Notes

Mathematics of Financial Derivatives

MA537 & MA837
Continuous Processes

Pradip Tapadar

[email protected]
Room 258, Sibson Building

Weeks 13 – 24
Spring Term

P Tapadar ([email protected]) MA537 & MA837 Spring Term 1 / 50

Continuous Processes Brownian Motion

What is a continuous process?

Three small-scale principles guide us:
1 Value can change at any time and from moment to moment.
2 Any real number can be taken as a value.
3 Process changes continuously without making instantaneous jumps.
FTSE100 (02-04-1984 – 13-02-2007):

Can we come up with a

model, which will produce
a share process with
similar characteristics?

P Tapadar ([email protected]) MA537 & MA837 Spring Term 2 / 50

 Continuous Processes Brownian Motion

Standard Brownian Motion Notes

Definition
Standard Brownian Motion, SBM, is a stochastic process
{Bt : t ≥ 0}, with the following defining properties:
1 B0 = 0.
2 Independent increments: Bt − Bs is independent of {Br : r ≤ s},
where s < t.
3 Stationary increments: Distribution of Bt − Bs depends only on
(t − s), where s < t.
4 Gaussian increments: Bt − Bs ∼ N(0, t − s).
5 Continuity: Bt has continuous sample paths.

Q: Can we use Standard Brownian Motion to model stock markets?

A: Possibly!

P Tapadar ([email protected]) MA537 & MA837 Spring Term 3 / 50

Continuous Processes Brownian Motion

Modelling Share Prices Using Brownian Motion

Modelling Stock Markets

In the next slide:
The first graph shows the plot of FTSE-100 from 02-Apr-1984 to
12-Feb-2007.
The second graph shows a simulation of SBM, Bt .
The thirds graph shows a simulation of Wt = W0 + σBt + µt.
The final graph shows a simulation of St = eWt . (St is also known as
Geometric Brownian Motion (GBM).)
Spot the similarities and differences between the graphs.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 4 / 50

 Continuous Processes Brownian Motion

Modelling Share Prices Using Brownian Motion Notes

Standard Brownian motion Brownian motion with drift and noise

0
0

Geometric Brownian motion FTSE 100

0 0

P Tapadar ([email protected]) MA537 & MA837 Spring Term 5 / 50

Continuous Processes Stochastic Calculus

Stochastic Calculus: Tools

Global to Local Behaviour: Ito’s Lemma

Let Xt be a stochastic process satisfying dXt = Yt dBt + Zt dt and let f (t, Xt )
be a real-valued function, twice partially differentiable with respect to x and
once with respect to t. Then:
" #
∂f ∂f ∂f 1 ∂2f 2
df (t, Xt ) = Yt dBt + + Zt + Y dt.
∂Xt ∂t ∂Xt 2 ∂Xt2 t

Local to Global Behaviour: Integration

If interval (a, b) is divided into n sub-intervals a = s0 < s1 < . . . < sn = b, then
Z b n
X
g(s)dh(s) = lim g(si−1 ) × (h(si ) − h(si−1 ))
a n→∞
i=1

P Tapadar ([email protected]) MA537 & MA837 Spring Term 6 / 50

 Continuous Processes Stochastic Calculus

Stochastic Calculus: Applications Notes

Building blocks in Stochastic calculus

(Global) Bt ⇔ dBu = Bu+du − Bu (Local)
(Global) Wt = W0 + σBt + µt ⇔ dWu =? (Local)
(Global) St = eWt ⇔ dSu =? (Local)

P Tapadar ([email protected]) MA537 & MA837 Spring Term 7 / 50

Continuous Processes 5-step method in continuous time

Theorem — The continuous counterpart.

Theorem (Self-financing hedging strategies in continuous time)

Given a stochastic process St for share prices and a cash or bond
process Bt = B0 ert for a constant continuously compounded risk-free
rate of interest r , then (φt , ψt ) is a self-financing strategy to construct a
derivative payoff X at maturity (time T ), if:
1 both φt and ψt are previsible;
2 the change in value Vt of the portfolio defined by the strategy
obeys:
dVt = φt dSt + ψt dBt ;
3 VT = φT ST + ψT BT is identically equal to the claim X ;
4 Vt is the fair price of the derivative at time t.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 8 / 50

 Continuous Processes 5-step method in continuous time

Proof outline Notes

Proof (5-step method in continuous time).

