Corresponding author: Aizhan Bolatbayeva ( aizhan.bolatbayeva@nu.edu.kz ) © 2021 Nonprofit partnership “Voprosy Ekonomiki”.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BYNCND 4.0), which permits to copy and distribute the article for noncommercial purposes, provided that the article is not altered or modified and the original author and source are credited.
Citation:
Bolatbayeva A (2021) A multicountry macroeconometric model for the Eurasian Economic Union. Russian Journal of Economics 7(4): 354370. https://doi.org/10.32609/j.ruje.7.72368

This paper introduces a multicountry macroeconometric model for the Eurasian Economic Union (EAEU). The model consists of five singlecountry models of the union member states: Armenia, Belarus, Kazakhstan, Kyrgyzstan, and Russia. The purpose of the research is to explain the structural relationship between the economies, evaluate the impact of internal and external shocks, and analyze the transmission mechanism of shocks across countries. The singlecountry models are linked to each other by the equations of bilateral trade and bilateral exchange rate. We find that the model fits actual data on main macroeconomic indicators of the countries in a dynamic expost simulation over 2004–2018. We also evaluate the effect of world trade and monetary policy shocks on the economies of the EAEU member states.
Eurasian Economic Union, EAEU, structural macroeconomic model.
The role of regional economic integration strengthens due to the opportunities created from the establishment of common markets and the removal of trade barriers. As a result, it is important to analyze the depth of the economic interrelationship between member states of an economic union. Macroeconomic modeling of economic integration turns out to be one of the most useful tools for this purpose as it allows one to examine the structure of interrelated economies and the effects of internal and external shocks on policy decisions of all member states. The present paper builds a multicountry macroeconometric model for the Eurasian Economic Union (EAEU) which consists of the following member states: Armenia, Belarus, Kazakhstan, Kyrgyzstan, and Russia. The EAEU aims to ensure the freedom of movement of goods, services, capital and labor, as well as implement a joint policy for various sectors of the economy.^{1} The main purpose of the model we build for the union is to bring in greater clarity about the structural interrelationship among the EAEU members.
There has been an increasing interest in the literature on the properties of multicountry macroeconometric models due to the rising number of regional economic unions. One of the vital contributions in this field was made by
The multicountry models are widely used at central banks and research organizations. The Federal Reserve Board developed a MCM in the late 1970s as a system of macroeconometric models of the USA, FR Germany, Canada, Japan and the UK (
In this paper, we build a multicountry macroeconometric model based on the framework of five structural singlecountry macroeconometric models of Armenia, Belarus, Kazakhstan, Kyrgyzstan, and Russia. Each singlecountry model consists of the following blocks: aggregate supply, goods market, labor market, prices, financial markets, and the government sector. We use quarterly data starting from 2001 to 2018. In total, the multicountry model includes 357 equations and identities. The equations are estimated in the form of error correction model, ARIMA model and Tobit regression. The countries are linked through bilateral trade and bilateral exchange rate. The properties of the multicountry model are examined by conducting an expost simulation for the period from 2004 to 2018. The results indicate good performance of the model in fitting actual data on the main macroeconomic indicators of Kazakhstan and Russia. However, the simulation results are less accurate for the rest of the countries in the model. In addition, we use the model to evaluate the effect of various shocks on important macroeconomic variables across the member states of the union. The model provides us with a clearer picture on the propagation mechanisms of various shocks across the EAEU countries. We conduct two scenario analyses with world trade and monetary policy shocks. In the first scenario, we assume a 10% contraction in world trade and produce a dynamic simulation of the model. In the second scenario, we evaluate 2 percentage points cut of the key rate by the Central Bank of Russia. An analysis of the influence of the latter shock is important as Russia has the largest economy among the EAEU member states.
