Empirical economic modelling in Russia is a complicated task. One of the most important limitations comes from the insufficiently long time series that make estimation of a comprehensive econometric model virtually impossible. Researchers therefore have to rely on parsimonious model specifications in their work. One example is traditional macroeconometric models (e.g. Benedictow et al. (2013)) consisting of a large number of pre-specified simultaneous equations. As regards a more flexible vector autoregression (VAR) approach, a typical model for Russia would comprise an ad-hoc selection of variables (often no more than five indicators in total) that either represents a theoretical long-term macroeconomic relationship (Korhonen and Mehrotra (2010), Mehrotra and Ponomarenko (2010)), or is sufficient to identify predetermined types of economic shocks (via a structural identification scheme (Korhonen and Mehrotra (2009)) or sign restrictions on impulse response functions (Granville and Mallick (2010), Mallick and Sousa (2013))), or simply comprises the indicators that are assumed to be the most important determi-nants of the modelled process (Rautava (2013)).