Arbitrage Pricing

Paul Schneider

Università della Svizzera Italiana

February 13, 2023

1 / 41
 Introduction

Goals of the Class

Basic instruments
Forwards
Futures
Options
No-arbitrage in discrete time with discrete states
Discrete-time models
Continuous-time models
Purpose
Construction
Applications
To get there
Review: random variables, probability, statistics
Estimation, calibration
Monte-Carlo simulation

2 / 41
 Introduction

Personal & Assessment

Personal
[email protected]
[email protected]
Office hours on demand
Modus
Theory (Paul)
Practice (Marc)
Keep checking icorsi for info and lecture slides

3 / 41
 Introduction

Literature

Main textbook:
Stochastic Calculus for Finance I,II, Steven E. Shreve, 2004 Springer
Useful references:
Arbitrage Theory in Continuous Time, Tomas Björk, 1998 Oxford
University Press
Interest Rate Models – Theory and Practice, 2006 Springer
Statistical Inference, George Casella, Roger L. Berger, 2002 Duxbury
Advanced Series

4 / 41
 Introduction

Basic Ideas
Instruments
Payoffs
Strategies
No-arbitrage
Discrete and finite state-sapce
Basic principle
Structure that comes with no-arbitrage
Linearity
Convexity
Stochastic processes
Discrete-time
Continuous-time models
Stochastic processes and no-arbitrage
Basic construction
No-arbitrage in continuous-time
Black-Scholes
Affine term structure models
5 / 41
 Introduction

Risk and Uncertainty

1600

1400

1200

1000

800

600

400

200
1990 1994 1998 2002 2006

6 / 41
 Introduction

VXO Implied Volatility

0.5

0.45

0.4

0.35

0.3

0.25

0.2

0.15

0.1

0.05
1990 1994 1998 2002 2006

7 / 41
 No Arbitrage

Portfolios and Payoffs

Start from J securities (assets)
There is today and tomorrow
There are S possible states
In state s, asset j pays xjs tomorrow
Denote by xj ∈ RS the payoffs of asset j
The payoff matrix  
x1
 .. 
D :=  . 
xJ
describes the possible states of the world
A portfolio is composed of a vector h of holdings in the J securities
and pays
hD ∈ RS
8 / 41
 No Arbitrage Portfolio Example

A Simple Market with a Stock and an Option

Example 1 (Simple Market).

Take a market X := R3 and assume there is one risky asset in the market
with payoff
x = [1, 2, 4] (1)
and a European call option on x with strike K = 2 and payoff

y = (x − K )+ = [0, 0, 2] (2)

In this economy the set of reachable claims

H = {z ∈ X : z = αx + βy } (3)

Definition 2 (Portfolio subspace).

Denote by [D] := hD : h ∈ RJ the set of all possible portfolio payoffs.


9 / 41
 No Arbitrage Portfolio Example

A Simple Market Illustration

The below picture shows the hyperplane spanned by the payoffs from
the previous slide as a function of the portfolio weights α and β

12
10
8
6
4
2
0
4
3.5
3
0 2.5
2
0.5 1.5 β
1 1
α 1.5 0.5
2 0
10 / 41
 No Arbitrage Complete Markets

Definition 3 (Complete Markets).

A market is complete, if [D] = RS

If markets are complete, any element of RS can be reached from a

portfolio position
An asset is redundant if its payoffs can be generated as a portfolio of
other assets
Market (in)completeness is empirically hard to test

11 / 41
 No Arbitrage Law of One Price

Two Assets and Three States Positive Portfolios

The below picture shows the payoffs from a three state - two asset
economy with payoff vectors [1, 2, 3] and [3, 2, 1]
On the x and y axes are portfolio weights. The red lines bound the
portfolio weights with positive payoffs in all states
3 state 1
pos. port.
2.5

2 state 2

1.5

1 state 3
β

0.5

-0.5

-1

-1.5
-1 0 1 2 3 4
α

12 / 41
 No Arbitrage Law of One Price

The Law of One Price

Denote by p : RS 7→ R the price of a payoff Z ∈ [D]

Definition 4 (Law of One Price (LOOP)).

The law of one price is satisfied if p is linear

p(αx + βy ) = αp(x) + βp(y ), (4)

for x, y ∈ [D] all α, β ∈ R

The LOOP says that the price of a portfolio is the sum of the prices
of the constituents
This is a realistic assumption most of the times

13 / 41
 No Arbitrage Arbitrage

Arbitrage

Definition 5 (Arbitrage).

A payoff Z ∈ [D] such that (p(Z ) ≤ 0 and Z > 0) or (p(Z ) < 0 and
Z >= 0) is called an arbitrage.