Step 1 Establish a measure Q for which Dt = e−rt St is a
Q-martingale.
Step 2 Define Vt = e−r (T −t) EQ [X |Ft ].
Step 3 Et = e−rt Vt = e−rT EQ [X |Ft ] is a Q-martingale.
Step 4 dEt = φt dDt : Martingale representation theorem.
Step 5 Define ψt = Et − φt Dt , and conclude the proof.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 9 / 50

Continuous Processes 5-step method in continuous time

Step 1: Tool: Martingales.

Definition
A continuous-time stochastic process Xt is said to be a martingale if:
 
1 E |Xt | < ∞ for all t.
 
2 E Xt Fs = Xs for all s < t.

Fact (Martingale and driftlessness)

A martingale is not expected to drift upwards or downwards.
Fact: If dXt = σXt dBt , then Xt is a martingale.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 10 / 50

 Continuous Processes 5-step method in continuous time

Step 1: Tool: Cameron-Martin-Girsanov. Notes

Theorem (Cameron-Martin-Girsanov theorem)

If Bt is a P-Brownian motion and γt is a previsible process, then there exists
an equivalent measure Q such that B̃t given by:

d B̃t = dBt + γt dt

is a Q-Brownian motion. In other words, Bt is a drifting Q-Brownian motion

with drift −γt at time t.

Definition (Equivalence of measure)

Two measures P and Q are equivalent if they operate on the same sample
space and agree on what is possible. So, if A is any event in the sample
space,
P(A) > 0 ⇔ Q(A) > 0.
In other words, if A is possible under P then it is possible under Q, and if A is
impossible under P then it is also impossible under Q. And vice versa.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 11 / 50

Continuous Processes 5-step method in continuous time

Step 1: In search of Q for Geometric Brownian Motion.

Drift of a Geometric Brownian Motion
Let St be a share process such that:
 

1 2

dSt = St σdBt + µdt ⇔ St = S0 exp σBt + µ − σ t ,
2

where Bt is a P-Brownian motion.

Question: Let Dt = e−rt St . Is D(t) driftless under P?

Answer: By Ito’s lemma: dDt = Dt σdBt + (µ − r )dt · · · Exercise!
Unless µ = r , Dt is not driftless.
Question: Can we make D(t) driftless?
Answer: Re-write dDt :
   
µ−r
dDt = σDt dBt + dt .
σ

P Tapadar ([email protected]) MA537 & MA837 Spring Term 12 / 50

 Continuous Processes 5-step method in continuous time

Step 1: In search of Q for Geometric Brownian Motion. Notes

Drift of a Geometric Brownian Motion . . . contd.

By CMG there is a new measure Q such that
t
µ−r µ−r
Z
B̃t = Bt + ds = Bt + t
0 σ σ

is a Q-Brownian motion.
Then under Q: dDt = σDt d B̃t . Hence Dt is driftless.
By the Fact in Martingale and driftlessness, Dt is a martingale.

Direct checks:
Question: Is D(t) a Martingale
 under measure P?
Answer: Find EP Dt F0 .
Question: Is D(t) aMartingale
 under Q?
Answer: Check EQ Dt Fu for any u < t.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 13 / 50

Continuous Processes 5-step method in continuous time

Steps 2 and 3 – Vt and Et

Step 2: Define Vt = e−r (T −t) EQ [X |Ft ].

Step 3: Et = e−rt Vt = e−rT EQ [X |Ft ] is a Q-martingale.

For any u < t, by tower law:
 
EQ Et Fu = EQ e−rT EQ X Ft Fu = e−rT EQ X Fu = Eu
    

So, Et is a Q-martingale.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 14 / 50

 Continuous Processes 5-step method in continuous time

Step 4: dEt = φt dDt Notes

Theorem (Martingale representation theorem)

If Et and Dt are Q-martingales, then there exists a previsible process
φt such that, dEt = φt dDt .

