MFE230Q: Final Exam, May 21, 2012

NAME: ID:

Please motivate your answers. Please underline your final answers.

Question Score
1
2
3
4
Total:

1. Discrete model (25 points): Consider the following three-date market, with three assets, B,
S, P , which can be interpreted as a risk-free bond, a stock, and a put option on the stock,
respectively. The assets are traded at t = 0, 1, 2. There are five states, ω1 , . . . , ω5 . At time
t = 1, at which points the assets are traded, it is determined whether the economy is in a
boom (u) or in a recession (d). Given that the economy is in a boom, the stock price can move
up (in state ω1 ) or down (in state ω2 ). In either case the option expires out of the money. If
the economy is in a recession, on the other hand, there are three possible outcomes for the
stock. It can go up, moderately down, or significantly down to the point that the underlying
firm defaults and becomes worthless. These are states ω3 , ω4 , and ω5 , respectively. In states
ω4 and ω5 , the option expires in the money. The probabilities for the different states are
P(ω1 ) = 0.1, P(ω2 ) = 0.2, P(ω3 ) = 0.5, P(ω4 ) = 0.1, and P(ω5 ) = 0.1. The prices of the three
assets in different states and times are summarized below.
B(0) S(0) P(0) B(1) S(1) P(1) B(2) S(2) P(2) State
 144 240 0 ω1
u 120 150 0
  144 120 0 ω2
100 100 20
  120 200 0 ω3
d 120 100 40 → 120 100 20 ω4
 120 0 120 ω5

1
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
(a) Define the filtration for this market.

(b) Is the market complete?

(c) An insurance company is considering introducing a credit default swap on the firm’s
default risk. Specifically, they would sell an insurance contract that pays a hundred
dollars at t = 2 in the case that the (stock-S) firm defaults (in state ω5 ), and makes no
payment otherwise. What is the t = 0 market price of such a credit default swap?

(d) Assume that the insurance company wishes to hedge the risk of the credit default swap
in the market. How could it do this with a dynamic portfolio trading strategy?

Solution:

(a)

F0 = {∅, Ω},
F1 = {∅, {ω1 , ω2 }, {ω3 , ω4 , ω5 }, Ω},
F2 = 2Ω .

(b) Remember from class that the (finite horizon) multi-period market is complete if is
complete period-by-period and state-by-state. It follows immediately that the market is
complete between t = 0 and t = 1, and that it is complete at t = 1 after an up move (in
state {ω1 , ω2 }), since two states are spanned by one risky and one risk-free asset. What
remains is to show that the market is complete at t = 1 after a down move. Since there
are three states and three assets, a necessary and sufficient condition is that the payoff
matrix ⎡ ⎤
120 120 120
D = ⎣ 200 100 0 ⎦
0 20 120
nonsingular. An easy way to verify nonsingularity is to check that the determinant is
nonzero, |D| = 0. Specifically, for the determinant of this 3 × 3 matrix, we have:
⎡ ⎤
 d11 d12 d13       
   d12 d13   d11 d13   d11 d12 
|D| = ⎣ d21 d22 d23 ⎦ = d31   − d32   
 d21 d23  + d33  d21 d22 

 d31 d32 d33  d22 d23 

= 0 − 20(120 × 0 − 200 × 120) + 120(120 × 100 − 200 × 120)

= −960, 000 = 0.

So, the market is indeed complete.

2
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
(c) Compute the state prices between period 0 and 1

ψu 1/3
= ,
ψd 1/2

and the state prices between period 1 and 2 in the down state in period 1
⎛ ⎞ ⎛ ⎞
ψω3 1/4
⎝ψω4 ⎠ = ⎝1/2⎠ ,
ψω5 1/4

Then the price of the CDS is equal to ψd × ψω5 × 100 = 1/2 × 1/4 × 100 = 12.5.
Alternatively, you could use to results from (d) to immediately determine the price of
the portfolio as hP0 × P (0) = 5/8 × 20 = 12.5.
d,P 
(d) After a down move, the portfolio (hd,B d,S
1 , h1 , h1 ) should be chosen such that the payoffs
in states ω3 , ω4 and ω5 should be 0, 0, and 100, respectively. This leads to the equations