Absence of arbitrage is a natural requirement for any market

Rational investors would take infinite positions if there were an
arbitrage
Excess demand would force the prices to be positive

14 / 41
 No Arbitrage Arbitrage

Fundamental Theorem of Asset Pricing

Here p : RJ × RS 7→ RJ is the vector version of p : RS 7→ R from

slide 13

Theorem 2.1 (Fundamental Theorem of Asset Pricing).

h is trading strategy
There does not exist a portfolio rule h such that

hD  0 and hp(D) ≤ 0

if and only if there exists M in RS such that

p(D) = DM and M  0.

15 / 41
 No Arbitrage Arbitrage

Two Assets and Three States Positive Portfolios

The below picture shows the region of portfolio with negative prices
(shaded)
x axis and y axis are portfolio weights as before

1.5

1 p = (p1 , p2 )

0.5

0
β

-0.5

-1

-1.5

-2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
α

16 / 41
 No Arbitrage Arbitrage

Two Assets – No Arbitrage

The below picture shows the region of portfolio with negative prices
(shaded green)
x axis and y axis are portfolio weights as before

3 state 1
p
2 state 2

1 state 3

0
β

-1

-2

-3

-4
-4 -3 -2 -1 0 1 2 3 4
α

17 / 41
 No Arbitrage Arbitrage

Two Assets – Weak Arbitrage

The below picture shows the region of portfolio with negative prices
(shaded green)
x axis and y axis are portfolio weights as before

3 state 1

2 state 2

1 state 3 and price

0
β

-1

-2

-3

-4
-4 -3 -2 -1 0 1 2 3 4
α

18 / 41
 No Arbitrage The Pricing Kernel

Vectors and Random Variables

Connect payoffs in vectors with random variables

Y = [y1 , . . . , yS ] ⇔ random variable Ỹ

with realizations y1 , . . . , yS that obtain with probability 1/S

Example 6.

Denote by Ms the value of M in state s in Theorem 2.1

Compute M0 = M1 + . . . + MS
Set M̂ = (M1 , . . . , MS )/M0
Then we can write for any payoff Z ∈ [D]
S
p(Z ) X h i
= M̂s Zs = Ê Z̃
M0
i=1

19 / 41
 No Arbitrage The Pricing Kernel

The Pricing Kernel

Normalized variable M̂ from Theorem 2.1 delivers a probability
representation of prices!
The variable M̂ is called the pricing kernel or state price density
Suppose the world is governed by an objective distribution P with
probabilities p1 , . . . , pS
We can then write
 
P dQ
h i
p(Z ) = M0 Ê Z̃ = M0 E Z̃ . (5)
dP
where the entries of M̂ denote the probabilities of distribution Q with
probabilities q1 , . . . , qS
dQ
dP denotes the likelihood ratio of Q with respect to P

dQ M̂s qs
(s) = =
dP ps ps
Write M̃ := M0 dQ
dP
20 / 41
 No Arbitrage Example: Binomial Model

Arbitrage Free Pricing in One Period Models

Example 7 (Replicating Portfolio).

Consider a market where there are only two points in time, 0 and t > 0
and two instruments, a bank account B and a stock S. The spot rate R is
known and the world evolves to two possible states, an upstate and a
downstate.

B0 = 1
Bt = 1 + R
S0 = s
(
u with probability pu
St = sZ , Z =
d with probability pd

21 / 41
 No Arbitrage Example: Binomial Model

Replicating Portfolio

Consider now a portfolio where you buy x of the bond and y of the
stock, then the value V h of your portfolio with strategy h at time 0
and at time t will be given by

V0h = x + ys
Vth = x(1 + R) + ysZ

Let us think about a clever way how to choose x and y in order to

learn about the arbitrage-free economy (cf. Definition 5)

22 / 41
 No Arbitrage Example: Binomial Model

Arbitrage Bounds in the Binomial Model

Let us now construct a portfolio that is worth 0 at time 0. We can

achieve that by setting x + ys = 0. Using this relation and by
plugging into the portfolio value function we get
(
ys(u − (1 + R)) if Z = u
Vth =
ys(d − (1 + R)) if Z = d

We notice immediately that for y > 0 and u > 1 + R, d > 1 + R, and

similarly for y < 0 and u < 1 + R, d < 1 + R there are arbitrage
opportunities in the market. To exclude both cases we get the very
important relation
d ≤1+R ≤u

23 / 41
 No Arbitrage Example: Binomial Model

Risk Neutral Probabilities in the Binomial Model

From the last slide we can express 1 + R as a convex combination of

u and d.
1 + R = qu u + qd d
Fix qu and qd to be positive and qu + qd = 1 (probabilities).
With qu and qd being probabilities and the above equation we can
write
Risk Neutral Probabilities
1 1 1
s= s(1 + R) = (qu su + qd sd) = EQ [St ]
1+R 1+R 1+R

