Harvesting Risk: Applying GLMs to Agricultural Insurance Pricing

Agricultural insurance plays a fundamental role in mitigating the financial risks faced by farmers due to adverse climatic events. One of the key elements of any insurance product is the estimation of the pure premium, which reflects the expected cost of claims. This study aims to evaluate and improve the pure premium estimation process currently used by Atlas MGA, a portuguese insurance company specialized in agriculture insurance. At present, the company adopts an approach to premium pricing that is based on historical data (from IFAP) on a key ratio. In this work, we propose the application of Generalized Linear Models (GLMs) as a more flexible and more robust alternative. The pure premium is calculated through two components: the expected frequency of claims and the expected severity each claims. The models were implemented using real agricultural insurance data provided by IFAP and the company, and the results were compared with the method currently used by the company. The findings show that GLMs produce more accurate and consistent premium estimates, and suggest that the traditional method may systematically underestimate the pure premium in certain cases. This study highlights the potential of GLMs to enhance actuarial pricing models in agricultural insurance and provides a foundation for future improvements, including the incorporation of additional variables and the use of more advanced predictive techniques, such as neural networks and machine learning.