JID: FRL

ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

Finance Research Letters 0 0 0 (2016) 18

Contents lists available at ScienceDirect

Finance Research Letters

journal homepage: www.elsevier.com/locate/frl

How fundamental is the one-period trinomial model to

European option pricing bounds. A new methodological
approachR
Yann Braouezec
IESEG School of Management, LEM CNRS (UMR 8179), Paris campus, France

a r t i c l e i n f o a b s t r a c t

Article history: We offer a new simple approach to price European options in incomplete markets using
Received 25 May 2016 the sole no-arbitrage principle and this only requires to make use of a one-period model;
Revised 13 October 2016
introducing a stochastic process is unnecessary. We show that determining the range of
Accepted 1 November 2016
arbitrage-free prices with a trinomial model only consists in locating two points on a tri-
Available online xxx
angle. As this range of prices may be lower than the classical ones, the parameters of the
JEL classication: model can be implied from the quoted bid and ask prices of liquid European options, used
G12 in turn to estimate the volatility bounds. A simple example is provided using options on
G13 the S & P 500.

Keywords: 2016 Elsevier Inc. All rights reserved.

Incomplete markets
No arbitrage
Option pricing bounds
Bid-ask spread
Volatility bounds

1. Introduction

A European option (call, put...) written on a nancial asset such as a stock or a stock index is a contract that gives its
holder at maturity the right but not the obligation to buy or to sell the stock (or the index) at a pre-determined price called
the strike price. From the seminal paper of Merton (1973), it is common to value options (and more generally derivatives)
using the so-called no-arbitrage principle, which basically states that an investment strategy that can be implemented at no
cost should not generate positive gains only (see e.g., Varian, 1987; Staum, 2007). When one considers a specic probabilistic
model (popular ones are binomial or Black-Scholes), the no-arbitrage alone leads to a unique price because the option payoff
can be perfectly replicated using the basic traded assets. Markets are said to be complete. Perfect replication is however not
possible in general and the application of the no-arbitrage principle leads to a range of prices rather than to a unique price.
Markets are said to be incomplete.
In practice, the causes of market incompleteness are related to trading constraints, for instance the impossibility (or the
diculty) to sell short a security (see Staum, 2007; Sonono and Mashele, 2016). In theory, the causes of market incom-
pleteness are generally not explained; they are simply related to the characteristics of the underlying stochastic process.

R
This paper has been presented in the 28th European Conference on Operational Research, Poznan, July, 2016 and in the seminar of nance of PULV,
September 2016. I want to thank an anonymous reviewer for her/his comments and remarks on a previous version of this paper. The usual disclaimers
apply.
E-mail address: y.braouezec@ieseg.fr

http://dx.doi.org/10.1016/j.frl.2016.11.001
1544-6123/ 2016 Elsevier Inc. All rights reserved.

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

2 Y. Braouezec / Finance Research Letters 000 (2016) 18

For instance, in the standard environment composed of a risky stock and a default risk-free asset, markets are (in general)
incomplete when the stock price is assumed to follow a (discrete-time) trinomial process (see e.g., Van der Hoek and Elliott,
2006, Appendix B) or a (continuous-time) Levy process (see e.g., Cont and Tankov, 2004, chapter 10).
In the literature, there are various pricing/hedging topics (e.g., option pricing bounds, indifference pricing, quadratic
hedging...) that are analyzed under a different set of assumptions but their common point is that a particular stochastic pro-
cess for the risky asset is given. For instance, in their seminal theory paper, Eberlein and Jacod (1997) consider a Levy process
and exhibit the set of arbitrage-free prices for European options with convex payoffs. In their numerical methods oriented
paper, Xiao and Ma (2016) also consider a Levy process but to price a double barrier option. Brown et al. (2015) consider
the determination of the pricing bounds of path-dependent European options using a trinomial process while Fard and Siu
(2013) consider the valuation of European options using a Markovian regime switching binomial process. Sundaram and
Das (2015, chapter 16) presents a lucid and very readable overview of the various types of (discrete time/continuous-time)
processes that includes stochastic volatility models (see also Shi et al., 2016 where a non-ane stochastic volatility model
is considered).
In this paper, we want to dispute the need to introduce a stochastic process (and even a probabilistic model) for stan-
dard European options valuation. Building on the methodology introduced recently in Braouezec and Grunspan (2016) but
restricted to standard European options (i.e., non path-dependent), the determination of the option pricing bounds only re-
quires to specify the set of possible stock (and option) prices at maturity T; intermediate values are irrelevant. We focus
here on the simplest set that contains three possible prices (hence the name of trinomial model) and we show that the
determination of the option pricing bounds simply consists in locating two points on a triangle. The interesting point is
that the resulting pricing bounds may be much lower than the classical ones so that they may coincide with observed bid
and ask prices. For instance, (Eberlein and Jacod, 1997) note that the no arbitrage principle alone, when the stock price is
assumed to follow a Levy process, does not suce to value contingent claims because the range of (arbitrage-free) prices is
much larger than the interval corresponding to the bid-ask spread. We shall actually show that this property is due to the
implicit strong assumption that the set of possible stock prices at time T is the set of real numbers, i.e., the lower bound is
equal to zero and there is no upper bound. As we work with a one-period trinomial model, such a problem does not exist,
and the bounds of the stock price can indeed be calibrated to the observed option prices (bid/ask). A simple example along
this line is provided using options written on the S&P 500. Once this is done, one may use this information to estimate the
implied volatility bounds for a given option contract.
The remainder of this paper is organized as follows. The second section is devoted to the assumptions and discussion.
The third and fourth section are devoted respectively to the option pricing bounds and to a simple example that shows how
to imply the volatility from the quoted option prices.

