Ruin Probabilities: Smoothness, Bounds, Supermartingale Approach

Ruin Probabilities

Smoothness, Bounds, Supermartingale Approach

Yuliya Mishura

Olena Ragulina

Series Editor

Nikolaos Limnios

Table of Contents

Cover image

Title page

Copyright

Preface

Part 1: Smoothness of the Survival Probabilities with Applications

1: Classical Results on the Ruin Probabilities

Abstract

1.1 Classical risk model

1.2 Risk model with stochastic premiums

2: Classical Risk Model with Investments in a Risk-Free Asset

Abstract

2.1 Description of the model

2.2 Continuity and differentiability of the infinite-horizon survival probability

2.3 Continuity of the finite-horizon survival probability and existence of its partial derivatives

2.4 Bibliographical notes

3: Risk Model with Stochastic Premiums Investments in a Risk-Free Asset

Abstract

3.1 Description of the model

3.2 Continuity and differentiability of the infinite-horizon survival probability

3.3 Continuity of the finite-horizon survival probability and existence of its partial derivatives

4: Classical Risk Model with a Franchise and a Liability Limit

Abstract

4.1 Introduction

4.2 Survival probability in the classical risk model with a franchise

4.3 Survival probability in the classical risk model with a liability limit

4.4 Survival probability in the classical risk model with both a franchise and a liability limit

5: Optimal Control by the Franchise and Deductible Amounts in the Classical Risk Model

Abstract

5.1 Introduction

5.2 Optimal control by the franchise amount

5.3 Optimal control by the deductible amount

5.4 Bibliographical notes

6: Risk Models with Investments in Risk-Free and Risky Assets

Abstract

6.1 Description of the models

6.2 Classical risk model with investments in risk-free and risky assets

6.3 Risk model with stochastic premiums and investments in risk-free and risky assets

6.4 Accuracy and reliability of uniform approximations of the survival probabilities by their statistical estimates

6.5 Bibliographical notes

Part 2: Supermartingale Approach to the Estimation of Ruin Probabilities

7: Risk Model with Variable Premium Intensity and Investments in One Risky Asset

Abstract

7.1 Description of the model

7.2 Auxiliary results

7.3 Existence and uniqueness theorem

7.4 Supermartingale property for the exponential process

7.5 Upper exponential bound for the ruin probability

7.6 Bibliographical notes

8: Risk Model with Variable Premium Intensity and Investments in One Risky Asset up to the Stopping Time of Investment Activity

Abstract

8.1 Description of the model

8.2 Existence and uniqueness theorem

8.3 Redefinition of the ruin time

8.4 Supermartingale property for the exponential process

8.5 Upper exponential bound for the ruin probability

8.6 Exponentially distributed claim sizes

8.7 Modification of the model

9: Risk Model with Variable Premium Intensity and Investments in One Risk-Free and a Few Risky Assets

Abstract

9.1 Description of the model

9.2 Existence and uniqueness theorem

9.3 Supermartingale property for the exponential process

9.4 Upper exponential bound for the ruin probability

9.5 Case of one risky asset

9.6 Examples

Appendix: Mathematical Background

A.1 Differentiability of integrals

A.2 Hoeffding’s inequality

A.3 Some results on one-dimensional homogeneous stochastic differential equations

A.4 Itô’s formula for semimartingales

Bibliography

Abbreviations and Notation

Index

Copyright

First published 2016 in Great Britain and the United States by ISTE Press Ltd and Elsevier Ltd

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced, stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms and licenses issued by the CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the undermentioned address:

ISTE Press Ltd

27–37 St George’s Road

London SW19 4EU

UK

www.iste.co.uk

Elsevier Ltd

The Boulevard, Langford Lane

Kidlington, Oxford, OX5 1GB

UK

www.elsevier.com

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

For information on all our publications visit our website at http://store.elsevier.com/

