Time Series Analysis of Soybeans Prices in Pakistan Applying Symmetric and Asymmetric GARCH Models with Normal and Non-Normal Innovations

Abstract

Forecasting and volatility modeling are important tools for all agricultural and financial sectors. The core aim of this study is to compare the performance of symmetric and asymmetric Generalized Autoregressive Conditional Heteroscedasticity (GARCH) models under both normal and non-normal error distributions to forecast soybean daily prices in Pakistan. The ARIMA models are applied as mean models along with four GARCH models (GARCH, EGARCH, TGARCH and PARCH) with three error distributions (Normal, Student-t and GED). The ACF and PACF of residual and squared residuals are used as diagnostics to check the appropriateness of the models. Based on the empirical findings it is concluded that improvement in the overall estimation is achieved using asymmetric GARCH models as the conditional variance. Moreover, TGARCH model with student-t distribution outperforms the other models in forecasting soybean prices in Pakistan. These results provide valuable insights for stakeholders and policymakers in choosing a suitable forecasting model for soybean prices in Pakistan. The government should take steps to develop high-yielding latest production technology and strategies to increase the production of soybeans.