Analyzing and forecasting of tea price at the Colombo tea auction using financial time series

Abstract

Tea industry plays major role in national economy of Sri Lanka and it is the world’s fourth-largest tea producer. Tea auction market in Sri Lanka is mature and therefore, the leading tea producing countries such as China, India, Indonesia and Kenya concern about Sri Lanka. Sri Lankan tea is sold through the Colombo tea auction. This study conducted in order to analyze and model the weekly tea auction prices at the Colombo tea auction based on time series analysis. The secondary data of weekly tea auction prices over the period 2015 to 2021 which collected from official website of Sri Lanka Tea Board used for this study. The tea dataset is specific to the black tea prices. The analysis is done by using R Programming. This paper employs asymmetric GARCH model to investigate the volatility and model the volatility on the tea prices at the Colombo tea auction. First of all, descriptive analysis was performed using seasonal decomposition, revealing that the data exhibits only a trend pattern without any seasonal or cyclical components. Then, the stationarity is checked by using ADF test and conditional mean model AR (1) model is fitted to differenced series. After checking residuals of the AR(1) model, conditional heteroscedasticity in residuals of fitted AR (1) model is detected using ARCH LM test. Under this effect, conditional variance model GARCH (1, 1) built to forecast the tea prices. Finally, the AR(1) - APARCH(1,1) is selected as the best for modeling tea prices by comparing information criteria (AIC, BIC and HQIC) and checking parameters and residual of the model. The accuracy of the model is measured using MAPE. MAPE of this model is 13.79% . This model has capable to capture the volatility on the weekly tea auction prices at the Colombo tea auction.