Structural Breaks, Automatic Model Selection and Forecasting Wheat and Rice Prices for Pakistan

Abstract

Structural breaks and existence of outliers in time series variables results in misleading forecasts. We forecast wheat and rice prices by capturing the exogenous breaks and outliers using Automatic modeling. The procedure identifies the outliers as the observations with large residuals. The suggested model is compared on the basis of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) with the usual ARIMA model selected ignoring the possible breaks. Our results strongly support that forecasting with breaks by using General to Specific (Gets through Autometric) model performs better in forecasting than that of traditional model. We have used wheat and rice price data (two main staple foods) for Pakistan.