Insurance: Mathematics and Economics 5 (1986) 103-112 103

North-Holland

Risk theorv and serendi&v J I r/

Karl BORCH working on risk theory. They may have come up

Norwegian School of Busrnexr. 5035 Bergen, Nornyv with few good answers, but they have certainly
formulated many hard questions. Some of these
questions are such that it seems inconceivable that
The paper gives a short presentation of a model which
they could be asked by anybody from the in-
determines equilibrium premiums in an insurance market, and
a series of examples and potential applications. The paper also surance world, but they fascinate an esoteric group
reviews De Finetti’s model of optimal dividend policies. and just because they are hard.
argues that this model gives the utility functions required by
the equilibrium model. 1.3. In the actuarial risk theory one usually studies
risks completely detached from any economic con-
Keywurds: Market equilibrium, Risk aversion. Expected text. It is then only to be expected that much of
utility, Multi-period models
the work seems to lack any clear purpose. In this
paper we shall try to connect the risk problems
with economic reality, and thus give the theory a
1. Introduction
purpose.
The next section gives a brief presentation of
1.1. De Finetti (1959) was the first to use the word
the standard model of general economic equi-
serendipity to characterize certain statistical activi-
librium, in a form suitable for the study of in-
ties. The word was coined by Horace Walpole in
surance. Section 3 gives some examples of how the
1754 in some comments on the fairy tale ‘The
model can be applied and indicates how the results
Three Princes of Serendip’. One of the ancient
may be useful in practice. The following section
names of Sri Lanka is Serendip, and its three
presents a model due to De Finetti, which derives
princes set out on a series of travels with very
the utility functions essential in the equilibrium
definite purposes in mind. For various reasons the
model, from a dynamic model. Section 5 gives a
princes never achieved what they set out to do, but
short discussion of the practical application of the
by good luck or accident they made important
different models.
discoveries and did many useful things.
The parallels between the fairy tale and risk
theory in the twentieth century are obvious. The 2. A model of the insurance market
theory was developed to solve practical problems
in insurance. Nobody can seriously claim that the 2.1. An insurance contract is a contingent claim.
actuaries who developed the theory were success- When the contract is concluded an amount of
ful in this respect, but by pure serendipity they cash, the premium is paid and the buyer will get a
made discoveries which turned out to be useful in random amount in return - as settlement of the
many unexpected ways. For instance, there is little claims he can make under the contract. The trans-
doubt that the collective risk theory has brought action at this abstract level is of exactly the same
modern mathematics into actuarial curricula, and form as the purchase of a share in a risky business.
this is useful, even if the theory is not. The price of such shares is presumably determined
by supply and demand in stock markets, and it is
1.2. One can look at the subject less kindly than natural to assume that insurance premiums are
De Finetti does. In a survey of contemporary determined in a similar way. In the following we
mathematics Mac Lane expresses some fear that shall therefore give a brief outline of the standard
one may tend to take ‘too much of a Hungarian equilibrium model, essentially due to Arrow (1953).
view of mathematics - that the science consists not
in good answers but in hard questions’. This re- 2.2. It is convenient to interpret the model as a
mark could well have been addressed to actuaries reinsurance market, in which n insurance compa-

0167-6687/86/$3.50 C 1986, Elsevier Science Publishers B.V. (North-Holland)

