Alexander Shirov, Alexey Yantovsky
The new version of the RIM[1] model
Abstract
This article is dedicated to the experience of development of the Russian economy’s I-O model. In current paper are represented general scheme of calculations as well as specifications for the main model parts.Special attention is paid to the problems of employment and labor productivity forecasting. A possible changes in labor productivity and its effects on Russia’s economic growth are estimated.
Keywords: I-O models, Russian economy, long term forecast, labor productivity, employment
JEL classification: E2,E3
The first version of the RIM model was created in 2000.A detailed description of the previous version of the model can be found in Georgy Serebryakov [1] and Marat Uzyakov[2].This was probablythe first dynamicinterindustry modelof Russian economydeveloped after the collapse of the Soviet Union.And this,of course,was the firstRussian Inforum-type model. For a long timemodelRIMhas been successfully usedfor practical calculations on the Russian economy.For example, it was used for the purposesof development of long-term forecasts of the Russian economy, the analysis of
the Russian government policy, assessing the impact of Russia's WTO accession.
But in 2004, Russian Statistics Service (Rosstat) has changed the basic principles ofindustrial classification, havingharmonized classifications adopted inthe European Union.In this regard, Input-Output tables were not being developedin Russiafrom 2004 till 2011.This posed a clear problem for the development of inter-industrymodels.
In 2008-2010 in Institute of economic forecasting Russian academy of sciences our team under the leadership of Marat Uzyakov worked on the creation of symmetric Input-Output tables for 2000-2009.A large quantity of statistical informationused in development of the I-O tablesallowed us to create balance sheets in 45 sectors classification.
Development of adequate statistical base allowed moving to the development of a new version of the RIMmodel.In addition, we paid special attention to the behavioralmodelsin thesystem of the macroeconomic calculationsused inour Institute.Our position isthatfor complexmacro-economicresearchis not enoughto have onlyone model.We need to usecomplexmodelswhich complementeach other'scapabilities. In this complex the RIMmodel should be atop-level model, defining the mainparameters of the forecast, forming akey restrictionfor economic development.
More specificobjectivesand detailedforecastsare madeat lower levels. It usesin other inter-industrymodel, named CONTO, as well as models of separate sector or regions. Some of the tasksof short-andmedium-term forecastare solvedby means of macroeconomic models. The task of developing and matching the key parameters ofthe scenario (external economic conditions, exchange rates etc.) can be solvedonthe basis of specificmodels.
Fig.1 Current system of IEF RAS models
Statistical base of the RIM model are rows of I-O tables of Russian economy in constant and current prices for years 1980-2009 constructed. These tables were developed in Institution of Economic Forecasting from data provided by Russian Federal State Statistics Service, customs statistics about foreign trade, statistical forms describing structure of production costs and other sources. Also during our work on the model we used following sources such as national accounts for years 2002-2009, institutional accounts for 2002-2009, matrixes of trade and transport margins, tax matrix, rows of sector investment and estimations of fixed capital, consolidated budget data. In the model the whole economy is divided in 45 sectors, represented in the table 1.
Agriculture / Automobiles, highway transport equipmentPetroleum extraction / Sea transport equipment and its repair
Natural gas extraction / Airplanes, rockets, and repair
Coal mining / Railroad equipment and its repair
Other Fuels, incl. nuclear / Recycling
Ores and other mining / Electric, gas, and water utilities
Food, beverages, tobacco / Construction
Textiles, apparel, leather / Wholesale and retail trade
Wood and wood products / Hotels and restaurants
Paper and printing / Transport and storage
Petroleum refining / Communication
Chemicals / Finance and insurance
Pharmaceuticals / Real estate
Plastic products / Equipment rental
Stone, Clay, and Glass products / Computing service
Ferrous metals / Research and development
Non-ferrous metals / Other business services
Fabricated metal products / Government, defense, social insurance
Machinery / Education
Computers, office machinery / Health services
Electical apparatus / Other social and personal services
Radio, television, communication equipment / Private households with employed persons
Medical, optical, and precision instruments
Fig.1 The economy sectors in the RIM
The general algorithm of model consists in iterative procedure of calculations with step of one year. For each year, a cycle of calculations is repeated until a convergence criteria is met. Before the start of the iterative process, each endogenousvariable is assigned a value, acquired on the previous step. At first step elements of final demand are calculated inconstant prices. Using price index acquired on previous iteration the respective vectors are calculated in current prices.On the next step gross outputs for every industry are estimated by leontivian model. Then production capacities are calculated from volumes of investment and obtained outputs are verified not to be exceeding restrictions imposed by fixed capital. On further step of calculationswe estimate the elements of gross value-added, such as wages, profits and taxes. Next, using price leontivian model current prices are calculated. Further incomes of population, business and government are estimated and used for calculation of final demand. Thereby closing a cycle of calculation. Before starting next cycle the convergence criteria, such as difference of GDP amounts obtained on twosuccessive iterations, is verified.
Fig.3 The principal scheme of the algorithm
This scheme is rather general for I-O models and the main interest lies in ways of forecasting of separate elements of the model. The core of the model consists in calculations of final demand and value added. These blocks of the model have endured much discussion before we came to a current scheme of calculations.
