Predict the Parallel Market’s Exchange Rate Volatility of Kip Per Dollar by ARCH Family Models

Abstract

The foreign exchange rate is one of the important factors of macroeconomic. If the policymaker does not have an enough efficient tools to manage the monetary which will trigger the emergence of the black market or parallel market especially when the demand of foreign currency is higher than the supply. Therefore, forecasting the exchange rate volatility will be useful for the decision of holding the assets. To study the prediction of the parallel market’s exchange rate volatility of kip per dollar by the Autoregressive Conditional Heteroscedasticity (ARCH) family models, the researchers used the daily times series data obtained between January 02, 2019 to April 2021. The empirical results expressed that ARCH (1) has the coefficient of mean equation of 1.8007 and time varying volatility in variance equation include constant and plus with the past error term. In terms of Generalized Autoregressive Conditional Heteroscedasticity (GARCH) model especially the GARCH (1,1) model, all coefficients are positive and statistically significant at 1% level, the mean equation include constant and its past value significantly predicts the present series by 0.45 points. All coefficients of the conditional variance specification meet the stability conditions of the model, but sum of the ARCH and GARCH parameters is very small. For the Threshold GARCH or TGARCH (1,3) model, the mean equation includes constant and its past value also able to predict the present series by 0.494 points, but the asymmetric of the variance equation is negative and statistically significant 1 % level, meaning that the good news has larger effects on the volatility of the market than bad news. According to the Exponential GARCH (1,1) or EGARCH (1,1) model, the mean equation includes constant value and its past value significantly predicts current series by 0.253 points and all the coefficients of the asymmetric term of variance equation is negative and statistically significant at 1% level, indicated that bad news has larger effects on the volatility of the market than good news. However, the constant coefficient is negative and the previous volatility is insignificant at 5% level statistically so this model is not satisfied. In model’s selection, we found that the AIC and SIC value of the ARCH model is lower than others therefore we conclude that ARCH (1) model is an appropriate model to predict the parallel market’s exchange rate volatility in Lao PDR especially the exchange of kip per UD dollar.