Modelling non-linear exchange rate data in Sri Lanka that exhibit disturbed trends over time

Abstract

Most free market economies worldwide depend heavily on demand and supply, which is influenced by exchange rates. As a developing country, Sri Lanka should emphasise exchange rates more than other factors to maintain its economic sustainability. It is important to study the characteristics and behaviour of exchange rates to maintain sustainability. Studies show that monetary and non-monetary factors affect the exchange rate. As a result, currency exchange rates frequently exhibit nonlinear trends, including sharp fluctuations induced by events such as the recent scenario of the sudden devaluation of the Sri Lankan rupee in early 2022. Modelling exchange rate data under these nonlinear tendencies is difficult, and this difficulty is made challenging by the abrupt shifts in the overall trend brought on by unforeseen disruptions. On that account, the combination of semi-parametric and interrupted time series models was proposed to develop a flexible model that describes the exchange rate movement in Sri Lanka. The Bayesian approach was employed for model fitting with R2Winbugs because it enables rigorous handling of uncertainty. The next value of the exchange rate was obtained by applying first-order bivariate Taylor series approximation to the mean response of the model. The result from the posterior predictive evaluation shows that the developed model accurately captured the trends and variability of the exchange rate. Furthermore, all model parameters are significant, indicating that the monetary (inflation, trade balance, tourist arrivals) and nonmonetary (sudden devaluation of Sri Lankan rupee) variables are important for studying exchange rate variabilities. The findings suggest that the developed model provides information about predicting and forecasting the future exchange rate movement in Sri Lanka. These predictions and forecast values would help to balance the economic sustainability in Sri Lanka.