Estimating Price Volatility Structure in Iran's Meat Market: Application of General GARCH Models

Over the past few years, the price volatility of agricultural products and food markets has attracted attention of many researchers and policy makers. This growing attention was started from the food price crisis in 2007 and 2008 when major agricultural products faced accelerated price increases and then rapidly decreased. This paper focused on the price volatility of major commodities related to three market levels of Iran’s meat market, including hay (the input level), calf and sheep (the wholesale level) and beef and mutton (the retail level). In particular, efforts will made to find more appropriate models for explaining the behavior of volatility of the return series and to identify which return series are more volatile. The effects of good and bad news on the volatility of prices in each return series will also be studied.
Different GARCH type models have been considered the best for modeling volatility of return series. Nonlinear GARCH models were introduced to capture the effect of good and bad news separately. The paper uses some GARCH type models including GARCH, Exponential GARCH (EGARCH), GJR-GARCH, Threshold GARCH (TGARCH), Simple Asymmetric GARCH (SAGARCH), Power GARCH (PGARCH), Non-linear GARCH (NGARCH), Asymmetric Power GARCH (APGARCH) and Non-linear Power GARCH (NPGARCH) to model the volatility of hay, calf, sheep, beef and mutton return series. The data on hay, calf, sheep, and beef and mutton monthly prices are published by Iran’s livestock support firm. The paper uses monthly data over the sample period of the May 1992 to the March 2014.
Descriptive statistics of the studied return series show evidence of skewness and kurtosis. The results here show that all the series has fat tails. The significant p-values for the Ljung-Box Q-statistics mean that the auto-correlation exists in the squared residuals. The presence of unit roots in the return series is confirmed by the results of the ADF and PP unit root tests. Different GARCH type models mentioned in materials and method were fitted to the return series and then have been compared based on 7 loss functions MSE_2, MSE_1, PSE, QLIKE, R2LOG, MAD_2, MAD_1, two information criteria AIC and BIC and log likelihood. The selected models for modeling the behavior of volatility in the hay, calf, sheep, beef and mutton return series are SAGARCH (1,1) with a t distribution, NGARCH (1,1), TGARCH (1,1), SAGARCH (1,1) and EGARCH (1,1) all with Gaussian distribution. The coefficient of asymmetry (γ) in all models shows signs of asymmetric behavior in volatilities so that for all of the return series except hay returns positive shocks have more effect on volatility relative than negative shocks of the same size. This evidence is vice versa for the hay return, in which negative shocks have more effect on volatility. The (α β) in all models are greater than 0.7 which means the high persistence of shocks to volatilities. In other words, shocks might die out very gradually. This feature is more pronounced in the case of beef and calf return series with α β greater than 0.9. Sensitivity of the current volatility to the new shock or news, α, in calf (0.76) and beef (0.71) returns are greater than that of others. The low sensitivity to the news is related to the sheep returns (0.16). The effect of current conditional variance for the next month conditional variance, β, in sheep (0.55) and mutton (0.42) returns are relatively high. Minimal β (0.14) is related to the calf returns.
The paper attempts to study persist shocks to volatility as well as how positive (good) or negative (bad) shocks (news) may have an asymmetric effect on the volatility of a return series of hay, calf, sheep, beef and mutton prices in Iran. The findings show signs of asymmetry and persistence in volatilities. The sensitivities of price were also, volatility to the news in the calf and beef markets is greater than other return series. By the way, the effect of current conditional variance of the next month conditional variance in sheep and mutton returns is greater than others. This finding indicates that when new shocks occurs in the meat market calf and beef returns are more influenced by them and sheep and mutton returns highly transmit the current volatility in the future. This suggests less political tensions in the country as much as possible to calm the economic and political space.