The Panjer recursion (also known as Panjer algorithm ) is an algorithm for the probability distribution of a special composite random variable
S.
: =
∑
i
=
1
N
X
i
: =
∑
n
=
0
∞
χ
{
ω
∈
Ω
|
N
(
ω
)
=
n
}
∑
i
=
1
n
X
i
{\ displaystyle S: = \ sum _ {i = 1} ^ {N} X_ {i}: = \ sum _ {n = 0} ^ {\ infty} \ chi _ {\ {\ omega \ in \ Omega | N (\ omega) = n \}} \ sum _ {i = 1} ^ {n} X_ {i}}
to calculate. Here and are random variables which form a collective model and denote the indicator function .
N
{\ displaystyle N}
X
i
{\ displaystyle X_ {i}}
χ
{\ displaystyle \ chi}
The algorithm was first published in a publication by Harry Panjer . It is widely used in insurance .
Preconditions
We are interested in the special composite random variable , where and must meet the following preconditions:
S.
=
∑
i
=
1
N
X
i
{\ displaystyle \ textstyle S = \ sum _ {i = 1} ^ {N} X_ {i}}
N
{\ displaystyle N}
X
i
{\ displaystyle X_ {i}}
Number of claims distribution
N
{\ displaystyle N}
is a "damage number distribution", i. H. . is independent of .
N
∈
N
0
{\ displaystyle N \ in \ mathbb {N} _ {0}}
N
{\ displaystyle N \,}
X
i
{\ displaystyle X_ {i} \,}
There must also be an element of the Panjer class . The Panjer class consists of all random variables with values in which satisfy the following relation:
with and for and with . The value is determined so that is satisfied.
N
{\ displaystyle N}
N
0
{\ displaystyle \ mathbb {N} _ {0}}
p
k
=
(
a
+
b
k
)
⋅
p
k
-
1
{\ displaystyle \ textstyle p_ {k} = \ left (a + {\ frac {b} {k}} \ right) \ cdot p_ {k-1}}
k
≥
1
{\ displaystyle ~~ k \ geq 1}
a
{\ displaystyle a}
b
{\ displaystyle b}
a
+
b
≥
0
{\ displaystyle a + b \ geq 0}
p
0
{\ displaystyle p_ {0} \,}
∑
k
=
0
∞
p
k
=
1
{\ displaystyle \ textstyle \ sum _ {k = 0} ^ {\ infty} p_ {k} = 1}
Sundt proved in the paper that only the binomial distribution , the Poisson distribution and the negative binomial distribution are in the Panjer class. You have the parameters and values as described in the following table, where the probability generating function is designated.
W.
N
(
x
)
{\ displaystyle W_ {N} (x) \,}
distribution
P
[
N
=
k
]
{\ displaystyle P [N = k]}
a
{\ displaystyle a}
b
{\ displaystyle b}
p
0
{\ displaystyle p_ {0}}
W.
N
(
x
)
{\ displaystyle W_ {N} (x)}
E.
[
N
]
{\ displaystyle E [N]}
V
a
r
(
N
)
{\ displaystyle Var (N)}
Binomial
(
n
k
)
p
k
(
1
-
p
)
n
-
k
{\ displaystyle {\ binom {n} {k}} p ^ {k} (1-p) ^ {nk}}
p
p
-
1
{\ displaystyle {\ frac {p} {p-1}}}
p
(
n
+
1
)
1
-
p
{\ displaystyle {\ frac {p (n + 1)} {1-p}}}
(
1
-
p
)
n
{\ displaystyle (1-p) ^ {n} \,}
(
p
x
+
(
1
-
p
)
)
n
{\ displaystyle (px + (1-p)) ^ {n} \,}
n
p
{\ displaystyle np \,}
n
p
(
1
-
p
)
{\ displaystyle np (1-p) \,}
Poisson
e
-
λ
λ
k
k
!
{\ displaystyle e ^ {- \ lambda} {\ frac {\ lambda ^ {k}} {k!}}}
0
{\ displaystyle 0 \,}
λ
{\ displaystyle \ lambda \,}
e
-
λ
{\ displaystyle e ^ {- \ lambda} \,}
e
λ
(
x
-
1
)
{\ displaystyle e ^ {\ lambda (x-1)} \,}
λ
{\ displaystyle \ lambda \,}
λ
{\ displaystyle \ lambda \,}
Negative binomial
Γ
(
r
+
k
)
k
!
