SOVSIM: A MODEL OF THE SOVIET ECONOMY
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP08S01350R000100110001-4
Release Decision:
RIPPUB
Original Classification:
K
Document Page Count:
52
Document Creation Date:
December 21, 2016
Document Release Date:
September 12, 2008
Sequence Number:
1
Case Number:
Publication Date:
February 1, 1979
Content Type:
REPORT
File:
Attachment | Size |
---|---|
![]() | 2.19 MB |
Body:
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
National
Foreign
Assessment
Center
SOVSIM: A Model of
The Soviet Economy
A Research Paper
ER 79-10001
February 1979
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
National
Foreign
Assessment
Center
SOVSIM: A Model of
The Soviet Economy
Comments and queries on this unclassified report
are welcome and may be directed to:
Director for Public Affairs
Central Intelligence Agency
Washington D.C.,20505
(703) 351-7676
For information on obtaining additional copies,
see the inside of front cover.
ER 79-10001
February 1979
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Model Comparisons
5
Historical Error Analysis, 1965-77
6
General Tracking Record
7
Average Errors for Key Variables
8
A Short Run Forecasting Experiment
10
SOVSIM Impact Analysis, 1978-85
11
Impacts of Three Hypothetical Shifts
14
A Preliminary Assessment
15
Appendix A. Specification of SOVSIM
18
Appendix B. Variable Lists
38
Table 1. Comparison of Actual and Projected GNP Growth
Table 2. Average Simulation Errors of Key Variables, 1965-77
Table 3. SOVSIM Forecasts of Key Variables for 1976
12
Table 5. Impact Analysis with SOVSIM: Effects on Selected
Economic Variables
16
Appendix Tables
Table Al. Equation List for the Production Block
19
Table A2. Equation List for the Consumption Block
22
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Table A3.
Equation List for the Investment Block
24
Table A4.
Equation List for the Capital Formation Block
26
Table A5.
Equation List for the Energy Block
30
Table A6.
Equation List for the Employment Block
33
Table A7.
Equation List for the Trade Block
36
Table B1.
Variable List for the Production Block
38
Table B2.
Variable List for the Consumption Block
39
Table B3.
Variable List for the Investment Block
40
Table B4.
Variable List for the Capital Formation Block
41
Table B5.
Variable List for the Energy Block
42
Table B6.
Variable List for the Employment Block
43
Figure 1.
General Flow Diagram of the Soviet Economic Model
vi
Figure 2.
Condensed Model Structure
3
Figure 4.
SOVSIM: Simulation of Private Consumption
8
Figure 5.
SOVSIM: Simulation of New Fixed Investment
8
Figure 6.
SOVSIM: Simulation of Net Exports of Fuel for Hard Currency
9
Figure 7.
SOVSIM: Simulation of Hard Currency Net Exports of Oil
9
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
SOVSIM: A Model of
The Soviet Economy
Econometric models have become conventional tools
for analyzing Western economies during the past
decade. They provide a convenient mechanism for
looking at the interactions of many factors simulta-
neously and for studying the potential impacts of
policies and economic events on the path the economy
is expected to follow.
The use of models in the study of centrally planned
economies (CPEs) has lagged behind Western applica-
tions, however. Western models are essentially descrip-
tions of the structure and sources of demand, which in
turn determine the levels of production, employment,
and prices. Little of the understanding of the economic
structure gained from Western modeling research can
be transferred to the description of supply-oriented
CPEs, where resources are more or less fully employed
and use is determined by both availabilities and
relative priorities.
This paper describes the present version of sovsIM. The
first section discusses the structure of the model in
general and schematic terms. The second section
reviews the performance properties of the model in
historical simulations, and the third looks at the model
as a short-run forecasting tool. The fourth section
illustrates the use of the model in impact analysis of
Soviet growth prospects to 1985, and the final section
gives a preliminary assessment of our research. We
have also included appendixes detailing the model's
structure and listing all of the variables.
sovsim is the outgrowth of a continuing effort to
develop a model of the Soviet economy. The structure
of sovsim reflects the fundamental production focus of
a CPE. Capacity of the capital goods industries
determines investment, which in turn establishes the
pattern of growth in the stock of productive capital.
Demographically determined employment together
with the capital stock set the achievable level of
production. This output is then divided among compet-
ing uses based on availabilities and relative priorities,
with private consumption generally taken as the
residual claimant.
The primary purpose of sovsim is to support studies of
growth prospects for the Soviet economy, especially the
influence that certain constraints on the supply side
could have on these prospects over the next decade.
Consequently, the structure of the model is designed to
accommodate analysis of the impact of policy shifts
and contingent events in areas like labor supply,
energy, investment, and foreign trade.
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
rn
L
Approved For Release 2008/09/12 : CIA-RDP08SO 1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
SOVSIM: A Model of
The Soviet Economy
sovsim is an annual model consisting of 207 equations
connecting a like number of endogenous variables with
67 exogenous variables. Thirty-five of the equations
involve econometric estimates of parameters; 90 of
them use nonstatistical procedures to estimate struc-
tural parameters and 82 are accounting identities.
The general structure of the Soviet model is shown in
figure 1. Since sovstm is basically supply driven, most
of the model is devoted to describing resource avail-
abilities and production relationships. This includes
the effects of (a) investment on capital formation, (b)
labor and capital on output, and (c) energy on capital
utilization and foreign trade. A much smaller portion
of the model is devoted to estimating the components of
demand other than investment.
? Investment. The allocation of available investment
resources among competing uses is set by policy
decision.
? Government Spending. This group includes the level
of personnel and the growth rate in nonpersonnel
expenditures for defense, and the shares of administra-
tion and research and development in gross national
product.
The model can be used to project seven groups of
economic variables-the model's endogenous
variables:
? Production. Outputs of 13 producing sectors are
computed in terms of value added and then are
summed to obtain GNP.
Model Variables
All projections from the model are conditioned by
assumptions regarding six groups of external or
exogenous variables:
? Energy. These variables include projected gross
outputs of fuels and electric power, the energy
allocation policy, and the capital flexibilities for each
producing sector.
? Population and Manpower. Projections of the able-
bodied population and the number of pensioners are
inputs to the model. Participation rates and employ-
ment rates, as well as the distribution of employment
by sector, must also be established using outside
information or by assumption.
? Weather. Weather conditions are defined by indexes
of precipitation and temperature.
? Foreign Trade. Nonfuel exports to the West depend
primarily on external economic conditions and are an
input to the model. Energy exports to Eastern Europe
are considered a function of both political and eco-
nomic factors and are therefore set outside the model.
Gold sales, arms sales, and new credit drawings also
fall in this category.
? Consumption. Separate calculations are made for
four categories of public consumption: administration
and R&D, which are scaled as fixed shares of GNP;
nonpersonnel defense expenditures, which are com-
puted from an exogenous growth rate; and personnel
defense expenditures, which are the product of as-
sumed manpower and imputed wage rates. Private
consumption is determined as the residual claimant on
output after deductions are made for public consump-
tion, investment, and foreign trade.
? Investment. The model computes investment in each
of the 13 producing sectors plus housing and capital
repair.
? Capital Formation. New additions to the stock of
productive capital, retirements, and the gross stock of
productive capital are estimated for each producing
sector.
? Energy. Nominal requirements and actual deliveries
of fuels and electric power are computed for each
producing sector. Utilization rates of capital and hence
the effective or active capital stock are also estimated.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
? Employment. This group includes the civilian labor
force and the level of employment in each sector.
? Trade. Exports and imports are calculated sepa-
rately for trade with Communist, hard currency, and
other countries. Debt, debt service, grain imports, and
net exports of fuels are for hard currency also
estimated within the model.
Model Structure in Condensed Form
The structure of sovsiM has been condensed into a set
of 19 equations to facilitate discussion of the functional
relationships among model variables (see figure 2).
