Step and Fly

Health Microinsurance Model

Prepared for Step and Fly Microinsurance Prepared by Matthew Tudball Last updated on 28 May, 2013

Foreword
Step and Fly: Health Microinsurance Model is an internal document for Step and Fly Microinsurance. It mathematically constructs the model used in the firm's actuarial calculations. Although it is a working model, it is still in the process of being improved and adapted.

Supply-Side Model
Exogenous variables: N, q, n, k0, e, Endogenous variables: b, g, P, xi, Suppose that there are N clients enrolled in the micro-insurance programme, each of whom are contributing some premium q to the communal insurance pot each period, such that the pot P = N q at the beginning of one period. In the simple model, we will assume that there is only one period. Suppose that the N clients are subdivided into g groups, where each group contains n people, so N = g n. Suppose again that the clients can only make one claim per period. Assuming that there is no risk of fraud in this model, if a client has medical expenses e that exceed some pre-set and exogenously defined upper bound, they can petition to the micro-insurance programme and will receive some benefit B, which is defined as B = min{e, b}, where b will be defined endogenously. Assume that e b in this model. In this model, there are two states of the world, the good state where the client is healthy, and the bad state where the client is ill. The probability that a single client will become ill is 0 < < 1, and this is constant across all clients. Assume for now that is inter-temporally constant. Given that there are N clients in the scheme, the probability of illness follows a binomial distribution, and can be written as follows. Suppose that S represents the number of clients who have fallen ill. Define K as the number of claims in the insurance scheme itself, and k as the number of claims per group. N 0 N P ( S = 0) = ( N ) ( 1 ) = ( 1 ) 0 ( N 1) 1 ( N 1) N P ( S =1 ) = ( ) ( 1 ) = N ( 1 ) 1 ... ( N K) K N P ( S = K ) = ( ) ( 1 ) K 0 N N P ( S = N ) = ( N ) ( 1 ) = N For each P(S = k), there is an associated value of the communal insurance pot, which can be defined as: For S = 0, P 0 = P For S =1, P 1 = P b For S = k , P k = P K b For S = N , P N = P N b Since we have the associated value for each case, we can calculate the expected value of the pot. Intuitively, this is the weighted average of the possible values that the communal insurance pot P can take.

E ( P ) = P K ( N )( 1 )( N K ) K K K=0
Few Claims Bonus Sub-Model Suppose that we want to introduce a group-based few claims bonus X, whereby if the number of claims per group is less than some exogenously defined k0 , the group will be distributed some amount x k based on how many claims were put forward. P ( k = 0 ) = ( 1 )n n 1 P ( k = 1) = n ( 1 )

P ( k = k 0 ) = ( n )( 1 )n k k k0
0

... P (k = n ) = n The associated values of xk are: For 0 claims put forward , X = x 0 For 1 claim put forward , X = x1 ... For k 0 claims put forward , X = x k For k 0+1 claims put forward , X =0
For n claims put forward , X = 0

Where x 0 > x 1 > ... > x k > 0 So the expected value of the few claims bonus per group is:
k0

E ( X ) = x k ( n )( 1 ) k k =0

n k

We can now expand this few claims bonus on a scheme-wide scale. Consider the following matrix, where the rows represent combinations of claims, and the columns represent the list of possible combinations.

0 0 0 0

0 0 0 1

0 0 0 0 0 0 0 0 1 2 n 0

0 0 1 1

0 0 0 0 0 0 1 1 2 2 n 0

0 0 2 1

0 0 0 0 2 2 2 n

n n n 0

n n n 1

n n n n n n 2 n

There are ( N ) columns in the matrix and g rows. Since order does not matter, we can calculate the expected value of the few claims bonus from the matrix. P ( g 0 claims)=( 1 )g n g n 1 P ( g 1 0 claims , 1 1 claim ) = g n ( 1 ) g n P ( g n claims ) = The associated value of the few claims bonus is: For g 0 claims , x = g x 0 For g 1 0 claims , 1 1 claim , x = ( g 1 ) x 0 + x1
For g n claims , x = 0

The general form of the few claims bonus will be calculable once n and g are defined. For now, we will

denote the scheme-wide expected value of the few claims bonus as E ( X ) . For simplicity, however, we can take the expected value of the few claims bonus per group as a single payment for all groups whose claims per period were less than k0. The probability that k is less than k0 is:
k0

P ( k is less than k 0) = ( n )( 1 ) k =0 k

n k

So the probability that groups will be eligible for the bonus is:

P ( 0 groups ) = ( g )( 1 ( n )( 1 )n k k ) 0 k =0 k P ( 1 group ) = ( g )( 1 ( n )( 1 ) 1 k =0 k
k0 k0 n k

k0

g 1

( ( n )( 1 ) k =0 k
k0

k0

n k

)
h

n k k P ( h groups ) = ( g )( 1 ( n )( 1 ) ) h k =0 k ... n k k P ( g groups ) = ( g )( ( n )( 1 ) ) g k =0 k k0 g

g h

( ( n )( 1 ) k =0 k

n k

The expected value is therefore:

E ' ( X ) = ( h x k ( n )( 1 )nk k )( g )( 1 ( n )( 1 )n k k ) k h h= 0 k =0 k =0 k
Supply-Side Model Continued

k0

k0

g h

( ( n )( 1 )n k k ) k =0 k

k0

Since we know the expected value of the pot and the approximate expected value of the few claims bonus, we can subtract the latter from the former to calculate the net expected value of the pot.

