Chapter 18

The Lognormal Distribution

Question 18.1.
−7+8 −11+8 −3+8 2+8
The five standard normals are √ = .2582, √ = −.7746, √ = 1.291, √ = 2.582, and
15 15 15 15
−15+8
√ = −1.8074.
15

Question 18.2.

If z is standard normal, µ + σ × z is N µ, σ 2 hence our five standard normals can be use

! "

to create the desired properties: .8 + 5 (−1.7) = −7.7, .8 + 5 (.55) = 3.55, .8 + 5 (−.3) = −0.7,
.8 + 5 (−.02) = .7, and .8 + 5 (.85) = 5.05.

Question 18.3.

x1 + x2 is normally distributed with mean −1 and variance 5 + 2 + 2(1.3) = 9.6. x1 − x2 is nor-

mally distributed with mean 3 and variance 5 + 2 − 2(1.3) = 4.4.

Question 18.4.

Sums and differences of two random variables are normally distributed hence x1 + x2 is normally
distributed with mean µ1 + µ2 = 10 and variance
# √ $
σ12 + σ22 + 2ρσ1 σ2 = 0.5 + 14 + 2 × (−0.3) × 0.5 × 14 = 10.3

The difference is normally distributed with mean µ1 − µ2 = −6 and (higher) variance

# √ $
σ12 + σ22 − 2ρσ1 σ2 = 0.5 + 14 − 2 × (−0.3) × 0.5 × 14 = 18.7.

Question 18.5.

x1 + x2 + x3 ˜N (5.5, 47.8), x1 + 3x2 + x3 ˜N (9.5, 123.4), and x1 + x2 + .5x3 ˜N (4.25, 30.65).

Question 18.6.
2
If x˜N µ, σ 2 then E (ex ) = ex µ+.5σ ; using the given numbers, E (ex ) = e2+2.5 = 90.017. There
! "

is a 50% probability x is below its mean of 2 hence the median of ex is e2 = 7.3891.

244
 Chapter 18 The Lognormal Distribution

Question 18.7.

Denote the stock price of month i by Si and let the continuous return of month i be denoted as
Ri = ln (Si+1 /Si ).

a) The average of the monthly continuous returns is zero for both stock A and B; the annual
return is also zero. Note the average simple return will be much higher for stock B although both
stocks have had the same 4 month holding period return.

b) The standard deviations, since R = 0, is 14 i Ri2 . For A, this monthly standard deviation
!
√
is 0.167% which is a yearly standard deviation of 12 (.00167) = .578%. For B, the monthly
standard deviation is 8.014% which is a yearly standard deviation of 27.76%. For this example we
did not adjust for degrees of freedom (i.e. divide the sum by 3 instead of 4).

c) No matter what time unit we use, the (annualized) mean continuous return will be ln (ST /S0 ) /T .
This is due to the intermediate observations cancelling. For example, suppose we have 2 years of
data observed at the end of the month (for 24 months of data). The mean monthly continuous return
is
24 24
1 " 1 " 1
ln (Si+1 /Si ) = ln (Si ) − ln (Si−1 ) = [ln (S24 ) − ln (S0 )] .
24 24 24
i=1 i=1

The intermediate end of month prices are irrelevant. This average, when annualized (i.e. multiply
the above by 12), is the same average we would get if we would use the average of two yearly
continuous returns. For standard deviation, there will be squared deviations which lead to cross
terms. The sampling interval will, therefore, give different estimates of volatility.

Question 18.8.
√ $
Since St = S0 exp α − 12 σ 2 t + σ tz where z is standard normal,
## $

1 2 √ 105
%% & % &&
P (St > 105) = P α − σ t + σ tz > ln
2 100
# 105 $ #
− α − 21 σ 2 t
' (
ln 100
$
=P z> √
σ t

− ln 105 1 2
' $ (
+ −
# $ #
100 α 2 σ t
= P −z < √ = N (d2 )
σ t
$ * # √$
where d2 = ln (100/105) + α − 12 σ 2 t / σ t . Using the given parameters, d2 = .045967and
) #

N (d2 ) = .4817. For this parameter specification, the probability St > 105 increases with t and

245
Part 5 Advanced Pricing Theory

decreases with σ . Analytically, since N ′ (d2 ) > 0, the derivative will have the same sign as ∂d2 /∂t
and ∂d2 /∂σ . Specifically,

− 2 /2 − ln (S /K) /t
# $
∂P (St > K) α σ 0
= N ′ (d2 ) √ >0
∂t 2σ t

since α − σ 2 /2 = .035 > 0 and ln (S0 /K) < 0. As an example, if t is 5 years, there is a 57.46%
chance of being greater than 105. For volatility, let t = 1. Then

α − ln (S0 /K) 1
% &
∂P (St > K) ′
= −N (d2 ) + < 0.
∂σ 2σ 2 4

Question 18.9.

