Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
14
Forecasting Thai Baht vs U.S. Dollar Rates Using the ARIMA Model
...............
Department of Business Economics
School of Economics, Bangkok University
E-mail: [email protected]
ARIMA . . 55
. . 55
9-
ARIMA (1, , ) (1, 1, )
. . 55
-
15
Abstract
Since Thailand adjusted to implement a floating exchange rate regime, the exchange rate
of the Thai baht against the U.S. dollar has been changing all the time. This has affected
loss or gain profits for international business transactions as a result of foreign exchange
rate fluctuations. This study aims to create a model to estimate and forecast the exchange
rate of the Thai baht against the U.S. dollar based on analysis of the time series using the
ARIMA model and a data collection from anuary 11 to ovember 1 . The findings
reveal that during the study period the THB/USD exchange rate fell between THB 9-
/ USD. The findings also indicate that the appropriate ARIMA model for forecasting
the exchange rate of THB/USD is (1, , ) (1, 1, ). When using this model to predict the
exchange rate of THB/USD in the second half of the year 15, it turns out that the
exchange rate tends to be stronger, reaching THB - / USD. This currency movement
is consistent with the fact that the U.S. economy is slowly recovering. Therefore, business
transactions dealing with USD should be done with extreme care. Traders, especially
importers, are recommended to have risk assessment and preventive measures because
the U.S. currency is likely to become stronger.
THB/USD Exchange Rate, ARIMA Model, Forecasting
16
(Managed Float) . . 5
5
5
. . 5
- 5
-
1
: , 55
17
(
, 55 : 11 -11 ;
, 555: 115-1 )
“ ”
(The urchasing ower arity Theory: )
18
(The Absolute
urchasing ower arity)
(The Relative
urchasing ower arity)
1.
1
Si
i /
i
Si
1
i
i
.
= ih-i
f
e , e1
1
ih, i
f
(Time Series Data)
( 55 )
5 - 55
1
eural etworks
e1-ee
19
Autoregressive Integrated Moving
Average (ARIMA) Generalized
AutoRegressive Conditional Heteroskedasticity
(GARCH)
Mean Absolute ercentage Error (MA E)
ARIMA( , , )
MA E .1
ARIMA ( , , ) with GARCH-M (1,1)
GARCH-M
MA E .1 5
eural etworks
1 hidden layer
MA E . 9
MA E
ARIMA with GARCH-M
ARIMA eural etworks
( 55 )
5 9 1
55 1,1 9
1)
Augmented Dickey-Fuller Test
(ADF)
(ARMA (p,q)) )
ector
Autoregressive Moving Average-GARCH
( ARMA-GARCH) ector Autoregressive
Moving Average Asymmetric GARCH
( ARMA-AGARCH) Constant
Conditional Correlation (CCC)
ARMA AR( )
MA( )
ARMA
AR(1) MA(1)
ARMA-AGARCH (1,1)
CCC
ARMA-
AGARCH
20
(ccc)
Rout, et al. ( 1 ) “Forecasting of
currency exchange rates using adaptive ARMA
model with differential evolution based training”
1-1-19 1-1 - 5, 1-1-19 1 1-1- 5
1-1-19 1 1-1- 5
ARMA
Mammadova ( 1 ) “Forecasting
exchange rates using ARMA and neural network
model” Brazilan
Real US Dollar
1999 1
ARMA
eural etwork Baharumshah, and Sen
( ) “The predictability of the
ASEA -5 exchange rates”
5
1 19 1
1999 Autoregressive
Integrated Moving Average (ARIMA) (p, d, q)
ARMA
Forward-Backward Least-Mean-Square
(FBLMS)
eural
etwork
ARMA ARIMA
-5
199
(
, 55 : 1-1 )
ARIMA
- (Box- enkins)
. . 55 . . 55
21
. . 55 - 55
(t)
-
(Box, enkins, and Reinsel, 199 )
Y t (Yt)
Y
(Yt-1, Y
t--, ..)
Yt Y
t-1, Y
t- ..
Yt 1
, Yt
, .
-
1.
(Stationary)
(Autocorrelation
function: ACF) t
( artial autocorrelation function: ACF)
(Yt)
.