104

nies trade among themselves. We shall take as (i) the market is complete, in the sense that it
given (for Y = 1,2,. . . , n), will assign a unique value P{ y } to an arbitrary
(i) the risk attitude of company r, represented final portfolio. described by a stochastic variable
by the Bernoulli utility function u,.( .), with the I?
properties U: > 0 and u:’ < 0, (ii) the value operator P{ .} is a linear func-
(ii) the initial portfolio of company r, repre- tional of y(x).
sented by the stochastic variable x,. From the Riesz representation theorem it then
In the market these n companies exchange parts follows that there exists a function G(x) such that
of their initial portfolios among themselves. As a
result of these exchanges company r obtains a
PC y } = (+“y(x)dG(x).
J-CC
final portfolio. represented by the stochastic varia-
Any actuary worth his salt believes that he can
ble ,Y~, r=l,2 ,..., n.
compute the premium for any risk with known
If the companies cannot trade with outsiders,
stochastic properties, so to him the first assump-
we have
tion will be trivial. Some economists, i.a. Hirshleifer
n n
(1970) seem to doubt that markets in the real
c y,= c x, = x,
world are complete, and the study of ‘incomplete
r=l r=l
markets’ has become fashionable. These studies
where x is the sum of the stochastic variables are however not relevant in, the present context,
representing the initial portfolios. since no restrictions are placed on the exchange
At this stage it is necessary to make an assump- arrangements which the companies are allowed to
tion of homogeneous beliefs, i.e., that all companies make.
hold the same opinion on the joint density If a derivative of G(x) exists, we can write,
f(x ,,..., x,,), and hence on f(x). without loss of generality, G’(x) = V’(x)f(x), so
that the formula above takes the form
2.3. Any set of exchanges which satisfies (1) is
feasible. The subset of Pareto optimal exchanges is P{ Y > =j+=y(xP”(x)f(x) dx
determined by (1) and the conditions
= E;;,(x)V’(x)J. (4)
k.~:(y~)=k,ut.(y,). r, .~=1,2 . . . . . n, (2)
In (4) y represents a Pareto optimal portfolio,
where k, and k, are arbitrary positive constants. and hence it is a function y(x) of the variable x.
The result has been proved explicitly by Borch A non-optimal initial portfolio is described by the
(1962) but is really contained in earlier work by variable x,., which is stochastically dependent on
Arrow. x. Analogy with (4) suggests that we can write
It is easy to see that (1) and (2) will determine
the Pareto optimal exchanges as n functions P{xr) = E{xY’(x)}
J,,(X), . . , y,?(x) of the stochastic variable x. These = x, x,) dx dx,.
+%+)f( (5)
functions will contain the arbitrary constants / -YZ
k,, . , k, as parameters, and yr(x) must be inter- Clearly (4) is a special case of (5).
preted as payoff to company r, if total payoff is x.
It is easy to see that (2) can be written in the
2.5. To find the competitive equilibrium in the
form
conventional way, we note that the problem of
Q&,(x)) = u’(x), (3) company r is

where the function U’(X) in a sense represents maxE{ti,.(y,)}, r=1,2 ,..., n,

aggregate marginal utility in the market. It will of subject to
course depend on the parameters k,, . , k,.
P{y,}=P{x,.}, r=1,2 ,..., n. (6)

2.4. To determine the competitive equilibrium of The conditions (6) say simply that the market
the market in the usual way, we need two assump- value of a portfolio must remain constant if all
tions: exchange transactions are settled at market prices.
 K. Bonh / Serendipity 105

The problem is equivalent to The condition U: > 0 is satisfied only when

x < c,., so the analysis is valid only in this domain.
max lt”{ u,.(y,.(x)) +x,.(E{x,.V’(x)} If (Y is an odd integer we can take u:.(x) = (c,. -
--3o
x)“. We shall use this simpler notation, and fall
-~,(+“(x))}f(x) dx, back on the complete notation only when it is
which can be seen as a problem in the calculus of necessary to avoid misunderstanding.
variation. The Euler equation is In the example, (3) takes the form

4(.&(x)) = X,0x), (7) k,(c, -y,)” = u’(x), (3’)

and from (3) it follows that this can be written as which can be written

u’(x) = X,k,V’(x). c,.-y,.= (k;‘u’(x))?

There is no loss of generality if we take X, = k,: ’ Summing these equations over all r, and using
and write v’(x) = u’(x). It then follows that the (1) we obtain
assumption that the function G(x) is differentia-
ble is equivalent to assuming that the aggregate i cr-x=(u’(x))“” 5 k;“,.
utility function is differentiable. r=l r=l

We can now sum up the results.

