MODELING THE SOVIET ECONOMY: SOVSIM AFTER SIX YEARS
Document Type:
Collection:
Document Number (FOIA) /ESDN (CREST):
CIA-RDP86T00591R000300340001-4
Release Decision:
RIPPUB
Original Classification:
C
Document Page Count:
31
Document Creation Date:
December 22, 2016
Document Release Date:
March 22, 2010
Sequence Number:
1
Case Number:
Publication Date:
July 1, 1985
Content Type:
REPORT
File:
Attachment | Size |
---|---|
![]() | 1.42 MB |
Body:
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Directorate of Confidential
Intelligence
Modeling the Soviet Economy:
SOVSIM After Six Years
A Technical Intelligence Report
Confidential
SOV 85-10131
July 1985
488
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Directorate of
Intelligence
SOVSIM After Six Years
Modeling the Soviet Economy:
Economic Performance Division, SOYA,
Office of Soviet Analysis. Comments and queries are
welcome and may be directed to the Chief,
This paper was prepared by
25X1
25X1
25X1
Confidential
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Modeling the Soviet Economy:
SOVSIM After Six Years
Although econometric modeling of market economies has been an accepted
analytic tool since the late 1960s, the utility of similar models of centrally
planned economies was not demonstrated until recently. SOVSIM, the
Central Intelligence Agency's model of the Soviet economy, was developed
in the late 1970s and has been revised, refined, and extended to increase its
capabilities and to improve its performance. This report, which describes
the latest version of the model and its operation and capabilities,
SOVSIM's primary purpose is to support studies of growth prospects for
the Soviet economy, giving special consideration to how certain policy
decisions might affect these prospects over the next decade. Consequently,
the model is designed to facilitate analysis of the impact of policy shifts and
contingent events in areas like investment, energy, and foreign trade. F-
25X1
25X1
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 1
SOVSIM: Summary of the Soviet Economy
Producing sectors
Industry
Capital
stock
Output uses
Consumption
Foreign trade and
finance
Hard currency
payments
position
Utilized
capital
Agriculture
Construction
Transportation
and communications
Domestic trade
and services
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Overview
Information available
as of 15 June 1985
was used in this report.
Modeling the Soviet Economy:
SOVSIM After Six Years 25X1
SOVSIM is a model of the Soviet economy. Using information and
assumptions about the structure of the economy, resource availability, and
economic policies, it can simulate the economy's operation and project its
growth. It is designed primarily to make medium- to long-term projec-
tions-typically three to 10 years beyond the current year. Other tech-
niques are more appropriate for assessing near-term economic prospects.
Soviet energy production prospects.
We have used SOVSIM since 1979 to examine Soviet economic issues of
the 1980s. Besides developing baseline views of economic prospects, we
have also used the model as a scenario analysis tool by altering equations or
data assumptions to reflect new policy initiatives or alternative technologi-
cal developments. This analysis permits us to study questions such as the
impact on future economic performance of postulated resource shifts into
or out of defense, of possible Western credit constraints, or of alternative
consistency than would be possible using alternative approaches.
The model also provides other distinct research and analytical benefits:
? It gives a discipline and organization to the central storage of large
amounts of data on the Soviet economy that facilitates their retrieval and
use.
? Construction and maintenance of its data base and multidisciplinary
analysis with the model encourage greater interaction among analysts
than would probably occur otherwise.
? Its use as an analytical tool forces assumptions to be made explicit so that
they can be examined and debated.
? The analytical product flowing from the model has much greater internal
not likely to change significantly in the period to be projected.
The model consists of over 400 equations, each describing a particular
facet of the Soviet economy. Some are merely definitional; others express
behavioral and technological relationships among elements of the economy.
The specific form of each behavioral and technological equation is
periodically estimated econometrically from relevant historical data. Pro-
jections based on such equations assume that the historical relationship is
SOVSIM groups the individual equations according to the main compo-
nents of the Soviet economy. Resource inputs are described by equations
detailing the availability and use of labor and the flow of investment goods
Confidential
SOV 85-10131
July 1985
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
into the capital stock. Still others lay out energy production and demand to
allow for possible shortages that could constrain the use of capital stock. E
The producing sectors are described by another group of equations.
Industrial production is disaggregated into specific sectors, such as ma-
chine building and industrial materials. Output of oil, gas, coal, and
electricity is treated separately. Agricultural production is disaggregated
into grain, nongrain crops, and livestock products. Output of specific
services, such as construction, transportation, and trade, is also considered.
Foreign trade and finance are treated by other equations. These account
for trade with Eastern Europe, the West, and the Third World and
describe financial transactions with hard currency countries that determine
the Soviet hard currency payments and debt position.
Finally, the uses of output are described through a set of equations that
represent the national economic accounts. These pull together all of the
production and trade results and describe the uses to which the goods and
services available to the economy are put: consumption, investment in
either new plant and machinery or repairs to existing production facilities,
defense, and exports to pay for imports or to reduce foreign debt.
Most of the equations in the model are linked. The results obtained,
therefore, are more comprehensive and internally consistent than those
that would be obtained by analysis of individual sectors of the economy,
and they include both the direct and indirect effects of events affecting the
economy. This makes SOVSIM especially useful for studying how one
aspect of the economy, say energy development, might affect another, such
as the hard currency payments position. Such links allow errors in the
description of one sector to be transmitted to related sectors and therefore
must be handled with care; however, they reflect the interrelationships that
actually exist among the various components of the Soviet economy. =
SOVSIM also captures dynamic relationships operating in the Soviet
economy by linking the flows of resources, outputs, and other economic
values over a number of years. For example:
The active capital stock in any year depends on investment in machinery
and energy made in previous years.
? The hard currency debt at any time is a function of past borrowing and
repayments.
Confidential vi
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
These dynamic relationships facilitate examination of the growth prospects
of the Soviet economy and the long-run consequences of events and
policies. The calculations for each individual year in turn provide a
snapshot of the economy's future size and structure.
The limitations of such models should not be ignored. Results can be no
better than the quality of the underlying structure and data. If little
institutional, technological, or political change is expected in the economy,
a model based strictly on historical economic experience may yield
sufficiently good results. If change is expected, however, the structure of
the model and the required data inputs must be adjusted to reflect where
the economy is going, not where it has been. Only then will it yield
projections that can serve as useful descriptions of possible future courses
for the economy.