1
ACF ACF
. ARIMA (p, d, q) ( ,
S, )
ACF ACF
.
5.
ARIMA (p, d, q) ( , S, ) (Seasonal
Autoregressive Integrated Moving Average)
.
ARIMA
. . 55
. . 55 (t, Y)
t Y
9-
22
ACF ACF
ARIMA
ARIMA (1, ,1) (1, 1, 1)
ARIMA (1, , ) (1, 1, )
23
ACF ACF
24
ARIMA (1, , ) (1, 1, )
AR(1) . SAR(1)
- .5 . 5 R-squared
. MA E 1.5 5
1
1.
ARIMA (1, , ) (1, 1, )
.
.
- .5
.
.
1.5 5
. 99
. 9
.19
1.1 9
. 55
- .
.
.
. 1
25
.
Kolmogorov
. 5
ormal Distribution
. ACF
ACF ( 5)
lag
White oise
ACF ACF Box- enkins
(THB/USD)
. . 55
. . 55
. . 55 . . 55
26
( ) .95, .5 , . , 1.
1. THB/USD
55
( ) .55, . 5, . 5,
.9 , . , . , . . THB/USD
ARIMA (1, , ) (1, 1, )
t 1
Yt-1 t-1
- 1Y
t-2 1Y
t-12 -
1 1
Yt-13
- 1Y
t-13 1 1Y
t-14
t t-1 t-2
-
0.523Yt-12 t-13
- 0.437Yt-14
t
.
AR(1)
.
AR, Seasonal(1)
- .5
t
(THB/USD) . . 55
.95 .5 . 1. 1. .55 . 5 . 5 .9 . . . .
. . 55
27
. . 55
. . 55
ARIMA (p, i, q) ( , S, )
ARIMA (1, , ) (1, 1, )
. . 55
55
ARIMA
(Ex-post Forecast)
Baharumshah, A. and Sen, L. .
nline]. Available: http://econwpa.
repec.org/eps/if/papers/ / .pdf
Bank of Thailand. 15.
nline]. Available: http://www .bot.or.th/
statistics/B TWEBSTATaspx reportID=
1 language=TH (in Thai).
. 55 .
]. :
h t tp : / /www .bo t . o r . t h / s ta t i s t i cs /
B TWEBSTATaspx report ID=1
language=TH
Box, G.E. ., enkins, G.M., and Reinsel, G.C.
199 . rd. ed. Englewood Cliffs, :
rentice-Hall.
Hatchavanich, D. 1 . “A Comparison of
Forecasting Models for the Monthly
Consumer rice Index: Box- enkins
and Exponential Smoothing Models.”
, : 1 -11 . (in Thai).
28
. 55 . “
:
-
.”
, : 1 -11 .
ansod. A. . “Accuracy Comparison in
Foreign Exchange Rate Forecasting
Between eural etworks and ARIMA
GARCH-M Models.” Master’s Thesis,
Department of Economics, Chiang Mai
University. (in Thai).
. 55 . “
.”
.
Mammadova, G. 1 .
Master’s Thesis, Department of
Economics, Western Illinois University.
Rattanapongpinyo. T. 1 . “A Study of Factors
Affecting the Short-term Movement of the
Thai Baht vs the US Dollar.”
, 1: 1-1 . (in Thai).
. 55 . “
.”
, 1:
1-1 .
Rout, M., et al. 1 . “Forecasting of Currency
Exchange Rates Using and Adaptive
ARMA Model with Differential Evolution
Based Training.”
, 1: -1 .
Saothayanun, L., et al. 1 . “A Comparison of
the Forecasts for Rubber rices Using
ARIMA and GARCH Models.”
, : 115-1 . (in Thai).
, . 555. “
ARIMA GARCH.”
, :
115-1 .
Sinchaikit, S. 11. “Modeling of Exchange Rate
and Gold rice olatilities of Thailand Using
Bivariate GARCH.” Master’s Thesis,
Department of Economics, Chiang Mai
University. (in Thai).
. 55 . “
.”
.
29
Channarong Chaiphat received his Master of Economics from Kasetsart
University, Thailand. He is currently an assistant professor at the School
of Economics, Bangkok University. His main interest is in International
Monetary Economics.