A s t m B u l l e t i n 11 (198o) 29-34

PREMIUI~( R A T E S

UNDER

INFLATIONARY

CONDITIONS

H. R. WATERS *
Heriot-Watt Univermty, Edinburgh

I. INTRODUCTION

This short note has as its starting point an interesting artlcle b y TAYLOR (1979) m which he considered the effects of inflation on a risk process. T a y l o r showed t h a t ~f the p r e m i u m density increased at the same rate as the cost of individual claims then, under certain conditions, ultimate rmn was certain. This raises a natural question, vlz. "If the cost of i n d l w d u a l clam~s is increasing how should the premiums be increased m order to keep the probability of ruin under control ?" It ~s this question t h a t we shall be considering in this note. In the next section we define the risk process t h a t we shall be s t u d y i n g for the remainder of thts note. Our process is essentially a c o m p o u n d Poisson process except t h a t we allow the distribution funcUon of an indiwdual clann to d e p e n d on the time at which the claim occurs. We start the t h i r d section b y deriving, with the help of a general result of GERBER (1973), a formula for the futuxe p r e m i u m density t h a t will keep the p r o b a b i h t y of r u m for our process below a p r e d e t e r m i n e d bound. We then derive a simple a p p r o x i m a t i o n to this formula t h a t shows more clearly how we require the p r e m m m density to change in relation to the change in claims costs. Finally we show t h a t if we consider the same process with annual premiums t h e n the p r o b a b i l i t y of u l t i m a t e ruin wall be kept below a p r e d e t e r m i n e d b o u n d if the annual p r e m i u m is calculated according to the principle of zero utility with an exponential utility function or, as a first a p p r o x i m a t i o n , according to the variance principle.
2. THE RISK PROCESS

In this section we describe the risk process t h a t we shall be s t u d y i n g in this note. We assume claims are i n d e p e n d e n t of each other and occur as the points of a Poisson process with a m e a n rate of p claims each year. The a m o u n t of a single claim occurring at time t years has distribution function Fe where
(1)

F t ( x ) = Fo(X/g(t) )

t ~ o

and where ~(t) (> o) is a non-stochastic index of claims inflation at time t and
* T h e a u t h o r is g r a t e f u l t o t h e r e f e r e e for s o m o u s e f u l c o m m e n t s o n a n e a r l i e r d r a f t of thas paper

30 i(o) = that (2)

H. R WATERS 1. If we denote the m o m e n t g e n e r a t i n g function of F~ b y rnt it is clear

redO ) = mo(Oi(t) )

W e shall a s s u m e t h a t too(0 ) is finite for all values of 0 b u t note the r e m a r k s following the proof of t h e t h e o r e m in the n e x t section. W e denote b y X t the a c c u m u l a t e d clainas in the i n t e r v a l [o, t~. T h e m o m e n t g e n e r a t i n g function of X t can be s h o w n to be (S)

Mr(O)

= e x p {lbt [lit

S m, (O)ds0

11}

F o r a d e r i v a t i o n of this f o r m u l a see either B/0HLMANN (1970 , p. 60) or TAYLOR (1979 , p. 153 ) . F o r n = 1, 2, 3 . . . we define (4) so t h a t (5) Y~ = X ~ - X n - 1

Yn

is t h e t o t a l claims in t h e n - t h year. Using (3) it is e a s y to show t h a t

E[Yn~ = p m

~ i(s) ds
#-1

(6)

Var [Yn~ = pcc i i2(s) ds

it-|

where m a n d 0t are the first a n d second m o m e n t s of F o a b o u t the origin respectively. W e denote b y U the insurer's free reserves at t= o. W e m a k e no specific allowance for i n v e s t m e n t income to be a d d e d to t h e insurer's reserves b u t following the r e m a r k s on p. 161 of TAYLOR (1979) we can regard i(t) as il(t)/ i2(t ) where il(t ) is a true index of claims inflation at t i m e t a n d i2(t ) is the a c c u m u l a t e d a m o u n t at t i m e t of a unit s u m i n v e s t e d at t i m e o. TAYLOR (1979) has shown t h a t if the t o t a l p r e m i u m income in ~o, t~ is (7)

