Description
This paper examines whether the exchange rates of the Asia-Pacific countries can be captured by the
Markov switching model (MSM). Using data from January 2000 to December 2011, the real interest
differential (RID) model is tested first. However, supporting evidence is limited, and results are markedly
different across the countries. It is worth noting that the signs on the coefficients for fundamental factors
are mostly wrong from the RID model perspective. By using the MSM-RID model, the results identify that
two regimes exist and persist, which is consistent with earlier literature indicating that there are
complex influences in exchange rate determination. This leads to the conclusion that the results are
strongly in favor of a nonlinear relationship between exchange rate volatility and fundamental factors.
When the probabilities transition matrixes of MSM are allowed to change [MSM-RID-time varying
transition probabilities model (TVTP)], it is found that MSM-RID-TVTP outperforms the MSM-RID model.
MSM classifies the currencies regimes and provides information about the change of currency prices in
some Asian-Pacific currencies.
Markov regimes switching with monetary fundamental-based exchange rate
model
Jing-Tung Wu
*
Department of Finance, Ming Chuan University, Taiwan
a r t i c l e i n f o
Article history:
Received 4 July 2012
Accepted 25 September 2013
Available online 16 April 2015
Keywords:
RID
Exchange rate
Markov switching model
a b s t r a c t
This paper examines whether the exchange rates of the Asia-Paci?c countries can be captured by the
Markov switching model (MSM). Using data from January 2000 to December 2011, the real interest
differential (RID) model is tested ?rst. However, supporting evidence is limited, and results are markedly
different across the countries. It is worth noting that the signs on the coef?cients for fundamental factors
are mostly wrong from the RID model perspective. By using the MSM-RID model, the results identify that
two regimes exist and persist, which is consistent with earlier literature indicating that there are
complex in?uences in exchange rate determination. This leads to the conclusion that the results are
strongly in favor of a nonlinear relationship between exchange rate volatility and fundamental factors.
When the probabilities transition matrixes of MSM are allowed to change [MSM-RID-time varying
transition probabilities model (TVTP)], it is found that MSM-RID-TVTP outperforms the MSM-RID model.
MSM classi?es the currencies regimes and provides information about the change of currency prices in
some Asian-Paci?c currencies.
© 2015 College of Management, National Cheng Kung University. Production and hosting by Elsevier
Taiwan LLC. All rights reserved.
1. Introduction
The Asia-Paci?c economies use a variety of exchange rate sys-
tems; they commonly peg their exchange rate to the US Dollar.
After the 1997 Asian crisis, their monetary authorities loosened the
control of exchange rates and let thembecome more “?oating”. The
linkage between exchange rates and fundamentals has been
extensively analyzed in international ?nance literature. However,
earlier studies found little evidence of linear serial dependence in
exchange rates. Exchange rates are often regarded as moving
accidentally, because no permanent factors can be detected.
The failure of exchange rate models initiated the use of the
Markov switching model (MSM) by Engle and Hamilton (1990) .
They developed a new statistical model of exchange rate dynamics
as a sequence of stochastic, segmented time trends. The conditional
distribution of quarterly exchange rate returns simultaneously
switches in both mean and variance. They found long swings in the
data, and the model generated better forecasts than a randomwalk.
Marsh (2000) added the fundamental factor to the model, using
interest rates to investigate the daily variation of exchange rates.
Dewachter (2001) extended the MSM by introducing separate and
independence latent variables for the dynamics in mean and vari-
ance. The applications of the MSM captured some major dynamics
that characterize exchange rate behavior, although the structure
may vary over time. Caporale and Spagnolo (2004) modeled East
Asian exchange rates using the MSM, and their results shown that
the behavior of exchange rates is nonlinear. Fr€ ommel, MacDonald,
and Menkhoff (2005a, 2005b) modi?ed the real interest differen-
tial (RID) model with a switching process in the underlying fun-
damentals. They found that the factors that were proved to be
closely related to regime switches were short-term interest rates,
in?ation differentials, and differences in economic growth. The
forecasting ability of MSM is good. Kumah (2011) used a three-
regime MSM to capture the Kyrgyz Republic foreign exchange
market pressure.
In this paper, I suggest that the exchange rates of Asia-Paci?c
economies are generated by a two-regime Markov switching
autoregressive model. Hence, this paper extends earlier literature
and goes beyond the content of Engle and Hamilton (1990) and
Fr€ ommel et al. (2005a, 2005b) by using the Asia-Paci?c market
currencies in comparing the difference between them.
