| dc.description.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 | |
