Analysis of Influential Factors of Building Damage from Natural Disasters Based on Compound Poisson Model

The purpose of this study is to construct statistical models for building damage from natural disasters and analyze the regional pattern of the damage. Both a compound Poisson model (CPM) and a zero-inflated compound Poisson model (ZICPM) have been developed to explain the amount for building damage from natural disasters, while adopting disaster region, disaster duration, and disaster type as explanatory variables. After estimating the above two regression models with the mentioned variables, the ZICPM is verified to be more appropriate than the CPM for building damage data from 1994 to 2015. Results show that amount and probability of building damage for provincial area are much greater than those for the Seoul Metropolitan Area. Additionally, damage from typhoon and heavy rain is statistically proved to be greater than that from strong winds. Accordingly, detailed measures and strategies based on disaster region and type should be prepared to mitigate building damage. Keywords: Natural Disaster, Damage Amount, Compound Poisson, Zero-inflated Compound Poisson