* Corresponding author. 250 Sec. 5, Chung Shan N. Road, Taipei, Taiwan.
E-mail address: [email protected].
Peer review under responsibility of College of Management, National Cheng
Kung University.
HOSTED BY
Contents lists available at ScienceDirect
Asia Paci?c Management Review
j ournal homepage: www. el sevi er. com/ l ocat e/ apmrvhttp://dx.doi.org/10.1016/j.apmrv.2014.12.009
1029-3132/© 2015 College of Management, National Cheng Kung University. Production and hosting by Elsevier Taiwan LLC. All rights reserved.
Asia Paci?c Management Review 20 (2015) 79e89
The remainder of the paper is structured in the following
manner. Section 2 presents the traditional monetary fundamental-
based exchange rate model. In Section 3, I introduce the MSM.
Section 4 presents the empirical analysis. Finally, Section 5 con-
cludes this paper.
2. Monetary fundamental-based exchange rate model
The monetary exchange rate model relied on a single state
relationship between fundamentals and the exchange rate. One
application of the monetary exchange rate model is the UIP hy-
pothesis. It stated that the change in the exchange rate should
incorporate any interest rate differentials between the two cur-
rencies. UIP suggested that when we borrowed money in our
home country and lent it to another country with a higher (or
lower) interest rate, we should expect a zero return due to the
appreciation (or depreciation) in the domestic currency. Frankel
(1979) assumed that higher interest rates re?ect lower money
demand. Therefore, a higher domestic interest rate is related to an
increase of the price of foreign currency. The UIP hypothesis stated
that high interest rate currencies appreciate over time and
therefore pay a positive expected return. The relationship can be
expressed as follows:
De
t
¼ i
t
Ài
*
t
(1)
where De
t
denotes the log of the exchange rate volatility at time t, i
t
is the interest rate, and the asterisk refers to a foreign country.
The monetary view may use several fundamental factors to
describe the ?uctuation of exchange rates. Frenkel (1976) and
Mussa (1976) brought up this idea and applied it to their research.
Its fundamental building block is absolute purchasing power parity.
It starts fromthe de?nition of the exchange rate as the relative price
in terms of the relative supply and demand for their monies.
Frankel (1979) extended other versions of the monetary models,
and proposed the RID model:
De
t
¼aþb
1
_
Dm
t
ÀDm
*
t
_
þb
2
_
Dy
t
ÀDy
*
t
_
þb
3
_
Dsi
t
ÀDsi
*
t
_
þb
4
_
Dli
t
ÀDli
*
t
_
þ?
t
(2)
where m
t
is the log of the money supply at time t, y
t
is the log of
national production, si
t
is the short-term interest rate, and li
t
is
the long-term interest rate. The RID model combined some
traditional concepts on determinations of exchange rates. It is
referred to as a hybrid monetary model by MacDonald and Marsh
(1999). Earlier empirical studies of the RID model have found that
its performance is good, especially when applied to free ?oating
currencies.
Each currency return and fundamental factor is calculated as the
difference of the log percentage change with the last year. Taking
money supply as an example, the calculated equation is as follows:
Dm
t
¼
m
home
t
Àm
home
tÀ12
m
home
tÀ12
À
m
foreign
t
Àm
foreign
tÀ12
m
foreign
tÀ12
(3)
The use of 1-year changes could reduce the seasonal effect and
noise from short-term exchange rate ?uctuation. Central bank and
government statistics also rely on this approach to monitor the
change of fundamental factors. There is another method that uses
the difference of the change with last month. The bene?t is that it
adopts recent information, but the shortage is it causes a seasonal
effect. In this paper, I show the results of yearly change only and
omit the results of monthly change data.
Monthly data for six major Asia-Paci?c currencies are employed:
Japanese Yen (JPY), Hong Kong Dollar (HKD), South Korea Won
(KRW), New Taiwan Dollar (NTD), Chinese Yuan (RMB), and
Singapore Dollar (SGD), over the period from January 2000 to
December 2011, for a total of 144 observations. The exchange rates
are all against the US Dollar (USD). Data are obtained from the
Taiwan Economic Journal (TEJ), speci?cally for JPY (January
2000eDecember 2011), HKD(January 2001eDecember 2011), KRW
(October 2001eDecember 2011), NTD (January 2000eDecember
2011), RMB (July 2005eDecember 2011), and SGD (January
2000eDecember 2011). The estimation process was performed us-
ing MATLAB (3 Apple Hill Drive Natick, Massachusetts 01760 USA).
