Forecasting retail prices of the most commonly used rice in Sri Lanka

Abstract

This thesis focuses on Modeling and forecasting Maximum Retail Price (MRP) of Samba, Nadu, Kekulu White and Kekulu Red rice in Sri Lanka using Univariate and Multivariate Time Series approaches. Sri Lanka is a developing country with population of 21.4 million as estimated in 2020. Rice is the most commonly used food in Sri Lanka. Thus, the finding a model for the forecasting prices is most economical advantage for Sri Lankan government. The fluctuations of the prices of rice making a great risk of investing, buffer stock maintaining, international trade and other associated actions. Thus, it is vital to forecast future prices for decision making purposes. Our objective is to forecast the average weekly prices of selected four products. In this study, we consider weekly average retail prices of Samba, Nadu, Kekulu White and Kekulu Red from September 2017 to March 2019. Thus, each series consists of 93 data points. The missing values are estimated using expectation maximization algorithm. Data is collecting from Central Bank of Sri Lanka. First 83 data points are used to build the model and remaining 10 data points are used to validate the forecasting model. To select the best model, selection criteria based on the Akaike information criterion (AIC). We observe that the best model for the Samba prices is exponential smoothing. Nadu price is ARIMA (2,1,0) and best model for Kekulu White and Kekulu Red are ARIMA (1,1,0) and ARIMA (1,1,0) respectively. Then, the testing data set is used to validate the prediction. Since there is a strong correlation between prices, we consider vector auto regression (VAR) model to improve the forecasts. Among several plausible models VAR of order 2 results in the best model. Nadu prices is independent of other three prices. VAR models provides better forecasts for prices of Nadu, Kekulu White, Kekulu Red.