Here we write
Conditions (1) and (3) determine the form of
the functions y,(x).
The equations (6) determine the parameters
2 c,.=c and i k;‘,‘n = K-l/“,
i-=1 r=l
k,, . . , k,, in the functions y,(x).
so that the equation takes the form
For the market value of an arbitrary portfolio
we find u’(x) = K(c - x)?
P{ xr} = E{ X$‘(X)}. (8) From (3’) we then find

2.6. We so far discussed only market values of k,(c,-y,(x))“=K(c-x)“,

portfolios. To obtain market premiums for in- and the explicit expression for y,(x),
surance contracts, we note that the initial portfolio
of company Y consists of assets R,, and of liabili- y,.(x)=c,-(c-x)(Kk,‘)l’a
ties under the insurance contracts held by the = c, -(c - X)h,,
company. Let the non-negative stochastic variable
z, represent claim payments under the contracts, where h,. = ( Kk;l)l/a, and Ch,. = 1.
and write The equations (6) then take the form

x, = R, - z,. KE{(c,-(c-x)h,)(c-x)a}

If for the sake of simplicity we assume that the =KE{x,(c-x)~}, (6’)

assets are risk-free, we have
and the parameters are determined by the equa-
P{xr} = R,-P{z,}. (9) tions
Formula (9) says simply that the market value
of the company’s portfolio is equal to risk-free
h = Ek--xJc-4)” r=l 2

,,...,n
r 1

assets. less the market premium for insurance of E{(c-x)~+‘} ’

the liabilities.
3.2. The right-hand side of (6’) gives the market
value of an arbitrary portfolio, so we can write
3. Examples and applications P{xr} = KE{x,(c-x)~}. (10)

3.1. As a simple example we shall take The special case (Y= 1 has led to models which
are widely used in practical financial analysis, so it
u:(x)=sgn(c,-x)l(c,-x)a(, r=l,...,n. may be useful to discuss this case in some detail.
106 K. Borch / Serendipity

For (Y= 1 the equation (10) takes the form statement has led to a number of hard questions
which must delight any actuary who takes the
P{xl} =KcE{x,} -KE{xx,}
‘Hungarian’ view. In their efforts to answer
=K(<,-E{x})E{x,.}-Kcovxx,.. Cramer’s questions these actuaries have by seren-
If we here take x, = x, we obtain in the same dipity made contributions of some importance to
way the development of probability theory. It is how-
ever doubtful if their work has had any noticeable
P{x}=K(c-E{x})E{x}-Kvarx. effect on insurance practice.
Combining these two equations, we obtain Economists working on finance seem to have
taken the opposite attitude. They have found an
P{x,.} =K/IE{x,.} +(P{x} -I(AE{x})z easy answer in CAPM, which may not be a par-
ticularly good one, and put it to application. One
can discuss how useful this model really has been
where c- E(x) =A. in practice. It is however certain that the applica-
It is worth noting that the arbitrary constant K tion of CAPM has, possibly by serendipity, led to
can be interpreted as a risk-free rate of interest in deeper insight into the functioning of financial
the market. markets.
Formula (11) is the well known ‘Capital Asset
3.4. Let us now return to the more general version
Price Model’ or CAPM, due to Mossin (1966) and
of CAPM given by (10). From (9) it follows that
others. The model has found widespread applica-
we have
tions in practice, and is discussed in virtually every
textbook in finance - in spite of its obvious P{z,.}=R,.-P{x,.}=R,-KE{x,.(~-x)~}.
shortcomings.
Here we substitute
We derived (11) by assuming that marginal
utilities were linear, i.e., that all utility functions x,.= R,.-z,. and x= R-z,
were quadratic. This assumption seems to be too and find the following expression for the pre-
strong for most economists. The equivalent as- mium:
sumption, implied by (ll), that values are de-
termined by the two first moments of a stochastic P{z,}=R,.{l-KE{(c-R+z)~}
variable, seems however to be acceptable. A multi-
+KE{z,(c-R-tz)“},
variate normal distribution is completely described
by means and covariances. If the variables z,, where R = CR, and z = Cz,.
zz,...>z,,, are normally distributed, the variable For the degenerate case z,. = 0, consistency re-
z = t,z, + . . , t,,lZ,,, defining a portfolio, is also quires that P{ z,.} = 0, so that we must have
normally distributed. This seems to be the current
K~‘=E{(c-R+z)‘~}.
justification for the practical application of CAPM.
The normal is the only distribution with finite Hence the premium formula takes the form
variance which has the ‘stability’ property sketched
P{z,)=KE{z,(c-R+L)~}
above. This means that strictly speaking CAPM
can assign a value only to portfolios defined by = E{z,(c-R+z)~}
normally distributed variables, and that may be
E{(c- R+z)~} ’
the explanation of the interest in ‘incomplete’
markets. or, written in full,

