Applied Mathematical Sciences, Vol. 9, 2015, no.

120, 5979 - 5986

HIKARI Ltd, www.m-hikari.com
http://dx.doi.org/10.12988/ams.2015.57467

The Fundamental Theorem of Asset Pricing

in L1-valued Stochastic Integrals
Christos E. Kountzakis

Department of Mathematics
University of the Aegean
Samos 83200-Greece

Copyright
c 2015 Christos E. Kountzakis. This article is distributed under the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduc-
tion in any medium, provided the original work is properly cited.

Abstract
This paper is devoted to the study of the stochastic integration from
the aspect of order completeness. It studies an L1 -valued stochastic
integral and whether the related markets admit a pricing functional,
which is the analog of the equivalent martingale measure.

Mathematics Subject Classification: 46A40; 60H05; 91B25

Keywords: order-completeness; stochastic integration; KB-space; sublat-

tice

1 Introduction
The presence of heavy-tails in continuous time models, in order to fit the
modelling requirements poses the question of more general versions of the
two FTAPs, mostly relied on the geometry of these spaces. We recall that
seminal references about FTAP are [1, 2]. The need for more general versions
of FTAP- and especially of the First one- is actually the main reason for
the consideration of an L1 -valued stochastic integral (with respect to some
filtered probability space (Ω, F, P), F = (Ft )t∈[0,T ] ) in the present paper. We
recall one of the main results in [3]: (Order 1st Fundamental Theorem of Asset
Pricing) [3, Th.20] Let E be a Banach lattice and M be a sublattice of E. If M
admits a strictly positive projection, then every strictly positive and continuous
5980 Christos E. Kountzakis

functional f : M → R, admits a strictly positive, continuous extension on E.

Also, if E is a Banach lattice and M is a sublattice of E such that every strictly
positive and continuous functional f : M → R, admits a strictly positive,
continuous extension on E, then M admits a strictly positive projection. In this
paper, we apply [3, Th.20] on certain stochastic finance models, relying on the
properties of Riesz Spaces. We introduce an L1 -valued stochastic integral and
we prove that the well-known properties of the Riemann integral hold, thanks
to the order-completeness of the Lp -spaces 1 ≤ p < ∞. We first remind of the
notion of order -completeness. A Riesz space (vector lattice) is order complete
if every non-empty, order bounded from above subset of it has a supremum. If
E is a band in an order complete vector lattice, then B is a projection band,
namely E = B ⊕ B d , (Riesz Thm. [4, Th.8.20]). The projection PB : E → B
is strictly positive (for the notion of strictly positive projection, see [3, Def.7]),
since it is positive and PB (x) = 0, x ∈ E+ , implies that since x = x1 +x2 , where
x1 ∈ B and x2 ∈ B d , x1 = 0. x2 ∧ 0 = 0 and x2 ≥ 0, hence x2 = 0 and finally
x = 0. The same situation is valid for Kantorovich -Banach spaces (or else
KB -spaces), in which E ∗∗ = E ⊕ E d . Such examples of spaces are reflexive
Banach lattices like Lp (Ω, F, P), 1 < p < ∞ and AL-spaces. For the notions
regarding [3], the reader of this paper may see the Section of Notions and
Definitions of that paper. About rest notions rgarding ordered linear spaces
and moreover, Riesz Spaces, the reader may see the Appendix at the end of
this paper.

2 Processes of Stochastic Integrals in L1

We take L1 as a case of a non-reflexive KB space, containing most of the heavy-
tail distributed random variables in L0 (Ω, F, P). According to the Hölder
Inequality, the product of a variable yt in Lp (or a variable which arises from
taking suprema and infima on a stochastic process y) and a variable which is
the increment of another process xt in Lq , (where p1 + 1q = 1, p, q ≥ 1), is a
variable in L1 . We also consider a time horizon [0, T ] and a partition

