Stochastic mortality and securitization of longevity risk Pierre DEVOLDER ( Université Catholique de Louvain) Belgium devolder@fin.ucl.ac.be Purpose of the presentation Suggestions for hedging of longevity risk in annuity market Design of securitization instruments Generalization of Lee Carter approach of mortality to continuous time stochastic mortality models Application to pricing of survival bonds Edinburgh 2005 Devolder 2 Outline 1. Securitization of longevity risk 2. Design of a survival bond 3. From Lee Carter structure of mortality… 4. …To continuous time models of stochastic mortality 5. Valuation of survival bonds 6. Conclusion Edinburgh 2005 Devolder 3 1. Securitization of longevity risk Basic idea of insurance securitization: transfer to financial markets of some special insurance risks Motivation for insurance industry : - hedging of non diversifiable risks - financial capacity of markets Motivation for investors : -risks not correlated with finance Edinburgh 2005 Devolder 4 1. Securitization of longevity risk 2 important examples : CAT derivatives in non life insurance Longevity risk in life insurance THE CHALLENGE : - Increasing move from pay as you go systems to funding methods in pension building - Importance of annuity market - Continuous improvement of longevity Edinburgh 2005 Devolder 5 1. Securitization of longevity risk Evolution of qx in Belgium ( men) -return of population : 1880/90 1959/63 2000 x= 20 0.00688 0.0014 0.001212 x= 45 0.01297 0.00491 0.002866 x= 65 0.04233 0.03474 0.0175 x= 80 0.1500 0.11828 0.0802 Edinburgh 2005 Devolder 6 1. Securitization of longevity risk Hedging context : L x initial cohort of annui tan ts aged x at time t 0 Initial total lump sum : P Lx a x Lx x t 1 r t p v t x where : p r reference survival probabilit y Edinburgh 2005 Devolder 7 1. Securitization of longevity risk -cash flow to pay at time t : CFt L x t p px ( p p actual survival probabilit y stochastic process) -cash flow financed by the annuity : CFt * L x t prx Edinburgh 2005 Devolder 8 1. Securitization of longevity risk Longevity risk at time t ( « mortality claim » ): LRt Lx (t ppx t prx ) Random variable Edinburgh 2005 Devolder Initial Life table 9 1. Securitization of longevity risk Decomposition of the longevity risk : LR Diversifiable part ( number of annuitants) General improvement of mortality Edinburgh 2005 Devolder Specific improvement of the group 10 1. Securitization of longevity risk -Hedging strategy for the insurer/ pension fund : - selling and buying simultaneously coupon bonds: Floating leg: Index-linked bond with floating coupon SURVIVAL BOND Edinburgh 2005 Devolder Fixed leg: Fixed rate bond with coupon CLASSICAL BOND 11 2. Design of a survival bond Classical coupon bond : t=0 t=n k k k 1+k Survival index-linked bond : t=0 t=n k1 k 2 k 3 Edinburgh 2005 Devolder 1 k n 12 2. Design of a survival bond Definition of the floating coupons : Hedging of the longevity risk LR -General principle : the coupon to be paid by the insurer will be adapted following a public index yearly published by supervisory authorities and will incorporate a risk reward through an additive margin Transparency purpose for the financial markets : hedging only of general mortality improvement Edinburgh 2005 Devolder 13 2. Design of a survival bond Form of the floating coupons: The coupon is each year proportionally adapted in relation with the evolution of the index. k t k (1 t prx I t ) k * Initial life table Mortality Index Edinburgh 2005 Devolder Additive margin 14 2. Design of a survival bond Valuation of the 2 legs at time t=0 : Principle of initial at par quotation : n n k P(0, t ) P(0, n ) E t 1 t 1 Zero coupon bonds structure Q (k t ) P(0, t ) P(0, n ) Mortality risk neutral measure Edinburgh 2005 Devolder 15 2. Design of a survival bond Value of the additive margin of the floating bond : n k* k r ( E ( I ) p Q t t x ) P(0, t ) t 1 n P(0, t ) t 1 1° model for the stochastic process I 2° mortality risk neutral measure Edinburgh 2005 Devolder 16 3.From classical Lee Carter structure of mortality…. Classical Lee Carter approach in discrete time: (Denuit / Devolder - IME Congress- Rome- 06/2004 submitted to Journal of risk and Insurance) p x ( t ) Probability for an x aged individual at time t to reach age x+1 p x (t ) exp( x ( t )) Time series approach Edinburgh 2005 Devolder 17 3.From classical Lee Carter structure of mortality…. Lee Carter framework : x ( t ) exp ( x x t x t ) Initial shape of mortality Mortality evolution Edinburgh 2005 Devolder ARIMA time series 18 4….To continuous time models of stochastic mortality Continuous time model for the mortality index : t I t exp( x (s) ds ) 0 x (s) stochastic mortality force at age x s at time s Edinburgh 2005 Devolder 19 4….To continuous time models of stochastic mortality Example of stochastic one factor model 4 requirements for