Modeling and Forecasting Cereal Price in Ethiopia: an Application of ARIMA and GARCH Models
No Thumbnail Available
Date
2013-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Cereal production and marketing constitute the sing le largest sub- sector in Ethiopian eco nomy. It
acco unts fo r ro ughly 60 percent of rural emp loyment, 73% of total culti vated land and 68.3% of
total o utput, 46 percent of a typ ical ho usehold's food expenditure more than 60% of caloric/
intake. According to ava ilable est imates, cereals production represe nts about 30 percent of gross
domestic product (GOP). In this study we attempted to model cereal price and obtain forecasts at
nat io nal leve l. The data used are mo nt hly cereal price obtained from the Centra l Stat istical
Agency (CSA) fo r the periods from September 1996 to July 201 2.
Seasonal ARIMA and GARCH were employed to ana lyze the mo nthly cerea l price data. It was
found that the Seasona l ARlMA(O, I, I)*(O, I, I) and ARMA(2, 1)-GARCH(I , I) were adequate
models for the data considered in this stud y. In the GARCH mode l, the va lue of the GARCH
term for the return of cereal price is close to one indicating slow convergence of vo latility to a
steady state and high persistence in vo latility. In additio n, the constant term in the mean equation
was significant and thus it fo llows an ARMA (2, I) model. The po int forecast results showed a
very clo ser match with the pattern of the actual data and better forecasting accuracy in validation
period. Almost all the in-sample forecast eva luations statistic indicated that the Seaso nal ARIMA
mode l is better in comparison to GARCH Model. However, almost all the out -sample forecast
eva luation stati stic shows the superiorit y of GARCH ( I, I) mode l over the Seaso nal ARIMA
model.