Ouka Cha
Ouka Cha
PROBABILITY-BASED APPROACH
LEILA OUKACHA
Contents
1. Preliminaries 2
1.1. Probability Spaces 2
1.2. Random Variables 2
1.3. Distributions 3
2. Martingales 5
2.1. Conditional Expectation 5
2.2. Martingale 6
3. Application to arbitrage and no-arbitrage pricing in finance 9
3.1. Arbitrage 9
3.2. Arbitrage-free single-period market 10
3.3. The Martingale Representation Theorem and Hedging in a multi-
period market 13
Acknowledgments 15
4. Bibliography 15
References 15
1. Preliminaries
1.1. Probability Spaces.
Definition 1.1.1. A probability space is a triple (Ω, F, P) where Ω is a set of
“outcomes”, F is a set of “events”, and P : F → [0, 1] is a function that assigns
probability to events.
Definition 1.1.2. (Ω, F) is a measurable space. A measure is a nonnega-
tive countably additive set function such that µ : F→ R satisfies the following
properties:
(i) µ(A) ≥ µ(∅) = 0 for all A ∈ F. S P
(ii) If Ai ∈ F is a countable sequence of disjoint sets, then µ ( i Ai ) = i µ(Ai ).
(iii) If µ(Ω) = 1, then µ is defined as a probability measure usually denoted
by P.
Example 1.1.3. Discrete probability spaces. Let Ω be a finite or infinite
countable set and F be the set of all subsets of Ω. Also, assume that A ⊂ F.
P P
Then, P (A) = ω∈A p(ω), where p(ω) ≥ 0 and ω∈Ω p(ω) = 1.
Proposition 1.1.4. A sigma-algebra F on a set Ω is a collection of subsets
of Ω satisfying the following properties:
(i) Ω ∈ F.
(ii) If A ∈ F, then Ω \ A ∈ F. S
∞
(iii) If A1 , A2 , A3 , . . . ∈ F, then i=1 Ai ∈ F.
Definition 1.1.5. The Borel sigma-algebra B(R) on R is the smallest sigma-
algebra containing all open subsets of R.
1.2. Random Variables.
Definition 1.2.1. A function X : Ω → S is said to be a measurable map from
(Ω, F) to (S, S) if X−1 (B) ≡ {ω : X(ω) ∈ B} ∈ F, ∀B ∈ S.
Definition 1.2.2. If (S, S) = (R, B(R)), then X is called a random variable.
Example 1.2.3. The indicator function of a set A ∈ F is a random variable:
(
1 if ω ∈ A
1A (ω) =
0 if ω ∈ /A
Theorem 1.2.4. If {ω : X(ω) ∈ A} ∈ F for all A ∈ A and A generates S such
that S is the smallest σ-field that contains A, then X is measurable.
Proof. Letting {X ∈ B} be equivalent to {ω : X(ω) ∈ B}, we have
∈ Bi } and {X ∈ B c } = {X ∈ B}c .
S S
{X ∈ i Bi } = i {X
So, by Theorem 1.2.5, f (X1 , . . . , Xn ) is a measurable map from (Rn , Rn ) → (R, R),
which proves that f (X1 , . . . , Xn ) is a random variable. □
Theorem 1.2.7. If X1 , . . . , Xn are random variables, then so are inf n Xn , supn Xn ,
lim supn Xn , and lim inf n Xn .
S
Proof.SFirst, observe that {inf Xn < a} = n {Xn < a} ∈ F. Similarly, {sup Xn >
a} = n {Xn > a} ∈ F.
Similarly,
lim sup Xn = inf sup Xm
n→∞ n m≥n
Proof.
Observe that
• If ω ≤ F (x), then X(ω) ≤ x, since x ∈/ {y : F (y) < ω}.
• On the other hand, if ω > F (x), then, since F is right continuous, there
exists ϵ > 0 such that F (x + ϵ) < ω and X(ω) ≥ x + ϵ > x.
2. Martingales
2.1. Conditional Expectation.
Definition 2.1.1. Let G be a sub-σ-algebra of F, and X ∈ L1 be a random
variable. Then the random variable ξ, denoted by E[X | G], is the conditional
expectation of X with respect to G if
(i) ξ ∈ L1 ,
(ii) ξ is G-measurable,
(iii) E[ξ · 1A ] = E[X · 1A ], for all A ∈ G.