P = {t0 = 0 < t1 < t2 < ... < tn−1 < tn = T },

of it. We also consider a filtration F = (Ft )t∈[0,T ] -under the usual sense and a
pair (y, x) of F- adapted stochastic processes y, x : [0, T ] × Ω → R. Given a
partition P,
mi = inf{yt |t ∈ [ti , ti+1 ], i = 0, 1, ..., n − 1},
Mi = sup{yt |t ∈ [ti , ti+1 ], i = 0, 1, ..., n − 1} ∈ Lp (Ω, FT , P),
thanks to the order completeness of Lp with respect to the usual partial or-
dering, which makes it a vector lattice. Hence we obtain the following
The fundamental theorem of asset pricing in ... 5981

Definition 1. The lower stochastic integral of y with respect to x under

P, is equal to
n−1
X
L(y, x, P) = mi (xti+1 − xti ) ∈ L1 (Ω, FT , P).
i=0

Definition 2. The upper stochastic integral of y with respect to x under

P, is equal to
n−1
X
U (y, x, P) = Mi (xti+1 − xti ) ∈ L1 (Ω, FT , P).
i=0

Again, thanks to the order completeness of Lp with respect to the usual

partial ordering, which makes it a vector lattice, we may take the following:
Definition 3. The lower stochastic integral of y with respect to x, is
equal to
L(y, x) = supP L(y, x, P) ∈ L1 (Ω, FT , P),
where P is a partition of [0, T ].
Definition 4. The upper stochastic integral of y with respect to x, is
equal to
U (y, x) = inf U (y, x, P) ∈ L1 (Ω, FT , P),
P

where P is a partition of [0, T ].

We notice that
L(y, x) ≤ U (y, x).
The next Criterion of Existence of the Stochastic Integral may be called
Riemann-Riesz Criterion, both due to the similarity of it to the Riemann inte-
grability criterion for real functions and due to the reference to Riesz Spaces.
Theorem 5. (Riemann-Riesz) The process y is integrable with respect
to x, namely
U (y, x) = L(y, x),
if for any sequence (yn )n∈N ⊆ L1 (Ω, FT , P), yn ↓ 0, a partition Pn of [0, T ]
exists, such that
U (y, x, Pn ) − L(y, x, Pn ) ≤ yn .
y
Proof: If U (y, x) = L(y, x), then U (y, x)+ n21 > U (, y, x, Pn1 ) and L(y, x)−
yn2
2
< L(y, x, Pn2 ). For any n0 ≥ n1 , n2 , since yn ↓ 0,
yn0 yn0
U (y, x, Pn ) − L(y, x, Pn ) < (U (y, x) − L(y, x)) + ( + ) = yn0 .
2 2
5982 Christos E. Kountzakis

RT RT RT
Theorem 6. (Linearity) If 0 yt dxt , 0 wt dxt exist in L1 , then 0 (yt + wt )dxt
RT RT
also exists in L1 . Furthermore, if λ ∈ R, then 0 (λ · yt )dxt = λ · 0 yt dxt .

Proof: For the case of a partition Pn of [0, T ] with n + 1 points {0 = t0 <

... < tn−1 < tn = T }, mi (y)+mi (w) ≤ mi (y+w) ≤ Mi (y+w) ≤ Mi (y)+Mi (w),
for any i = 1, 2, ..., n. These order relations imply L(y, x, Pn ) + L(w, x, Pn ) ≤
L(y + w, Pn ) ≤ U (y + w, Pn ) ≤ U (y, x, Pn ) + U (w, x, Pn ) < U (y, x, Pn ) +
U (w, x, Pn ) + gn , where gn ↓ 0. Hence from the application of the Riemann-
Riesz Criterion for y, w, we take
Z T Z T Z T
(yt + wt )dxt = yt dxt + wt dxt .
0 0 0

If λ ≥ 0 then mi (λ·y) = λ·mi (y), and Mi (λ·y) = λ·Mi (y), for any i = 1, 2, ..., n.
These equalities imply λ · L(y, x, Pn ) = L(λ · y, x, Pn ) ≤ U (λ · y, x, Pn ) =
λ · U (y, x, Pn ) < λ · L(y, x, Pn ) + λ · yn , yn ↓ 0, since y is integrable with respect
to x, from the Riemann-Riesz Criterion. If λ < 0, λ · L(y, x, Pn ) + λ · yn <
λ · U (y, x, Pn ) = L(λ · y, x, Pn ) ≤ U (λ · y, x, Pn ) = λ · L(y, x, Pn ). Hence in any
case, from Riemann -Riesz Criterion again,
Z T Z T
λ yt dxt = λ · yt dxt .
0 0

