model
Single-period binomial model Two-period binomial model
Introduction to discrete-time market models:
the binomial model
80-646-08 Stochastic calculus
Geneviève Gauthier
HEC Montréal
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
The riskless security I
The market model
The stochastic process n
St(1) :t =0,1o
represents the evolution of the share price of a riskless security (a bond, for example).
Let’s assume that time is measured in periods (a period may be one year, one month, one day, one hour, one second, etc.)
And let’s assume that theperiodic interest rate r is constant during the process observation,
then
8ω2Ω, S1(1)(ω) =S0(1)(ω) (1+r).
We can assume that8ω2Ω,S0(1)(ω) =1 which implies that
8ω2Ω, S1(1)(ω) = (1+r).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
The risky security I
The market model
The second stochastic process n
St(2) :t =0,1o , for its part, represents the evolution of the share price of a risky security, St(2) being the price of a share at time t.
Since the current share price of that security is known with certainty,
8ω2 Ω, S0(2)(ω) =s0 2R+= (0,∞).
We will assume that, in the next period, the share price of the risky security can only take one of two values, say s11
ands12 (s11,s12 2R+,s11 <s12).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Measurable space I
Question. What is the …ltered measurable space that allows us to model such a situation?
Answer. In this example, the random experiment is not clearly de…ned. However, if we consider the
two-dimensional stochastic process
!S
t =n St(1),St(2) :t =0,1o ,
we note that such a process has two di¤erent paths only, i.e.
S0(1),S0(2) > S1(1),S1(2) >
trajectoire #1 (1,s0)> (1+r,s11)>
trajectoire #2 (1,s0)> (1+r,s12)>
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Measurable space II
As a consequence, if Card(Ω)>2, then there would be some identical paths, i.e.
9ω1,ω2 2Ωsuch that 8t 2 f0,1g, !S
t (ω1) =!S
t(ω2) and, in such a case, even our largest source of information,
F1 =σ S0(1),S0(2),S1(1),S1(2) ,
could not distinguish between ω1 andω2 being realized.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Measurable space III
For that reason, we choose Card(Ω) =2 and
ω S0(1)(ω),S0(2)(ω) > S1(1)(ω),S1(2)(ω) >
ω1 (1,s0)> (1+r,s11)>
ω2 (1,s0)> (1+r,s12)>
This is not the only possible form for Ω, but this is a convenient one.
Grouping all undi¤erentiated ω under the same name is common practice. This is also one of the reasons why many people mix up the concepts of sample space and random variable.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Measurable space IV
By choosing such an Ω= fω1,ω2g, we will be able to distinguish which of its elements has occured after observing the process until maturity, i.e. untilt =1.
That’s why,
F1 = F
= fall events inΩg
= f?,fω1g,fω2g,Ωg. Since the process S is constant at time t =0,F0 =f?,Ωg.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
The portfolio I
Let
φ1 = the number of shares of the riskless security being held φ2 = the number of shares of the risky security being held.
De…nition The pair !
φ = (φ1,φ2)2R Ris called aportfolio.
Ifφi is negative, its means a short sale of φi shares of securityi has occurred.
The value of such a portfolio at timet is
Vφ(t,ω) =φ1St(1)(ω) +φ2St(2)(ω).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
The portfolio II
Since Vφ(0,ω) =φ1+φ2s0, we will need to pay out an amount of φ1+φ2s0 in order to acquire the portfolio ! φ ifφ1+φ2s0<0, we will receive an amount of
(φ1+φ2s0)in acquiring the portfolio ! φ.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Arbitrage opportunity
De…nition We say that !
φ is an arbitrage opportunity if (A1) 8ω 2Ω, Vφ(0,ω) =0 (A2) 8ω 2Ω, Vφ(1,ω) 0 (A3) 9ω2 Ω, Vφ(1,ω)>0,
i.e., starting from a zero investment(A1), we are certain not to incur a loss(A2) and we have a positive probability to make a gain (A3).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model I
Condition (A1)implies that, for !
φ = (φ1,φ2)to be an arbitrage opportunity, we must have
φ1 = φ2s0 (1) since 0=Vφ(0,ω) =φ1+φ2s0.
It is to be noted as of now that condition (A3)ensures that (0,0)is not an arbitrage opportunity. As a result, ( φ2s0,φ2)cannot be an arbitrage opportunity ifφ2 =0.
