Economic Modelling 37 (2014) 296305

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.elsevier.com/locate/ecmod

Pricing foreign equity options with regime-switching

Kun Fan a,d, Yang Shen b,d, Tak Kuen Siu c,d,, Rongming Wang a
a
School of Finance and Statistics, East China Normal University, Shanghai 200241, China
b
School of Risk and Actuarial Studies and CEPAR, Australian School of Business, University of New South Wales, Sydney, NSW 2052, Australia
c
Cass Business School, City University London, 106 Bunhill Row, London EC1Y 8TZ, United Kingdom
d
Department of Applied Finance and Actuarial Studies, Faculty of Business and Economics, Macquarie University, Sydney, NSW 2109, Australia

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

Article history: In this paper, we investigate the valuation of two types of foreign equity options under a Markovian regime-
Accepted 1 November 2013 switching mean-reversion lognormal model, where some key model parameters in the dynamics of the foreign
equity price and the foreign exchange rate are modulated by a continuous-time, nite-state Markov chain. A fast
JEL classication: Fourier transform (FFT) approach is applied to provide an efcient way to evaluate the option prices. Numerical
F31
analysis and empirical studies are provided to illustrate the practical implementation of the proposed pricing
G13
model.
Keywords: 2013 Elsevier B.V. All rights reserved.
Foreign equity option
Regime-switching
Mean-reversion
Fast Fourier transform

1. Introduction
Due to recent technological advance and trade liberalization, the
growth of globalization has been accelerated and the economic growth
has been boosted unprecedentedly. In the global nancial markets, for-
eign exchange risk arising from uctuations in foreign exchange rate
has received much attention, especially since the currency crises in
emerging markets. To hedge and manage foreign exchange risk, both
academic researchers and industry practitioners have proposed a varie-
ty of currency options. Partly attributed to globalization, many rms and
households are massively involved in investment activities of foreign
assets. There are two key sources of risk arising in investment on foreign
assets, namely foreign exchange (FX) risk and asset's price risk. Effective
management of these two sources of risk is the key to successes in for-
eign assets investments. Foreign equity options provide a possible way
to manage or hedge both the FX risk and the equity price risk. According
to the denition in Kwok and Wong (2000), the currency-translated
foreign equity options are contingent claims whose payoffs are deter-
mined by nancial prices or indices denominated in one currency but
the actual payouts are settled in another currency. As its name implies,
the underlying asset of a foreign equity option is a foreign equity. There
Corresponding author at: Cass Business School, City University London, 106 Bunhill
Row, London EC1Y 8TZ, United Kingdom; Department of Applied Finance and Actuarial
Studies, Faculty of Business and Economics, Macquarie University, Sydney, NSW 2109,
Australia. Tel.: +61 2 98508573; fax: +61 2 98509481.
E-mail address: ktksiu2005@gmail.com (T.K. Siu).
are two main tempting features of foreign equity options. Firstly, foreign
equity options provide investors with a variety of exible ways to deal
with the multidimensional risks, mainly the foreign equity price uctu-
ation risk and the foreign exchange risk. There exist a variety of types of
foreign equity options with different payoff functions. Seen from this as-
pect, foreign equity options could provide investors with more invest-
ment and risk management choices. Exchange-trade is the second
advantage of foreign equity options, which means this kind of nancial
product enjoys a higher degree of liquidity. Furthermore, the regula-
tions of clearinghouse help investors reduce or avoid some risks, such
as counterparty risk.
Since the pricing model of foreign equity options needs to depict the
joint dynamics of the exchange rate and foreign equity prices, there are
some literature about the valuation of foreign equity options under
different models. Early works usually consider the valuation of foreign
equity options in the BlackScholes framework. Kwok and Wong
(2000) investigated the valuation of foreign equity options with path
dependent features. Examples of pricing foreign equity options beyond
the traditional BS framework include a multi-dimensional Lvy process
to depict the dynamics of both the exchange rate and the foreign equity
prices in Huang and Hung (2005). Xu et al. (2011a) considered the
valuation of foreign equity option under a stochastic volatility model
with double jumps. To incorporate the impacts of skewness and kurtosis
on foreign equity option prices, the GramCharlier series expansion
approach was adopted by Xu et al. (2011b).
It is known that certain vital features of nancial time series cannot
be depicted by the classical Black-Scholes models. Among the models
extending the classical BlackScholes model, the ability to incorporate

