Academia.eduAcademia.edu

Modern actuarial risk theory

2001, Journal of Natural History - J NATUR HIST

Rob Kaas • Marc Goovaerts Jan Dhaene • Michel Denuit Modern Actuarial Risk Theory Using R Second Edition fyj Springer Contents There are 1011 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers — Richard Feynman (1918-1988) 1 Utility theory and insurance 1.1 Introduction 1.2 The expected utility model 1.3 Classes of utility functions 1.4 Stop-loss reinsurance 1.5 Exercises 1 1 2 5 8 13 2 The individual risk model 2.1 Introduction .1 2.2 Mixed distributions and 2.3 Convolution 2.4 Transforms 2.5 Approximations 2.5.1 Normal approximation 2.5.2 Translated gamma approximation 2.5.3 NP approximation 2.6 Application: optimal reinsurance 2.7 Exercises 17 17 18 25 28 30 30 32 33 35 36 3 risks Collective risk models 3.1 Introduction 3.2 Compound distributions 3.2.1 Convolution formula for a compound cdf 3.3 Distributions for the number of claims 3.4 Properties of compound Poisson distributions 3.5 Panjer's recursion 3.6 Compound distributions and the Fast Fourier Transform 3.7 Approximations for compound distributions 3.8 Individual and collective risk model 3.9 Loss distributions: properties, estimation, sampling 3.9.1 Techniques to generate pseudo-random samples 3.9.2 Techniques to compute ML-estimates 41 41 42 44 45 47 49 54 57 59 61 62 63 Contents 3.9.3 Poisson claim number distribution 3.9.4 Negative binomial claim number distribution 3.9.5 Gamma claim severity distributions 3.9.6 Inverse Gaussian claim severity distributions 3.9.7 Mixtures/combinations of exponential distributions 3.9.8 Lognormal claim severities 3.9.9 Pareto claim severities 3.10 Stop-loss insurance and approximations 3.10.1 Comparing stop-loss premiums in case of unequal variances 3.11 Exercises 63 64 66 67 69 71 72 73 76 78 Ruin theory 4.1 Introduction 4.2 The classical ruin process 4.3 Some simple results on ruin probabilities 4.4 Ruin probability and capital at ruin 4.5 Discrete time model 4.6 Reinsurance and ruin probabilities 4.7 Beekman's convolution formula 4.8 Explicit expressions for ruin probabilities 4.9 Approximation of ruin probabilities 4.10 Exercises 87 87 89 91 95 98 99 101 106 108 Ill Premium principles and Risk measures 5.1 Introduction 5.2 Premium calculation from top-down 5.3 Various premium principles and their properties 5.3.1 Properties of premium principles 5.4 Characterizations of premium principles 5.5 Premium reduction by coinsurance 5.6 Value-at-Risk and related risk measures 5.7 Exercises 115 115 116 119 120 122 125 126 133 Bonus-malus systems 6.1 Introduction 6.2 A generic bonus-malus system 6.3 Markov analysis 6.3.1 Loimaranta efficiency 6.4 Finding steady state premiums and Loimaranta efficiency 6.5 Exercises 135 135 136 138 141 142 146 Ordering of risks 7.1 Introduction 7.2 Larger 7.3 More dangerous 7.3.1 Thicker-tailed 149 149 152 154 154 risks risks risks Contents xvii 7.3.2 Stop-loss order j ^ 159 7.3.3 Exponential order 160 7.3.4 Properties of stop-loss order 160 7.4 Applications 164 7.4.1 Individual versus collective model 164 7.4.2 Ruin probabilities and adjustment coefficients 164 7.4.3 Order in two-parameter families of distributions 166 7.4.4 Optimal reinsurance 168 7.4.5 Premiums principles respecting order 169 7.4.6 Mixtures of Poisson distributions 169 7.4.7 Spreading of risks 170 7.4.8 Transforming several identical risks 170 7.5 Incomplete information 171 7.6 Comonotonic random variables 176 7.7 Stochastic bounds on sums of dependent risks 183 7.7.1 Sharper upper and lower bounds derived from a surrogate .. 183 7.7.2 Simulating stochastic bounds for sums of lognormal risks .. 186 7.8 More related joint distributions; copulas 190 7.8.1 More related distributions; association measures 190 7.8.2 Copulas 194 7.9 Exercises 196 8 Credibility theory 1 8.1 Introduction ...'. 8.2 The balanced Buhlmann model 8.3 More general credibility models 8.4 The Biihlmann-Straub model 8.4.1 Parameter estimation in the Biihlmann-Straub model 8.5 Negative binomial model for the number of car insurance claims . . . 8.6 Exercises 9 Generalized linear models 9.1 Introduction 9.2 Generalized Linear Models 9.3 Some traditional estimation procedures and GLMs 9.4 Deviance and scaled deviance 9.5 Case study I: Analyzing a simple automobile portfolio 9.6 Case study II: Analyzing a bonus-malus system using GLM 9.6.1 GLM analysis for the total claims per policy 9.7 Exercises 10 IBNR techniques 10.1 Introduction 10.2 Two time-honored IBNR methods 10.2.1 Chain ladder 203 203 204 211 214 217 222 227 231 231 234 237 245 248 252 257 262 265 265 268 268 xviii Contents V 10.2.2 Bornhuetter-Fergujspn 10.3 A GLM that encompasses various IBNR methods 10.3.1 Chain ladder method as a GLM 10.3.2 Arithmetic and geometric separation methods 10.3.3 De Vijlder's least squares method 10.4 Illustration of some IBNR methods 10.4.1 Modeling the claim numbers in Table 10.1 10.4.2 Modeling claim sizes 10.5 Solving IBNR problems by R 10.6 Variability of the IBNR estimate 10.6.1 Bootstrapping 10.6.2 Analytical estimate of the prediction error 10.7 An IBNR-problem with known exposures 10.8 Exercises 270 271 272 273 274 276 277 279 281 283 285 288 290 292 11 More on GLMs 11.1 Introduction 11.2 Linear Models and Generalized Linear Models 11.3 The Exponential Dispersion Family 11.4 Fitting criteria 11.4.1 Residuals 11.4.2 Quasi-likelihood and quasi-deviance 11.4.3 Ex'tended quasi-likelihood 11.5 The canonical link 11.6 The IRLS algorithm of Nelder and Wedderburn 11.6.1 Theoretical description 11.6.2 Step-by-step implementation 11.7 Tweedie's Compound Poisson-gamma distributions 11.7.1 Application to an IBNR problem 11.8 Exercises 297 297 297 300 305 305 306 308 310 312 313 315 317 318 320 The 'R' in Modern ART A.I A short introduction to R A.2 Analyzing a stock portfolio using R A.3 Generating a pseudo-random insurance portfolio 325 325 332 338 Hints for the exercises 341 Notes and references 357 Tables 367 Index 371