The Black Scholes model

Antonino Zanette
University of Udine
[email protected]

1
Continuous Stochastic Processes

The origin of stochastic processes can be traced back to the field of statistical
physics. A physical process is a physical phenomenon whose evolution is
studied as a function of time.

In a financial framework, the idea is to give a model of stock price

fluctuations in continuous time.

Definition
Let (Ω, F , P) be a probability space. A continuous-time stochastic process is a family (Xt )t≥0 of
R-valued random variables on (Ω, F , P).
• the index t stands for the time.
• for each time t fixed:
Xt : Ω −→ R

• for each ω ∈ Ω the map t −→ Xt (ω) is called the path of the process.

2
Brownian motion
In finance, the most common models are constructed on the Brownian
motion.
Definition
A Brownian motion is a real-valued, continuous stochastic process (Xt )t≥0 with indipendent,
normally distribuited and stationary increments. In other words :
P1 B0 = 0.
P1 the function s 7→ Bs (ω) is a continuous function.
P2 indipendent increments : for each k, 0 ≤ t0 < t1 < . . . < tk , the increments
Bt0 , Bt1 − Bt0 , Bt2 − Bt1 , .., Btk − Btk−1 are indipendent.

P3 for each t > s ≥ 0, Bt −hBs ∼ N (0, ti− s) ⇒

EP (Bt − Bs ) = 0 and EP (Bt − Bs )2 = t − s.
In particular for s = 0 it follows that EP (Bt ) = 0 e V ar(Bt ) = t.
A particle that is undergoing a Brownian motion Bt has the following property.
P2 Absence of memory.
P3 The mean of Bt − Bs is 0, so there is not a privilegiate direction.
The variance of the particle movement is proportional to the observed time.

3
Brownian motion
The path of the Brownian motion are continuous, but not differentiable.

0.5

0 0.5 1

4
Financial example 2
h i
−rT σB
P =e EP (K − e T )+ .

with BT B.M a time T .

We consider a put option with underlying asset
σBt
St = e

√
Under P, BT ∼ N (0, T ). So BT = g T con g ∼ N (0, 1).
We can approximate the put price with

−rT f (X1 ) + · · · + f (Xn )

P ≈e
n
X1 , .., Xn ∼ N (0, T ).

5
Monte Carlo algorithm

main()
{
double mean_price,mean2_price,brownian,price,price_sample,error_price,inf_price,sup_price;
mean_price= 0.0;
mean2_price= 0.0;
for(i=1;i<=N;i++)
{
/*Brownian motion simulation*/
brownian=gaussian()*sqrt(T);

price_sample=MAX(0.0,K-exp(sigma*brownian));

mean_price= mean_price+price_sample;
mean2_price= mean2_price+SQR(price_sample);
}
/* Price */
price=exp(-r*T)*(mean_price/N);
error_price= sqrt(exp(-2.*r*T)*mean2_price/N - SQR(price))/sqrt(N-1);
inf_price= price - 1.96*(error_price);
sup_price= price + 1.96*(error_price);
}

6
Simulation of Brownian motion path
T
Let [0, T ] be divided using N time intervals of lenght ∆T = N .
N
X N
X
BT = BN ∆T = (Bk∆T − B(k−1)∆T ) = ∆Bk
k=1 k=1

The increments ∆Bk ∼ N (0, ∆T ) are indipendent and normally distribuited :

√
∆Bk = g ∆T

with g ∼ N (0, 1).

T
Start t0 = 0, B0 = 0, ∆T = N
for k = 1, . . . , N
BEGIN;
tk = tk−1 + ∆T ;
simulation of g ∼ N (0, 1);
√
Bk∆T = B(k−1)∆T + g ∆T ;
END;

7
Algorithm

main()
{
double k,T,brownian,B_T,time;
int N;

k=T/N;
brownian=0.;
time=0.;
for(i=1;i<=N;i++)
{
/*Time*/
time+=k;

/*Brownian path simulation*/

brownian=brownian+gaussian()*sqrt(k);
}

/* B_T */
B_T=brownian;
}

8
Brownian motion and random walk
One of the standard way used to approximate a Brownian motion is to use a random walk. Here
we use the standard symmetric random walk.
Proposition
Let (Xi , i ≥ 1) be a sequence of independent random variables such that P(Xi = ±1) = 1/2.
Set Sn = X1 + · · · + Xn .
Let ∆T = T /N be the time step. Set
√
BN = ∆T SN .

