23.3.
CONDITIONAL PAYOFFS USING THE LOGNORMAL MODEL 331
Table 23.2: Derivation of Partial Expectation of Lognormal
We want the partial expectation of a lognormal random variable X with parameters m and v, given that
it is less than k. The probability density function of a lognormal with parameters m and v is
2 2
e −(ln x−m) /2v
f (x; m, v) √
xv 2π
Therefore, the partial expectation is
∫ k 2 2 ∫ k
xe −(ln x−m) /2v dx 1 2 /2v 2
PE[X | X < k] √ √ e −(ln x−m) dx
0 xv 2π v 2π 0
Substitute y ln x − m:
x e y+m
dx e y+m dy
As part of the substitution, translate the bounds of the integral from x to ln x − m.
0 → −∞
k → ln k − m
and the expression becomes
∫ ln k−m ∫ ln k−m
1 2 /2v 2 em 2 /2v 2
√ e −y e y+m dy √ e −y e y dy
∫ ln k−m
v 2π −∞ v 2π −∞
em 2 −2v 2 y)/2v 2
√ e −(y dy
v 2π −∞
We want the numerator of the combined exponent to be a negative square. To complete the square, we need
to turn y 2 − 2v 2 y into (y − b)2 for some b. We have 2b y 2v 2 y, so b v 2 and
y 2 − 2v 2 y (y − v 2 )2 − v 4
The resulting expression for the partial expectation becomes
∫ ln k−m ∫ ln k−m
em −(y−v 2 )2 /2v 2 v 2 /2 m+0.5v 2 1 2 2 2
√ e e dy e √ e −(y−v ) /2v dy
v 2π −∞ −∞ v 2π
The integrand is now the density of a normal distribution with mean v 2 and variance v 2 , so the integral is the
probability that such a variable is less than ln k − m, which is N (ln k − m − v 2 )/v . The partial expectation is
therefore
2 ln k − m − v 2 ln k − m − v 2
PE[X | X < k] e m+0.5v N E[X]N
v v
IFM Study Manual—1st edition 2nd printing
Copyright ©2018 ASM
