SF2975 Financial Derivatives - Exercise session 5

Assume we have some 𝑇 −claim 𝜑 (𝑋𝑇 ) that we want to price. If we are in an incomplete market,
there are many possible Q−measures that would generate no arbitrage via their pricing function
 
𝜑 (𝑋𝑇 )
Π(𝜑 (𝑋𝑇 )) = E 𝐵𝑡
Q
|F𝑡 (1)
𝐵𝑇
Take a fixed Q. It is given by a Girsanov transformation.

𝑑𝑊𝑡P = 𝑑𝑊𝑡Q − 𝜆𝑑𝑡 ⇒ 𝑑𝑋𝑡 = 𝜇𝑑𝑡 + 𝜎 (𝑑𝑊𝑡Q − 𝜆𝑑𝑡) = (𝜇 − 𝜆𝜎)𝑑𝑡 + 𝜎𝑑𝑊𝑡Q

Assume that Π𝑡 (𝜑 (𝑋𝑇 )) = 𝐹 (𝑡, 𝑋𝑡 ). Then we get by FK that F solves

𝐹𝑡 + (𝜇 − 𝜆𝜎)𝐹𝑥 + 12 𝜎 2 𝐹𝑥𝑥 − 𝑟 𝐹 = 0
(

𝐹 (𝑇 , 𝑥) = 𝜑 (𝑥)

Chapter 9: A primer on Incomplete Markets1

Exercise 9.1
Consider a claim 𝜑 (𝑋𝑇 ) with pricing function 𝐹 (𝑡, 𝑥). Prove that the process 𝐹 (𝑡, 𝑋𝑡 )/𝐵𝑡 is a
𝑄−martingale.
Solution:
By Π𝑡 (𝜑 (𝑋𝑇 )) = 𝐹 (𝑡, 𝑋𝑡 ) and 𝜑 (𝑋𝑇 ) = 𝐹 (𝑇 , 𝑋𝑇 ) in (1) we get
 
Q 𝜑 (𝑋𝑇 )
𝐹 (𝑡, 𝑋𝑡 ) = E 𝐵𝑡 |F𝑡
𝐵𝑇
 
𝐹 (𝑡, 𝑋𝑡 ) Q 𝜑 (𝑋𝑇 )
=E |F𝑡
𝐵𝑡 𝐵𝑇
𝐹 (𝑡,𝑋𝑡 )
So 𝐵𝑡 is a martingale under Q and it follows that 𝐹 (𝑡, 𝑋𝑡 ) must have drift 𝑟 .

1 Bjork, T. (2020). Arbitrage Theory in Continuous Time. Oxford University Press, USA.

1
Exercise 9.3
Consider the same model and a fixed claim Γ(𝑋𝑇 ). Choose an arbitrary market price of risk
of the form 𝜆(𝑡, 𝑥) and define the pricing function 𝐺 (𝑡, 𝑥) as the solution of the corresponding
pricing PDE. Assume that the volatility function 𝜎𝐺 (𝑡, 𝑥) is non-zero. We now expect the market
(𝐵, 𝐺) to be complete. Show that this is indeed the case, i.e. show that every simple claim of
the form 𝜑 (𝑋𝑇 ) can be replicated by a portfolio based on 𝐵 and 𝐺.

Solution:
We have that the traded asset 𝐺𝑡 = 𝐺 (𝑡, 𝑋𝑡 ). We get via Ito that

𝜕𝐺 1 𝜕 2𝐺
 
𝜕𝐺 𝜕𝐺
𝑑𝐺𝑡 = +𝜇 + 2
𝑑𝑡 + 𝜎 𝑑𝑊𝑡P
𝜕𝑡 𝜕𝑥 2 𝜕𝑥 𝜕𝑥
2
 
𝜕𝐺 𝜕𝐺 1𝜕 𝐺
𝜕𝑡 + 𝜇 𝜕𝑥 + 2 𝜕𝑥 2
𝜕𝐺
= 𝐺𝑡 𝑑𝑡 + 𝜎 𝜕𝑥 𝐺𝑡 𝑑𝑊𝑡P
𝐺𝑡 𝐺𝑡
P
=𝛼𝐺 𝐺𝑡 𝑑𝑡 + 𝜎𝐺 𝐺𝑡 𝑑𝑊𝑡

