Evaluating the Predictive Ability of Seasonal Autoregressive Integrated Moving Average (SARIMA) Models When Applied to Food and Beverages Price Index in Kenya

Abstract

Price instability has been a major concern in most economies. Kenya's commodity markets have been characterized by high price volatility affecting investment and consumer behaviour due to uncertainty on future prices. Therefore, precise forecasting models can help consumers plan for their expenditure and government policymakers formulate price control measures. Due to the seasonality of Kenya's food and beverage price indices, the current study postulates that the Seasonal Autoregressive Integrated Moving Average (SARIMA) model can best be the best fit model for the data. The study used secondary data on Kenya's monthly food and beverage prices index from January 1991 to February 2020 to examine the predictive ability of the possible SARIMA models based on the minimization of the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). A first-order differenced SARIMA (1,1,1) (0,1,1)12 minimized these model evaluation criteria (AIC = 1818.15, BIC =1833.40). The cross- validation test results of 6, 12, 18, 24, 30, and 36 step-ahead forecasts demonstrated that SARIMA models are unstable for use in forecasting over a long-time period with a tendency of increasing prediction errors with an increase in the forecast period. It is anticipated that the findings of the current study will provide necessary valuable information to the policymakers and stakeholders to understand future trends in commodity price.