Modelling Open Market Retail Price of Red Onions in Colombo using ARIMA-GARCH Mixed Model

Abstract

Red onion is an important commodity in Sri Lankan culture which subject to high frequent fluctuations in retail price due to government policies, trade agreements and weather conditions like heavy rainfall. Objective of this study is to find more accurate time series model to forecast future prices of red onions. This study considers weekly average retail prices (WARP) of red onions in Colombo main markets from January 2014 to April 2019. Several models were fitted and based on model selection criterions, ARIMA(1, 1, 1) was identified as the best model. As the residuals of the model were heteroscedastic, ARCH(9) and GARCH(9,1) models were fitted. According to the literature, WARP of red onion price shows drastic increase in 2017 as a result of production fall. Thus, using change point analysis, series was divided into 3 windows and ARIMA(0,1,0) model was suggested as the best model for each window. Finally using all four models; ARIMA(1,1,1),ARIMA(1,1,1)+ARCH(9), ARIMA(1,1,1)+GARCH(9,1)and ARIMA(0,1,0) price was forecasted for the year 2019 using two methods; static and dynamic forecasting. Forecasting accuracy of the models measured using root mean squared error (RMSE). As a conclusion ARIMA(1,1,1)–GARCH(9,1) model was chosen as the most suitable model with 6.26 RMSE to forecast WARP of red onions. Though there are several studies carried out on behaviour of red onion price in Sri Lanka, no specific model was suggested so far. Therefore this model can be used by the cultivators, intermediaries and government in decision making on production quantity, pricing and import/export regulations