FINAL REPORT T O

NATIONAL COUNCIL FOR SOVIET AND EAST EUROPEAN RESEARC H

TITLE :

AUTHOR :

CONTRACTOR :

A Macroeconometric Model of a
Centrally Planned Econom y
With Endogenous Plans :
The Hungarian Cas e

Edward A . Hewet t

University of Texas at Austi n

PRINCIPAL INVESTIGATOR :

COUNCIL CONTRACT NUMBER :

Edward A . Hewet t

620- 7

The work leading to this report was supported in whole or i n

part from funds provided by the National Council for Sovie t
and East European Research .

Ed . A . Hewet t

A Macroeconometric Model of a Centrally Planne d

Economy with Endogenous Plans :
The Hungarian Cas e

SUMMARY

There is a substantial and increasing interest amoung various group s

in the West in developing a capability to predict the economic performanc e
of the Soviet and Eastern European countries . Banks have a great dea l
invested in those countries, are being asked to invest more, and nee d
objective information on their economic performance . Governments ,
frequently asked to quarantee loans to Soviet and East European countries ,
share similar concerns . In addition, Western governments have an inheren t
interest in Soviet and East European economic performance grounded in th e
importance of economic performance for the political stability of thes e
regimes, for economic and political relations between Eastern Europe an d
the Soviet Union, and for the capability of those countries as a group t o
sustain large and growing military expenditures .
All of these concerns coincide with the interests of some Wester n
economists in using macroeconometric models to better understand th e
operation of Soviet and East European centrally planned economies (CPE's) ,
and recent years have seen a substantial and increasingly sophisticate d
effort to use econometric techniques to generate predictions of economi c
performance in CPE's similar in scope and precision to predictions common place for western economies . The SOVMOD series of models (now under th e
direction of Wharton's CPE Projects) is the most well-know and ambitiou s
of such efforts, but there are also others, both for the Soviet Unio n
(for example the CIA's SOVSIM), and for many East European countries .

In predicting the economic performance of a CPE over, say, a fiv e

year period, one mus t - start with a prediction of what planners themselve s
will do, then decide how their actions will affect economic performance ,
and finally decide how economic performance might feed hack on planners '
actions . Our ability to predict planners' behavior, and to build thos e
predictions into a model of a CPE is still rather limited . To date n o
CPE model has successfully captured the richness and magnitude of planners '
roles in determining macroeconomic aggregates in CPE's .
This paper attempts to improve on that situation by constructin g
a macroeconomic model of a CPE which fully incorporates a model of planners '
behavior into the system . The three basic assumptions behind this effor t
arc :

1) that planners are the major adjustment mechanism in the CPE ,

thus no major change in economic activity occurs unless they initiate it ,

2) that planners monitor key performance variables in the economy (th e
balance of payments, the level of consumption, and so on), and that when thos e
variables deviate from the values planners desire, they take action t o
bring the economy back on track, and 5) that planners act in a regular an d
predictable fashion when making adjustments in their policies in respons e
to changes in various performance variables . The equations summarizing ho w
planners respond to changes in economic activity are called "reactio n
functions", and they serve to predict changes in economic policy resultin g
from changes in economic performance . The remaining equations in the mode l
are similar to those in any macroeconomic model :

they explain variation s

in economic performance using many variables, including the economic policy

variables determined by the planners reaction functions . Section II of th e

paper discusses the structure of the model in detail .
Sections III-IV report on the application of this approach in th e
construction of a macroeconometric model of the Hungarian economy (th e
Hungarian Econometric Model #1, or HEM-1) . Data on economic activity an d
economic policy are relatively good for Hungary (compared to the Sovie t
Union and the other Eastern European economies) ; consequently they provid e
a good test of the approach under relatively favorable conditions .
HEM-1 is a small annual model [21 equations] which highlights major inter connections between economic policy and economic performance . It predict s
the plans for the key macroeconomic aggregates for which the Hungarian s
provide data (consumption, investment, GNP), and it predicts those aggregate s
themselves . One of the most im p ortant determinants of the plans is recen t
economic performance ; and one of the most important determinants of consumption, investment, government spending and foreign trade is the plans .
The details of the specification and estimation of the model are discusse d
in Section III .
Section IV uses the model to explore the impact on Hungary of th e
changes in its terms of trade flowing from the OPEC price increases . Durin g
the years 1973-7S, Hungary's terms of trade in convertible currency fel l
about 20 percent ; its terms of trade in rubles fell about 15 percent . Th e
comparison of the economic performance of Hungary as modelled by HEM-1 wit h
and without those terms of trade changes suggests that all of Hungary's shor t
term dollar borrowing (about $1 .5 billion) over 1973-78 was a result o f
those terms of trade changes . On the other hand, Hungarian planner s
did respond to the terms of trade changes by diverting GNP from consumption ;

-4 government spending, and investment to net exports ; and conse q uentl y

they borrowed much less than they would have if they had not reacted a t
al] to the crisis .
In the 1980's as economic conditions grow more severe in the Sovie t
Union and Eastern Europe, it is extremely important that we in th eWst
develop our own capability to analyze the economic performance of thos e
systems, and the actions planners will take in response to changes i n
economic performance . Because econometric models rely heavily on the pas t
they cannot form the sole source for our analyses, but they nevertheles s
provide one indispensable component in any analysis of present and futur e
economic performance in those countries . The experience with HEM- 1
unambiguously supports the notions that planners do react with regularit y
to changes in the economic performance of their systems, and that wher e
data permit, reaction equations modelling planners behavior should play
key role in macroeconometric models of CPE's .