Table 1 shows the results. HKDis the only currency for which the
coef?cients have the same sign as the RIDmodel expected. KRWhas
two coef?cients performing as the same sign, which are industrial
productionand long-terminterest rate. However, all the coef?cients
of JPY are against the RID model expectations. There is only one
coef?cient that performs as the RID model expected, the industrial
production, while the signs of the currencies are all the same.
The explanation power of the RID model is higher than the UIP
hypothesis. Of the six currencies (except the JPY) which have an R
2
above 0.250, the highest is RMB, for which the value is 0.405.
Compared with the UIP hypothesis, the RID model performs much
better.
The simplicity of the monetary model is very attractive. How-
ever, it requires many assumptions, such as assuming perfect sub-
stitutability of domestic and foreign assets, free adjustment of the
exchange rate to equilibrate supply, and demand in the foreign
exchange market. From Table 1, it can be seen that the RID model is
not easily applied to all countries. It should be noted that the signs
of the coef?cients are mostly not in accordance with the expected
values from the RID. Every currency faces different economic sit-
uations and their behaviors are not alike. This may be the reason
why Meese and Rogoff (1983) mentioned that the exchange rate
changes cannot be forecast by fundamentals at horizons of less than
a year. Engle (2000) had stated that money demand, purchasing
power parity, and UIP do not work well. The monetary funda-
mentals do not help to predict exchange rates retain conventional
wisdom. This can be seen as another motivation for applying MSM.
Indeed, Engle and Hamilton (1990) have also challenged this,
reporting evidence in favor of a Markov switching regimes process
for exchange rate changes.
Table 1
The real interest differential (RID) model.
Currency JPY HKD KRW NTD RMB SGD Expected signs
Intercept e4.813 0.171 2.111 e0.644 0.602 e1.672 n.a.
Money supply e0.550 0.015 e0.529 e0.371 >0
Industrial production n.a. e0.439 e0.136 e0.159
This paper examines whether the exchange rates of the Asia-Pacific countries can be captured by the
Markov switching model (MSM). Using data from January 2000 to December 2011, the real interest
differential (RID) model is tested first. However, supporting evidence is limited, and results are markedly
different across the countries. It is worth noting that the signs on the coefficients for fundamental factors
are mostly wrong from the RID model perspective. By using the MSM-RID model, the results identify that
two regimes exist and persist, which is consistent with earlier literature indicating that there are
complex influences in exchange rate determination. This leads to the conclusion that the results are
strongly in favor of a nonlinear relationship between exchange rate volatility and fundamental factors.
When the probabilities transition matrixes of MSM are allowed to change [MSM-RID-time varying
transition probabilities model (TVTP)], it is found that MSM-RID-TVTP outperforms the MSM-RID model.
MSM classifies the currencies regimes and provides information about the change of currency prices in
some Asian-Pacific currencies.
Markov regimes switching with monetary fundamental-based exchange rate
model
Jing-Tung Wu
*
Department of Finance, Ming Chuan University, Taiwan
a r t i c l e i n f o
Article history:
Received 4 July 2012
Accepted 25 September 2013
Available online 16 April 2015
Keywords:
RID
Exchange rate
Markov switching model
a b s t r a c t
This paper examines whether the exchange rates of the Asia-Paci?c countries can be captured by the
Markov switching model (MSM). Using data from January 2000 to December 2011, the real interest
differential (RID) model is tested ?rst. However, supporting evidence is limited, and results are markedly
different across the countries. It is worth noting that the signs on the coef?cients for fundamental factors
are mostly wrong from the RID model perspective. By using the MSM-RID model, the results identify that
two regimes exist and persist, which is consistent with earlier literature indicating that there are
complex in?uences in exchange rate determination. This leads to the conclusion that the results are
strongly in favor of a nonlinear relationship between exchange rate volatility and fundamental factors.
When the probabilities transition matrixes of MSM are allowed to change [MSM-RID-time varying
transition probabilities model (TVTP)], it is found that MSM-RID-TVTP outperforms the MSM-RID model.
MSM classi?es the currencies regimes and provides information about the change of currency prices in
some Asian-Paci?c currencies.