SOVSIM can be evaluated in several different ways, including ex post and
ex ante projections of historical periods and comparison with other models.
The model well describes ex post the trend of a number of Soviet economic
activities over the 1975-83 period. Estimates of annual GNP (in 1970
rubles) average less than 0.5 percent below actual values. Corresponding
average errors in estimates of yearly output in various industrial, agricul-
tural, and service sectors and annual consumption and investment are no
more than 3 percent. Errors for the trade sector are greater, with grain im-
port estimates averaging 6 percent below and balance-of-payments mea-
sures averaging up to 11 percent above actual values for 1975-83.
The record of ex ante projections of Soviet economic growth for the 1981-
84 period is varied because of the wide fluctuations in actual Soviet
performance. Specifically, projections of average industrial growth rates
for 1983-85, based in part on the poor record of the preceding three years,
turned out to be too low. On the other hand, projections of average
agricultural growth have improved in recent years as actual performance
moved back toward trend levels. Projections of average consumption
growth rates have been consistent, if on, the low side, while those for
average investment growth have varied more widely; the most recent
projection is close to the actual average. In recognition of these differences
between past model projections and actual values, present analysis of
economic growth is conditioned on a variety of alternative scenarios. This
approach yields a better appreciation of the range of possible outcomes and
their implications for the economy.
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
SOVSIM projections can be compared with those made by other models of
the Soviet economy. Projections made with SOVMOD, the most important
of these models, are more optimistic-largely because of more favorable
assumptions regarding productivity, investment, and agricultural and
energy output. SOVMOD, for example, has assumed that the higher labor
and capital productivities of the late 1960s and early 1970s reflect likely
trends for the rest of the 1980s. SOVSIM projections, on the other hand,
assume that these higher productivities represent the upper end of a range
of likely levels. The track records for both models are equally mixed.
Differences between the two models are not necessarily bad, however, for
they sharpen the debate about the performance of the Soviet economy and
help to suggest limits to development possibilities.
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
SOVSIM: A Description of the Soviet Economy
1
Energy in the Production Process
2
Analysis of Historical and Current Developments
10
3.
Comparison of SOVSIM Projections for 1981-85 With Actual
Values for 1981-84, Average Annual Growth Rates
16
4.
SOVSIM and SOVMOD Assumptions and Results, Average
Annual Growth Rates, 1985-89
17
5.
Performance in Projecting 1982 Economic Activities
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Modeling the Soviet Economy:
SOVSIM After Six Years P
SOVSIM: A Description of the Soviet Economy
The Soviet economy is the second largest in the
world-some 150 million persons working at hun-
dreds of thousands of locations to produce annually
over 20 million different goods and services worth
almost $2 trillion. SOVSIM describes only the most
important general features of this economy. It groups
all Soviet economic activities into 15 major producing
sectors, treating the output of each as a "single good,"
and then allocates this output to four major uses-
consumption, investment in future growth, defense,
and exports (inset).
Resource Inputs
For most sectors of the economy, output is estimated
from the labor and capital available in the sector by a
"production function," an equation that describes the
relationship between the quantities of these inputs
required to produce a unit of output. Production may
be reduced, however, by energy shortages.
Producing Sectors
Industry
Machinery
Chemicals
Ferrous metals
Nonferrous metals
Construction materials
Forest products
Consumer goods
Energy
Coal
Gas
Petroleum
Electric power
Agriculture
Construction
Transportation and communications
Trade and services
Labor. The size of the Soviet labor force is derived
from population and labor growth estimates prepared
by the Center for International Research of the US
Bureau of the Census (figure 2). Projections of mili-
tary manpower, provided by CIA military analysts,
allow the model to estimate the civilian labor force.
Deduction of agricultural employment from Census
projections then yields the nonagricultural work force.
Annual increases in nonagricultural employment are
allocated to each relevant sector on the basis of the
marginal investment share and the marginal capital-
labor ratio believed to hold in that sector.
Capital. The Soviet machine-building and construc-
tion sectors are the sources of the stock of plant and
equipment that are used to produce goods and services
for the entire economy (figure 3). Some of their output
is devoted to military needs. Some is sold to consum-
ers as durable goods. Some machinery is also export-
ed, although these exports are more than balanced by
imports of equipment from other countries. What
remains is used for investment in the Soviet economy.
Uses of Output
Consumption
Investment in future growth
Defense
Exports
Some investment is used to repair machinery and
facilities, and some replaces equipment that is being
retired because of old age. Most investment, however,
is used to augment the stock of productive machinery
and buildings and hence to increase its ability to
produce goods and services in the future.
Estimates of military procurement and construction,
machinery imports, and output of durable consumer
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 2
SOVSIM: Labor Sector Flowchart a
Civilian Military
labor force manpower
Agricultural Nonagricultural Investment
employment employment shares
Nonagricultural Sectoral
sectoral
employment
a Variables in bold boxes are inputs to the model.
Other variables are computed by the model.
goods are made by CIA analysts. Estimates of ma-
chinery exports have averaged about 6 percent of final
sales of machinery in recent years. These estimates
are subtracted from the value of machinery and
construction output, as computed in the model's in-
dustrial production block, to arrive at investment.
This is then allocated among the sectors of the
economy in the model on the basis of recent trends,
the latest annual or five-year plan, or the expected
relative priorities of investing in the different sectors.
Scenarios may involve various allocations, each repre-
senting a different set of priorities, to uncover the
economic potential and possible bottlenecks inherent
in the Soviet economy. It is also possible to turn the
model around to examine instead the consequences of
alternative-assumed investment growth rates on de-
fense procurement or consumer durable production.
Energy in the Production Process. Relative energy
availability in a given sector may affect output in that
sector. If energy shortages occur, resources are not
fully utilized and output is reduced. For this reason,
the determination of an overall energy balance and
Figure 3
SOVSIM: Capital Sector Flowchart a
Military
needs
Machinery
output
L Total
Machinery Investment
imports I
Machinery
exports
Durable
IN. goods
Capital Sectoral A
r investment
r- air investment
Sectoral
shares
Sectoral
retirement
Sectoral
capital
formation
Sectoral
capital
stock
a Variables in bold boxes are inputs to the model.
Other variables are computed by the model.
the distribution of energy to productive sectors, con-
sumers, and the export market are critical factors.