© ISTE Press Ltd 2016

The rights of Yuliya Mishura and Olena Ragulina to be identified as the authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library

Library of Congress Cataloging in Publication Data

A catalog record for this book is available from the Library of Congress

ISBN 978–1-78548–218-2

Printed and bound in the UK and US

Preface

Yuliya Mishura; Olena Ragulina, Kyiv

Risk theory is traditionally considered as a branch of insurance mathematics and its active area of research. Since Filip Lundberg [LUN 03] introduced the collective risk model in 1903, the estimation of ruin probabilities has been one of the central directions for investigations in risk theory. The infinite- and finite-horizon ruin probabilities are the probabilities that the surplus of an insurance company becomes negative on infinite and finite time intervals, respectively. The infinite-horizon ruin probability is usually considered as a function of the initial surplus, and the finite-horizon ruin probability is assumed to be a function of the initial surplus and time interval. Instead of ruin probabilities, one sometimes deals with the corresponding survival probabilities, which are the probabilities that the surplus of the insurance company does not become negative. So, the survival of the insurance company at least implies that it is able to reimburse claims and meet its obligations. Although other characteristics have also been proposed and investigated (e.g. distribution of the ruin time, surplus prior to ruin, deficit at ruin and Gerber-Shiu expected discounted penalty functions), the ruin or survival probabilities are still the main objects of interest in risk theory.

In this book, we will consider continuous-time risk models. Due to the obvious relationship between ruin and survival probabilities, from the mathematical viewpoint it makes no difference which of these two characteristics is studied. Since sometimes it is more convenient to deal with the ruin probabilities and sometimes with the survival probabilities, in what follows we consider both of them. To be more precise, we investigate both survival probabilities in Chapters 2–6 and ruin probabilities in Chapters 7–9.

This monograph mainly provides the results obtained by the authors recently in this field. It is a revised and extended version of the material published in [AND 07a, AND 07b, BON 12, MIS 15, RAG 10, RAG 11, RAG 12, RAG 14, RAG 15].

The book deals with several aspects of risk theory. The first of them is the smoothness of the survival probabilities. Particularly, we investigate the continuity and differentiability of the infinite- and finite-horizon survival probabilities in detail for different risk models. The initial motivation for this investigation is the reasonable derivation of integro-differential equations for these functions. These equations turn out to be very useful for calculation of the survival probabilities. In some cases, one can find their closed-form solutions. In other cases, the integro-differential equations provide some analytic information concerning the survival probabilities. In particular, they are used for obtaining bounds and approximations for the survival probabilities or investigation of their asymptotics. Note that it is not a problem to write down the integro-differential equations for the survival probabilities and most of the equations we derive have already been obtained and used for further investigation. Furthermore, one can easily write down all these equations in terms of infinitesimal generators for stopped surplus processes, which are Markov processes (see, e.g. [ASM 10, p. 248] or [SCH 08b, pp. 220, 225]). Nevertheless, usually one has to make the additional assumption concerning the differentiability of the survival probabilities or conditions when these functions that are really differentiable are not studied. Thus, investigation of the continuity and differentiability of the survival probabilities is of great interest. As we will see, these functions are not always differentiable in the classical sense. Moreover, sometimes points of discontinuity are also possible. So, the continuity and differentiability of the survival probabilities must be investigated separately for each risk model, and we study these properties for different models in this book.

Next, we concentrate on some possible applications of the results concerning the smoothness of the survival probabilities. Firstly, we find analytic expressions for the infinite-horizon survival probabilities in a few cases when the c.d.f. of the insurance compensation is a sum of absolutely continuous and discrete components. Secondly, we solve the problems of optimal control by franchise and deductible amounts in the classical risk model from the viewpoint of survival probability maximization. Thirdly, we obtain relationships connecting accuracy and reliability of uniform approximations of the survival probabilities from their statistical estimates.

