Authors: William Lubowa; James Wokadala; Tom Makumbi Nyanzi
Addresses: Budget Office, Parliament of Uganda, P.O. Box 7178, Kampala, Uganda ' College of Business and Management Sciences, P.O. Box 7062, Kampala, Uganda ' College of Business and Management Sciences, P.O. Box 7062, Kampala, Uganda
Abstract: The study was set to fit appropriate auto-regressive integrated moving average (ARIMA) and vector auto-regressive (VAR) models for forecasting Uganda's core inflation; to compare their forecasting capabilities and establish whether the forecasts produced were significantly different. The study employed Box and Jenkins (1976) and Sims (1980) modelling techniques on monthly time series data for core inflation index, broad money supply (M2), nominal official exchange rate and short-term 91-day Treasury bill rates for the period July 2005 to December 2014. The appropriate ARIMA model was evaluated and found to be ARIMA(3, 1, 3). Pairwise Granger causality test confirmed that money supply, nominal Ushs/USD exchange rate and short-term Treasury bill rate caused core inflation. Thus, the appropriate restricted VAR (VECM) model was estimated and found to produce more accurate core inflation forecast for longer time horizons while ARIMA was better for shorter forecast horizons. However, the forecasts from both models were not significantly different.
Keywords: inflation forecasting; auto-regressive integrated moving average; ARIMA; vector auto-regressive; VAR; Uganda; modelling.
International Journal of Sustainable Economy, 2017 Vol.9 No.2, pp.121 - 141
Received: 16 Jan 2016
Accepted: 16 Jul 2016
Published online: 24 Mar 2017 *