Nowcasting of Key Macroeconomic Indicators of Russian Economy: Approach Comparison

The paper is devoted to assessing the accuracy of scientists for the main macroeconomic indicators of the Russian economy: GDP according to the methods of calculating "expenditure" and "income", the GDP deflator index and GDP components according to the methods of calculating "expenditure" and "income". We consider models that can predict less operational low-frequency (quarterly) data based on high-frequency data (monthly). The quality of the predictions tested on 10 and 20 points for a wide range of models is compared: different specifications of MIDAS-models (models with mixed data), Mixed-Frequency Bayesian VAR with Minnesota prior on lags from 3 to 8 and linear models with regularization in two forms: without selection and with selection of variables. As a set of explanatory variables, we use a set of indices describing the components of GDP for production, as well as the main statistical indicators: unemployment, consumer price index, Brent crude oil price and official US dollar-Russian ruble exchange rate. The paper shows the effectiveness of MFBVAR models and their ability to build monthly science scientists for quarterly dependent variables. An accuracy assessment is also given for each of the models considered.