22
The fair valuation problem of guaranteed annuity options: The stochastic mortality environment case Laura Ballotta ,Steven Haberman

1. introduction 2. A valuation approach for guaranteed annuity options

  • Upload
    burke

  • View
    37

  • Download
    0

Embed Size (px)

DESCRIPTION

The fair valuation problem of guaranteed annuity options: The stochastic mortality environment case Laura Ballotta ,Steven Haberman. 1. introduction 2. A valuation approach for guaranteed annuity options 3. A stochastic approach to mortality risk: the basic model and its extensions - PowerPoint PPT Presentation

Citation preview

Page 1: 1. introduction 2. A valuation approach for guaranteed annuity options

The fair valuation problem of guaranteed annuity options:

The stochastic mortality environment case

Laura Ballotta ,Steven Haberman

Page 2: 1. introduction 2. A valuation approach for guaranteed annuity options

• 1. introduction• 2. A valuation approach for guaranteed annuity

options• 3. A stochastic approach to mortality risk: the

basic model and its extensions• 4. A model for the financial risk and the GAO

valuation formula• 5. Numerical calculations and sensitivity analysis

Page 3: 1. introduction 2. A valuation approach for guaranteed annuity options

Guaranteed annuity option

• Guaranteed annuity option(GAO) is a contract giving the holder the right to receive at retirement the greater of

(a) a cash payment equal to the current value of the investment in the equity fund, S, (b)the expected present value of the life annuity obtained by converting this investment at the guaranteed rate.

Page 4: 1. introduction 2. A valuation approach for guaranteed annuity options

• Assumption:1. The mortality risk is independent of the financial

risk.2. Single premium S0(ignore any expense)3. The market is frictionless and competitive market

with continuous trading• Model1. Heath-Jarrow-Morton for interest rate2.Bullotta and Haberman for mortality intensity

Page 5: 1. introduction 2. A valuation approach for guaranteed annuity options

riskmortality andrisk financial:risk of sources by two affected iscontract GAO that theshow (3)~Eqs.(1)

where

(3) ˆ

is xageder then policyhold aby 0 at time enteredcontract GAO theof Tt0 at time valueThe

y. ageder policyhold a of lifetime remaining the

ngrepresenti r.v. a is andpayment annuity theof times theare T where

(2) ˆ

factor'annuity ' therepresents , 1/gK where

(1)

:x-N=Tmaturity at out pays GAO the age, retirement normal theis N 0, at time xaged iser policyhold theIf

GAOfor approach n A valuatio 2.

0

0

yj

0

0

0

0

0

)0(0

0

00

txx

]|FIC[eEt),t,T(xV

]|FIe[Ea

a

K)(agS)Sa(gSC

(t)

tt}T{τT

dur(t)

x

)x(Tw

jTT}T{τ

dur

Tx

Tx

TxTTTxTT

(t)x

T

tu

jTx

jT

Tu

Page 6: 1. introduction 2. A valuation approach for guaranteed annuity options

(8) .50,50

70)( (7) ,0

and rend,forecast t in the s variationrandom model tointroduced is which ,,on process stochastic a is 0 where

(6) e)0,(

),()0,(),( ,

(5) eu)RF(y,

:structure following thereduced bemay UK which in the spopulationpensioner andannuitant toeappropriat models of series a derive (2000) al.et Sithole (1996), al.Renshawet

s)(continuou ),()0,(),((discrete) ),()0,(),(

modelfactor reduction theis ratesmortality projectingfor model actuarial usedA widely

(4) ]|[p

)|(p years sleast at survive will t whoat time x agedperson ay that probabilit thep

extensions its and model basic the:riskmortality oapproach t stochasticA 3.