We can use the FTAP II which says that the market is complete iff there exists a unique mar-
tingale measure Q. This means that we want 𝐺𝐵𝑡𝑡 to be a martingale i.e. we want 𝐺𝑡 to have drift
𝑟.
𝑑𝐺𝑡 = 𝛼𝐺 𝐺𝑡 𝑑𝑡 + 𝜎𝐺 𝐺𝑡 𝑑𝑊𝑡P = 𝑟𝐺𝑡 𝑑𝑡 + 𝜎𝐺 𝐺𝑡 𝑑𝑊𝑡Q
∫ 𝑡
Q P 𝑟 − 𝛼𝐺 Q P 𝑟 − 𝛼𝐺
⇒ 𝑑𝑊𝑡 = 𝑑𝑊𝑡 − 𝑑𝑡 ⇒ 𝑊𝑡 = 𝑊𝑡 − 𝑑𝑠
𝜎𝐺 0 𝜎𝐺
n∫ o
1 𝑡 2
So a Q that works is given by the Girsanov kernel 𝜑 = 𝑟 −𝛼 i.e. exp
𝑑Q 𝑡 ∫
𝜎𝐺
𝐺
|
𝑑P F𝑡 = 0 𝑠
𝜑 𝑑𝑊𝑠 − 𝜑
2 0 𝑠 𝑑𝑠
and it is unique. By FTAP II, [𝐵, 𝐺] is thus complete.
Exam 2021-10-25 Exercise 1b

Consider the standard Black Scholes model with one riskless and one risky asset and fixed
parameters 𝑟, 𝛼, 𝜎 > 0. Let 𝜎 = 𝑇 = 1 and 𝑟 = 12 . Find today’s arbitrage-free price of the claim

max 𝑆𝑡 − 𝑆𝑇
0≤𝑡 ≤𝑇

Solution:

max 𝑆𝑡 − 𝑆𝑇 = 𝑆¯𝑇 − 𝑆𝑇
0≤𝑡 ≤𝑇
Consindering the BS model
𝜎2 Q Q
𝑆𝑡 = 𝑆 0𝑒 (𝑟 − 2 )𝑡+𝜎𝑊𝑡 = 𝑆 0𝑒 𝜎𝑊𝑡

2
 ¯ −𝑟𝑇 Q ¯
 
Q max0≤𝑡 ≤1 𝑊𝑡Q Q 𝑊𝑇Q
Π0 (𝑆𝑇 − 𝑆𝑇 ) = 𝑒 E [𝑆𝑇 − 𝑆𝑇 ] = 𝑆 0𝑒 −𝑟𝑇
E [𝑒 ] − E [𝑒 ]
Now since 𝑇 = 1
1 2 1
EQ [𝑒𝑊𝑇 ] = EQ [𝑒𝑊1 ] = 𝑒 2 1 = 𝑒 2
Q Q