As household’s consumption we forecast personal consumption per capita which it multiplied by population. Personal consumption of some industries goods is predicted as function from total personal consumption, population income per capita, relative prices and difference in the level of become between present and previous years. But choice of such set of factors for regression equations as well as period of 2000-2010 years for estimation of equation’s coefficients has a certain drawback. More precisely it causes excessive growth in consumer’s demand of food and textile industries goods. From our point of view structure of household’s consumptionshould be changing towards increase of share of services and durable goods. To solve this problem, we intend to implement saturation function for personal consumption forecasting. By usage of such functions we can receive more plausible dynamics of household’s consumption structure. Parameters of these saturation functions are estimated by comparisons of international data about personal consumption and income level.
pceRpc[i] =a1 +a2*pceRTpc + a3*dinc+a4*moneyinc+a5*rprices
where moneyinc – population incomes per capita,
rprices – relative prices
pceRTpc = pceRT/pop – total personal consumption per capita
dinc = pceRTpc - pceRTpc[1]- increase in personal consumption per capita in comparison with previous year
Governmental consumption still remains one of most unfinished blocks of the model. At present time we assume that product structure of government consumptions is set exogenously or remains constant on the whole forecast period. The total amount of government expenditures depend on tax incomes. Along with estimation of budget incomes we estimate amount of tax payments which is transferred into reserve funds. In current version reserve funds are replenished from mineral extraction tax payments and export duties on oil. Financial assets, accumulated in reserve fund are available for usage in later periods if needed. There is determined by special exogenous parameter, which has meaning of minimal level of governmental expenditures. If current budget income is lower then this preset value and also there is a positive amount of money in reserve funds, a deficient amount is transferred from reserves into budget. This functional allows us to obtain more smooth dynamics in case of abrupt changes in external prices and export amounts.
For forecasting of the fixed capital we use a bit complex scheme. We calculate amount of investments made by purchaser in constant prices and then use “investment bridge matrix” to allocate the product composition of investment. Obtained values are aggregated over sectors and vector of investment by products in made. For calculation of investment by purchaser we use regression equations. A set of factor in these equation include amount of replacement of retired fixed capital, first difference of industry’s output, profits and percentage of used production capacities.
capinv = a *replace +a2* dif +a3*profit/invD +a4*out/capstock,
where replace=replaceRate*capstock –replacements of retired fixed capital,
dif =outR[t]-outR[t-1] - first difference of the industry's output,
outR/capstock – level of production capacity usage,
profit- amount of received profits in given industry
invD = invT[t]/invRT[t] – index of prices on investment assets,
invT – investment in current prices,
invRT – investment in constant prices.
Amount of replacement is calculated from capital stock by exogenously set retirement rate. Ratio between output and fixed capital stock allows estimating level of production capacities usage and thus necessity for investment in given sector. Achieved profits describe in the first place amount of financial resources available for financing of investments and secondly interest in given sector’s development. After estimation of investments made by given sector as purchaser obtained amount is divided using matrix of investment’s technological structure by sectors. This structure consists of construction, machinery and equipment and other sectors (for example structure of capital investments in oil extraction is 32% construction, 23% machinery, 44% others). Elements of technological structure are aggregated over sectors and we receive amount of investment expenditures on construction, machinery and other sectors in whole economy. Each of these elements in turn has its own structure by products. By summing them we finally obtain vector of investments by product.
inv[i] = ∑j (capinv[j] * ∑ InvTS[j][n]*InvEl[i][n])
InvTS – matrix of technological structure of investment made by purchaser.
InvEl –matrix of coefficients, which show product structure of construction, machinery and production of other investment goods.
Inventory changes are calculated from output amounts. This block remains second undeveloped part of the model. Assumption that inventory changes can be calculated as shares of outputs on whole forecasting period is an arguable one. Another considered method of receiving inventory changes is exogenously set their amounts. Currently we decide in favour of former way.
Exports are forecasted from outputs, volume of internal consumption and world economy growth rate, which is an exogenous variable.
exR=a1+a2*outR + a3*intCons+ a4*exR[1]*worldrate;
where worldrate – world economy growth rate
intCons- internal consumption
For several industries, such as oil and gas extraction, ferrous and non-ferrous metals production and chemicals production amounts of export are determined as residue between outputs and internal consumption. It is possible because outputs for these industries are calculated by production functions rather than by leontivian model.
As for imports we calculate a share of import goods in internal consumption. This variable depends on its value in previous year, exchange rate and share of new production capacities in fixed capital stock. Thepurpose of latter factor is to describe decrease of share of import in internal consumption as economy grow and bind it to variable with economic meaning instead of using simple time dependency. In this case we assume that newly created manufactories will be competitive with foreign producers at least on domestic market.