Asset pricing formula is independent of the objective probabilities pu

and pd from slide 21. Using the no arbitrage property, it gives us the
price of an asset as the expectation under a probability measure Q

24 / 41
 No Arbitrage Pricing Kernels and Measure Changes

From Eq. (5) h i

p(Z ) = EP M̃ Z̃ (6)
Price a claim W that pays 1 in every state of the world

p(W ) = M0

p(W ) is the price of a zero bond

If there is time value of money it is reasonable to assume M0 ≤ 1
Suppose asset Y > 0 a.s. then p(Y ) > 0
" #
h i M̃ Ỹ M̃ Ỹ dY
p(Y ) = EP M̃ Ỹ ⇔ 1 = EP ⇔ =:
p(Y ) p(Y ) dP

Rewrite Eq. (6)

  " #
P dY p(Y ) p(Z ) Y Z̃
p(Z ) = E Z̃ ⇔ =E
dP Ỹ p(Y ) Ỹ

25 / 41
 No Arbitrage Pricing Kernels and Measure Changes

Fundamental Pricing Equation

From the previous slide written explicitly with time indices t ≤ T

 
pt (ZT ) Y ZT
= Et (7)
pt (YT ) YT

Asset Y is arbitrary, as long as it is strictly positive

Asset Y is called the numeraire asset
Which measure/numeraire to take for pricing?
Subsequent slides switch to index variables with time rather than
periods
First introduce some basic instruments

26 / 41
 No Arbitrage Widely Used Pricing Measures

Forward Measure
The probability measure associated with the zero bond price pt,T as
numeraire is called the QT forward measure
Writing the asset pricing equation (7) with the zero bond numeraire
 
Xt QT XT
= Et = EQ T
t [XT ] (8)
pt,T pT ,T
In this expectation we have just XT – no dependence problems
Exercise 2.1.
Show that Q = QT if and only if interest rates are deterministic.

Exercise 2.2 (First Glimpse at Forward Measure).

There are two points in time, today and tomorrow. There are two assets in
the market. A risky stock S which is worth 100 today and either 120 or 80
w.pr. 1/2, and a bond p which is worth 99 today and either 99.5 or 98.9

2
tomorrow. Price a derivative claim which pays Sp tomorrow.
27 / 41
 No Arbitrage Widely Used Pricing Measures

Change of Measure

Using the expectations with bank account and bond as numeraire

 
QT Q BT
Et [XT ] = Et · XT
BT pt,T
 
− tT ru du
R (9)
e
= EQt
 h i · XT

− tT ru du
R
EQ
t e

The process RT
e − t ru du dQT
h RT i =:
Q
Et e − t ru du dQ

is a Q martingale

28 / 41
 No Arbitrage Basic Instruments

Forward Price

Forwards are (over-the-counter) OTC contracts. This means you need

to search for a counter party to enter one
Futures are institutionalized, where everything goes through the
clearing house and no search is necessary
You can buy forwards on commodities, interest rates, exchange rates,
pork bellies ...
The cashflow of a forward on an underlying X in a long position looks
like this
Position cashflows today final (net) cashflows
long forward on X0 0 XT − F
But how is the value of F determined? Consider the arbitrage
portfolio strategy on the next slide...

29 / 41
 No Arbitrage Basic Instruments

Forward Price II

We assume that shortselling is allowed and denote as before by pt,T

the value of a zero coupon bond at time t with maturity T and face
value 1
Position cashflows today final (net) cashflows
long forward on X0 0 XT − F
Lend X0 CHF −X0 X0 /p0,T
Short commodity X0 −XT
Sum 0 X0 /p0,T − F
The above payoff is completely deterministic and with X0 /p0,T > F it
would be strictly greater than zero. Does this tell us anything about
the value of F ? If we assume that the markets are arbitrage free then
the answer is yes!

30 / 41
 No Arbitrage Basic Instruments

Short Rates and the Bank Account

Definition 8 (Short Rate and bank Account).
Denote by pt,T the price of a zero bond at time t, paying one unit
currency at time T with certainty. Define implicitly the simply
compounded rate L through
1
pt,T = (10)
1 + (T − t)L(t, T )

The short rate r (t) is defined as

rt := lim L(t, T ) (11)

T →t

Denote with Bt a continuously compounded bank account. Starting with

one unit currency (B0 = 1) we have
Rt
ru du
Bt = e 0 (12)
31 / 41
 No Arbitrage Basic Instruments

Instantaneous Forward Rates

Definition 9 (Simply-compounded Forward Rates).

The simply-compounded forward rate F (t, T , S) is defined as
pt,T
= 1 + (S − T )F (t, T , S)
pt,S

Definition 10 (Instantaneous Forward Rates).