Then, the sequence BN converges in distribution to BT .

Consequently, if f is a bounded continuous function then
h i h i
EP f (BN ) converges to EP f (BT ) .

In the financial example 2, f (BT ) = (K − eσBT )+ . We use f (BN ) = (K − eσBN )+ , or

√
σ ∆T SN
g(SN ) = (K − e )+

9
Besides, E(g(SN )) can be computed as follows :

 u(N ∆T, x) = g(x),
1 1
 u(n∆T, x) = u((n + 1)∆T, x + 1) + u((n + 1)∆T, x − 1).
2 2

10
Exercise
Compute
h i
−rT
P =e EP (K − eσBT )+ .

with K = 100, r = 0.03, σ = 0.2 using

• Monte Carlo algorithm
• Tree method

11
Brownian motion with drift
Xt = µt + σBt
with Bt standard brownian motion, with µ and σ costants.

50.3

50.25

50.2

50.15

50.1

50.05

50

49.95

0 0.5 1

A french matematician Bachelier introduces it in first years of 1900 for modelling stock prices.
But there is the problem of the negative prices:

Xt ∼ N (µt, σ 2 t).

12
Geometric Brownian motion
Definition
A Geometric Brownian motion St is a continuous stochastic process such that:
P1 S0 = x.
1 2 )t+σB
P2 St = S0 e(µ− 2 σ t

Bt standard Brownian motion, µ and σ costant.

65

60

55

50

0 0.5 1

St
The log-returns log S0 have normal distribution (the returns are normal).

St 1 2
S0 is log-normal of parameters (µ − 2 σ )t and σ 2 t.

13
14
Property of the GBM

P1 Consider s<t. Then

St 1
log( ) ∼ N ((µ − σ 2 )(t − s), σ 2 (t − s))
Ss 2
Expectation

St
E( ) = eµ(t−s)
Ss
Variance

St 2µ(t−s) σ 2 (t−s)
V ar( )=e (e − 1)
Ss
P2 for each 0 ≤ t0 < t1 < . . . < tn , the relative increments Stk /Stk−1 are indipendent and
have common law.

15
Financial example 3
h i
−rT
P =e EP (K − ST )+ ,

with ST value of the GBM at time T T .

We consider a put option with underlying asset
1 2 )t+σB
St = S0 e(µ− 2 σ t

The payoff can be written

1 2 )T +σB
h(BT ) = (K − S0 e(µ− 2 σ T )+

Then
−rT h(X1 ) + · · · + h(Xn )
P ≈e
n
X1 , .., Xn ∼ N (0, T ).

16
Monte Carlo algorithm

main()
{
double mean_price,mean2_price,brownian,price,price_sample,error_price,inf_price,sup_price;
mean_price= 0.0;
mean2_price= 0.0;
for(i=1;i<=N;i++)
{

brownian=gaussian()*sqrt(T);

price_sample=MAX(0.0,K-x*exp((mu-0.5*sigma*sigma)*T+sigma*brownian));

mean_price=mean_price+price_sample;
mean2_price=mean2_price+SQR(price_sample);
}

price=exp(-r*T)*(mean_price/N);
error_price= sqrt(exp(-2.*r*T)*mean2_price/N - SQR(price))/sqrt(N-1);
inf_price= price - 1.96*(error_price);
sup_price= price + 1.96*(error_price);
}

17
Simulation of Geometric Brownian motion path
T
Let [0, T ] be divided using N time intervals of lenght ∆T = N .
N
Y N
Y
Sk∆T 1 2 )∆T +σ∆B
ST = SN ∆T = S0 = S0 e(µ− 2 σ k,

k=1
S(k−1)∆T k=1

with
1 2 1 2 √
(µ − σ )∆T + σ∆Bk = (µ − σ )∆T + σg ∆T con g ∼ N (0, 1)
2 2
Simulation of the GBM path (St )0≤t≤T :

T
Start t0 = 0, S0 = x, ∆T = N
for k = 1, . . . , N
BEGIN;
tk = tk−1 + ∆T ;
simulation of g ∼ N (0, 1); √
1 σ 2 )∆T +σg
(µ− 2 ∆T
Sk∆T = S(k−1)∆T e ;
END.