The variables and parameters appearing in these
equations are also defined in figure 2. This way of
reviewing the model focuses on the primary endog-
enous linkages; the full specifications of the model
equations are given in appendix A. The model
variables are listed in appendix B:
? Production. There are constant-returns-to-scale
Cobb-Douglas production functions for each
nonenergy producing sector (equation 1). Value added
in the energy sectors is scaled from gross output, which
is exogenous for these sectors. GNP is obtained by
summing value added in the 13 producing sectors
(equation 2).
? Consumption. Government expenditures (equation
3) include exogenous defense spending and an endog-
enous component scaled from the level of GNP.
Private consumption (equation 4) is calculated as the
residual claimant of GNP.
? Investment. The supply of capital goods available for
domestic investment is the residual of deliveries of
machinery and construction output to final demand,
after deductions are made for deliveries to defense,
exports, consumption, and capital repair (equation 5).
Equation 6 distributes new fixed investment to each
producing sector and housing with shares set outside
the model.
? Capital Formation. Net additions to the productive
capital stock are estimated from past investment and
assumed depreciation rates (equation 7). Identity
equations then link capital stock to the previous year's
capital stock and net capital formation (equation 8).
? Employment. The labor force is estimated from the
able-bodied population and participation rates (equa-
tion 9). Total employment (equation 10) depends on
the labor force and employment rates, and sector
employment levels (equation 11) follow from the total
employment and labor allocation shares.
? Energy. Equation 12 estimates nominal demands for
oil, gas, coal, and electric power in each consuming
sector from the capital stock and energy-use coeffi-
cients tied to the capital stock of the given sector.
Actual deliveries (equation 13) are determined by a
combination of nominal requirements and assumed
allocation policy. Equation 14 calculates domestic
energy residuals by subtracting domestic deliveries
from gross domestic output. Depending on its sign, the
residual indicates either a capacity for net exports or a
need for net imports. Equation 15 calculates the
fraction of sector energy requirements, in terms of
standard fuel units, actually met by deliveries. To-
gether with an elasticity of active capital with respect
to energy input, this fraction determines the rate of
capital utilization and thus the active capital stock in
each sector (equation 16). Any shortfall in meeting
nominal domestic requirements for energy leads to a
reduction in capital utilization. The degree of reduc-
tion for a given shortfall varies by sector depending on
the value of the capital elasticity and the relative
contribution of the type of energy in short supply to the
particular sector's energy consumption.
? Foreign Trade. Net exports of fuels to hard currency
countries (equation 17) are the difference between the
domestic energy residuals and exogenous net exports to
Communist and other countries. Net exports of fuels to
hard currency countries, along with other variables
that represent sources of hard currency, feed into a
calculation of the hard currency import capacity
(equation 18), which in turn drives imports from hard
currency countries (equation 19). If these imports fall
below a specified floor, domestic energy use is reduced
by reducing eig. and energy exports are increased (or
energy imports are reduced) until sufficient import
capacity exists to meet the import minimum.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Energy
12. Ei,I=K.Xdi.l
13. Ei.J = Ei.l X ei.l
14. RJ=QJ - Ei.J- Efd,l
15. Di=(I EiJXhl)/(2; E..JXhl)
j j
16. Ki=K.Xi(I - giX(I - D,))
Foreign Trade
17. EHJ = R1 - EC1
18. MH, _ I EHJ + T
j
19. MH = f (MH,, MH)
Ni Employment in sector i
ds POP Able-bodied population
Qj Gross output of energy type j
Rj Residual of domestic production of energy
type j after deduction for domestic
deliveries
Rk Capital repair
T Net earnings of hard currency (other than
through trade in fuels) and net credit
drawings
es Xi Value added in sector i
icy Xk Value added in machinery and construction
sectors
p Participation rate
ri Depreciation rate of capital in sector i
t Share of output devoted to administration and
research and development.
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Figure 2
Condensed Model Structure
A. Equations
Production
Capital Formation
Energy
1. X, = f (K,, N, )
7. KF1=f(II,I(-1),,r,)
12. E,j =K1Xd,,j
2. GNP=EX,
8. K1=K,(-1)+KF,(-1)
13. E,,j = E.,j X elj
i
Consumption
Employment
9. LF = p X POP
14. RJ=Qj -E,,;-Efd.J
3. G=tXGNP+DF
15. D, = (E E1
j X h,) /(I E,
1 X h
)
4. C=GNP-I-G-(Ex-M)
5. I=aXXk - Ck - Gk - EXk - Rk
6. I,=b,XI
10. N=erXLF
11. N,=c,XN
,
,
j
j j
16. K1= K,X,(I -g1X (1 -D,))
Foreign Trade
17. EHj = Rj - ECG
18. MH,=ZEHj + T
j
19. MH = f (MH,, MH)
C
Private consumption
G
Government expenditures
N;
Employment in sector i
Ck
Expenditures on consumer durables
Gk
Defense expenditures on capital goods
POP
Able-bodied population
D;
Deliveries of fuels and power to sector i as a
GNP
Gross national product
Qj
Gross output of energy type j
percent of nominal requirements
I
Total investment
Rj
Residual of domestic production of energy
DF
Defense spending
I;
Investment in sector i
type j after deduction for domestic
E.,j
Nominal requirements of energy type j in
K;
Nominal capital stock in sector i
deliveries
sector i
K ,
Active capital stock in sector i
Rk
Capital repair
E
Deliveries of energy type j to sector i
KF1
Net capital formation in sector i
T
Net earnings of hard currency (other than
Efdj
Deliveries of energy type j to final demand
LF
Civilian labor force
through trade in fuels) and net credit
ECj
Net exports of energy type j to Communist
M
Imports
drawings
and other countries
MH
Imports from hard currency countries
Xi
Value added in sector i
EHj
Net exports of energy type j to hard currency
MR
Minimum imports from hard currency
Xk
Value added in machinery and construction
countries
countries
sectors
Ex
Exports
MHc
Hard currency import capacity
Exk
Machinery exports
N
Total employment
C. Parameters
a
Ratio of deliveries to final demand of machin-
e1,
Deliveries of energy type j to sector i as a
p
Participation rate
ery and construction to value added in these
percent of nominal requirements
r;
Depreciation rate of capital in sector i
sectors
er
Employment rate
t
Share of output devoted to administration and
bi
Share of total investment going to sector i
Elasticity of active capital with respect to
research and development.
c;
Share of total employment in sector i
input of energy in sector i
d .,j
Input of energy type j per unit of capital in
hj
Units of standard fuel per unit of energy type j
sector i
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Data Sources
The empirical relations in sovsim were estimated using
a data base covering 1960-77. While some data series
covering the 1950s were available, the 1960-77 period
was the longest for which a consistent data set could be
compiled. Data sources included official statistical
publications of the USSR, national income accounts
estimated by the Office of Economic Research (OER),
and input-output tables reconstructed by the Foreign
Demographic Analysis Division (FDAD) of the US
Department of Commerce.
Output in the Production Block was described by
sector-of-origin valued-added indexes of Soviet GNP
accounts developed by OER. The corresponding OER
estimates of GNP end-use accounts were the primary
basis for estimating the Consumption Block.
Data for total investment, new fixed investment, net
additions to livestock and capital repair in the Invest-
ment Block came from OER end-use accounts for
GNP. In addition, various issues of the official Soviet
economic handbook-Narodnoye Khozyastvo SSSR
(Narkhoz)-were the source of the sector investment
used to establish sector shares in new fixed invest-
ments. Capital stock data for the Capital Formation
Block came from OER indexes of the fixed capital
stock of the USSR. Estimates of imported capital were
compiled by Green and Levine ' and depreciation rates
were estimated by Green.'
