8/8/2019 FXX
1/68
8/8/2019 FXX
2/68
8/8/2019 FXX
3/68
8/8/2019 FXX
4/68
8/8/2019 FXX
5/68
8/8/2019 FXX
6/68
8/8/2019 FXX
7/68
8/8/2019 FXX
8/68
8/8/2019 FXX
9/68
8/8/2019 FXX
10/68
8/8/2019 FXX
11/68
8/8/2019 FXX
12/68
8/8/2019 FXX
13/68
8/8/2019 FXX
14/68
8/8/2019 FXX
15/68
8/8/2019 FXX
16/68
8/8/2019 FXX
17/68
8/8/2019 FXX
18/68
8/8/2019 FXX
19/68
8/8/2019 FXX
20/68
8/8/2019 FXX
21/68
8/8/2019 FXX
22/68
8/8/2019 FXX
23/68
8/8/2019 FXX
24/68
8/8/2019 FXX
25/68
8/8/2019 FXX
26/68
8/8/2019 FXX
27/68
8/8/2019 FXX
28/68
8/8/2019 FXX
29/68
8/8/2019 FXX
30/68
k l f $
8/8/2019 FXX
31/68
Risk-Neutral Measure for $
Theorem: Let QA be the risk-neutral probabilitymeasure for the US Dollar investor, and QB therisk-neutral measure for the UK Pound Sterling investor.Unless = 0 (that is, unless the exchange rate is purelydeterministic), it must be the case that
QA = QB
Stochastic Calculus p. 14/2
Ri k N l M f $
8/8/2019 FXX
32/68
Risk-Neutral Measure for $
Theorem: Let QA be the risk-neutral probabilitymeasure for the US Dollar investor, and QB therisk-neutral measure for the UK Pound Sterling investor.Unless = 0 (that is, unless the exchange rate is purelydeterministic), it must be the case that
QA = QB
This is a special case of a more general phenomenon:
Stochastic Calculus p. 14/2
N i Ch g
8/8/2019 FXX
33/68
Numeraire Change
Suppose that a market has tradeable assets A, B withshare price processes S At and S Bt (evaluated in a common
numeraire C ). Let QA
and QB
be risk-neutral measuresfor numeraires A, B , respectively.
Stochastic Calculus p. 15/2
N meraire Change
8/8/2019 FXX
34/68
Numeraire Change
Suppose that a market has tradeable assets A, B withshare price processes S At and S Bt (evaluated in a common
numeraire C ). Let QA
and QB
be risk-neutral measuresfor numeraires A, B , respectively.
Theorem: QA = QB if and only if S At /S Bt is a constantrandom variable. Furthermore, in general, for any nitetime T ,
dQB
dQA F T =S BT S AT
S A0S B0
Stochastic Calculus p. 15/2
Consequence
8/8/2019 FXX
35/68
Consequence
In the foreign exchange context, the riskless assets forthe two numeraires are US Money Market and UKMoney Market, with share prices (in $)
At = exp {r A t}B t = exp {r B t}/Y t
Stochastic Calculus p. 16/2
Consequence
8/8/2019 FXX
36/68
Consequence
In the foreign exchange context, the riskless assets forthe two numeraires are US Money Market and UKMoney Market, with share prices (in $)
At = exp {r A t}B t = exp {r B t}/Y t
Therefore, the likelihood ratio between the risk-neutralmeasures for and $ investors is
dQB
dQA F T =
Y T Y 0
1
exp{(r B r A)T }
Stochastic Calculus p. 16/2
Consequence
8/8/2019 FXX
37/68
Consequence
In the foreign exchange context, the riskless assets forthe two numeraires are US Money Market and UKMoney Market, with share prices (in $)
At = exp {r A t}B t = exp {r B t}/Y t
Therefore, the likelihood ratio between the risk-neutralmeasures for and $ investors is
dQB
dQA F T =
Y T Y 0
1
exp{(r B r A)T }
Stochastic Calculus p. 16/2
Likelihood Ratio Identity
8/8/2019 FXX
38/68
Likelihood Ratio Identity
Let V it be the time- t share price of any contingent claimin numeraire i = A,B,C . These share prices satisfy:
Stochastic Calculus p. 17/2
Likelihood Ratio Identity
8/8/2019 FXX
39/68
Likelihood Ratio Identity
Let V it be the time- t share price of any contingent claimin numeraire i = A,B,C . These share prices satisfy:
V At = V C t /S AtV Bt = V
C t /S
Bt
Stochastic Calculus p. 17/2
Likelihood Ratio Identity
8/8/2019 FXX
40/68
Likelihood Ratio Identity
Let V it be the time- t share price of any contingent claimin numeraire i = A,B,C . These share prices satisfy:
V At = V C t /S AtV Bt = V
C t /S
Bt
The time-zero share price is the discounted expectedvalue of the time t share price for each of thenumeraires A, B . The discount factors are 1, so
Stochastic Calculus p. 17/2
Likelihood Ratio Identity
8/8/2019 FXX
41/68