RT RU RU
Theorem 7. (Additivity) If 0 yt dxt , T yt dxt exist in L1 (Ω, FU , P), then 0 yt dxt
RT RU RU
also exists and 0 yt dxt + T yt dxt = 0 yt dxt .

Proof:
For the case of a partition Pn1 of [0, T ] consisted by n1 + 1 points {0 = t1 <
... < tn1 −1 < tn1 = T },

1
U (y, x, Pn1 ) − L(y, x, Pn1 ) < dn1 ,
2
while for a partition of [T, U ] consisted by n2 + 1 points {T = t1 < t2 < ... <
tn2 −1 < tn2 = U },

1
U (y, x, Pn2 ) − L(y, x, Pn2 ) < dn2 .
2
We also consider the partition P = Pn1 ∪ Pn2 of [0, U ]. For the partition P,
we get
U (y, x, Pn1 +n2 −1 ) = U (y, x, Pn1 ) + U (y, x, Pn2 ),
L(y, x, Pn1 +n2 −1 ) = L(y, x, Pn2 ) + L(y, x, Pn2 ),
The fundamental theorem of asset pricing in ... 5983

hence we obtain the conclusion from Riemann-Riesz Criterion, if we suppose

that both y, x on [0, T ], [T, U ] are restrictions on y, x on [0, U ], respectively.
Hence,
U (y, x, Pn1 +n2 −1 ) − L(y, x, Pn1 +n2 −1 ) < dn1 +n2 −1 ,
since dn ↓ 0, dn ∈ L1 (Ω, FU , P) for any n ∈ N.
RT RT
Proposition 8. If 0 yt dxt exist in L1 (Ω, FT , P), then 0 |yt |dxt also exists.
Proof:
For the case of a partition Pn of [0, T ] with n + 1 points {0 = t1 < ... <
tn−1 < tn = T }, if s, u ∈ [ti , ti+1 ], i = 0, ..., n, we have that |ys | − |yu | ≤
|ys −yu | ≤ Mi (y)−mi (y). Hence, Mi (|y|)−mi (|y|) ≤ Mi (y)−mi (y), i = 0, ..., n.
RT
Hence since by the Riemann -Riesz Criterion, the stochastic integral 0 yt dxt
RT
exists, hence the integral 0 |yt |dxt exists in L1 . This is true since we may
select
kPn k = sup{|xti+1 − xti |, i = 0, 1, ..., n} = dn ∈ Lq (Ω, FT , P) ↓ 0,
hence
U (|y|, x, Pn ) − L(|y|, x, Pn ) ≤ U (y, x, Pn ) − L(y, x, Pn ) ≤
Xn
≤ (Mi (y) − mi (y))dn = an ↓ 0,
i=0
1
where an ∈ L (Ω, FT , P).
RT
Corollary 9. The space of the stochastic processes (yt )t∈[0,T ] such that 0 yt dxt
exist in L1 (Ω, FT , P), is a sublattice of the space F-adapted processes.
Proof:R
T RT RT RT RT
Since 0 (yt +wt )dxt = 0 yt dxt + 0 wt dxt , 0 λ·yt dxt = λ· 0 yt dxt , while
RT
0
|yt |dxt exists, too. Hence the stochastic integrals for y ∨ w = 21 (y + w + |y −
w|), y ∧ w = 21 (y + w − |y − w|) exist, with respect to x over [0, T ].
Definition 10. A selection of intermediate points from a partition P
of [0, T ] is a set of points E = {ξ1 , ξ2 , ..., ξn }, such that ti ≤ ξi ≤ ti+1 , i =
0, 1, 2, ..., n. The Riemann -Riesz Sum for P and E is the element of
L1 (Ω, FT , P):
Xn
S(y, x, P, E) = yξi (xti+1 − xti ).
i=0