We find that macroeconomic indicators of the EAEU countries respond to the world trade shock in different ways, both in terms of magnitude and timing. We find that real GDP in Armenia falls below the baseline more significantly than in other countries, while the Russian economy is less affected by the shock. In the case of an expansionary monetary policy shock in Russia we find that it causes domestic output and inflation rate to temporarily increase. Nevertheless, the ruble’s appreciation and strong aggregate demand raise total imports resulting in lower real GDP growth. The monetary policy shock in Russia exhibits the largest effect on Belarus among all other countries which partially reflects a very high share of the Russian economy in the exports and imports of Belarus. Most macroeconomic indicators in Belarus experience a more prolonged return to the baseline level than the same indicators for other countries. At the same time, the results reveal a negligible effect of the shock on the economy of Kyrgyzstan.
The paper is organized as follows. The following section presents the data description and discusses adjustments that have been applied to the data. Section 3 outlines the structure of singlecountry models and discusses channels that link the economies of the member states. Section 4 examines the ability of the multicountry model to fit actual dynamics of the main macroeconomic indicators. Section 5 provides scenario analyses with world trade and monetary policy shocks. Section 6 makes concluding remarks.
The multicountry model is based on quarterly data from 2001 to 2018. In total, the data contains 405 variables for the five countries which include both countryspecific and foreign variables. The data for the EAEU members are collected from national statistics agencies, central banks and finance ministries of the member countries. All other variables are retrieved from the IMF and Bloomberg databases. There is only annual data available for some variables which we convert to quarterly frequency via extrapolation. Quarterly data on capital stock is not available for all countries during the sample. Therefore, we apply the Perpetual Inventory Method (PIM) to calculate suitable capital stock series. We have chosen the first quarter of 2010 as the base year for national income account variables in our data set. That is, the data on GDP and its expenditure components have been adjusted for each country to calculate them in 2010 prices. The approach used to adjust the data for Kazakhstan is presented in detail in
As in the twocountry model by
This section briefly describes the structure of a macroeconometric model for the EAEU. Singlecountry models are constructed with the same underlying principles. Each model is divided into six blocks: supply side, goods market, labor market, prices, financial market, and government sector (Table
Country i  Block  Country j  
Exogenous  Endogenous  Endogenous  Exogenous  
Working age population  Potential output  Supply  Potential output  Working age population  
Labor force  Labor force  
NAIRU  NAIRU  
TFP  TFP  
Government consumption  Private consumption  Demand 
Private consumption  Government consumption  
Oil price  Investment  Investment  Oil price  
World trade  Export to other countries  Export to other countries  World trade  
Export to country j  Export to country i  
Import from country j  Import from country i  
Import from other countries  Import from other countries  
Nominal wages  Labor market  Nominal wages  
Employment  Employment  
Foreign CPI  CPI  Prices  CPI  Foreign CPI  
GDP deflator  GDP deflator  
GDP expenditure components deflators  GDP expenditure components deflators  
Nominal exchange rate with foreign currency  Real effective exchange rate  Financial market  Real effective exchange rate  Nominal exchange rate with foreign currency  
UIP condition  Nominal interest rate  
Nominal interest rate  
Central Bank policy rate  Government sector  Central Bank policy rate  
Income tax rate  Personal income taxes  Personal income taxes  Income tax rate  
Corporate tax rate  Corporate income taxes  Corporate income taxes  Corporate tax rate  
VAT rate  VAT revenues  VAT revenues  VAT rate  
Excise taxes  Excise taxes  
Other taxes  Other taxes 
The multicountry model includes 357 equations and identities, which are presented in Appendix S2 and Appendix S3 (see Supplementary material
We start building a multicountry model by modeling aggregate supply for each country. Potential output in each economy is estimated using the Cobb–Douglas production function with constant returns to scale:
Y_{t} = BA_{t} K_{t}^{α} L_{t}^{1–α}, (1)
where A_{t} stands for total factor productivity (TFP) and B represents some normalizing constant. Capital and labor shares in Kazakhstan are fixed at 0.44 and 0.56 respectively whereas they are equal to 0.4 and 0.6 for Russia (see
The goods market of singlecountry models consists of equations for private consumption, gross capital formation, imports and exports. Government consumption is treated as an exogenous variable in all singlecountry models. Equations for private consumption are built in the form of Keynesian consumption function in which it depends on current disposable income. We also resort to permanent income hypothesis in specifying the consumption equation where we use the interest rate as an explanatory variable on the righthand side. However, the effect of the interest rate is significant only for Belarus, Kyrgyzstan, and Russia. Therefore, we drop the interest rate from the righthand side of private consumption equation for the two remaining countries. In addition, we include lagged consumption on the righthand side to capture the effect of habit persistence. We model investment as depending on real domestic demand and longterm real interest rate for Belarus, Kazakhstan, and Russia. At the same time, investment in Armenia and Kyrgyzstan is modeled according to the neoclassical theory of investment whereby the user cost of capital is included as an explanatory variable in the regression equation for investment.