2. Assumptions

We consider a standard nancial market in which there are two basic traded assets, a default risk-free asset, whose
(deterministic) interest rate is equal to r > 0, and a risky stock. Only two dates will be considered, the current time t = 0
and the T > 0. Let ST be the stock price at time T.

Assumption 1. The stock price at time T > 0 takes three different values

ST ( ) = S0  := {d, m, u} (1)
and each state of the world has a strictly positive probability.

The above assumption thus means that the underlying probability measure P is such that P({} ) := p > 0 for each
{d, m, u}. In what follows, we shall assume that d, m and u are three positive numbers such that d < m < u. Since the
probability measure P is not fully specied (or not fully known), it is usual to talk about Knightian uncertainty. Under such
a situation, it is thus not possible to compute the volatility of the stock. Let g(ST ) be the payoff of the European option with
maturity T and consider the three following points.

D = (dS0 ; g(dS0 )), M = (mS0 ; g(mS0 )), U = (uS0 ; g(uS0 )) (2)

A given point simply represents the stock price and the options payoff at time T > 0 in given state of the world.

Assumption 2. The polygon spanned by the three points D, M and U forms a triangle

The polygon is either a triangle or a segment. From a nancial point of view, it is only when this polygon is a triangle
that markets are incomplete, and this explains the above assumption. It is interesting at this stage to understand what
happens when one adds new states of the worlds. Consider the general case in which  = {1 , . . . , n } (with i < i+1 for
each i = 1, . . . , n 1, n > 3) so that we now have n points; Pi = (i S0 , g(i S0 )), i = 1, . . . , n. The polygon spanned by these
n points is the smallest convex polygon that contains these n points.1 When the European option g under consideration is a

1
It is called more technically the convex hull, see Braouezec and Grunspan (2016). For an arbitrary set of n points of R2 (or Rd with d > 2), the
computational complexity (of the convex hull) increases with n and depends on the algorithm which is used. However, in our nancial setting, the n

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

Y. Braouezec / Finance Research Letters 000 (2016) 18 3

Fig. 1. The triangle  formed by the three points D, M and U.

call or a put option, this convex polygon is either a segment (if g(i S0 )) > 0 for each i = 1, . . . , n so that the option price
is unique), a triangle (if there exists j {2, . . . , n 1} such that j S0 = K ) or a convex quadrilateral (the general case). For
such vanilla options, it is thus never interesting to consider more than four states of the world. As we shall see, considering
three states of the world (i.e., a trinomial model) is enough for our purpose.

3. Option pricing bounds: locating two points on a triangle

For concreteness and to facilitate the presentation, we shall consider the case of a (European) call option with maturity
one year for which the payoff is equal to
C1 ( ) = max{S0 K; 0}, {d, m, u} (3)
Remark 1. Everything could be done for an arbitrary (European) payoff g(ST ) and an arbitrary maturity T > 0. It suces to
set = zT or = ezT with z {d, m, u}.