Projections of energy exports, based on both domestic
needs and likely trade policies, are supplied to
SOVSIM by CIA energy analysts (figure 4). From
these and computed production of energy, SOVSIM
calculates the availability of energy to domestic users.
CIA analysts also estimate the energy needs of the
various sectors of the Soviet economy as well as the
domestic distribution of available energy. This infor-
mation suggests the relative satisfaction of sectoral
needs.
25X1
25X1
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 4
SOVSIM: Sectoral Fuel Deliveries and
Value Added Impacts
Net fuel Thermal
and power 4 electricity
exports output
Gas and
power deliveries
Gas and power
needs
Capital Value 10. stock added
Oil and coal
needs
Nonthermal
electricity
output
Fuel and power
p availability
a Variables in bold boxes are inputs to the model.
Other variables are computed by the model.
Because current projections reveal no indication of
electricity or gas shortages, SOVSIM assumes that
availability of these energy sources meets require-
ments and equates their use to total production less
net exports.' Evidence does indicate the possibility of
oil and coal shortfalls through 1995. The model allows
for the possibility of such shortages and rationing
among the various economic sectors.
The Production Process
SOVSIM groups economic output into 15 different
sectors. For each, output is measured as the total
value added by the sector to the products and service
it produces.' In addition, physical output is estimated
for the energy sectors, grain, and meat.
' There have been sporadic electric power outages because of
weather conditions, fuel transport bottlenecks, and other transitory
phenomena. There is, however, no shortage of electric power
capacity, and more capacity is expected to be installed over the
' Value added is the difference between the value of the sectoral
output and that of its material and energy inputs. It constitutes the
value added to these raw materials by the capital and labor used in
period that is to be analyzed with SOVSIM.
the sector to produce output.
Oil and coal 4/ Oil and coal
deliveries shortfall shares
Agriculture. SOVSIM models Soviet grain, nongrain
crops, and livestock production separately (figure 5).
Grain production is especially important:
? As a direct source of food for the Soviet consumer.
? As feedgrain that, along with available roughages
and nongrain concentrates, determines the growth
of livestock herds and hence the output of meat and
other livestock products.
Soviet leaders generally rely on grain imports to make
up for shortfalls in domestic supply, but such imports
require hard currency payments. A good grain har-
vest, therefore, not only facilitates food production,
which raises consumer well-being, but also eases the
demand for hard currency.
In SOVSIM, Soviet grain output is projected by trend
grain production as estimated by CIA analysts.' Grain
25X1
25X1
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 5
SOVSIM: Agriculture Sector Flowchart a
Trend grain Grain
output imports
Grain
availability
Livestock
herd
Livestock
output
a Variables in bold boxes are inputs to the model.
Other variables are computed by the model.
Value of
agriculture
output
Nongrain Agricultural Agricultural
concentrates capital employment
Agricultural
land
Value of nongrain
crop output
imports are projected by the model from an historical-
ly derived equation relating imports to domestic pro-
duction and poultry stocks. CIA analysts also project
nonfeedgrain uses on the basis of expected supply and
demand. Feedgrain availability is estimated as a
residual. Over the long run, changes in grain stocks
are assumed to be small on average.
Meat output is computed separately for beef, pork,
and poultry from the historical relationship between
these commodities and feed supplies, livestock herd
size, and changes in the herd size. Estimates of other
meat products are provided by CIA agricultural
analysts. SOVSIM projects nongrain crop output
from the historical relationship between such output
and agricultural capital and employment and land
sown to nongrain crops.
Expected prices for grain and meat products are used
to arrive at an estimate of the total value of grain and
meat production, respectively. The value of total
agricultural output is computed by adding these to the
value of nongrain crop output. Deducting the estimat-
ed value of agricultural purchases from other sectors
of the economy yields agricultural value added.
Energy. Oil output is by far the most important
product of the energy sector. It declined slightly in
1984 and may well continue to decline over the next
decade, creating the possibility of energy shortages.
The Soviets themselves are aware of this danger. They
are expanding gas production as rapidly as possible,
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 6
SOVSIM: Oil Output Flowchart
depletion of resources at existing wells, these then
determine total output in western Siberia. Oil output
in other regions is estimated by CIA energy analysts.
Exploration Development
costs drilling
Depletion of
existing West
Siberian wells
Drilling West
outside Siberian
West Siberia output
Output
outside
West Siberia
a Variables in bold boxes are inputs to the model.
Other variables are computed by the model.
but gas cannot easily replace oil in many uses. The
Soviets are also investing even more heavily in the oil
sector to try to sustain output. Maintaining oil pro-
duction, however, is becoming increasingly difficult,
as output in the large fields of western Siberia begins
to taper off. Moreover, no other readily accessible
large fields are being developed. The cost of develop-
ing new production, therefore, is increasing and plac-
ing a considerable strain on Soviet investment alloca-
tion.
The importance of oil production in the Soviet econo-
my is mirrored in SOVSIM by a detailed description
of that sector. Overall exploration needs, as estimated
by CIA analysts, determine the investment available
for development drilling (figure 6). The drilling that
occurs in western Siberia fixes the number of new
wells in the region and their flow rate. Along with
Output of other fuels and power is determined by
investment and labor inputs in these sectors. Energy
production is converted to value added in 1970 rubles
according to the output-value-added ratio for that
year.
Industrial Production and Services. Other Soviet
production originates in a number of industrial and
service sectors (first inset). The industrial sectors
include machine building, ferrous metallurgy, nonfer-
rous metallurgy, construction materials, forest prod-
ucts, consumer goods, and chemicals. The main ser-
vice sectors are construction, transportation and
communications, and trade and services.
SOVSIM estimates the output of each of these sectors
by means of a production function, an equation
relating the annual capital and labor available and the
output produced in a given sector. The exact form of
the function used for each sector is estimated by 25X1
examining the historical relationship between labor
and capital availability and output in that sector. F_~
Projections based on such functions are valid to the
extent that the historical conditions underlying past
production in the sector continue to hold during the
projection period. In fact, these conditions will
change, but they may not shift radically over the next
10 years:
? The Soviets seem reluctant to scrap existing equip-
ment; even if new technology were to appear in a
given sector, it would take a long time for the more
productive equipment to make up enough of a share
of total capital in the sector to alter average produc-
tive performance significantly.
