Improving Ethiopia coffee productivity through price volatility forecasting

Purpose
The general objective of this study is to model the future likely coffee prices paths to help producers make informed investments decisions and effectively manage their price risk exposure. An inherent price risk otherwise known as volatility in the coffee market possess serious challenges to more than 4 million smallholder farmers (hereafter referred to as producers) and approximately 15 million Ethiopians who directly and indirectly depend on coffee production as a means of livelihood.

Design/methodology/approach
We evaluate the forecast ability of the general autoregressive conditional heteroscedasticity (GARCH) and autoregressive conditional heteroscedasticity (ARCH) in terms of their ability to forecast seven-month-ahead conditional variance. The performance of the GARCH (1, 1) model is compared with that of the ARCH (1). We compare the models using monthly coffee spanning between the periods January 1976 to July 2018. The models are evaluated out-of-sample using five different loss functions namely the Mean Error (푀퐸), the Mean Squared Error (푀푆퐸), the Mean Absolute Error (푀퐴퐸), the Root Mean Squared Error (푅푀푆퐸) and the Standard Error form the regression model(푆퐸푅). Forecasting of primary agricultural commodities have been in previous studies to model rice yield and price forecasting (Shabri et al., 2009), price forecasting and milk productivity (Novotorv & Brikach, 2016), modelling and forecasting of edible oil price and cotton price using GARCH, EGARCH and ARIMA (Lama et al., 2015), improving rice productivity and crop rotation in Thailand (Ruekkaseem & Sasananan, 2018) hence, this study adopted similar research approach.

Findings
The result shows the persistence of volatility and takes a long time to die off and no remarkable difference between the predictive accuracies of both models. Using the standard error of the regression, it appears that the GARCH model might produce more robust forecasts than the ARCH model and can capture future price volatility.

Research limitations/Implications
The methodology employed in this paper can also be replicated in forecasting other agricultural commodities that exhibit similar volatility.

Practical Implications
Comparing the forecasted price volatility and actual prices indicates that producers can effectively manage price volatility by relying on forecasted volatility price to hedge price risk and plan coffee production.

Social Implications
Effective management of price volatility will improve producer’s returns, reduce poverty and contribute to coffee production sustainability.

Originality/Value of a paper
Contributing to the extant literature on coffee volatility in Ethiopia were gap was identified in forecasting future coffee price volatility.