The Energy Block uses a combination of time series
and intersectoral transactions data. Production figures
for fuels and power were taken from the Narkhoz, and
the amount of fuels in foreign trade from the official
Soviet foreign trade handbook. Sector shares in the
allocation of fuels and power available for domestic use
were estimated for 1972 from a preliminary recon-
struction of the 1972 Soviet input-output table in
producers' prices. These figures for 1972, along with
estimates of apparent consumption (gross production
minus net exports) and sector capital stock for all
years, were used to estimate changes in the allocation
pattern over time.
' See Donald W. Green and Herbert S. Levine, "Soviet Machinery
Imports," Survey, Spring 1977-78, p. 114.
2 See Donald W. Green, "Capital Formation in the USSR, 1959-
74," Review of Economics and Statistics, February 1978, p. 40.
Data for the Employment Block came primarily from
FDAD projections and compilations.' The official
Soviet foreign trade handbook was the major source of
trade data for the Foreign Trade Block and OER
estimates were used for hard currency debt and
drawdowns of credit.
Model Comparisons
As a model of a centrally planned economy, sovsim
bears little resemblance to conventional econometric
models of Western economies. The impressive
econometric research effort in the West has focused on
giving empirical content to an essentially Keynesian
view of Western macroeconomies. This means giving
great attention to the determination of the structure
and level of effective demand and little concern for real
constraints on growth.
Obviously, the latter issue-real constraints on
growth-is the core of any analysis of Soviet growth
prospects. This requires giving much greater attention
to descriptions of production and resource availability
than to competing uses for the output produced. This
was true of the earliest efforts by Niwa ? to construct a
very small model of the Soviet economy and continues
with both sovsim and the SOVMOD series of large-scale
econometric models developed by the combined efforts
of SRI International and the Wharton Econometric
Forecasting Associates (SRI-WEFA). 5
Both sovsim and SOVMOD are driven by a series of
sector production functions. Each uses Cobb-Douglas
specifications rather than more complex forms because
the statistical basis for rejecting the simpler form is, at
best, weak at the sectoral level. In other areas, though,
the specifications differ in critical ways.
'See Stephen Rapawy, Estimates and Projections of the Labor
Force and Civilian Employment in the USSR: 1950-1990, Foreign
Economic Report No. 10, September 1976.
' Haruki Niwa, "An Econometric Analysis and Forecast of Soviet
Economic Growth" in P. J. Wiles, The Prediction of Communist
Economic Performance (Cambridge, 1977).
' The original model is given in Donald W. Green and Christopher 1.
Higgins, SOVMOD I: A Macroeconometric Model of the Soviet
Union (Academic Press, 1977). Later versions of SOVMOD have been
described in working papers published by SRI-WEFA.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Most of these specification differences reflect a differ-
ence in the fundamental analytical objectives and
assumptions underlying the development of the
models. The SRI-WEFA research has assumed that
any regularities in the decisions of planners that could
be isolated from historical data were a preferred basis
for projecting Soviet growth prospects. Consequently
in certain parts of SOVMOD, important roles are played
by estimated trade-offs between alternate patterns of
resource use, based on indexes of planners' behavior.
The sovsim research has focused, instead, on the roles
that resource constraints play in shaping the pattern of
Soviet growth and has assumed that past trade-off
responses of the planners are not necessarily the
preferred basis for judging the future. As a result, the
specification of sovsim generally has less behavioral
content but more effectively takes resource availabil-
ities into account in projecting Soviet growth.
These differences are especially pronounced in
four areas:
? Investment. sovsim constrains current new fixed
investment to equal the available output of the
machinery and construction sectors and allocates it
among sectors by share trends, or shares determined
exogenously. The SRI-WEFA approach looks at
investment from a behavioral perspective that tries to
link realized investment with resource competition
with defense, but does not limit realized investment to
the investment that could be carried out with the
resources available.
? Energy. sovsim was developed to examine the effects
of shifts in resource supplies, especially energy re-
sources, on growth potential. The sovsim specification,
therefore, gives explicit consideration to sector de-
mands for energy and links energy supplies to both
domestic production and foreign trade possibilities.
This integration of energy into the model is missing
in SOVMOD.
? Foreign Trade. Soviet imports from the West depend
on Soviet hard currency import capacity, which in turn
is limited by Western credits and hard currency export
earnings. The sovsim specification, by explicitly link-
ing Western imports to import capacity, import
capacity to fuel exports, and fuel exports to a set of
overall fuel balances, more fully integrates trade into
Soviet growth analysis.
? Employment. The sovsim specification of employ-
ment is essentially an accounting framework based on
outside projections of population, participation rates,
and sector employment shares. The SRI-WEFA speci-
fication is a much more ambitious attempt to estimate
the influences on urban-rural migration and the
impact this process has on overall employment. It fails,
however, to impose realistic upper bounds on Soviet
participation rates, which are already the highest in
the industrialized world. Under certain conditions, this
can lead to an upward bias in long-term growth
analysis.
sovsim is then much more eclectic than conventional
econometric models, even of the Soviet Union. It tries
to deal with the real constraints on Soviet growth while
reflecting the rather underdeveloped theoretical un-
derstanding of the process of central planning and the
behavior of the planners who shape it.
Historical Error Analysis, 1965-77
One measure of the performance of an econometric
model is its ability to reproduce the historical (ob-
served) growth pattern of key variables. This test,
however, assumes knowledge of some data-such as
nonfuel exports to the West, weather indexes, and
defense spending-not available in making a forecast.
Moreover, future structural changes could alter
underlying functional relationships. Historical error
analysis should therefore be considered only in con-
junction with other tests in establishing the usefulness
of sovsim in the analysis of Soviet growth potential.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
USSR: Comparison of Actual and
Projected GNP Growth
Figure 3 Table 1
I I I I I I I I I I I 1 1
1965 66 67 68 69 70 71 72 73 74 75 76 77
General Tracking Record
The historical and simulated paths of GNP growth in
1965-77 are compared in figure 3. The model projec-
tions follow the historical record very closely, never
deviating by more than 1.5 percent from the official
figures. A more demanding comparison is that between
actual and projected growth rates (table 1).
The projected growth rates are reasonably close to
historical figures-the correlation coefficient between
them is 0.54-and the direction of year-to-year
changes is correctly projected in 10 out of 12 cases.
Private consumption and new fixed investment are
shown in figures 4 and 5. Both figures again show a
close match, but the simulation errors in a given year
are in opposite directions. This reflects the model's
specification of private consumption as the residual
claimant on GNP. Since GNP is simulated with little
1965
6.8
7.6
1966
6.4
7.0
1967
5.2
4.0
1968
6.1
5.8
1969
3.0
4.9
1970
7.7
5.2
1971
4.5
3.9
1972
1.9
2.7
1973
7.2
6.3
1974
4.1
9.1
1975
2.2
2.8
1976
4.0
4.0
1977
3.6
5.1
error, when investment is overestimated, consumption
as the residual use of GNP must be underestimated
and vice versa. Nonetheless, the projection error in
neither variable seems to be biased since the projec-
tions are not consistently above or below actual values.
The value of net exports of oil, gas, and coal to hard
currency trading partners is a key foreign trade
variable calculated in the model. Since these exports
are estimated as residuals in the fuels balance equa-
tions, while gross production of fuels and exports to
other countries are exogenous, the ability of the model
to track hard currency fuel exports depends on the
ability to project domestic use of fuels. Figure 6 shows
that the model accurately tracks the value of hard
currency fuel exports, including the rapid acceleration
that occurred after 1973. This indicates the general
validity of the underlying energy balance computations
in the model.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
SOVSIM: Simulation of
Private Consumption
Figure 5
IIIIIIIIIIIII I I I I I I I I I I I I I
The dominant variable feeding into the calculation of
net exports of fuels to hard currency countries is net
exports of oil. The results depicted in figure 7 show
that the changing Soviet capacity to export oil to the
West was captured by the model, although the
historical shift was sharper than the simulations
indicated. Nonetheless, there is no particular bias in
the projection errors over the full period.