as 𝑊1Q ∼ N (0, 1). While

¯
EQ [𝑒 max0≤𝑡 ≤1 𝑊𝑡 ] = EQ [𝑒𝑊1 ]
Q

Formula sheets gives that 𝑊¯ 1 has cdf

(
Φ(𝑥) − Φ(−𝑥), 𝑥>0
𝐹𝑊¯ 1 (𝑥) = P(𝑊¯ 1 ≤ 𝑥) =
0 𝑥≤0

pdf is then
( (
𝑑 𝜙 (𝑥) − 𝜙 (−𝑥)(−1), 𝑥>0 2𝜙 (𝑥), 𝑥 > 0
𝑓𝑊¯ 1 (𝑥) = 𝐹𝑊¯ 1 (𝑥) = =
𝑑𝑥 0 𝑥≤0 0 𝑥≤0
∞ ∞
1 1
∫ ∫ ∫ ∞ 2
∫ ∞
𝑊¯ 1 1 2
Q
E [𝑒 ] = 𝑥
𝑒 𝑓𝑊¯ 1 (𝑥)𝑑𝑥 = 𝑒 2𝜙 (𝑥)𝑑𝑥 = 2 √
𝑥
𝑒 𝑒 𝑑𝑥 = 2 √
𝑥 − 𝑥2
𝑒 − 2 (𝑥 −2𝑥)𝑑𝑥
0 2𝜋 2𝜋 0
−∞
∫ ∞ 0
1 1 1
∫ ∞ ∫ ∞
− 12 (𝑥 2 −2𝑥+1)+ 21 1 1 2 1 1 2
= 2√ 𝑒 𝑑𝑥 = 2𝑒 2 √ −
𝑒 2 (𝑥−1)
𝑑𝑥 = 2𝑒 2 √ 𝑒 − 2 𝑦 𝑑𝑦
2𝜋 0 2𝜋 0 2𝜋 −1
∫ ∞
1 1
= 2𝑒 2 𝜙 (𝑦)𝑑𝑦 = 2𝑒 2 Φ(1)
−1
where we completed the square, changed the variable as 𝑦 = 𝑥 − 1 and used the symmetries of
𝜙: 𝜙 (𝑥) = 1 − 𝜙 (−𝑥).
Plugging everything together
1 1 1
Π 0 (𝑆¯𝑇 − 𝑆𝑇 ) = 𝑒 −𝑟𝑇 EQ [𝑆¯𝑇 − 𝑆𝑇 ] = 𝑆 0𝑒 − 2 (2𝑒 2 𝜙 (1) − 𝑒 2 ) = 𝑆 0 (2Φ(1) − 1)

Exam 2020-10-29 Exercise 3

Consider the market consisting of a domestic bank account 𝐵 (denoted in SEK) and a foreign
stock 𝑆 (denoted in EUR) where the currencies are related via the exchange rate 𝑋 (denoting
SEK/EUR). Let 𝐵, 𝑆, and 𝑋 follow the dynamics

𝑑𝐵𝑡 = 𝑟𝐵𝑡 𝑑𝑡, 𝐵 0 > 0,

𝑑𝑆𝑡 = 𝛼𝑆𝑡 𝑑𝑡 + 𝜎𝑆𝑡 𝑑𝑊𝑡 , 𝑆 0 > 0,
𝑑𝑋𝑡 = 𝛼𝑋 𝑋𝑡 𝑑𝑡 + 𝜎𝑋 𝑋𝑡 𝑑𝑊𝑡 , 𝑋 0 > 0,

where 𝑊 is a Brownian motion and 𝑟, 𝛼, 𝜎, 𝛼𝑋 , 𝜎𝑋 > 0 are given constants. This model is free
of arbitrage and complete.

3
 a) Provide today’s no-arbitrage price of the claim

𝑋 = (ln 𝑋𝑇 − ln 𝑋 0 ) 2 .

b) Suppose that the dynamics of 𝑋 is instead given by

𝑑𝑋𝑡 = 𝛼𝑋 𝑋𝑡 𝑑𝑡 + 𝜎𝑋 𝑋𝑡 𝑑𝑊
e𝑡 ,

where 𝑊 and 𝑊 e are two independent Brownian motions. Determine whether or not the
market is free of arbitrage and whether it is complete or incomplete.
c) Consider the market set out in b). Depending on your answer in b), address one of the
following two questions (only one is appropriate):

• Provide an arbitrage portfolio.

• Consider the claim 𝑋 given in a) and provide the smallest value 𝑥¯ such that the
following holds for any 𝑥 > 𝑥:¯ The possibility of buying today the claim 𝑋 for the
price 𝑥 does not enable the buyer to create arbitrage.

