for insurance companies Non-dangerous risky investments Manfred - - PDF document

for insurance companies non dangerous risky investments
SMART_READER_LITE
LIVE PREVIEW

for insurance companies Non-dangerous risky investments Manfred - - PDF document

for insurance companies Non-dangerous risky investments Manfred Sch?il 6, (e-mail: schael@uni-bonn.de) Angew. Math., Univ. Bonn, D-53115 Bonn, Wegelerstr. Inst. dynamic programming Key words. ruin probability, investment, financial market,


slide-1
SLIDE 1

Non-dangerous risky investments for insurance companies

Manfred Sch?il

Inst.

  • Angew. Math., Univ. Bonn, D-53115 Bonn, Wegelerstr.

6, (e-mail: schael@uni-bonn.de)

Key words. ruin probability, investment, financial market, stochastic dynamic programming The problem of controlling ruin probabilities is studied in a Cram6r-Lundberg model where the claim process is described by a compound Poisson process with claim size Y n at claim time Ti. The number Ni of claims in (0,t] is a Poisson process with claim intensity 1,. There is a premium (income) rate c which is fixed. The insurance company can invest the capital (surplus/risk reserve) in a financial market where 2 assets can be traded. One of them is called the bond and is described by the interest rate which here is assumed w.l.o.g. to be zero. The other asset is called stock. It is described by a l-dimensional price process {Sn, n > 1} where Sn is the price of one share

  • f the stock at time Tn. The price process

will be driven by a compound Poisson process which can be defined by the sequence {Tn, n>l } of market jump times and the sequence

  • f returns

{Rn, n21}, where 1 + Rn > 0 and ( l ) S = $ . . ( l + R ) . n n-r n' We write Nt for the number of market jumps in (0,t] where {N,} is a Poisson process with market intensity

  • v. In general,

v will be much larger than 1,. Thus the price process is driven by a Lövy process as in the Black-Scholes

  • model. However,

a compound Poisson process is chosen in place of a Wiener process. Moreover, only a moment condition is assumed for the distribution of Rn. Thus, the model for the financial market is quite general and flexible. The main advantage

  • f the Black-scholes model is the completeness
  • f the

financial market. But this property is not needed in the present control problem. We define the Poisson process {Nr} bf superposition: (2) N, := N, + N, is a Poisson process with parameter l, + v and jump times Trr, n ) 1.

P[Rn<0] >0,E[Rn] >0 and E[Rfr] <"",

P [ K n - 1 ] = # = 1 - P [ K n = o ] ,

q ,= 3.E [yn]

We write Kn = 1 if the jump at Tn is caused by the financial market and Kn = 0 if the jump is caused by a claim. Then we make the following assumption: Model Assumption: All random variables Zn,= Tr, - Tn_', Yrr, Rrr, Krr, n)1, are independent. The (Zn) are üd and have an exponential distribution with parameter ?t+v; the (Yn) are iid and positive; the (F.n) are iid as well as the (Kn).We assume

(3)

< | for the classical ruin probability qwith start in 0.

slide-2
SLIDE 2

2

In the joint model of insurance and finance, {Tn, n20} are the decision times and the real-valued discrete-time proces. {Xrr, n)0} describes the risk process (surplus process) immediately after time Tn. A dynamic portfolio specifies a portfolio 0n e R at any time Tn. There 0n represents the amount

  • f capital invested

in the stock. We have the following law of motion:

(4) Xn+1 = Xn * c.Zn+l* 0n.Rn+l for Kn*, = 1, Xn 2 0. Xn+l = Xr, * c'zn+l- Yn+l for Kn*t = o' Xn 2 o'

slide-3
SLIDE 3

For technical reasons we set Xn+l r € for Xn < 0. Once in state x e (----oo,Q;, the system moves to the absorbing state

  • "" in the next step.

We write O(x) for the set

  • f all portfolios 0 admissible

at x which is assumed to be

O(x) = [0,^] ,x ) 0. We define @(x) := {0i for x < 0. For 0 e is invested in the bond, i.e., which is not negative amounts 0rr, thus excluding short A number x < 0 represents a state

  • f ruin.

tp such that q(x) c @(x) for all x. Then q(Xrr) specifies the portfolio 0r, e @(Xrr) for the period

  • , ,l. We will sometimes

write for the state (risk) process

n + l r

= Xx'Q ' n n Our performance criterion is the ruin probability: (6) V(x,q) '= P [Xä'9 < 0 for some n] At first view, the ruin probability is not a classical performance criterion for control problems. However, one can write the ruin probability as some total cost in an embedded discrete-stage model where one has to pay one unit of cost when entering the ruin state. After this simple

  • bservation,

results from discrete-time dynamic programming for minimizing costs apply. Lundberg inequalities will be derived for the controlled model which extend the classical inequalities for the uncontrolled model. The present paper is related to Gaier, Grandits & Schachennayer (2003) where a continuous-time control model is studied for a Black-Scholes market by different methods. If ty(x) is the classical ruin probability for an initial reserve x, then V(x) = V(x,qg) where go(x) = 0. If ro > 0 is the classical Lundberg exponent then a classical theorem says 0 < C^.e-ro'^ ( V(*) = V(x,qg) ( e-ro'X for some constant C^ > 0.

