Probabilistic Insights and Uncertainty in Advanced Time-Series Forecasting Models for Ginger Supply Response

Abstract

Despite the increasing demand and growth potential of Nigeria’s ginger production, the country has yet to fully harness the economic benefits of expanding value addition and market diversification. This study aims to advance ginger supply forecasting by employing cutting-edge time-series techniques to model and predict future ginger production while emphasizing credibility through uncertainty and probability estimates. Cointegration analysis and the Vector Error Correction Model (VECM) were utilized to model the supply response, while Exponential Smoothing (ETS), ARIMA, and Facebook Prophet models were employed to forecast ginger production over a 10-year horizon. The results revealed seven cointegrating relationships, with all predictors being highly significant. Notably, disease prevalence and input usage had the highest long-run elasticities, at -93.96 and 15.74, respectively, while inflation was significant in the short run. The forecast results indicate a high probability of increased production, albeit with greater uncertainty in the long term. Performance evaluation showed that the ARIMA (0,1,0) model outperformed other models with a Mean Absolute Scaled Error (MASE) of 0.77. These findings underscore the need for effective planning and strategies to manage ginger supply and enhance its production potential, thereby mitigating the risk of undersupply and market failure. Policymakers and stakeholders can utilize these forecasts to
develop short- and long-term strategies for optimizing ginger production.