Likelihood Ratio Identity
Let V it be the time- t share price of any contingent claimin numeraire i = A,B,C . These share prices satisfy:
V At = V C t /S AtV Bt = V
C t /S
Bt
The time-zero share price is the discounted expectedvalue of the time t share price for each of thenumeraires A, B . The discount factors are 1, so
V A0 = V C 0 /S A0 = E AV C t /S At
V B
0 = V C
0 /S B0 = E
B
V C
t /S Bt
Stochastic Calculus p. 17/2
Likelihood Ratio Identity
8/8/2019 FXX
42/68
Likelihood Ratio Identity
It follows that for every contingent claim V with shareprice V C t (in numeraire C ),
S A0 E A(V C t /S At ) = S B0 E B (V C t /S Bt )
Stochastic Calculus p. 18/2
Likelihood Ratio Identity
8/8/2019 FXX
43/68
Likelihood Ratio Identity
It follows that for every contingent claim V with shareprice V C t (in numeraire C ),
S A0 E A(V C t /S At ) = S B0 E B (V C t /S Bt )
Apply this to the contingent claim with payoff V C T S BT attime T to obtain the following identity, valid for allnonnegative random variables V C T measurable F T :
E B V C T = E AV C T S
BT S A0S AT S B0
Stochastic Calculus p. 18/2
Likelihood Ratio Identity
8/8/2019 FXX
44/68
Likelihood Ratio Identity
It follows that for every contingent claim V with shareprice V C t (in numeraire C ),
S A0 E A(V C t /S At ) = S B0 E B (V C t /S Bt )
Apply this to the contingent claim with payoff V C T S BT attime T to obtain the following identity, valid for allnonnegative random variables V C T measurable F T :
E B V C T = E AV C T S
BT S A0S AT S B0
This is the dening property of a likelihood ratio.Stochastic Calculus p. 18/2
Exponential Martingales
8/8/2019 FXX
45/68
Exponential Martingales
Let W t be a standard Wiener process, wth Brownianltration F t , and let t be a bounded, adapted process.Dene
Z t = exp t
0s dW s
t
02s ds/ 2
Stochastic Calculus p. 19/2
Exponential Martingales
8/8/2019 FXX
46/68
Exponential Martingales
Let W t be a standard Wiener process, wth Brownianltration F t , and let t be a bounded, adapted process.Dene
Z t = exp t
0s dW s
t
02s ds/ 2
Fact: Z t is a positive martingale.
Stochastic Calculus p. 19/2
Exponential Martingales
8/8/2019 FXX
47/68
po e t a a t ga es
Let W t be a standard Wiener process, wth Brownianltration F t , and let t be a bounded, adapted process.Dene
Z t = exp t
0s dW s
t
02s ds/ 2
Fact: Z t is a positive martingale. Proof: It!
Stochastic Calculus p. 19/2
Exponential Martingales
8/8/2019 FXX
48/68
p g
Let W t be a standard Wiener process, wth Brownianltration F t , and let t be a bounded, adapted process.Dene
Z t = exp t
0s dW s
t
02s ds/ 2
Fact: Z t is a positive martingale. Proof: It!
dZ t = Z t t dW t Z t 2
t dt/ 2+ Z t 2
t dt/ 2= Z t t dW t =
Z t = Z 0 +
t
0 Z s s dW sStochastic Calculus p. 19/2
Girsanovs Theorem
8/8/2019 FXX
49/68
Girsanov s Theorem
Because Z t is a positive martingale under P with initialvalue Z 0 = 1 , for every xed time T the random variableZ T is a likelihood ratio: that is,
Q(F ) := E P (I F Z T )
denes a new probability measure on the algebra F T of events F that are observable by time T .
Stochastic Calculus p. 20/2
Girsanovs Theorem
8/8/2019 FXX
50/68
Girsanov s Theorem
Because Z t is a positive martingale under P with initialvalue Z 0 = 1 , for every xed time T the random variableZ T is a likelihood ratio: that is,
Q(F ) := E P (I F Z T )
denes a new probability measure on the algebra F T of events F that are observable by time T .Theorem: Under the measure Q, the process
{W t t
0 s ds}0 t T is a standard Wiener process.
Stochastic Calculus p. 20/2
Exchange Rates
8/8/2019 FXX
51/68
g
Consider again the $ and currencies. Assume that eachhas a riskless Money Market, and that the rates of returnr A , r B are constant. Assume that the exchange rate Y tobeys
dY t = ( r B r A)Y t dt + Y t dW t
where W t is a standard Wiener process under therisk-neutral probability QB for investors. Thus,
Y t = Y 0
exp{(r B r A 2
/ 2)t + W t }.