Proof.
We will prove the theorem for stochastic processes (diffusion models)
with SDE of the form:

dEt = σ(Et , t) d B̃t + µ(Et , t) dt

dDt = σ(Dt , t) d B̃t + µ(Dt , t) dt

Under Q, both Et and Dt are driftless, so µ(Et , t) = µ(Dt , t) = 0.

σ(Et ,t)
So, dEt = σ(Et , t)d B̃t = φt σ(Dt , t)d B̃t = φt dDt , where φt = σ(Dt ,t) .

And φt is previsible!
P Tapadar ([email protected]) MA537 & MA837 Spring Term 15 / 50

Continuous Processes 5-step method in continuous time

Step 5: Conclude the proof

dEt
Recall: φt = dDt and define: ψt = Et − φt Dt .

Consider a portfolio (φt , ψt ) of shares and cash at time t.

The value of the portfolio at time t (assuming B0 = 1):

φt St + ψt Bt = ert φt Dt + ψt = ert Et = Vt .
 

The value of the portfolio at time t + dt:

= er (t+dt) φt Dt+dt + ψt
 
φt St+dt + ψt Bt+dt
= er (t+dt) φt Dt + φt dDt + ψt
 

= er (t+dt) Et + dEt
 

= er (t+dt) Et+dt = Vt+dt

P Tapadar ([email protected]) MA537 & MA837 Spring Term 16 / 50

 Continuous Processes 5-step method in continuous time

Step 5: Conclude the proof · · · Notes

So,

dVt = Vt+dt − Vt



= φt St+dt + ψt Bt+dt − φt St + ψt Bt
= φt dSt + ψt dBt .

Moreover as,

= e−r (T −t) EQ X Ft ]

Vt
= e−r (T −T ) EQ X FT ] = EQ X FT ] = X .
 
⇒ VT

So, Vt is the fair price at time t for this derivative contract (otherwise
arbitrage opportunities!).

P Tapadar ([email protected]) MA537 & MA837 Spring Term 17 / 50

Continuous Processes 5-step method in continuous time

Summary

What is this φ?
5-step method proved existence of a self-replicating strategy
(φt , ψt ) for certain (well-behaved!) stochastic processes.
But, we have not said what φt is or how it works!
Replicating strategy:
1 Start with V0 invested in φ0 shares and ψ0 cash at time 0.
2 Continuously re-balance the portfolio to hold exactly φt shares and
ψt cash at time t.
3 Precisely replicate the derivative payoff at time T .

P Tapadar ([email protected]) MA537 & MA837 Spring Term 18 / 50

 Continuous Processes 5-step method in continuous time

Complete market Notes

Definition
The market is complete if for any contingent or derivative claim X
there is a replicating strategy (φt , ψt ).

Example
1 The Binomial Model.
2 The continuous time log-normal model for share prices, i.e.
  
1 2
St = S0 exp σBt + µ − σ t
2

where Bt is a P-Brownian motion.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 19 / 50

Continuous Processes 5-step method in continuous time

Applying 5-step Method to Price Derivatives

Share prices following Geometric Brownian Motion
If the underlying share price process follows a GBM:
  
1
St =P S0 exp σBt + µ − σ 2 t under P
2
   
µ−r 1 2
St =P S0 exp σ Bt + t + (r − σ )t under P
σ 2
 
1 2
St =Q S0 exp σ B̃t + (r − σ )t under Q
2
   
˜

1 2
ST Ft =Q St exp σ BT − B̃t + r − σ (T − t) under Q
2
   
1 2 2
log ST Ft ) ∼Q N log St + r − σ (T − t), σ (T − t)
2
since B˜T − B̃t ∼Q N(0, T − t)

P Tapadar ([email protected]) MA537 & MA837 Spring Term 20 / 50

 Continuous Processes 5-step method in continuous time

Applying 5-step Method to Price Derivatives Notes

If the underlying share price process follows a GBM, find the price, at time t,
of the following derivatives, where the payoff is at time T .
1 XT = K .
2 XT = ST .
3 XT = ST − K .
4 XT = STn .
5 XT = log ST .