Proposition 2.1.2. Properties of Conditional Expectation
Let X ∈ L2 (Ω, F, P ) and let G be a σ-algebra contained in F. Then Conditional
Expectation satisfies the following properties:
Since B ∈ G, we have by Definition 2.1.1 and the Tower Property (4) in Proposi-
tion 2.1.2,
E(X · 1B ) = E(E(X | G) · 1B ) and E(Xn · 1B ) = E(E(Xn | G) · 1B ).
But, by Theorem 2.1.3,
E(X · 1B ) = lim E(Xn · 1B ) and E(V · 1B ) = lim E(E(Xn | G) · 1B ),
n→∞ n→∞
Thus, P (B) = 0. □
2.2. Martingale.
Definition 2.2.1. Let {Fn } denote the information in A1 , A2 , . . . , A∞ such that
information refers to the collection of data or events available up to a given time
t. Then a filtration of a set Ω (finite or infinite) is defined to be a collection Ft ,
indexed by a time parameter t (discrete or continuous), such that
(1) Each Ft is a σ-algebra of subsets (events) of Ω;
(2) If s < t, then Fs ⊆ Ft .
Example 2.2.2. Suppose X1 , X2 , . . . are independent, identically distributed ran-
dom variables with mean µ.
(1) E(Mn | Fm ) = Mm .
Equivalently,
(2) E(Mn − Mm | Fm ) = 0.
Example 2.2.4. “Martingale betting strategy”
Then think of the random variables Xi as the results of a game where one flips
a coin where:
To beat the game, we will keep doubling our bet until we eventually win. At
this point, we stop.
Let Wn denote the winnings (or losses) up through n flips of the coin using this
strategy and let W0 = 0. Whenever we win, we stop playing, so our winnings stop
changing such that
P{Wn+1 = 1 | Wn = 1} = 1.
Now, suppose that the first n flips of the coin have turned up tails. After each flip,
we have doubled our bet, so we have lost 1 + 2 + 4 + . . . + 2n−1 = 2n − 1 dollars
and Wn = −(2n − 1). At this time, we double our bet again and wager 2n on the
next flip. This gives
1
P Wn+1 = − 2n+1 − 1 | Wn = − (2n − 1) = .
2
Therefore,
E(Wn+1 | Fn ) = Wn .
Hence, by Definition 2.2.3, Wn is a martingale with respect to Fn .
8 LEILA OUKACHA
Definition 2.2.5. Let (Ω, F, P ) be a probability space and (Ft )0≤t≤T or (Ft )0≤t<∞
be a filtration by sub-σ-algebras of F. An adapted sequence Xt of integrable random
variables is defined to be a
• Martingale if E(Xt+1 | Ft ) = Xt , ∀t.
• Submartingale if E(Xt+1 | Ft ) ≥ Xt , ∀t.
• Supermartingale if E(Xt+1 | Ft ) ≤ Xt , ∀t.
Example 2.2.6. Let {Xn }n≥0 be a martingale relative to the filtration {Fn }n≥0 ,
and ϕ : R → R be a convex function such that E[ϕ(Xn )] < ∞ for each n ≥ 0. Then
the sequence {Zn }n≥0 defined by
Zn = ϕ(Xn )
is a submartingale relative to the filtration {Fn }n≥0 by Definition 2.2.5.
This is a consequence of Jensen’s inequality and the martingale property of {Xn }n≥0 :
E[Zn+1 | Fn ] = E[ϕ(Xn+1 ) | Fn ]
≥ ϕ(E[Xn+1 | Fn ])
= ϕ(Xn ) = Zn .
Proof. We will show is that for every 0 < a < b < ∞, the probability that the mar-
tingale fluctuates infinitely often (ie: the probability that the martingale diverges)
between a and b is 0.
1) Think of Mn as giving the cumulative results of some fair game and Mn+1 − Mn
as being the result of the game at time n + 1.
3) Stop betting until Mn < a again and return to betting 1. Continue this process,
changing the bet to 0 when Mn > b and changing back to 1 when Mn < a.