? Given the educational and training systems used by
the Soviets, the quality of labor is also unlikely to
improve rapidly.
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
? Nor are Soviet administrative and management
practices likely to change sharply.
? Other changes, such as those in weather conditions,
may occur but are as likely to retard, as they are to
improve, productive performance.
On balance, then, the net effect on the conditions
underlying productive performance in the various
sectors is not likely to change very much over the next
five to 10 years.
Where a significant departure from past conditions is
expected, the parameters of the equation in question
may be altered on the basis of the analyst's judgment
to reflect the new situation. Where uncertainty exists
as to which of alternative possibilities may hold,
alternative sets of parameters, each reflecting one of
the possible situations, may be used to arrive at the
range of possible outcomes. The latest projection of
the Soviet economy uses this approach because of
uncertainty as to the future productivity of Soviet
capital and labor.
Using the production function in output projections
allows us to estimate the contribution of both capital
and labor to output in each sector and hence the
impact of changes in capital and labor on sectoral
output. We can also examine the consequences of
separate changes in capital and labor efficiency and
technological change on sector output. Because these
areas are critical to the development of the Soviet
economy, SOVSIM's production functions provide a
tool for exploring questions that concern the Soviet
leadership.
International Economic Relationships
Hard currency imports have been particularly signifi-
cant in several key sectors of the Soviet economy:
? Grain purchases have helped to ease the conse-
quences of poor harvests and permitted expansion of
Soviet livestock herds and meat production.
? Machinery imports, while small compared to total
Soviet investment, have been important to the ex-
pansion of energy and chemicals production and
transportation services.
Hard currency imports are projected by CIA trade
analysts. In the model, they can also be set as a policy
variable to determine their impact on the balance-of-
payments position and the rest of the economy.
Figure 7
SOVSIM: Overall Hard Currency
Balance of Payment Flowchart a
Government-
backed credit
Medium- and Short-term
long-term commercial
commercial credit credit
Gold Total credit
sales drawings
I
Total hard
currency capital
availability
Accounting
Net borrowin
Current
Capital
Principal
T
g
accounts
accounts
repayment
balance
balance
Gr
oss
Foreign
debt
Overall
asset
balance of
purchases
payments
Net
debt
Other uses:
political activities,
export credits,
aid and trade
a Variables in bold frames are inputs to the model.
Other variables are computed by the model.
The USSR's ability to pay for these imports depends
on the hard currency it receives from exports, gold
sales, credit drawings, and interest earnings on and
the sale of foreign asset holdings (figure 7). These
receipts are projected by CIA analysts. The analysts
also project government-backed and short-term com-
mercial credit drawings. Given projected import
needs, medium- and long-term credit drawings re-
quired to clear accounts are computed by SOVSIM.
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 8
SOVSIM: Trade and Current Account Balancea
Current account
balance
Net earnings from
tourism and
transport services
Interest
payments
Soviet nonenergy
exports
a Variables in bold frames are inputs to the model.
Other variables are computed by the model.
bSoft currency, non-Communist countries are represented
in the model, but they account for a small part of Soviet
trade and are therefore not included in this chart.
Soviet energy
exports
The model also computes various statistics describing
the Soviet hard currency position, including the hard
currency current and capital account balances, debt,
debt service, and the debt service ratio. Alternatively,
SOVSIM can solve for hard currency import capacity
on the basis of projected credit drawings, or, given
constraints on the debt service ratio, the model can
compute the level of one or more import components.
SOVSIM assumes that intra-CEMA trade is handled
largely on a barter basis. Nonenergy exports to the
Communist Bloc depend on East European economic
growth and the price of oil charged to Communist
countries, as indicators of need and ability to pay,
respectively (figure 8). In turn, Soviet imports from
Communist countries depend on total Soviet exports
to those countries. Both of these relationships are
estimated from historical data.
Users of Goods and Services in the Soviet Economy
Each sector of the Soviet economy requires inputs
from other sectors. If these inputs are subtracted from
the gross output of the sector, the remainder measures
the value added, the new output, produced by that
sector. In SOVSIM the output of all sectors except
agriculture is computed directly as value added. Agri-
cultural value added is obtained by subtracting the
value of inputs from gross output. Military value
added is obtained as military wages, projections of
which are obtained from SOVA military analysts. F_
The total value added by all sectors is the gross
national product of the economy. Some of this output
is used to repair or replace equipment and facilities.
What remains is the net national product (NNP) of
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 9
SOVSIM: National Income Accounts Flowchart a
Military Electric
value added power
Consumer
goods
Capital
repair
Tourism and
transport
a Variables in bold frames are inputs to the model. Imports
Other variables are computed by the model.
Gross
national
product
New fixed
investment
the economy.' Some of the NNP is sold or traded to
other countries. In return, the Soviet Union receives
other goods and services, some of which are used as
inputs in Soviet production. The remainder and the
NNP that is not traded away are available to the
Soviet economy for final use:
? Consumption.
? Investment.
? Government activities, including defense, research
and development, and administration.
' The Soviets structure their national income accounts differently.
The definitions in SOVSIM are those of the Western market
economies and are used to facilitate a better understanding of
Soviet economic activities and comparison of these activities to
those in the West.
Gas Machine Chemicals Industrial
building materials
Other
production
Transport and
communication
Arms
sales
Government activities are projected by CIA analysts.
Investment may be either specified to determine its
impact on consumption and other variables, or pro-
jected by the model as the output of the machine-
building and construction sectors-adjusted for de-
fense procurement and construction, production of
consumer durables, net exports of machinery and
equipment, and capital repairs.
Consumption is computed as a residual (figure 9),
after other final uses are subtracted from the goods
and services available for final domestic use. It is
adjusted by an estimate of the additional value gained
by the export of gold and energy products. Since the
mid-1970s, rising world prices for these exports have
25X1
25X1
Trade and
services
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
increased their value far above the factor costs of
production, causing Soviet hard currency earnings to
soar. The result has been a growth in Soviet hard
currency imports that has benefited both investment
and consumption.
SOVSIM simulates the economy one year at a time.
For each year, estimates for each output variable in
the model are computed from model relationships and
available data. Some of these data may have been
computed by the model as results for preceding years
or for other equations. Other required data are not
computed by the model at all but must be supplied to
it for each year of the projection. Several different
types of data are needed:
? The estimated parameters of those relationships
that are derived econometrically from historical
observations. The model assumes that the underly-
ing Soviet behavior and technical relationships will
not change during the projection period.
