Forecasting Quarterly Russian GDP Growth with Mixed-Frequency Data

Abstract

This paper presents a pseudo real‐time out‐of‐sample forecast exercise for short‐term forecasting and nowcasting quarterly Russian GDP growth with mixed‐frequency data. We employ a large set of indicators and study their predictive power for different subperiods within the forecast evaluation period 2008–2016. Four indicators consistently figure in the list of top-performing indicators: the Rosstat key sector economic output index, the OECD composite leading indicator for Russia, household banking deposits, and money supply M2. Aside from these indicators, the top indicators in the 2008–2011 evaluation period are traditional real‐sector variables, while those in the 2012–2016 evaluation period largely comprise monetary, banking sector and financial market variables. We also compare the forecast accuracy of three different mixed‐frequency forecasting model classes (bridge equations, MIDAS models, and U-MIDAS models). Differences between the performance of model classes are generally small, but for the 2008–2011 period MIDAS models and U-MIDAS models outperform bridge equation models.