The aim of this master thesis is to derive methods that assesses the impact the deductiblehas on the risk premium of an insurance contract. The additive structure of a deductiblenecessitates approaches beyond treating it as a regular covariate in a generalized linearmodel for predicting the risk premium. Using simulated data, three methods areimplemented to estimate a parameter, denoted as βd ∈ [0,1], which quantifies the proportionof the risk premium that remains after imposing a deductible d on the insurance contract.The three implemented methods involve: 1) Regularized generalized linear models, 2)Utilizing the cumulative density function of the insurance contract, and 3) Estimating adiscrete probability distribution with K-means clustering and a classifier. To compare theperformance of these methods, they are tested against each other through a competition.The results reveal that the method employing the fitting of a discrete distribution yieldedthe best performance. / I den här uppsatsen härleds tre olika metoder där syftet är att skapa en oberoendevariabel βd ∈ [0,1] som beskriver hur stor del av ett försäkringskontrakts riskpremiesom kvarstår, efter att en självrisk d ålagts på försäkringskontraktet. Självrisken kaninte användas som vilken oberoende variabel som helst på grund av den additivastruktur som självrisken innehar. De tre olika metoderna som har implementerats skiljersig åt genom att 1) använder sig av upprepade regulariserad GLM-modeller för olikasjälvrisknivåer, 2) nyttjar försäkringskontraktets fördelningsfunktion, och 3) skattar endiskret sannolikhetsfördelning med hjälp av klustring och en klassificerare. De tre olikametoderna testades sedan mot varandra i ett spel. Metod tre presterade bäst i spelet, dåden hade lägst procentuell avvikelse från den sanna riskpremien.

Title: Traditional and modern approaches to pricing in nonlife insurance Abstract: This thesis deals with the theory and implementation of generalized linear models in the area of pricing of non-life insurance and subsequent optimalization of rates. Using the generalized linear models it is possible to estimate expected value and variance of compound distribution of total claims made according to insurance policy during definite time period. The next step is to build an optimalization model and describe several methods how to determine rates that lead to optimal distribution of safety margins within insurance policies in particular risk groups. Represented approaches how to calculate insurance premiums are numerically illustrated on simulated data in concluding parts of the thesis.