? The most recent historical values of those variables
that the model is to project over time; these are a
starting point for model calculations.
? Projections of values of those variables that may
change over time but are not computed by the
model-policies, technological conditions, economic
conditions in other countries, natural phenomena;
these are needed for each of the years to be
simulated and are obtained from experts and other
outside sources.
Equation parameters, the first type of data, are
estimated periodically by straightforward, well-known
statistical techniques. Their values involve some de-
gree of uncertainty because of the quality of the
underlying data and limitations inherent in the statis-
tical estimation procedure. Nevertheless, these data
are believed to be relatively accurate. Initial condi-
tions for the model, the second type of data, are also
reasonably accurate, given the quality of historical
data available on the Soviet economy.
Major Data Requirements
Sector/Item
Agriculture
Grain output
Livestock herd
Energy
Trade
Distribution priorities
Sectoral energy/output rates
Expected production
Expected investment
Investment distribution
Population
Age-sex population levels
Sectoral employment rates
Sectoral labor distribution
Trade and international
Gold sales
Arms sales
Foreign assets
Imports and exports
Government and pipeline credit drawings
Defense
Procurement
Construction
Pay rates
Other
Research and development
Government administrative expenses
Estimating the future course of the other variables
found in the model is a large and difficult task. Here
the SOVSIM effort relies heavily on:
? Extrapolation of likely trends established by previ-
ous analysis.
? The judgment and research findings of other experts
on the Soviet economy both within the CIA and at
other government and academic and research insti-
tutions (inset).
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
These projections involve differing degrees of certain-
ty. For example, estimates of labor force growth are
relatively firm, because population changes occur
slowly over time and participation rates are already so
high in the Soviet Union that they are not sensitive to
policy shifts. On the other hand, the distribution of
GNP among primary uses-consumption, investment,
defense, and net exports-although subject to some
rigidities, is ultimately subject to policy choices that
could change significantly over the period to be
simulated.
In general, we have more certainty about input values
that are subject to little, if any, policy manipulation or
are clearly reflections of long-term trends not likely to
be quickly reversed. We have less certainty about
those values that can be strongly influenced by such
factors as policy decisions and international market
forces. Where the uncertainties are greatest, we may
try alternative projections to gauge the sensitivity of
the model's results. Furthermore, where gaps in our
knowledge of the Soviet economy exist, SOVSIM
provides a means of estimating unknown values by
showing the range over which they would be consis-
tent with what is known of Soviet economic activities.
Thus, SOVSIM brings together a diverse body of
information and provides a means of checking the
consistency of data obtained from a variety of sources.
This helps us to assess the value of existing intelli-
gence and research activities and to suggest what
additional efforts may be needed.
? How might this projection change if one or more of
the assumed natural conditions or policies is
changed?
? What requirements are needed to carry out specific
Soviet policies or reach given goals?
Analysis of Historical and Current Developments
Although we already know a great deal about Soviet
economic activities since the mid-1970s, SOVSIM
can be used to help us understand why certain
developments occurred and what impact they had on
the economy. For example, the model can be used to
examine the decline in the rate of growth of Soviet
GNP during the past decade, from 3.7 percent per
year during the 1971-75 period to 2.5 percent per year
in 1976-82. While the causes for this decline are
shrouded in some uncertainty,' analysis with
SOVSIM may be able to suggest some of the factors
that could be involved.
Sharp reductions in the rates of growth of oil output
and the labor force also took place during the past 10
years, that for oil falling from 6.9 percent a year
during the first period to 2.9 percent in the second and
that for labor from 1.6 percent a year from 1970 to
1975 to 1.1 percent during 1980-82. If Soviet oil
output had continued to grow at the earlier rate (all
other things, particularly oil industry investment,
remaining equal), GNP would have grown almost 0.3
percentage points per year faster after 1975 than it
actually did and would have been almost 2 percent
greater in 1982 than it was.6 If the labor force had
continued to grow at the 1971-75 rate, GNP growth
would have been about 0.25 percentage points per
year faster after 1975, and 1982 GNP would have
SOVSIM's output is a quantitative description of the
domestic and international sectors of the Soviet econ-
omy (table 1). The model can be used in various ways
to study and help answer different questions about the
economy:
? How well has the economy performed over the
recent past?
? How well is it now operating? What problems does
it face and how might they be solved?
? Under the conditions that will probably prevail, how
is the economy likely to develop over the next three
to 10 years?
`This estimate is conservative. SOVSIM results indicate that at the
higher rate of growth of oil production, oil supplies would have
exceeded the economy's direct needs and resulted in a considerable
surplus. In the oil-short world of the late 1970s, this surplus could
have been exported for hard currency to finance the purchase of
investment goods that could have supported additional growth of
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Table 1
Types of Model Output
Sector/Item Output Invest- Employ- Capital Fuel Foreign trade Hard Currency Communist Soft Currency
ment ment Stock Require-
ments
Meat VA/Q Balance of V V V
Nongrain VA
crops
Energy VA/Q V Q V Q Imports V V
Coal VA/Q V Q V Q Interest V
Gas VA/Q V Q V Q Gold sales V
Electricity VA/Q V Q V Q Capital account V
balance
Ferrous VA
metals
Non- VA
ferrous
metals
V Q V Q
Construc- VA
tion ma-
terials
Forest VA
products j
Consumer VA
goods
Construc- VA V Q V Q
tion
Transpor- VA
tation and
communi-
cations
Trade and VA V Q V Q
services
Housing V V Q
(construc-
tion)
Capital V
repair
Aggregate V
consumption
Q - quantity
V - value
VA - value added
Debt service Q
ratio
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
been about 1.6 percent greater. It appears that these
two factors alone could explain almost half of the
decline in the GNP growth rate since 1975. If so,
however, other factors must also have contributed to
the decline in GNP growth.
SOVSIM is also useful for evaluating current Soviet
economic developments and problems. One area of
interest is the rate of growth of the industrial sector of
the economy that, until 1983, had been declining
quite steadily since the 1950s. Since the mid-1970s
this decline has been exacerbated by a decline in the
rate of growth of capital and labor productivity.
Analysis with SOVSIM reveals that, if Soviet indus-
try were making as efficient use of capital and labor
today as it did in the early 1970s, industrial output
would be growing about 50 percent faster than it did
during the early 1980s.