18
Algorithm

main()
{
double k,T,w_derive,s,S_N,mu=0.1,sigma=0.2,time;
int N;

k=T/N;
s=50.;
time=0.;
for(i=1;i<=N;i++)
{
/*Timew*/
time=time+k;
/*Geometric Brownian simulation*/
s=s*exp((mu-0.5*sigma*sigma)*k+sigma*gaussian()*sqrt(k));
}
/* S_T */
S_T=s;
}

19
Differential property of the Brownian motion
h i
2
E (Bt − Bs ) = t − s

Let us consider the random variable.

 2
X = Bt+∆t − Bt

Then h i
E X = (t + ∆t) − t = ∆t

and h i
2
V X = 2(∆t)

When ∆t is close to zero the r.v. X is “not to much random” and is very close to his mean ∆t:
 2
X= Bt+∆t − Bt ≈ ∆t

We write  2
dBt = dBt dBt = dt

and
√
dBt = dt

20
The quadratic variation of the Brownian motion in [0, T ] is equal to his variance.

Let (Bt , t ≥ 0) be a standard Brownian motion. For each T > 0 and partition
n n
0 = tn n n
0 < t1 < · · · < tn = T so that π = sup (ti − ti−1 ) goes to zero when n → ∞ :
i≤n

n 
X 2
n n
B(ti ) − B(ti−1 ) → T,
i=1

in the sense if the quadratic mean, for n → ∞

Proof
n 
hX 2 i
n n
E B(ti ) − B(ti−1 ) =T
i=1
 2
The random variables B(tn
i ) − B(tn
i−1 ) , i = 1, 2, . . . , n are indipendent.
" n # n  
X 2 X 2
Var B(tn
i ) − B(tn
i−1 ) = Var B((tn
i ) − n
B(ti−1 )
i=1 i=1
n
X
2
= 2 tn n
i − ti−1 ≤ 2πT.
i=1

This variance goes to 0 when n → ∞.

21
Stochastic integral
Consider the stochastic integral
Z T
f (t, Bt )dBt .
0

RT
We can define Xt = 0
f (t, Bt )dBt as the limit of discrete sums of the type

n−1
X n
Xn = f (tj , Btn )(Btn − Btn ),
j j+1 j
j=0

as n goes to infinity.

When can think Xn as a "Riemann sum" in which the representative point inside each
subinterval is the left-most point.

This definition of the stochastic integral is called the Ito integral.

Of course, conditions on f are necessary to ensure that Xn converge in a reasonable sense and

that the limit does not depende on the sequence on meshes tn

i .

22
Example
Z T
dBs = BT
0

Example
Z T 1 1 2
Bs dBs = − T + BT
0 2 2

23
Z T n−1
X  
Bs dBs = lim Btj · Btj+1 − Btj
0 n→∞
i=0
n−1
X 
= lim Btj Btj+1 − Bt2
n→∞ j
i=0

X
n−1
1 2 1 2 1 2

= lim − Btj+1 − Btj − Bt + Bt
n→∞
i=0
2 2 j 2 j+1
" n−1 n−1
#
1 X 2 X 2 2

= lim − Btj+1 − Btj + Bt − Bt
n→∞ 2 j+1 j
i=0 i=0
1 1 2
= − T + BT .
2 2

24
Ito integral property
Z T
f (t, Bt )dBt .
0

- Linearity
- Expectation
hZ T i
E f (t, Bt )dBt = 0.
0

- Quadratic mean
h Z T 2 i hZ T i
2
E f (t, Bt )dBt =E f (t, Bt )dt .
0 0

25
Stochastic differential equations
Definition A process (Xt )t≥0 which satisfies
Z t Z t
(1) Xt = x + µ(s, Xs )ds + σ(s, Xs )dBs ,
0 0

is called a solution of the stochastic differential equation with coefficient µ and σ, intial
condition x and Brownian motion (Bt )t≥0 .

(Xt )t≥0 is called the diffusion process corresponding to the coefficients µ and σ. We can write
the differential simbolic notation
(
dXt = µ(t, Xt )dt + σ(t, Xt )dBt
X0 = x.

Example

The standard Brownian motion, the Brownian motion with drift and the geometric Brownian

motion are solution of particular s.d.e.

26
Example : Brownian motion with drift
The Brownian motion with drift is solution of the following s.d.e.
(
dXt = µdt + σdBt
X0 = x.

