BMAN 70141

Derivative Securities
Session 4: Black-Scholes-Merton Formula

Dr Yuekun Liu
Office: AMBS 4.031, office hours: Wednesday (morning)
e-mail: <yuekun.liu@manchester.ac.uk>
...
Accounting & Finance Group, Manchester Business School
The University of Manchester
...
AMBS 3.006, 17 October 2024, 2-4 pm
session recording plus pre-recorded video on Blackboard
Content

▶ the previous session taught us how to value options using a

discrete-time approach (namely, the binomial tree approach);
▶ in this session, we turn to continuous-time approaches;
▶ to this end, we will deal with the following topics first:
1. an overview of Wiener processes;
2. modelling stock prices using Wiener processes;
3. deriving the Black-Scholes-Merton partial differential equation;
4. introduction to the Black-Scholes-Merton formula;
▶ I then provide examples on how to use the Black-Scholes-Merton
formula (one with and one without dividends);
▶ you’ll notice that this is important: without practising, you won’t be
able to correctly use the formula in the exam;
Stochastic Processes

▶ assumption: stock prices evolve randomly over time;

▶ the reason is that stock prices incorporate expectations about the
future ⇒ as a result, they only change with news;
▶ news, however, cannot be forecasted (by definition) and arrives
randomly over time, leading to random stock price changes;
▶ we hence model the stock price using a stochastic process:
▶ stochastic processes describe the law of motion according to
which a random variable (i.e., the stock price) evolves over time;
▶ they often include both a deterministic and a random component;
▶ two examples:
Case 1: each day, the stock price increases by $1 with
probability 30%, stays the same with probability 50%,
and decreases by $1 with probability 20%;
Case 2: each day, the stock price change is drawn from a normal
distribution with a mean of zero and some variance;
Markov Processes

▶ Markov processes only depend on the current level of the random

variable – historical values are of no importance for the future;
▶ do the two following variables follow a Markov process:
▶ the temperature in London;
▶ the stock price of Barclay’s Bank;
▶ modelling the stock price using a Markov process is consistent
with the assumption of a weakly efficient financial market:
▶ weak-form efficiency implies that historical prices contain no
information about future prices (“efficient market hypothesis”);
▶ in a weakly efficient market, technical trading is therefore futile;
▶ a good example of a Markov process is a random walk:
yt = yt −1 + εt ;
A Wiener Process
▶ assume that φ(µ, ν) denotes a normal distribution with
expectation equal to µ and variance equal to ν;
▶ a random variable z follows a Wiener process if its change,
denoted by ∆z, over a small time interval ∆t fulfills:
√
1. ∆z = ε ∆t, where ε ∼ φ(0, 1);
2. the values of ∆z for any two different (i.e., non-overlapping) time
periods are independent from one another;
▶ assuming z (0) = 0, it directly follows that:
▶ mean of (z (T ) − z (0)):
E [(z (T ) − z (0))] = E [∆z (1) + ∆z (2) + . . . + ∆z (T )]
= E [∆z (1)] + E [∆z (2)] + . . . + E [∆z (T )] = 0 + 0 + . . . + 0 = 0;
▶ variance of (z (T ) − z (0)):
var [(z (T ) − z (0))] = var [∆z (1) + ∆z (2) + . . . + ∆z (T )]
= var [∆z (1)] + var [∆z (2)] + . . . + var [∆z (T )] = 1 + . . . + 1 = T ;
√
▶ the standard deviation is the square root of the variance ( T );
An Example of a Wiener Process

▶ the value of a stochastic variable is currently 40;

▶ the variable follows a standard Wiener process;
▶ at the end of year 1, the mean of the variable is 40 and its
standard deviation is 10 (hence, the variance is 102 = 100);
▶ the share price of the end of the year can be written as:

ST = S0 + ∆z (1) + ∆z (2) + . . . + ∆z (100);

▶ what is the mean/variance after:

1. two years have passed?
2. one half year has passed?
3. a quarter of a year has passed?
4. ∆t of a year has passed?
▶ carefully note that variances are additive over time, while
standard deviations are clearly not additive;
A Generalized Wiener Process

▶ in a standard Wiener process, each increment of the process has

a mean equal to zero and a variance equal to one;
▶ in a generalized Wiener process, the drift rate and the variance
rate can be set to any chosen constants (with the variance > 0);
▶ the generalized Wiener process can be written as:
√
∆x = a ∆ t + b ε ∆t ,

where a represents the mean change in x per time unit, b2 the variance
of the change per time unit, and ∆t is the time interval (in years);
▶ if we let the time increment ∆t become smaller and smaller:
√
dx = adt + bε dt = adt + bdz .