120hd,B
1 + 200hd,S d,P
1 + 0h1 = 0
120hd,B
1 + 100hd,S d,P
1 + 20h1 = 0
120hd,B
1 + 0hd,S
1 + 120hd,P
1 = 100

From the third equation, 1.2hd,P

1 = 1 − 1.2hd,B
1 . Similarly, from the first equation,
2hd,S
1 = −1.2hd,B d,B
1 . Plugging these into the second equation leads to h1 = −5/12,
which in turn leads to

d,P  
(hd,B d,S
1 , h1 , h1 ) = (−5/12, 1/4, 5/4) .

The price of this portfolio at t = 1 is −5/12 × 120 + 1/4 × 100 + 5/4 × 40 = 25.
The payoffs in states ω1 and ω2 are 0, so trivially, a replicating portfolio after an up
move is
u,P  
(hu,B u,S
1 , h1 , h1 ) = (0, 0, 0)

(since the price of the option is 0 after an up move, arbitrary other positions in the
option are of course also possible).
Now, at 0 a hedging portfolio (hB S P 
0 , h0 , h0 ) should be chosen such that the payoffs 0
and 25 should be replicated after an up and down move, respectively. Since there are
more assets than states, there are again several ways of doing this, but since the option
between t = 0 and t = 1 is basically an Arrow-Debreu security on the down state, it is
clear that
P 
(hB S
0 , h0 , h0 ) = (0, 0, 5/8)

is an especially simple such portfolio.

3
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
2. Black-Scholes (20 points): Consider the standard Black-Scholes economy with a risky and a
risk-free asset,
dBt
= rdt,
Bt
dSt
= μ̂dt + dW.
St
All the standard assumptions (no transaction costs, no arbitrage, etc.) are satisfied. Assume
that an investor wishes to create a simple contingent claim with payoff Φ(ST ) at time T , by
using dynamic portfolio trading.

(a) Formulate a dynamic, self-financing, trading strategy, ht = (hB S 

t , ht ) , 0 ≤ t ≤ T , that
allows the investor to replicate the payoff of the contingent claim, by trading in the bond
and the stock.
(b) Prove that the trading strategy in (a) does indeed replicate the payoff of the contingent
claim.
(c) What is the time-0 price in this market of a so-called “power” contingent claim, that
makes the terminal payoff Φ(ST ) = ST2 ?

Solution:

(a) This is basically Theorem 8.5 in Björk (see also the no-arbitrage section in part 1.2 of
the slides). The theorem states that if F solves the PDE
1
Ft + rSFS + σ 2 S 2 FSS − rF = 0,
2
F (T, S) = Φ(S),

then the portfolio strategy


S  F − SFS
ht = (hB
t , ht ) = , FS
B
is self-financed, has value V (t, St ) = F (t, St ), and replicates the payoff VT = Φ(ST ) at
T.

t Bt + ht St = F − SFS + SFS = F ,
(b) That V (t, St ) = F (t, St ) follows immediately from hB S

which of course also ensures that the final payout is correct, VT = V (T, ST ) = F (T, ST ) =
Φ(ST ).
What remains to show is that the strategy is self financed, i.e., that the instantaneous
cash flows dF h generated by the portfolio is 0. We have (see, e.g., the continuous time
portfolio model section in Part 1.2 of slides)
F − SFS
dF h = ht (dB, dS) − dV = dB + FS dS − dV.
B

4
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
 Now, since V = F , it follows that dV = dF = Ft dt + FS dS + 12 σ 2 S 2 FSS dt, which when
plugged into the equation yields

h F − SFS 1 2 2
dF = dB + FS dS − Ft dt + FS dS + σ S FSS dt
B 2

1 2 2
= (F − SFS )rdt − Ft dt + σ S FSS dt
2
1
= −(Ft + rSFS + σ 2 S 2 FSS − rF )dt
2
= −0 × dt.

So, the portfolio strategy is indeed self financed and the proposition follows.
(c) The easiest way to calculate the price is by using the risk neutral expectation:

   
P0 = E0Q e−rT Φ(ST ) = e−rT E0Q ST2 .