0264-9993/$ see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.econmod.2013.11.009
 K. Fan et al. / Economic Modelling 37 (2014) 296305
structural changes in economic conditions makes regime-switching
models one of the most practically useful models in nancial economet-
rics. These changes, which may be attributed to changes in economic
fundamentals or business cycles, represent an additional source of risk
to which an additional amount of risk premium may be required to
compensate. Furthermore, the risk brought by these changes can be
hardly diversied since it is more likely to be regarded as a systematic
risk. Since regime-switching models provide a natural and convenient
choice to model the structural changes in economic conditions, espe-
cially due to nancial crises, this class of models will enjoy more and
more popularity. The seminal work of Hamilton (1989) popularized ap-
plications of regime-switching models in nancial econometrics. Typi-
cally, the so-called modulated by a Markov chain means the model
dynamics or parameters will change when the underlying Markov
chain changes from one state to another. The states of the Markov
chain represent the states of an economy. Since the last decade or so,
there has been an interest on studying option valuation problems in
regime-switching models (see Bufngton and Elliott (2002), Elliott
et al. (2005), Siu (2008), Yuen and Yang (2010), Shen et al. (2013),
Shen and Siu (2013), etc.). Considering the increasingly changing for-
eign exchange market, there is a considerable interest to investigate
the valuation of currency options under regime-switching models, in-
cluding Bollen et al. (2000), Siu et al. (2008), Bo et al. (2010). Empirical
studies in Bollen et al. (2000) veried that trading strategies under
regime-switching models can gain higher prot and be more attractive
to investors. That also indicates the potential practical value of regime-
switching models.
However, relatively little attention has been given to pricing foreign
equity options in the context of regime-switching models. In this paper,
we investigate the valuation of foreign equity options under a Markovian
regime-switching mean-reversion lognormal model, which extends
the mean-reversion lognormal model for foreign exchange rate.
More specically, the model parameters, including the risk-free do-
mestic interest rate, the volatility of the foreign equity, the mean-
reversion level and the volatility of the foreign exchange rate, as
well as the instantaneous correlation coefcient between the foreign
equity and the exchange rate, are modulated by a continuous-time,
nite-state, observable Markov chain. To apply the fast Fourier trans-
form (FFT) approach to discretize the integral pricing formula, we
need to rst calculate the characteristic function of the logarithmic
underlying equity price. For the valuation of the foreign equity op-
tion with strike price in the foreign currency (FEOF 1), we rst
apply a measure change technique and use a version of the Bayes'
rule to derive the conditional characteristic function of the logarith-
mic equity price under the new measure. For the valuation of the for-
eign equity option with strike price in the domestic currency (FEOD),
we calculate the characteristic function of the summation of the log-
arithmic foreign equity price and the logarithmic foreign exchange
rate under the risk-neutral probability measure. Then, we derive
the Fourier transform of the foreign equity option price in these
two cases. To illustrate the pricing of foreign equity options, we pro-
vide a numerical analysis using the FFT method. Finally, an empirical
application is provided, revealing that the regime-switching model
outperforms the model with a single regime in terms of lower tting
errors and prediction errors. The main contributions of this paper are
as follows. (1) We investigate the valuation of foreign equity options
under a regime-switching mean-reversion lognormal model. The
main feature of our model is that it combines the advantages of
both regime-switching models and mean-reversion lognormal
models. The mean reversion feature of foreign exchange rates has
been well-documented. (Jorion and Sweeney (1996); Sweeney
(2006)) Wong and Lau (2008); Wong and Lo (2009); Wong and
1
Following the notation in Xu et al. (2011b), let FEOF and FEOD represent the foreign eq-
uity option with strike price in the foreign currency and the foreign equity option with
strike price in the domestic currency, respectively.
297
Zhao (2010); Leung et al. (2013)). (2) By applying a measure change
technique, the Fourier transform of the FEOF option price can be cal-
culated more easily. Then, we adopt the FFT approach in Carr and
Madan (1999) and Liu et al. (2006) to derive a pricing formulae for
the foreign equity options.
The rest of the paper is organized as follows. The next section pre-
sents the model dynamics. In Section 3, we derive the pricing formulae
of FEOF and FEOD under the Markovian, regime-switching, mean-
reversion lognormal model, respectively. Section 4 presents numerical
examples. An empirical application of our model is provided in
Section 5. The nal section concludes the paper.
2. The model dynamics
In this section, we consider a continuous-time economy with a nite
time horizon T : 0; T , where T b . Let (, F P) be a complete prob-
ability space. In the literature about foreign exchange rate modeling, it is
customary to assume that P is a risk-neutral probability measure (See
Wong and Lau (2008)). To describe the evolution of the state of an econ-
omy over time, we consider a continuous-time, N-state, observable
Markov Chain X : fXt jtT g . The N different states of the chain
may represent N observable different states of an economy or different
stages of a business cycle. Without loss of generality, using the conven-
tion in Elliott et al. (1994), we assume the chain X has a canonical state
space E : = {e1,e2,,eN} N, where the j-th component of ei is the
Kronecker delta ij, for each i,j = 1,2,,N. Let Q: = [qij]i,j = 1,2,,N de-
note the generator or rate matrix of the chain X under P , where qij is
the transition intensity of the chain X from state ei to state ej. Then the
following semimartingale dynamics for the chain X were obtained in
Elliott et al. (1994):
Z t
Xt X0 QXsds Mt ; tT :
0
Here fMt jtT g is an N-valued, ( FX, P)-martingale, where FX is
the right-continuous, P -complete, natural ltration generated by the
chain X.
We now specify the Markovian regime-switching models for
the dynamics of the foreign equity and the foreign exchange
rate. Let S : fSt jtT g and Z : fZ t jtT g denote the price process
of the foreign equity and the logarithmic foreign exchange rate process
respectively. Let y be the transpose of a vector or a matrix y, , be the
scalar product in N, and diag(y) be the diagonal matrix with diagonal
elements being given by the components of the vector y. For each tT ,
let r(t) and (t) be the domestic, instantaneous continuously compounded,
interest rate and the volatility of the equity at time t, respectively. We
assume that r(t) and (t) are determined by the value X(t) of the
chain at time t as:
r t : hr; Xt i;
t : h; Xt i;
where r: = (r1,r2,,rN) N with ri N 0 and : = (1,2,,N) N
with i N 0 for each i = 1,2,,N.
Let f t jtT g and ft jtT g be the mean-reversion level and vol-
atility of the process Z. Again we suppose that
t : h; Xt i;
t : h; Xt i;
where : = (1,2,,N) N and : = (1,2,,N) N
with i N 0, for each i = 1,2,,N. The parameter , controlling the
speed of mean reversion for the logarithmic foreign exchange rate pro-
cess, is assumed to be a positive constant.
298 K. Fan et al. / Economic Modelling 37 (2014) 296305