▶ allowing a and b to be deterministic functions of x and t, we

obtain an Itô process (i.e., dx = a(x , t )dt + b(x , t )dz).
Why a Generalized Wiener Process is Inappropriate for
Modelling the Evolutions of Stock Prices
▶ do not use a generalized Wiener process to model stock prices:
▶ the stock price cannot become negative due to limited liability;
▶ the expected percentage change (i.e., return) should remain
constant over time — not the actual change;
▶ the variance of the actual change should depend on the stock
price level (greater price → greater variance);
▶ the following Itô process, often referred to as Geometric Brownian
motion (GBM), makes more sense for our purposes:

dS = µSdt + σSdz .

▶ the expectation and the variance of its change (dS) converge to

zero as the stock price (S) moves down to zero;
▶ zero is an absorbing barrier for that stochastic process;
▶ the discrete-time equivalent is:
√
∆S = µS ∆t + σS ε ∆t .
One Possible Sample Path for the GBM Process
More Than One Sample Path For the GBM Process
The Purpose of Itô’s Lemma

▶ we now introduce a mathematical rule allowing you to derive the

differential of a function depending on a stochastic variable;
▶ important for the derivation of the Black-Scholes-Merton formula;
▶ suppose that we have a function dependent on a deterministic
variable (time t) and a stochastic variable (the GBM x);
▶ an example: G(x , t ) = log (x ), with dx = µxdt + σxdz;
▶ how does the function evolve over time (i.e., what is dG)?
▶ to find out, use a Taylor series expansion of G(x , t ):
∂G ∂G 1 ∂2 G ∂2 G 1 ∂2 G 2
∆G = ∆x + ∆t + 2
∆x 2 + ∆x ∆t + ∆t + . . .
∂x ∂t 2 ∂x ∂x ∂t 2 ∂t 2
▶ in words: the change in G(x , t ) is determined by changes in x
and t, and the sensitivities of the function to these variables;
▶ now let ∆t become “very close to zero” to obtain dG;
The Derivation of Itô’s Lemma I
▶ dt being very close to zero (i.e., ∆t → dt) implies that we can
treat everything closer to zero than dt as zero;
▶ in ordinary calculus (i.e., if x did not evolve according to a
stochastic process), this would mean that:

∂G ∂G
dG = dx + dt .
∂x ∂t
▶ but when x evolves according to a GBM, then one of the
higher-order terms does not disappear;
√
▶ plug ∆x = µx ∆t + σx ε ∆t into the Taylor-series expansion:
∂G √ ∂G 1 ∂2 G √
∆G = (µx ∆t + σx ε ∆t ) + ∆t + (µx ∆t + σx ε ∆t )2
∂x ∂t 2 ∂x 2
| {z }
squared term involving x

∂2 G √ 1 ∂2 G 2
+ (µx ∆t + σx ε ∆t )∆t + ∆t + . . .
∂x ∂t 2 ∂t 2

and let’s look more carefully at the squared term involving x;

The Derivation of Itô’s Lemma II
▶ consider the squared term involving x in the prior sum:
∂2 G √ ∂2 G  3

2
(µx ∆t + σx ε ∆t )2 = 2 µ2 x 2 ∆t 2 + σ2 x 2 ε2 ∆t + 2µσx 2 ε∆t 2 ;
∂x ∂x
▶ letting ∆t → dt, we can ignore all terms smaller than dt because
(intuitively speaking) dt is as close to zero as possible:
∂2 G 2 ∂2 G 2 2 2

dx = σ x ε dt ;
∂x 2 ∂x 2
▶ for the same reason, the third term in the above sum disappears
as well as all higher order terms not explicitly shown (in “. . .”);
▶ finally, notice that because ε ∼ φ(0, 1):
E [ε] = 0;
var[ε] = E [ε2 ] − E [ε]2 = E [ε2 ] = 1;
var[ε∆t ] = ∆t 2 var[ε] → var[εdt ] = dt 2 var[ε] = 0;

▶ we can hence replace ε2 with 1 in the above equation;

Itô’s Lemma
▶ what we are left with is:
∂G ∂G 1 ∂2 G 2
dG = dx + dt + dx ;
∂x ∂t 2 ∂x 2
▶ plugging in for dx, we obtain:
∂G ∂G 1 ∂2 G
dG = (µxdt + σxdz ) + dt + (µxdt + σxdz )2
∂x ∂t 2 ∂x 2
∂G 1 ∂2 G 2 2
 
∂G ∂G
= µx + + σ x dt + σxdz ;
∂x ∂t 2 ∂x 2 ∂x

▶ an example: how does G(x ) = x 2 evolve over time when x

follows Geometric Brownian motion (i.e., dx = µxdt + σxdz)?
2
▶ we have: ∂∂Gx = 2x, ∂∂Gt = 0, and ∂∂xG2 = 2;
▶ and so:
1 1
dG = 2xdx + 0dt + 2dx 2 = 2x (µxdt + σxdz ) + 2(µxdt + σxdz )2
2 2
1
= 2x µxdt + 2x σxdz + 2σ2 x 2 dt = 2µx 2 + σ2 x 2 dt + 2σx 2 dz ;


