Title: An improved approach for estimating large losses using the g-and-h distribution
Authors: Marco Bee - University of Trento (Italy) [presenting]
Luca Trapin - University of Bologna (Italy)
Julien Hambuckers - University of Liege - HEC Liège (Belgium)
Abstract: The g-and-h distribution finds applications in modeling highly skewed and fat-tailed data, like extreme losses in the banking and insurance sector. Given the lack of a closed-form density, two estimation methods are introduced: a maximum likelihood technique based on a numerical approximation of the likelihood function, and an indirect inference approach with a bootstrap weighting scheme. A realistic simulation study suggests that indirect inference is computationally more efficient and provides better estimates in case of extreme features of the data, whereas maximum likelihood is preferable in terms of root-mean-squared-error when the data are less skewed and heavy-tailed. Empirical illustrations on insurance and operational losses illustrate these findings.