You are on page 1of 30

Home Page

Option Pricing
Title Page

Contents

JJ II

J I
December 11, 2019

Page 1 of 30

Go Back

Full Screen

Close

Quit
1. Introduction

Home Page
An option is a contract giving the buyer the right, but not
the obligation, to buy (call) or sell (put) the underlying as-
Title Page set at a specific price on or before a certain date.
Contents
Options are known as derivatives because they derive their
JJ II value from an underlying asset.
J I
European call option on stock with strike K and expiration
Page 2 of 30
T : Holder can buy one share of stock for price K at time T .
Go Back Payoff is (ST − K)+ .
Colab link for payoff function
Full Screen

Close

Quit
Fundamental Theorem

Home Page
If there is no arbitrage and r is the interest rate, then there
exists measure (called a risk-neutral measure Q) such that
Title Page the price of the call option is
Contents
EQ e−rT (ST − K)+ . (1)
JJ II
In fact the first fundamental theorem of asset pricing asserts
J I
that a market is arbitrage free iff there exists a risk neutral
Page 3 of 30 measure. The latter is defined as a measure that is equiva-
lent to the physical measure under which discounted prices
Go Back of all traded assets are martingales.
Full Screen
HW: Show that C + e−rT K < S0 admits arbitrage. Here
Close
C is the price at time 0 of a European CALL option with
Quit strike K maturity T . The price of the underlying STOCK
at time 0 is S0 and the risk free interest rate is r. Trading
is allowed at times 0 and T in STOCK, BOND and CALL.
Black-Scholes set-up

The dynamics of the underlying asset is represented by the


Home Page
geometric Brownian motion
Title Page dSt = µSt dt + σSt dBt
Contents while interest rates follow the following non-stochastic dif-
ferential equation
JJ II
dRt = rRt dt.
J I Here Bt is standard Brownian motion. This allows the ex-
plicit solution
Page 4 of 30
St = S0 exp (µ − σ 2 /2)t + σBt .

(2)
Go Back
Let C(t, x) be the function that describes the price of a
Full Screen contingent claim. Then, C(t, x) solves the following (non-
stochastic) partial differential equation:
Close
1
Quit
rC(x, t) = Ct (x, t) + rxCx (x, t) + σ 2 x2 Cxx (x, t).
2
This is obtained from Ito calculus and the first fundamental
theorem.
Under the risk neutral measure, µ = r.

2 /2
Home Page HW: Show this using EQ (e−rt St = S0 and mgf(u)=eu for
a standard normal random variable.
Title Page

Contents Under the Black Scholes set-up, the price of Call option is
JJ II S0 Φ(d1 ) − Ke−rT Φ(d2 )
J I 2

where d1 = (log(S
√ 0 /K) + (r + σ /2)T )/(σ T)
Page 5 of 30 and d2 = d1 − σ T .
Φ is the standard normal cdf.
Go Back

Full Screen HW: Show this using (2), (1) and µ = r.


Close Colab link for option pricing using formula, R function and
simulation
Quit
Examples of Options

American option can be exercised at any time before the


Home Page maturity date.
Title Page

A basket option is an option on the weighted average of sev-


Contents
eral underlyings.
JJ II

J I
A lookback option is a path dependent option where the
option owner has the right to buy (sell) the underlying in-
Page 6 of 30 strument at its lowest (highest) price over some preceding
period.
Go Back

Full Screen An Asian option is an option where the payoff is not de-
Close
termined by the underlying price at maturity but by the
average underlying price over some pre-set period of time.
Quit

A barrier option involves a mechanism where if a ’limit price’


is crossed by the underlying, the option either can be exer-
cised or can no longer be exercised.
2. The Greeks

Home Page
Delta is the amount an option price is expected to move
based on a $1 change in the underlying stock. Calls have
Title Page positive delta, between 0 and 1. That means if the stock
price goes up and no other pricing variables change, the
Contents
price for the call will go up. Puts have a negative delta,
JJ II between 0 and -1.
Colab link for computing Theta with Centered differencing
J I and Richarson’s extrapolation
Page 7 of 30

Delta is also important in Hedging as it equals the num-


Go Back
ber of shares of stock to hold in a replicating self financing
Full Screen portfolio.

Close

Quit
Gamma is the second-derivative of the option price with
respect to the stock price. Gamma indicates the amount
the delta would change given a unit move in the underlying
Home Page
security.
Title Page Theta is the amount the price of calls and puts will decrease
Contents
(at least in theory) for a one-day change in the time to expi-
ration. Each moment that passes causes some of the option’s
JJ II time value to “melt away.” So theta is usually positive.
J I

Page 8 of 30

Go Back

Full Screen

Close

Quit
Vega is the amount call and put prices will change for a unit
change in volatility. Typically, as volatility increases, the
value of options will increase. That’s because an increase in
Home Page
volatility suggests an increased range of potential movement
Title Page for the stock.