We have d1 = .2540, and using equation (18.30),

N (d1 ) .6003
E (S1 |S1 > 105) = 100e.08 = 100e.08 = 135.
N (d2 ) .4817

Table One shows how changes in t and σ affect this conditional expectation. We see it increases
with time and volatility.

Table One (Problem 18.9)

t E(St|St>105) σ E(S1|S1>105) r E(S1|S1>105)
0.25 117.19 0.10 115.15 0.01 131.63
0.50 123.98 0.20 124.65 0.02 132.08
0.75 129.74 0.30 135.00 0.03 132.54
1.00 135.00 0.40 146.25 0.04 133.01
1.25 139.97 0.50 158.48 0.05 133.49
1.50 144.76 0.60 171.78 0.06 133.98
1.75 149.42 0.70 186.26 0.07 134.49
2.00 154.01 0.80 202.04 0.08 135.00
2.25 158.54 0.90 219.26 0.09 135.53
2.50 163.05 1.00 238.06 0.1 136.07
2.75 167.54 1.10 258.61 0.11 136.63
3.00 172.04 1.20 281.11 0.12 137.19
3.25 176.54 1.30 305.77 0.13 137.78
3.50 181.06 1.40 332.82 0.14 138.37
3.75 185.61 1.50 362.55 0.15 138.98
4.00 190.19 1.60 395.25 0.16 139.60
4.25 194.81 1.70 431.27 0.17 140.24
4.50 199.48 1.80 471.02 0.18 140.90
4.75 204.19 1.90 514.93 0.19 141.57
5.00 208.95 2.00 563.51 0.2 142.25

246
 Chapter 18 The Lognormal Distribution

Question 18.10.

We have

1 2 √ 98
%% & % &&
P (St < 98) = P α − σ t + σ tz < ln
2 100

ln 100 − α − 12 σ 2 t
' # 98 $ # $ (
=P z< √ = N (−d2 )
σ t

with −d2 = (ln (98/100) − .035) /.3 = −.18401. Hence P (St < 98) = N (−.18401) = 42.70%.

Question 18.11.

Using equation 18.28, E(S1 |S1 < 98) = 100e.08 N(−.484)/N(−.184) = 79.71. Similarly,
E(S1 |S1 < 120) = 100e.08 N(.1911)/N(.4911) = 90.62. See Figure One for the negative change in
both expectations as we increase t. When σ = .1, E(S1 |S1 < 98) = 92.99 and E(S1 |S1 < 120) =
105.35. Both significantly increase.
Figure One (Problem 18.11)
105

100

95

90 (St|St<120)

85

80
(St|St<98)

75

70

65
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Question 18.12.

See Figure Two on the next page. Option prices depend on the conditional (risk neutral) expectation,
not the probability the option is in the money. As T increases, the likelihood that ST > KT may be

247
Part 5 Advanced Pricing Theory

lower; however, the payoff depends on the conditional expectation (since the option does not pay
a constant amount). The increased dispersion offsets the lower probability (for the call option).
Figure Two (Problem 18.12)
0.9

0.8

0.7
(ST < 100e.08T)

0.6

0.5

(ST > 100e.08T)

0.4

0.3

0.2
0 5 10 15 20 25

Question 18.13.

An example is with K = 80. When t is less than approximately 6 years, P (St < 80) is increas-
ing# in t; however, if t is larger, the probability decreases. For example, if t = #1/12 we have
P S1/12 < 80 = 0.45%. If we increase the time by one month (i.e. t #= 2/12) P S$1/6 < 80 =
$ $

3.08%. When t = 10, P (S10 < 80) = 27.29%; with t = 10 + 1/12, P S121/12 < 80 = 27.27%.
The effect on the conditional expectation is unambiguous (it decreases).

Question 18.14.

The mean should be varying year by year; whereas, the standard deviation should be more stable.

Question 18.15.

Although both data should appear non (log) normal, the weekly data should be closer to normality.

248