Panel Data Regression with Feasible Generalized Least Squares for Analyzing School Participation Rate Among Adolescents in Indonesia

Abstract

School participation among adolescents aged 16-18 years varies across provinces in Indonesia, reflecting disparities in educational access. This study examines the associations of the Unemployment Rate (UR), Average Hourly Earnings of Employees (AHE), Mean Years of Schooling (MYS), and Percentage of Poor Population (PPP) with the School Participation Rate (SPR) with a panel dataset comprising 34 provinces observed from 2018 to 2023. Selection between the panel specifications was based on the Chow and Hausman tests, with the results favoring the Fixed Effects Model (FEM). Diagnostic examination identified heteroskedasticity, serial correlation, and cross-sectional dependence. The model was therefore re-estimated using the modified Feasible Generalized Least Squares (FGLS) estimator proposed by Bai, Choi, and Liao, which accounts for these error dependencies. Results show that AHE, MYS, and PPP were positively associated with SPR, with significant relationships at the 5% level, whereas UR was positively but not significantly associated with SPR. These findings indicate that household economic capacity, educational attainment, and poverty levels are significantly associated with differences in school participation across provinces, while provincial unemployment conditions do not show a statistically significant association. The findings provide evidence on socioeconomic factors associated with school participation among adolescents aged 16-18 years in Indonesia.