  • Now let Q

be a stationary investment plan such that the decision maker invests a constant fraction y of capital at any (decision) time, i.e. 0(x) = T.x, then it was shown by Paulsen & Gjessing (1997) and Frovola, Kabanov & Pergamenshchikov (2002) that the asymptotic behaviour of the ruin probability is completely different under the investment plan Q. In fact, in the latter case the ruin probability has a polynomial decay (as function of the initial reserve) even if the financial market is described by a stock price process with high returns. Therefore, this plan is called dangerous in Frovola, Kabanov & Pergamenshchikov (2002). However, if Q is an investment plan such that a constant amount A is invested in the stock independently

  • f the current risk reserve,

i.e. Q(x) = A, then one can find some A and some exoonenti>r suchthat @(x), x - 0 represents the amount of the capital which invested in the stock. In this model we do not allow for selling

  • f the stock.

A stationary (investment) plan is a measurable function

(T .T ' n ' (5)

slide-4
SLIDE 4

(7) V(x,,Q) < "-t't (8) V(x,q) > Ö'"-i '* fo, some 0 < Ö < 1 and for every investment plan g (see Gaier, Grandits & Schachermayer 2003, Schäl 2005). A plan with (7) for some i t ro will be call profitable. The plan rQr may however be not admissible when the insurance company is poor, since rQ(x) = A G @(x) for x < A. In the present paper we study stationary investment plans <p* such that (9) q*(x) = min (x, A) for some A > 0. Such a plan is admissible. THsoneI\,I There exist a plan q* of the form (9) and some r* > rosuch that (9) V(x,q*) < "-t*'* . Then q* is admissible and profitable. Of course in view of (8), we have rx ( i in (10). THeoneIvI One even can choose rx = i if the market intensity v is high. If v is high which is a natural condition, then the model is close to a continuous-time model. 'PARADoxoN' Assume (i) Yn -F.2, i.e. the claims have an Erlang distribution; (ii) the price process is a martingale, i.e. E [Rn] = 0; (iii) v is large. Then there exist some stationary plan <p such that q(x) = x for small values x and V(x,q) ) V(x,<po) = V(x) and y(x,<p) > V(x,<po) = V(x) for small x. Hence, if the system is close to ruin, it may be good to invest all the capital in a martingale. The paper builds on methods from discrete-time control / stochastic dynamic programming / Markov decision processes. References Frovola A G, Kabanov Yu M, Pergamenshchikov S M (2002). Finance and Stochastics 6: 221135 Gaier J, Grandits P, Schachenneyer W (2003). Annals of Applied Probability 13: LO54-1016 Paulsen J, Gjessing HK (1997). Advance in Applied Probability 29:965-985 Schäl M (2004). Scand. Actuarial

  • J. 3: 189-210

Schäl M (2005). Math.Meth.Oper.Res. 62: l4l-158 Schmidli H (2OOZ). Annals of Applied Probability 12: 890-907

slide-5
SLIDE 5

N

  • f

dong,^or^s a'irhr.1 tnsc.s+rilrnh

ün thc insrncrrcr' buslruss h ourhs.l

Schä [

"F"SrJ. t'lath.

Bonyr

ln.ßt.

Uniy.

slide-6
SLIDE 6
  • ConhoL
  • [ ^uin P^obald

Lrf iis

by clircnc{e' trrrr., tnocshnGrh t1 qn hecl SchäL Inrt, An$cto. f1 4[h,

LLnr'v. Bonn

Papr^s tn ecnFi narou{ Htrnc "

IS rrorrn s ,

  • S. (1f

qS) I G cr'wr.1,, Grtenclih.P.r SchoGlA{ *1Y,n,U. l.+t,ppfc.rPLuwr.t'l Cl,ooo) (?oo3 I t*iPP,C,, Sc.hniielLirhl . (a0o3) Sehmi cllt.+{. ( ZOO4)

P s ulscn & 6iess irrj ( tI t? lrt)

f4ouolcr , Kgb crnov 8,

P-gavilensh chi hov [ 2 Oof )