The three points dened in Eq. (2) become

D = (dS0 ; C1 (d )), M = (mS0 ; C1 (m )), U = (uS0 ; C1 (u )) (4)
Since C1 (u) > C1 (m) > C1 (d), there are more than one way to obtain a triangle. We shall here assume (see Fig. 1) that
C1 (d ) = 0, C1 (m) > 0 and C1 (u) > 0 but nothing is changed if C1 (m ) = 0.

3.1. Existence of option pricing bounds

Let us call the point F the forward point dened as follows

F = ((1 + r )S0 ; (1 + r )C0 ) (5)
where C0 is a given price of our European call option at time t = 0. The following result provides a simple link between
no-arbitrage and the location of the forward point F in the triangle .

Proposition 1. No-arbitrage is equivalent to F , that is, the forward point F must lie in the interior of the triangle  :=
DMU.

Proof. Let us consider the if part and assume that F is located outside  or on the boundary of . We shall now show
that this generates an arbitrage opportunity. From Fig. 2, it is easy to see that if S0 (1 + r ) < dS0 or if S0 (1 + r ) > uS0 , F lies
outside the triangle  and this generates an elementary arbitrage strategy independently of the option price. As a result,
it must be the case that d < (1 + r ) < u. We now assume that d < (1 + r ) < u. From Fig. 2, one can also clearly see that
/ , i.e., if F is located on the boundary or outside , one can nd a line (indeed innitely many) that separates2 the
if F
forward point F from the triangle . Let V1 ( ) = aS0 + b the equation of such a line called V1 . Saying that this line strictly
separates the forward point from the triangle means that, for all R+ , V1 () > C1 (). Since the model only considers

points are not arbitrary and depend in a simple way of the payoff g of the option. For a put or a call option, the determination of the convex hull is a
trivial problem so that the computational complexity essentially remains invariant with respect to n.
2
The separation is strict if F is located outside .

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

4 Y. Braouezec / Finance Research Letters 000 (2016) 18

Fig. 2. The triangle  separated by a line.

three states of the world at time T = 1, it thus follows that V1 () > C1 () for each {d, m, u}. We shall analyze the
case depicted in Fig. 2 in which the option is overpriced but the case in which the option is underpriced is similar. We thus
assume that V1 () > C1 () for each {d, m, u}.
By construction, the forward point F is located on this line V1 (see Fig. 2) which means that (1 + r )C0 = a(1 + r )S0 + b .
Since b is arbitrary at this stage, let us assume that b = (1 + r )b so that the equation of our line becomes (1 + r )C0 =
a(1 + r )S0 + (1 + r )b and reduces to aS0 + b C0 = 0. From a nancial point of view, this equation provides the composition
of the arbitrage portfolio. At time t = 0, by being long a units of stock, short on one call option, and long or short b =
C0 aS0 units of cash (the position depends on the sign of b), we obtain a portfolio whose value 0 is equal to zero at
time t = 0 and is equal to 1 ( ) = aS1 ( ) + b(1 + r ) C1 ( ) at time T = 1 in the state of the world . Recalling that
V1 ( ) = aS1 ( ) + b(1 + r ), since V1 () > C1 () for each {d, m, u} by assumption, it thus follows that 1 () > 0 for
each {d, m, u} and this is an arbitrage opportunity.3 This concludes the proof of the if part.
To prove the converse, assume that there is an arbitrage opportunity. Since it is assumed that d < (1 + r ) < u, the arbi-
trage only comes from the mis-pricing of the option. There are two possibilities regarding the arbitrage opportunity.4 For
each {d, m, u}, the arbitrage portfolio can be such that i) 0 = 0 and 1 () 0 with at least one for which the
inequality is strict, or such that ii) 0 < 0 and 1 ( ) = 0. Regarding arbitrage ii), 1 ( ) = aS1 ( ) + b(1 + r ) C1 ( ) = 0
is equivalent to solve a system of three linear equations with two unknowns, a and b. However, since  is a triangle, it
is not dicult to show that this system does not admit a solution and this means that arbitrage ii) is not feasible. The
only possibility thus is arbitrage i). But then, this means that there exists a line that separates the forward point from the
triangle, that is, F must be located on the boundary or outside the triangle  and this concludes the proof of the only if
part.
To sum up, we have shown that the existence of an arbitrage opportunity is equivalent to F located on the boundary,
or outside the triangle . No-arbitrage thus is equivalent to F located strictly inside , and this concludes the proof of
Proposition 1 . 

