Professional Documents
Culture Documents
Dorje C. Brody
Department of Mathematics,
Imperial College London,
London SW7 2AZ
www.imperial.ac.uk/people/d.brody
-2 -
18 June 2010
c DC Brody 2010
-3 -
18 June 2010
1 Q
PuXudu ,
S t = Et
Pt
t
where
(1)
Pu = exp
rsds
(2)
c DC Brody 2010
-4 -
18 June 2010
t = t
PuXudu + Bt,
(3)
(4)
c DC Brody 2010
-5 -
18 June 2010
As a simple model for the commodity convenience benefit, let us consider the
case where {Xt} is given by an Ornstein-Uhlenbeck (OU) process.
Then we have
dXt = ( Xt)dt + dt,
where {t} is a Brownian motion that is independent of {Bt}.
(5)
Here is the mean reversion level, is the mean reversion rate, and is the
volatility.
We shall be looking at the constant parameter case first, and then extend the
results into time-dependent situation.
A standard calculation making use of an integration factor shows that
Xt = e
X0 + (1 e
) + e
esds.
(6)
Thus, starting from the initial value X0, the process tends in mean towards the
level , and has the variance
2
Var[Xt] =
1 e2t
(7)
2
Practical Quantitative Analysis in Commodities
c DC Brody 2010
-6 -
18 June 2010
(8)
eudu.
(9)
(10)
c DC Brody 2010
-7 -
18 June 2010
for T > t.
Interestingly, there is another orthogonal decomposition as well, this time for Xt,
which plays a crucial role in what follows.
This decomposition is given by
Xt =
et et
XT
Xt T
e eT
et et
XT .
+ T
e eT
(11)
et et
btT = Xt T
XT
T
e e
is an Ornstein-Uhlenbeck bridge (OU bridge).
An alternative way of expressing the OU bridge is to write
sinh(t)
btT = Xt
XT .
sinh(T )
Clearly we have b0T = X0 and bT T = 0.
Practical Quantitative Analysis in Commodities
(12)
(13)
c DC Brody 2010
-8 -
18 June 2010
E[btT ] =
sinh(T )
sinh(T )
and variance
2 cosh(T ) cosh((T 2t))
var[btT ] =
.
2
sinh(T )
The mean and variance of the OU bridge are plotted in Figure 1.
(14)
(15)
E
t
PuXudu t, Xt . (16)
(17)
c DC Brody 2010
-9 -
18 June 2010
OU Bridge
0.7
data1
data2
Mean
Variance
0.6
0.5
0.4
0.3
0.2
0.1
-0.1
-0.2
50
100
150
200
250
300
350
Time
Mean (red) and variance (blue) of the OU bridge. The parameters are set as = 0.15, T = 1, X0 = 0.6,
= 1.4, and = 0.5.
Figure 1:
c DC Brody 2010
-10 -
where
t = t
18 June 2010
PuXudu + Bt.
(18)
t = t t
PuXudu.
(19)
Ft = t,
t
s
0<st
, {Xs}0st .
(20)
t s
Q [T < x|{s}0st] = Q T < x t,
t
s
= Q [T < x|t] .
0<st
(21)
This follows from the fact that t and T are independent from Gt, where
t s
Gt =
,
t
s 0st
which follows in turn from properties of the standard Brownian motion.
Practical Quantitative Analysis in Commodities
(22)
c DC Brody 2010
-11 -
18 June 2010
c DC Brody 2010
-12 -
18 June 2010
Pt St = E
t
= E
t
= E
t
PuXudu Ft
PuXudu t, Gt, {Xs}0st
PuXudu t, {Xs}0st
PuXudu t, {Xs}0st .
(24)
(25)
= E
t
t PuXu du.
PuXudu t, Xt .
(26)
t
c DC Brody 2010
-13 -
18 June 2010
PuXudu =
Pu Xu e(ut)Xt du
Pue(ut)du Xt.
