Modeling USD/ LKR exchange rate using time series techniques

Abstract

Modeling Exchange Rate is one of the most vital and essential parts of financial econometrics. This study aimed to formulate time series models using both univariate and multivariate time series techniques. The macro-economic factors of Inflation Rate, Trade balance, Net Foreign Assets, and Foreign Remittances have been used in this study which were chosen by an extensive literature survey. Data from January 2015 to October 2022 was used in this study. A univariate model was formulated using ARIMA and ARCH/GARCH modeling techniques. The GARCH model with the mean equation which is estimated by an ARIMA (1,1,1) Model along with a dummy variable and the variance equation which is estimated by GARCH (1,0) Model was selected as the best-fitted univariate model with a R-square value 81.8%. Using ARDL modeling technique associated with ARDL bound tests for cointegration was used to formulate the multivariate model. The ARDL (3,4,3,0,0) was selected as the best-fitted model which satisfied all the assumptions along with model specifications. It was identified that there exists significant co-integration among the exchange rate and the selected macro-economic factors. Error correction model was formulated based on the selected ARDL model and according to the results, the speed of adjustment is -5.78% which means that the short-run model will be converged to the long-run equilibrium by 5.78% for each period moving forward. Among the two models selected from univariate and multivariate analysis, ARDL (3,4,3,0,0) model was selected as the final best fitted model. Key Words: Exchange Rate, Advanced Time Series Analysis, ARIMA, ARCH/GARCH, ARDL, Cointegration