Finally, we introduce and apply the supermartingale approach to obtain upper exponential bounds for the infinite-horizon ruin probabilities in different generalizations of the classical risk model with risky investments. This approach generalizes the martingale one introduced by Gerber [GER 73] to obtain the famous exponential bound for the infinite-horizon ruin probability in the classical risk model, which is known as the Lundberg inequality.

The book is not a textbook in risk theory and does not provide general results. Nevertheless, we give some basic results concerning ruin probabilities in the classical risk model and the risk model with stochastic premiums in Chapter 1. The reader can find the general results in risk theory in [ASM 10, BEA 84, BÜH 70, DIC 10, GER 79, GRA 91, ROL 99]. Furthermore, we refer the reader to [SCH 08b] for a systematic description of optimal control methods applied to insurance problems and [AZC 14] for the viscosity approach to optimal control problems in insurance.

This work is intended for researchers in probability theory, actuarial sciences and financial mathematics, as well as for graduate and postgraduate students.

Though the book is mainly theoretical, it is also accessible to practitioners who want to extend their knowledge in insurance mathematics.

The book is self-contained and does not require any complementary material. All its main results are provided with proofs. It is assumed that the reader has a basic knowledge in probability theory, stochastic processes, stochastic calculus, ordinary and stochastic differential equations and calculus. We provide some well-known mathematical results which are often used in the main text in the Appendix. Moreover, we refer the reader to [DOO 53, ELL 82, JAC 06, KAR 91, ØKS 07, ROG 00, ROL 99, SHR 04] for basic notions of the theory of stochastic processes and stochastic calculus and to [CHE 05, IKE 89, KAR 91, ØKS 03] for basic facts concerning stochastic differential equations.

The main material of the book is divided into two parts and organized as follows. In Part 1, we investigate the continuity and differentiability of survival probabilities for different risk models and concentrate on some possible applications of these results. In Chapter 1, we give some basic results concerning ruin probabilities in the classical risk model and the risk model with stochastic premiums. In Chapter 2, we deal with a generalization of the classical risk model where an insurance company invests all surplus in a risk-free asset. We investigate the continuity and differentiability of infinite- and finite-horizon survival probabilities in detail. In Chapter 3, we obtain the corresponding results for the risk model with stochastic premiums and investments in a risk-free asset. In Chapter 4, we consider the classical risk model under the additional assumption that an insurance company applies a franchise and a liability limit. Assuming that claim sizes are exponentially distributed we find analytic expressions for the infinite-horizon survival probabilities and investigate how a franchise and a liability limit change the survival probability for small and large enough initial surpluses. In Chapter 5, we solve the problems of optimal control by franchise and deductible amounts in the classical risk model from the viewpoint of survival probability maximization. In Chapter 6, we extend the results of Chapters 2 and 3 to a case where an insurance company invests all surplus in risk-free and risky assets proportionally and the price of the risky asset follows a jump process. Moreover, we obtain relationships connecting accuracy and reliability of uniform approximations of the survival probabilities by their statistical estimates.

In Part 2, we apply the supermartingale approach to obtain upper exponential bounds for infinite-horizon ruin probabilities in different generalizations of the classical risk model where a premium intensity depends on a current surplus of an insurance company. To this end, in Chapter 7 we assume that all surplus is invested in one risky asset, the price of which follows a geometric Brownian motion, and the premium intensity grows rapidly with increasing surplus. Furthermore, we investigate the question concerning the probability of explosion of the surplus process between claim arrivals in detail. In Chapter 8, we suppose that all surplus is also invested in one risky asset, the price of which follows a geometric Brownian motion, but an insurance company stops its investment activity when the price of the risky asset goes below some fixed level or outside some fixed interval. In Chapter 9, we assume that the surplus is invested in one risk-free and a few risky assets, the prices of which follow geometric Brownian motions, and borrowing is also possible.

We conclude the preface by thanking all those without whose help and encouragement this book would never have been written. First of all, it is our pleasure to thank Hanspeter Schmidli,