We rewrite ST = S0 eyT , where under the risk neutral probability measure yT ∼ N ((r −
σ 2 /2)T, σ 2 T ). Also, we define xT = 2yT ∼ N (2(r − σ 2 /2)T, 4σ 2 T ) and we then have
ST2 = S02 exT . Standard formulas of log-normal distributions imply that
2 /2)T + 1 4σ 2 T 2T
E0Q [exT ] = e2(r−σ 2 = e2rT +σ .

So, in total we get

  2
P0 = e−rT E0Q ST2 = e−rT S02 e2rT +σ T
2 )T
= S02 e(r+σ .

Note that this is a higher price than the “naive” price P0 = S02 , which would be the price of
buying S0 shares of the stock at t = 0. Such a strategy would lead to the terminal payoff
S0 ST , not ST2 .

One could also find the solution by solving the Black-Scholes PDE in (a). Specifically, we
2 )(T −t)
verify that the function F (t, St ) = St2 e(r+σ satisfies the PDE in (a). We have P0 =
F (0, S0 ) = S02 e (r+σ2 )T , and PT = F (T, ST ) = ST2 . Further, the PDE is satisfied, since
1 2 )(T −t) 1 2
Ft + rSFS + σ 2 S 2 FSS − rF = −(r + σ 2 )F + rS2Se(r+σ + σ 2 S 2 2e(r+σ )(T −t) − rF
2 2
= −(r + σ 2 )F + 2rF + σ 2 F − rF
= 0.

Thus, F is a solution to the PDE, and P0 is therefore the price of the power claim.

5
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
3. Dividends (20 points): Consider the “Black Scholes” economy where the stock pays a constant
dividend yield δ.
dB
= r dt, r > 0,
B
dS + δS dt
= μ̂ dt + σ dW.
S
where μ̂, σ, and r are all constant. Now, consider a call option that has no maturity date,
but has strike K and will be exercised the first time the stock price reaches S ∗ . Hence, the
cash flow when this first hitting time occurs is (S ∗ − K) Here, S ∗ has been chosen such that
it is greater than K.

(a) Determine the differential equation that the value of this call, C(S), satisfies.
(b) Solve for the call price.
(c) Now assume that the buyer of the option is allowed to choose for herself the S ∗ at which
she will elect to exercise it. Determine this optimal S ∗ .

Solution:

(a) Under the risk-neutral measure, the stock price process follows

dS = (r − δ) S dt + σ S dz.

Since the call pays no dividend, we have

rC = EQ
t
[dC]
σ2 2
= Ct + (r − δ) SCS + S CSS . (1)
2
However, since there is no explicit time-dependence in the state variable dynamics, and
no explicit time-dependence in the payoff, it follows that this call will have the same
value each time the same value of S is reached. It thus follows that Ct = 0, implying
that its dynamics reduce to
σ2 2
rC = (r − δ) SCS + S CSS . (2)
2
(b) Assuming C(S) ∼ S α , we find

σ2 α
rS α = (r − δ) αS α + α(α − 1) S (3)
2
or equivalently that

σ2 σ2
0= α2 + α(r − δ − )−r (4)
2 2

6
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
 with solutions

  
1 σ2 σ2 2
α± = −(r − δ − ) ± (r − δ − ) + 2rσ 2 . (5)
σ2 2 2

Note that since r > 0, the term inside the square root is larger than the term outside. As
such, we have α+ > 0, α− < 0. Furthermore, α+ = 1 when the dividend payout δ = 0.
∂α+
Moreover, it is straightforward to demonstrate that α+ is increasing in δ in that ∂δ > 0.

Thus, we know that the call price is of the form

C(S) = AS α+ + BS α− (6)

The boundary conditions are

C(S = 0) = 0 (7)
C(S = S ∗ ) = S ∗ − K (8)

Solving for A and B, we find

α+
∗ S
C(S) = (S − K) . (9)
S∗

(c) We solve the first-order condition:

∂C
0 =
∂S ∗
= (1 − α+ )(S ∗ ) −α+ + α+ K(S ∗ )−α+ −1 (10)

implying that

∗ α+
S =K (11)
α+ − 1

As a side note: recall that α+ > 1 when δ > 0, but approaches one as δ ⇒ 0. Thus, as
δ ⇒ 0, we find S ∗ ⇒ ∞, implying that it is always better to wait. This is consistent with
the fact that, for finite maturity American call options, it is never optimal to exercise
early if the dividend is zero.

