An Analysis of Private Tutoring Participation and Its Predictors Among One-Year-Old Infants in South Korea

This study investigated the current status of private tutoring participation among one-year-old infants in South Korea and identified key predictors for participation and associated expenditures using machine learning techniques. Data from 2,555 infants from the second wave (2023) of the Korea Early Childhood Education and Care Panel Study (KECECPS) were analyzed using XGBoost and SHapley Additive exPlanations (SHAP), an explainable AI method. The analysis revealed that the private tutoring participation rate for one-year-olds was 14.3%, with participating infants spending an average of 115,700 KRW per month. According to the machine learning results, satisfaction with childcare support policies emerged as the most significant factor predicting both participation and cost. Notably, while higher satisfaction with in-kind support, such as vouchers, was associated with lower participation and costs, higher satisfaction with cash-based support (e.g., parental allowances and utility discounts) tended to increase the likelihood and expenditure of private tutoring. Furthermore, although positive perceptions of early childhood institutions (daycare centers and kindergartens) suppressed tutoring demand, infants' cognitive and language development levels, household income, and housing size were positively correlated with such demand. Based on these findings, the study suggests that policy designs should prioritize improving the quality of childcare services and expanding in-kind support rather than focusing solely on cash assistance to address the issue of infant private tutoring. Finally, the study discusses the necessity and potential of utilizing machine learning for data-driven policymaking in the fields of early childhood education and care.