CMStatistics 2017: Start Registration
View Submission - CFE
Title: Spatio-temporal modeling of yield-weather dependence Authors:  Guillaume Bagnarosa - Rennes School of Business (France) [presenting]
Suikai Gao - Rennes School of Business (France)
Gareth Peters - University College London (United Kingdom)
Matthew Ames - Institute of Statistical Mathematics (Japan)
Tomoko Matsui - The Institute of Statistical Mathematics (Japan)
Abstract: Weather risk represents today one of the main challenge for the farming businesses and, consequently, for the insurance companies proposing them hedging solutions. However, the lack of data at the farm level, and the ubiquitous moral hazard associated to crop insurance make the pricing of insurance policies quite perilous and requires thus new approaches to solve both problems. We propose to first combine a farm acreage weighted Gaussian Process with a SARIMA dynamic to cope with the weather spatio-temporal joint distribution which is then used to model the distribution at the county level of the crop yields. We apply our new methodology to a large database of Romanian farms and demonstrate that our approach outperforms other methods commonly used in the industry for determining insurance premium rate.