So, let’s assume φ2 6=0.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model II
From condition (A2), it follows that( φ2s0,φ2)is an arbitrage opportunity only if
Vφ(1,ω1) = φ2s0S1(1)(ω1) +φ2S1(2)(ω1)
= φ2s0(1+r) +φ2s11
= φ2(s11 s0(1+r)) 0 and
Vφ(1,ω2) = φ2s0S1(1)(ω2) +φ2S2(1,ω2)
= φ2s0(1+r) +φ2s12
= φ2(s12 s0(1+r)) 0.
Therefore, arbitrage will be possible only if
s12 >s11 s0(1+r)or ifs11 <s12 s0(1+r).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model III
Adding the third condition (A3) implies that, if
s12 >s11 s0(1+r),then, for allφ2>0, the portfolio ( φ2s0,φ2)is an arbitrage opportunity since
Vφ(1,ω1) = φ2(s11 s0(1+r)) 0;
Vφ(1,ω2) = φ2(s12 s0(1+r))>0 and if s11 <s12 s0(1+r)then, for allφ2 <0,the portfolio ( φ2s0,φ2)is an arbitrage opportunity since
Vφ(1,ω1) = φ2(s11 s0(1+r))>0;
Vφ(1,ω2) = φ2(s12 s0(1+r)) 0.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model IV
In other words, if s12 >s11 s0(1+r) then one just has to
sell shortns0 shares of the riskless security (borrow an amount ofns0 dollars)
and buynshares of the risky security for the same amount.
The amount to be paid out isVφ(0) = ns0 1+n s0 =0.
Att =1, we sell the shares of the risky security (we get nS1(2)(ω)dollars)
and pay back our loan, with interest (which amounts to ns0(1+r)dollars).
We therefore get a net amount of
nS1(2)(ω) ns0(1+r) =V( ns0,n)(1,ω)
= ns11 ns0(1+r) =n(s11 s0(1+r)) 0 if ω=ω1
ns12 ns0(1+r) =n(s12 s0(1+r))>0 if ω=ω2.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model V
Ifs11 <s12 s0(1+r)then we can sell shortnshares of the risky security and
and invest the amount ofns0 thus obtained in purchasing ns0 shares of the riskless security:
Vφ(0) =ns0 1 n s0 =0.
At timet =1,an amount ofns0(1+r)dollars is obtained from selling the riskless security and we buy at a cost of nS1(2)(ω)thenshares of the risky security which we had sold short.
The net proceeds from such a transaction are
ns0(1+r) nS1(2)(ω)
= V(ns0, n)(1,ω)
= ns0(1+r) ns11=n(s0(1+r) s11)>0 siω=ω1
ns0(1+r) ns12=n(s0(1+r) s12) 0 siω=ω2.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model VI
By contrast, if s11 <s0(1+r)<s12,there are no portfolios providing arbitrage opportunities since
Vφ(0) =0)φ1 = φ2s0.
Ifφ2=0,then Vφ(1,ω1) =Vφ(1,ω2) =0 and,
ifφ2 6=0,then Vφ(1,ω1) andVφ(1,ω2)cannot be both non-negative since
Vφ(1,ω1) = φ2(s11 s0(1+r))
| {z }
<0
andVφ(1,ω2) = φ2(s12 s0(1+r))
| {z }
>0
.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
What are the arbitrage opportunities in the model VII
In what follows, we will assume that our model contains no arbitrage opportunities, i.e.
s11 <s0(1+r)<s12. (2)
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Contingent claim I
De…nition
Acontingent claim is a contract between two parties, a seller and a buyer, the value of which will depend on the state of the market during the contract validity period. It is like an
insurance contract.
Mathematically speaking, a contingent claim C is any non-negative (Ω,F) random variable.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Contingent claim II
For example, a put option is a contingent claim entered into by the two parties at time t =0, and allowing the buyer to sell, at time t=1,a share of the risky security at a price determined as of now, sayK,which is called the strike price.
Thus, ifS1(2)<K then the buyer will exercise his option and sell his risky security share for an amountK greater than the price he would obtain on the market.
By contrast, ifS1(2) K, the buyer will not exercise his option, since he can obtain on the market a better price than the one set by the contract.
Hence, the value of the put option is max K S1(2),0 . Obviously, the contingent claim seller will not o¤er such a bene…t to the buyer without being compensated. The amount paid out at timet =0 by the buyer to the seller in order to acquire the option is the contingent claim price.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Pricing
Contingent claim
Question. What is the contingent claim price?
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Pricing I
The seller
Can the portfolio be chosen in such a way that its value at t=1 matches the contingent claim value? In the case when there are as many scenarios as securities, it is possible to obtain a single solution under certain conditions.