E ( P )= P K ( N )( 1 )( N K ) K K K =0
(h
h= 0 k =0 g k0 n k k n k k x k ( n )( 1 ) )( g )( 1 ( n )( 1 ) ) k h k =0 k k0 g h

( ( n )( 1 ) k =0 k

k0

n k

From here, we can define the good state as one where K y. This is an arbitrary condition to impose, but it means that when we calculate prices, the upper bound of the benefit and the few claims bonus, if total claims are less than y, the pot will expect to have a positive amount in it.

E ( P )= P K ( N )( 1 )( N K ) K + P K ( N )( 1 )( N K ) K E ' ( X ) K K K =0 K = y +1
Since we do not want to expect bad state losses, we can set:
N

P K ( N )( 1 )( N K ) K =0 K K = y + z+1

Where z is another arbitrary value. And we can define the expected value of the few claims bonus in terms of some segment of the expected value of the pot. Once again, this is arbitrary, but simplifies the calculations later on.

)( 1 )n k k )( g )( 1 ( n )( 1 )n k k ) (h x k (n k h h= 0 k =0 k =0 k

k0

k0

g h

( ( n )( 1 )n k k ) k =0 k

k0

P K ( N )( 1 ) K K = y +1

y +z

(N K)

Since there are k potential x values defined in one expression, when it comes to calculations, trial and error will probably yield a set of xi's which adhere to the condition above. Isolating b, we have calculated the upper bound as a function of the number of clients N, the insurance premium q and the risk of illness , such that:

N q b ( N , q , ) =

( N )( 1 )( N K ) K K = y + z+ 1 K

K ( N )( 1 )( N K ) K K K = y + z+1

Theorem 1.1:

N q
N

lim b ( N , q , ) = lim

K =1 N

)( 1 )( N K ) K (N K

K =1

)(1 )( N K ) K K ( N K lim N q lim =

Proof of Theorem 1.1:

N q lim
N N

)( 1 )( N K ) K (N K
K =1

)( 1 )( N K ) K (N N K
K =1

K =1 N

)( 1 )( N K ) K K ( N K

N K = 1

lim

)(1 )( N K ) K K ( N K

i)

( N )( 1 )( N K ) K =1 N K lim
K =1 N

ii) lim

N K =1

)( 1 )( N K ) K K ( N K ( N 1)( 1 )( N K ) K 1 N K 1
K =1 N N

= lim N lim
N N

Since

lim ( N 1)( 1 )( N K ) K 1 = 1, lim K ( N )( 1 )( N K ) K = lim N N K = 1 K 1 N K =1 K N N lim N q lim ( N )( 1 )( N K ) K lim N q N N K = 1 K q N = = , QED N lim N lim K ( N )( 1 )( N K ) K N N K =1 K

This is an interesting theorem, since it demonstrates that the supply-side model can provide actuarially fair insurance a broadly demand-side concept when the number of policyholders approaches infinity. The reconciliation of the supply- and demand-side models is an interesting consideration when calculating benefit caps.

Now that we have defined the few claims bonus and the upper bound of the benefit so that the pot will always expect to carry some money over from period to period, we can look at defining the model intertemporally.

E ( P 1) = ( P K b ) ( N )( 1 )( N K ) K K K=0 E (P2) =
y

)( 1 )( N K ) K K b ) ( N )( 1 )( N K ) K ( P + ( P K b) ( N K K
K =0 K =0 y 2

= ( P K b ) ( N )( 1 )( N K ) K + ( P K b ) [( N )( 1 )( N K ) K ] K K K =0 K =0 ... E ( P t )=
t =1 t0

)( 1 )( N K ) K ] ( P K b ) [( N K K =0
t

The inter-temporal expected value of the insurance pot is therefore:

E ( P t )=
t =1

t0

)( 1 )( N K ) K ] E ( X t ) ( P K b ) [( N K K =0

Where E(X) can again be defined arbitrarily in terms of some segment of the expected value of the pot. Appendix I It is worth considering, as an addendum, the relationship between the hitherto undiscussed period length, the per period premium and the benefit. Until now, we have assumed that the period length is some arbitrary interval of length l = 1. We can incorporate it into the formula for the benefit b in a more general form.

N l q b ( N , q , ) =
N

( N )( 1 ( l )( 1 )lm m ) K = y + z+ 1 K m= 1 m

NK

m= 1 l

l )( 1 )l m m (m
K

)(1 ( l )( 1 )lm m ) K ( N K K = y + z+ 1 m= 1 m

NK

l )( 1 )l m m (m m= 1