2
Towards the Black-Scholes(-Merton) Formula:

▶ armed with Itô’s Lemma, we derive the famous Black-Scholes

formula (often called the Black-Scholes-Merton formula);
▶ the idea behind the BS formula is identical to that underlying the
binomial tree approach we discussed in the last session:
▶ form a portfolio consisting of some units of the stock and the
option held short which completely eliminates uncertainty;
▶ set the return of this portfolio equal to the risk-free rate;
▶ this yields a partial differential equation;
▶ to this end, note that stock value evolves according to:
∆S = µS ∆t + σS ∆z ;
▶ following from Itô’s lemma, the option value (a function of the
stock price and time) evolves following:
∂f 1 ∂2 f 2 2
 
∂f ∂f
∆f = µS + + σ S ∆t + σS ∆z ;
∂S ∂t 2 ∂S 2 ∂S
The Replication Portfolio
▶ buy δ units of the stock and short one option, denoting portfolio
value by Π and its change by ∆Π = δ∆S − ∆f :
∂f 1 ∂2 f 2 2
 
∂f ∂f
∆Π = δµS ∆t + δσS ∆z − µS + + σ S ∆t − σS ∆z
∂S ∂t 2 ∂S 2 ∂S
∂f 1 ∂2 f 2 2
   
∂f ∂f
= δµS − µS − − σ S ∆t + δ − σS ∆z ;
∂S ∂t 2 ∂S 2 ∂S
▶ the ∆z terms are responsible for the uncertainty in portfolio value
(the other terms are deterministic functions of time);
▶ choose δ = ∂∂Sf , so that the ∆z terms cancel out;
▶ a risk-free portfolio should pay the risk-free rate of return:
 
∂f ∂f
∆Π = r Π∆t ⇔ −∆f + ∆S = r −f + S ∆t ;
∂S ∂S
▶ plug in for ∆f and ∆S to obtain the BS differential equation:
∂f ∂f 1 ∂2 f
+ rS + σ2 S 2 2 = rf ;
∂t ∂S 2 ∂S
The Black-Scholes Differential Equation

▶ any security whose price is dependent on the stock price and

time satisfies the partial differential equation on the former slide;
▶ the particular security being valued is determined using
“boundary conditions” (e.g., the payoffs at maturity);
▶ when valuing a long forward, the boundary condition is:

f = S −K when t = T ;

▶ the unique solution to the equation is:

f = S − Ke−r (T −t ) .

▶ in general, it is extremely hard to solve partial differential

equations (but easier to solve ordinary differential equations);
The Black-Scholes Formula (No Dividends)

▶ Fisher Black and Myron Scholes managed to solve the partial

differential equation for European call and put options;
▶ if the underlying asset pays no dividends, their solutions are:
▶ call: c = S0 N (d1 ) − Ke−rT N (d2 );
▶ put: p = Ke−rT N (−d2 ) − S0 N (−d1 );
▶ where:
▶ d1 = ln(S0 /K )+(r +σ2 /2)T
√
σ T
ln(S0 /K )+(r −σ2 /2)T √
▶ d2 = √ = d1 − σ T
σ T
▶ we can interpret N (d2 ) (N (−d2 )) as the risk-neutral probability of
the call (put) option ending up in-the-money;
▶ swapping the risk-free rate for the stock’s actual expected return,
d2 would become a real-world probability;
The N(x) Function

▶ N(x) is the probability that a normally distributed random variable

with a mean of 0 and a standard deviation of 1 is less than X :

▶ to determine N(x), see tables at the end of the book;

▶ alternatively, type normdist(x,0,1,1) into any cell in EXCEL;
Cumulative Normal Distribution Table

2
▶ you can also use the approximation N (x ) ≈ e2(0.7988x (1+0.04417x ))
2 ;
1+e2(0.7988x (1+0.04417x ))
▶ as example, assume that x = −1:
▶ 0.7988(−1)(1 + 0.04417(−1)2 ) = −0.834082996;
2(−0.834082996)
▶ so that: N (−1) ≈ e 2(−0.834082996) = 0.158668853;
1+e
Practicing the Black-Scholes Model
▶ assume the following:
▶ a stock whose current value is 30 and whose volatility is 25%;
▶ a European (or American) call option written on the stock with
strike price equal to 36 and 16 months to maturity;
▶ the continuously-compounded risk-free rate is 2% p.a.;
▶ to find the value of the option:
▶ calculate d1 and d2 :
√
d1 = (ln(S0 /K ) + (r + 0.5σ2 )T )/(σ T )
p
= (ln(30/36) + (0.02 + 0.5 · 0.252 ) · (16/12))/(0.25 16/12)
= −0.394866789;
√ p
d2 = d1 − σ T = −0.394866789 − 0.25 16/12 = −0.683541924;