Faninde? Gf o[, ( r Po'r?}

slide-7
SLIDE 7

Qene^ctl

(l

Ca

[n

  • wrc\,Jo^h-

Ion ltn av\c, vY\q^hßl

S ^ - S o

t

€.Le stoch p^Lce

t Lü] Ld, vy p4ocess

Ex 1 BLqck- $c,ho[es

' - o - ' U t + ( F - + c r a ) ' t

,r^BS

S ; ' - S o ' I { g U r + [ ' t t )

€x 2 eo wtoolrncl Fo isson

L L = t Nrl Poisson wi{h 4crk v

=+ S; : So'TTH (r + Rr,)

\Jecrk con\r4^qencq

{ s; } *t s:t } ( v-'oo)

LLt t zJl B: j -tuur+(F- !o)r]

Lr

=+

€ + J r N 1

1z- n=4 R: \ -t to-ur -+

rl\

slide-8
SLIDE 8
  • b

Neqessc{zr{

r,):. B: :=

z[: R-l

  • SernL-knq^t=i

t4 3c,t-e- clectemP.

44 = v' ELBUI'I G..(p-Lcl)'t

qondihiovrs Io4 convezr.gel4c€

m B + A T

i l L

X f + A F

E E

v.Et Rol +. 1 n. t

  • v. Et (R")tl .t -r

qz I

r1

A - . -

L

\ | | a t

l_

TxsrnPLe Dengb^ (tq(6) Coaneu Diss.

?aiqer,t* )

Retg.^ !+ | (2oo\) J, E-unprnical T,n,

Sccr iLIef l P C E: : b" ): p' , PC ß: E

  • bu)=1-po

v be ': crz

( e p - 1 ) v ' b ! : f - + t r ?

erLso s urttdcte.n+

slide-9
SLIDE 9

O c

LepL cAu f-aqruf erntryr g(s),: Ef e,- tR J v.[ 1

  • B(s)l

:- S. u.Et Rl \ sl .V.E[Rt] +

  • CtJ

#

I

't' A F (

  • --+) s.F - +- sa

g-e

V + s r

S=a g J*o* v'[ 1a e.Cs)]

  • +

q^he} ptice of aiste-

( r t {t

rt

%

slide-10
SLIDE 10

1

X = x

()

cop,toL

s Iope

I

I

I

sto ch

C.tam.e4 i Luncllr*g

cLqtrn

ttrirr l.

' l i

"?e11fu/.vr

S^+r= 5; (rl + Rn+r)

Yn*n

I

I

I

vLce

I

I

sn

I

l

  • b

I

I I

n

I

I

I

lio

I

I

I I

cl

I

I

I

  • l

I

P

I

i i

Suprn

l l

t l

P.o iss

  • n/tok V

I

t

  • T. T.

a1-J

zn Polss

  • n/4qk \+

= 1- PI Kh=

  • J

v

r r V

  • A + v

Pt Kn=11

slide-11
SLIDE 11

I

X,, c q ptt a[ of Tn

9n qmount ol cqpü+ol i,nuesle,cf

X r . * t = X r , + e o Z n * , +

0,,,. Rr.,+1.Kn+1

  • x+1 ,

(1 - K,.,+1)

ed nisslble qclion-s e fr^ T Xn >o

plon lconhol

0n: ?^(X^) e O(X",)

l r - ( . ? o , y r , g a , , . . )

s+ s fi

  • n

aÄy

T r = ( g , ' g r . . . ) : i g *

+ iplce[ 1 yo = Q (x) = tr.x dcrng urorrs !

+ ( x ) = A n o t o d m i s s i b h ,

g ( x ) = r n l n ( x , A ) .'tuLn p^obcr\ri Li { y

rP(n,x) '= PtXÄ'*

Xssufip{, \.8tyl

slide-12
SLIDE 12

e A

clcrssLceL

$oa some Co) O

  • <,zo

Lu nd b*g ex ponen{

P cuLscn & 6lessing ( trq +l ?e )

:f4ouoLc , Kobqnov & Prrrgcrlilsnsh ehi hov ( lOOf )

S e x ) = ö , x + Y ( F*,t)

$ o a s o m e ö > o r s > o G cr i cru, G ^on 4 i ta r S ehod/iAln4.f 0A (loGI !

l+ipF, Sehrrri,clu (zoo3) sc,h. (zoosJ

$ ( x ) = A

Hr'net g(x) = rnin(x,A) ) 3A >0,474o s{,

v ( T : x )

slide-13
SLIDE 13

3

narthov Dec'isi

  • n Pao

cess

[o^ rnini wrizivr5 (,+o+ß[) cosls

  • nc-s+.p

cor\ g(x) = 1a_ e*ot(*)

N n+1 =

  • o.