Contents
HW:
√ Show that for European CALL option “vega” equals
JJ II S0 T φ(d1 ), where φ is the normal pdf.
J I
These relations are also called sensitivities.
Page 9 of 30

Go Back Colab link for plotting option price vrs volatility. Also all
Greeks using R function.
Full Screen

Close

Quit
3. Volatility

Home Page
The σ in the option price is the variance/volatility of the
stock price.
Title Page

Contents The value of σ obtained by inverting the observed option


price is called the Implied Volatility.
JJ II

J I The variance of stock price obtained from observed stock


prices is Realized Volatility.
Page 10 of 30

Go Back The difference is called Market Price of risk. This is


Full Screen
generally positive, that is the observed price of option es-
timates volatility more than it is in the stock prices. This
Close is the price that the option market demands for providing
protection against extreme movements of the stock.
Quit
Compute Implied Vol

Home Page
1. Compute implied volatility by bisection method

Title Page
2. General Newton-Raphson update

Contents x1 = x0 − f (x0 )/f 0 (x0 ). (3)

JJ II This amounts to approximating the function as being


linear and using the root of the approximation as the
J I
updated guess. This is a faster root-finding algorithm
Page 11 of 30 than bisection.
Write the program in R for computing implied volatil-
Go Back ity using Newton Raphson method.
Full Screen
guess=guess-call/vega
Close

3. A similar idea that does not require the computa-


Quit
tion of derivative is to keep track of two most recent
(x,f(x)) and use the difference quotient. Write the pro-
gram in R for computing implied volatility using Se-
cant method.
vega=(call-prior_call)/(guess-prior_guess)

Home Page

Title Page

Contents

JJ II

J I

Page 12 of 30

Go Back

Full Screen

Close

Quit
The values of implied volatility obtained by inverting op-
tions of different maturities and strikes are not same. One
observes low IV when K is close to S0 and higher values far
Home Page
away, leading to the nomenclature of IV Smile.
Title Page These are often combined to form a single index, the most
Contents
common measure being the VIX followed by CBOE.
Relevance of Options
JJ II

J I 1. Trade options to protect yourself against extreme move-


ments of the market.
Page 13 of 30

2. Use option prices to estimate the distribution and dy-


Go Back
namics of the underlying stock price process.
Full Screen
3. Use option price together with stock price to estimate
Close the market price of risk.

Quit
4. Solution of PDEs for American
options
Home Page

Black-Scholes PDE was stated as


Title Page

1
Contents rC(x, t) = Ct (x, t)+rxCx (x, t)+ σ 2 x2 Cxx (x, t). (x, t) ∈ D,
2
JJ II
(4)
where t is a time variable, x is a state variable, and C(x, t)
J I is an unknown function.
Page 14 of 30
Note that with a change of variables this is same as the heat
Go Back equation
∂u ∂ 2u
Full Screen = c2 2 , (5)
∂t ∂x
Close

Quit
To find a well-defined solution, we need to impose the final
condition
C(x, T ) = CT (x)
Home Page
and, if D = [a, b]X[0, T ], the boundary conditions
Title Page

C(a, t) = ga (t) and C(b, t) = gb (t)


Contents

JJ II where CT , ga , gb are continuous functions.

J I
The diffusion equation (4) with the initial condition (4) and
Page 15 of 30 the boundary conditions (4) is well-posed, i.e. there exists a
unique solution that depends continuously on u0 , ga and gb .
Go Back

Full Screen

Close

Quit
Grid Points

Home Page
To find a numerical solution to equation (4) with finite dif-
ference methods, we first need to define a set of grid points
Title Page in the domain D as follows:
Choose a state step size ∆x = (b − a)/M
Contents
and a time step size ∆t = T /N (M and N are an integers).
JJ II

J I Draw a set of horizontal and vertical lines across D, and get


all intersection points (xj , ti ), or simply (j, i),
Page 16 of 30 where xj = a + j∆x, j = 0, · · · , M
and ti = i∆t, i = 0, 1, · · · , N
Go Back