)12(21

)))(((

)(

),(

xs

xs

xs

2321

0

xzxR

eRa) =az,μ(x

F,P)(Ω):t(Yzx

ztzxRFzxztzxSo

uyRFyuyuyRFyquyq

FeE

FsP

RbRbb

t

Yztzx

uy

t

dzztzx

tx

zth

s

Page 7: 1. introduction 2. A valuation approach for guaranteed annuity options

1

let

mindet )D timesides(Both dY

:market. financial in the existing randomness of

sources theoft independenmotion,Brownian -P ldimensiona-one standard a is 0)t:(X where

90

equation, aldifferenti stochastic following thesatisfies Y process thes,other wordin process;Uhlenbeck -Ornstein

an by governed is Y that assume weconcerned, is (6) Eq. ofcomponent stochastic theasfar As

0

00

0000

t

t

t

0

with De D

dtaD dDdDYdY D)Y(*) d(D

dXe dXe e Y

dXe dXe YDY D

dXedXD)Y d(D (*)

istic)er (DdXDdtYaDdY DdXdtaY

sol

) ( ,Y

dXdtaYdY

att

tt

tttttt

t

vv)a(tt

vavat

t

t

vavt

vav

tt

tat

tttt

ttttttt

tt

tttOrnstein-Uhlenbeck model has the desirable property of mean reversion

Page 8: 1. introduction 2. A valuation approach for guaranteed annuity options

aliainter (2002), Pitacco and Olivieri ofapproach the followingby risk longevity eincorporat order toin above described model theenhance We

enhanced. are tsimprovemen subsequentfor potential then thens,expectatio behind fall tsimprovemen if Similarly,

reduced. be willtsimprovemenfurther for potential theafterwards then point, someat nsexpectatio exceed tsimprovemenmortality if that impliesreversion Mean

zero torun values-long a toreversion Mean 存機率最逐漸收斂到零該性質可確保高齡者生

是一項很好的性質

Page 9: 1. introduction 2. A valuation approach for guaranteed annuity options

motion.Brownian ˆ ldimensiona-one standard a is 0ˆand where

S

ˆdS

equation. aldifferenti stochastic following by the described is dynamic neutral-risk whosefund,equity an intocontract theofstart at the

er policyholdeach by paid premium single theinvestsinsurer that theAssume

formula valuationGAO theandrisk financial for the modelA 4.

0

t

P):tZ(Rσ

R

ZdSdtSr

ts

ttstr

Page 10: 1. introduction 2. A valuation approach for guaranteed annuity options

T

tduutf

tT

t

tt

tt

t

tt)λ(TT

t

t)λ(ut)λ(T

eTTtf

W):tW(

WW

dtZdWd

):tW(

Wdσe)dtduee(σdf(t,T)

),(

tt

'

'2t

2

)(P and ),(limr

rates.interest and valuesfundequity between t coefficienn correlatio therepresents Hence,

ˆ oft independenmotion Brownian P ldimensiona-one standard a is 0ˆ where

,ˆ1ˆZ implies This

0 ˆˆ that so

,Z with correlatedmotion Brownian P ldimensiona-one standard a is 0ˆ where

ˆ

bygiven is rate forward theof dynamic neutral-risk thes,other wordIn structure. decayinglly exponentiaan hasy volatilitrate forward the

in which case specific heconsider t weandframework HJMfactor -single aby given is ratesinterest of structure term theofevolution that theassume weabove, mentioned As

Page 11: 1. introduction 2. A valuation approach for guaranteed annuity options

.P~ measurey probabilit adjusted-risk-stock under then expectatio thedenotes E~ where