Singlecountry models form the multicountry framework through the channels of bilateral trade and bilateral exchange rate. We construct four bilateral export equations for each singlecountry model to determine a country’s exports to its trading partners within the EAEU. Bilateral export equations are estimated in an error correction form where exports depend on the lagged value of itself, foreign demand and the bilateral exchange rate. The world trade index is used as an explanatory variable in some equations for bilateral exports. The country’s exports to the rest of the world are also modeled within the multicountry framework. The lagged dependent variable, the real effective exchange rate (REER) of a shipping country and the world trade index are used as explanatory variables in the equations for exports to the rest of the world. We also use the oil price as an independent variable on the righthand side of regression equations for the exports of Kazakhstan and Russia to the rest of the world. The total exports of each country are determined by summing the bilateral exports and exports to the rest of the world. The modeling approach of imports is simple as we define bilateral imports with respect to bilateral exports using the following identity:
M_{ij} = X_{ji} E_{ij}, (2)
where M_{ij} represents bilateral imports of a receiving country, X_{ji} stands for bilateral exports of a shipping country and E_{ij} is the exchange rate of a receiving country’s currency in terms of the currency of a shipping country. At the same time, an equation for imports from the rest of the world is estimated in an error correction form in each singlecountry model. The variable is mainly determined by the lagged dependent variable, domestic demand and REER.
The bilateral exchange rates are key elements of the financial markets in this model. Together with the trade equations, it forms the main channel through which the singlecountry models affect each other. We estimate 10 bilateral exchange rate equations and define 10 corresponding bilateral exchange rate identities in the multicountry model. As in
${E}_{t}\left(\frac{{S}_{t+k}}{{S}_{t}}\right)={i}_{t}{i}_{t}^{*}+{\u03f5}_{t}$, (3)
where S_{t} stands for the bilateral exchange rate in terms of domestic currency, i_{t} refers to the domestic interest rate and i_{t}^{*} is the foreign interest rate and ϵ_{t} is the risk premium. We also specify a regression equation for REER where it depends on bilateral nominal exchange rates. REER in Russia is also determined by real GDP, while REER in Belarus and Kyrgyzstan is significantly affected by the domestic price levels. We also build equations for medium term government bond rates in Armenia, Belarus, Kazakhstan, and Russia, whereas for Kyrgyzstan we model the household saving rate.
The labor market in each singlecountry model is characterized by the equations of nominal wages and labor demand. Nominal wages are primarily determined by the lagged value of itself and unemployment rate. A negative sign of the coefficient of unemployment rate in the nominal wage equations is in line with the bargaining model. Domestic inflation is also used as an explanatory variable in the nominal wage equation in all singlecountry models except for Armenia. Nominal wages in Armenia, Kyrgyzstan, and Russia are also determined by labor productivity. All nominal wage equations are specified in an error correction form. At the same time, the employment equation in each singlecountry model is estimated via Tobit regression. Labor demand in Armenia, Kazakhstan, and Russia is explained by the lagged dependent variable, domestic output and real wages. Employment in Kyrgyzstan depends on unit labor costs, while employment in Belarus is determined by the lagged dependent variable and GDP.
