Comparison of methods for mixed data sampling (MIDAS) regression models to forecast Indonesian GDP using agricultural exports

Indonesia is an agrarian country and has agricultural land and abundant resources. Therefore, it is expected that Indonesia can utilize existing natural resources to increase the export value of agricultural products, which will impact the value of the GDP. This can improve the welfare of Indonesian society in general, and farmers in particular. However, the data to forecast the growth of the Indonesian GDP using the export value of the agricultural sector has unequal frequency data. Therefore, a special regression model, namely, the Mixed Data Sampling (MIDAS) regression model introduced by Ghysels, Santa-Clara, and Valkanov (2004) is applied. The advantages of MIDAS, in addition to overcoming the problem of data with mixed frequency, is to minimize the number of estimated parameters and make the regression model simpler. A weighting function is used to reduce the number of parameters in the MIDAS regression. The weighting function can have a number of functional forms. Ghysels, Santa-Clara, and Valkanov suggest the Exponential Almon function and the Beta function, then compare their performance with the distributed lag model. This research proves that, based on the Root Mean Square Error, the MIDAS Beta regression model yields a better model estimation than either the MIDAS Exponential Almon or the distributed lag model in the case of forecasting the growth of the Indonesian GDP using the export value of the agricultural sector.