(16) ,ˆ

~

processdensity by the P~ P~ measurey probabilit a Define

)(Dahl,2004 market. financial therisk tomortality transfer toneed Then we

yprobabilit survival neutral"risk the"ngrepresenti where

)(

)( ˆ

ˆ ˆ

ˆ

ˆ

:follows asrewritten becan )2.(

0

0

0

),(

0

),(

0

),(

0

0

)(

0

)(

00

)(

00

)(

00

)(

0

)0(0

SSe

PdPd:η

p

TPp

TP]|F[eE

]|F[eE]|F[eE

]|Fee[E

]|FIe[Ea

Eq

Tdur

F

T

xTT

jT

)x(Tw

jxTT

jT

)x(Tw

jT

dzzTzx

T

dur)x(Tw

jT

dzzTzx

)x(Tw

jT

dzzTzxdur

)x(Tw

jTT}T{τ

dur

Tx

T

u

T

Tj

Tj

TjT T

jT

Tu

TjT T

TjT T

jT

Tu

jTx

jT

Tu

對於評價與死亡力相關的商品,不能在” risk neutral measure under financial risk”,而是應該找出一個” risk neutral measure under mortality risk”

Page 12: 1. introduction 2. A valuation approach for guaranteed annuity options

(14). Eq. asspirit same in the calculated are ,ies, probabilit survival retirement-post theandP, measurey probabilit real under the dynamic its equals process rate hazard theof dynamic- P~ theTherefore,

risk.mortality therespect to with neutral completely ismarket the that thatassume weearlier, discussed weas However,.

termmortality by the measure,-P~ under the ,'discounted' is payoff GAO terminalheequation t previous in the that Note

(17) ~

ˆ]|ˆ

ˆ

(15) ˆ

ˆ

ˆ

ˆ

is t at time GAO theof value theEq.(3) from ly,Consequent

0

00

00

00

00

0

0

0

0

0

0

0

0

(T)j

tT (t)

(T)j

tT (T)

(T)j

tT (T)

(T)j

T

tutT (T)

T

tu

(T)j

tT (t)

T

tu

tT (t)

T

tu

tT (t)

T

tu

(t)x

T

tu

xT T

z)dzz,tμ(x

tjT

)x(Tw

jxT T

z)dzz,Tμ(x

t

tjT

)x(Tw

jxT T

z)dzz,Tμ(x

Ttt

tjT

)x(Tw

jxT T

dur

Tt

z)dzz,Tμ(xdur

tjT

)x(Tw

jxT TT

z)dzz,tμ(xdur

tTxT

z)dzz,tμ(xdur

tT

z)dzz,tμ(xdur

tt}T{τT

dur(t)

x

p

e

] |FK))(TP p([eEgS

]|FK))(TP p(e[ηEF[SEg

]|FK))(TP p(eηSe[eEg

] |FK))(TPp(Se[eEg

]|FK)(agSe[eE

]|FCe[eE

] |FIC[eEt),t,T(xV

Page 13: 1. introduction 2. A valuation approach for guaranteed annuity options

measure-P~,GAO,

.0YdY

;50,50

70)(

,0 , e)0,(),(

(9)-(6) Eqs.in described process rate hazard theof P) measureder dynamic(un on the depend

: iesprobabilit survive retirement-post

:factor mortality retirement-pre

~

:Recallanalysisy sensitivit and nscalculatio Numerical 5.

'22

22

2

'222

2

000

2321

0

0

00

0

ˆ1ˆ)2

)(1()2

(

)ˆ1ˆ(2

ˆ2

ˆ2

0T

0

t

)12(21

)))(((

0

TsTsss

TTss

TssTs

sT

u

T

u

T

u

zth

TjT (T)

(T)j

tT (t)

(t)

(T)j

tT (t)