Ifci =C (ωi)then
8ω 2Ω,Vφ(1,ω) =C(ω)
, φ1(1+r) +φ2s11 =c1 andφ1(1+r) +φ2s12 =c2 , φ1
φ2 =
s12c1 s11c2
(s12 s11)(1+r) c2 c1
s12 s11
!
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Pricing II
The seller
Thus, the value at timet =0 of such a portfolio is Vφ(0,ω)
= s12c1 s11c2
(s12 s11) (1+r)+ c2 c1 s12 s11s0
= c1 1+r
s12 s0(1+r) s12 s11
+ c2 1+r
s0(1+r) s11
s12 s11
. (3)
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure I
So far, we have been working on a …ltered measurable space, and it has not been spoken of probability measure yet.
Let’s set
q = s12 s0(1+r) s12 s11
and note that the no-arbitrage condition is equivalent to the fact thatq is comprised between 0 and 1:
s11 <s0(1+r)<s12 ,0<q <1.
Thus, if Q(ω1) =q and Q(ω2) =1 q then Qis a probability measure built on our measurable space (Ω,F).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure II
It is important to note that Q(ω1)likely does not correspond to the ”real” probability that the event ω1
occurs.
Generally, that probability is not known.
Let’s call P the probability measure that associates to each ω in the sample space the ”real” probability that such an event occurs (P(ω1) =p and P(ω2) =1 p).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure III
As we have established in equation (3), the contingent claim price C is
c1
1+r
s12 s0(1+r) s12 s11 + c2
1+r
s0(1+r) s11
s12 s11
= c1 1+r
s12 s0(1+r) s12 s11 + c2
1+r 1 s12 s0(1+r) s12 s11
= c1
1+rq+ c2
1+r (1 q)
= EQ C
1+r = 1
1+rEQ[C].
So, the price of any contingent claim C is the expectation, under the measure Q, of that contingent claim discounted value 1C+r
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure IV
Wonderful! Rather than having to solve an optimization problem in order to determine a contingent claim price, we just have to calculate an expectation, provided the
appropriate probability measure to be placed on our
…ltered measurable space is known.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure I
Why such a name?
Why such a name?
The return on a security during the time period going fromt 1 tot is de…ned as
RS(t,ω) = St(ω) St 1(ω) St 1(ω)
whereSt(ω)denotes the share price of the security at time t.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure II
Why such a name?
For any probability measure Pb, the expected return on the riskless security over the time period[0,1] is
EPb[RS(1)(1)] =
∑
ω2Ω
S1(1)(ω) S0(1)(ω) S0(1)(ω)
Pb(ω)
=
∑
ω2Ω
(1+r) 1 1 Pb(ω)
=
∑
ω2Ω
rPb(ω)
= r.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure III
Why such a name?
The expected return on the risky security, under the measurePb, is
EPb[RS(2)(1)] =
∑
2 i=1S1(2)(ωi) S0(2)(ωi) S0(2)(ωi)
Pb(ωi)
= s11 s0
s0 Pb(ω1) +s12 s0
s0 Pb(ω2).
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure IV
Why such a name?
So, the ”true” expected return on the risky security is the one calculated under the measure P and it is equal to
EP[RS(2)(1)] = s11 s0
s0 P(ω1) + s12 s0
s0 P(ω2)
= s11 s0 s0
p+s12 s0 s0
(1 p)
= s11 s0 s0
s12 s0
s0 p+ s12 s0 s0
= s12 s0 s0
s12 s11 s0
p
and this latter quantity will be, in general, di¤erent fromr.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Risk-neutral measure V
Why such a name?
By contrast, the expected return on the risky security, calculated under the measure Q, is
EQ[RS(2)(1)]
= s11 s0
s0 Q(ω1) +s12 s0
s0 Q(ω2)
= s11 s0 s0
s12 s0(1+r) s12 s11 +s12 s0
s0
1 s12 s0(1+r) s12 s11
= r
which means that, on the space (Ω,F,Q), there is no bene…t associated with risk, the expected return on the risky security is the same as the one on the riskless security.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Martingale measure I
Why such a name?
Why such a name?
Because, on the …ltered probability space(Ω,F,F,Q), the discounted price processes of the securities S(
i) t
(1+r)t :t=0,1 are martingales.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Martingale measure II
Why such a name?
Proof. In the case of the riskless security,
EQ
"
S1(1) 1+r F0
#
= EQ 1+r 1+r F0
= 1
= S
(1) 0
(1+r)0.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Martingale measure III
Why such a name?
