A STUDY OF ARIMA AND GARCH MODELS TO FORECAST CRUDE PALM OIL (CPO) EXPORT IN INDONESIA

Indonesia is the biggest exporter Crude Palm Oil (CPO) country in the world. In 2013, export of Indonesian CPO reach $ 15.84 billion. CPO is important for various things, such as cooking oil, vegetable fat for milk and ice cream, raw materials of soap or cosmetics industry, or alternative fuels. The export of CPO is very important for the national economic development since it contributed significantly to the trade balance of Indonesia. Therefore, in order to support the government in making the right policy, forecasting export of CPO is necessary. In this study, a series of CPO export data was analyzed using time series analysis and forecasting techniques. Two classes of models were investigated, namely ARIMA models and GARCH models, and the comparison of the models was reviewed. The data used in this study were monthly export registration data, available at various ports for export across Indonesia from 1996 to 2013. The results showed that the first model was suitable for time series data with constant variance and hence if the problem of variance heterogeneity existed in the data, then a transformation to the data was required. The second model, on the other hand, was suitable for time series data having serious problems with heterogeneity of variance. However, in this study it was shown that a simple transformation of ARIMA models has fitted the data properly and could overcome the heterogeneity of variance problems. Key words: CPO, ARIMA, log transformation, GARCH, forecastin