It is the diffusion process corresponding to the coefficients µ(t, Xt ) = µ and σ(t, Xt ) = 1.

Stochastic Integral
Z t Z t
Xt = x + µds + σdBs
0 0

The solution is
Xt = x + µt + σBt

27
Example : Geometric Brownian motion
The Geometric Brownian motion with drift is solution of the following s.d.e.
(
dSt = µSt dt + σSt dBt
S0 = x.

Stochastic Integral
Z t Z t
(2) St = S0 + µSu du + σSu dBu
0 0

The solution is
(µ− 1 σ 2 )t+σBt
St = xe 2

The results is obtained using the Ito’s Lemma.

28
Ito’s Lemma
Lemma
Let (Xt )t≥0 the solution of

dXt = µ(t, Xt )dt + σ(t, Xt )dBt

X 0 = x0

and let f (t, Xt ) be a real-valued function of class C 1,2 .

Then

∂f (t,Xt ) ∂f (t,Xt ) 1 2 ∂ 2 f (t,Xt ) ∂f (t,Xt )

df (t, Xt ) = ( + µ(t, Xt ) + σ (t, X t ) )dt + σ(t, Xt ) dBt
∂t ∂Xt 2 ∂Xt2 ∂Xt

We can write
∂f
df (t, Xt ) = α(t, Xt )dt + dXt
∂Xt
with
∂f σ 2 (t, Xt ) ∂ 2 f
α(t, Xt ) = +
∂t 2 ∂Xt2

29
Example

dSt = µSt dt + σSt dBt

S0 = x
Using Ito’s lemma
1 2 )t+σB
St = S0 e(µ− 2 σ t

Let us consider Xt = Bt and

1 2 )t+σB
St = f (t, Bt ) = S0 e(µ− 2 σ t

Ito’s lemma implies that

 1 2 1 2 
dSt = df (t, Bt ) = (µ − σ )St + σ St dt + σSt dBt = µSt dt + σSt dBt
2 2

30
On the contrary, let is consider Xt = St and

f (t, St ) = ln(St )

Ito’s lemma implies that

1 2
dln(St ) = df (t, St ) = (µ − σ )dt + σdBt
2
or in the integral form
Z T Z T Z T
1 2
1 dln(St ) = (µ − σ ) dt + σ dBt
0 0 2 0
1
[ln(St )]T
0 = (µ − σ 2 ) [t]T T
0 + σ [Bt ]0
2
ST 1 2
ln( ) = (µ − σ )T + σ(BT − B0 )
S0 2
 
ST 1 2
= exp (µ − σ )T + σBT
S0 2

Then
 
1 2
(3) ST = S0 exp (µ − σ )T + σBT .
2

31
Example
Consider f (t, x) = x2 and Xt = Bt . Then
2
f (t, Bt ) = Bt

Ito’s lemma implies that

1 
dBt2 = df (t, Bt ) = 2 dt + 2Bt dBt = dt + 2Bt dBt
2
In the integral form
Z t Z t
2 2 1
Bt = B0 + 2du + 2Bu dBu
0 2 0

so that Z t 1 2
Bu dBu = (Bt − t).
0 2

32
Theorem (Existence and Uniqueness)
Z t Z t
(4) Xt = x + µ(s, Xs )ds + σ(s, Xs )dBs ,
0 0

If µ and σ are continuous functions, and if there exists a constant K < +∞, such that :
1. |µ(t, x) − µ(t, y)| + |σ(t, x) − σ(t, y)| ≤ K|x − y|

2. |µ(t, x)| + |σ(t, x)| ≤ K(1 + |x|)

then, for any T ≥ 0, (4) admist a unique solution in the interval [0, T ].
Moreover, this solution (Xs )0≤s≤T satisfies :
!
2
E sup |Xs | < +∞
0≤s≤T

33
Simulation diffusions paths
Euler Discretization Scheme

∆Xt = Xt+∆T − Xt = µ(t, Xt )∆t + σ(t, Xt )∆Bt

X 0 = x0

T
Start t0 = 0, x0 , ∆T = N
for k = 1, . . . , N
BEGIN;
tk = tk−1 + ∆T ;
simulation of g ∼ N (0, 1);
√
xk∆T = x(k−1)∆T + µ(x(k−1)∆T , tk−1 )∆T + σ(x(k−1)∆T , tk−1 )g ∆T
END;

34
Brownian motion

Definition
A Brownian motion is a real-valued, continuous stochastic process (Xt )t≥0 with indipendent,
normally distribuited and stationary increments. In other words :
• B0 = 0.
• continuity.
• indipendent increments : if s ≤ t, Bt − Bs is indipendent of Fs = σ(Bu , u ≤ s).