© 2015 College of Management, National Cheng Kung University. Production and hosting by Elsevier
Taiwan LLC. All rights reserved.
1. Introduction
The Asia-Paci?c economies use a variety of exchange rate sys-
tems; they commonly peg their exchange rate to the US Dollar.
After the 1997 Asian crisis, their monetary authorities loosened the
control of exchange rates and let thembecome more “?oating”. The
linkage between exchange rates and fundamentals has been
extensively analyzed in international ?nance literature. However,
earlier studies found little evidence of linear serial dependence in
exchange rates. Exchange rates are often regarded as moving
accidentally, because no permanent factors can be detected.
The failure of exchange rate models initiated the use of the
Markov switching model (MSM) by Engle and Hamilton (1990) .
They developed a new statistical model of exchange rate dynamics
as a sequence of stochastic, segmented time trends. The conditional
distribution of quarterly exchange rate returns simultaneously
switches in both mean and variance. They found long swings in the
data, and the model generated better forecasts than a randomwalk.
Marsh (2000) added the fundamental factor to the model, using
interest rates to investigate the daily variation of exchange rates.
Dewachter (2001) extended the MSM by introducing separate and
independence latent variables for the dynamics in mean and vari-
ance. The applications of the MSM captured some major dynamics
that characterize exchange rate behavior, although the structure
may vary over time. Caporale and Spagnolo (2004) modeled East
Asian exchange rates using the MSM, and their results shown that
the behavior of exchange rates is nonlinear. Fr€ ommel, MacDonald,
and Menkhoff (2005a, 2005b) modi?ed the real interest differen-
tial (RID) model with a switching process in the underlying fun-
damentals. They found that the factors that were proved to be
closely related to regime switches were short-term interest rates,
in?ation differentials, and differences in economic growth. The
forecasting ability of MSM is good. Kumah (2011) used a three-
regime MSM to capture the Kyrgyz Republic foreign exchange
market pressure.
In this paper, I suggest that the exchange rates of Asia-Paci?c
economies are generated by a two-regime Markov switching
autoregressive model. Hence, this paper extends earlier literature
and goes beyond the content of Engle and Hamilton (1990) and
Fr€ ommel et al. (2005a, 2005b) by using the Asia-Paci?c market
currencies in comparing the difference between them.
* Corresponding author. 250 Sec. 5, Chung Shan N. Road, Taipei, Taiwan.
E-mail address: [email protected].
Peer review under responsibility of College of Management, National Cheng
Kung University.
HOSTED BY
Contents lists available at ScienceDirect
Asia Paci?c Management Review
j ournal homepage: www. el sevi er. com/ l ocat e/ apmrvhttp://dx.doi.org/10.1016/j.apmrv.2014.12.009
1029-3132/© 2015 College of Management, National Cheng Kung University. Production and hosting by Elsevier Taiwan LLC. All rights reserved.
Asia Paci?c Management Review 20 (2015) 79e89
The remainder of the paper is structured in the following
manner. Section 2 presents the traditional monetary fundamental-
based exchange rate model. In Section 3, I introduce the MSM.
Section 4 presents the empirical analysis. Finally, Section 5 con-
cludes this paper.
2. Monetary fundamental-based exchange rate model
The monetary exchange rate model relied on a single state
relationship between fundamentals and the exchange rate. One
application of the monetary exchange rate model is the UIP hy-
pothesis. It stated that the change in the exchange rate should
incorporate any interest rate differentials between the two cur-
rencies. UIP suggested that when we borrowed money in our
home country and lent it to another country with a higher (or
lower) interest rate, we should expect a zero return due to the
appreciation (or depreciation) in the domestic currency. Frankel
(1979) assumed that higher interest rates re?ect lower money
demand. Therefore, a higher domestic interest rate is related to an
increase of the price of foreign currency. The UIP hypothesis stated
that high interest rate currencies appreciate over time and
therefore pay a positive expected return. The relationship can be
expressed as follows:
De
t
¼ i
t
Ài
*
t
(1)
where De
t
denotes the log of the exchange rate volatility at time t, i
t
is the interest rate, and the asterisk refers to a foreign country.
The monetary view may use several fundamental factors to
describe the ?uctuation of exchange rates. Frenkel (1976) and
Mussa (1976) brought up this idea and applied it to their research.
Its fundamental building block is absolute purchasing power parity.
