FW: Hossak Pollard Zehnwirth

NAUSS, JOHN ( (no email) )
Fri, 21 Aug 1998 14:48:45 -0700

Dimitra,
We're talking about a collective risk model here (a review of Bowers
readings for 5A may clarify this somewhat).
In any event, the HPZ passage on p.157 gives a bit of a clue when it
indicates that "The conditional formulae (4.4.3) and (4.5.2) can then be
used to show that..."

What it boils down to is that if the Number of claims N is a random
variable with a Poisson distribution and mean and variance both equal to
n, and if the size of a claim is another random variable X (and N and X
are independent)
then the formula for the variance of C = X1 + X2 +....Xn is:

Var (C) = E[N] * Var (X) + Var (N) * E[X]^2
Var (C) = n*s^2 + n*m

where m is the mean of the size of the claim and s^2 is the variance of
the size of the claim distribution.

John Nauss
johnau@safeco.com

-----Original Message-----
From: Dimitra Roidakis [SMTP:dr@g-g-a.com]
Sent: Friday, August 21, 1998 9:22 AM
To: studygroup4b@lists.casact.org
Subject: Hossak Pollard Zehnwirth

I have a question concerning Full Credibility theory, page 157 of
HPZ. When they state that Var(C)=n(s^2+m^2)

To me Var(C)=Var(X1+X2+...Xn)=Var(X1)+..Var(Xn)=nS^2
(were S^2 stands for sigma square)
My question is where does the nm^2 comes from?

Dimitra
dr@g-g-a.com