• stationary increments : if s ≤ t, Bt − Bs and Bt−s have the same law.

35
Continuous-time martingale

Let us consider a probability space (Ω, A, P) and a filtration F := (Ft , t ≥ 0) on this space.

Definition An adapted family (Mt , t ≥ 0) of integrable random variables is a (Ft )-martingale if

for each s ≤ t,
E (Mt |Fs ) = Ms .

It follows from the definition that if (Mt )t≥0 is a martingale, then E(Mt ) = E(M0 ), for each t.

Example

Bt is an Ft -martingale.

36
Markov property

The intuitive meaning of the Markov property is that the future behaviour of the process
(Xt )t≥0 after t depends only on the value Xt and is not influenced by the history of the process
before t.

Mathematically speaking, (Xt )t≥0 satisfies the Markov property if, for any function f bounded
and measurable and for any s and t, such that s ≤ t, we have :

E (f (Xt ) |Fs ) = E (f (Xt ) |Xs ) .

This property is satisfied for a solution of the equation (4).

This is a crucial property of the Markovian model and it will have great conseguences in the
pricing of options.

37
Black-Scholes model
- Risk-free asset (
dSt0 = rSt0 dt
S00 = 1.

- Risk asset (
dSt = µSt dt + σSt dBt
S0 = x.

with (Bt )t≥0 standard brownian motion under the historical probability P.
- The short-term interest rate is known and is costant through time.
- The stock pays no dividends or other distributions.
- There are no penalties to short-selling.
- It is possible to borrow any fraction of the price of a security to buy it or to hold it, at the
short-time interest rate.
- Absence of arbitrage opportunities.

38
Financial interpretation of the parameters
- r istantaneous interest rate : [0%, 12%]

- µ expected return of the risky asset.

St µt
E( )=e
S0

- σ is the volatility σ.
This is vey important parameters : [30%, 70%] in the equity market.
- risk premium λ
µ−r
λ=
σ
Then
µ = r + λσ

The expected return µ of the risky asset is the sum of the return of the no-risky asset plus
something proportional to σ.

39
We can write
dSt = rSt dt + σSt (dBt + λdt)
The Girsanov theorem gives
bt = dBt + λdt
dB

bt standard Brownian motion under the risk neutral probability Q.

with B

40
Dinamics under the risk neutral probability measure
(
bt
dSt = rSt dt + σSt dB
(5)
S0 = x.

bt standard Brownian motion under Q.

with B
The solution(5) is
(r− 1 σ 2 )t+σ B
b
t
St = xe 2

Then  
ST r(T −t)
EQ |Ft =e
St

41
Radon-Nikodyn Theorem
Let P and Q be two probabilty measure on (Ω, F )
If Q is absolutely continuous with respect to P, (A ∈ F , P(A) = 0 → Q(A) = 0), then there
existe a unique r.v. X ≥ 0, F -misurable such that
Z
Q(A) = XdP
A

The random variable X is commonly written as

dQ
=X
dP

X is called the Radon-Nikodyn derivative.

42
Change of probability measure in the Gaussian case
Let use consider Z ∼ N (µ, 1) under P.
Then there exists Q so that Z(0, 1) under Q where
1 2
dQ = e−µZ+ 2 µ dP.

In fact
Z Z z −(x−µ)2
1
P(Z ≤ z) = dP(ω) = √ e 2 dx,
{ω:Z(ω)≤z} −∞ 2π
and
Z Z Z
−µZ(ω)+ 1 µ2
z 1 −x2
Q(Z ≤ z) = dQ(ω) = e 2 dP(ω) = √ e 2 dx.
{ω:Z(ω)≤z} {ω:Z(ω)≤z} −∞ 2π

Moreover it holds
h i h i h i h i
µZ− 1 µ2 −µZ+ 1 µ2
EP f (Z) = EQ f (Z)e 2 and EQ f (Z) = EP f (Z)e 2 .