( qbsolbitttr ) tf K r, <O

  • ns- Stf
  • F{ ato,t T

[ o . n V : l R 1 t 0 , o o )

Tqv (x ),= E[v(Xn{,r) [Xn= )(, 0.= f (XJJ ,rr0

9 $on (q,.?- ) i€ ( q ,ql ,T,,,, )

\+, (q,.?')

  • TS U(t*)

T * v i = i n [ . e T ? V

T h c n T . ? \ , ( g o a ) B V ( T - )

T n E o R e m ( v z 5 )

s ) T + v s V = + V r ? * t

( * ) E t v ( X ; ' " ) r r i ( X ; " ) l

  • v x V r r

LE r-rMA

  • er)

Pt 0( X^o< I'l ü.o.] =O Vn >o,V,.,ttr

b ) ( x ) h o [ d s t l v ( o " ) : Q , v ( x ) = \ V x < 0

le) T*v

slide-14
SLIDE 14

S c t v ( x ) : m l n ( e t x , 1 ) T.rt usuaL oss'

  • Y1

rn(s) :: EI es Y]

( sunatJ' cLaims co,se)

3 e

I s s . 1 + R . > o , P E R <

  • l

> o , E ' t R l > O l.(s) := Et e--tRl

A !: sutrg>O v'U- e(s))

1. emriA

Let 4 be

s* )o

, A > o

posilive

(X)x'(rn(a)- 1) := (clasricats v=O + L errr*l A ä ex\sts ,

  • DEF. A,= "*/ i

I

soLulion

A + c'4- A = Q = +

^

4

t

= lol

.p(x)

  • rrrin

(x,A) v " ( 1'- s,( s't)) (1- ecsl)

slide-15
SLIDE 15

Tn eoR,en

( o r ) 1 4 e , ( t . o , t ] s u c h { t a o {

\*( T*,x) S e-- t-K r x zo

(b ) \p( gTx ) ( e. ä-("-A) = J e-- tx , xv^

rl

(*rf) ).( 1 +ä ..ä) e.r*< A + \ + c t

Convc^3u^(4, te Btach - sct^o\rr

T f,.r{r?- CX ) =iCX )u -

(,r()s ) (vn(e)- 1) : ,0ä), + c.i-

is sqme e.cf r^ql'iovt lol ä cß qorr_ ßto cl'L

  • Sct4
  • Lrr n4a k4l

C cp- d a i c,t I erasn clik / süoünl*or€A.)

1loaq-orl-u\ (X X ) =.,(X X)t--

(rtr()-\ r^n/vz -1 ^ '

,\.lu'/ - 1l

S B

slide-16
SLIDE 16

IT)

. \ , \

4 = a(v) + As u^t h.ene rn (I ) < oe V t

Yn Ct) -

  • e
  • frrl. t - 4.e

C uru a L qssump{. wi{ü0<a *äe€) t+.tlai ä*(oo I

Exqmptt Y,-, fr, ) r*- 13

L o L l e R ß o u N D v ( x ) ? € C , d ä x

T * v

4

  • ^

,

= : r = s u P y z o E I e ä ( Y - Y ) l y > y ]

C t y Z O

  • b -

' f

Tn EoRErr .

V ( r r , x )

slide-17
SLIDE 17

+

q) Tqv

b ) T * v

A p F L f c A T t o N , v : V ( . ? - )

+ T*v

imphouelnc,n{ lfut€,:f'q =o (+ v clagiceL rrqin prr&)

  • $\
  • f

(a ) €t Houa^d

( b) e> \tt rLtic,ohion {hcoa ewl

t.c. T* v =v + yi ts optiüal

  • DEF. cp** optimol Flo$qrcl

irnpnoucilent

  • [?o'

3e)

E xemPLEs

q ' = X , E T Y J I C

I ) Y ^ ' E , l

V ( 9 ; , x ) : q . e - t o ' X

n ) y ^ ' E a

V( Y.-, x) = (q+ f )d te'x - f e- Ü''x

O< {o ( ü [uh{qeLJ Lundb*$ cor,lf.

  • < l -
slide-18
SLIDE 18

E

iU

E r R l > o

uhac

?*t oplimaL 1, epr (x) =

3 A r

= O

'1 F

  • I

^ r

I ) Y ' w E 4

EtR, 1<0 - qk op+r'ileL

  • 1+"t l
  • u

q^d Lmploucntcn{

S c x ) > o V x 7 0

f.f

  • u a^d i wrpfo\rtm3n{

rntn (x , A* )

A

JI) Ytu Ea

E I R I

v

(c- \)1

  • e

\tJ ( Etyl

  • -l

)

  • e opFimaL

lf uua^d trnp^oeemqd .?,'-

$ h ü ( ? t t ( x ) = X V x 4 x r

( 1 x s > o )