Seasonal Variations and Prediction of Annual Non-Oil Export: A Mixed Data Sampling Approach

The aim of this paper is to specify and estimate such a model for non-oil export that allows to revise the previous predicted volume of annual non-oil export, as soon as new seasonal data for the explanatory variables are released. This cannot be achieved unless a model is constructed in such a way that annual non-oil export is set as a function of some seasonal explanatory variables. By utilizing a very recent method of regression specification, namely MIDAS regression, this article sets annual non-oil export as a function of seasonal real GDP, seasonal real exchange rate and seasonal real exchange rate fluctuations. To provide more accurate predictions, as compared to traditional methods, this regression model is capable of providing a revised prediction as soon as new information concerning explanatory variables are released. The specified regression model, which is estimated by using time series data within the period 1989 – 2014, predicts the real amount of annual non-oil export for 2015 with a small error of almost 2%. So, this prediction is considered to be very close to the reality.