The instantaneous forward rate ft,T is defined as

∂ log pt,T
ft,T ≡ lim F (t, T , S) = − (13)
S→T ∂T

Note that the above implies that

RT
pt,T = e − t ft,u du

32 / 41
 Options

Puts and Calls

Definition 11.
A European option on an underlying S pays at maturity T
(ST − K )+ call option
(K − ST )+ put option.
Denote the price at time t ≤ T of a European call option with strike price
K by Ct,T (K ), that of a European put option by Pt,T (K ).

European puts and calls contain a lot of information about the QT

distribution
We will see that we can replicate (almost) every claim from option
portfolios

33 / 41
 Options

Put Call Parity I

100
long call
short put

50
Payoff

-50

-100
0 50 100 150 200
S

34 / 41
 Options

Put Call Parity II

Put Call Parity arises from the mathematical identity

(S − K )+ = (K − S)+ − K + S (14)

An immediate consequence of the above is that we can replicate the

underlying from two European options

Proposition 3.1 (Put Call Parity).

Denote by Ft,T the forward price of ST at time t. If there is no arbitrage

then
Ct,T (K ) = Pt,T (K ) − K pt,T + Ft,T pt,T (15)

Exercise 3.1.
Prove Proposition 3.1 by using the FTAP under the forward measure

35 / 41
 Options

Breeden Litzenberger

Proposition 3.2 (Breeden-Litzenberger).

Suppose the market is arbitrage-free and the QT measure admits a density

qt,T . Then
∂ 2 Ct,T (K )
= p(t, T )qt,T (K ). (16)
∂K 2

Exercise 3.2.
Prove Proposition 3.2 and derive an expression in terms of put prices.
Hint: Start from
Z ∞
QT  +

Ct,T (K ) = pt,T Et (ST − K ) = pt,T (s − K )qt,T (s) ds.
K

36 / 41
 Options

Convexity of Options

Proposition 3.3 (Convexity).

Suppose the market is arbitrage-free. Then European call and put options
are convex in strike.

Exercise 3.3.
Prove Proposition 3.3. Hint: Reuse the derivatives you have taken in
Exercise 3.2.

The convexity property of option prices sets arbitrage bounds for the
relation between puts and call options. For example

Pt,T (K /2) ≤ Pt,T (K )/2

37 / 41
 Options

Convexity II

Pt,T (K1 )
Pt,T (K )
Pt,T (K0 )

The picture shows the admissible region (in blue) that is determined
by Pt,T (0) = 0, Pt,T (K0 ) and Pt,T (K1 ).

Exercise 3.4.
Show that
0
Pt,T (K ) ≥ Pt,T (K0 ) + Pt,T (K0 )(K − K0 ) (17)
38 / 41
 Options

Continuous Option Portfolios

Proposition 3.4 (Carr and Madan (2001)).

Suppose the market is arbitrage-free and Φ : R+ 7→ R is twice

differentiable almost everywhere. Assume that a forward contract F on S
and options (on S) for all strikes K ∈ R+ are traded. Then
(Z
Ft,T
1
QT
Et [Φ(FT ,T ) − Φ(Ft,T )] = Φ00 (K )Pt,T (K )dK
p(t, T ) 0
Z ∞ ) (18)
00
+ Φ (K )Ct,T (K )dK
Ft,T

39 / 41
 Options

Exercises for Carr and Madan (2001)

Exercise 3.5.
Show that the forward price Ft,T from Proposition 3.4 above is a QT
martingale.

Exercise 3.6.
Show that for Φ twice differentiable almost everywhere

Φ(x) = Φ(y ) + Φ0 (y )(x − y )

Z y Z ∞ (19)
00
+ +
Φ (K )(K − x) dK + Φ00 (K )(x − K )+ dK
0 y

Hint: Use put-call parity Eq. (14) and start from first-order Lagrange
remainder

Exercise 3.7.
Prove Proposition 3.4 using Exercises 3.5 and 3.6. 40 / 41
 Options

Exercises
Exercise 3.8 (Simply-Compounded Forward Rates).
Use the FTAP (7) to show that the simply compounded forward rate
F (t, T , S) is under no-arbitrage given by
pt,T
= 1 + (S − T )F (t, T , S)
pt,S

Exercise 3.9 (LIBOR Rates).

The LIBOR rate L is defined as the simple rate L(Ti ) := L(Ti , Ti+1 ) with
Ti < Ti+1 as the solution of
1
pTi ,Ti+1 = , δi ≡ Ti+1 − Ti . (20)
1 + δi L(Ti )

Show that the forward LIBOR rate Ft (Ti ) := F (t, Ti , Ti+1 ), t ≤ Ti is a

martingale under the QTi+1 forward measure.
41 / 41