Abstract:
Recently, modelling and forecasting of high frequency data (such as daily commodity price
volatility) using GARCH-MIDAS component attracts the attention of many researchers.
Following the same line, the objective of the present study is to apply Multiplicative GARCH MIDAS two component model for selected daily ECX commodities (Harar coffee, Jimma
coffee and sesame) price return volatility over the period of 4-11-2009 to 30-12-2016. The
GARCH-MIDAS component model decomposes the conditional variance as short run
component which follows a mean-reverting GARCH(1,1) process and long run component
which considers different frequencies of macroeconomic variables (in this study REER, fuel
oil price and inflation rate) via Mixed Interval Data Sampling (MIDAS) specification using
beta weight function. The results of ARCH effect tests on the residuals from the mean models
revealed the existence of time varying conditional variance. Among the conditional variance
models, GARCH (1, 1), GARCH (2, 1) and GARCH (1, 2) were identified for Harar coffee,
Jimma coffee and sesame price return volatility, respectively. Engle and Ng tests show the
insignificance of the asymmetric term, while Lundbergh & Terasvirta LM and the Li-Mak
portmanteau tests from the residuals of GARCH models show the existence of time varying
unconditional variance. From the result of the estimated GARCH-MIDAS component models,
REER, fuel oil price and inflation rate were found to be the best drivers of Harar coffee,
Jimma coffee and sesame price return volatility, respectively. Moreover, the estimated
GARCH-MIDAS component models were used for out-of-sample forecasts by incorporating
relevant macroeconomic variables. Finally, the MSE, MAE and DM test were used for
evaluating and comparing the forecasting ability of GARCH-MIDAS component models
against standard GARCH models. The results show that including low-frequency
macroeconomic variables improves the forecasting ability of volatility component models.