WWTT

WWTZTZTdurdurT

dur

jT

tt

RbRbbYztzx

T

z)dzz,Tμ(x

xTT

z)dzz,tμ(x

xt T

tjT

)x(Tw

jxT T

z)dzz,tμ(x

t(t)

x

e

eeeeSSe

)(TP

dXdtaYxzxR

eRa) =az,μ(xzxztzx

]FeE[p

ep

] |FK))(TP p([eEgSt),t,T(xV

下的利率分佈。或是找出在就得先算出因為我們想要評價

Page 14: 1. introduction 2. A valuation approach for guaranteed annuity options

motions.Brownian standard-P~ are 1ˆW~

ˆW~

thatimplies theoremGirsanov By the

)P~,F,(on motion Brownian a is )W~(, )[0,T fixedeach for Then

ˆ~

by )W~(W~ process theDefine

martingale a is 21ˆexpˆ

~Then

]E[e satisfying process adapted-F ,measurable a is )(XXLet

Theorem Girsanov

2''t

t

ˆ1ˆ)2

)(1()2

(

T

TT0t

0

0tt

0

2

0

|X| 21

t

'22

22

2

2T

0 u

tW

tW

e

duXWW

)duXWdX(PdPdη

st

st

WWTT

Tt

t

utt

t

u

t

uut

du

TsTsss

Page 15: 1. introduction 2. A valuation approach for guaranteed annuity options

~] )1(2

)[1(),0(),(limr

~)()](21)(1[),0(

~)(),0(

~)()1(),0(

~)(),0(),(

:

~])1(2

)[1(),0(r

bygiven is rateshort that theimplieswhich

,~)(

ˆ

thenis rate forward theof dynamic-P~ The

02

2

t

0

222

00

22

00

000

02

2

t

2

t

vv)λ(Tsλtλt

tT

t

vv)λ(TλTt)λ(TsλTt)λ(TλTt)λ(T

t

vv)λ(TλTt)λ(Tst v)λ(Tv)λ(T

t

vv)λ(TλTt)λ(Tst v)λ(Tv)λ(T

t

vv)λ(Tt

sv)λ(Tt T

v

v)λ(uv)λ(T

v

t v)λ(Tstt

tt)λ(T

s

T

t

t)λ(ut)λ(T

tt)λ(TT

t

t)λ(ut)λ(T

Wdσeσeλ

etfTtf

Wdσeeeσeeλ

eeλλ

Tf

Wdσeeeσdveeλ

Tf

Wdσeeeσdveλ

σeTf

WdσedvσedvdueσeTfTtf

pf

Wdeeetf

Wdσedt)due(σσ

Wdσe)dtduee(σdf(t,T)

Page 16: 1. introduction 2. A valuation approach for guaranteed annuity options

])1(2

)[1()(with

))(),((~),0(r

contract theof value theofn computatio for the s techniqueCarlo Monteadopt we,e

factor mortality theof propertieson distributi about the knowledge oflack theand process rate hazard theof naturedependent -path Given the

2

)1( )( and )1(),( where

)()(

)(

),(,..,2,1,Tat maturing bondcoupon zero a of T at time price ingcorrespond The

2

2

2T

),(

222

)(

)),0()(,()(),()21(

0

0

~)()](21)(1[),0(),(

0j

0

22

022

2

sλTλTr

rr

dzztzx

λT

r

TTλ

j

TfrTTTTTj

Wdσeeeσ

eeλ

eeλλ

tfdttTf

jT

σeλ

eTm

TTmNTf

λeT

λeTT

eTPTP

eeTP

xTj

s

j

Tjrj

jT

T

Tv

v)λ(tλtT)λ(tsλtT)λ(tλtT)λ(tjT

T

Page 17: 1. introduction 2. A valuation approach for guaranteed annuity options

))(,0( follows and Y oft independen is

integral stochastic the,particularIn integral stochastic the,particularIn

Y

][

Y

t,uany for 2

)1()( with )),(,0(~Y that implies ,Y

Y, process thegoverningequation aldifferenti stochastic thetosolution that theobserve weconcerned, is price GAO theofcomponent mortality theasfar As

