Andrey Polbin: Standard methods of co - integration analysis between the GDP and oil prices seem to be non-applicable.

On April 10, Andrey Polbin, Head of Macroeconomic Modeling Department, delivered a presentation at the XX April NRU HSE Conference: “Building – up and assessment of the model of unobservable factor for the Russian Federation amid high dependence on oil prices”.
Central banks rely on formal economic and mathematical models in modern reality as well as on a wide range of indicators of economic activity for developing measures of monetary/credit policy, and one of the most important indicators is the output gap or its cyclic component interpreted as deviation of the actual level of output from its potential value. As a rule, when GDP is decomposed into trend and cyclic components, dependence of domestic production on trading conditions and oil prices is not taken into consideration. Russia has its place among leading exporters of hydrocarbons and that is why high dependence of domestic macroeconomic indicators on oil prices is unlikely to be doubted.

By now, academic literature has offered quite a lot of assessments of cyclical component of the output of the Russian economy. In the study presented by A. Polbin, he proposed an extension of the model of unobservable components for decomposition of Russia's GDP into trend, cyclic and seasonal components. It was suggested that level of potential output of the domestic economy depends on terms of trade (oil prices are used as proxies) while cyclical component of the output is correlated with shocks of trading terms, and the long-term growth rates of potential GDP are described by a random volatility. Availability of the latter element leads to the fact that time series of GDP turns out to be of the second order of integration. Accordingly, standard methods of co-integration analysis between GDP and oil prices are not applicable.

Expert underlined that abovementioned study allowed to identify a significant slowdown of long-term growth, confirm the relevance of  model for forecasting and outline further areas of research: evaluation of alternative specifications, analysis of the model's predictive properties, consideration of structural changes and multidimensional expansion of the model.

Presentation to the report


Friday, 19.04.2019