ImShare =ImShare[1](a1 +a2*rateusd/rateusd[1]+ a3* capinv[1]/capstock[1])
capinv[1]/capstock[1]- share of new facilities in fixed capital
rateusd- exchange rate
Exchange rate in turn depends on ratio between export and import, consumer price index and external oil prices.
rateusd[t] = rateusd[t-1]*(1+a1*(CPI[t]-brent[t]/brent[t-1])+a2*(imT[t]/exT[t] -1)
Outputs for most industries are calculated using leontivian using with the exception of as oil and gas extraction, ferrous and non-ferrous metals production and chemicals production. For mentioned industries outputs are estimated by means of production function. For example, output of oil extraction is calculated by function
outR[2] = 0.94*outRlag[1][2]+(0.11*capinv[2]+0.27*capinvlag[1][2])* /capintensity [t];
where outRlag[1][1]- output of oil extraction in previous year
capinv[2]- investment made by oil extraction as purchaser
capintensity- capital intensity of oil extraction, which depend from accumulated output of oil extraction
Output of ferrous metals production is estimated via function
outR[16] = 0.97*outRlag[1][16]+(0.08*capinv[16]+0.18*capinvlag[1][16]);
Similar method is used when we determine current prices. For some industries, which are oriented on export, for example oil and gas extraction, or are natural monopolies, such as transport and power generation, prices are set exogenously. For others prices are received bysolving price leontivian model. For that we need preliminary estimate elements of gross value added. Wages are calculated from outputs in current prices, received on previous iteration, and consumer price index. For industries with exogenously set prices profits are calculated as difference between gains and costs. Industries with “unfixed” prices try to achieve level of profitability determined by special regression equation or simply inherited from previous iteration. Taxes are calculated from amount of outputs, exports and profits – depending on type taх. Besides taxes on products, which are calculated as a share of outputs in current prices, in the model are represented value added tax, mineral extraction tax, export duties, profit tax.
After estimation of gross value added and its elements we are able to determine current prices and income of population, government and business. Latter values are used from calculation of final demand elements.
Along with described main cycle of calculations there are several important features, which also influence model results. First of them are productivity functions. These functions are used in the model as upper constraints for gross outputs. The main factors used in them are amount of fixed capital, investments and exogenously set efficiency of primary resources usage.
From our point of view efficiency of primary resources usage is important element reflecting of key technologies change. Under primary resources we understand all set of raw materials used for manufacture. They are energy resources and products of metallurgy, woodworking and chemistry. Efficiency of primary resources usage is measured as a ratio of total output and used primary resources costs. The higher its value is, the more is share of value added by manufacture. Given indicator shows us a level of economic and technological development of economy. Increase in efficiency of primary resources usage may be caused by fixed capital renovation. High enough investment behavior provides achievement of more effective technological structure of production and reducing of intermediate consumption of primary resources, first of all energy.
Next feature of the model are functions of labor productivity and employment. Changes of labor productivity on forecast are firstly caused by optimization of structural and organizational components and secondly by improvement in used technologies due to renovation of fixed capital. The former factor is exogenous and based on our estimations derived from labor productivity comparison between different countries, such as Russia, United States, Japan, Germany and CzechRepublic. The latter one is calculated inside the model with the help of regression equations, which use amounts of fixed capital and investments as factors.
Thereupon employment is calculated as ratio of gross output and labor productivity. Growth of labor productivity may cause significant decrease of employment. To avoid this problem professor Clopper Almon suggested to use regression equations forecasting the logarithm of the employment/output ratio as a function of time and the change in the logarithm of output.
We used our model to make a forecast and estimate possible restrictions from labor on economy growth. It showed that in scenario of economy average growth of 6 per cent, several industries of manufacture will encounter deficit of available labor resources in 2020-2022. These industries are in first place machinery, electrical apparatus production, production of Radio, television, communication equipment, vehicles industry. Partly this is caused by the fact that the given industries need highly skilled specialists. Their shortage could not been compensated by attraction of migrants. Slowdown of machinery growth complicates buildup of investments in economy. It leads to decrease in growth rates of whole economy.
Not only amounts of required labor resources change, but also their structure does. Share of specialists increases in half and share of maintenance and scientific personal almost triples. At the same time percentage of skilled workers
and operators of plant and equipment slightly decreases, whereas shares of executive officers and unskilled workers are reduced by half.
The availability ofa workingversion of the modelhas made possibleto make a numberofpractical calculations.In particular,a forecastof possible dynamics of employment and labour productivity in the Russian economy tillyear 2030 was made. For assessingthe prospects forRussian economic developmentwe should take into accountitstransientnature.At presentthe distinctive features of the Russian economy include:
•For a long period Russian economy could grow without high share of investments in GDP (in 1999-2006 about 16% of GDP);
• In period 2000-2008 Russian economy had average growth rate of GDP about 7%
•This growthwas basedon the use ofthe old Sovietcapital;
• Old capital was created in old conditions: chip energy, low restrictions for labor forces ect. Old plants keep old system of management;
The presence in the economy of a large amount of production capacities, which were created in planned economy andare characterized by low efficiency values,is a certain restriction for economic development. Thisproblemcanbesolvedintwoways. The first one –by investments in fixed capital, the second – by application of new methods in management.The second way is not so expensive and allows to increase efficiency of production quickly enough. For example: growth of labor productivity in last years was connected with changing in structure of business and not so much depended on technological changes. We have to take this factorinto account in our models and forecasts.