43
Girsanov’s Theorem
Let Bt be a Brownian motion under (Ω, F , P) adapted to the filtration Ft .
Let (Zt )0≤t≤T be the process defined by :
 1 2 
Zt = exp − λBt − λ t .
2

Then, under the probability measure Q with density ZT with respect to P

dQ = ZT dP

bt )0≤t≤T given by B
the process (B bt = Bt + λt, is a standard Brownian motion under Q.

44
Risk neutral pricing formula
 
ST r(T −t)
EQ |Ft =e
St
This holds for each asset:
hV i
T r(T −t)
EQ |Ft = e
Vt
Equivalently
 
−r(T −t)
Vt = EQ e VT |Ft .

The price of a contingent claim is the expected value of the discounted payoff.
 
−rt −rT
e Vt = EQ e VT |Ft .

Discounted prices are martingales.

45
Black-Scholes formula for European Call options
The price at time t of an European Call option in the Black-Scholes model
   
−r(T −t) −r(T −t) (r− 1 σ 2 )(T −t)+σ(BT −Bt )
Ct = EQ e CT |Ft = EQ e (St e 2 − K)+

is given by
−r(T −t)
Ct = St N (d1 ) − Ke N (d2 )
with    
St σ2
log K + r + 2 (T − t) p
d1 = √ d2 = d1 − σ T −t
σ T −t

and N (x) the distribution function of the standard Gaussian variable

Z d
1 2 /2
N (d) = √ e−x dx.
2π −∞

46
Black-Scholes formula for European Put options
The price at time t of an European put option in the Black-Scholes model
   
−r(T −t) −r(T −t) (r− 1 σ 2 )(T −t)+σ(BT −Bt )
Pt = EQ e PT |Ft = EQ e (K − St e 2 )+

is given by
−r(T −T )
Pt = Ke N (−d2 ) − St N (−d1 )

47
Implementation of the formula
The price of the call option depends on six parameters.

C = C(St = x, t, T, K, σ, r)

- The strike K and the maturity T are specified in the contract.

- r is constant. But in general this is not true (Vasicek or CIR model).

- The volatility cannot be observed directly. In practice, two methods are used to evaluate σ
– The historical method: in the BS model, σ 2 T is the variance of log(ST ) and the
variables log(S∆T /S0 ), log(S2 /S∆T ), . . . , log(SN ∆T /S(N −1)∆T ) are i.i.d random
variables.
Therefore, σ can be estimated by statistical means using past observations of the asset
price.
– the ”implied volatility” method: some options are quoted on organized markets; the
price of options being an increasing function of σ, we can associate an ”implied”
volatility to each quoted option, by inversion of the Black-Scholes formula.
Obs
C (S0 , 0, T, K) = C(S0 , 0, T, K, Σ(K, T ), r)

Σ is called implied volatility. Due to the market imperfections Σ has a typical

dependence on K called SMILE EFFECT.

48
 n = m = 6, 30 iter.

0.3

0.25
0.3
0.2

0.2
0.15

0.1
0.1

0 0.05

−0.1 0

−0.2 −0.05
2
−0.1
1.5
800
1 750
700 −0.15
650
0.5 600
550 −0.2
t (an) 0 500
S

49
Approximating the distribution function of g ∼ N (0, 1)

1
Set t = 1+px , then:

1 − √1 exp(− x2 )(b1 t + b2 t2 + b3 t3 + b4 t4 + b5 t5 ) if x≥0
2π 2
N (x) = 2
 √1 exp(− x )(b1 t + b2 t2 + b3 t3 + b4 t4 + b5 t5 ) if x<0
2π 2

with the following constants :

p = 0.2316419;
b1 = 0.319381530;
b2 = −0.356563782;
b3 = 1.781477937;
b4 = −1.821255978;
b5 = 1.330274429;

50
Approximating the distribution function of g ∼ N (0, 1)

double N(double x)
{ const double p= 0.2316419;
const double b1= 0.319381530;
const double b2= -0.356563782;
const double b3= 1.781477937;
const double b4= -1.821255978;
const double b5= 1.330274429;
const double one_over_twopi= 0.39894228;
double t;
if(x >= 0.0)
{
t = 1.0 / ( 1.0 + p * x );
return (1.0 - one_over_twopi * exp( -x * x / 2.0 )
* t * ( t *( t * ( t * ( t * b5 + b4 ) + b3 ) + b2 ) + b1 ));
}
else
{/* x < 0 */
t = 1.0 / ( 1.0 - p * x );
return ( one_over_twopi * exp( -x * x / 2.0 ) *
t * ( t *( t * ( t * ( t * b5 + b4 ) + b3 ) + b2 ) + b1 ));
}
}

