioral and family-centered strategies influence high-risk children’s outcomes. By integrating these advanced methods with the rich structure of ECLS-K:2011, we extend beyond descriptive profiling to evaluate the effectiveness of policy levers, such as early screening protocols, targeted classroom accommodations, and family outreach programs, on narrowing achievement gaps. The findings can provide educators, policymakers, and practitioners with specific information about efficient utilization of allocated funds and design suitable interventions to improve educational outcomes for all children.