f{~~)=KJo~~~z’(~-R+‘)“f(z; z,)dzdz,..
3.3. Actuaries seem to consider CAPM as too
primitive for their purposes. They may find some
support for this view in a statement by Cramer (12)
(1930) who wrote: ‘ . .in many cases the ap- Formula (12) shows how the premium for a risk
proximation obtained by using the normal func- which can lead to losses represented by the sto-
tion is not sufficiently good to justify the conclu- chastic variable z,., depends on
sions that have been drawn in this way’. The (i) the stochastic properties of the risk itself,
 107

(ii) the stochastic relationship between the par- where

ticular risk and claims in the market as a whole,
x4,-=1 and cc,.=O.
described by the joint density f( z, z?),
(iii) the attitude to risk in the market as a The sharing rules have this form only if all
whole, represented by c, utility functions belong to one of the following two
(iv) the total assets of all insurance companies classes:
in the market, R.
A realistic theory of insurance premiums must (i) U:.(X) = (x + L.,.)“, or
of course take all these four elements into account.
(ii) U:(X) = ee”,‘.
This is however rarely done in actuarial risk the-
ory. The recent book by Goovaerts et al. (1983) The result is well known, and has been proved
covers only the first of the four elements. by many authors, i.a. by Borch (1968).
One could consider as a fifth element the inter- If the sharing rules are linear, it is possible for
est earned on the premium before claims are paid. the companies to reach a Pareto optimum through
This is regularly done in financial theory, but is proportional reinsurance arrangements. Other
rather trivial in the present context. One has just forms of reinsurance are widely used, so there
to multiply the premium by the appropriate dis- should be a need for studying utility functions
count factor. which do not lead to linear sharing rules. The
problem is difficult, and it is surprising that it has
3.5. As another simple example we shall take not been taken up by those who cherish ‘hard
questions’ for their own sake. We shall not discuss
Ui(x)=exp -$ , r=1,2 ,..., n.
1 i- 1 the problem in any detail, but it may be useful to
study a simple example.
Equation (3) then takes the form
Consider two persons with the utility functions
k,. exp{ -y,.tiJ’} = u’(x).
u,(x)=1 --e-l, and u,(x)=log(c+x).
Taking logarithms on both sides, we obtain
Condition (2) then takes the form
log k,-y,.:,a;’ = log u’(x).
Multiplication by (Y, and summation over all Y kle-“i = k,,
gives c+.Yz
&,logk,-x=logu’(x)Ca,. As y, + y1 = x, we can write this as
Writing (c+yz)e~“-‘2)=k,

5 a\- log k,s = K and i (Y,= A, Clearly we must have yZ > -c. Hence under a
.s=1 \=I
Pareto optimal rule of loss-sharing, person 2 can
we obtain under no circumstances pay more than c, so that
the rule will correspond approximately to a stop-
U’(x)=exp{-71
loss insurance arrangement.
and
(YK 3.7. Let us now return to the general case (3), and
L;.(x)=y+a,logk,-*. consider a risk described by the stochastic variable
‘.
We can then proceed as in the first example.
From (8) it follows that the insurance premium
From (8) we determine the value of an arbitrary
for this risk is
portfolio in the market, and hence also the market
premium for an arbitrary insurance contract. P{ z} =~r~zz~‘(x)/(x, z)dxdz. (13)