Note that
Wn ≥ (b − a)Un − |Mn − a|,
where Un denotes the number of times that the martingale goes between a and b
(ie: the number of upcrossings) and Mn − a gives an estimate for the amount lost
in the last interval. Since Wn is a martingale, we have
(1) E(W0 ) = E(Wn ) ≥ (b − a)E(Un ) − E(|Mn − a|).
By the triangle inequality,
(2) E(|Mn − a|) ≤ E(|Mn |) + a ≤ C + a.
Since A1 is riskless by Definition 3.2.2, the share price S11 of A1 in any scenario ωi
where r is the riskless rate of return is defined as follows:
S11 (ωi ) = er ∀i = 1, 2, . . . , N.
Definition 3.2.4. Portfolios. A portfolio is a vector
θ = (θ1 , θ2 , . . . , θK ) ∈ RK
of K real numbers such that θj is the number of shares of asset Aj that are owned.
If θj < 0, then the portfolio is said to be short |θj | shares of asset Aj .
ARBITRAGE AND MARKET DYNAMICS: A PROBABILITY-BASED APPROACH 11
Proposition 3.2.5.
Suppose, for the sake of contradiction, that there exists a point y ∈ F with a
nonpositive first coordinate. Then the line segment L has endpoints v and y such
that L ⊂ F . Because this line segment must cross the hyperplane consisting of
points with first coordinate 0, we may suppose that y has the form
y = (0, y2 , y3 , . . . , yk ),
12 LEILA OUKACHA
Let L consist of all points of the form y(ϵ) := ((1 − ϵ)a, ϵy2 , ϵy3 , . . . , ϵyk ), where
0 ≤ ϵ ≤ 1.
Now, the closest point to the origin on L must be v. But, for all sufficiently
small ϵ > 0, the point y(ϵ) is actually closer to the origin than v, which is a con-
tradiction. □
Theorem 3.2.9. Fundamental Theorem of Arbitrage Pricing. There exists
an equilibrium measure if and only if arbitrages do not exist.
Proof.
In the case for an arbitrage portfolio, if V1 (θ; ωi ) > 0 for every market sce-
nario ωi , then (1) implies that V0 (θ) > 0, and so θ cannot be an arbitrage
by Definition 3.2.6. Thus, arbitrages do not exist.
We must show that if the market does not admit arbitrages, then it has an
equilibrium measure π, that is, a probability distribution π(ωi ) on the set
Ω of market scenarios ωi such that (1) in Definition 3.2.7 holds.
First, for j = 1, asset 1 is the riskless asset, so its share price at time
t = 0 is S01 = 1 and its share price at time t = 1, under any scenario ωi , is
er . So for any probability distribution π on the set of market scenarios,
N
X N
X
1 = S01 = e−r π(ωi )er = e−r π(ωi )S11 (ωi ).
i=1 i=1
We can choose a real number −θ1 that lies between these two values. Then
adding θ1 to both sides of the previous inequality shows that for every
market scenario ωi ,
K
X K
X
e−r θj S1j (ωi ) > 0 > θj S0j .
j=1 j=1
For each t ≤ T ,
Ft = {all events determined in the first t trading periods}.
The finite sequence (Ft )0≤t≤T is a filtration (ie: Definition 2.2.1) of the space Ω of
market scenarios.
Remark 3.3.3. The share prices of assets in a multiperiod market depend on
market scenarios, but evolve in such a way that their values at any time t, being
observable at time t, do not depend on the unobservable post-t futures of the
scenarios.
14 LEILA OUKACHA
(i) For each scenario, the first (respectively, second) entry indicates whether the
share price of the asset stock increased or decreased in the first (respectively, sec-
ond) trading period.
(iii) The only events that are determined before the first trading period are the
trivial events ∅ and Ω.
F0 = {∅, Ω},
F1 = {∅, Ω, F+ , F− },
Proposition 3.3.6. The market M has scenario space Ω = {+, −}T , the set of all
sequences of pluses and minuses of length T . Moreover, there is a riskless asset
bond with rate of return r, and a risky asset stock whose price process evolves
according to the rule
ξt (ω1 ω2 . . . ωT ) = ωt · 1 for t = 1, 2, . . . , T,
Ft = σ(ξ1 , ξ2 , . . . , ξt ).