Γ
(
r
)
p
r
(
1
-
p
)
k
{\ displaystyle {\ frac {\ Gamma (r + k)} {k! \, \ Gamma (r)}} \, p ^ {r} \, (1-p) ^ {k}}
1
-
p
{\ displaystyle 1-p \,}
(
1
-
p
)
(
r
-
1
)
{\ displaystyle (1-p) (r-1) \,}
p
r
{\ displaystyle p ^ {r} \,}
(
p
1
-
x
(
1
-
p
)
)
r
{\ displaystyle \ left ({\ frac {p} {1-x (1-p)}} \ right) ^ {r} \,}
r
(
1
-
p
)
p
{\ displaystyle {\ frac {r (1-p)} {p}} \,}
r
(
1
-
p
)
p
2
{\ displaystyle {\ frac {r (1-p)} {p ^ {2}}} \,}
Distribution of individual losses
We assume that there are identically distributed independent random variables that are independent of . Furthermore, it must be distributed on a grid with grid length .
X
i
{\ displaystyle X_ {i} \,}
N
{\ displaystyle N \,}
X
i
{\ displaystyle X_ {i} \,}
H
N
0
{\ displaystyle h \ mathbb {N} _ {0}}
H
>
0
{\ displaystyle h> 0 \,}
f
k
=
P
[
X
i
=
H
k
]
.
{\ displaystyle f_ {k} = P [X_ {i} = hk]. \,}
Recursion
The algorithm uses recursion to calculate the probabilities .
G
k
=
P
[
S.
=
H
k
]
{\ displaystyle g_ {k} = P [S = hk] \,}
The starting value is:
G
0
=
W.
N
(
f
0
)
{\ displaystyle g_ {0} = W_ {N} (f_ {0}) \,}
with the special cases
G
0
=
p
0
⋅
exp
(
f
0
b
)
For
a
=
0
,
{\ displaystyle g_ {0} = p_ {0} \ cdot \ exp (f_ {0} b) {\ mbox {for}} a = 0, \,}
and
G
0
=
p
0
(
1
-
f
0
a
)
1
+
b
/
a
For
a
≠
0.
{\ displaystyle g_ {0} = {\ frac {p_ {0}} {(1-f_ {0} a) ^ {1 + b / a}}} {\ mbox {for}} a \ neq 0.}
The following values can be calculated as follows:
G
k
=
P
[
S.
=
H
k
]
=
1
1
-
f
0
a
∑
j
=
1
k
(
a
+
b
⋅
j
k
)
⋅
f
j
⋅
G
k
-
j
.
{\ displaystyle g_ {k} = P [S = hk] = {\ frac {1} {1-f_ {0} a}} \ sum _ {j = 1} ^ {k} \ left (a + {\ frac {b \ cdot j} {k}} \ right) \ cdot f_ {j} \ cdot g_ {kj}.}
example
The example shows the approximated density function of where and . The individual damage distribution was discretized with a grid width (see also Fréchet distribution ).
S.
=
∑
i
=
1
N
X
i
{\ displaystyle \ textstyle S \, = \, \ sum _ {i = 1} ^ {N} X_ {i}}
N
∼
NegBin
(
3.5
,
0.3
)
{\ displaystyle \ textstyle N \, \ sim \, {\ text {NegBin}} (3.5,0.3)}
X
∼
Cheeky
(
1.7
,
1
)
{\ displaystyle \ textstyle X \, \ sim \, {\ text {Frechet}} (1.7,1)}
H
=
0.04
{\ displaystyle h = 0.04}
See also
literature
Schmidt, Klaus D .: Actuarial Mathematics , Springer Dordrecht Heidelberg London New York 2009, ISBN 978-3-642-01175-7 .
Individual evidence
^ Harry H. Panjer: Recursive evaluation of a family of compound distributions. . (PDF) In: ASTIN Bulletin . 12, No. 1, 1981, pp. 22-26. doi : 10.1017 / S0515036100006796 .
^ B. Sundt and WS Jewell: Further results on recursive evaluation of compound distributions . (PDF) In: ASTIN Bulletin . 12, No. 1, 1981, pp. 27-39. doi : 10.1017 / S0515036100006802 .
<img src="https://de.wikipedia.org//de.wikipedia.org/wiki/Special:CentralAutoLogin/start?type=1x1" alt="" title="" width="1" height="1" style="border: none; position: absolute;">