6 XT = K × I ST > K , where I(·) is the indicator function.


7 XT = ST × I ST > K , where I(·) is the indicator function.
8 XT = max[ST − K , 0].

P Tapadar ([email protected]) MA537 & MA837 Spring Term 21 / 50

Continuous Processes Black-Scholes model

Assumptions

Assumptions underlying the Black-Scholes model

1 The share process follows a geometric Brownian motion.
2 The underlying market is complete.
3 The risk-free rate is constant and same for borrowing and lending.
4 No risk-free arbitrage opportunities.
5 Negative holdings are allowed.
6 No taxes and transaction costs.
7 Underlying assets can be traded continuously . . .
8 . . . and in infinitesimally small number of units.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 22 / 50

 Continuous Processes Black-Scholes model

How realistic are these assumptions? Notes

Criticisms of the assumptions:

1 Share prices can jump - invalidates assumptions 1 and 2.
2 Risk-free rate of interest can vary.
3 Unlimited negative holding may not be allowed.
4 Shares are traded in integer multiples of units . . .
5 . . . not continuously . . .
6 . . . and attract taxes and transaction costs.
7 Distributions of share returns tend to have fatter tails.
However,
Models are only approximations.
Good approximations provide valuable insight.
Black-Scholes model is a good model since it gives us prices
which are close to what we observe in the market.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 23 / 50

Continuous Processes Black-Scholes model

Block-Scholes formula for a European call option

Theorem
Consider a European call option on a share at time T ,
so that the payoff at time T is X = max{ST − K , 0}.
Then the price of the European call option at time t is given by:

= St Φ(d1 ) − Ke−r (T −t) Φ(d2 )

Vt
 
log SKt + r + 12 σ 2 (T − t)
where d1 = √ ;
σ T −t
 
St 1 2
log K + r − 2 σ (T − t) √
d2 = √ = d1 − σ T − t; and
σ T −t
Z x
Φ(x) = φ(t)dt, . . . φ(t) is the Normal density function.
−∞

P Tapadar ([email protected]) MA537 & MA837 Spring Term 24 / 50

 Continuous Processes Black-Scholes model

Block-Scholes formula for a European put option Notes

Theorem
Consider a European put option on a share at time T ,
so that the payoff at time T is X = max{K − ST , 0}.
Then the price of the European put option at time t is given by:

Vt = Ke−r (T −t) Φ(−d2 ) − St Φ(−d1 )

where d1 and d2 have the same formula as for the European call
option price.

Exercise.
1 Prove the theorem from first principle (as for the call option).
2 Verify using put-call parity.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 25 / 50

Continuous Processes Black-Scholes model for dividend-paying shares

Dividend-paying shares
Fact
Share-process St defined so far assumed no dividend payments.
Q: How would the derivative prices change for dividend-paying shares?

Modified SDE
Assumption: Suppose dividends are paid continuously at a constant
rate of q per annum per share.
Then, the SDE of the total return share process S̃t is:



d S̃t = S̃t σdBt + µdt + q S̃t dt = S̃t σdBt + (µ + q)dt .

By Ito’s lemma, the total return share process is given by:

 
1 2

S̃t = S̃0 exp σBt + µ + q − σ t .
2
P Tapadar ([email protected]) MA537 & MA837 Spring Term 26 / 50
 Continuous Processes Black-Scholes model for dividend-paying shares

5-step method revisited for dividend-paying shares Notes

5-step method in continuous time.

Step 1 Establish a measure Q̃ for which Dt = e−rt S̃t is a
Q̃-martingale.
Step 2 Define Vt = e−r (T −t) EQ̃ [X |Ft ].
Step 3 Et = e−rt Vt = e−rT EQ̃ [X |Ft ] is a Q̃-martingale.
Step 4 dEt = φt dDt : Martingale representation theorem.
Step 5 Define ψt = Et − φt Dt , and conclude the proof.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 27 / 50

Continuous Processes Black-Scholes model for dividend-paying shares

Step 1: In search of Q̃.

Step 1: Find Q for which Dt is a Q̃-martingale
Given the total return share process St :
 
1

Dt = e−rt S̃t = S̃0 exp σBt + µ + q − r − σ 2 t .
2

By Ito’s lemma:


dDt =P Dt σdBt + (µ + q − r )dt · · · Exercise!
   