Average Errors for Key Variables
The model's ability to replicate history can be best
measured by computing average simulation errors for
1965-77. No single error index can describe the
predictive power of the model reliably. Three conven-
tional indexes, however, taken together give a rounded
picture of how well the model projections match the
historical data:
Figure 4 SOVSIM: Simulation of New
Fixed Investment
? Mean Percentage Error (MPE) 6-The MPE will be
smaller to the degree that annual or individual errors
are of opposite signs and therefore offsetting.
? Mean Absolute Percentage Error (MAPS) 7-This
error index is useful because it counts individual errors
without regard to their signs and therefore does not
allow for offsetting effects. If the absolute value of the
MPE is close to the MAPE value and annual
percentage errors are generally of similar size, annual
errors tend to be of the same sign, perhaps signifying
some built-in model bias.
6 MPE = I E (Estimated-Actual)X 100
-t
N Actual
' MAPE = 1 (Estimated - Actual I X 100
N t Actual
Approved For Release 2008/09/12 : CIA-RDP08SO135OR000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
SOVSIM: Simulation of Net Figure 6
Exports of Fuel for Hard Currency
SOVSIM: Simulation of Hard
Currency Net Exports of Oil
Figure 7
IIIILIIIIIIII
? Root-Mean-Squared Percentage Error (RMSE) The RMSE like the MAPE, ignores the signs of
individual errors. However, an individual error re-
ceives more weight in calculation of this index accord-
ing to the square of its size. This index has preferred
statistical properties but can easily become distorted.
One or two historical figures lying far from the values
projected by the model can cause large RMSE values.
The average simulation errors for key production, end-
use, and foreign trade variables are summarized in
table 2. The main characteristic of the production and
end-use variables themselves is that they all show a
strong time trend. With the exception of agricultural
I E (Estimated-Actual)X 100
N Actual
output, they are not especially volatile, but they tend to
grow fairly steadily from year to year. The statistical
equations used in projecting the output variables
generally exhibit extremely good fits. Consequently,
average simulation errors tend to be small for this
group of variables. Of the 12 production and end-use
variables listed in table 2, seven have RMSEs of less
than 3 percent and only one has an error larger than
5 percent.
The trade variables are fundamentally more volatile
than the production and end-use variables. Part of the
explanation for this is that trade depends on conditions
outside the control of Soviet policymakers. Further-
more, those aspects of trade subject to control can
change dramatically from one year to another as
Soviet trade policies vary.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Average Simulation Errors
of Key Variables, 1965-77
Table 2 shows that average simulation errors for the
trade variables tend to be considerably larger than
those for the production and end-use variables. None-
theless, exports of oil and total fuels to the West,
nongrain imports from the West, and Communist
exports and imports are tracked fairly well over the
historical period-their average errors are generally
less than 20 percent. The trade balances show larger
errors because they are calculated as residuals of
balance relations in which the residual is much smaller
in absolute terms than the variables involved in the
calculation. As a consequence, small percentage errors
in these variables can result in large percentage errors
in the balance residuals. Just as in Western economies,
prediction of Soviet trade balances probably will
remain one of the areas of projection most prone to
error. Comparison of the RMSEs and MAPEs for the
balance variables also shows that aberrations in one or
two years are a major source of distortion in the RMSE
index.
Mean
Error
Mean
Absolute
Error
Root-
Mean-
Squared
Error
Gross National Product
-0.1
0.8
1.0
Total Consumption
-0.3
2.0
2.3
Total New Fixed Investment
0.2
2.0
2.4
Actual Agriculture Output
-0.5
3.6
4.3
Consumer Goods Output
-0.2
2.5
3.1
Industrial Materials Output -1.8 2.3
2.7
Other Industry Output
-0.6
1.5
1.7
Chemicals Output
5.5
8.0
9.0
Construction Output
-0.6
1.5
1.8
Machinery Output
1.8
3.2
3.6
Transport and Communi-
cations Output
-3.1
3.3
3.5
1.4
Hard Currency Balance of
Trade
- 13.1
74.4
Nonoil,Nongrain Imports
-4.4
12.0
18.7
Hard Currency Net Exports
of Fuel
9.1
20.0
28.0
Communist Balance of
- 156.9
165.8
316.6
Communist Nonfuel
Imports
0.0
3.7
6.8
Balance of Trade With
World
-143.5
158.8
392.8
Hard Currency Net Exports
of Oil
6.1
Hard Currency Net Exports 17.7 69.4
of Coal
Hard Currency Net 103.9
Exports of Gas
The relatively large average-percentage error in
projected gas exports primarily reflects the small
absolute errors in estimating gas exports in the early
years of the simulation period when they were practi-
cally negligible. Although less pronounced, a similar
situation exists in the case of coal exports. In both
cases, simulation errors in the later years were
substantially below the average.
A Short Run Forecasting Experiment
Historical simulations can give an unrealistic impres-
sion of the strengths or weakness of an econometric
model. They presume the existence of some data-for
both exogenous variables and structural parameters-
that would not actually be available when using the
model to look into the future. They also place in a
favorable light model specifications that are peculiar to
the historical period but that are not necessarily the
preferred way of viewing more general, long-term
trends.
As a second test of sovsIM, we have constructed a short
run forecasting experiment free of these defects. We
use data for 1960-75 to forecast Soviet economic
growth in 1976 and 1977, using projections of the
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
exogenous variables for 1976-77 based on growth rates
of preceding years. Comparison of the simulation
results for 1976 and 1977 with actual values for the
endogenous variables suggests the order of magnitude
of forecasting errors one might expect in using sovsIM
as a short run forecasting tool. It also highlights the
parts of the model in which accuracy of exogenous data
is crucial to forecasting performance and tests the
general stability of the underlying structural
specifications.
As part of the analysis, the net forecast error was
partitioned into (a) an error reflecting the approximate
nature of the model's structure as a description of the
Soviet economy and (b) an error reflecting the
approximate nature of the exogenous variables that
were extrapolated based on their 1960-75 values. All
endogenous variables were first forecast for 1976 and
1977 using parameters estimated from the full histori-
cal data through 1977 and actual observations of
exogenous variables also through 1977. Comparing
forecasts based on full historical information (through
1977) with actual observations for the endogenous
variables in 1976 and 1977 gives the "model error," the
forecasting error due solely to the model's approximate
specification. The "net error," the difference between
actual observations and forecasts using only data
through 1975, reflects both exogenous data error and
"model error." Therefore, the "data error" can be
estimated as the difference between the "net error"
and the "model error." In general, if the "model error"
and the "data error" have the same signs, they
reinforce each other. If not, they partially offset each
other.
Results for 1976
For 1976, the projections of the model (table 3) were
more accurate for the production and end-use variables
than for the trade variables, but this is to be expected
for reasons already discussed. Most of the errors are in
the range of 1 to 3 percent. In the group of production
and end-use variables, only value added in chemicals
exhibits a net error greater than 3 percent in the
experimental forecast. The estimate falls short of the
actual figure because of,roughly equal negative errors
due to model specification and data error. The branch
output forecasts show no consistent bias; the net error
in 1976 GNP is only 0.1 percent. For 1976, neither the
error due to model deficiencies nor the error due to
data errors consistently dominates as the source of
forecasting error.
Results for 1977
A comparison of net percentage errors for 1976 and
1977 (tables 3 and 4) indicates that the projections of
the simulated forecast tend to be further off the mark
in 1977. The difference is most pronounced for the
trade variables. Six of the 12 production and end-use
variables have greater net percentage errors in 1977
than in 1976 compared with seven out of the 11 trade
variables. Moreover, the differences in net percentage
errors between the two years are much greater for the
trade variables than for the production and end-use
variables. In 1977 the model error is generally lower
than the data error (both measured in absolute values)
for the trade variables. This means that forecasts of
external events affecting trade are a more severe
constraint than model structure on the ability of the
model to make accurate projections more than a year
ahead.