(27)
Pue(ut)du Xt + E At tAt + Bt ,
(28)
+
t
Pt St =
t
where
At =
t
Pu Xu e(ut)Xt du,
(29)
c DC Brody 2010
-14 -
18 June 2010
But now the problem is essentially solved, since the remaining expectation is of
the form
E [A|A + B] ,
(30)
where A and B are independent Gaussian random variables each with a known
mean and variance.
That is to say, we have:
A=
t
Pu Xu e(ut)Xt du
(31)
and
Bt
B= .
t
(32)
(33)
Writing
we observe in particular that A + B and (1 x)A xB are orthogonal and
hence independent if we set
Var[A]
x=
.
(34)
Var[A] + Var[B]
Practical Quantitative Analysis in Commodities
c DC Brody 2010
-15 -
18 June 2010
This then enables us to work out the expectation to determine the value of the
commodity.
We proceed as follows.
First we note from (9) that
Xu e(ut)Xt = (1 e(ut)) + eu
esds.
(35)
Therefore, we have
A =
t
Pu Xu e(ut)Xt du
=
t
Pudu et
u=t
Pueu
Pueudu
esdsdu.
(36)
s=t
It follows that
E[A] = pt etqt ,
Practical Quantitative Analysis in Commodities
(37)
c DC Brody 2010
-16 -
18 June 2010
where
pt =
Pudu
(38)
Pueudu.
(39)
and
qt =
t
Pu e
e ds du =
Pueudu ds,
(40)
u=s
s=t
s=t
u=t
s=t
es
Pueudu ds
u=s
esqsds.
(41)
Var[A] = 2
t
Practical Quantitative Analysis in Commodities
e2sqs2ds.
(42)
c DC Brody 2010
-17 -
18 June 2010
We also have
1
Var[B] = 2 .
t
The commodity price can then be worked out as follows.
(43)
We have
Pt St = E
PuXudu t, Xt
(44)
(45)
E[A] = pt etqt .
(46)
t
PtSt = (1 xt) pt + e qt(Xt ) + xt .
t
(47)
c DC Brody 2010
-18 -
18 June 2010
2 2t t e2sqs2ds
.
xt =
2s 2
2
2
1 + t t e qs ds
(48)
Thus we see that for large and/or large the value of x tends to unity.
On the other hand, for small and/or small the value of x tends to zero.
Hence, if the market information has a low noise content (high ), then the
market information is what mainly determines the price of the commodity.
On the other hand, if the volatility of the benefit is high, then market
participants also rely heavily of the latest information in their determination of
prices.
The other term in the expression for St is essentially an annuitised valuation of a
constant benefit rate set at the mean reversion level, together with a correction
term to adjust for the present level of the benefit rate.
This term dominates in situations when the market information is of low quality.
Practical Quantitative Analysis in Commodities
c DC Brody 2010
-19 -
18 June 2010
c DC Brody 2010
-20 -
18 June 2010
(49)
1 rt
e ,
r
qt =
1 (r+)t
e
,
r+
(50)
and
2 2t
.
xt =
2r(r + )2e2rt + 2 2t
A short calculation then shows that
1
t
1
1
St = (1 xt)
Xt + xt ert .
+
r r+
r+
t
(51)
(52)
c DC Brody 2010
-21 -
18 June 2010
Sim
Sim
Sim
Sim
Sim
Market
110
100
90
80
70
60
50
40
30
20
50
100
150
200
250
300
350
time
Figure 2:
Sample paths for the price process (colour) vs the market price for crude oil (black).
c DC Brody 2010
-22 -
18 June 2010
(53)
Recall that the commodity price process {St} in the OU model is a linear
function of the convenience yield
Xt = etX0 + (1 et) + et
esds,
(54)
t = t
eruXudu + Bt.
(55)
ru
Xudu},
t e
and {Bt} at
Since the sum of Gaussian processes is also Gaussian, the evaluation of the call
price reduces to the determination of the mean mT and the variance T2 of ST .