Then, under the risk-neutral probability P, the dynamics of S and Z payoff functions of the two kinds of foreign equity options are
are given by given by

dSt rt St dt t St dW 1 t ; 1 FEO F F T ST K F ;

and and

dZ t t Z t dt t dW 2 t ; 2 FEOD F T ST K D ;

where fW 1 t jtT g and fW 2 t jtT g are two standard Brownian mo- where KF and KD are the strike prices in the foreign currency and in
tions with respect to their respective right-continuous, P-complete, nat- domestic currency, respectively. Note that the payoff is represented
ural ltrations under P . Furthermore, we suppose that the two in the domestic currency for both the FEOF option and the FEOD
Brownian motions W1, W2 are correlated and that the instantaneous option.
correlation coefcient at time t is given by:
Z t
3.1. Valuation of an FEOF option
hW 1 ; W 2 it sds ;
0 Considering the particular payoff function of an FEOF option, the fol-
lowing pricing formula is standard:
where (t) = ,X(t) and : = (1,2,,N) N with 1 b j b 1
for j = 1,,N; with a slight abuse of the notation, fhW 1 ; W 2 it jtT g de- 2 Z T 3
notes the predictable quadratic covariation process between W1 and W2. r t dt
6 7
For each tT , write Y(t): = ln(S(t)) and F(t): = eZ(t) for the loga- C F 0; T; K F E4e 0 F T ST K F 5 ; 3
rithmic foreign equity price and the foreign exchange rate at time t, re-
spectively. Note that in Shen and Siu (2013), a Markovian regime-
switching Hull-White model was used for modeling stochastic interest where E is an expectation under the risk-neutral measure P. Let
rate, so there is a positive probability that the interest rate goes negative. kF = ln(KF) be the logarithmic strike price. The modied FEOF option
Here the logarithmic foreign exchange rate is modelled by a Markovian price is dened by
regime-switching, mean-reverting process, so that foreign exchange
rate stays positive. Applying It's differentiation rule, the risk-neutral c F 0; T; k F e
aFkF
C F 0; T; K F ;
dynamics of Y and F are given by
  where aF is a predetermined positive constant such that cF(0, T, kF) is
1 2
dY t rt t dt t dW 1 t ; square integrable in kF over the entire real line. As in Carr and Madan
2
(1999), the Fourier transform of cF(0, T, kF) is as follows:
and Z
iuk F
  F 0; T; u e c F 0; T; k F dk F : 4
dF t 1 2
t t lnF t dt t dW 2 t :
F t 2
      The following proposition gives an integral representation for the
Let FS : F S t jtT , FZ : F Z t jtT and FX : F X t jtT
price of the FEOF option.
be the right-continuous, P -complete, natural ltrations generated by
processes S, Z and X, respectively. Furthermore, we dene the enlarged Proposition 3.1. For each j = 1, 2,,N, let
ltration G : fGt jtT g by the minimal -eld containing F S(t),
F Z(t) and F X(t). That is, 1 2Tt 2 1
Tt 2 2
g j t; u : r j e j e j uia F 1 j
S Z X 2 2 
Gt : F t F t F t ; tT : 1 2 Tt
iuia F 1 r j j e j j j :
2
For each tT , Gt represents publicly available market information
up to time t.
Write
Remark 2.1. The economic motivation of using the regime-switching
N
model may be illustrated by a utility maximization problem of a rep- gt; u : g 1 t; u; g 2 t; u; ; g N t; u ;
resentative agent's intertemporal consumption. If the agent receives
a stream of Markov-modulated dividends from holding a single secu- where is the complex space and N is the N-fold product of .
rity, it can be shown that the maximization of an expected utility of Then under the Markovian regime-switching mean-reversion log-
the agent leads to a regime-switching model dynamics for the secu- normal model, the price of the FEOF option is given by the following in-
rity. Interested readers may refer to Di Graziano and Rogers (2009) tegral formula:
for details. Due to various economic and nancial factors, the stream Z
of dividends may change with the current economic conditions or ea F k F
iuk F
C F 0; T; K F e F 0; T; udu ;
market modes. The regime-switching model provides a natural way 0

to describe these changes, so it may provide a better matching of

where
the agent's expectation towards market trends than its non-regime
switching counterpart. n o
exp eT Z 0 iu a F 1Y 0
3. Pricing foreign equity options F 0; T; u
Z T a F a F u i2a F 
1u
2 2

Typically, there are two types of foreign equity options, which are X0exp diag gt; udt Q T ; 1 :
0
classied by the strike prices of the underlying foreign equities. The
 K. Fan et al. / Economic Modelling 37 (2014) 296305 299

(t s)
Before proving Proposition 3.1, several useful results are given. First- By Girsanov's theorem, WQ t
1 (t): = W1(t) 0 (s)e (s)ds
t (t s)
ly, we introduce a probability measure Q equivalent to P on GT by the and WQ2 (t): = W2(t) 0 e (s)ds are two standard Brownian
D E
following RadonNikodm derivative: motions under Q. It is also obvious that W Q Q
1 ; W 2 t hW 1 ; W 2 it

dQ

Z T t
e 0 udu.

