Inflation Forecasting Models and Forecasting Combination Analysis: The Case of Ethiopia
Abstract
The main objective of this study is to compare different inflation forecasting models and combinations techniques that best fit for Ethiopian inflation forecasting. In particular, the random walk model, ARIMA, ECM, VECM, Phillips curve and BVAR model was employed. Since Ethiopian CPI data does not follow random walk process using statistical analysis Augmented Dickey-Fuller test it was excluded in forecast performance evaluation and forecasting combination analysis. Therefore, in model comparison only five models have been compared using RMSE for both in-sample and pseudo out of sample forecasting. The empirical finding shows that, using both in-sample and pseudo out of sample forecast accuracy ARIMA model performs best than other models. Next to ARIMA model ECM and BVAR model performs best as compared to VECM and Phillips curve. On the other hand VECM performs worst than other models compared up to eight period ahead forecasts. In the study different forecast combination techniques were compared. From those forecasting combination techniques Winsorized Mean, Median and Trimmed Mean respectively performs best than Bats/Granger Method, Equal Weight and OLS. Compared to VECM model forecast combination leads best in a reduction of forecast error, although some of the individual models like ARIMA, ECM and BVAR perform better than forecast combinations.
Key words: Inflation, forecasting, forecast combination, ARIMA, BVAR and VECM, Forecast Evaluation;
DOI: 10.7176/JESD/14-15-03
Publication date:October 31st 2023
![Download the IISTE publication guideline!](https://www.iiste.org/sub/iiste_footer_journal.jpg)
To list your conference here. Please contact the administrator of this platform.
Paper submission email: JESD@iiste.org
ISSN (Paper)2222-1700 ISSN (Online)2222-2855
Please add our address "contact@iiste.org" into your email contact list.
This journal follows ISO 9001 management standard and licensed under a Creative Commons Attribution 3.0 License.
Copyright © www.iiste.org