Exploring Cognitive and Social-Emotional Impact Patterns in a Randomized Controlled Trial of Math Tutoring

• In Event: Heterogeneity in Educational Intervention Fadeout and Long-run Mechanisms: Three Analyses of Block-Level Impact Patterns

Abstract

Objectives and Theoretical Framework
High-dosage tutoring is among the most effective interventions for improving student mathematics achievement (Kraft, 2015), yet the persistence of these effects varies (Authors, 2023). In some cases, interventions yield long-term outcomes despite short-term fadeout (Chetty et al., 2011). The Large Interconnected Network Theory (LINT; Bailey et al., 2024) proposes that interventions generate “ripple effects” across interconnected developmental domains, producing long-term effects.
To examine this hypothesis, we focus on variation at the tutor level within a multi-site randomized controlled trial (RCT). Each tutor delivers the same curriculum but may differ in effectiveness (Hashim et al., 2025). These differences represent natural variation in how the intervention plays out and provide a powerful opportunity to test LINT’s predictions about cross-domain transfer. We treat each tutor as a unique intervention and estimate tutor-specific impacts. Using meta-analytic techniques, we test whether tutors who produce stronger posttest effects in cognitive outcomes also generate larger follow-up social-emotional impacts or vice versa, thus bridging methodological innovation with a theoretically grounded model of developmental change.
Methods
Data came from a Grade 3 math intervention, with Grade 4 follow-up (Author, 2022). 449 students with mathematics difficulty were randomized to a word-problem intervention or a business-as-usual control. The pretest cognitive composite included math, executive function, and reading. Posttest and follow-up composites included math subtests. Pretest social-emotional functioning was assessed with the Mathematics Anxiety Rating Scale, and posttest/follow-up outcomes with a Math Feelings Scale.
We constructed tutor blocks by grouping treatment students with the same tutor and control students from the same classroom(s). Tutor-level treatment effects were estimated using linear models with tutor-by-treatment interactions, controlling for pretest scores. Standard errors were clustered by tutor and classroom. Balance checks and falsification tests suggest potential violation of the exogeneity assumption, so we controlled for pretest differences and treated estimates as quasi-experimental. We then examined whether tutors with larger posttest effects in one domain produced larger follow-up effects in another using regression models. A key assumption is that follow-up impacts are fully mediated by posttest impacts. While we cannot test for unmeasured pathways, we mitigate this concern by controlling for pretest differences.
Results
We found substantial variation in impacts across tutors (Figure 1 and Table 1). Average posttest cognitive effects were 1.06 (p < .001), τ = 054, Q(23) = 736.55, p < .001. Follow-up cognitive effects were 0.20 (p = .07), τ = 0.54, Q(23) = 914.82, p < .001. Posttest social-emotional effects averaged 0.20 (p = .03), τ = 0.43, Q(23) = 547.36, p < .001, with follow-up effects at 0.20 (p = .18), τ= 0.59, Q(23) = 82.07, p < .001.
Evidence favored cognitive-driven over social-emotional-driven transfer (Table 2). Cognitive posttest effects predicted social-emotional follow-up effects (b = .73, p = .04), but not vice versa (b = -.06, p = .80).
Significance
Findings suggest that cognitive improvements may support sustained emotional growth, consistent with LINT’s framework of developmental transfer. This study contributes to understanding how educational interventions can have lasting effects beyond their immediate targets.

Authors

• Siling Guo, University of California - Irvine

• Sarah R. Powell, University of Texas at Austin

• Emma Hart, Boston College

• Caroline Botvin, Columbia University

• Xin Lin, University of Macau

• Tyler Watts, Teachers College, Columbia University

• Drew Bailey, University of California - Irvine