GT :  Z T X  :
dP
E e jF T To apply the fast Fourier transform approach (Carr and Madan
(1999)), we need to calculate the characteristic function of the logarith-
The idea of introducing a probability measure Q here may not be unlike mic terminal spot price of the foreign equity. Here, due to the measure
that of introducing a forward measure in a Markovian regime- change, the conditional characteristic function of the logarithmic termi-
switching, mean-reverting process, (see, for example, Shen and Siu nal foreign equity price under the probability measure Q should be de-
(2013)). rived rst.
Then, by a version of the Bayes' rule, Eq. (3) becomes
Lemma 3.2. The conditional characteristic function of Y(T) given F X(T)
2 2 Z T 33 under Q is calculated as:
6 6 r t dt
X 77 h i
C F 0; T; K F E4E4e 0 F T ST K F
F T 55 Q Q iuY T X
Y T j F X T u : E e j F T
2 Z 3  Z T  Z T
T 1 2 1 2 2
r t dt h i h i exp iuY 0 iu r t t dt u t dt
6 X Q X 7 Z T 0 2  2 0
E4e 0 E F T jF T E ST K F jF T 5; Tt
iu e t t t dt
0

where EQ represents an expectation under the measure Q.

where EQ denotes an expectation under the probability measure Q.
Lemma 3.1. The RadonNikodm derivative is given by:
Proof. It is easy to nd that under the probability measure Q,

 Z Z 
dQ

1 T
2Tt 2
T
Tt Z T   Z

GT exp e t dt e t dW 2 t : 1 2 T
Tt
dP
2 0 0 ST S0exp r t t dt e t t t dt
Z T 0  2 0
Q
t dW 1 t :
Then, 0
Z t
Q ts
W 1 t : W 1 t se sds Consequently, conditional on F X(T), Y(T) is normally distributed
0
with the following mean and variance:
and Z T 
h i 1 2
Q X
Z t
E Y T jF T Y 0 r t t dt
Q ts Z T 0 2
W 2 t : W 2 t e sds
Tt
0 e t t t dt ;
0
are two standard Brownian motions under Q. The instantaneous corre- and
Q Q
lation coefcient between W 1 and W 2 at time t is still (t).
h i Z T
Q X 2
Proof. A direct calculation to Eq. (2) gives Var Y T jF T t dt :
0
Z T
Z T
T Tt Tt
Z T e Z 0 e t dt e t dW 2 t : Then the conditional characteristic function of Y(T) is easy to com-
0 0
pute.
For notational simplicity, write:
It is easy to see that given F X(T), the conditional distribution of Z(T)
is a normal distribution with the following mean and variance: Z T
RT : r t dt ;
h i Z T 0
X T Tt
E Z T jF T e Z 0 e t dt ; Z Z
i Z T
T 1 T
h 0
LT : e
Tt
t dt
2 Tt
e
2
t dt :
X 2 Tt 2
Var Z T jF T e t dt ; 0 2 0
0
The result presented in the following lemma resembles to those in
so that Lemma 4.1 and Lemma 4.2 in Shen and Siu (2013).
h i Lemma 3.3. Let F Q y be the conditional distribution function of
Z T X
E e jF T Y T j F X T
 Z T
Z T
 Y(T) given F X(T) under Q. Then, the Fourier transform of the price of the
T Tt 1 2Tt 2
exp e Z 0 e t dt e t dt : FEOF option is given by:
0 2 0

Consequently, F 0; T; u 
RT LT eT z0 Q
E e Y T j F X T uia F 1
dQ eZ T
 Z T X  a2F a F u2 i2a F 1u
dP E e jF T
n o Z T 
 Z T
Z T
 exp e
T
Z 0 iu a F 1Y 0 X0 exp diagg t; udt Q T ; 1
1 2 Tt 2 Tt 0
exp e t dt e t dW 2 t : :
2 0 0 a2F a F u2 i2a F 1u
300 K. Fan et al. / Economic Modelling 37 (2014) 296305

where QY T j F X T
u denotes the conditional characteristic function of Y(T) Proof of Proposition 3.1. The proof is standard. Applying the inverse
given F X(T) under the probability measure Q. Fourier transform to Eq. (4), the following equation can be derived:

Proof. The proof resembles to the proofs of Lemma 4.1 and Lemma 4.2
a F k F Z
in Shen and Siu (2013), which are standard. We present the proof here a F k F e iuk F
C F 0; T; K F e c F 0; T; k F e F 0; T; udu :
for the sake of completeness. Let kF = ln(KF), 0
Z The result can be obtained from Lemma 3.3 immediately.
iuk F
F 0; T; u e c0; T; k F dk F
Z

e
a F k F iuk F
C 0; T; K F dk F
e 3.2. Valuation of an FEOD option
Z

  
a F k F iuk F R Y T k
e e E e T F T e e F dk F The pricing formula for an FEOD option is given by

Z


   
a F k F iuk F RT Y T kF X
E e e E e F T e e jF T dk F
Z

2 Z 3
h i     T
a F k F iuk F RT X Q Y T k X r t dt
E e e e E F T jF T E e e F jF T dk F 6 7
C D 0; T; K D E4e 0 F T ST K D 5 ; 6
Z
   
T
a F k F iuk F RT LT e Z 0 Q Y T k X
E e e e E e e F jF T dk F

"Z Z #
T  
a F k F iuk F RT LT e Z 0 y k Q
E e e e e e F F Y T j F X T dydk F where E denotes an expectation under the risk-neutral mea-
kF
Z T  Z y   sure P. Let k D = ln(K D ). The modied FEOD option price is de-
R L e Z 0 ya F iuk F 1a F iuk F Q
E e T T e e dk F F Y T j F X T dy ned by
" Z