Nonetheless, even for forecasts two years beyond the
1960-75 period covered by the data base used in this
experiment, net errors were generally only about 1
percentage point worse for production and end-use
variables and around 10 to 15 percentage points worse
for most trade variables than the one-year forecast
errors. The 1977 results also vividly show that a model
will never be able to predict the kind of dramatic
turnaround in Soviet trade balances that occurred in
that year. As the error decomposition indicates, this
failure in 1976-77 is essentially a reflection of trade
variable sensitivity to shifts in behavior that are not
anticipated in either the underlying model parameters
estimated for 1960-75 or the crude extrapolations of
exogenous data also based only on preceding years.
SOVSIM Impact Analysis, 1978-85
All sovsiM projections depend on assumptions regard-
ing the exogenous variables and key coefficients in the
model. These variables and coefficients define the
internal and external economic, political, and techno-
logical environments. Domestic and foreign policies or
economically significant events can be described by
combinations of these variables and coefficients. The
reaction of sovsiM to hypothetical policy changes or
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
SOVSIM Forecasts of
Key Variables for 1976
Gross National Product
483.4
482.9
0.1
-0.5
0.6
Total Consumption
273.9
277.9
-1.4
0.2
-1.6
Total New Fixed Investment
108.6
111.7
-2.7
-2.5
-0.2
Actual Agriculture Output
69.6
68.7
1.3
1.1
0.2
Consumer Goods Output
62.3
61.2
1.8
-0.5
2.3
Industrial Materials Output
25.6
25.9
-1.3
-3.1
4.4
Other Industry Output
8.0
8.1
-0.4
-1.5
1.1
Chemicals Output
12.2
12.9
-5.7
-3.3
-2.4
Construction Output
33.9
33.9
-0.2
-1.7
1.5
Machinery Output
62.1
63.7
-2.6
-0.7
-1.9
Transport and Communications Output
46.9
45.6
3.0
0.4
2.6
Trade and Services Output
72.2
73.1
-1.2
-0.7
-0.5
Million US $
Hard Currency Balance of Trade
-5,106.0
-5,516.0
-7.4
-18.7
11.3
Nonoil, Nongrain Imports
11,416.0
12,051.0
-5.3
-13.4
8.1
Hard Currency Net Exports of Fuel
4,708.0
4,683.0
0.5
-7.2
7.7
Communist Balance of Trade
1,014.0
1,787.0
-43.3
-17.1
-26.2
Communist Nonenergy Exports
14,969.0
16,114.0
-7.1
-5.1
-2.0
Communist Nonfuel Imports
19,642.0
19,652.0
-0.1
-2.7
2.6
Balance of Trade With World
-4,119.0
-3,797.0
8.5
-19.2
27.7
Million Metric Tons
36.7
41.2
-11.0
-10.9
-0.1
11.1
8.8
26.8
-3.0
29.8
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
SOVSIM Forecasts of
Key Variables for 1977
Gross National Product
504.3
500.2
0.8
1.0
-0.2
Total Consumption
284.5
289.2
-1.6
3.2
-4.8
Total New Fixed Investment
114.1
115.0
-0.8
-2.3
1.5
Actual Agriculture Output
70.6
70.4
0.3
8.6
-8.3
Consumer Goods Output
64.8
63.3
2.4
-0.5
2.9
Industrial Materials Output
26.7
26.4
1.4
-1.0
2.4
Other Industry Output
8.2
8.2
0.3
-0.8
1.1
Chemicals Output
12.9
13.6
-5.3
-3.3
-2.0
Construction Output
35.3
34.7
1.6
-0.9
2.5
Machinery Output
66.1
67.5
-2.1
-0.6
-1.5
Transport and Communications Output
50.5
47.4
6.5
2.8
3.7
Trade and Services Output
74.6
76.1
-1.9
-2.0
0.1
Million US $
Hard Currency Balance of Trade
-5,354.0
-2,431
120.2
9.7
110.5
Nonoil, Nongrain Imports
13,291.0
10,229.0
29.9
-3.4
33.3
Hard Currency Net Exports of Fuel
6,260.0
6,180.0
-1.3
-17.9
19.2
Communist Balance of Trade
1,174.0
2,718.0
-56.8
-38.9
-17.9
Communist Nonenergy Exports
16,733.0
16,950.0
-1.3
0.2
-1.5
Communist Nonfuel Imports
22,739.0
20,368.0
11.6
5.3
6.3
Balance of Trade with World
-4,255.0
93.8
-4,636.0
-1,337.0
-3,259.0
Billion Cubic Meters
13.8 14.8
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
external events described by shifts in particular
variables and parameters can be extremely useful in
evaluating the path of the economy's adjustment to
such changes as well as in validating the model itself.
Such projections are called conditional forecasts.
Baseline Case
The potential impact of particular events or policy
changes is conventionally assessed by comparing two
model projections, a reference case and a case incorpo-
rating the given changes in terms of shifts in param-
eters or exogenous variables.
As a reference case, we developed a baseline projection
of Soviet growth in 1978-85 that assumes a
continuation of present Soviet policies and no change
in the trends of critical variables like participation
rates and weather:
? Exports of Fuels. Net exports of oil, coal, and gas for
hard currency are the residual from domestic produc-
tion after domestic deliveries and net exports to
Communist countries are covered. Net exports of oil to
Communist countries increase to 95 million metric
tons in 1980 and stay at this level through 1985.
? Other Hard Currency Trade. Hard currency exports
of commodities other than fuels grow at 9 percent a
year in real terms, and new drawdowns of medium-
and long-term credits increase at a real rate of 5
percent a year. Both rates are consistent with recent
trends. A floor was placed on the value of hard
currency imports other than oil and grain; their share
of GNP in any year was not allowed to fall below one-
half the 1977 figure.
? Allocation of Fuel Supplies. Sectors producing fuels
and power, and public and private consumption are
given priority when oil deliveries are insufficient to
meet the demands of all sectors. They are always
allocated 100 percent of their nominal oil demand.
? Production of Fuels. Oil production peaks at 590
million metric tons in 1980 and falls to 500 million
metric tons in 1985. This is at the high end of the
production range we have estimated.9 Gas output
grows at 6 percent a year and coal production at
2 percent.
? Defense Spending. The real value of Soviet defense
expenditures rises at an annual rate of 4 percent-
consistent with the trend over the last decade.
? Population and Employment. Total population
grows at 1 percent a year, but growth in the able-
bodied population slows dramatically through 1985
because of demographic factors. Agriculture's share of
the labor force falls from 24 percent in 1978 to 20
percent in 1985, while participation rates essentially
remain at their current high levels.
In the baseline case, these assumptions lead to a Soviet
oil shortage in the 1980s and a deceleration in Soviet
growth rates. The average annual rate of GNP growth
falls to 2.5 percent in 1981-85, more than 1 percentage
point below the average annual growth in 1976-80. The
Soviets would export oil to the West until 1981, after
which they would become net importers of Western oil.
Soviet net oil exports to the world would remain
positive through 1985, however, because of our as-
sumption that exports to Eastern Europe and other
Communist countries continue at a level of 95 million
metric tons in 1981-85.
Impacts of Three Hypothetical Shifts
To illustrate how sovsIm can be used to make
conditional forecasts, we have resimulated the model
over the 1978-85 period after changing-one at a
time-three of the assumptions underlying the
baseline solution:
? Labor force participation rates rise by 1985 by
1 percentage point for the able-bodied population and
by 2 percentage points for pensioners. The increase in
participation rates implies that the Soviets make
greater use of incentives and manpower regulations
than assumed in the baseline to improve the tight labor
situation over the next decade.