µ+q−r
=P σDt dBt + dt .
σ
=Q̃ σDt d B̃t · · · under Q̃ by C-M-G.

Then under Q̃: dDt = σDt d B̃t . Hence Dt is driftless.

By Fact 2: Martingale and driftlessness, Dt is a Q̃-martingale.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 28 / 50
 Continuous Processes Black-Scholes model for dividend-paying shares

Steps 2, 3 and 4 – Vt , Et and φt Notes

Step 2: Define Vt = e−r (T −t) EQ̃ [X |Ft ].

Step 3: Et = e−rt Vt = e−rT EQ̃ [X |Ft ] is a Q̃-martingale.

For any u < t, by tower law:
 
EQ̃ Et Fu = EQ̃ e−rT EQ̃ X Ft Fu = e−rT EQ̃ X Fu = Eu
    

So, Et is a Q̃-martingale.

Step 4: dEt = φt dDt

By Martingale representation theorem, if Et and Dt are Q̃-martingales,
then there exists a previsible process φt such that:

dEt = φt dDt .
P Tapadar ([email protected]) MA537 & MA837 Spring Term 29 / 50

Continuous Processes Black-Scholes model for dividend-paying shares

Step 5: Define ψt = Et − φt dDt and conclude the proof

Consider a portfolio (φt , ψt ) of shares and cash at time t.

The value of the portfolio at time t (assuming B0 = 1):

φt S̃t + ψt Bt = ert φt Dt + ψt = ert Et = Vt .

 

The value of the portfolio at time t + dt:

φt S̃t+dt + ψt Bt+dt = er (t+dt) φt Dt+dt + ψt = Vt+dt

 

So, dVt = Vt+dt − Vt = φt d S̃t + ψt dBt .

Moreover, Vt = e−r (T −t) EQ̃ X Ft ] ⇒ VT = X .



So, Vt is the fair price at time t for this derivative contract.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 30 / 50

 Continuous Processes Black-Scholes model for dividend-paying shares

Garman-Kohlhagen formula for a European call option Notes

Theorem
Consider a European call option on a dividend-paying share at time T ,
so that the payoff at time T is X = max{ST − K , 0}.
Then the price of the European call option at time t is given by:

Vt = St e−q(T −t) Φ(d1 ) − Ke−r (T −t) Φ(d2 )

 
log SKt + r − q + 12 σ 2 (T − t)
where d1 = √
σ T −t
 
St 1 2
log K + r − q − 2 σ (T − t) √
and d2 = √ = d1 − σ T − t.
σ T −t
Note: The formula uses the share price St , and NOT the total return
index S̃t . Why?
P Tapadar ([email protected]) MA537 & MA837 Spring Term 31 / 50

Continuous Processes Black-Scholes model for dividend-paying shares

Garman-Kohlhagen formula – The algebra!

Proof.
Using S̃t = St eqt and K̃ = KeqT , we have:

Vt = e−r (T −t) EQ̃ max(ST − K , 0)

 

= e−qT × e−r (T −t) EQ̃ max(S̃T − K̃ , 0)

 

= e−qT × S̃t Φ(d1 ) − K̃ e−r (T −t) Φ(d2 )

 

= St e−q(T −t) Φ(d1 ) − Ke−r (T −t) Φ(d2 )

 
S̃t 1 2
log K̃ + r + 2 σ (T − t)
where d1 = √
σ T −t
 
St 1 2
log K + r − q + 2 σ (T − t)
= √
σ T −t
√
and d2 = d1 − σ T − t.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 32 / 50

 Continuous Processes Black-Scholes model for dividend-paying shares

Garman-Kohlhagen formula for a European put option Notes

Theorem
Consider a European put option on a dividend-paying share at time T ,
so that the payoff at time T is X = max{K − ST , 0}.
Then the price of the European put option at time t is given by:

Vt = Ke−r (T −t) Φ(−d2 ) − St e−q(T −t) Φ(−d1 )

where d1 and d2 have the same formula as for the European call
option price for dividend-paying shares.