Corollary 1. There exists two critical prices C0 and C 0 such that if C0 (C 0 ; C 0 ), then, C0 is an arbitrage-free price.

These two critical prices located on the boundary of the triangle  are usually called the option pricing bounds and the
interval R := (C 0 ; C 0 ) denes the range of prices. Fig. 3 allows one to visualize, up to the term (1 + r ), the range of prices
given by the vertical segment [F, F].

Remark 2. Proposition 1 assumes that the option under consideration is written on a single stock. It would be interesting, as
a further research, to consider the same methodology but applied to options written on more than one assets. The simplest
case would the spread call, whose payoff is max{ST1 ST2 K; 0}, where STi is the price of stock i at maturity T.

3.2. Computing the option pricing bounds

We shall now explicitly show how to explicitly compute these two (call) pricing bounds. Let
F = ((1 + r )S0 ; (1 + r )C 0 ) (6)

3
If the point F lies on the boundary of the triangle, say on the segment DU, there is a unique line V1 and it is easy to see that this still generates an
arbitrage opportunity since V1 ( ) = C1 ( ) if {d, m} and V1 (m) > C1 (m).
4
Using the terminology of Carr and Wu (2016), this is a type 1 arbitrage since it only involves the underlying asset and cash.

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

Y. Braouezec / Finance Research Letters 000 (2016) 18 5

Fig. 3. The lower and the upper bound of the call option.

F = ((1 + r )S0 ; (1 + r )C 0 ) (7)

be two points located respectively on the segment DU and MU, see Fig. 3. As a result, F (F) can be expressed as a convex
combination of the points D and U (M and U). Let X = (x1 , x2 ) denotes a generic vector. In what follows, the transpose of
this generic vector X will be denoted by X , i.e., X thus is a column vector. Moreover, to simplify the notation, the option
payoff in state will now be denoted by C instead of C1 ().

Fact 1. The critical prices C 0 and C0 can be computed as follows.


To calculate C 0 , it suces to nd the weight qu (0, 1 ) that solves F = qu U + (1 qu )D , where is such that qu = 1 qd .
To calculate C0 , it suces to nd the weight qu (0, 1) that solves F = qu U + (1 qu )M , where is such that qu = 1 qm .

The next proposition uses the above fact to derive the option pricing bounds.

Proposition 2. The pricing bounds of a call option are given below

quCu + (1 qu )Cd 1+rd
C0 = where qu = (8)
1+r ud

quCu + (1 qu )Cm 1+rm

C0 = where qu = (9)
1+r um
   
 ( 1 + r )S0 uS0
Proof. From the denition of U, D and F, note that F = qu U + (1 qu )D is equivalent to = qu + (1
(1 + r )C 0 Cu
 
dS0
qu ) and yields the following linear 2 2 system.
Cd

(1 + r )S0 = qu uS0 + (1 qu )dS0 (10)

(1 + r )C 0 = quCu + (1 qu )Cd (11)

1+rd
From Eq. (11), by dividing by 1 + r, we immediately obtain C 0 in Eq. (8). Solving now Eq. (10) yields qu = ud
. Eq. (9) is
obtained in the same way.5 

Remark 3. The methodology offered should be of pedagogical interest. Due to its simplicity,6 it allows an instructor to start
introductory courses on option pricing directly in incomplete markets. This is indeed simpler than the classical approach
since it does not require to introduce a stochastic dynamics nor does it requires to introduce the notions of replication or of
risk neutral valuation. The binomial model could be introduced as the special case in which the points D, M and U form a

5
Note that since Cm > 0 by assumption, it is not dicult to show that solving F = qu U + (1 qu )M yields C 0 = S0 K
1+r
.
6
Compare with Van der Hoek and Elliott (2006, Appendix B). The authors also consider the option pricing bounds of a call. However, as they build on
the duality theorem of linear programming, the determination of the no-arbitrage bounds of a call is (much) more involved.

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

6 Y. Braouezec / Finance Research Letters 000 (2016) 18

segment so that the option price is unique. It is now time to explain that it is so because the option payoff can be perfectly
replicated using the basic traded assets. Once again, as long as the three points from a triangle, there is no need to introduce
the notion of replication to derive the range of prices.