? Western economic growth is slower than the baseline
case assumes for the next decade, and therefore
nonfuel exports to hard currency countries grow at
only one-half the assumed baseline rate.
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
? Oil exports to Eastern Europe fall from 95 million
metric tons in 1980 to 45 million metric tons in 1985,
as the Soviets attempt to relieve economic pressures
associated with oil shortfalls in the early 1980s.
The more aggressive manpower policy would have an
immediate positive effect on GNP that would cumu-
late over the period (table 5), eventually adding a little
less than 1 percentage point to GNP by 1985. The
gains in private consumption would amount to about
half the gains in GNP-implying a stable share for
consumption in final demand. The small increase in oil
exports possible in the early years comes from the
excess of new domestic production over new domestic
demand resulting from the extra labor. The extra hard
currency imports are simply a reflection of the extra
hard currency earnings resulting from these oil ex-
ports. The improved labor situation would have little
impact on the hard currency and oil problems
projected for 1982-85. As table 5 shows, hard currency
imports in those years would not change from their
baseline levels, which are the minimum allowed under
the baseline conditions. The extra GNP during
1982-85 comes from two sources-primarily the in-
creased labor supply and to a much smaller extent
added energy production from domestic sources.
Slower Western growth would cut hard currency
imports immediately because of the lower-than-
baseline Soviet import capacity. The early effects of
trade on GNP are passed through very slowly because
of the small role that imported capital goods play in the
total Soviet economy. The large negative impact in the
later years is predominantly due to a reduced ability to
finance oil imports from the West, which is reflected in
lower capital utilization rates. Imports would be
restricted to almost 20 million metric tons below
baseline levels in 1985 because of the need to devote
increasingly scarce hard currency to keep other
imports from the West at least at minimum levels.
With other oil trade fixed, this means higher Soviet net
oil exports by the same amount. These shifts in trade
would improve the Soviet trade balance and simulta-
neously reduce the level of private consumption
compared with baseline levels. Consequently, con-
sumption as the residual end use would absorb a larger
share of the projected GNP fall-about two-thirds-
than the projected 50-percent share in the GNP
increase in the previous case where trade balances
were essentially stable between the baseline and
resimulation.
In the last case, it was assumed that Soviet oil exports
to Eastern Europe would fall only after pressures on oil
supplies began to mount in the early 1980s. The
potential impact of this policy shift is substantial-it
would add more than 1 percent to GNP by 1985-
because it would directly ease oil shortages, projected
in the baseline for the later years. The emergence of oil
shortages would be delayed, as the hard currency
import constraint in the baseline is now binding in only
the last three years.
The shifts in the pattern of trade are now more
complicated, however. When the Soviets are projected
in the baseline case to have sufficient hard currency
earnings to finance practically all import needs-as in
1981 and 1982-the oil released from export to
Eastern Europe is used domestically in place of oil
imports from the West and the hard currency savings
are diverted to finance more nonoil imports from hard
currency countries. Net oil exports to the world would
then be unchanged from baseline levels. When, in-
stead, the Soviets are projected to be under a hard
currency constraint in the baseline-as in 1983-85-
the oil diverted from Eastern Europe is used along with
baseline projections of Western oil imports to ease
internal oil shortages. In these years, imports of
Western oil and other goods would be unchanged from
the baseline, but net Soviet oil exports to the world
would fall by the amount of reduced oil exports to
Eastern Europe.
Our experience with sovslM-in historical analysis, in
forecasting experiments, and in impact analysis-
indicates that a Soviet macroeconometric model of this
kind can be a reliable and useful tool in studying Soviet
growth prospects. It provides a consistent framework
for investigating the effects of alternative sets of
analytical assumptions and for examining the linkages
between sets of interrelated issues.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Impact Analysis With SOVSIM:
Effects on Selected Economic Variables
Changes Due to Higher
Labor Participation Rates
GNP (Billion 1970 Rubles)
0
1.0
1.9
2.5
3.5
4.2
5.1
6.2
Consumption (Billion 1970
Rubles)
0
0.6
1.0
1.3
1.9
2.2
2.7
3.3
Net Hard Currency Oil
Exports (Million Metric
Tons)
0
0.4
0.7
0.9
0
0
0
0
Net Oil Exports (Million
Metric Tons)
0
0.4
0.7
0.9
0
0
0
0
Hard Currency Imports
Other Than Grain and Oil
(Billion US $)
0
0.1
0.2
0.2
0
0
0
0
Changes Due to Slower
Western Growth
GNP (Billion 1970 Rubles)
0
0
0
0
-1.8
-2.3
-2.7
-3.1
Consumption (Billion 1970
Rubles)
0
0
0
0
-1.1
-1.4
-1.6
-1.8
Net Hard Currency Oil Ex-
ports (Million Metric Tons)
0
0
-0.1
-0.2
10.7
13.4
15.2
17.0
Net Oil Exports (Million
Metric Tons)
0
0
-0.1
-0.2
10.7
13.4
15.2
17.0
Hard Currency Imports
Other Than Grain and Oil
(Billion US $)
-0.2
-0.5
-0.8
-1.3
0
0
0
0
Changes Due to Lower Oil
Exports to Eastern Europe
GNP (Billion 1970 Rubles)
0
0
0
0
0.3
4.8
6.5
8.1
Consumption (Billion 1970
Rubles)
0
0
0
0.1
0.3
3.0
4.1
5.1
Net Hard Currency Oil
Exports (Million Metric
Tons)
0
0
0
10.0
18.1
0
0
0
Net Oil Exports (Million
Metric Tons)
0
0
0
0
-1.9
-29.0
-39.3
-48.0
Hard Currency Imports
Other Than Grain and Oil
(Billion US $)
0
0
0
1.5
2.9
0
0
0
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Applications of sOvsIM to growth studies should help
improve our understanding of both the long-term
prospects for Soviet economic development and the
methods we use to analyze them. Therefore we tend to
focus on the use of sovsim in the next phase of our
research. This means applying the present version of
the model to analysis of trends in such areas as labor
supply, energy, and foreign trade, and their impacts on
Soviet growth potential in the 1980s. These applica-
tions studies should serve as a comprehensive test of
the usefulness of a Soviet macroeconometric model in
improving our perceptions of the range of Soviet
growth options, Soviet flexibility in the face of
expected shifts in resource growth patterns, and the
general interrelatedness of issues facing Soviet
policymakers.
In the long run, however, the usefulness of sovsim can
be enhanced by further development, especially in
several specific areas. Production functions play a
crucial role in any Soviet model and sovsim now con-
ventionally employs a straightforward Cobb-Douglas
specification. Considerable research has been done
since 1970 on the application of constant elasticity of
substitution (CES) production functions to highly
aggregated Soviet data. A potentially fruitful area for
future research is the application of CES or other more
complex specifications to the description of Soviet
production on a sector basis. These changes would be
aimed at an improved depiction of the substitutability
of capital and labor, and its variation across producing
sectors. However, applications of more sophisticated
techniques to highly disaggregated data are certain to
be plagued by greater data shortcomings than aggre-
gate analysis, because of the lack of offsetting error
possibilities. Such problems are not fatal to
disaggregated estimates of the less sensitive Cobb-
Douglas function, but they will become important in
estimating more sensitive functional forms for highly
disaggregated sectors.
There are also several other areas of sovsim that
demand further development. More detail must be
given to the description of production in the agricul-
tural sector. At least a partial behavioral foundation
must be established for projections of private consump-
tion. It may be possible to explain changes in participa-
tion rates through changes in real incomes. Also, the
process of adjustment to intersectoral disequilibriums,
which is the focus of current SRI-WEFA research,
may be amenable to simulation within the framework
of a macroeconometric model and will be incorporated
in our own work if present experiments prove
successful.
