Proceedings of the IConSSE FSM SWCU 2015, pp. MA.93–97 ISBN: 978-602-1047-21-7
SWUP
MA.93
Combination of volatility and Markov-switching models for financial crisis in Indonesia based on real exchange rate indicators
Sugiyanto
†
and Etik Zukhronah
Department of Mathematics, FMIPA UNS Ir. Sutami Street 36A Surakarta, Central Java, Indonesia
Abstract
Financial crisis that happened in 1997 to 1998 resulted in decline of the economy in Indonesia. Therefore in this study will be built a model of the financial crisis in Indonesia
which can be used in anticipation of or preparation for what to do in the future. The model is based on indicators of the real exchange rate from January 1990 to December
2014 using a combination of volatility and Markov_Switching models. The results showed that the model SWARCH2,4 and SWARCH3,4 can catch a signal crisis in
Indonesia on February 1998 and July 1998.
Keywords financial crisis, volatility model, Markov-Switching model, real exchange rate, SWARCH
1. Introduction
The crisis that hit the Indonesian economy in 1997 to 1998 is caused by the fall of bath Thailand exchange rate in July 1997. The impact of the crisis in 1997 to 1998 are quite severe
makes researchers interested in establishing a model that illustrates the crisis. The model can be built through the monitoring of a number of macroeconomic indicators Abimanyu
Imansyah, 2008. According to Ford et al. 2007, one of the indicators that can be used to detect the currency crisis in Indonesia is the real exchange rate indicators.
The real exchange rate is the relative price of goods between the two countries Mankiw, 2003. The real exchange rate can be calculated from the nominal exchange rate
multiplied by the price level. Price level used is the consumer price index CPI. Kaminsky et al. 1997 states that a high real exchange rate will reduce exports so as to reduce the supply
of foreign currency is entered. That led to the weakening of the domestic currency and will likely result in a huge crisis.
Monthly data of real exchange rate is a time series data. Cryer 1986 introduced one model of stationary time series data, namely autoregressive moving average ARMA. ARMA
models have the assumption of constant residual variance homoscedasticity. Real exchange rate data have indicated volatility or volatility clustering. Volatility clustering of data that is
gathering a bunch of great value and followed a group of small-value data. These circumstances indicate that the variance is not constant, so the real exchange rate data does
not meet the homoscedasticity assumptions.
Engle 1982 introduced a model of autoregressive conditional heteroscedasticity ARCH which can model the data that has heteroscedasticity. However, the financial time
series data can undergo structural changes caused by changes in policy, war or natural
†
Corresponding author. E-mail address: sugiy50yahoo.co.id
Combination of volatility and Markov-switching models for financial crisis in Indonesia based on real exchange rate indicators
SWUP
MA.94
disasters and ARCH model does not take into account the structural changes that occur in such volatility.
Hamilton 1989 introduced Markov switching models as an alternative modelling of time series data that undergo structural changes. Hamilton combined Markov switching and
autoregressive models resulting in a Markov switching autoregressive models SWAR. Hamilton Susmel 1994 introduced a model that combined ARCH models and
Markov_Switching models then called Markov Switching ARCH SWARCH. SWARCH model can explain changes in the structure and illustrates the volatility. Some researchers have
applied SWARCH model to detect crisis that occur in a country. Among these researchers were Chen Lin 2000 which apply SWARCH models to identify the stock market volatility
in Taiwan. Then Chang et al. 2010 also apply SWARCH models to identify the volatility of the stock market and the exchange rate in Korea as well as the financial crisis global.
This research will be carried out modeling of the financial crisis in Indonesia is based on indicators of the real exchange rate using combined of volatility and Markov switching
models. The real exchange rate data that indicated heteroscedasticity and undergo structural changes can be modeled by SWARCH model of two and three states.
2. Materials and methods