In the literature, it is usual to derive the option pricing bounds by taking the inmum and the supremum over the set
of pricing measures, see e.g., Staum (2007). Within our approach, we have been able to determine the range of prices in
Proposition 2 without introducing such a pricing measure. However, inspecting Eqs. (8) and (9) suggests that each pricing
bound could be seen as a discounted expectation of the option payoff under some probability measure. Let
Q = ( qd ; 0 ; qu ) (12)

Q = ( 0 ; qm ; qu ) (13)
be two probability measures, that is, qd + qu = 1 and qm + qu = 1. As usual in nance, we shall say that a probability measure
is a pricing measure if it yields an arbitrage-free price. We shall denote E0 (. . . ) to mean E(. . . |S0 ).

Fact 2. The two prices given in Eqs. (8) and (9) can be written as a discounted expectation under the probability measure
Q and Q respectively.
qdCd + quCu C1
 
C0 = = EQ (14)
1+r 0
1+r

qmCm + quCu  C 
Q 1
C0 = = E0 (15)
1+r 1+r
It is however important to note that Q and Q are not pricing measures since the two prices C 0 and C0 are consistent
with arbitrage. To obtain a pricing measure, it suces to consider a convex combination of these two probability measures.

Proposition 3. Let (0, 1) and consider the probability measure Q dened as

Q := Q + (1 )Q = ( qd ; (1 )qm ; qu + (1 )qu ) (16)
The option price under the probability measure Q is equal to
 C 
EQ = C 0 + (1 )C 0
1
(17)
0
1+r
and is arbitrage-free.
 
qd Cd +qm Cm +qu Cu
Proof. Let Q = (qd ; qm ; qu ) be a probability measure and note that by denition EQ
C1
0 1+r = 1+r . Assume now
that Q = Q + (1 )Q. From the rhs of Eq. (16), qd = qd , qm = (1 )qm and qu = ( qu + (1 )qu ) so that
 C  qdCd + (1 )qmCm + [ qu + (1 )qu ]Cu
EQ
1
= (18)
0
1+r 1+r
Using now C 0 and C0 as given in Eqs. (14) and (15), it is easy to see that C 0 + (1 )C 0 is equal to the rhs of Eq. (18).
From Corollary 1, since (0, 1), the option price C 0 + (1 )C 0 thus is arbitrage-free. 

Since the option price obtained in Eq. (17) is arbitrage-free, the probability measure Q thus is a pricing measure. In
Braouezec and Grunspan (2016), it is shown that Q is the barycentric coordinates7 of the forward point F. However, even for
a trinomial model, computing these barycentric coordinates still require to compute a 3 3 determinant. Proposition 3 ac-
tually provides an easier and more direct way to compute Q which does not require to use the barycentric coordinates of
F. Proposition 3 also shows that there is actually a one-to-one correspondence between the set of arbitrage-free prices and
the set of pricing measures. As a result, to be a pricing measure, since (0, 1), each component of the pricing measure Q
must be strictly positive, i.e., q > 0 for all {d, m, u}. Since it was assumed that p > 0 for all {d, m, u}, to use the
classical terminology in quantitative nance, the pricing measure Q is said to be equivalent to the statistical (or historical)
probability measure P. Note that within our approach, it is not possible to provide a link between P and Q as P is not
specied.

Fact 3. For each (0, 1), the probability measure Q as dened in Eq. (16) is a pricing measure such that
 S 
EQ
1
= S0 (19)
0
1+r
S1
Fact 3 is an elementary consequence of our methodology and says that 1+r can be thought of as a martingale under Q.

7
See Vince (2013, chapter 11) for an introduction to barycentric coordinates.

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

Y. Braouezec / Finance Research Letters 000 (2016) 18 7

3.3. On classical option pricing bounds

From the seminal paper of Merton (1973), many textbooks (e.g., Jarrow and Turnbull, 1996, chapter 3) derive the pricing
bounds of options in a model-free way. For a (European) call option with strike K and maturity one year, the option pricing
bounds are given by
 K

C 0 = max 0, S0 and C 0 = S0 (20)
1+r
K
In the particular case in which S0 < 1+ r , the range of prices thus is the interval (0, S0 ). As observed by many authors
(see e.g., Eberlein and Jacod, 1997), this interval is clearly too large to be realistic. It is fundamental to understand that
while this interval (0, S0 ) is derived in a (probabilistic) model-free way, it is still implicitly assumed that the set of possible
prices at maturity is R+ . And this is a very strong assumption. To see this within our trinomial model, consider the case
K K
in which S0 < 1+ r and assume m = S so that Cm = 0. It is easy to show that the resulting triangle implies that C 0 = 0 and
 1+rd
uS0 K
 0

that C 0 = 1+r ud
ud
. It is only when d = 0 and u that the pricing bounds (of our trinomial model) coincide
with the classical ones. However, when d > 0 and/or u < , the pricing bounds may be much lower.