Full Screen

Close

Quit
Finite Differences

2
Home Page
The partial derivatives Cx := ∂C
∂x
and Cxx := ∂∂xC2 are always
approximated by central difference quotients, i.e.
Title Page
i i i
Cj+1 − Cj−1 Cj+1 − 2Cji + Cj−1
i
Contents Cx ≈ and Cxx ≈
2∆x (∆x)2
JJ II
at a grid point (j, i). Here Cji = C(xj , ti ). Depending on how
J I Ct is approximated, we have three basic schemes: explicit,
implicit, and Crank–Nicolson schemes.
Page 17 of 30

Go Back

Full Screen

Close

Quit
Explicit Scheme

Home Page
If Ct is approximated by a backward difference quotient

Title Page Cji − Cji−1


Ct ≈
∆t
Contents

at (j, i), then the corresponding difference equation to (4)


JJ II
at grid point (j, i) is
J I
Cji−1 = A(j)(Cj−1
i
+ B(j)Cji + C(j)Cj−1
i
,
Page 18 of 30
rx σ 2 x2
where A(j) = ∆t(− 2∆x + 2(∆x)2
).
Go Back
σ 2 x2
B(j) = 1 − ∆t(r + (∆x)2
)
Full Screen rx 2
σ x 2
C(j) = ∆t( 2∆x + 2(∆x)2
).
Close

Quit
The difference equations 4 together with terminal and bound-
ary conditions can be solved explicitly, starting from the end
time T and going backwards. Colab link for explicit scheme
American Option

Home Page
For an American option, if the price of Put option at any
point of time is lower than the value( (K − S)+ ), then one
Title Page can buy the put and execute at once to make guaranteed
profit. This is an arbitrage situation.
Contents

JJ II So we need to introduce the extra condition that at every


J I step, P ≥ (K − S)+ .

Page 19 of 30
It can be shown that this is indeed a sufficient condition.
Go Back Colab link for pricing American option using explicit scheme

Full Screen

Close

Quit
Implicit Scheme

If Ct is approximated by a forward difference quotient


Home Page

Cji+1 − Cji
Title Page
Ct ≈
∆t
Contents
at (j, i), then the corresponding difference equation to (4)
JJ II at grid point (j, i) is
J I i
D(j)(Cj+1 i
+ E(j)Cji + F (j)Cj−1 = Cji+1 ,
Page 20 of 30 In matrix form
 
Go Back F (1) G(1)
 E(2) F (2) G(2) 
Full Screen
 C = b.
 
 .. ..
 . . 
Close
E(M ) F (M )
Quit
Since the matrix is in tri-diagonal form, the difference equa-
tions (4), together with the initial and boundary conditions
as before, can be solved using the Crout algorithm or the
SOR algorithm.
Crank–Nicolson Scheme

Home Page
The Crank–Nicolson scheme is the average of the explicit
scheme at (j, i) and the implicit scheme at (j, i + 1). The
Title Page resulting difference equation is
!
Contents
Cji+1 − Cji i
Cj−1 − 2Cji + Cj+1
i i+1
Cj−1 − 2Cji+1 + Cj+1
i+1
= σ 2 x2 +
JJ II ∆t (2∆x)2 (∆x)2
J I
The difference equations (4), together with the initial and
Page 21 of 30 boundary conditions as before, can be solved using Crout
algorithm or SOR algorithm.
Go Back

Full Screen

Close

Quit
Local Truncation Errors

Home Page
These are measures of the error by which the exact solution
of a differential equation does not satisfy the difference equa-
Title Page tion at the grid points and are obtained by substituting the
exact solution of the continuous problem into the numerical
Contents
scheme.
JJ II

J I A necessary condition for the convergence of the numerical


solutions to the continuous solution is that the local trunca-
Page 22 of 30 tion error tends to zero as the step size goes to zero. In this
case the method is said to be consistent. It can be shown
Go Back
that all three methods are consistent. The explicit and
Full Screen implicit schemes have local truncation errors O(∆t, (∆x)2 ),
while that of the Crank–Nicolson scheme is O((∆t)2 , (∆x)2 ).
Close

Quit
Stability

Home Page
A scheme is stable if roundoff errors are not amplified in the
calculations.
Title Page

Contents The Fourier method can be used to check if a scheme is sta-


ble following von Neumann stability condition.
JJ II

J I It can be shown that the explicit scheme is stable if and


∆t
only if λ = (∆x) 2 ≤ 1/2, called conditionally stable, and
Page 23 of 30
the implicit and Crank–Nicolson schemes are stable for any
Go Back values of λ, called unconditionally stable.

