Nowcasting of GRP of Russian Regions

In the paper the nowcast system of low-frequency variable, such as Gross Regional Product (yearly data), is constructed for 76 regions of Russia. The nowcast is performed on the base of high-frequency variables (quarterly data) using three models, build for work with such types of data – DFM (Dynamic Factor Model), MIDAS (Mixed Data Sampling), MFBVAR (Mixed Frequency Vector Autoregression) with Minnesota prior. The main purpose of the work is to estimate the accuracy of the models and compare their predictive power. For the first two models nowcast is made on two points (two years). To use the last model the method of linear interpolation is used to transform yearly data into quarterly data. After that, the nowcast is made on 8 points (8 quarters, or 2 years). Due to the features of MFBVAR, only DFM and MIDAS are comparable between themselves. Main economic indicators of Russian regions are used as explanatory variables in the models. As a result, the accuracy of the predictions is quite high, even though there is limited amount of information available for the dependent variable and the explanatory variables.