! #
RT LT e
T
Z 0 e1a F iuy e1a F iuy Q aD kD
E e F Y T j F X T dy cD 0; T; kD e C D 0; T; kD ;
a F iu 1 a F iu
h T
i
RT LT e Z 0 Q
E e Y T j F X T uia F 1
where a D is a predetermined positive constant such that cD (0, T,
a2F a F u2 i2a F 1u
n o Z T  k D ) is square integrable in k D over the entire real line. As in Carr
exp eT Z 0 iu a F 1Y 0 E exp hgt; u; Xt idt and Madan (1999), the Fourier transform of c D (0, T, k D ) is as
0
: follows:
a2F a F u2 i2a F 1u
Z
ivkD
Dene D 0; T; v e cD 0; T; kD dkD : 7

Z t

t : Xt exp hgs; u; Xsids ; tT : Dene a process fGt jtT g with G(t): = ln(F(t)S(t)) for each tT .
0
By direct calculation,
Applying It's differentiation rule to (t),
Z T
 
T 1 2 Tt
GT e Z 0 lnS0 r t t e t dt
Z  Z Z 0 2
t T T
Tt
d t hgt; u; Xt i t dt exp hgs; u; Xsids dXt t dW 1 t e t dW 2 t :
0 Z  0 0
t
diaggt; u Q t dt exp hgs; u; Xsids dMt :
0
5 Then, the conditional characteristic function of G(T) given F X(T)
under the risk-neutral probability measure P is given by

Taking expectation on both sides of Eq. (5) under P gives: h i

X
GT j F X T v E expfivGT gjF T :
dE t  diag gt; u Q E t dt :
The following proposition gives an integral representation for the
Solving gives price of the FEOD option. This result resembles to that of Proposition 3.1.

Proposition 3.2. For each j = 1, 2,,N, let

Z T 
E XT exp hgt; u; Xt idt  
1 2 Tt
Z0 T  h j t; v : r j iviaD 1 r j j e j
2
X0exp diaggt; udt Q T :  
1 2 2 2 Tt 2 Tt
0 viaD 1 j e j 2e j j j :
2

Consequently, Write

N
ht; v : h1 t; v; h2 t; v; ; hN t; v :
Z T

E exp hgt; u; Xt idt Then under the Markovian regime-switching mean-reversion log-
0
normal model, the price of the FEOD option is given by
hE T ; 1i
 Z  Z
eaD kD
T
X0exp diag gt; udt Q T ; 1 : C D 0; T; K D e
ivkD
D 0; T; vdv ;
0 0
 K. Fan et al. / Economic Modelling 37 (2014) 296305 301

where Table 2
Option prices calculated via the FFT.

FEOF FEOD
D 0; T; v n  o
Z 0  Z 
T
exp iv aD Y 0 e T k State 1 State 2 State 1 State 2
X0exp diaght; vdt Q T ; 1 :
a2D aD v2 i2aD 1v 0 0.3 0.5614 0.4554 0.9849 0.8450
0.2 0.4672 0.3611 0.9022 0.7713
0.1 0.3772 0.2698 0.8164 0.6932
Proof. The proof here resembles to that in Lemma 3.3. Let F GT j F X T g 0 0.2947 0.1882 0.7282 0.6112
denote the conditional distribution function of G(T) given F X(T) under 0.1 0.2225 0.1225 0.6388 0.5269
P. Then, the Fourier transform of the FEOD option price is given by: 0.2 0.1623 0.0753 0.5500 0.4426
0.3 0.1146 0.0450 0.4637 0.3610
Z
aD kD ivkD
D 0; T; v e e C D 0; T; K D dkD
The derivation of the second equation is similar with the proof in
Z
  
aD kD ivkD R GT k Lemma 3.3.
e e E e T e e D dkD

Z
   
aD kD ivkD R GT k X 4. Numerical examples
E e e E e T e e D j F T dkD

"Z Z #
  In this section, we perform a numerical analysis for pricing the
aD kD ivkD RT g k
E e e e e e D F GT j F X T dgdkD FEOF option and the FEOD option under our regime-switching
kD
Z
Z g    mean-reversion lognormal model. To simplify our computation,
RT aD kD ivkD g k
E e e e e e D dkD F GT j F X T dg we consider a two-state Markov chain X. For each tT ,

Z Z g    X(t) = (1,0) and X(t) = (0,1) are State 1 and State 2,
R a ivkD g a 1ivkD
E e T e D e D dkD F GT j F X T dg respectively.
h
i The rate matrix of the Markov chain X under P is assumed to be
R
E e T GT j F X T viaD 1
:  
a2D aD v2 i2aD 1v q q
Q ;
q q
Since
Z T  where q takes values in [0,1]. Intuitively, with a larger q, the economy
T 1 2 Tt
GT e Z 0 lnS0 r t t e t dt will display a more volatile feature. We consider the following congu-
Z T 0 Z T 2 rations of other parameters values given in Table 1.
Tt
t dW 1 t e t dW 2 t ; Here, the domestic interest rate, the mean-reversion level and vola-
0 0
tility of the exchange rate, the volatility of the foreign equity as well as
then the instantaneous correlation coefcient between the foreign equity
price and the exchange rate take different values when the states of
RT the economy change. Table 2 presents the prices of the FEOF option
e GT j F X T viaD 1
 Z T Z T 
T Tt and the FEOD option with different modied strike levels under the
exp iv aD r t dt iv aD 1 Y 0 e Z 0 e t dt
0 0 regime-switching mean-reversion lognormal model, where we assume
Z T Z T 
1 2 2Tt 2 Tt S(0) = 1, F(0) = 1, T = 1 and q = 0.5. The FFT method is applied to
viaD 1 e t dt 2 t t t e dt
2 
  0 Z T 0  calculate the option prices (see also Carr and Madan (1999), Lee
1 1 2 1 2
iv aD v aD aD 1 t dt : (2004), Liu et al. (2006), Wong and Guan (2011) and Kwok et al.
2 2 2 0
(2012)).