Exercise.
Verify using put-call parity.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 33 / 50

Continuous Processes Black-Scholes PDE

Black-Scholes PDE
Setting the scene
Consider a non-dividend-paying share process St for which:
 
Local behaviour: dSt = St σdBt + µdt .
Global behaviour: St = S0 exp σBt + µ − 12 σ 2 t .



Consider a derivative on this share process, whose price at time t is f (t, St ).

The SDE for f – By Ito’s lemma:

∂2f
 
∂f ∂f ∂f 1
df = σSt dBt + + µSt + σ 2 St2 2 dt
∂St ∂t ∂St 2 ∂St
∂f ∂f 1 ∂2f
= dt + dSt + σ 2 St2 2 dt
∂t ∂St 2 ∂St
1 2 2 ∂2f
 
∂f ∂f
−df + dSt = − − σ St dt.
∂St ∂t 2 ∂St2
P Tapadar ([email protected]) MA537 & MA837 Spring Term 34 / 50
 Continuous Processes Black-Scholes PDE

The Risk-free Portfolio Notes

Making a portfolio risk-free.

Consider the following portfolio Π consisting of
∂f


∂St shares; and


− 1 derivative.
The value of the portfolio Π at time t:

∂f
Π(t, St ) = −f + St
∂St

So, the change in the portfolio value, dΠ, over the interval (t, t + dt) is:

1 2 2 ∂2f
 
∂f ∂f
dΠ = −df + dSt = − − σ St dt.
∂St ∂t 2 ∂St2

Important point: The portfolio is risk-free – no random element dSt !

P Tapadar ([email protected]) MA537 & MA837 Spring Term 35 / 50

Continuous Processes Black-Scholes PDE

The Black-Scholes PDE

A risk-free portfolio must instantaneously earn the risk-free rate of re-
turn r . Otherwise there will be an arbitrage opportunity. So,
 
∂f
dΠ = r Πdt = r − f + St dt.
∂St

Now, we have two sets of equations in dΠ, which we can equate:

1 2 2 ∂2f
   
∂f ∂f
− − σ St dt = r − f + St dt.
∂t 2 ∂St2 ∂St

Rearranging, we get the Black-Scholes PDE:

∂f ∂f 1 ∂2f
+ rSt + σ 2 St2 2 = rf .
∂t ∂St 2 ∂St
1 2 2
Θ + rSt ∆ + σ St Γ = rf .
2
P Tapadar ([email protected]) MA537 & MA837 Spring Term 36 / 50
 Continuous Processes Black-Scholes PDE

Application of Black-Scholes PDE Notes

Fact
To verify that a proposed price of derivative f is correct, we need to
check that f satisfies:
the Black-Scholes PDE.
the boundary conditions as t → T .

Exercise
For each of the following check that the Black-Scholes PDE and the
boundary condition are satisfied:
f = Ke−r (T −t) .
f = St .
f = St − Ke−r (T −t) .
f = St Φ(d1 ) − Ke−r (T −t) Φ(d2 ).

P Tapadar ([email protected]) MA537 & MA837 Spring Term 37 / 50

Continuous Processes Black-Scholes PDE

Black-Scholes PDE for dividend-paying shares

∂f



Consider the same portfolio of ∂S t
shares and − 1 derivative.
The change in the portfolio value, dΠ, over the interval (t, t + dt) is:

1 2 2 ∂2f
 
∂f ∂f ∂f
dΠ = −df + (dSt + qSt dt) = − + qSt − σ St dt.
∂St ∂t ∂St 2 ∂St2

But, to avoid arbitrage opportunity:

 
∂f
dΠ = r Πdt = r − f + St dt.
∂St

From these two equations we get:

∂f ∂f 1 ∂2f
+ (r − q)St + σ 2 St2 2 = rf .
∂t ∂St 2 ∂St

Check that the boundary conditions are met for European call option
price given by Garman-Kohlhagen formula for dividend-paying shares.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 38 / 50
 Continuous Processes Black-Scholes PDE

Reconciling Black-Scholes PDE with 5-step Method Notes

Fact (The product rule)

If Xt is a stochastic process while Yt is not, then
d(Xt Yt ) = Xt dYt + Yt dXt .