a one factor model : 1° generalization of deterministic and Lee Carter models; 2° …taking into account dramatic improvement in mortality evolution ; 3° …in an affine structure ; 4°… with mean reversion effect and limit table . (+strictly positive process !!!!!!!!!!) Edinburgh 2005 Devolder 20 4….To continuous time models of stochastic mortality Step 1 : static deterministic model : Initial deterministic force of mortality : x s x s (0) exp ( x s ) ( classical life table = initial conditions of stochastic differential equation) Edinburgh 2005 Devolder 21 4….To continuous time models of stochastic mortality Step 2: dynamic deterministic model taking into account dramatic improvement in mortality evolution : x (s) exp( x s x (s) s) x s exp( x (s) s) (prospective life table ) Edinburgh 2005 Devolder 22 4….To continuous time models of stochastic mortality Step 3:stochastic model with noise effect – continuous Lee Carter : x (s) x s exp( x (s) s) (s) with exp onential martingale : s 1 2 (s) exp( ( u )dz ( u ) ( u ) du ) 2 0 Edinburgh 2005 Devolder z= brownian motion 23 4….To continuous time models of stochastic mortality This stochastic process is solution of a stochastic differential equation : d x (s) x (s) x s x (s) s x (s) ds (s) x (s) dz (s) x s Classical model Time evolution Edinburgh 2005 Devolder Randomness 24 4….To continuous time models of stochastic mortality Step 4: affine continuous Lee Carter ( Dahl) : d x (s) x (s) x s x (s) sx (s) ds (s) x (s) dz (s) x s Change in the dimension of the noise Edinburgh 2005 Devolder 25 4….To continuous time models of stochastic mortality Step 5: affine continuous Lee Carter with asymptotic table : We add to the dynamic a mean reversion effect to an asymptotic table ~ Deterministic force of mortality x s Introduction of a mean reversion term : ~ (s)) k( x s x Edinburgh 2005 Devolder 26 4….To continuous time models of stochastic mortality ~ (s)) ds d x (s) x (s) x s x (s) s x (s) ds k( x s x x s (s) x (s) dz (s) Mean reversion effect Edinburgh 2005 Devolder 27 4….To continuous time models of stochastic mortality Step 6: affine continuous Lee Carter with asymptotic table and limit table : Introduction of a lower bound on mortality forces: ~ * x s x s x s Present life table Expected limit Biological absolute limit Edinburgh 2005 Devolder 28 4….To continuous time models of stochastic mortality ~ (s)) ds d x (s) x (s) x s x (s) sx (s) ds k ( x s x x s (s) x (s) *x s dz (s) …in the historical probability measure… Edinburgh 2005 Devolder 29 4….To continuous time models of stochastic mortality Survival probabilities : T Tt p x t ( t ) E P (exp x (s) ds t ) t In the affine model : Tt p x t (t ) exp( A(t, T, x ) B(t, T, x ) x ( t )) Edinburgh 2005 Devolder 30 4….To continuous time models of stochastic mortality Particular case : - initial mortality force : GOMPERTZ law: x s b cx s - constant improvement coefficient : -constant volatility coefficient : Edinburgh 2005 Devolder 31 4….To continuous time models of stochastic mortality Explicit form for A and B : 2(e ( T t ) 1) B( t , T, x ) ( )(e ( T t ) 1) 2 with : ln c k 2 2 2 T 1 2 ~ A( t , T, x ) (k x s B(s, T, x ) *x s B2 (s, T, x )) ds 2 t Edinburgh 2005 Devolder 32 5. Valuation of survival bonds Introduction of a market price of risk for mortality : Equivalent martingale measure Q t z Q ( t ) z ( t ) h (s, x (s)) ds 0 Valuation of the mortality index : t E Q ( I t ) E Q ( exp x (s) ds ) 0 Edinburgh 2005 Devolder 33 5. Valuation of survival bonds Affine model in the risk neutral world: h (s, x (s)) x s x s *x s *x s x s *x s Mortality index : E Q (I t ) exp( AQ (0, t, x ) BQ (0, t, x ) x (0)) Edinburgh 2005 Devolder 34 5. Valuation of survival bonds Valuation of the additive margin : n k* k Q Q r (exp( A ( 0 , t , x ) B ( 0 , t , x ) ( 0 )) p x t x ) P (0, t ) t 1 n P(0, t ) t 1 Interpretation : weighted average of mortality margins Edinburgh 2005 Devolder 35 5. Valuation of survival bonds n k* k ( MM t 1 t P(0, t )) n P(0, t ) t 1 Decomposition of the mortality margin : MMt MM(t1) MM(t 2) Edinburgh 2005 Devolder 36 5. Valuation of survival bonds MM(t1) exp(A(0, t, x) B(0, t, x)x (0))t prx = longevity pure price MM (t 2 ) exp( A Q (0, t , x ) BQ (0, t , x ) x (0)) exp( A(0, t , x ) B(0, t , x ) x (0)) =market price of longevity risk Edinburgh 2005 Devolder 37 5. Valuation of survival bonds Particular case : GOMPERTZ initial law and constant , , ~ (Tt ) 2( e 1) B ( t , T, x ) ~ ~ ~ ( T t ) ( )(e 1) 2~ with : ~ ln c k ~ ~ 2 2 2 Q Edinburgh 2005 Devolder 38 5. Valuation of survival bonds T ~ * )B(s, T, x ) ds A Q ( t , T, x ) ( * k x s x s t T 1 2 *x s B2 (s, T, x ) ds t 2 Edinburgh 2005 Devolder 39 6. Conclusions Next steps : Calibration of the mortality models on real data Estimation of the market price of longevity risk Other stochastic mortality models for the valuation model Edinburgh 2005 Devolder 40