It starts fromthe de?nition of the exchange rate as the relative price
in terms of the relative supply and demand for their monies.
Frankel (1979) extended other versions of the monetary models,
and proposed the RID model:
De
t
¼aþb
1
_
Dm
t
ÀDm
*
t
_
þb
2
_
Dy
t
ÀDy
*
t
_
þb
3
_
Dsi
t
ÀDsi
*
t
_
þb
4
_
Dli
t
ÀDli
*
t
_
þ?
t
(2)
where m
t
is the log of the money supply at time t, y
t
is the log of
national production, si
t
is the short-term interest rate, and li
t
is
the long-term interest rate. The RID model combined some
traditional concepts on determinations of exchange rates. It is
referred to as a hybrid monetary model by MacDonald and Marsh
(1999). Earlier empirical studies of the RID model have found that
its performance is good, especially when applied to free ?oating
currencies.
Each currency return and fundamental factor is calculated as the
difference of the log percentage change with the last year. Taking
money supply as an example, the calculated equation is as follows:
Dm
t
¼
m
home
t
Àm
home
tÀ12
m
home
tÀ12
À
m
foreign
t
Àm
foreign
tÀ12
m
foreign
tÀ12
(3)
The use of 1-year changes could reduce the seasonal effect and
noise from short-term exchange rate ?uctuation. Central bank and
government statistics also rely on this approach to monitor the
change of fundamental factors. There is another method that uses
the difference of the change with last month. The bene?t is that it
adopts recent information, but the shortage is it causes a seasonal
effect. In this paper, I show the results of yearly change only and
omit the results of monthly change data.
Monthly data for six major Asia-Paci?c currencies are employed:
Japanese Yen (JPY), Hong Kong Dollar (HKD), South Korea Won
(KRW), New Taiwan Dollar (NTD), Chinese Yuan (RMB), and
Singapore Dollar (SGD), over the period from January 2000 to
December 2011, for a total of 144 observations. The exchange rates
are all against the US Dollar (USD). Data are obtained from the
Taiwan Economic Journal (TEJ), speci?cally for JPY (January
2000eDecember 2011), HKD(January 2001eDecember 2011), KRW
(October 2001eDecember 2011), NTD (January 2000eDecember
2011), RMB (July 2005eDecember 2011), and SGD (January
2000eDecember 2011). The estimation process was performed us-
ing MATLAB (3 Apple Hill Drive Natick, Massachusetts 01760 USA).
Table 1 shows the results. HKDis the only currency for which the
coef?cients have the same sign as the RIDmodel expected. KRWhas
two coef?cients performing as the same sign, which are industrial
productionand long-terminterest rate. However, all the coef?cients
of JPY are against the RID model expectations. There is only one
coef?cient that performs as the RID model expected, the industrial
production, while the signs of the currencies are all the same.
The explanation power of the RID model is higher than the UIP
hypothesis. Of the six currencies (except the JPY) which have an R
2
above 0.250, the highest is RMB, for which the value is 0.405.
Compared with the UIP hypothesis, the RID model performs much
better.
The simplicity of the monetary model is very attractive. How-
ever, it requires many assumptions, such as assuming perfect sub-
stitutability of domestic and foreign assets, free adjustment of the
exchange rate to equilibrate supply, and demand in the foreign
exchange market. From Table 1, it can be seen that the RID model is
not easily applied to all countries. It should be noted that the signs
of the coef?cients are mostly not in accordance with the expected
values from the RID. Every currency faces different economic sit-
uations and their behaviors are not alike. This may be the reason
why Meese and Rogoff (1983) mentioned that the exchange rate
changes cannot be forecast by fundamentals at horizons of less than
a year. Engle (2000) had stated that money demand, purchasing
power parity, and UIP do not work well. The monetary funda-
mentals do not help to predict exchange rates retain conventional
wisdom. This can be seen as another motivation for applying MSM.
Indeed, Engle and Hamilton (1990) have also challenged this,
reporting evidence in favor of a Markov switching regimes process
for exchange rate changes.
Table 1
The real interest differential (RID) model.
Currency JPY HKD KRW NTD RMB SGD Expected signs
Intercept e4.813 0.171 2.111 e0.644 0.602 e1.672 n.a.
Money supply e0.550 0.015 e0.529 e0.371 >0
Industrial production n.a. e0.439 e0.136 e0.159