Projections of the Soviet Economy and Scenario
Analyses
Since 1979 SOVSIM has been used extensively to
provide a complete, integrated view of Soviet econom-
ic prospects. Research papers in 1979, 1982, and 1984
presented full baseline projections of the Soviet econo-
my in the 1980s.' SOVSIM has also been used to
explore the consequences of changes in the energy,
agricultural, and trade sectors; the structure of invest-
ment; productivity levels; and end-use orientations of
the Soviet economy-either because of policy shifts or
other conditions. The model has been useful in evalu-
ating the potential impact on the Soviet economy of
Western policy options such as embargoes of grain or
energy equipment and credit restrictions.
Such studies allow us to evaluate the impact of
different policies on various economic outcomes and
thus provide a better understanding of the economy's
inherent potential and the barriers and bottlenecks
that may appear under different circumstances. For
example, a 1982 study 8 suggested that significant
reductions in defense spending during the 1980s could
allow Soviet GNP growth to increase and would
probably enhance the availability of consumer goods
in the economy. In the important agricultural sector,
Soviet grain output has recently deviated significantly
from trend levels, with production averaging 8 percent
below trend from 1979 through 1982 and 16 percent
above from 1976 through 1978. A 1984 SOVSIM
analysis' suggests that, given Kremlin goals related to
consumption, such deviations could seriously affect
the hard currency balance-of-payments position and,
depending on the reaction of Soviet leaders, indebted-
ness to Western countries or merchandise imports
from the West.
SOVSIM can also be used to determine the resources
and activities needed to achieve specific economic
goals or to carry out certain policies. Such analyses
reveal the feasibility of these goals and policies and
the obstacles that may impede their achievement and
execution. They indicate what they will cost the
Soviet Union and provide a means of comparing the
costs of alternatives to arrive at an estimate of the
relevant trade-offs.
Limitations
SOVSIM is subject to various limitations. Some of
these apply to all economic models; others are unique
to SOVSIM. Any economic model contains only those
features that the user feels are important. Therefore,
it cannot be used to explore problems involving other
components of the economy. Moreover, because each
model is generally structured to perform a particular
task-such as simulating activities, finding optimal
solutions, or determining operational levels of activi-
ties-it can be used to perform other tasks only poorly
or not at all. Finally, the validity of results from any
model depends on the quality of its construction and
supporting data base. A model that does not describe
reality well or that uses inaccurate data is unlikely to
yield good results.
25X1
25X1
25X1
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
As with any tool, SOVSIM is suited to certain jobs
but not to others. Because it is a simplified description
of the Soviet economy, the model lacks the interrela-
tionships and complexity needed to deal with many
interesting problems. For example, SOVSIM:
? Does not embody a full input-output table. There-
fore, it can address "balance" issues only in selected
areas-energy, construction, agriculture (grain), and
machine building. Nor can it ensure that each
producing sector receives all of the different kinds of
intermediate products that it requires. Therefore, it
cannot deal with specific bottlenecks.
? Reflects a particular structure of the Soviet econo-
my and is not applicable to projections of situations
involving significant changes in that structure.
? Is not meant to arrive directly at optimal answers to
problems. Other models exist for such analyses.
?. Is not suitable for examining short-run problems. It
lacks many of the variables that determine short-
run variations in Soviet economic activities. More-
over, the model's structure has been estimated from
long-run trends and is geared to the study of longer
run developments.
SOVSIM does not make predictions; rather it projects
trends based on specific assumptions and the data
used. Each projection involves some degree of
uncertainty:
? The estimated structure of the Soviet economy that
is made from historical data involves some statisti-
cal imprecision.
? It is impossible to foresee technical developments
with certainty.
? It is also impossible to ascertain in advance the
policies to be implemented by the Soviet leadership.
The uncertainty is greater in some areas than in
others. Therefore, model output must be interpreted
carefully to take account of the alternative possibili-
ties involved in a particular problem. Often this
interpretation requires assessment of a range of possi-
ble outcomes rather than a point estimate.
SOVSIM can be evaluated in several different ways.
Assessing its ability to describe the Soviet economy is
relatively easy: the model can be used to simulate past
or present economic activities and the results checked
against known historical values. It is.more difficult to
check SOVSIM's ability to project future economic
developments:
? We can examine old projections that were actually
made to see how accurate they turned out to be.
? SOVSIM projections can also be compared with
those obtained from other analyses.
? Sources of error can be examined by comparing
projections based on full historical data with those
based partially on data derived in other ways. F_
SOVSIM is primarily designed to project medium- to
long-run trends. Year-to-year values are strongly af-
fected by changes in weather, international political
and economic conditions, and national policies that
are not included in the model. We expect, therefore,
that SOVSIM will be more accurate in projecting
trends and average behavior over a number of years.
Historical Tracking
To test the ability of the model to describe the Soviet
economy, SOVSIM was used to simulate its activities
during the 1975-83 period. The results (figure 10)
suggest that SOVSIM can approximate recent and
current aggregate economic trends, although it may
deviate from actual performance for specific years or
activities. As expected, the long-term trends suggested
by SOVSIM output do reflect the historical trends.
The model output, however, also appears to follow
year-to-year changes in the historical values of these
variables reasonably well. In only a few cases do
model projections of such short-term changes run
counter to what actually happened.
Most model output nearly matches the corresponding
historical values. Significant deviations appear to
occur only for certain years for a few variables-such
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Figure 10
Comparison of 1975-83 SOVSIM Output for Key
Variables With Corresponding Actual Values
- Actual GNP Growth Rate
SOVSIM
Transportation and
Communications
0 1975
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
as GNP growth rates, grain imports, and the debt
service ratio-that are computed as residuals. For
these variables, relatively small errors in component
values can give rise in some years to relatively large
Table 2
Simulation Errors for
Key Results, 1975-83
errors in the calculated residuals.