ARBITRAGE AND MARKET DYNAMICS: A PROBABILITY-BASED APPROACH 15
Proof. Fix a scenario ω = ω1 ω2 . . . ωT ∈ Ω and let Gt (ω) be the set of all scenarios
whose first t entries are ω1 ω2 . . . ωt . Since the sequence (Yt )0≤t≤T is a martingale,
it follows that for every t < T and every ω ∈ Ω,
(1) E[Yt+1 1Gt (ω) ] = E[Yt 1Gt (ω) ].
Moreover, since the sequence (Yt )0≤t≤T is adapted to the natural filtration, the
value Yt+1 (ω) − Yt (ω) depends on the scenario ω = ω1 ω2 . . . ωT only through its
first t + 1 entries. Thus, equation (1) implies that, for each t and each ω,
(2) pYt+1 (ω1 ω2 . . . ωt +) + qYt+1 (ω1 ω2 . . . ωt −) = Yt (ω1 ω2 . . . ωt ).
Equation (2) also holds if Yt+1 and Yt are replaced respectively by St+1 and St
because (St )0≤t≤T is also a martingale. Solving both equations for −q/p leads to
the relation:
Yt+1 (ω1 ω2 . . . ωt +) − Yt (ω1 ω2 . . . ωt ) q St+1 (ω1 ω2 . . . ωt +) − St (ω1 ω2 . . . ωt )
=− = ,
Yt+1 (ω1 ω2 . . . ωt −) − Yt (ω1 ω2 . . . ωt ) p St+1 (ω1 ω2 . . . ωt −) − St (ω1 ω2 . . . ωt )
which in turn implies that
Yt+1 (ω1 ω2 . . . ωt +) − Yt (ω1 ω2 . . . ωt ) Yt+1 (ω1 ω2 . . . ωt −) − Yt (ω1 ω2 . . . ωt )
=
(3) St+1 (ω1 ω2 . . . ωt +) − St (ω1 ω2 . . . ωt ) St+1 (ω1 ω2 . . . ωt −) − St (ω1 ω2 . . . ωt )
:= βt (ω1 ω2 . . . ωt ).
Remark that the common value of the fractions on the two sides of equation (3)
depends only on ω1 ω2 . . . ωt , so the definition of βt is valid, which proves Theo-
rem 3.3.8. □
Acknowledgments
It is my pleasure to thank my mentor, Jakob Wellington, for his invaluable
guidance throughout the writing of this paper. I would also like to extend my
thanks to Professor Rudenko for conducting apprentice lectures, especially in group
theory. Special thanks to Professor Lawler for delivering lectures on probability and
analysis. Finally, my sincere appreciation to Professor May for organizing the Math
REU, which allowed me to gain valuable insights into the experience of conducting
mathematical research for the first time.
4. Bibliography
References
[1] Lawler, Gregory F. Introduction to Stochastic Processes. Chapman and Hall/CRC, 1995.
[2] Durrett, Rick. Probability: Theory and Examples. 5th ed. Cambridge: Cambridge University
Press, 2019.
[3] Lalley, Steven P. Conditional Expectation. University of Chicago, 2015. http://galton.
uchicago.edu/~lalley/Courses/383/ConditionalExpectation.pdf
16 LEILA OUKACHA
[4] Zitkovic, Gordan.“Lecture Notes Page.” University of Texas at Austin. Accessed June 23,
2024. https://web.ma.utexas.edu/users/gordanz/lecture_notes_page.html.
[5] Kwok, Yue Kuen. An Introduction to the Mathematics of Financial Deriva-
tives. Elsevier, 2008. https://www.sciencedirect.com/book/9780123846822/
an-introduction-to-the-mathematics-of-financial-derivatives.
[6] Harvard Business School. Online Business Insights.
[7] Lalley, Steven P. “Lectures 3 and 4: Martingales.” University of Chicago. http://galton.
uchicago.edu/~lalley/Courses/390/Lecture3.pdf
[8] Lalley, Steven P. “The Fundamental Theorem of Arbitrage Pricing.” University of Chicago.
http://galton.uchicago.edu/~lalley/Courses/390/Lecture1.pdf