Application of MF-PVAR Model for Nowcasting Gross Regional Products in Russia

Abstract

The significant lags in Rosstat’s publication of estimates of annual gross regional product (GRP) severely constrain the timely analysis of economic dynamics across the regions of Russia, heightening the demand for nowcasting techniques. This study uses a mixed-frequency panel vector autoregression (MF-PVAR) model for GRP nowcasting that integrates heterogeneous data and accounts for spatial heterogeneity and cross-sectional dependence among the regions. The sample used is a balanced panel of 68 Russian regions covering 2010–2022 and includes annual growth rates of real GRP together with monthly growth rates of sectoral indicators. The model is estimated over 2010–2018 and validated on 2019–2022 data. The results show that the precision of the MF-PVAR forecast is higher compared to the naive forecast, Ridge regression, and dynamic panel models based on generalised method of moments.