Example

St = ert Dt ⇒ dSt = d ert Dt

 

= rert Dt dt + ert dDt

= rSt dt + ert dDt

Et = e−rt Vt ⇒ dEt = d e−rt Vt

 

= e−rt dVt − re−rt Vt dt

P Tapadar ([email protected]) MA537 & MA837 Spring Term 39 / 50

Continuous Processes Black-Scholes PDE

Reconciling Black-Scholes PDE with 5-step Method

∂ 2 Vt
 
∂Vt 1 ∂Vt
dVt = + σ 2 St2 dt + dSt . . . from slide 36
∂t 2 ∂St2 ∂St
1 2 2 ∂ 2 Vt
 
∂Vt ∂Vt 
rSt dt + ert dDt

= + σ St 2
dt +
∂t 2 ∂St ∂S t

1 2 2 ∂ 2 Vt
 
∂Vt ∂Vt ∂Vt rt
= + σ St 2
+ rS t dt + e dDt
∂t 2 ∂St ∂St ∂St
1 2 2 ∂ 2 Vt
 
−rt ∂Vt ∂Vt ∂Vt
dEt = e + σ St + rSt dt + dDt − re−rt Vt dt
∂t 2 ∂St2 ∂St ∂St
1 2 2 ∂ 2 Vt
 
−rt ∂Vt ∂Vt ∂Vt
= e + σ St 2
+ rSt − rVt dt + dDt
∂t 2 ∂St ∂St ∂St
φt = ∂V
(
t
∂St
But, dEt = 0 dt + φt dDt ⇒ 1 2 2 ∂ 2 Vt
rVt = ∂V ∂t + 2 σ St ∂S 2 + rSt ∂St
t ∂Vt
t

which is the Black-Scholes PDE.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 40 / 50
 Continuous Processes Black-Scholes PDE

Martingale versus PDE approach Notes

Advantages of the Martingale approach

Much more clarity of the process.
An expectation which can be evaluated explicitly.
Provides a replicating strategy.
Can be applied to any FT -measurable derivative payment.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 41 / 50

Continuous Processes Calibrating Binomial Models

Calibrating Binomial Model

Binomial Tree and Geometric Brownian Motion
It is often convenient when calibrating the binomial model to have the mean
and variance implied by the binomial model correspond to the mean and
variance of a log-normal distribution.

Geometric Brownian Motion

Consider the time interval (t, t + δt) over which a share price process follows
a geometric Brownian motion. So, under the risk-neutral measure Q:
   
1 2
S(t + δt) = S(t) exp σ(B̃t+δt − B̃t ) + r − σ δt ;
2
r δt
 
E S(t + δt)/S(t) = e ;

= σ 2 δt;


Var log S(t + δt)/S(t)

2
≈ σ 2 δt . . . ignoring (δt)2 terms.

E log S(t + δt)/S(t)

P Tapadar ([email protected]) MA537 & MA837 Spring Term 42 / 50

 Continuous Processes Calibrating Binomial Models

Calibrating Binomial Model Notes

Binomial Branch
Consider the time interval (t, t + δt) over which a share price process follows
a binomial branch model. So, under the risk-neutral measure Q:

u with probability q
S(t + δt)/S(t) =
d with probability (1 − q)

So,  
E S(t + δt)/S(t) = qu + (1 − q)d.
Calibrating this with the mean obtained for GBM, we get:

er δt − d
q= ,
u−d
which is indeed the risk-neutral probability for the binomial model.

P Tapadar ([email protected]) MA537 & MA837 Spring Term 43 / 50

Continuous Processes Calibrating Binomial Models

Calibrating Binomial Model

Calibrating Volatility
Considering the second order moment for the binomial branch, we get:

2
= q(log u)2 + (1 − q)(log d)2 .

E log S(t + δt)/S(t)

If we make an additional assumption ud = 1, the expression simplifies to:

2
= (log u)2 .

E log S(t + δt)/S(t)

Calibrating this with the second order moment obtained for GBM, we get:

(log u)2 = σ 2 δt.