2t

)(

)(t

)(

)(

0

)()(

0

)()(

0

)(

0

)(u

222

0 t)(

t

tuN

dXe

dXee

dXedXee

dXeedXedXe

aettNdXe

u

t ssua

u

t ssuatua

u

t sstat

sstatua

u

sstatuau

ssttuau

ssua

att

ssta

Page 18: 1. introduction 2. A valuation approach for guaranteed annuity options

5070)(R

)0,(

,)0,(),(

as generated becan rate hazard theofpath thely,Consequent on.distributi

normal edstandardiz a from samples randomt independen of sequence a is }where{z

)(Y

as Y processUhlenbeck -Ornstein theofpath thegenerate westep,each At .,...,2,1,0, define we

and,n

t-Tt fixed of lssubinterva equaln into t]-T[0, period timeobserved thesubdivide we

,e

factor mortality retirement-pre theofn computatio for the Thus,

)(

i

)12(21

)(

)))((()()(

n1,.....,i

)(t

i

),(

2321

)(

i

1i

0

)(

it

RbRbbii

t

Ytxi

tii

t

tta

dzztzx

x

eRaax

extx

ztYe

niti

ii

ithiit

ii

tTt

Page 19: 1. introduction 2. A valuation approach for guaranteed annuity options

))). ξ(T N(0,~ Y (i.e.on distributi its from generated isor 0), = Y (i.e. mean value itsby edapproximat iseither Y andignored, is Y process theof

history past hein which t case in the also iesprobabilit survival retirement-post theof simulation out the carried have weprocedure,ion approximat theof

ion specificatlast thisgardingprocess.Re retirement-pre theofn calculatio thefrom resulting Y of valuesrun theeach at using above, described

procedure same by the generated is rate hazard retirement-post theHence,

py probabilit survival retirement-post

])t,(x2)t,(x)t,(x[2

z)dztz,(x

factormortality retirement-prerule ltrapezoida

TT

T

T

xT

1-n

1kkk

(t)nn

(t)00

(t)

t-T

0

(t)

(T)j T

t

接下來我們需要計算

來逼近接著我們使用

Page 20: 1. introduction 2. A valuation approach for guaranteed annuity options

2100

210021122

10010011

x

2

1

whereet)m(x,

modelCarter and

1

0

10

1

0

11

1002

100

112

001

εεxβεxβkxβcxβxcβxα

xxxxxx

xxxxxxxx

ttx

e

eee)m(x,

eee)m(x,

εk c

εkc k εεkcc

ε)εk(cc εkc k εkck

Lee

ε)ε)εk(c(cβαε)εk(cβαεkβα

)εβ(εβkcβαε)εk(cβαεkβα

t

jj

jtt-

j

tj

ttt

k

Page 21: 1. introduction 2. A valuation approach for guaranteed annuity options

aettN

dXe,Y

dXdtaYdYand

etxm

wherekNtxm

txm

at

t

sstattt

Yk

t

j

jxrr

tjx

k

k

k

trtj

x

tt

jj

jtx

tjx

tt

jj

jttjx

ttx

2)1()( with )),(,0(~Y that implies

,Y 0

),(

)(1 )),c((~),(ln

e

e

e),(

222

t

0

)(t

0

)c(

1

0

22220

1-t

0jx

)c(

)c(

01-t

0jx

1

0

10

1-t

0jx

1

0

10

1-t

0jx

x

1

0

10

1

0

t

jj

jtt-

j

tjt εkc k

Page 22: 1. introduction 2. A valuation approach for guaranteed annuity options

xxxx

0

1

00x

1

p

)( t)()( p)(

t )(

2

))(1ln(

)( 1)(

22

22

2

2

22

222

2007) (BAJ ,assumption UDD

t

xt

txt

t

xxsxt

x

kkx

x

xx

x

x

xx

x

xx

x

xxx

x

xx

x

x

x

qttqdsspq

Method

q

Method

m(x,t)m(x,t)q(x,t)

q(x,t)q(x,t)

ldl

ld

dld

dl

d)d(ll

dll

dLdm(x,t)

(x,t)m(x,t)Under

固定假設各區間的死亡力皆連續型離散型校正預測死亡率