Theorem 11. (Interpolation) If y is integrable with respect to x, then this is

equivalent to the following: For any sequence dn ↓ 0, where (dn )n∈N ⊆ L1 , and
any selection of intermediate points En of a partition Pn
Z T
|S(y, x, Pn , En ) − yt dxt | < dn .
0
5984 Christos E. Kountzakis

Proof:
If y is integrable with respect to x, from the Riemann -Riesz Criterion
and for the sequence (dn )n∈N ⊂ L1 (Ω, FT , P), dn ↓ 0, a sequence of partitions
(Pn )n∈N of [0, T ] exists, such that U (y, x, Pn ) − L(y, x, Pn ) < dn , which is
preserved if Pn ⊂ Pn1 , namely U (y, x, Pn1 ) − L(y, x, Pn1 ) < dn1 ≤ dn , because
n ≥ n1 in this case. Let En1 some selection of intermediate points of the
partition Pn1 . In this case,
Z T
L(y, x, Pn1 ) ≤ yt dxt ≤ U (y, x, Pn1 ),
0

and
L(y, x, Pn1 ) ≤ S(y, x, Pn1 , En1 ) ≤ U (y, x, Pn1 ).
Hence,
Z T
|S(y, x, Pn1 , E) − yt dxt | ≤ U (y, x, Pn1 ) − L(y, x, Pn1 ) < dn1 .
0

RT
For the inverse, we begin from |S(y, x, Pn , En ) − 0 yt dxt | < dn and we are
going to apply the Riemann -Riesz Criterion. We consider two selections of
intermediate points {ξ1 , ξ2 , ..., ξn }, {η1 , η2 , ..., ηn } with respect to Pn , such that
yξi < mi (y) + dn , Mi (y) − dn < yηi , i = 0, 1, 2, ..., n. Then, we obtain
n
X n
X
U (y, x, Pn ) − L(y, x, Pn ) = Mi (xti+1 − xti ) − mi (xti+1 − xti ) ≤
i=0 i=0

≤ ni=0 yηi )(xti+1 −xti )+dn ni=0 (xti+1 −xti )− ni=0 yηi )(xti+1 −xti )+dn ni=0 (xti+1 −
P P P P
xti ) =

Xn n
X
( yηi )(xti+1 − xti ) − I) + θn + (I − yηi )(xti+1 − xti ) + θn < 4θn ,
i=0 i=0

RT
if I = 0 yt dxt , θn = dn ni=0 (xti+1 − xti ) ↓ 0, while dn ↓ 0 in L1 , since
P
E(|xti − xti−1 |) exist for i = 1, 2, ..., n − 1. Then the proof is complete.

3 The FTAP for spaces of L1- Stochastic Inte-

grals
Since the classic conditional expectation E(.|Ft ) is a strictly positive projection,
[3, Th.20] is applied, in case where F = (Ft )t∈[0,T ] denotes the classic notion
The fundamental theorem of asset pricing in ... 5985

of filtration. For this application, we have to deduce whether the space of the
marketed contingent claims
Z T
1
CT = {cT ∈ L (Ω, FT , P)|cT = y0 + yt dxt },
0

is a sublattice of L1 (Ω, FT , P).

Theorem 12. CT is a sublattice of L1 (Ω, FT , P), hence again a space of the

form L1 (Ω, A, P), where A is a sub-σ-algebra of FT , if the constant random
variable 1 is replicated by some (yt )t∈[0,T ] .