4. Implied volatility bounds : an example

For standard European options, the ask price and the bid price are quoted on the market. If we assume that the interval
formed by these two observed prices is the set of arbitrage-free prices, one may imply the parameters of the model from
these prices.8 Once this is done, the volatility bounds of the stock price can be estimated. We provide a simple example
K ask (K, 1Y) := Cask be the quoted bid and ask price of the call option with
assuming that S0 < 1+ r . Let Cbid (K, 1Y) := Cbid and C
maturity 1Y (i.e., one year) for a given strike K and let
d:= d (K ), m
 := m(K ), 
u := u(K ) (21)
be the parameters of the trinomial model implied from the option prices. There are several ways to imply these parameters
and as in the literature on implied trees, depending on the method chosen, the implied parameters may still be consistent
with arbitrage.9 In what follows, we assume that d = 0 so that only u and m have to be calibrated.
K
Proposition 4. Consider a call option for which the strike K is such that S0 < 1+r and let Cbid = C 0 and C ask = C 0 . Assume that
d = 0 and m < 1 + r. From Proposition 2, the parameters implied from the observed bid and ask prices, Cbid and Cask , are equal
to
K

u= (22)
S0 C ask

(1 + r )(C ask Cbid )

=
m (23)
(1 + r )Cbid [ S0 C
ask
C ask K
]
K
Proof. Since d = 0, Cd = 0 and note that since S0 < and m < 1 + r, Cm = 0. Using Proposition 2 with d = 0 and Cd = 0,
1+r
solving Eq. (8) reduces to solve one equation with one unknown, i.e., u. It is easy to show that 
u= K
ask
and this proves
S0 C
(1+r )[uS0 KuCbid ]
Eq. (22). Solving Eq. (9) in m yields m = uS0 K(1+r )Cbid
:= h(u ). Inserting now 
u as dened in Eq. (22) gives h(
u) = m
 and
this proves Eq. (23). 

For the calibration exercise, we consider the case of few liquid European call options written on the stock index S&P 500.
In Eq. (22), only the rate r is not directly observable. As a proxy, we shall use the Libor rate with maturity one year in the
relevant currency (dollar). Given a proxy for r, once the parameters u and m are calibrated (d = 0), the implied volatility
under the probability measures Q and Q now can be computed.

Q ( K ) = u (u 1 r ) + (1 q
q  2
u )(0 1 r )
2 (24)


Q ( K ) = q
u
(
u 1 r )2 + (1 q
u
)(m
 1 r )2 (25)

As long as the volatility lies in the interval (Q (K ), Q (K )), it is consistent with no-arbitrage. Let t = 0 be September,
28, 2016. At time t = 0, the value of the S & P 500 is equal to S0 = 2171.37 and the Libor rate in dollar with maturity one
year10 is equal to 1.5513%. The following table reports the bid and ask prices for few call options written on the S & P 500
(Table 1).

8
Sonono and Mashele (2016) consider the opposite problem for interest rate options. Their aim is to estimate the bid-ask prices from the mid-price.
9
Implied trees can generate negative (risk-neutral) probabilities, which means that the model is consistent with arbitrage, see Derman et al. (1996) and
Moriggia et al. (2009).
10
See the web site of Ice Benchmark Administration, https://www.theice.com/marketdata/reports/170.

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001
JID: FRL
ARTICLE IN PRESS [m3Gsc;November 3, 2016;0:20]

8 Y. Braouezec / Finance Research Letters 000 (2016) 18

Table 1
Call option written on the S & P 500
index (maturity 1Y).

Sep 28, 2016, S0 = 2171.37

Strike Open int Bid Ask

2300 623 54.4 56.1

2425 20 0 0 18.8 22.9
2450 758 11.2 17.8

Table 2
Calibration results, d = 0.

