An Enhanced Value-Added Model Using Seemingly Unrelated Regression: Social-Emotional Learning, Teacher Effectiveness, and Students' Academic Outcomes in Kwara State, Nigeria

Abstract

Traditional Value-Added Models (VAMs) used to evaluate teacher and school effectiveness rely almost exclusively on cognitive, test-score-based outcomes, thereby overlooking the socio-emotional and non- cognitive dimensions of student development that strongly predict long-term success. In addition, when several related outcomes are modelled separately using Ordinary Least Squares (OLS), the contemporaneous correlation among their disturbances is ignored, producing statistically inefficient estimates. This st udy develops an Enhanced Value -Added Model (E -VAM) within a Seemingly Unrelated Regression Equations (SURE) framework to jointly estimate the contributions of teacher effectiveness, school factors, parental involvement, and demographic characteristics to s tudents' academic performance, non -cognitive skills, and social -emotional learning. A quantitative research design was adopted, combining a Monte Carlo simulation with an empirical application to survey data obtained from 500 students drawn from public and private secondary schools across the senatorial districts of Kwara State, Nigeria, using stratified random sampling. Residual -correlation analysis and the Breusch-Pagan test for independence were used to justify joint estimation, after which the system of equations was estimated by Feasible Generalized Least Squares (FGLS). Both the simulation and the empirical results showed statistically significant positive effects of teacher effectiveness, school factors, parental involvement, and discipline on academi c performance, non -cognitive skills, and social - emotional learning, with the SURE estimates outperforming independent OLS estimates in efficiency, as evidenced by non-zero residual covariances (0.048-0.054) and a highly significant Breusch -Pagan statistic (LM = 335.44, p < 0.05). Teacher effectiveness emerged as the strongest and most consistent predictor across all three outcome domains. The findings demonstrate that integrating cognitive and non-cognitive outcomes within a jointly estimated E-VAM framework yields a more efficient, holistic, and policy-relevant assessment of teacher and school contributions than conventional single -outcome VAMs, with implications for teacher evaluation, professional development, and education policy in Nigeria and similar contexts.