The model consists of the following price variables: CPI, GDP deflator, consumption deflator, investment deflator, government consumption deflator, export and import deflators. The dependent variable in the equation for CPI is measured as annual CPI inflation. An inflation persistence parameter in the estimated CPI equations has a wide range as it varies from 0.45 in Armenia to 0.87 in Kazakhstan. CPI inflation in Kazakhstan is largely determined by domestic output and US inflation. Inflation rates in Belarus and Russia depend on real GDP and the exchange rate of a domestic currency visavis the US dollar. Consumption deflator and real GDP are used as explanatory variables in modeling CPI inflation in Armenia. At the same time, CPI inflation in Kyrgyzstan depends on domestic output and CPI inflation in Russia and China. On the other hand, the GDP deflator is determined by the domestic and foreign CPI inflation for all five countries. Other deflators are modeled in a similar fashion.
Equations for revenue components of the government sector are also estimated for each singlecountry model. We specify the equations in an error correction form for personal income taxes, corporate income taxes, VAT revenues, excise taxes and other taxes. The government sector includes an equation for a central bank’s policy rate. A policy interest rate in each country is explained by the lagged dependent variable and the inflation rate. The key rate in Russia also depends on the output gap and REER. The interest rate in Belarus mainly depends on the output gap, while the base rate in Kazakhstan is set with a look towards a nominal exchange rate.
All dependent variables are specified either in logarithmic differences or in levels, meaning that in the former case we use the growth rate of a dependent variable. As a result, a constant term in a regression equation represents the deterministic trend growth of the dependent variable when the latter is in growth rates. At the same time, a constant coefficient captures the constant mean of a dependent variable when it is specified in levels.
An insignificant constant coefficient in a regression with a dependent variable in growth rates implies that there is no deterministic trend growth in the dependent variable, and its trend, if there is any, is purely driven by the independent variables on the righthand side. For example, price deflator equations in the model are specified in growth rates, and constants in these regressions tend to be insignificant which means they do not have a deterministic trend. However, it is a wellknown empirical regularity that prices tend to rise over time, but in this case the trend in prices is driven by independent variables on the righthand side. The same logic applies to all other dependent variables specified in growth rates. For example, a constant term in the consumption equation for Belarus is insignificant, but it does not mean that there is no trend growth in consumption. It simply means that the variable does not exhibit a deterministic time trend, but it exhibits trend growth due to the rising disposable income appearing on the righthand side as an explanatory variable.
The equations with dependent variables specified in levels have insignificant constant coefficients, if they do not have constant means or their constant means are already captured by some of the independent variables on the righthand side. For example, equations for interest rates tend to have insignificant constant coefficients, and we also include the lagged interest rate as an explanatory variable. Therefore, a constant mean of an interest rate is usually captured by its own lag because the mean of the interest rate does not change much from one period to the next. In general, we do not drop constant coefficients from the regressions, even if they are insignificant because in case residuals have a nonzero mean the intercept of the regression absorbs the constant mean of the residuals.