Practical Quantitative Analysis in Commodities
c DC Brody 2010
-23 -
18 June 2010
(56)
and that
T2
2
e2rT
2
2T
2
1e
+ xT
+ 2 .
=
2
2
2(r + )
2r(r + )
T
(57)
PuXudu = At +
Pue(ut)du Xt
= At +
1 rt
e Xt ,
r+
(58)
Xt + xt e At + xt e
.
+
r r+
r+
t
(59)
However, the Gaussian processes {Xt}, {At}, and {Bt} are independent.
Practical Quantitative Analysis in Commodities
c DC Brody 2010
-24 -
18 June 2010
rT
2T
(z mT )2
dz.
(z K) exp
2T2
(60)
If we write
x
1
N (x) =
exp 12 z 2 dz
2
for the cumulative normal density function, then we obtain
(mT K)2
T
exp
C0 = e
2T2
2
for the price of a commodity option.
rT
+ (mT K)N
(61)
mT K
T
(62)
c DC Brody 2010
Figure 3:
-25 -
18 June 2010
Option price surface as functions of the initial asset price and option maturity.
c DC Brody 2010
-26 -
18 June 2010
10
12
14
The call option prices as functions of the initial asset price in the OU model. The parameters are set as
= 0.15, = 0.25, X0 = 0.6, = 0.15, r = 0.05, and K = 10. The three maturities are T = 0.5 (blue), T = 1.0
(green), and T = 3.0 (brown).
Figure 4:
c DC Brody 2010
-27 -
18 June 2010
(63)
sds,
ft =
(64)
we find, by use of the standard method involving an integration factor, that the
solution to (63) is given by
Xt = e
ft
fs
e ssds +
X0 +
0
efs sds .
(65)
c DC Brody 2010
-28 -
18 June 2010
tT s ds
Xt + e
0t s ds
u
0 s ds
uudu
+e
0t s ds
u
0 s ds
udu .
(66)
tT s ds
Xt + e
tT s ds
Xt
(67)
for T > t.
Practical Quantitative Analysis in Commodities
c DC Brody 2010
-29 -
18 June 2010
As before, there is another orthogonal decomposition, this time for Xt, which
plays a crucial role in the time-inhomogeneous setup.
This decomposition is given by
Xt =
Xt
eft
efT
t 2fs 2
0 e s ds
T 2f 2
s
0 e s ds
XT
eft
efT
t 2fs 2
0 e s ds
T 2f 2
s
0 e s ds
XT .
(68)
eft
efT
t 2fs 2
0 e s ds
T 2f 2
s
0 e s ds
XT
(69)
c DC Brody 2010
-30 -
18 June 2010
1
PuXudu t, Xt .
St = E
(70)
Pt
t
We can use the orthogonal decomposition (67) to isolate the dependence of the
commodity price on the current level of the benefit rate Xt.
As before, this turns out to be linear in the benefit rate:
PuXudu =
Pu Xu e(fuft)Xt du
Pue(fuft)du Xt.
(71)
Pue(fuft)du Xt + E At tAt + Bt ,
(72)
+
t
Pt St =
t
where
At =
t
Pu Xu e(fuft)Xt du,
(73)
c DC Brody 2010
-31 -
18 June 2010
(74)
where and > 0 are constants, and {t} is a standard Q-Brownian motion.
Assuming for simplicity that the short rate {rt} is also a constant given by r, we
have, for the cumulative dividend, the expression
Z = X0
1 2 )s
es(r+ 2
ds.
(75)
(76)
c DC Brody 2010
-32 -
18 June 2010
2X0
2
z 1e /(2X0z)
,
()
(77)
where = 1 + 2 2(r ).
The price process of the asset is then obtained by defining the information
process {t} of (18) according to
t = tZ + Bt.
Zt = X0
s (r+ 12 2)s
ds.
(78)
(79)
(80)
c DC Brody 2010