▶ look up N (d1 ) and N (d2 ) using a distribution table:

N (−0.394866789) ≈ 0.34827 and N (−0.683541924) ≈ 0.24825;

▶ plug into BS formula (S0 N (d1 ) − Ke−rT N (d2 )):

30 · 0.34827 − 36 · e−0.02(16/12) 0.24825 = 1.746270458;

The Impact of Dividends on Black-Scholes Pricing

▶ dividends change the stock price distribution at maturity;

▶ in an efficient market, a dividend per share of $5 reduces the
stock price by $5 (. . . taxes reduce this drop in the share price);
▶ dividend payouts can be modelled in two ways:
1. continuous payouts occurring at a rate of q;
2. discrete payouts occurring at specific points in time;
▶ in each case, the dividend payouts shift downwards the
distribution of the stock price at maturity;
▶ instead of decreasing the stock value at the time dividends occur,
decrease the initial stock price by the PV of the dividends;
1. continuous payouts: S0 e−qT ;
2. discrete payouts: S0 − PV(Dividend payouts);
▶ value European options by decreasing the initial stock prices and
then pretend that there are no dividend payouts until maturity;
The Black-Scholes Formula (Continuous Dividends)

▶ if there are dividends, the BS formulae become:

▶ call: c = S0 e−qT N (d1 ) − Ke−rT N (d2 );
▶ put: p = Ke−rT N (−d2 ) − S0 e−qT N (−d1 );
▶ where:
r −q +σ2 /2)T −qT r +σ2 /2)T
▶ d1 = ln(S0 /K )+(√ = ln(S0 e /σK√)+(
σ T T
r −q −σ2 /2)T √
▶ d2 = ln(S0 /K )+(√ = d1 − σ T
σ T
▶ we can use these formulae to value European index options (set
S0 to the current index level and q to the average dividend yield);
▶ we can also use them to value European currency options (set S0
to the FX rate, and q to the foreign interest rate);
Practicing the Black-Scholes Model With Dividends
▶ stock now pays a $2 dividend after ten months;
▶ calculate the present value (PV) of the dividend:

PV (Div ) = 2 · e−0.02(10/12) = 1.96694290;

▶ the new stock price is thus 30 − 1.96694 = 28.03305709;

▶ to find the value of the option:
▶ calculate d1 and d2 :
√
d1 = (ln((S0 − D )/K ) + (r + 0.5σ2 )T )/(σ T )
p
= (ln(28.03305/36) + (0.02 + 0.5 · 0.252 ) · (16/12))/(0.25 16/12)
= −0.629778641;
√ p
d2 = d1 − σ T = −0.629778641 − 0.25 16/12 = −0.918453776;

▶ look up N (d1 ) and N (d2 ) using a distribution table:

N (−0.629778641) ≈ 0.26435 and N (−0.918453776) ≈ 0.17879;

▶ plug into BS formula ((S0 − D )N (d1 ) − Ke−rT N (d2 )):

28.03305 · 0.26435 − 36 · e−0.02(16/12) 0.17879 = 1.143466863;

Risk-Neutral Valuation

▶ the expected stock return does not appear in the BS formula;

▶ the solution to the differential equation is therefore the same in a
risk-neutral world as it is in the real-world;
▶ to value the derivative in the real-neutral world:
1. assume that the expected stock return is the risk-free rate;
2. calculate the expected payoff from the option;
3. discount at the risk-free rate;
▶ an example: the payoff of a forward contract is: ST − K ;
▶ the expected payoff in the risk-neutral world is S0 erT − K ,
discounted at the risk-free rate of return:

e−rT [S0 erT − K ] = S0 − Ke−rT .

▶ the details are in BMAN 71541: “Stochastic Calculus”;

Synopsis

▶ in this session, we learned about valuing options using the

continuous-time approach of Black, Scholes, and Merton;
▶ we hence reviewed the following topics:
▶ basic principles of stochastic processes;
▶ Itô’s Lemma helping us to determine the change in a function
dependent on a stochastic variable (the stock price) and also time;
▶ replication portfolios in continuous-time;
▶ the intuition of continuous-time approaches is similar to that
underlying simpler binomial tree valuation approaches:
1. we form a dynamic riskfree portfolio of option and stock;
2. the portfolio needs to fulfill a certain partial differential equation;
3. we solve that equation subject to boundary conditions;
▶ the solution is the famous Black-Scholes option pricing formula;
▶ the best-known equation in the whole realm of finance;
▶ Myron Scholes and Robert Merton won the Nobel price for it;