With respect to the risky security,
EQ
"
S1(2) 1+r F0
#
= EQ
"
S1(2) 1+r
#
=
∑2 i=1
S1(2)(ωi) 1+r Q(ωi)
= s11
1+rQ(ω1) + s12 1+rQ(ω2)
= s11 1+r
s12 s0(1+r) s12 s11
+ s12 1+r
s0(1+r) s11 s12 s11
= s0
= S
(2) 0
(1+r)0.
model
Single-period binomial model
Prices Measurable space Portfolio Arbitrage Contingent claim Pricing Risk-neutral measure
Martingale measure Discount factor Two-period binomial model
Discount factor
We have seen that, if the model contains no arbitrage opportunities, then there exists a probability measureQ such that the contingent claim price at time 0 is the expectation of the discounted value of its future cash ‡ows
Price=EQ C 1+r .
But, under the « true» probability measure P,there is no reason to discount the future cash ‡ows with a discount factor based on the riskless interest rate since the contingent claim is a risky asset.
The contingent claim price could be obtained by taking the expectation, under the measure P, of the discounted value of its future cash ‡ow.
The problem in this case is to determine the discount factor.
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
The riskless security
The model
The stochastic process n
St(1) :t =0,1,2o
represents the evolution of the share price of a riskless security.
Let’s assume that the periodic interest rater is constant during the process observation and that
8ω 2Ω, S0(1)(ω) =1, then
8ω 2Ω, St(1)(ω) = (1+r)t.
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
The risky security I
The model
The second stochastic process n
St(2) :t =0,1,2o , for its part, represents the evolution of the share price of a risky security, St(2) being the share price of the security at time t.
Since the current share price of such as security is know with certainty,
8ω2 Ω, S0(2)(ω) =s0 2R+= (0,∞).
We assume that, in the following period, i.e. at t =1, the share price of the risky security can only take one of two values, says11 and s12 (s11,s12 2R+,s11 <s12).
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
The risky security II
The model
We assume that, in the second period, i.e. att =2, there are only four possible values for the share price of the security, says21,s22 (if S1(2) =s11), s23 ands24 (if S1(2) =s12) subject to the constraintss21,s22,s23, s24 2R+,s21 <s22,s23 <s24.
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
The risky security III
The model
ω S0(1)(ω) S0(2)(ω)
!
S1(1)(ω) S1(2)(ω)
!
S2(1)(ω) S2(2)(ω)
!
ω1 (1,s0)> (1+r,s11)> (1+r)2,s21
>
ω2 (1,s0)> (1+r,s11)> (1+r)2,s22 >
ω3 (1,s0)> (1+r,s12)> (1+r)2,s23
>
ω4 (1,s0)> (1+r,s12)> (1+r)2,s24 >
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
Filtered measurable space
The model
As we have previously argued, we can justify choosing as sample space Ω=fω1,ω2,ω3,ω4g.
The sigma-algebra F is the set of all events in Ω and the …ltrationF =fFt :t =0,1,2gconsists of the following sigma-algebras of F:
F0 =f?,Ωg,
F1 =σffω1,ω2g,fω3,ω4ggand F2 =F.
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
Example
The model
Ifr =10%then ω S0(1)(ω)
S0(2)(ω)
!
S1(1)(ω) S1(2)(ω)
!
S2(1)(ω) S2(2)(ω)
!
ω1 (1;2)> (1,1;2)> (1,21;1)>
ω2 (1;2)> (1,1;2)> (1,21;3)>
ω3 (1;2)> (1,1;4)> (1,21;1)>
ω4 (1;2)> (1,1;4)> (1,21;5)>.
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
Notation
The riskless security
De…nition
A stochastic processX =fXt :t =0,1,2, ...g built on a measurable space(Ω,F) is said to bepredictable with respect to the …ltrationfFt :t =0,1,2, ...g ifX0 is a (Ω,F0) random variable and 8t 2 f1,2, ...g,Xt is a (Ω,Ft 1) random variable.
model
Single-period binomial model Two-period binomial model
The model Measurable space Example Predictable process Trading strategy Example Arbitrage Risk-neutral measure Pricing
Trading strategy I
The predictable stochastic processes n
φ(t1) :t=0,1,2o and n
φ(t2):t =0,1,2o
represent the evolution of our portfolio over time:
φ(1)t = number of shares of the riskless security held during the time period(t 1,t],
φ(2)t = number of shares of the risky security held during the time period(t 1,t].
Exceptionally, the portfolio !
φ0 = φ10,φ20 is only held at time t =0. In practice, we will set !
φ0 = ! φ1.