1 . Introductio n

This paper discusses a macroeconometric model of Hungary which utilise s

planners' reaction function to explore the link between economic activit y
and plans, and plans and economic activity . This approach pulls togethe r
two separate, but interrelated, lines of research on centrally planne d
economies (CPE's) . There is on the one hand research beginning with Kalecki ,
then Goldman and Kouba, and, more recently,

and Lacko, whic hsuget

that planners in CPE's regularly behave in such a fashion that they produc e
investment cycles . Two points are im p ortant here : (1) there is a com p onen t
of macroeconomic planning in CPE's which is regular and predictable, an d
(2) the plans resulting from that predictable behavior may contribute t o
cycles in the economic system . Many of these studies rely o nqualitve
or exceedingly crude empirical techniques to test their hypotheses, whic h
means that they are suggestive, but hardly definitive . Furthermore thei r
scope is limited in that they fail to consider the possibility of regula r
planners' behavior in determining macroeconomic aggregates other
nivesttha
and they fail to incorporate what information they have on planner s
behavior into a macroeconomic model of the CPE, in order to explore the ful l
macroeconomic consequence of that behavior . Kornai (1971, 1976) has provide d
some useful concepts for a more comprehensive approach in his discussions o f
planner s ' reaction functions, and what is needed now is to estimate empirica l
versions of those functions . .
The second related line of research has been the construction in recen t
years of numerous macroeconomctric models of CPE's (Shapiro ; Shapiro an d
Halabuk ; Green, et . al . ; Green and Higgins ; U .S . C .I .A . ), including several

2.

medium-sized models of the Hungarian economy (Halabuk,

Hunyadi, et

t . al . ; Simon ;

al. ) . These models are extremely valuable for exploring inter -

relationships in CPE's, and also for the discipline they impose on the model builder as he/she must write out an internally consistent set of equations .
Existing macroeconometric models

CPE's fail to systematically tak e

account of the regular elements of planners' behavior . This is understand able for the models constructed by Soviet and East European economists, sinc e
those models are usually built for planners, and one doubts that planner s
would appreciate receiving a model including planners' reaction functions .
But even major western macroeconometric models of CPE's, such as SOVMOD ,
although it was pathbreaking in many ways, treat planners' behavior casually ,
leaving most of that behavior outside the model :
A complete macroeconometric model of a CPE must start from, and buil d
on, both of these strands of research . On the one hand it must take accoun t
the role [.p lanners themselves play as a key macroeconomic adjustment mechanis m
in the system . In a CPE, when the terms of trade deteriorate, and the trad e
balance and foreign exchange reserves begin to fall, it is planners (and no t
automatic mechanisms such as the price system) which move to bring the econom y
back into balance through a myriad of measures beginning with postponing ad justment by borrowing, then introducing policy changes (including changes i n
centrally-determined prices), which in turn affect real variables (exports ,
imports, the growth rates

of

final demand elements, and so on) . A model o f

CPE without specific planners' reaction functions is a model without th e

major adjustment mechanism operating in CPE's . Predictions or countrerfactua l
simulations in such a system arc impossible, unless the model builder i s
willing to provide the missing link by "playing planner ." When CPE planner s
themselves are using a forecasting model built for their use, that may be a

perfectly appropriate strategy . But when outsiders build the model, the y
should try as much as possible to integrate the regular elements of planners '
behavior into the model .
A macroeconometric model with endogenous p lanners ' behavior cannot, o f
course, capture all that trans p ires in macroeconomic policy and performanc e
in a particular CPL . Planners do have tremendous power to change key macro economic parameters without very much warning, and they can do so in way s
which are difficult to predict . The working hypothesis behind the curren t
effort is not that all planners' behavior can be explained, but that enoug h
of that behavior is regular and explainable so that it can and should pla y
an integral role in the macro model .
A macroeconometric model of a CPE which includes planners' reactio n
functions makes it possible to obtain answers to several important question s
which are quite difficult to address in any other way .

For exam p le, how hav e

planners reacted to the OPEC induced terms of trade deterioration in th e

CPE's, how has that affected. economic performance and the distribution o f
aggregate demand, and how in turn has that affected plan s-? Is there evidenc e
that planners act as a destabilizing influence in the system? How do th e s e
systems respond to other exogenous shocks such as had weather, which drive s
down agricultural output, national income, and the balance of trade? Th e
answers to these and other questions are important both for increasing ou r
understanding of how central planning actually works in CPE's, and also fo r
enhancing our ability to predict the performance of these economies in th e
future .
Section II discusses a conceptual framework for constructing a macro model with planners' reaction functions, and the impediments to such a n
approach presented by data typically available on CPE plans and performance .