Solution:

a) We choose to express all prices in the domestic currency; writing 𝑆˜𝑡 = 𝑆𝑡 𝑋𝑡 , we obtain
the following description of our market:
𝑑𝐵𝑡 = 𝑟𝐵𝑡 𝑑𝑡,
𝑑 𝑆˜𝑡 = 𝑆˜𝑡 ((𝛼 + 𝛼𝑋 )𝑑𝑡 + (𝜎 + 𝜎𝑋 )𝑑𝑊𝑡 ) .

Using Itô’s lemma, we get

𝑑 𝑆˜𝑡 𝑆˜𝑡
= ((𝛼 + 𝛼𝑋 + 𝜎𝜎𝑋 − 𝑟 )𝑑𝑡 + (𝜎 + 𝜎𝑋 )𝑑𝑊𝑡 ) .
𝐵𝑡 𝐵𝑡

We thus see that the unique (equivalent) measure 𝑄 making the discounted prices a mar-
tingale is the measure under which
𝑄
𝑊𝑡 = 𝑊𝑡 − 𝜑𝑡

is a Brownian motion, where

𝑟 − (𝛼 + 𝛼𝑋 + 𝜎𝜎𝑋 )
𝜑= .
𝜎 + 𝜎𝑋

By substituing:
1 2 1 2
𝑋𝑡 = 𝑋 0𝑒 𝜎𝑋 𝑊𝑡 + (𝛼𝑋 − 2 𝜎𝑋 )𝑡 = 𝑋 0𝑒 𝜎𝑋 𝑊𝑡 + (𝛼𝑋 − 2 𝜎𝑋 +𝜑𝜎𝑋 )𝑡 .
𝑄

4
 Hence, the price of the given claim is given by

Π 0 (𝑋 ) = 𝑒 −𝑟𝑇 E𝑄 (ln 𝑋𝑇 − ln 𝑋 0 ) 2
 

"  2#
1

= 𝑒 −𝑟𝑇 E𝑄 𝜎𝑋 𝑊𝑇 + 𝛼𝑋 − 𝜎𝑋2 + 𝜑𝜎𝑋 𝑇
𝑄
2
2 !
1

= 𝑒 −𝑟𝑇 𝜎𝑋2 𝑇 + 𝛼𝑋 − 𝜎𝑋2 + 𝜑𝜎𝑋 𝑇 2 ,
2
where 𝜑 is given by (2).
b) We choose again to express all prices in the domestic currency; writing 𝑆˜𝑡 = 𝑆𝑡 𝑋𝑡 , we
obtain the following description of our market:

𝑑𝐵𝑡 = 𝑟𝐵𝑡 𝑑𝑡,

˜ ˜
 
𝑑 𝑆𝑡 = 𝑆𝑡 (𝛼 + 𝛼𝑋 + 𝜎𝜎𝑋 )𝑑𝑡 + 𝜎𝑑𝑊𝑡 + 𝜎𝑋 𝑑𝑊𝑡 .
e

At this point, one might recognize the structure of the volatility matrix of this model and
deduce from one of the theorems we derived in class that the model is free of arbitrage
and incomplete.
Alternatively, we go ahead and search for martingale measures. To this end, we again
choose to use the bank account 𝐵 as the numeraire. We then have that

