Forecasting of Monthly Naira-Dollar Exchange Rate Series: Seasonal or Non-seasonal Univariate Time Series Model?

Abstract

Over the past eight decades, the United States Dollar has become the vehicle currency driving other major foreign currencies as well as minor currencies emerging in the financial markets. Consequently, the goal of developed and developing countries is to maintain good exchange rates with the USD through effective monetary policy management. To assess the monthly Naira-Dollar exchange rates (EXR) relationship, this study applied the non-seasonal Autoregressive Integrated Moving Average (ARIMA) and Seasonal ARIMA (SARIMA) models to examine the forecasting dynamics of the series. The pre-test results confirmed that EXR is a stationary difference process of order one {I(1)}, contains seasonality. SARIMA (0, 1, 1)(0, 1, 1)₁₂ was observed to outperform its ARIMA (3, 1, 0) counterpart. The preference of the Akaike Information Criterion (AIC) in selecting the model over Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) was highlighted. A diagnostic check was carried out on the identified model and it revealed that the residual of the model is a white noise since it is homoscedastic, stationary, and non-autocorrelated. Findings from this study established that the Naira is projected to continue to fluctuate and further lose value relative to the USD within the forecasted time frame.   Based on these findings, both the Government and policymakers need to formulate policies that will further enhance investment incentives, currency stabilization, and local production support. The practical implications of the forecast results on the country’s economy include rising costs of living, higher inflation, reduced foreign direct investment, job losses, worsening trade deficit, and skilled worker migration.