Of course, any modeling effort uses much outside
information to describe processes not embodied within
the model itself. In sovsIM, questions of labor supply
and demographic patterns, improvements in total
factor productivity at the sector level, and gains in
energy conservation are all handled outside the model,
but they obviously have a strong bearing on the
character of any analysis conducted with the model.
Improvements in our understanding of such issues is a
necessary ingredient in upgrading the quality of Soviet
growth analysis in general.
It is unlikely, though, that the near future will see
much of an improvement in our ability to analyze the
planning process.explicitly. Ideally a Soviet
macromodel should endogenize the reaction functions
of the planners and thus provide a description of the
planning process itself, including the effects of plan on
performance and the feedbacks of performance on
future plan formulation. The failure of Soviet
macroeconometric models to meet this requirement
has been the center of recent criticisms of current
approaches. 10
Given our limited theoretical understanding of planner
decisionmaking, the sharp discontinuities possible in
planning behavior, and the insufficient set of historical
precedents that exist, it is not likely that a purely
statistical solution to modeling planning behavior will
be found. A second best procedure may be combin-
ing, within a series of analyses or conditional forecasts,
the use of a macroeconometric model with sound
judgment on how underlying policies might change as
decisionmakers adapted to events projected by the
model. It is this approach that we will continue to
follow in our research.
10 See Richard Portes, Macroeconometric Modelling of Centrally
Planned Economies: Thoughts on SOVMOD I (Harvard Institute
for Economic Research, May 1978).
- Approved For Release 2008/09/12 : CIA-RDP08SO135OR000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
This appendix describes the structural equations in
sovsIM. The equations are divided into seven blocks,
corresponding to the groups of endogenous variables
discussed in the condensed version of the model. Brief
discussions of each block and tables listing the
equations are provided below. For regression equa-
tions, the RZ statistic is adjusted for degrees of freedom
while t-statistics appear in parentheses below esti-
mated parameters.
Production Block (Table Al)
There are 14 endogenous variables describing sector
production levels in this block and they are influenced
by eight exogenous variables. Value added in agricul-
ture is estimated in a two-step procedure. First, normal
agricultural output is estimated based solely upon
land, labor, and capital inputs. Then, actual output is
estimated from normal output and current indexes of
temperature and rainfall. Production functions for
agriculture and the eight nonagriculture, nonenergy
sectors were estimated using a constant-returns-to-
scale' Cobb-Douglas specification, implying an elas-
ticity of substitution among inputs equal to unity.
Other production-function specifications gave gener-
ally inferior statistical results. The nonagriculture
production functions were estimated using generalized
least squares (GLS) since results using ordinary least
squares (OLS) showed strong autocorrelation of the
residuals.
' The constant-returns-to-scale assumption requires only the capital
coefficient to be estimated statistically. The labor coefficient is
simply one minus the capital coefficient and it does not have an
associated t-statistic.
Because we expect over the next decade substantial
reductions from historical levels in the marginal
products of labor and capital in the energy sectors, we
have not used production functions estimated from
historical data for these sectors. Value added in the
energy sectors is calculated instead from scale relation-
ships that apply value-added weights to exogenous
indexes of gross output. Gross outputs of the energy
sectors are exogenous to the model, but provision exists
for computing marginal changes from reference output
levels in response to marginal shifts in labor and capital
inputs to these sectors.
The capital stocks for the machinery, chemicals, and
oil-producing sectors are separated into domestically
produced and imported components, although it was
not possible to estimate separate coefficients for each
type of capital.
Approved For Release 2008/09/12 : CIA-RDP08S01350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDPO8SO135OR000100110001-4
Equation List for the
Production Block
Equation
Number
Dependent
Variable
Equation
R 2
Durbin-
Watson
Statistic
Estimation
Method
1
Normal
agriculture
output
log (normal agriculture output) =
- 5.08 + 0.01 X time + 0.738
(-16.07) (12.82)
X log (agriculture employment) + 0.111
0.974
1.10
Ordinary
Least
Squares
(15.98)
Actual
agriculture
output
actual agriculture output =
normal agriculture output
0.969 + 0.140 X precipitation index + 0.519
(78.71) (5.10) (2.52)
0.699
1.32
Ordinary
Least
Squares
Construction
output
log (construction output) _
- 4.49 + 0.779 X log (construction employment)
(-54.77)
+ 0.221 X log (capital in construction sector)
(17.14)
0.978
1.59
Generalized
Least
Squares
4
Transport and
communications
output
log (transport and communications output) =
-1.87 + 0.173 X log (transport and
0.989
2.04
Generalized
Least
Squares
(-15.11)
communications employment) + 0.827
(31.74)
X log (capital in transport and communications sector)
5
Trade and
log (trade and services output) =
0.958
0.72
Generalized
services output
Least
- 5.33 + 0.836 X log (trade and services employment)
Squares
(-69.19)
+ 0.164 X log (capital in trade and services sector)
(11.97)
6
Industrial
log (industrial materials output) =
0.995
1.15
Generalized
materials
Least
output
- 3.10 + 0.538 X log (industrial materials employment)
Squares
(-40.71)
+ 0.462 X log (capital in industrial materials branches)
(30.36)
Approved For Release 2008/09/12 : CIA-RDPO8SO135OR000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Table Al (Continued)
Equation List for the
Production Block
Equation
Dependent
Equation
R 2
Durbin-
Estimation
Number
Variable
Watson
Statistic
Method
7
Consumer
goods
output
log (consumer goods output) _
- 2.97 + 0.635 X log (consumer goods employment)
0.975
1.39
Generalized
Least
Squares
(-17.84)
+ 0.365 X log (capital in consumer goods branches)
(12.73)
8
Machinery
output
log (machinery output) _
- 2.57 + 0.452 X log (machinery employment)
0.988
1.65
Generalized
Least
Squares
(-16.63)
+ 0.548 X log (domestic capital in machinery branch
(20.08)
9
Chemicals
output
log (chemicals output) =
- 2.92 + 0.507 X log (chemicals employment)
(-29.07)
+ 0.493 X log (domestic capital in chemicals branch
(23.65)
0.987
1.28
Generalized
Least
Squares
10
Other
industry
output
log (other industry output) =
-4.22 + 0.690 X log (other industry employment)
(-41.13)
+ 0.3 10 X log (capital in other industry)
(16.87)
0.982
1.00
Generalized
Least
Squares
I I
Gas output
Gas output = 1.798 X(gross gas output
197.9
12
Oil output
Oil output = 7.549 X (gross oil output
353.4
13
Coal output
Coal output = 4.945 X gross coal output
624.1
14
Electric power
Electric power output = 7.268 X (gross electric power output
output
` 740.9
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Consumption Block (Table A2)
The consumption block consists of nine endogenous
variables and is affected by 11 exogenous variables.
