Real-time assessment of regional economic activity in Russia using dynamic factor models

Official estimates of the gross regional product (GRP) in Russia are published annually and with a significant lag, which critically limits the analysis of regional heterogeneity and the effects of macroeconomic policy at business-cycle frequencies. This paper constructs real-time indicators of economic activity for 79 constituent entities of the Russian Federation at a monthly frequency. The indicators are estimated with a Bayesian dynamic factor model that combines low-frequency GRP data with available monthly and quarterly statistics characterizing the production, income and expenditure sides of the regional economies. The resulting indicator (“proxy GRP”) successfully identifies key macroeconomic episodes, including the 2014—2016 commodity-market downturn, the COVID-19 pandemic shock, and the structural transformation of the economy after 2022. The aggregated proxy GRP is highly correlated with the Rosstat index of output of basic industries and the composite index of regional economic activity (REA) proposed by Smirnov et al. (2017). In a pseudo out-of-sample exercise, the forecasts and nowcasts of GRP growth produced by the model are on average 50% more accurate than a naive random walk forecast and 30% more accurate than estimates based on the index of output of basic industries. The resulting estimates can be used for real-time monitoring of regional economies and for analyzing the transmission mechanisms of monetary and fiscal policy.