3.6. The two preceding examples are fairly simple

Here we have written x for total claims in the
because the ‘sharing rules’ are linear, i.e.,
market, and we have not brought the assets ex-
Yr(X) = qrx + c,, plicitly into the utility function.
108 K. Borch / Srrmdqqy

It is convenient to write (13) in the form 4. A dynamic approach to utility

P{ z) d+.
z} =~ru’(x){~mzf(x, 4.1. In the preceding sections we have studied
simple one-period decision problems, and assumed
Let us now write z = z, + z2, where z, = min(z, there was a utility behind the decision. In practice
D) and z2 = max(O, z - D). it may often be difficult to specify this utility
For the two components we find the following function - for good reasons. The utility a com-
premiums: pany assigns to the profits earned in one period

P{ z,} =~mu’(x,[s:)zt(~.
z) dz
will presumably depend on how that profit can be
applied in future periods. This suggests that the
0
problem should be studied in a dynamic or multi-
+D 6/(x, z) dz]dx, period framework. If the company has an overall,
s
long-term objective, this must contain the utility
P{z~}=~?(x)[/~~(z-I,)/(x; z)dz]dx. functions which govern the decisions in each
period. This overall objective may often be simpler
These two formulae can be given a number of to specify than the Bernoulli utility function.
different interpretations, i.a., One of the first to study this problem in an
- P{ z1 } can be seen as the reduction in the insurance context was De Finetti (1957) in one of
premium offered to a buyer, if he will accept a his pioneering papers. As a long-term objective he
deductible D. suggested that an insurance company should seek
_ P{ z2 } can be seen as the reinsurance premium to maximize the expected present value of its
for a stop-loss contract, under which the reinsurer dividend payments. In the following decades this
pays all claims in excess of the limit D. problem has been discussed by several other
We now write P{ z2} = P{ O}, and differentiate authors, i.a. by Borch (1967, 1969) and Biihlmann
twice with respect to D. This gives (1970).
De Finetti made his observation in a critical
study of the now obsolete ‘collective risk theory’.
This theory placed the focus on the probability
P”(D) = ~=?A’(
x)f( x, D)dx. (14) that an insurance company shall remain solvent
forever, provided that it does not change its oper-
If a reinsurer is willing to quote premiums for a ating procedures. This probability of achieving
number of stop-loss contracts, with different limits, eternal life will inevitably be zero, unless the com-
the left-hand side of (14) can be estimated for an pany allows its reserves to grow without limit. De
arbitrary D. If the joint density f(x, z) is known, Finetti pointed out that an insurance company
at least approximately, it may be possible to ob- could not really be interested in building up un-
tain an estimate of the kernel function u’(x). With limited reserves, and argued in fact that the margi-
this estimate, we can then compute the premium nal utility of accumulated profits (uincite utili)
for an arbitrary insurance contract in the market. must be decreasing. He illustrates his argument
It is obviously impossible to determine the with a very simple example, based on a two-point
function u’(x) from (3) i.e., from the preferences distribution. In the following we shall consider a
of all companies, and then use (8) to calculate the simple, slightly more general example.
market premium for an arbitrary insurance con-
tract. 4.2. Consider an insurance company which in each
It is however clear that u’(x) plays an im- successive operating period underwrites identical
portant part in determining market premiums. portfolios, and take the following elements as
Since premiums are observable, we can turn the given:
problem around, and estimate the function U’(X) S = the company’s initial capital,
from observations. This approach is widely used in P = the premium received by the company at the
economics. Demand curves and underlying utility beginning of each operating period,
functions cannot be observed, but must be esti- f(x) = the probability density of claims paid by
mated from observed behaviour. the company in each operating period.
 K. Borch / Sermdrprtj 109

