@ARTICLE{26543120_800824520_2022, 
	author = {Natalya Makeeva and Ivan Stankevich}, 
	keywords = {, nowcasting, Russian GDP, time series, mixed frequency models, MIDAS modelsforecasting},
	title = {Nowcasting of the Components of Russian GDP},
	journal = {HSE Economic Journal },
	year = {2022},
	volume = {26},
	number = {4},
	pages = {598-622},
	url = {https://ej.hse.ru/en/2022-26-4/800824520.html},
	publisher = {},
	abstract = {The paper discusses the problem of nowcasting the current growth rates of Russian GDP and its components using quarterly data. The quality of restricted and unrestricted MIDAS models (models with mixed data), MIDAS model with L1 regularisation and MFBVAR model (Bayesian vector autoregression of mixed frequency) are compared. The results are compared with classical autoregression to justify the need to use nowcasting models for the rapid assessment of macroeconomic indicators. Production indices for various industries and macro indicators characterising Russian GDP and its components were used as explanatory variables. The paper proposes a way to quickly assess the current state of the economy and proposes a nowcasting method based on data only for the first or first two months of the quarter under consideration. As a result, for each dependent variable, the best model for building a nowcast based on the last 12 points is selected based on the criterion of mean absolute error (MAE) and root mean square prediction error (RMSE).},
	annote = {The paper discusses the problem of nowcasting the current growth rates of Russian GDP and its components using quarterly data. The quality of restricted and unrestricted MIDAS models (models with mixed data), MIDAS model with L1 regularisation and MFBVAR model (Bayesian vector autoregression of mixed frequency) are compared. The results are compared with classical autoregression to justify the need to use nowcasting models for the rapid assessment of macroeconomic indicators. Production indices for various industries and macro indicators characterising Russian GDP and its components were used as explanatory variables. The paper proposes a way to quickly assess the current state of the economy and proposes a nowcasting method based on data only for the first or first two months of the quarter under consideration. As a result, for each dependent variable, the best model for building a nowcast based on the last 12 points is selected based on the criterion of mean absolute error (MAE) and root mean square prediction error (RMSE).}
}