C,-- c ~ i(s) ds
0

for some c o n s t a n t c t h e n u l t i m a t e r u m is certain for our risk process p r o v i d e d only t h a t t h e r e exists some c o n s t a n t k such t h a t (8)

Fo(c/k )<

1 a n d i * ( s ) ds~<
0

k*(t) forall

t1> o.

i.e. p r o v i d e d only t h a t sufficiently large claims are possible a n d t h a t the r a t e of inflation is large enough. T h e a b o v e conditions on F o a n d i(s) are not necess a r y for the results of t h e n e x t section.

P R E M I U M RATES U N D E R I N F L A T I O N A R Y C O N D I T I O N S

31

3" PREMIUM RATES

We denote by ct the insurer's instantaneous rate of premium income at time t. We start this section by showing how to determine ct in such a way that the insurer's probability of ultimate ruin can be kept below a predetermined bound. We do this in the following theorem.

Theorem
The insurer's probability of ultimate rum will be bounded above by exp { - RU} if ct is chosen so that (9)
c~ = p E m d R ) - I]/R

for t~> o

where R is any positive number.

Proof
We define the process {Zt} ** o by
$

(lo)

Z~ = f csds- Xe
0

(where cs is as defined in (9)) so that the insurer's reserves at time t are U+ Zt. This process has independent increments so we can use a result of GERBER (1973) which states that the probability of ultimate ruin for such a process is bounded above by (11) min exp {--rU} max
r Oct

E[exp {-rZt}]

But, using (3), we have for any r

t

(I2)

E[exp { - rZt}] = exp {- r f c8 ds) Ms(r)

0

exp{-r

i c, ds+bt[1/t i m,(r) ds-1]}

0 0

= exp { -

i ~ + rcs- pros(r)] ds}

0

By putting r=R in (12) and then using (9) we can see that (13) E[exp { - RZt)] = 1 for all t>~ o.

Our theorem is then a simple consequence of Gerber's result.

Remarks
1. In the special case i(s)= I for all s (i.e. for a standard compound Poisson risk process) the above theorem is nothing more than Lundberg's inequality

32

H. R WATERS

for the p r o b a b i l i t y of ruin since it is clear from (9) t h a t R is the insurer's insolvency constant. W h a t we have done is to e x t e n d this result to the case where the cost of a claim depends on the time at which it occurs and we have achieved this by requiring ct to be calculated m such a w a y t h a t the insurer's " i n s t a n t a n e o u s insolvency c o n s t a n t " at time t is held fixed at some value R > o. 2. An alternative i n t e r p r e t a t i o n of our result is t h a t we have chosen ct in such a way t h a t the process ( e x p ( - RZt) } t ~ o is a martingale. See GERBER (t 975). 3. The a s s u m p t m n t h a t m0(0 ) is finite for all 0 imphes t h a t m~(R) and hence c~ will be finite for a n y values of R and t. Suppose t h a t only the weaker condition mr(0) < oo for all o~< 0~< 0o and o~<t~<t o

holds where 00, to> o. We can then show t h a t the p r o b a b i h t y of rum before time t o is b o u n d e d above b y exp { - R U } p r o v i d e d R < 0o and provided ct is calculated as m (9) for o~< t~< l 0. The proof is as before except t h a t it reqmres the finite-time version of Gerber's result. See p. 207 of GERBER (1973). F o r m u l a (9) gives little indication of the way in which we require ct to change relative to i(t). We t r y to provide th_ls, at least for small values of l, in the following corollary.

Corollary I
The rate of p r e m i u m income ct speclhed b y (9) gives the following approximation for small values of t. (15) where X = ct =

Coi(l)

El+ x,(t)]/El+

Ro~/2m.

