You are on page 1of 4

Economics of Insurance

Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E.

RISK SPREADING: THE ARROW-LIND THEOREM

Here we investigate the effect on the premium of sharing the insured risk among an increasing number of insurers. We show that, as the number of risk-averse insurers sharing the risk increases, the premium paid by the insured will tend to the expected cost of the accident or event insured against, i.e. to the actuarially fair premium.

MODEL:

Let there are n members in a syndicate.

They are willing to pay to acquire a risky project for which they will equally share the gains and losses.

We assume that the syndicate member have identical preferences, probability beliefs and income. The income

of a representative member of the syndicate in state s is:

y s : original state-contingent income

Z s : payyoff from the risky prospect

y s +kZ s -p

------ (1)

k=1/n: fraction of the prospect received by each individual of the syndicate.

p:amount paid by each member for the right to share in the risky prospect.

& P = np is the amount that syndicate pays for the prospect and shares the cost equally amont the members.

We consider y s and Z s are statistically independent:

cov(y s , Z s )=0

----------(2)

The maximum price (p) each member of the syndicate would be willing to pay for her share that satisfies:

E [v(y s +kZ s -p)] = E[v(y s )]

-------------(3)

The expected value of the individual share in the prospect is: kE[Z s ].

Therefore, risk loading is:

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. RISK SPREADING:

l = kE[Z s ] – p

------------ (4)

(Excess of expected return over price).

Now, the prospect payoff in state s can be written as:

Z s = E(Z s ) +e s

-------- (5)

Where e s : random deviation with E(e s ) =0

Therefore, (3) can be written as:

E [v(y s +kZ s -p)] = E[v(y s +kZ s - kE(Z s )+l)] = E[v(y s +ke s +l)] = E[v(y s )]

------- (6)

Equation (6) bet which pays :

Equation (6) bet which pays :

the risk loading is the compensation required to make the individual accept the risky but fair ke s = kZ s - kE(Z s ) in state s.

1 | P a g e

Economics of Insurance

Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E.

Therefore, the risk loading depends upon the share of the risky prospect borne by each syndicate member:

l = l(k) = l(1/n)

------- (7)

Now, the maximum amount that the syndicate would pay for the risky prospect is:

P = np = n[kE(Z s )-l(k)] = nkE(Z s )-nl(k) = n.kE(Z s )-n.l(1/n)

----------- (8)

If n.l(1/n) risky prospect.

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. Therefore, the

0 as n becomes large: a large syndicate will be willing to pay the expected value of the

Using the Implicit Function Theorem 1 we can find the effect of the member’s share k on her risk loading from (6). Since R.H.S is independent of k, we can write:

. ,

,

---- (9)

Since, Z s =E(Z s ) +e s is not correlated with y s , e s is also uncorrelated with y s .hence the only sourse of correlation b/w (MU of income) and e s that as e s increases decreases 2 , i.e., , 0.

As n increases, k=1/n decreases

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. Therefore, the

l(k) decreases.

i.e,

/

------- (10)

Hence we can make the following proposition:

Total risk loading tends towards zero as n becomes large: . /

----- (11)

Since both the numerator and the denominator of l(k)/k goes to zero, we use L’Hôspital’s Rule

/

/

,

,

Hence (11) holds & so large syndicates of risk averse members will be willing to contract as though they are all risk-neutral.

The total risk loading n.l(1/n) tends to zero as the syndicate size increases b/e the risk loading of each member decreases faster than the rate at which the syndicate size n is increasing.

  • 1 Given F(y,x)=0, if an implicit function y = f(x) exist, then the implicit-function-rule gives : dy/dx = -F x /F y . See Fundamental Methods of Mathematical Economics, Alpha. C. Chiang, page 204-208.

  • 2 As

"

0, and no correlation between e s and y s , an increase in e s leads to a fall in .

Economics of Insurance

Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E.

RISK POOLING

Suppose there are n individuals identically and independently distributed.

y i : random income of i th indiv. without pooling that is continuously distributed within the interval [0,1] and has the distribution function F(y i ).

we now prove the following proposition:

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. RISK POOLING

Equal shares pooling scheme: the n members put their realized incomes y i into the pool and share out the total realized income equally. The random pooled income of each member can be written as:

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. RISK POOLING

--------- (1)

Alternative pooling scheme: the scheme gives some individual member of the pool the share k i of the income of the jth member of the pool, where k i (1/n) for some i 3 . The random income of such a member would be:

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. RISK POOLING

------- (2) 4

where we have substituted for y n from (1).

Hence in this alternative pooling scheme the income of a member = random income under equal share pooling [w(n)] + weighted sum of further n-1 variables.

Let y i -y n = difference between the random incomes of the i th and n th individual.

Let S in = sum of realizations of all n-2 random incomes except y i and y n so that

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. RISK POOLING
  • 3 E.g: sharing scheme of two individual: individual-1 gets k share of his realized income and (1-k) share of the realized income of individual-2. Therefore, w(k) = ky 1 +(1-k)y 2 = y 2 + k(y 1 -y 2 ).

  • 4 In this alternative scheme an n member pool decides to give the first n-1 members 1/n-1 of their total realized income and the n th member just y n identical to n-1 member pool with an optimal equal shares scheme for its members.

Economics of Insurance

Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E.

The probability that y i - y n = y given that the realized value of w(n)=w is:

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability
Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability
Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability

-------- (3)

⁄ . ⁄
⁄ . ⁄

------ (4)

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability

Now let us consider the probability that

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability
⁄ . ⁄ ------- (5)
⁄ . ⁄
------- (5)

Since y i and y n are identically distributed, (4) and (5) are equal and the events that

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability

Hence the expected value of y i – y n given w(n) is zero.

From (2), w(k) is sum of w(n) + n-1 random variable with zero mean for given w(n).

Economics of Insurance Ref. Gravelle & Rees: Microeconomics, Chapter 19, section D & E. The probability

So, risk averters prefer w(n) to w(k). Hence equal shares pooling scheme is optimal. : Proof of the proposition.