Proof: It is obvious that if the space of integrands y = (yt )t∈[0,T ] is a

sublattice R of the F -adapted stochastic processes y : [0, T ] × Ω → R such
that yt ∈ Lp (Ω, Ft , P), this is equivalent to yt ∨ wt ∈ Rt , yt ∧ wt ∈ Rt , where Rt
is the t-projection of R. For a given partition P = {0 = t1 < t2 < ... < tn−1 <
tn = T } of [0, T ]

mi (y ∨ w) ≤ mi (y) ∨ mi (w) ≤ Mi (y) ∨ Mi (w) ≤ Mi (y ∨ w),

mi (y ∧ w) ≤ mi (y) ∧ mi (w) ≤ Mi (y) ∧ Mi (w) ≤ Mi (y ∧ w)

hold. Since the stochastic integral of y ∨ w, y ∧ w with respect to x = (xt )t∈[0,T ]
exists, by the above order relations, implies that CT +R0 (the range subspace of
the stochastic integrals) is a sublattice, hence CT is a sublattice of L1 (Ω, FT , P).
If 1 is replicated, then by [4, Th.13.11], CT = L1 (Ω, A, P).

Theorem 13. If CT = L1 (Ω, A, P), then [3, Th.20] is applied on such a

market.

Proof: This implies that for the conditional expectation with respect to
A is a strictly positive projection P : L1 (Ω, FT , P) → L1 (Ω, A, P). For
any strictly positive functional f0 of L1+ (Ω, A, P), f0 admits a strictly positive
extension on L1 (Ω, FT , P), through P ∗ (f ) = g. g provides the Radon-Nikodym
derivative of what should be called equivalent martingale measure for CT .

References
[1] F. Delbaen, W. Schachermayer, A General Version of the Fundamental
Theorem of Asset Pricing, Mathematische Annalen, 300 (1994), 463-520.
http://dx.doi.org/10.1007/bf01450498

[2] F. Delbaen, W. Schachermayer, The Fundamental Theorem of Asset Pric-

ing for Unbounded Stochastic Processes, Mathematische Annalen, 312
(1998), 215-250. http://dx.doi.org/10.1007/s002080050220
5986 Christos E. Kountzakis

[3] C.E. Kountzakis, On the Order Form of the Fundamental Theorems

of Asset Pricing, Journal of Mathematical Finance , 4 (2014), 221-233.
http://dx.doi.org/10.4236/jmf.2014.44019

[4] C.D. Aliprantis, K.C. Border, Infinite Dimensional Analysis, A Hitch-

hiker’s Guide, (third edition), Springer, 2005.

Appendix
A partially ordered vector space E is a vector lattice if for any x, y ∈ E, the
supremum and the infimum of {x, y} with respect to the partial ordering de-
fined by P exist in E. In this case sup{x, y} and inf{x, y} are denoted by
x ∨ y, x ∧ y respectively. If so, |x| = sup{x, −x} is the absolute value of
x and if E is also a normed space such that k |x| k = kxk for any x ∈ E,
then E is called normed lattice. If a normed lattice is a Banach space, then
it is called Banach lattice. A Banach lattice E whose norm has the property
kx + yk = kxk + kyk, x, y ∈ E+ is called AL-space. A set S in a vector lattice E
is called solid if |y| ≤ |x| and x ∈ S implies y ∈ S. A solid vector subspace of a
vector lattice is called ideal. An ideal I is a sublattice of E, i.e. a subspace of E
such that x ∨ y ∈ I, x ∧ y ∈ I if x, y ∈ I respectively. A net {xa }a∈A in a vector
lattice E is order convergent to x if there is a net {ya }a∈A in E with ya ↓ 0,
o
such that |xa −x| ≤ ya for each a ∈ A. This convergence is denoted by xa → x.
o
A set D in E is order closed if {xa }a∈A ⊆ D and xa → x, implies x ∈ D. If D
is also an ideal, then D is called band. A Banach lattice has order continuous
norm, if for any net {xa }a∈A ⊆ E with xa ↓ 0, kxa k ↓ 0 holds. A Banach
lattice E which is a band in its second dual (in the sense of norm topology) is
called Kantorovich-Banach space. If S is a subset of a vector lattice E, then
its disjoint complement is the set S d = {x ∈ E : |x| ∧ |y| for any y ∈ S}. If
for a vector lattice E a band B satisfies the property E = B ⊕ B d , then B
is called projection band. For more details on the content of this Section, the
reader may see [4, Ch.8, Ch.9].

Received: July 19, 2015; Published: September 28, 2015