The Effects of Technical Indicators on Exchange Rates: Empirical Insights from Quantile Regression Models

Abstract

The modelling and forecasting of foreign exchange rates have proven challenging due to the prevailing extreme volatility and uncertain nature. Therefore, the primary objective of this investigation is to analyze and model the dynamics of exchange rates of the EURO, GBP, and USD against LKR using technical indicators of the previous day's low, high, and opening price, along with lagged and moving average (MA) values of closing prices. The generalized lambda distribution (GLD) regression models were employed in this study due to the non-normal behaviour exhibited by the error term. The GLD, being a versatile probability distribution, can encompass diverse distributional forms. Regarding the fitted GLD regression models, quantile regression (QR) models were utilised under two distinct conditions on closing price values of exchange rates: Case-I, coefficients were permitted to vary while maintaining a fixed intercept; Case-II, all coefficients were allowed to vary. The empirical study uses the daily data collected from the Yahoo Finance website from January 1, 2008, to February 28, 2022. Our findings show that the influence of technical indicators on exchange rate returns varies significantly across different quantiles. The models that demonstrated superior performance fall under Case-I, and based on the lower quantile of 0.1, for EURO/LKR with a mean absolute error (MAE) of 1.3246 and mean absolute percentage error (MAPE) of 0.0058, and for GBP/LKR with the minimum errors of MAE of 1.2253 and MAPE of 0.0045. For USD/LKR, the QR model fitted with the 0.5 quantile demonstrated the lowest errors with MAE of 1.1369 and MAPE of 0.0057. These findings hold significance as forecasts of exchange rates play an important role in financial decision-making processes.