GNP is computed as the sum of value added in the 13
producing sectors. Total private consumption is calcu-
lated as the residual claimant on GNP once the other
end uses (investment, public consumption, and the
balance of trade) are deducted. Reduced-form equa-
tions that allow for both demand and supply influences
were estimated for per capita food consumption and
per capita durable goods consumption. In both equa-
tions the demand influence is represented by national
income per capita. The supply influence in the food
consumption equation is lagged agricultural output per
capita, and in the durables equation it is machinery
output per capita. Unallocated production as a GNP
residual is estimated as a function of time and serves to
close the GNP accounts on a sector-of-origin basis.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Equation List for the
Consumption Block
Equation Dependent Equation R 2 Durbin-
Number Variable Watson
Statistic
Gross national Gross national product =
product industrial materials output + machinery output + chemicals output
+ consumer goods output + other industry output
+ gas output + oil output + coal output + electric power output
+ construction output + transport and communications output
+ trade and services output + agriculture output
+ (unit wage of military X military manpower)
2 Total Total consumption =
consumption gross national product - (total new fixed investment + capital repair
+ livestock investment) - (government administration share
+ government R&D share) X gross national product
- (nonpersonnel defense spending + (unit wage of military
X military manpower)) - (0.001 X (ruble balance of trade)
+ invisibles ruble/dollar ratio
X earnings on invisibles except official transfers) - end-use residual
3 Consumption Consumption per capita =
per capita total consumption
population
4 Food log (food consumption per capita) = 0.995 1.77
consumption
per capita -0.937 + 0.652
(-16.96) (37.56)
X log(gross national product
%\ population J
+ 0.132 X log/actual agriculture output(.,)
(3.57) 1\ population
5 Food Food consumption =
consumption food consumption per capita X population
6 Durables log (durables consumption per capita) = 0.987 0.56
consumption
per capita - 1.69 + 0.868 X log gross national productl
(-2.15) (1.69) population
+ 0.857 X log achinery output
(2.67) ` population
7 Durables Durables consumption =
consumption durables consumption per capita X population
8 Nonpersonnel Nonpersonnel defense spending=
defense (I + growth in nonpersonnel defense spending)
X nonpersonnel defense spending(_,,
9 Sector-of-origin Sector-of-origin residual = 0.997 2.36
residual 3.00 + 1.81 X time
(4.18) (56.43)
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Investment Block (Table A3)
Investment calculations involve 19 endogenous
variables and are affected by four exogenous variables.
New fixed investment is set by the supply of capital
goods and does not reflect demand considerations.
Deliveries to final demand from construction and
machine building are estimated by scaling value-added
figures that are derived from production functions.
Scaling coefficients are calculated from Soviet input-
output data. The claims of defense, capital repair,
consumer durables, and exports are deducted from
gross construction and machinery output to give net
domestic production available for new fixed invest-
ment. In addition, the calculation of construction
available for new fixed investment involves an adjust-
ment to maintain consistency between the estimated
new fixed investment component of the GNP accounts
and the estimated output of machinery and construc-
tion available for investment.' Total new fixed invest-
ment is then distributed among the 13 producing
sectors and housing according to given shares reflect-
ing historical trends or assumed policies. Investment
models using endogenous share calculations have also
been investigated, but they have a weak statistical
basis and generally lead to unacceptable projections in
long-run growth analysis.
2 Construction and machinery output data are not fully consistent
with independent figures on investment. To maintain correct
accounting in the model, we have developed an adjustment applied to
the construction series, where consistency is established with the
smallest distortion of original sector output figures.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Equation List for the
Investment Block
Equation Dependent Equation
Number Variable
Machinery for Machinery for new fixed investment =
new fixed
investment 1.238 X machinery output - (0.5 X capital repair
Durbin-
Watson
Statistic
2 Investment Investment adjustment factor = 0.893 0.98
adjustment
factor 0.877 + 0.0122 X time
(32.40) (10.03)
3 Construction Construction for new fixed investment =
for new fixed
investment 2.751 X investment adjustment factor X construction output
- (0.5 X capital repair + military construction)
4 Total new Total new fixed investment =
fixed
investment machinery for investment + construction
3
for investment +E (net imported capital formation in the ith sector
i=1
+ 0.06 X imported machinery in the ith sector)
5-17 New fixed New fixed investment in the ith sector =
investment
in the ith the ith sector share of total new fixed investment
sector X total new fixed investment
19 New fixed New fixed investment in the housing sector =
investment
in the the housing share of total new fixed investment
housing X total new fixed investment
sector
24
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Capital Formation Block (Table A4)
Capital formation in sovsiM is described by 32
endogenous variables and is influenced by one exoge-
nous variable. This process is concerned with the
increase in the productive capital stock that comes
from commissionings of new plants, which in turn
depends on expenditures on capital goods. The stock of
productive capital in a particular producing sector
during a given year is computed from an accounting
relationship involving last year's stock and last year's
net formation of productive capital. Net capital
formation equations are estimated for each producing
sector based on assumed depreciation rates and invest-
ment levels in the last two time periods. A plan cycle
variable representing the Soviet tendency to start a
wide range of capital projects during the early plan
years and to push projects to completion in the later
years was found to be a significant explanatory
variable in the gas and coal sectors and was therefore
also included in those net capital formation equations.
The estimated net capital formation equations reflect
two factors: the general failure of Soviet capital stock
census data to be fully consistent with data on annual
investment expenditures, and the variable gestation
periods and patterns of unfinished investment across
producing sectors. For the machinery, chemicals, and
oil sectors, distinctions are made between capital
produced domestically and capital imported from hard
currency and other Communist countries. For im-
ported equipment, we simply assume a one-year
average lag between the import of capital goods and
their inclusion in the stock of productive capital.
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Equation List for the
Capital Formation Block
Equation Dependent Equation R 3 Durbin-
Number Variable Watson
Statistic
Capital in the Capital in the agriculture sector =
agriculture
sector net capital formation in the agriculture sector t_11
+ capital stock in the agriculture sector t_,1
2-10 Capital in the Capital in the ith sector =
ith sector
net capital formation in the ith sector t_1l
+ capital stock in the ith sector (_,)
11-13 Domestic capital Domestic capital in the jth sector =
in the jth sector
14-16 Imported capital Imported capital in the jth sector =
in the jth sector
net imported capital formation in the jth sector t_~1
+ imported capital in the jth sector t_1l
17 Net capital Net capital formation in the agriculture sector
formation in + 0.05 X capital in the agriculture sector =
the agriculture
sector / investment in the agriculture sector
0.0312 + 0.796 X (`+ investment in the agriculture sector t_~)
(0.05) (17.86)
2
18 Net capital Net capital formation in the construction sector + 0.968 2.22
formation in 0.06 X capital in the construction sector =
the construction
sector / investment in the construction sector
0.0348 + 1.11 X + investment in the construction sector t_~1
(0.26) (22.19) 2
19 Net capital Net capital formation in the transport and communications sector + 0.025 0.926 1.99
formation in X capital in the transport and communications sector =
the transport and
communications investment in the transport and communications sector
sector 1.11 + 1.03 XC investment in the transport and communications sector t_il
(2.01) (14.19) 2
20 Net capital Net capital formation in the trade and services sector + 0.02 X 0.866 1.57
formation in capital in the trade and services sector =
the trade and
services sector / investment in the trade and services sector
- 0.873 + 1.01 X 1\+ investment in the trade and services sector t_I1 J
(-0.72) (10.22) 2
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Approved For Release 2008/09/12 : CIA-RDP08SO1350R000100110001-4
Table A4 (Continued)
Equation List for the
Capital Formation Block
Equation Dependent Equation
Number Variable
Durbin-
Watson
Statistic
21 Net capital Net capital formation in the industrial materials branches + 0.845 2.18
formation in 0.045 X capital in the industrial materials banches =
the industrial
materials investment in the industrial materials branches
branches -0.459 + 0.679 (+ investment in the industiral materials branches
(-0.99) (9.40) 2
22 Net capital Net capital formation in the consumer goods branches 0.904 1.38
formation in + 0.05 X capital in consumer goods branches =
the consumer ( investment in consumer goods branches
goods branches 0.209 + 0.935 X1} investment in consumer goods branches(_,)
(0.85) (12.29) \ 2
23 Net domestic Net domestic capital formation in the machinery branch + 0.05 0.901 2.36
capital X domestic capital in the machinery branch =
formation in investment in machinery goods branch \
the machinery 1.03 + 0.985 X( investment in machinery goods branch(_1~ J
branch (2.34) (12.08)
24 Net imported Net imported capital formation in the machinery branch + 0.06
capital X imported capital in the machinery branch =
formation in n nc V .,.