Dene
Option prices with modified strike k=0.2
  0.5
1 2 Tt
ht; Xt ; v rt iviaD 1 rt t e t 0.45
2
1  
prices

2 2 2 Tt 2 Tt
viaD 1 t e t 2e t t t : 0.4
2
0.35 State1
Then, State2
0 0.2 0.4 0.6 0.8 1
D 0; T;nv  o
exp iv aD 1 Y 0 e
T
Z 0 Z T  q
E exp hht; v; Xt idt
a aD v i2aD 1v
2 2
0
n D  o Option prices with modified strike k=0.1
Z 0  Z T 
T
exp iv aD 1 Y 0 e 0.5
X0exp diaght; vdt Q T ; 1 : State1
aD aD v i2aD 1v
2 2
0
State2
0.4
prices

0.3
Table 1
Assumptions of parameter values.
0.2
r 0 0.2 0.4 0.6 0.8 1
q
State 1 0.02 1 0.8 0.2 0.4 0.4
State 2 0.04 1 0.4 0.4 0.2 0.2
Fig. 1. Option prices corresponding to different q with k = 0.2,0.1.
302 K. Fan et al. / Economic Modelling 37 (2014) 296305

Option prices with modified strike k=0 Option prices under the RS model and the NRS model in State 2
0.35
State1 0.5
0.3 State2
0.45
prices

0.25
0.4
RS model
0.2 NRS model
0.35

0 0.2 0.4 0.6 0.8 1 0.3

prices
q
0.25

Option prices with modified strike k=0.1 0.2

0.4
State1
State2
0.15
0.3
prices

0.1
0.2
0.05
0.1
0
0 0.2 0.1 0 0.1 0.2 0.3
0 0.2 0.4 0.6 0.8 1 k
q
Fig. 4. Option prices calculated under the RS model and the NRS model in State 2.
Fig. 2. Option prices corresponding to different q with k = 0,0.1.

In the sequel, we use the valuation of an FEOF option as an example

As shown in Table 2, for both the FEOF option and the FEOD op-
tion, the option prices in State 1 are systematically higher than
those in State 2 when the strike level is xed. If the option valuation
is viewed from the perspective of a domestic investor, State 1 is a
Bad state while State 2 is a Good one. Seen from the foreign ex-
change rate aspect, the higher volatility of the foreign exchange rate
means higher potential prots when the income, denominated in
the foreign currency, translated into the domestic currency. On the
other hand, seen from the foreign equity aspect, the underlying for-
eign equity has a higher interest rate and a lower volatility in State
2. A lower volatility means less chance of the equity price being
very high or very low. In this case, the option will be less valuable.
Consequently, it is reasonable that the option prices in State 1 are
higher than the corresponding prices in State 2 due to the additional
amount of risk premium required to compensate for a disadvantage
economic condition. Note that the option prices converge quickly. In
our illustration, we always adopt the number of discretization
M = 4096.
Option prices under the RS model and the NRS model in State 1
0.6
0.55
for sensitivity analysis. The sensitivity analysis of the valuation of an
FEOD option can be conducted similarly.
4.1. The impact of q on option prices
Under our model, we assume S(0) = 1, F(0) = 1 and T = 1. To il-
lustrate the impact of q on option prices, we perform a sensitivity anal-
ysis for the option prices with respect to the rate of transition q.
From Figs. 1 and 2, when q increases, the option prices in State 1 and
State 2 display different trends. In State 1, the option prices decrease
while increase with q in State 2. Note that the value of q calibrates the
probability of the chain X transiting between State 1 and State 2, As ex-
plained earlier, the options are more expensive in State 1 and cheaper in
State 2. Consequently, the option price will decrease when q increases in
State 1, while the opposite trend is displayed in State 2. When q = 0,
the regime-switching effect will not exist. Under this degenerate case,
option prices are the highest in State 1 and lowest in State 2.
From Figs. 1 and 2, a particular attention is given to the case q = 0,
where the model dynamics for the foreign equity and the foreign ex-
change rate have no switching regimes. For simplicity, we denote our
Option prices against T and k in State 1

0.5
RS model
NRS model 0.8
0.45

0.4 0.6
Option prices
prices

0.35
0.4
0.3

0.25 0.2

0.2 0
1
0.15 0.4
0.2
0.1 0.5
0
0.2 0.1 0 0.1 0.2 0.3 0.2
k T 0 0.4 k

Fig. 3. Option prices calculated under the RS model and the NRS model in State 1. Fig. 5. Option prices corresponding to different T and k in State 1.
 K. Fan et al. / Economic Modelling 37 (2014) 296305 303

Option prices against T and k in State 2 Option prices against S(0) and F(0) in State 2

0.5 50

0.4 40

Option prices
Option prices

0.3 30

0.2 20

0.1 10

0 0
1 10
0.4 10
0.2 8
0.5 5 6
0 4
0.2 2
T 0 0.4 k 0 0
F(0) S(0)

Fig. 6. Option prices corresponding to different T and k in State 2. Fig. 8. Option prices corresponding to different S(0) and k in State 2.