51
Scilab

function [y]=Norm(x)
[y,Q]=cdfnor("PQ",x,0,1);
endfunction

52
Price of a Call option

main()
{
double sigmasqrt,d1,d2,delta,price;

sigmasqrt=sigma*sqrt(t);
d1=(log(s/k)+r*t)/sigmasqrt+sigmasqrt/2.;
d2=d1-sigmasqrt;
delta=N(d1);
/*Price*/
price=s*delta-exp(-r*t)*k*N(d2);
}

53
Price of a put option

main()
{
double sigmasqrt,d1,d2,delta,price;

sigmasqrt=sigma*sqrt(t);
d1=(log(s/k)+r*t)/sigmasqrt+sigmasqrt/2.;
d2=d1-sigmasqrt;
delta=N(-d1);

/*Price*/
price=exp(-r*t)*k*N(-d2)-delta*s;
}

54
Put-Call Theorem Parity
We have the following put-call parity between the prices of the underlying asset St and
European call and put options on stocks that pay no dividends:
−r(T −t)
Ct = Pt + St − Ke .

55
Payoff Call

100

50

0 50 100 150 200

Payof Put

90

80

70

60

50

40

30

20

10

0 50 100 150 200

56
Proof of the Black-Scholes formula
h 1 2 i
−r(T −t)
C(t, x) = EQ e (xe(r− 2 σ )(T −t)+σ(BT −Bt ) − K)+

Then
h √  i
− 1 σ 2 (T −t)+σ T −tg −r(T −t)
C(t, x) = EQ xe 2 − Ke
+

with g ∼ N (0, 1).

Set  
x

σ2
log K + r + 2 (T − t) p
d1 = √ d2 = d1 − σ T −t
σ T −t

 √  
σ (T −t)g− 1 σ 2 (T −t) −r(T −t)
C(t, x) = E xe 2 − Ke 1g≥−d2

Z  √  2
+∞
σ (T −t)y− 1 σ 2 (T −t) −r(T −t) e−y /2
= xe 2 − Ke √ dy
−d2 2π
Z  √  2
d2
−σ (T −t)y− 1 σ 2 (T −t) −r(T −t) e−y /2
= xe 2 − Ke √ dy.
−∞ 2π

57
 Z  √  2
d2
−σ (T −t)y−σ 2 (T −t)/2 −r(T −t) e−y /2
C(t, x) = xe − Ke √ dy.
−∞ 2π
p
The change of variable z = y + σ (T − t), gives :

C(t, x) = xN (d1 ) − Ke−r(T −t) N (d2 ),

with : Z
1 d 2 /2
N (d) = √ e−x dx.
2π −∞

58
Monte Carlo method in the Black-Scholes model
We want compute
h i
−rT
P =e EQ (K − ST )+ .

in the Black-Scholes model

(r− 1 σ 2 )T +σBT
ST = S0 e 2

The payoff function can be written in the following way

1 2 )T +σB
h(BT ) = (K − S0 e(r− 2 σ T )+

We can approximate the price with

−rT h(X1 ) + · · · + h(Xn )

P ≈e
n
X1 , .., Xn ∼ N (0, T ).

59
Monte Carlo algorithm
European Put in the Black-Scholes model

main()
{
double mean_price,mean2_price,brownian,price,price_sample,error_price,inf_price,sup_price;
mean_price= 0.0;
mean2_price= 0.0;
for(i=1;i<=N;i++)
{

brownian=gaussian()*sqrt(T);

price_sample=MAX(0.0,K-x*exp((r-0.5*sigma*sigma)*T+sigma*brownian));

mean_price= mean_price+price_sample;
mean2_price= mean2_price+SQR(price_sample);
}
/* Price */
price=exp(-r*T)*(mean_price/N);
error_price= sqrt(exp(-2.*r*T)*mean2_price/N - SQR(price))/sqrt(N-1);
inf_price= price - 1.96*(error_price);
sup_price= price + 1.96*(error_price);
}

60