Identifying Macro Determinants of Natural Disaster: Applying Machine Learning Approach

This study analyzes the macroeconomic determinants of human and economic losses caused by natural disasters using machine learning techniques, particularly the double lasso regression framework. By focusing on floods, storms, and earthquakes, we examine the complex relationships between economic development, geographical characteristics, and the severity of disaster impacts. Our findings indicate a U-shaped relationship between GDP per capita and economic losses, where initial GDP growth reduces losses until a threshold is reached, beyond which further economic development increases the financial damages. Human losses, on the other hand, show an inverse U-shaped relationship with GDP in the case of floods, highlighting that while higher GDP initially leads to more fatalities, continued economic growth eventually reduces mortality rates. Moreover, we find that coastal countries are more vulnerable to both human and economic losses compared to island and landlocked countries, though they demonstrate a greater capacity to leverage GDP growth in reducing economic losses. These results underscore the importance of tailored disaster risk management strategies that consider both economic development and geographic factors to mitigate the adverse impacts of natural disasters. JEL Codes: Q54, O44, C55