Examining the non-linear relationship between the residential environment and birth rate in Seoul, South Korea

Abstract

South Korea recorded the world’s lowest total fertility rate of 0.72 in 2023, highlighting an urgent socio-economic challenge with far-reaching implications such as labor shortages, population aging, and economic stagnation. While numerous studies have examined how the physical environment influences fertility, most have relied on linear assumptions that limit real-world applicability. This study adopts a non-linear, machine-learning-based approach to investigate how residential environmental factors—measured through the “N-minute city” framework—affect fertility rates in Seoul. Using XGBoost and SHAP analysis, the study evaluates the influence of spatial proximity to key urban functions—including childcare, education, healthcare, safety, and transportation—on birth counts, based on fine-grained 250m×250m grid-level data for 2023. The results reveal several threshold effects: fertility rates were positively associated with commuting times between 48–66 minutes, access to public daycare within 238 meters, clinics within 176 meters, academies within 197 meters, and elementary schools within 550 meters. Conversely, proximity to high-priced housing areas, subway stations, and parks—typically within 500 meters—was negatively correlated with fertility. The analysis also finds that walkable access to grocery stores and restaurants supports higher fertility rates, despite the widespread use of delivery services. By identifying non-linear thresholds and spatial dependencies in fertility-related behaviors, this study provides actionable insights for urban planning and residential policy. It suggests that policies aiming to promote childbearing should prioritize enhancing physical accessibility to essential services within walkable distances and consider both spatial equity and behavioral thresholds in design strategies.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 4 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.