For some variables, SOVSIM appears to yield biased
results. Estimates of consumption on the end-use side
and trade and services on the sector-of-origin side are
consistently biased downward over the entire histori-
cal projection. An upward bias appears in the case of
oil and agricultural output, the debt service ratio, and,
in more recent years, new fixed investment. Such
biases suggest the omission of one or more explana-
tory variables or relationships. Even in these cases,
however, SOVSIM captures the long-term trends, and
few large deviations from actual values occur. F_
A good summary of SOVSIM's performance in de-
scribing these individual activities over the entire
historical period is the average deviation of the annual
estimates from the actual values for those years (table
2). As might be expected, estimates for the GNP are
more accurate than those for each of the component
sources and end uses. Component errors tend to
average out at the more aggregate level. Even so, the
average error for many sectors-such as output of
meat and nongrain crops, equipment, and transporta-
tion and communications services-is small, because
production in these sectors is measured by statistically
derived equations that are, on the whole, relatively
accurate.
Results for the international sector are worse. These
are generally estimated as residuals of many variables
in which the small errors in components carry over as
relatively large errors. Therefore, estimates for indi-
vidual years may be considerably inaccurate, al-
though the model as a whole prevents serious errors
from persisting. By contrast, estimates of trade with
Communist countries are far more accurate because
these are made by statistically derived equations. F_
An accurate portrayal of past and current develop-
ments does not give complete assurance of the model's
validity in other uses. It can, however, reveal which
segments of the model describe the corresponding
components of the Soviet economy reasonably well
and which do not. This, in turn, gives us some
Average
Annual
Error
Largest
Annual
Error
Agriculture
2.4
4.3
Nongrain crop output
0.8
2.2
Meat output
-0.9
4.2
Consumer goods
2.0
3.9
Industrial materials
-3.1
-8.1
Chemicals
2.9
6.4
Construction
2.3
5.5
Machine building
0.6
2.8
Oil output
2.9
4.9
Transportation and communications
-0.9
-3.8
Nonenergy exports
-0.2
-3.8
Nonenergy imports
0.2
-6.9
a On the average, over the nine-year period, SOVSIM underesti-
mated actual GNP by 0.2 percent.
b The largest error in estimating GNP for each of the nine years
was an underestimation of 2.3 percent in 1975.
understanding of the nature and size of the errors that
may be involved with specific results and suggests
where additional intelligence, research, and modeling
efforts must be focused to make the greatest contribu-
tion to the performance of the model and the useful-
ness of its results.
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Table 3
Comparison of SOVSIM Projections
for 1981-85 With Actual Values
for 1981-84, Average Annual Growth Rates
How Well Has SOVSIM Performed?
The ability to track historical events may not be a
good indicator of how well future events can be
projected. One way to assess projection performance
is to review past projections of now historical periods.
As SOVSIM was developed, projections of Soviet
economic activities during the 1981-85 period were
made in mid-1978 and mid-1979 and again in 1982
and 1983.10 A comparison of the results of these
analyses with the corresponding historical values (in
this case data are available for the 1981-84 period
only-table 3) may provide insights into the strengths
and weaknesses of the model and suggest ways in
which it can be improved.
Initial projections of the 1981-85 average annual
growth rate of industrial production were quite close
to the actual rate. Those made in the early 1980s,
however, were relatively low. This deterioration in
projection accuracy might be associated with possible
overestimates of the decline in the productivity of
industrial capital and labor inputs or underestimates
of the ability of the Soviets to conserve use of energy
while maintaining industrial production. Current
analysis with SOVSIM explores a range of possible
productivity and energy-use levels.
Early projections of 1981-85 growth of agricultural
output suggested some improvement over the poor
performance of the late 1970s. They did not fully
anticipate the strength of the recovery that actually
occurred in 1981 and 1982, probably because they
were strongly colored by the poor Soviet weather of
the last half of the 1970s. When weather conditions
improved in 1982, revised climatic assumptions
caused projections of agricultural output to shift
markedly upward, perhaps overcompensating for the
earlier error. Although this portion of the model has
been much improved over time, projection of agricul-
tural output remains difficult because of the influence
of external factors, such as weather conditions. It may
be useful, therefore, to consider the various agricul-
tural policies available to Soviet leaders and to explore
the implications of different weather patterns for
agricultural output.
Projections of GNP growth rates for the 1981-85
period tend to average out the errors in the sector-of-
origin estimates. On the whole, however, they tend to
be somewhat conservative. This suggests that
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
SOVSIM is likely to provide better estimates at
higher levels of aggregation than for more specific
activities. As improvements are made in the estimates
and projections for the latter, aggregate variables are
likely to improve as well but to a smaller extent. I
More recent estimates of 1981-85 consumption
growth rates have been slightly pessimistic but rela-
tively consistent. The underestimation appears to have
resulted primarily from the relatively low estimates of
GNP. Because SOVSIM computes consumption as a
residual variable, errors in the estimation of the other
variables involved in the computation are also proba-
bly important. New approaches to estimating con-
sumption, to defining and estimating currently unde-
fined residuals such as inventory changes, and to
estimating the role of exports and imports may well be
helpful in improving model estimates of consumption.
Estimates of new fixed investment growth have been
more volatile, probably because of their dependence
on a number of assumptions that have been subject to
change over time. These include projections of defense
spending, trade, and energy availability. Here again,
improvements in these individual areas may help to
improve estimates of investment growth. This vari-
able, however, is greatly influenced by Soviet policies
concerning consumer durable production and invest-
ment allocation. Its estimation, therefore, would bene-
fit from an exploration of the options open to Soviet
leaders.
Comparison With Other Models and Projections
The validity of SOVSIM projections can also be
assessed against those made with other economic
models. SOVSIM differs from economic models of
Western, market-oriented economies that focus on
demand to determine levels of production, employ-
ment, and prices. In contrast, SOVSIM assumes that
Soviet output levels are determined by resource avail-
ability and the efficiency of production processes.
Demand is relatively unimportant, since taut central
planning ensures a perpetual state of excess demand.
There are several other Western models of the Soviet
economy. The best known of these, SOVMOD, con-
tinues to be more optimistic than SOVSIM in project-
ing the future of the Soviet economy, primarily
Table 4
SOVSIM and SOVMOD
Assumptions and Results,
Average Annual Growth
Rates, 1985-89
GNP
1.89 to 2.73
2.66
Industrial output
2.36 to 3.78
2.78
Agricultural output
1.40 to 1.42
2.89
1.81 to 2.21
2.33
Investment
2.00 to 4.00
6.42
Hard currency debt
8.59 to 20.67
10.35
a The lower bound for investment growth is obtained when
SOVSIM is allowed to solve for the rate of growth. The upper
bound was exogenously introduced to determine the implications of
more rapid investment growth. With SOVMOD, investment is
determined by the model; hence, no assumption is made as to the
rate of growth of investment.
because of important differences in the assumptions
about resource availabilities, policies, and structure of
the Soviet economy (table 4)." SOVMOD assumes
that:
? A larger share of total output is devoted to
investment.
