Estimating and Forecasting Meat Prices in Pakistan: A Comparative Study of ARIMA, GARCH and State Space ARIMA Models

Abstract

Forecasting plays essential role in making effective planning and decisions for a gainful business. Modeling monthly price series containing nonstationarity, seasonality, lag dependence, heteroscedasticity and structural changes is challenging. This leads to the applications of modeling and forecasting technique alternative to widely used conventional ARIMA (Autoregressive Integrated Moving Average) models and GARCH (generalized autoregressive conditional heteroscedastic) models to accommodate all these factors and capture the dynamics of the system. State space ARIMA models through Kalman Filter seem to be the appropriate candidates for this purpose. The main aim of this study is to investigate the worth of state space models in forecasting monthly meat prices in Pakistan under both homoscedasticity and heteroscedasticity. This study investigates a comparison of state space ARIMA model and ARIMA-GARCH models in the presence of heteroscedasticity and with simple ARIMA models after adjusting the heteroscedasticity through transformation. The empirical evidences are generated by applying these models to five monthly meat price series: chicken, mutton, beef, fish and shrimp in Pakistan. On the basis of the empirical results it is concluded that the state space ARIMA models outperform the ARIMA-GARCH models in the presence of conditional heteroscedasticity and simple ARIMA models in case of homoscedasticity.