To analyze the ability of the multicountry model to fit the historical growth path of the main macroeconomic indicators in the EAEU countries, we conduct a baseline simulation from 2004 to 2018. An assessment of the model’s ability to fit the actual data in the baseline simulation is an important procedure to validate the present multicountry model for conducting policy and shock analyses. In particular, we focus on the model’s ability to repeat the dynamics of real GDP, potential output, inflation rate and unemployment rate. The results of the model simulation for all countries are presented in Appendix
Appendix
The dynamics of macroeconomic indicators for Kyrgyzstan are presented in Appendix
In this section, we discuss two simulation exercises with two different shocks to illustrate the propagation mechanisms of shocks in the multicountry model. In particular, we consider the world trade shock and the positive monetary policy shock in Russia. Appendix
Appendix
Appendix
At the same time, the response of macroeconomic indicators in other countries to the monetary policy shock in Russia varies in terms of directions and magnitudes. Real GDP and the inflation rate in Kazakhstan respond to the shock negatively. The strong aggregate demand in Russia and the value effect of the local currency depreciation temporarily increase imports in Kazakhstan. Although exports of Kazakhstan also rise due to the currency depreciation, the effect of the shock on imports outweighs the impact of the increased exports, leading to a decline in GDP and the inflation rate. The impact of the shock on real GDP of Kyrgyzstan is negligible since bilateral trade between Kyrgyzstan and Russia is small. At the end of 2018, it accounted for 0.02% of Kyrgyzstan’s total trade volume. Real GDP in Armenia responds positively to the monetary policy shock in Russia. The shock affects macroeconomic indicators in Armenia mainly through the channel of the bilateral exchange rate. Belarus is the hardest hit economy by the shock among the EAEU countries. The Belarusian ruble depreciates and the strong aggregate demand in Russia raises exports of Belarus. However, the later slowdown in the aggregate demand in Russia exhibits a negative impact on real GDP of Belarus. The inflation rate in Belarus also falls since the negative monetary policy shock in Russia results in lower aggregate demand for Belarus. Macroeconomic indicators in Belarus and Russia experience a more prolonged return to the baseline level than in other countries.
This paper presents a multicountry macroeconometric model for the EAEU. It is constructed based on five macroeconometric models of Armenia, Belarus, Kazakhstan, Kyrgyzstan, and Russia. The research aims to describe the structural relationship between the economies, assess the effects of various shocks on main macroeconomic indicators, and investigate the transmission mechanisms of shocks across the EAEU member states. The singlecountry models are connected through equations of bilateral trade and bilateral exchange rates. We evaluate the goodness of fit of the multicountry model by conducting the expost simulation for the period from 2004 to 2018. The simulation results demonstrate that the model is able to match the observed data on the main macroeconomic indicators in Kazakhstan and Russia. At the same time, the results of the ex‑post simulation also show that the goodness of fit is satisfactory for Armenia, Belarus, and Kyrgyzstan but it is inferior to the fit of the model for Kazakhstan and Russia.
The multicountry model is used to analyze the responses of important macroeconomic indicators to foreign and domestic shocks. Therefore, we perform two scenario analyses with a negative world trade shock and a positive monetary policy shock in Russia. In the first scenario, we consider the effect of a 10% world trade contraction on macroeconomic indicators across the member states. The counterfactual analysis shows that all the countries experience an economic downturn in response to the world trade contraction. Nevertheless, we find that the impact of the shock differs across them. The hardest hit economy by the world trade contraction is Armenia, whose economy largely depends on tourism, while the least affected country is Russia. Real GDP in Armenia falls below the baseline level by 5 percentage points whereas domestic production in Russia falls only by 0.8 percentage points. This proves the vulnerability of the Armenian economy to external shocks negatively affecting the world trade whereas the Russian economy is relatively immune to international trade shocks.
In the second scenario, we analyze the impact of an expansionary monetary policy in Russia. The results indicate stronger influence of the Russian monetary policy shock on domestic output of Belarus relative to other countries. At the same time, the response of real GDP in Kyrgyzstan is negligible which implies that its economy is less dependent on the Russian economy compared with other EAEU countries. In general, the model allows us to evaluate the impact of changes in monetary or fiscal policies of trading partners on the domestic economies of the EAEU member states. Therefore, the multicountry macroeconometric model can be considered as a useful tool for simulating main macroeconomic indicators across the EAEU in response to changes in policy and global economic conditions.
The author gratefully acknowledges Nurdaulet Abilov (NAC Analytica, Nazarbayev University) and anonymous referees for their valuable comments and suggestions.