Prices of fruits and vegetables do not only fluctuate between years but also vary considerably between seasons. The purpose of this study is to investigate the patterns (characteristics) of price variation over time and find appropriate models for forecasting future prices.
The study deals with prices of four vegetables, two fruits, and four fruit-bearing vegetables. The study tests stationarity of price series and analyze seasonal effects in them. It applies mainly time-series methods - ARIMA, GARCH, VAR, and the transfer function approach - to model the movements of produce prices.
The main results are summarized as follows:
First, since systematic differences between monthly produce prices are found, price data should be differentiated between monthly figures (seasonal differences) on a year-to-year basis. And Chusok (the harvest festival) is found to have a holiday effect on the prices of apples and pears, whose prices increase by ten to twenty percent.
Second, the coefficients of seasonal autoregressive terms are found to be significantly negative and less than one in absolute value in most models. This implies that prices fluctuate but converge to the equilibrium level following the cobweb theory.
Third, according to ex-post forecasting, the study finds a model whose forecast errors are less than 15 percent for most items. In most cases, the ARIMA/GARCH and transfer function models outperform VAR models.
Lastly, the inclusion of an exogenous variable (supply quantity) into a pure time-series model increases the explanative and forecast power. This implies that the incorporation of a regression method into the time-series method would be beneficial in modeling the price of produce.
Researchers: Yong-Sun Lee, Byung-Ok Choi, and Song-Bo Sim
E-mail Address: yslee@krei.re.kr; bochoi@krei.re.kr; simsb@krei.re.kr