regime-switching mean-reversion lognormal model and the model

Option prices against 1 and k in State 1
without regime-switching as the RS model and the NRS model re-
spectively. Figs. 3 and 4 provide us with a visual comparison between
the option prices under the RS model and the NRS model, with the as-
sumption that q = 0.5 in the RS model. 0.8
As indicated in Figs. 3 and 4, the foreign equity option prices are
lower (higher) under the RS model than those under the NRS
0.6
Option prices

model in State 1 (State 2). This is intuitively clear if State 1 and

State 2 are interpreted as a Bad state and a Good one, respective-
0.4
ly. Compared with the NRS model, the possibility of regime shifts
from the current state to the opposite one in the RS model will inev-
0.2
itably lower option prices in a Bad state while higher those in a
Good one. In other words, ignoring the regime-switching effect
would result in the FEOF option being overpriced in State 1 and 0
0.4
being underpriced in State 2. 0.2 0.4
0 0.2
0
0.2 0.2
4.2. The impact of T and k on option prices
0.4 0.4
1 k
Figs. 5 and 6 depict the price of the FEOF option versus the modied
strike value k and the maturity time T. It is easy to see that the longer the Fig. 9. Option prices corresponding to different 1 and k in State 1.
maturity time is, the higher the option price when k is xed in both the
two states. On the other hand, when the maturity time remains the
same, the price of the FEOF option decreases when k increases.

Option prices against S(0) and F(0) in State 1 Option prices against 1 and k in State 2

60 0.5

50 0.4
Option prices
Option prices

40
0.3
30
0.2
20
0.1
10

0 0
10 0.4
10 0.2 0.4
8 0 0.2
5 6 0
4 0.2 0.2
2
F(0) 0 0 S(0) 1 0.4 0.4
k

Fig. 7. Option prices corresponding to different S(0) and k in State 1. Fig. 10. Option prices corresponding to different 1 and k in State 2.
304 K. Fan et al. / Economic Modelling 37 (2014) 296305

Fig. 11. Nikkei 225 index from September 2003 to September 2013.

4.3. The impact of initial equity price S(0) and F(0) on option prices 4.4. The impact of the correlation coefcient 1 on option prices

Figs. 7 and 8 illustrate the FEOF option prices versus different initial
equity price S(0) and different initial foreign exchange rate F(0) with
k = 1, q = 0.5 and T = 1. When the initial equity price S(0) or the ini-
tial exchange rate F(0) increases, the price of the FEOF option is more
likely to be higher due to the possible higher payoff. Note that the in-
creasing speed of the FEOF option price is faster against S(0) than that
against F(0).
Furthermore, we provide sensitivity analysis for the correlation co-
efcient 1 with different k in both State 1 and State 2 under the as-
sumption that S(0) = 1, T = 1, F(0) = 1 and q = 0.5. As illustrated
in Figs. 9 and 10, the foreign equity option prices will increase as 1
does given that other parameters are xed. This indicates that an addi-
tional amount of premium is required to compensate the correlation
risk between the foreign equity price and the exchange rate.