√ √
Hence, u = eσ δt
and d = e−σ δt
.
For a dividend-paying share, with ν as the continuously payable dividend rate,
the adjustments to the above formulae give:
√ √
u = eσ δt+νδt
and d = e−σ δt+νδt
.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 44 / 50
 Continuous Processes State-price deflators

State-price deflators Notes

For a non-dividend-paying share:

 
Under P, dSt = St σdBt + µdt .
 
Under Q, dSt = St σd B̃t + rdt ,
µ−r
. . . where d B̃t = dBt + σ dt.

Corollary (Corollary to the Cameron-Martin-Girsanov Theorem)

There exists a process ηt , such that for any (FT -measurable) derivative
payoff X at time T , we have:
 
  ηT
EQ X Ft = EP X Ft ,
ηt
1 2
where ηt = e−γZt − 2 γ t .

P Tapadar ([email protected]) MA537 & MA837 Spring Term 45 / 50

Continuous Processes State-price deflators

State-price deflators
Definition
State-price deflators A state-price deflator (also deflator, state-price
density, pricing kernel, stochastic discount factor) is defined as:

At = e−rt ηt

Price of a derivative using state-price deflator

EP e−rT ηT X Ft
   
−r (T −t)
  −r (T −t) ηT
Vt = e EQ X Ft = e EP X Ft =
ηt e−rt ηt
 
EP AT X Ft
Vt =
At
Important: The risk-neutral and the state-price deflator approaches
give the same price Vt .
P Tapadar ([email protected]) MA537 & MA837 Spring Term 46 / 50
 Continuous Processes The Greeks

Portfolio risk management Notes

Fact
Combinations of various derivatives and the underlying asset in a single
portfolio allow us to modify our exposure to risk.

Example
Call option allows exposure to upside movement.
Put option allows downside risk to be removed.

Definition (The Greeks)

The Greeks are a group of mathematical derivatives that can be used to
manage or understand the risks in our portfolio.

Notation
f or f (t, St ) be the value of a derivative at time t when the price of the
underlying asset is St . For example: ct , pt , Ct , Pt , etc.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 47 / 50

Continuous Processes The Greeks

Delta

Definition (Delta)
∂f
The delta for an individual derivative is: ∆ = ∂St .

Example
f = St ⇒ ∆ = 1. The derivative and asset price move together.
∆ = 0. The derivative is completely insensitive to underlying asset
price movement. For such a derivative, if we know its price at time
t, we can predict its value at t + dt with complete certainty, also
called instantaneously risk-free.
Call and put options: Given the shape of the ct and pt , how will the
graphs of ∆ look like for European call and put options?
(Exercise!)

P Tapadar ([email protected]) MA537 & MA837 Spring Term 48 / 50

 Continuous Processes The Greeks

Gamma Notes

Definition (Gamma)
∂2f
The gamma for an individual derivative is: Γ = ∂St2
. Γ is the rate of
change of ∆ with the price of underlying asset.

Example
f = St ⇒ Γ = 0.
Γ large means:
1 ∆ changes fast with St .
2 Frequent adjustments to portfolio for delta hedging.
3 Increased transaction costs.
4 Keep Γ small.
Call and put options. Given the shape of the ct and pt , how will the
graphs of Γ look like for European call and put options? (Exercise!)

P Tapadar ([email protected]) MA537 & MA837 Spring Term 49 / 50

Continuous Processes The Greeks

Theta, Vega, Rho and Lambda

Definition (Theta)
The theta for an individual derivative is: Θ = ∂f
∂t , where t is the time since the
start of the contract (and NOT time to maturity).

Definition (Vega)
∂f
The vega for an individual derivative is: V = ∂σ . This is the sensitivity of the
derivative price to changes in the assumed level of volatility of St .

Definition (Rho)
∂f
The rho for an individual derivative is: ρ = ∂r . Rho tells us about the
sensitivity of the derivative price to changes in the risk-free rate of interest r .

Definition (Lambda)
∂f
The lambda for an individual derivative is: λ = ∂q , where q is the continuous
dividend yield on the underlying security.
P Tapadar ([email protected]) MA537 & MA837 Spring Term 50 / 50