Predicting the Country Commodity Imports Using Mixed Frequency Data Sampling (MIDAS) Model

Predicting the amount of country imports toward assessing trade balance and its effect on the balance of payments (BOP) and finally money supply, general level of prices and the rate of economic growth is of paramount importance. Therefore, economic policymakers seriously need a model which cannot only predict the volume of imports well but also be capable of revising the initial prediction over time as soon as new data for the explanatory variables are available. To this purpose, mixed frequency data sampling model was used which allows time series variables with different annual, seasonal and even daily frequencies to be used in a single regression model. In estimating the model using the software R, annual real imports, real exports and quarterly of real GDP, real exchange rate and the volatilities of the real exchange rate in the range of 1988 to 2014 are used. Information related to 2014 is not used in preliminary estimation of relationship, so that the predictive power of the model outside of the estimated range can be tested. The proposed model predicts that real imports of goods as49948 million dollars for 2014 which is associated with an error of only41 million dollars, or about 8 percent, compared to its real amount achieved of49907 million dollars. The result suggests that the predictive power of the MIDAS model is very satisfactory.