Sep 28, 2016

K  (K )
m u (K )
 Q ( K ) Q ( K )
2300 0.32615 1.08873 22.5% 27.0%
2425 0.6956 1.12871 19% 33.9%
2450 0.85902 1.13765 13.825% 35.217%

Note importantly that we have chosen liquid contracts only, those for which the open interest (i.e., the number of out-
standing contracts) is high enough. For many contracts, the open interest is very low (between 0 and 10) so that it makes
no sense to calibrate the parameters of the model for such highly illiquid contracts.
The above table exhibits the following interesting properties (Table 2).
The implied volatility Q (K ) decreases with the strike price K and is consistent with the well-known stylized fact on
equity index, called the volatility skew (see e.g., Carr and Wu, 2016; Fengler, 2012, Hull, 2014, chapter 20. Since Q (K ))
also increases with K, the volatility bounds increases with K and this suggests that valuing call options deeply out-of-
the-money is more uncertain.
The implied volatility tends to be higher than the classical one obtained with the Black-Scholes model. As the implied
volatility is a convention, this difference is not an issue and just reects the fact that a different model is used.

5. Conclusion

This paper offers a new and simple methodology to determine the range of prices of standard European call (or put)
options using a one-period trinomial model. Building on the exibility of this trinomial model, it is shown that the pricing
bounds may be lower than the classical ones. As a result, the parameters of the trinomial model can be implied from the
bid and ask prices of European options and an implied volatility can be easily computed. As this approach to compute
the implied volatility is much simpler than the classical one, it could provide an interesting alternative. However, a (more)
complete empirical analysis is necessary to assess the quality and the (possible) interest of this alternative.

References

Braouezec, Y., Grunspan, C., 2016. A new elementary geometric approach to option pricing bounds in discrete time models. Eur. J. Oper. Res. 249 (1),
270280.
Brown, D.J., Ibragimov, R., Walden, J., 2015. Bounds for path-dependent options. Ann. Finance 11 (34), 433451.
Carr, P., Wu, L., 2016. Analyzing volatility risk and risk premium in option contracts: a new theory. J. Financ. Econ. 120 (1), 120.
Cont, R., Tankov, P., 2004. Financial Modelling with Jump Processes. Chapman & Hall.
Derman, E., Kani, I., Chriss, N., 1996. Implied trinomial tress of the volatility smile. J. Derivatives 3 (4), 722.
Eberlein, E., Jacod, J., 1997. On the range of options prices. Finance Stochastics 1 (2), 131140.
Fard, F.A., Siu, T.K., 2013. Pricing and managing risks of european-style options in a markovian regime-switching binomial model. Ann. Finance 9 (3),
421438.
Fengler, M.R., 2012. Option data and modeling black-scholes-merton implied volatility. In: Handbook of Computational Finance. Springer, pp. 117142.
Van der Hoek, J., Elliott, R.J., 2006. Binomial Models in Finance. Springer Science & Business Media.
Hull, J.C., 2014. Options, Futures, and Other Derivatives. Pearson Education.
Jarrow, R.A., Turnbull, S.M., 1996. Derivative Securities. South-Western Pub.
Merton, R.C., 1973. Theory of rational option pricing. Bell J. Econ. Manage. Sci. 4, 141183.
Moriggia, V., Muzzioli, S., Torricelli, C., 2009. On the no-arbitrage condition in option implied trees. Eur. J. Oper. Res. 193 (1), 212221.
Shi, G., Liu, X., Tang, P., 2016. Pricing options under the non-ane stochastic volatility models: an extension of the high-order compact numerical scheme.
Finance Res. Lett. 16, 220229.
Sonono, M.E., Mashele, H.P., 2016. Estimation of bid-ask prices for options on libor based instruments. Finance Res. Lett 19, 3341.
Staum, J., 2007. Incomplete markets. In: Handbooks in Operations Research and Management Science, vol. 15, pp. 511563.
Sundaram, R.K., Das, S.R., 2015. Derivatives: Principles and Practice. McGraw-Hill Irwin.
Varian, H.R., 1987. The arbitrage principle in nancial economics. J. Econ. Perspect. 1 (2), 5572.
Vince, J., 2013. Mathematics for Computer Graphics. Springer Science & Business Media.
Xiao, S., Ma, S., 2016. Pricing discrete double barrier options under lvy processes: an extension of the method by mariyan and aldo. Finance Res. Lett 19,
6774.

Please cite this article as: Y. Braouezec, How fundamental is the one-period trinomial model to European option pricing
bounds. A new methodological approach, Finance Research Letters (2016), http://dx.doi.org/10.1016/j.frl.2016.11.001