Proof
F o r m u l a (9) gives
(16) ct = Co[mr(R)-

Ill[too(R)- 1]

and we have (17)

mr(R)-1=

Ri(t)m+ c~(R,(t) )2 / 2+ Z (Ri(t) )Jo~j/7 .t

where ~j is the j - t h m o m e n t of F 0 a b o u t the origin. B y assumption, (17) is a convergent series. In practice R is hkely to be small so t h a t if i(t) is not large the first two terms on the right h a n d side of (18) should give a reasonable a p p r o x i m a t i o n to lint(R)- 1]. Making a similar a p p r o x i m a t i o n to Em0(R)- 1~ and p u t t i n g these two a p p r o x i m a t i o n s into (16) we obtain (15).

PREMIUM RATES U N D E R I N F L A T I O N A R Y C O N D I T I O N S

33

Remarks
a. It is interesting to compare (15) with (7). 2. It can be easily checked t h a t if i(l)>~ 1 and m, o~, =j~> o for j>~ 3 then the "-4-" sign in (15) can be changed to a ">~" sign. 3. The range of values of t for wlnch (15) is likely to be a reasonable approximation ~s not ammcdiately clear since it depends on the relationships between R, ,(l) and F o. However, in the special case where F 0 is a negative exponential distribution we can get a clearer idea of the accuracy of (15). Let us suppose then t h a t Fo(X) = I - e x p { - a x } for some a > o so t h a t m= 1/a and ~= 2/a =. F o r m u l a (9) gives (18) ct = #i(t)/[a- R,(t)] provided a > Ri(t).

Tins gives the exact relationship (19) c, = Coi(t)Et- R/a~/[1- Rz(t)/a]

F o r m u l a (15) gives the following approxnnation (20)

c, -- Coi(t)[l+ Ri(t)/a]/[l+ R/a 3

So far in tlns sectmn we have been concerned with ct, the instantaneous rate of premium income, which we have assumed to be continuously variable. We now suppose t h a t in the time interval I n - I, n], where n is a positive integer, a total premium Pn is payable at a constant rate t h r o u g h o u t the year.

Corollary 2
Assuming either t h a t i(t) is a non-decreasing function of t or t h a t " r u m " can only occur after an integral number of years, the probability of ultimate rmn for our risk process will be bounded above by exp { - R U } if P n is calculated b y the formtfla
)t

(21)

Pn=

P [ J" m d R ) a t - 1 ] / R

n=

1,2 . . . .

This formula for Pn gives the following approximation for small values of n (22)

Pn -- E[Yn] + (R/2) Var [Y,~]

Proof The first part of the corollary as easily proved since (21) can be written
(23)

Pn = S c~ dt

where ct is as in (9). Using similar approximations to those used in the proof of

34 Corollary 1 we have t h a t (24)

H.R. WATERS

P~ ~ P i [mi(t)+ o~Ri2 (t)/2] dt

n-I

and so (22) follows from (5) and (6).

Remarks

i(t)~> 1 and ~j>~ o for j>~3 the " ~ - " sign in (22) can be replaced b y a ">~" sign. 2. F o r m u l a (21) is equivalent to
I. If (25) P n = (I/R) In [E[ exp {RYn}]] In other words to keep the probability of ruin below e x p { - R U } t h e annual p r e m i u m should be calculated using the principle of zero utility with the utility function (26) See GERBER (1974).
REFERENCES

u(x) = (l/R) ( I - exp { - Rx})

~BUHLMANN, H. (t970). Mathenlattcal Methods in lhsk Theory, Springer, New York.

GERBER, H. U (t973). Martingales in Risk Theory, M~tte,lungen der Vere,mgung Schwe*zer~scher Verszcherungsmathemahker, 73, 205-216. G E R B E R , H U. (1974) On Additive Premium Calculation Principles, Ast,n Bullehn, 7, 215-222 GERBER, H U. (1975) The Surplus Process as a Fair Game--Utihtywise, Astzn Bullehn, 8, 307-322 TAYLOR, G. C. (1979). Probablhty of Ruin under Inflationary Condttlons or under Experience Rating, Astzn Bullet,n, 10, 149-162