Baby Steps: A Machine Learning Approach to Optimizing Pronatalist Policies in South Korea

Abstract

The Republic of Korea, otherwise known as South Korea, currently has one of the lowest recorded birth rates in the world. In addition to the troubling rate of projected population decline in the coming decades, this critically low fertility poses immense risks to national security and economic stability. Pronatalist policy interventions at the national level have been implemented since 2005, yet the birth rate has not seen the rebound necessary to avoid a population collapse. This research begins by discussing the cultural and historical factors that have contributed to this decline, arguing that the state's overreliance on pronatalist policies centered on financial incentives was, and to a lesser extent still is, fundamentally misaligned with the most significant barriers to fertility. This study brings a gender-inclusive perspective to the study of fertility decline in the Seoul Metropolitan Area of South Korea by utilizing data from an original survey of male and female residents of childbearing age. Analyzing this survey using Random Forest and XGBoost classification models, I identify the primary social determinants of fertility intentions. My findings indicate that monthly household income and concerns about the financial cost of childrearing are not robust determinants of fertility intentions, and that factors such as gender or perceptions of marriage are substantially stronger predictors. Based on these insights, I offer adjustments to improve the efficacy of current pronatalist policies in South Korea.