Full Screen

Close

Quit
5. Pricing Path dependent Options

Home Page
The payoff depends on the entire path of the underlying and
not only the value at the point of execution.
Title Page

Contents One way is through simulation of the entire path multiple


times.
JJ II
RT
J I Payoff of Asian Option is ( T1 0
St dt − K)+ .
Page 24 of 30
Discretize the time and approximate by ( n1 ni=1 Sti/n −K)+ ,
P
Go Back compute this over M realizations of the path and take aver-
Full Screen
age.
Colab link for pricing Asian option by simulating path and
Close
discretizing time. Then pricing using various approxima-
Quit tions.
Advisable to parallelize over the different paths.
Why does averaging work

Home Page
Consider the problem of estimating an expectation α=E[f (U )],
with U uniformly distributed between 0 and 1.
Title Page

Contents Suppose we have a mechanism for drawing points U1 , U2 , · · ·


independently and uniformly from [0, 1].
JJ II

J I Evaluating the function f at n of these random points and


averaging the results produces the Monte Carlo estimate
Page 25 of 30
n
Go Back 1X
α̂n = f (Ui ).
n i=1
Full Screen

Close
If f is integrable over [0, 1] then, by the strong law of large
numbers,
Quit
α̂n → α with probability 1 as n → ∞.
If f is in fact square integrable and we set
Z 1
2
Home Page
σf = (f (x) − α)2 dx
0
Title Page
then by the Central Limit Theorem, the error α̂n − α in the
Contents Monte Carlo estimate is approximately normally
√ distributed
with mean 0 and standard deviation σf / n, the quality of
JJ II this approximation improving with increasing n.
J I
The parameter σf would typically be unknown in a setting
Page 26 of 30
in which α is unknown, but it can be estimated using the
Go Back sample standard deviation
n
Full Screen 1 X
s2f = (f (Ui ) − α̂)2 .
n − 1 i=1
Close

Quit Thus, from the function values f (U1 ), · · · , f (Un ) we obtain


not only an estimate of α but also a measure of the error in
this estimate.

The form of the standard error σf / n is a central feature
of the Monte Carlo method.
Home Page
In contrast, the error in the simple trapezoidal rule is O(n−2 ),
Title Page at least for twice continuously differentiable f .
Contents

Monte Carlo is generally not a competitive method for cal-


JJ II
culating one-dimensional integrals.
J I

Page 27 of 30
The value of Monte Carlo as a computational tool lies in the
fact that its O(n−1/2 ) convergence rate is not restricted to
Go Back integrals over the unit interval.
Full Screen

Close

Quit
The problem is now to generate random vectors (SiT /n , i =
1, · · · , n) where S follows geometric Brownian motion
Home Page
dSt = µSt dt + σSt dBt
Title Page
Brownian motion is a process starting at zero with contin-
Contents uous paths and independent normal increments with mean
zero and variance equal to the time interval.
JJ II
BiT /n − B(i−1)T /n ∼ N (0, T /n) and independent over all i.
J I
p
Page 28 of 30 Hence BiT /n = B(i − 1)T /n + T /nZi where Z1 , · · · , Zn
are iid standard normal.
Go Back
p
Full Screen From (2), SiT /n = S(i−1)T /n exp((µ − σ 2 /2)T /n + σ T /nZi ).
Close Colab link for simulating geometric Brownian motion

Quit
Barrier option

Home Page
A barrier option is “knocked out” if the underlying asset
crosses a prespecified level.
Title Page

Contents For instance, a down-and-out call with barrier b, strike K,


and expiration T has payoff 1τ (b) > T (S(T ) − K)+, where
JJ II τ (b) = inf t : S(t) < b is the first time the price of the un-
J I derlying asset drops below b.

Page 29 of 30
A down-and-in call gets “knocked in” only when the under-
Go Back lying asset crosses the barrier. Up-and-out and up-and-in
calls and puts are defined analogously.
Full Screen

Close These are discretely monitored if t ∈ 0 ≤ t1 < · · · < tn ≤ T .


They can be priced by simulation through sampling of
Quit
S(t1 ), · · · , S(tn ), S(T ).
Lookback options

Home Page
Lookback puts and calls expiring at tn have payoffs

Title Page (maxi=1,··· ,n S(ti ) − S(tn ))

Contents and
(S(Tn ) − mini=1,··· ,n S(ti ))
JJ II
respectively.
J I

Page 30 of 30 A lookback call, for example, may be viewed as the profit


from buying at the lowest price over t1 , · · · , tn and selling at
Go Back
the final price S(tn ).
Full Screen

Close
Continuously monitored versions of these options are defined
by taking the maximum or minimum over an interval rather
Quit than a finite set of points.

Both lookback and barrier options can be valued by simula-


tion of the entire path.

You might also like