𝑑 𝑆˜𝑡 𝑆˜𝑡  
= (𝛼 + 𝛼𝑋 − 𝑟 )𝑑𝑡 + 𝜎𝑑𝑊𝑡 + 𝜎𝑋 𝑑𝑊𝑡 .
e
𝐵𝑡 𝐵𝑡

We thus see that if 𝜑 and 𝜑˜ are chosen such that

𝜎𝜑 + 𝜎𝑋 𝜑˜ = 𝑟 − (𝛼 + 𝛼𝑋 ),

and if 𝑄 makes 𝑊𝑡 − 𝜑𝑡 and 𝑊 ˜ independent Brownian motions, then the discounted

e𝑡 − 𝜑𝑡
prices are martingales under that 𝑄. We note that such an (equivalent) measure exists for
any solution (𝜑, 𝜑)˜ to (3); since (3) admits infinitely many solutions (for any value of 𝜑,
˜ the market is thus free of arbitrage and incomplete.
there is a feasible choice of 𝜑),
c) We are in the market set out of point b), so the market is free of arbitrage but not com-
plete. This means that we cannot find an arbitrage but we have an infinite amount of
EMM. The value 𝑥¯ we are looking for is the threshold price below which an arbitrage op-
portunity exists for the claim X. It represents the minimum fair value of the claim based
on its expected payoff, taking into account the risks and the incompleteness of the market.

5
Recall that a martingale measure (or risk-neutral measure) is a probability measure un-
der which 𝑆𝑡 /𝐵𝑡 behaves like a martingale. In simple terms, this means that under this
measure, the expected future prices of assets, when discounted to the present, equal their
current prices. To ensure that the claim is priced correctly without creating arbitrage
opportunities, we can minimize the expected discounted payoff across all possible mar-
tingale measures Q. This leads us to find the smallest price 𝑥¯ that satisfies the no-arbitrage
condition:

𝑥¯ = inf 𝑒 −𝑟𝑇 E𝑄 (ln 𝑋𝑇 − ln 𝑋 0 ) 2 .

 
EMM 𝑄

In fact, if the price chosen for the claim 𝑋 is smaller than the minimum price 𝑥¯ (the
smallest value ensuring no arbitrage), it creates an arbitrage opportunity. Here’s what a
trader can do

• Buy the claim 𝑋 at a low price.

• Hedge by creating a portfolio that mirrors the payoff of the claim 𝑋 in the future.
The trader knows that the actual expected value of the claim under the worst-case
¯ but they are paying less than this fair
(most conservative) martingale measure is 𝑥,
price.
• Receive the full payoff of the claim at time 𝑇 and profit from the difference be-
tween the true value and the lower purchase price.

Let’s say:

• The minimum fair price 𝑥¯ is determined to be 100 SEK.

• The trader manages to buy the claim for 90 SEK (which is below the minimum).

The trader knows that, under the worst possible scenario, the claim will still be worth at
least 100 SEK. So, by paying 90 SEK today, they are guaranteed a profit of 10 SEK (i.e.,
100 − 90) at time 𝑇 . This profit is risk-free and without requiring any additional capital,
hence an arbitrage opportunity.
Going back to the exercise, we note that there is a one-to-one correspondence between
martingale measures and choices of 𝜑; we therefore write 𝑄 𝜑 for the measure under which

(and 𝑊𝑡 − 𝜑𝑡 with 𝜑 specified by (3))

𝜑
𝑊𝑡 = 𝑊𝑡 − 𝜑𝑡

is a Brownian motion. We note that

1 2 1 2
𝑋𝑡 = 𝑋 0𝑒 𝜎𝑋 𝑊𝑡 + (𝛼𝑋 − 2 𝜎𝑋 )𝑡 = 𝑋 0𝑒 𝜎𝑋 𝑊𝑡 + (𝛼𝑋 − 2 𝜎𝑋 +𝜑𝜎𝑋 )𝑡 .
𝜑

Hence,
𝑥 = inf 𝑒 −𝑟𝑇 E𝑄 (ln 𝑋𝑇 − ln 𝑋 0 ) 2
𝜑  
𝜑 ∈R

6
 "  2#
1

= inf 𝑒 −𝑟𝑇 E𝑄 𝜎𝑋 𝑊𝑇 + 𝛼𝑋 − 𝜎𝑋2 + 𝜑𝜎𝑋 𝑇
𝜑 𝜑
𝜑 ∈R 2
2 !
1

= inf 𝑒 −𝑟𝑇 𝜎𝑋2 𝑇 + 𝛼𝑋 − 𝜎𝑋2 + 𝜑𝜎𝑋 𝑇 2
𝜑 ∈R 2

= 𝑒 −𝑟𝑇 𝜎𝑋2 𝑇 .