Study on Forecasting Soybean Production: An Application of ARIMA Model

This paper applies econometric modeling (Univariate Auto-Regressive Integrated Average Model) in forecasting soybean yield in Zambia. The time-series data for soybean yield analysis ranges from 1973 to 2016, collected from the FAO statistics and deliberated on the selected reliant model. In this analysis we employ a two-step Box-Jenkins method introduced by the statisticians George Box and Gwilym Jenkins. The sole purpose is to find the best fit of a time series model to preceding values of Soybean output. Additionally, in assessing the best fit of our model, we used Auto-Correlation, Partial Auto-Correlation, Regulated Bayesian Information Criterion, Dickey-Fuller test statistic, Ljung-Box-Pierce Q statistics as well as the projected residuals. Thus, using the ARIMA (p, d, q) model, the extrapolated statistics exposed that ARIMA (2, 2, 1) was the best model for forecasting soybean output. The results illustrate a cumulative trend in production with the projected value of 6611.7hg/ha in 2026. Thus, based on the selected model, it can be anticipated that in Zambia soybean output in 2020 would increase to 23430.3 hg/ha in comparison to 19624 hg/ha in 2016. Further, the study estimates a stable growth in the soybean yield in impending farming seasons.