Mixed-frequency data models and their application to real-time analysis and forecasting

Abstract

This paper presents monthly mixed-frequency models MIDAS, MIDAS-GETS and MF-VAR describing the dependence of the producer price index in industry of the Republic of Belarus on the Belarusian consumer price index and producer price index in industry of the Russian Federation. Daily growth rates of the Belarusian ruble against the US dollar and the Russian ruble are used as real-time data. The models demonstrate significantly higher accuracy of out-of-sample short-term forecasts compared to ARX and VARX models with the similar exogenous structure and average monthly exchange rates. The causal analysis based on the MF-VAR and VARX models allows us to conclude that the Russian index of industrial producer prices has a leading influence on the Belarusian indices