4. Term structure (20 points): Assume that the short rate follows the asset pricing dynamics
specified by the CIR model:
√
drt = a(b − r)dt + σ rdW Q .

7
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
This model belongs to the class of affine term structure models, implying that the price of a
T -bond is on the form
p(t, T |rt ) = eA(T −t)−B(T −t)rt .

(a) State the ODEs that determine the functions B(·) and A(·), respectively.
(b) Verify that the functions
2(eqx − 1)
B(x) = ,
(q + a)(eqx − 1) + 2q
 
2ab 2qe(q+a)x/2
A(x) = ln ,
σ2 (q + a)(eqx − 1) + 2q
√
where q = a2 + 2σ 2 , solve the ODEs stated in (a).

Solution:

(a) Using the same derivation as in class (i.e., plugging in the conjectured form of the solution
into the one-factor term structure PDE), it follows that A and B need to satisfy the two
ODEs:
1
B  + Ba + B 2 σ 2 = 1,
2

−A − Bab = 0.

The boundary conditions are B(0) = 0, A(0) = 0.

(b) We begin by noting that given the conjectured solutions for B and A,
4q 2 eqx
B  (x) = ,
((q + a)(eqx − 1) + 2q)2
 
2ab q + a (q + a)qeqx
A (x) = − .
σ2 2 (q + a)(eqx − 1) + 2q
We first verify the first ODE.

2
4q 2 eqx 2(eqx − 1) 1 2(eqx − 1)
2 + a qx − 1) + 2q
+ σ2 =1
((q + a)(e − 1) + 2q)
qx (q + a)(e 2 (q + a)(eqx − 1) + 2q

1
4q 2 eqx + a2(eqx − 1)((q + a)(eqx − 1) + 2q) + σ 2 (2(eqx − 1))2 = ((q + a)(eqx − 1) + 2q)2
2
= (q + a)2 (eqx − 1)2 + (2q)2
+ 2(q + a)(eqx − 1)2q
1
2a(eqx − 1)((q + a)(eqx − 1) + 2q) + σ 2 (2(eqx − 1))2 = (q + a)2 (eqx − 1)2
2
+ 2a(eqx − 1)2q
1
(2aq + 2a2 )(eqx − 1)2 + σ 2 (2(eqx − 1))2 = (q 2 + 2qa + a2 )(eqx − 1)2
2
(2aq + 2a2 ) + 2σ 2 = (q 2 + 2qa + a2 )

8
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.
In the last equation we use the fact that q 2 = a2 + 2σ 2 and we are done. Now let’s verify
the second ODE:
 
2ab q + a (q + a)qeqx 2(eqx − 1)
− + ab =0
σ2 2 (q + a)(eqx − 1) + 2q (q + a)(eqx − 1) + 2q
 
q+a (q + a)qeqx σ 2 (eqx − 1)
− + =0
2 (q + a)(eqx − 1) + 2q (q + a)(eqx − 1) + 2q
q+a
((q + a)(eqx − 1) + 2q) − (q + a)qeqx + σ 2 (eqx − 1) = 0
2
(q + a)2 (eqx − 1) − 2(q + a)q(eqx − 1) + 2σ 2 (eqx − 1) = 0
(q 2 + 2qa + a2 )(eqx − 1) − 2(q + a)q(eqx − 1) + 2(q 2 − a2 )(eqx − 1) = 0
(q 2 + 2qa + a2 ) − 2(q + a)q + (q 2 − a2 ) = 0
0 = 0.

2ab
Finally, it is immediately verified that B(0) = 0/(2q) = 0, and A(0) = σ2
ln(1) = 0. We
are done.

9
This document is authorized for use by Geoffrey Simmons, from 1/1/2013 to 6/22/2013, in the course:
MFE 230Q-1: Introduction to Stochastic Calculus (Spring 2013), University of California, Berkeley.
Any unauthorized use or reproduction of this document is strictly prohibited.