If S, is the company’s capital at the end of view on mathematics, a general solution appears
period t, and x,, , the claims paid by the com- uninteresting. The optimal dividend policy is given
pany during period t + 1, the company’s capital at by the value of Z, which maximizes V(S, Z). The
the end of period t + 1 will be existence of an optimal policy poses some hard
questions, which we shall not take up.
s r+l =$+P-x,+1.
Assume now that the company operates under 4.4. The main idea behind De Finetti’s paper can
the following conditions: be presented as follows.
(i) if S, < 0, the company is insolvent, or Assume an insurance company with equity
‘ruined’, and cannot operate in any of the follow- capital S is offered a premium Q, if it will under-
ing periods, write a one-period insurance contract described by
(ii) if S, > Z, the company pays a dividend the claim density g(x). If the company accepts the
s, = S, - Z. It is natural to assume that Z is cho- contract, expected present value of future dividend
sen by the management, because accumulated payments will be
profits beyond Z have lower utility than a paid
out dividend. u(s) = ptaV(S+Q-x)g(x)dx.
JO
The dividend payments s,, s2,. ., s,. . is a
sequence of stochastic variables, and we shall write Hence, given the company’s long-term objec-
V( S, Z), or when no misunderstanding is possible tive. the contract will be acceptable if and only if
V(S) for the expected discounted sum of the U(S) > V(S). This means that V(S) serves as the
dividend payments which the company makes be- utility function which determines the company’s
fore the inevitable ruin, i.e., decision, and that in (8) u’(x) can be replaced by
v’(x). This assumes of course that the company
I’(S)= E u’E{s,}, receives the offer after it has made its reinsurance
r=l arrangements, or that it takes the reinsurance pos-
where u is a discount factor. sibilities into account when deciding about the
particular offer.
4.3. To study the function V(S), we first note that De Finetti did not elaborate his suggestion, but
for S < Z < S + P it must satisfy the integral went in stead on to discuss other functions of an
equation arbitrary upper limit to accumulated profits, which
could serve as alternative objectives. His paper was
I’(S)=u/‘+‘{x-Z+ V(Z)}f(S+P-x)dx generally ignored for a decade, even in actuarial
z
circles. The influential paper on the subject is by
+u %‘(x)f(s+ P-x) dx. (15) Shubnik and Thompson (1959), who indepen-
/0 dently developed a model which is virtually identi-
The equation says that if claims x, are less than cal to that of De Finetti, but not interpreted in
S + P - Z, a dividend S + P - Z - x will be paid terms of insurance.
at the end of the period, and the company will
begin the next period with a capital Z. 4.5. De Finetti’s model implies that an insurance
If S + P - Z < x < S + P the company will pay company is basically risk-neutral. This may at first
no dividend, and begin the next period with a sight seem surprising, but it makes some sense if
capital S + P - x. If S + P < x, the company is one considers the alternatives. It seems artifical to
ruined, and cannot operate in any of the following assume that the board of a company will consider
periods. different utility functions, and possibly take a
In the interval stated equation (15) is an in- vote, to pick one which adequately represents the
tegral equation of Fredholm’s type, and it is known company’s attitude to risk. It seems more natural
that it has a unique continuous solution. For S c to assume that a board will settle for some simpler
Z - P, no dividend can be paid at the first period, rule, such as that of maximizing expected present
and some modifications are necessary. value of dividend payment. With this rule the
On the whole, equation (15) is difficult to han- company will behave as if it were risk-averse in its
dle, and unless one holds a strong ‘Hungarian’ underwriting and reinsurance transactions. This
does, however, throw some doubt on the relevance or
of the model in Section 2, but the model may have
W(Z)=cjo’+“(P-x)j(x)dx
been useful by serendipity, i.a. by providing a
simple generalization of CAPM.
At this stage we can see dimly the outline of an
/[l -vF(P+Z)] (16)
attractive unified theory of insurance:
4.7. The problem of the company is now to de-
(i) premiums and reinsurance arrangements are
termine the value of Z which maximizes (16), i.e.,
determined by the methods sketched in Section 2,
to find the optimal amount of capital which the
(ii) the utility functions which play an im-
owners should put at risk in their company.
portant role in Section 2, can be determined by the
The first-order condition, W’(Z) = 0, takes the
methods indicated in this section.
form
A number of hard questions must be answered
before a complete theory can take form. Before Z(1 -uF(P+Z)}=~Sol’+z(P-x)f(x)dx.
taking up the challenge, it may however be useful
to pause, and ask if it really is necessary to answer (17)
these questions.
From (17) Z can be determined. Comparison of
(16) and (17) shows that for the optimal Z we
4.6. The model leading to (15) specified when a have
dividend should be paid, but not when the com-
z= W(Z).
pany’s equity capital should be strengthened. Both
questions are of equal importance to management, This simple condition expresses the obvious,
and at least if the insurance company is owned by that the optimal capital to put into a venture is
a holding company, they will be considered to- equal to the expected present value of the return.
gether. Hence let us assume that the insurance In the discount factor u = (1 + i)-’ which occurs
company initially holds an equity capital Z, and in (16) and (17) i must be interpreted as the
consider the following policy: return on competing investments. This invites some
If claims in a period amounts to x, conventional comparative static analysis.
(i) the company pays out a dividend max( P - Integration of the right-hand side of (17) shows
x, 0) that the equation can be written in the form
(ii) an amount min(x - P, Z) is paid into the
company as new equity capital.
This policy implies that if the company is solvent
z =u
J‘+zF(x)
0
dx. (17’)