Fig. 12. USD/JPY exchange rates from September 2003 to September 2013.
 K. Fan et al. / Economic Modelling 37 (2014) 296305
5. Empirical studies
In this section, an empirical study of the regime-switching mean-
reversion lognormal model is provided to illustrate the practical imple-
mentation of the model. Here, we take the Nikkei 225 index as the for-
eign equity and the US dollar as the domestic currency from the
perspective of an US investor. Firstly we calibrate the model parameters
to the market prices of the European call options on Nikkei 225 index
and the exchange rates between the US dollar (USD) and the Japanese
Yen (JPY). By comparing the in-sample tting errors and out-of-
sample prediction errors, we illustrate how well the RS model might
t the market data and how the RS model might improve on the NRS
model. As in the last section, we assume that the Markov chain has
only two states and use the valuation of an FEOF option as an example.
Figs. 11 and 12 describe the value of the Nikkei 225 index and the ex-
change rate USD/JPY from September 2003 to September 2013. From
these gures, one may see that the stock index and the exchange rate
exhibit both the regime-switching and mean-reversion features. Our
dataset consists of European call option prices written on the Nikkei
225 index and the exchange rate of USD/JPY for seventeen consecutive
trading days from 1 October 2013 to 17 October 2013, obtained from
the Datastream Database of Reuters. For each trading day, the options
have 9 strikes ranging from 12,000 to 16,000. The in-sample data in-
clude the option prices and exchange rates from 1 October 2013 to 14
October 2013 and the out-of-sample data include the rest option prices
and exchange rates from 15 October 2013 to 17 October 2013.
Without loss of generality, we assume the domestic risk-free interest
rate to be r = (0.02, 0.04). In addition, we assume the rate matrix is not
necessarily symmetric. To test how well the RS model and the NRS
model t the market data, we employ the method of nonlinear least
squares by minimizing the sum of squared errors between the market
prices and model prices. Denote the model parameters as : = (1,
2, 1, 2, , 1, 2, 1, 2, q12, q21, p). Here the model prices are the
weighted averages of the option prices in State 1 and State 2, with
weights being given by p and 1-p, respectively. The parameter estimates
of the RS model are given by
0:650; 0:625; 0:346; 0:201; 0:902; 0:209; 0:102;
0:697; 0:101; 0:666; 0:334; 0:315 :
Furthermore, Table 3 reports the root mean square error (RMSE) for
the tting and prediction errors of both the RS model and the NRS
model based on the same in-sample and out-of-sample data. As indicat-
ed in Table 3, the RS model has lower RMSEs for both tting and predic-
tion errors than the NRS model. The results verify that the RS model
improve quite signicantly the performance of the NRS model in the de-
scription of the dynamics of the foreign equity price and the exchange rate.
From the perspective of an investor, using the RS model for the val-
uation of foreign equity options can provide more accurate option prices
and hence reduce model risks.
6. Conclusions
We considered the valuation of foreign equity options with strike
prices in both the foreign currency and the domestic foreign currency
under a regime-switching mean-reversion lognormal model. The pa-
rameters were assumed to be modulated by an observable, nite-state
Markov chain. The FFT approach was applied to price these two kinds of
foreign equity options. Numerical examples and empirical studies were
provided to illustrate the practical implementation of our methods.
305
Table 3
In-sample tting errors and out-of-sample prediction errors.
Errors RS model NRS model
In-sample 0.4356% 1.0286%
Out-of-sample 0.9529% 1.9859%
Acknowledgments
The authors thank an anonymous referee for his/her helpful com-
ments. R. Wang and K. Fan would like to acknowledge the National
Natural Science Foundation of China (11231005), Doctoral Program
Foundation of the Ministry of Education of China (20110076110004),
Program for New Century Excellent Talents in University (NCET-09-
0356) and the Fundamental Research Funds for the Central Universities.
References
Bo, L., Wang, Y., Yang, X., 2010. Markov-modulated jump-diffusions for currency option
pricing. Insur. Math. Econ. 46, 461469.
Bollen, N.P.B., Gray, S.F., Whaley, R.E., 2000. Regime switching in foreign exchange rate:
evidence from currency option prices. J. Econ. 94, 239276.
Bufngton, J., Elliott, R.J., 2002. American options with regime switching. Int. J. Theor.
Appl. Finan. 5, 497514.
Carr, P., Madan, D., 1999. Option valuation using the fast Fourier transform. J. Comput.
Finan. 2, 6173.
Di Graziano, G., Rogers, L.C.G., 2009. Equity with Markov-modulated dividends. Quant.
Finan. 9, 1926.
Elliott, R.J., Aggoun, L., Moore, J.B., 1994. Hidden Markov Models: Estimation and Control.
Springer, Berlin-Heidelberg-New York.
Elliott, R.J., Chan, L., Siu, T.K., 2005. Option pricing and Esscher transform under regime
switching. Ann. Finance 1, 423432.
Hamilton, J.D., 1989. A new approach to the economic analysis of nonstationary time se-
ries and the business cycle. Econometrica 57, 357384.
Huang, S.C., Hung, M.W., 2005. Pricing foreign equity options under Lvy processes.
J. Futur. Mark. 25 (10), 917944.
Jorion, P., Sweeney, R.J., 1996. Mean reversion in real exchange rates: evidence and impli-
cations for forecasting. J. Int. Money Financ. 15 (4), 535550.
Kwok, Y.K., Wong, H.Y., 2000. Currency-translated foreign equity options with path de-
pendent features and their multi-asset extensions. Int. J. Theor. Appl. Finan. 3 (2),
257278.
Kwok, Y.K., Leung, K.S., Wong, H.Y., 2012. Efcient option pricing using the fast Fourier
transform. In: Duan, J.C., Hrdle, W.K., Gentle, J.E. (Eds.), Handbook of Computational
Finance. Springer-Verlag, Berlin, pp. 579604.
Lee, R.W., 2004. Option pricing by transform methods: extensions, unication, and error
control. J. Comput. Finan. 7, 5186.
Leung, K.S., Wong, H.Y., Ng, H.Y., 2013. Currency option pricing with Wishart process.
J. Comput. Appl. Math. 238, 156170.
Liu, R.H., Zhang, Q., Yin, G., 2006. Option pricing in a regime-switching model using the
fast Fourier transform. J. Appl. Math. Stoch. Anal. Artic. ID 18109, 122.
Shen, Y., Siu, T.K., 2013. Pricing bond options under a Markovian regime-switching Hull
White model. Econ. Model. 30, 933940.
Shen, Y., Fan, K., Siu, T.K., 2013. Option valuation under a double regimes witching model.
J. Futur. Mark. http://dx.doi.org/10.1002/fut.21613.
Siu, T.K., 2008. A game theoretic approach to option valuation under Markovian regime-
switching models. Insur. Math. Econ. 42 (3), 11461158.
Siu, T.K., Yang, H., Lau, J.W., 2008. Pricing currency options under two-factor Markov-
modulated stochastic volatility models. Insur. Math. Econ. 43, 295302.
Sweeney, R.J., 2006. Mean reversion in G-10 nominal exchange rates. J. Financ. Quant.
Anal. 41, 685708.
Wong, H.Y., Guan, P., 2011. An FFT-network for Lvy option pricing. J. Bank. Finan. 35,
988999.
Wong, H.Y., Lau, K.Y., 2008. Path-dependent currency options with mean reversion.
J. Futur. Mark. 28 (3), 275293.
Wong, H.Y., Lo, Y.W., 2009. Option pricing with mean reversion and stochastic volatility.
Eur. J. Oper. Res. 197, 179187.
Wong, H.Y., Zhao, J., 2010. Currency option pricing: mean reversion and multi-scale sto-
chastic volatility. J. Futur. Mark. 30 (10), 938956.
Xu, W., Wu, C., Li, H., 2011a. Foreign equity option pricing under stochastic volatility
model with double jumps. Econ. Model. 28, 18571863.
Xu, W., Wu, C., Li, H., 2011b. Accounting for the impact of higher order moments in for-
eign equity option pricing model. Econ. Model. 28, 17261729.
Yuen, F.L., Yang, H., 2010. Option pricing with regime switching by trinomial tree method.
J. Comput. Appl. Math. 233, 18211833.