25X1
25X1
25X1
" Wharton Econometric Forecasting Associates, Centrally Planned
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
? The growth of energy output and use are more
favorable.
? Grain output will grow faster.
? Both exports to and imports from the West are
assumed to grow more rapidly but in such a way
that projections of the overall hard currency bal-
ance-of-payments position and debt are more
favorable.
Because of these differences, SOVMOD projects both
total output and consumption to grow faster than does
SOVSIM.
Soviet economists use similar models, constructed
with the same statistical procedures, to examine their
economy.12 They are forced, however, to use data that
Western economists believe to be misleading. Much of
their research is published openly, so that we can use
their perceptions in conjunction with SOVSIM re-
sults.
From time to time, results from the various models of
the Soviet economy are compared with those produced
by SOVSIM. Similarities reinforce our conclusions
about the Soviet economy. Differences, however, are
not necessarily bad because they often simply reflect
our limited knowledge about the Soviet economy and
different theories about its operation. As such, they
force us to consider more carefully our assumptions
and recheck our calculations.
A Test Projection
Projections require an estimate of the structure of the
Soviet economy, as reflected in the equations of the
model, and the performance of a variety of variables,
as provided by the data inputs. Historical observations
are available to facilitate the model's construction and
use. The economy's structure, however, may change
over time, requiring corresponding changes in the
model's structure. Moreover, data inputs for the
future period to be simulated must themselves be
projected by a variety of means and thus are subject
to various degrees of uncertainty.
" The CIA is now working to reconstruct several such models.
These will prove useful in assessing the Soviets' perceptions of their
SOVSIM projections, therefore, are subject to both
model and data errors. Model error involves inaccura-
cy in the model's equations as a description of the
future structure of the Soviet economy:
? Equations concerning significant economic activities
may be omitted from the model.
? An equation describing an activity may be incor-
rectly specified; it may include an incomplete or
incorrect set of explanatory variables.
? An equation may show an incorrect relationship
between the activity being explained and one or
more of the explanatory variables; the estimated
parameters of the equation may be incorrect.
Data error involves inaccuracy in the assumptions
about the future, as reflected in the data used in the
model:
? The most recent historical values of those variables
that the model is to project may be erroneous,
causing the projection to start at the wrong place.
? The projected values of data inputs, required by the
model to compute other values, may be incorrect.
Because the future is inherently unknowable, such
projected values are bound to involve some degree of
error that is passed on to the values computed by the
model.
To explore the relative importance of these two errors,
we used SOVSIM to make two projections of the
Soviet economy in 1982, the last year for which
complete historical data were available:
? For the first projection, we assumed that actual data
for 1982 were unavailable. Data inputs, therefore,
included values for 1982 that were projected on the
basis of average growth rates or trends for the
respective variables during a period of years up
through 1981. Moreover, model parameters were
estimated by using historical data from the 1970-81
period only.
? In the second run, the data inputs included actual
observations for 1982, and the model parameters
were estimated by using historical data from the
entire 1970-82 period.
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Table 5
Performance in Projecting
1982 Economic Activities
Percent
(unless otherwise noted)
Activities
Actual 1982 Value
Deviation From Actual Value Caused by
Total Error a
Model Error
Data Error
Domestic
Billion 1970 rubles
(unless otherwise noted)
Agriculture
78.0
6.2
-0.8
6.9
Meat output (million metric
tons)
15.4
2.7
-1.4
4.0
Transportation and
communication
63.7
3.6
1.9
1.7
International
Billion current dollars
(unless otherwise noted)
Trade balance
-0.99
155.2
-11.9
167.0
Credit drawings
2.45
-188.1
-109.8
-78.3
Debt
20.00
-26.6
-1.7
-24.9
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
The results from each simulation were then compared
to the corresponding actual historical values for 1982.
Any errors in the results of the first simulation reflect
total error. For the second simulation, they reflect
only model error. The data error alone can be deter-
mined by subtracting the model error of the second
simulation from the total error of the first.
A sample of the results of this analysis (table 5)
suggests that data error was larger than model error
in the international and agricultural sectors. Both
were in transition in 1982; agriculture had just come
through several bad years and the price of energy
exports had stopped climbing for the first time since
1978. Data projections for these sectors on the basis of
performance in the recent past were unlikely to be
good. In the energy sector, model error appears to be
larger than data error. In the other sectors of the
economy the relative importance of the two kinds of
error was mixed. For the economy as a whole and the
two end uses given, data error was much larger than
model error.
In the productive sectors, total error was relatively
small. Output in these sectors was estimated by finely
tuned production functions that generally yielded very
low model error. The nonagricultural sectors had been
fairly stable for several years prior to 1982, permitting
data projections to be relatively accurate. Total error
for GNP represents an averaging of the errors of the
individual productive sectors, but it appears that most
of it was due to data problems rather than to the
model.
It is not surprising that investment involves much less
error than does consumption. Both are estimated as
residuals. But the components of investment are both
well known and relatively stable. Consumption, by
contrast, is estimated using more activities, some of
which are difficult to measure accurately. Relatively
small errors in these components net out to a relative-
ly large error in the final estimate of consumption. F
Trade balance estimates are subject to the same
problem: small errors in the components net out to
relatively large errors in the residual balances. Be-
cause the model uses its hard currency trade balance
to estimate credit drawings, the model error for this
variable turned out to be quite large.
For similar reasons, the GNP growth rate was subject
to a large error. Small and unbiased errors in the
year-to-year estimates of GNP can result in a much
larger error in the percentage change. In our analysis,
however, most of the error resulted from data prob-
lems.
It should be noted that projection errors are peculiar
to the period simulated. Economic conditions change
constantly. Additional information and intelligence
are constantly being made available. Because esti-
mates of input data for a particular year improve as
time passes, projections of that year are likely to
improve as the year draws closer. Projection accuracy
is also enhanced by stability of economic activity. F_
25X1
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4
Confidential
Confidential
Sanitized Copy Approved for Release 2010/03/22 : CIA-RDP86T00591 R000300340001-4