at the end of a period, it will enter the next period From (17’) we find
with an equity capital Z.
Expected present value of the dividend pay-
ment at the end of the first period is
This again expresses the obvious. If return on
u
/0
P+Z(P-x)f(x)
dx,
competing investments decreases, i.e., if u in-
creases, more capital will flow into the insurance
industry. In the opposite case insurance companies
and the probability that the company shall be able
may be unable to attract new equity capital.
to operate in the second period is
Clearly F( P + Z) is the probability that an
Pr(x<P+Z)=F(P+Z). insurance company shall be able to pay its claims.
In most countries government regulations require
It then follows that the expected present value
that this probability shall be close to unity. In
of all payments is
practice this means that an insurance company is
only allowed to operate if it satisfies a solvency
condition of the form F( P + Z) > a. If there is a
general increase in returns on investments, it is
x c [UF(P-tZ)l”, evident that insurance companies must increase
n=O their premiums, in order to attract the capital
 K. Borch / Serendipr!v 111

necessary to satisfy the solvency condition. This without a costly checking of the statistics. The
may be a useful result. It shows that the govern- possibility that the ceding company cheats is al-
ment cannot decree both high solvency and ‘rea- ways there, so the reinsurer has the choice of either
sonable’ premiums in a free capital market. The expenses, or of being exposed to moral hazard. It
result should be obvious, but does not always seem is therefore natural that reinsurers should try to
to be so. cover a medium sized risk within a fairly small
group, and this may lead to some segmentation of
the market.
5. The model on the real world One way to model this would be to assume that
there are costs c involved in checking the portfolio
5.1. The model presented in Section 2 was inter- offered by a would-be ceding company. If n com-
preted as a reinsurance market, and we determined panies participate in an exchange arrangement,
the equilibrium premiums. One must expect that each checking the other, total costs will be n(n -
these premiums, at least in the longer run, will 1)~. With increasing n there will clearly be a point
have some influence on the premiums in direct where increasing costs will offset the gain by wider
insurance. diversification of the risk.
Insurance markets are far from perfect, and a
5.3. In Section 4 we have in a sense banished
company may well ~ for some time ~ be able to
utility and risk aversion from the supply side of
charge premiums above those determined by the
the insurance market. There are however some
model. This company can however earn a risk-free
reasons for assuming that insurance companies
profit if it reinsures its whole portfolio in the
should be risk-neutral. Insurance companies are
market, so the situation cannot be stable in the
intermediaries between those who want to buy
long run. One must also expect that the customers
insurance, and those who are willing to risk their
of this company eventually will discover that they
money in the underwriting of insurance contracts.
can buy their insurance cheaper elsewhere.
If these two parties are risk-averse, a set of Pareto
Similarly an insurance company may charge
optimal arrangements exists. The parties may how-
lower premiums than those determined by the
ever be unable to reach a Pareto optimum, if the
model. This company will probably find that rein-
surance costs more than it is worth, and carry transactions have to be carried out through an