Stochastic Calculus p. 21/2
Exchange Rates
8/8/2019 FXX
52/68
g
Since
dQA
dQB F T =Y T Y 0 exp{ (r B r A)T }
= exp {W T 2 T/ 2}
Stochastic Calculus p. 22/2
Exchange Rates
8/8/2019 FXX
53/68
Since
dQA
dQB F T =Y T Y 0 exp{ (r B r A)T }
= exp {W T 2 T/ 2}
Girsanov implies that under QA the process W t is aWiener process with drift . Thus, to the $ investor, itappears that the exchange rate obeys
dY t = ( r B r A 2 )Y t dt + Y t dW t
where W t is a standard Wiener process under QA .Stochastic Calculus p. 22/2
8/8/2019 FXX
54/68
Stochastic Calculus p. 23/2
Proof of Girsanov 1
8/8/2019 FXX
55/68
The statement that X is a standard Wiener process is anassertion that the increments of X are independentGaussian random variables with the correct variances.Lets show that under Q, the distribution of W T T isgaussian with var T (where T =
T 0 s ds).
Stochastic Calculus p. 24/2
Proof of Girsanov 1
8/8/2019 FXX
56/68
The statement that X is a standard Wiener process is anassertion that the increments of X are independentGaussian random variables with the correct variances.Lets show that under Q, the distribution of W T T isgaussian with var T (where T =
T 0 s ds). For this, it
sufces to show that for any real ,E Q exp{(W T T )} = exp {
2 T/ 2}
Stochastic Calculus p. 24/2
Proof of Girsanov 1
8/8/2019 FXX
57/68
The statement that X is a standard Wiener process is anassertion that the increments of X are independentGaussian random variables with the correct variances.Lets show that under Q, the distribution of W T T isgaussian with var T (where T =
T 0 s ds). For this, it
sufces to show that for any real ,E Q exp{(W T T )} = exp {
2 T/ 2}
To evaluate the expectation, change measure:
E Q exp{(W T T )} = E P exp{(W T T )}Z T
Stochastic Calculus p. 24/2
Proof of Girsanov 2
8/8/2019 FXX
58/68
Objective: Show that E P H T = 1 , where
H t = exp {(W t t ) 2 t/ 2}Z t
Stochastic Calculus p. 25/2
Proof of Girsanov 2
8/8/2019 FXX
59/68
Objective: Show that E P H T = 1 , where
H t = exp {(W t t ) 2 t/ 2}Z t
= exp { t
0(s + ) dW s +
t
0(s
2s / 2
2 / 2) ds}
Stochastic Calculus p. 25/2
Proof of Girsanov 2
8/8/2019 FXX
60/68
Objective: Show that E P H T = 1 , where
H t = exp {(W t t ) 2 t/ 2}Z t
= exp { t
0(s + ) dW s +
t
0(s
2s / 2
2 / 2) ds}
= exp { t0 (s + ) dW s t0 (s + )2 ds/ 2}
Stochastic Calculus p. 25/2
Proof of Girsanov 2
8/8/2019 FXX
61/68
Objective: Show that E P H T = 1 , where
H t = exp {(W t t ) 2 t/ 2}Z t
= exp { t
0(s + ) dW s +
t
0(s
2s / 2
2 / 2) ds}
= exp { t0 (s + ) dW s t0 (s + )2 ds/ 2}Thus, H t is an exponential martingale under P , and soits expectation is constant over time.
Stochastic Calculus p. 25/2
Proof of Girsanov 2
8/8/2019 FXX
62/68
Objective: Show that E P H T = 1 , where
H t = exp {(W t t ) 2 t/ 2}Z t
= exp { t
0(s + ) dW s +
t
0(s
2s / 2
2 / 2) ds}
= exp { t0 (s + ) dW s t0 (s + )2 ds/ 2}Thus, H t is an exponential martingale under P , and soits expectation is constant over time. A similar
calculation establishes the independence of the
increments.Stochastic Calculus p. 25/2
Scratch
8/8/2019 FXX
63/68
Stochastic Calculus p. 26/2
Scratch
8/8/2019 FXX
64/68
Stochastic Calculus p. 26/2
Scratch
8/8/2019 FXX
65/68
Stochastic Calculus p. 26/2
Scratch
8/8/2019 FXX
66/68
Stochastic Calculus p. 27/2
Scratch
8/8/2019 FXX
67/68
Stochastic Calculus p. 27/2
Scratch
8/8/2019 FXX
68/68
Stochastic Calculus p. 27/2