most of the risk itself. With good luck this may go intermediary that imposes its own risk aversion.
on for quite some time, and with ample reserves The problem has been discussed in economic liter-
the company will be able to satisfy the solvency ature, i.a. by Malinvaud (1972). We shall not
conditions laid down by the government. When discuss it further here, since a complete discussion
luck runs out, as it eventually must, the company will require detailed specification of the institu-
may well find itself insolvent, when its liabilities tional framework.
are valued at market prices. On the demand side risk aversion remains es-
sential, since a buyer of insurance is by definition
risk-averse. When he is faced by a market which
5.2. In Section 2 it was proved that a Pareto
can quote a premium for any kind of insurance
optimum could be reached only if every insurer
contract, he can compute or program his way to
participated in every risk in the market. This does
the contract which is optimal according to his
not happen in practice, and it is natural to seek the
preferences.
explanation in transaction costs. This is however
not entirely satisfactory, since it should not be
prohibitively expensive to carry out the calcula- References
tions necessary to divide a risk in a Pareto optimal
way. It seems more likely that some elements of Arrow, K.J. (1953). Le rBle des valeurs boursi&es pour la
moral hazard enter. It is always stressed that rein- &partition la meilleure des risques, CoNoques fnrer-
nationuux du CNRS 40, 41-48. Translated in 1964 as: The
surance is carried out under conditions of the
role of securities in the optimal allocation of risk-bearing,
‘utmost good faith’. In practice this means that a Review of Economrc Studies 31, 91-96.
reinsurer accepts the ceding company’s report on Borch, K. (1962). Equilibrium in a reinsurance market,
the stochastic properties of the risk covered, Econometrica 30, 424-444.
112 K. Aorch / Serendipi?)

Borch, K. (1967). The theory of risk, Journal oJ the Royal De Finetti, B. (1972). Prohahrlrt,v, Induction and Stcrtrstrcs.
Statrstical Society 29, Series B, 432-452. Wiley, New York.
Borch, K. (1968). General equilibrium in the economics of Goovaerts, M., F. De Vylder and J. Haezendonck (1983).
uncertainty, in: K. Borch and J. Mossin, eds., Ruk and Insurcrnce Premiums. North-Holland, Amsterdam.
Uncertainty. Macmillan, London/New York. Hirshleifer, J. (1970). Inuestment, Interest and Coprtul.
Borch, K. (1969). The static decision problem in its dynamic Prentice-Hall, Englewood Cliffs, NJ.
context, Colloques Internationaux du CNRS. no. 171, 53-62. Mac Lane, S. (1983). The health of mathematics, The Mathe-
Biihlmann, H. (1970). Mathematical Methods in Risk Theoty. mclriccrl Intelligencer 5, no. 4. 53-55.
Springer-Verlag, Berlin/Heidelberg. Malinvaud, E. (1972). The allocation of individual risks in large
Cram&, H. (1930). On the mathematical theory of risk, Skandia markets, Journal of Economic Theorv 4. 312-328.
Jubilee Volume, 7-84. Mossin, J. (1966). Equilibrium in a capital asset market,
De Finetti, B. (1957). Su un’impostazione altenativa della teoria Econometrica 34, 768-783.
colletiva de1 rischio, Transactions of the XV International Shubik, M. and G. Thompson (1959). Game of economic
Congress ofActuaries2, 433-443. survival, Naval Research Logistics Quarterly 6, 1 1 I- 123.
De Finetti, B. (1959). La probabiliti e la statistica nei rapporti
con l’induzione. Centro Internazionale Matematico Estiao,
Cremonese. English translation as Chapter 9 in De Finetti
(1972).