Section Ill presents a small macroeconometric

planners' reaction function . Section

model

IV

utilizes

for Hungary which include s

the model in a

counter-factual simulation which expl ores what might have hap pened to Hungarian micro economic plans and performance if the OPEC price changes, the resultin g . term s
of trade change

and the

world

recession, had n ot transpircd in the

. Section V briefly discusses future research possibilities.

mid-1970's

II .

Conceptual Framework and General Empirical Constraint s

The concern here is with macroeconomic policy and performance i n

CPL, which removes from direct consideration much of what is traditionall y
discussed as "central planning" in CPE's. The goal is to construct mode l
which explains movements in the macroecenomic agg r egates -- consumption, in .roAlepfanthsmyidcreonmgthsare ,ndthrolesag tplyindermgwhatplnersdo
vestment, government spending, and net exports, and which illuminates th e

functions which planners perform--determining individual investmen t

projects, the output of individual products, the structure of exports an d
imports, and so on--are not explicitly considered here, except where thes e
microeconomic decisions have macroeconomic consequences .
This section discusses the conceptual framework which motivates th e
specification of the macroeconometric model . This conceptual framework i s
suitable for constructing a macromodel of any CPE, even Hungary where th e
New Economic Mechanism implemented a major reform almost exclusively directe d
at microeconomic plannin g , while leaving the macroeconomic controls of th e
old system basically intact . The framework is discussed in terms of a linea r
model for convenience of exposition only . In the form discussed it is no t
estimatable using data publically available in any CPE, and this section end s
by outlinin g the compromises one must make in estimating this model for an y
CPE, and the.probable consequences of thos ecompris

Conceptual Framewor k
Planners in CPE's have a range of macroeconomic variables which they us e
as measures of the performance of the system, and as signals when somethin g
has gone awry . For a typical CPE the list of these variables would include

national income (Marxian-value added in the material sphere), consumption ,

investment, measures of supply/demand imbalance in those two areas ,
ments is balances with the Ruble (non-convertible currency) and Dollar
tlble currency) areas, and (an obviously related variable) foreign exchang
(conver
e
reserve -

These are, to borrow a term from control theory, the " stat e

variables" for planners, so-called because they are used by planners t o

measure the state of the system . Planners will set targets for these var iables, and when performance deviates from target, they eventually will (an d
usually must) do something . These state variables will be designated wit h
the n-element vector "s" ; the values of the state variables in the time " t "
are s

t
Planners in the CPE use various policy instruments to influence th e

values of the state variables . They control the level of investment throug h
various taxes which draw investable funds into the government budget, throug h
physical controls on the share of national income produced plus import s
actually devoted to investment, or (indirectly) through controls on persona l
incomes, hence consumption . Consumption itself can be controlled throug h
using many of the same: instruments, as well . as price and income policy (whic h
is related to tax policy) .

Net exports can be controlled through the use o f

exchange rates, subsidies and taxes on foreign trade, and direct controls .
Again, borrowin g from the terminology of control theory, I shall nam e
variables "control variables," so called because planners can manipulate th e
variables, and in so doing have an effect on ( " control " ) the states . Th e
control variables will be designated by the m-element vector "c", and th e
value of the control variables in year "t" as c t .
Finally, there are variables outside of planners' control, which nevertheless affect the state variables, hence the performance of the system .

7.

These include the weather, economic performance in other countries, the term s
of trade (which for most CPE's, in most of their trade relations, are given) ,
population, and so on . These I shall designate using the k-element vecto r
"e", whose values in year " t " are given as e

The definition of planners follows from what has been said above . the y
are the individuals with authority to set values for

the controls . In al l

CPE's this includes the central planning board, which is a committee report directly to the Council of Ministers .

It also will include variou s

other ministries and committees attached to the Council of Ministe rs wit h

general responsibility for various areas of the economy ; but i n the case

of

macroeconomic planning it excludes ministries solely responsible for th e

production and distribution of individual p roducts .
State variables are determined by the control variables and the exogenou s
variables .

(2 .1) s t

= Ec t + Oc t

The matrix "E " is nxm, representing the reduced form of equations describin g
the links .between planners' "m" control variables and the "n" stat evaribls
The
The matrix "O' ' is nxk . , giving the links between the "k"
"k "

exogenous variable

. For convenience the system is pictured

s,andthe" varibles
s a
linear, and simultaneity among the state variables is ignored .
A l though the Matrix E posits a link between controls and states, it i s
no way assumes that planners can exercise absolute controls over the states .
Planners in CPE's share their power with many other interest groups in th e
system (the ministerial bureaucracy, enterprises, individuals), in the mai n
because of the staggering information requirements associated with attempt s
at total central control . That planners share their power implies that, for

example, while they may move to restrain the level of investmen throug
various measures, other elements in the system (en t e rprises and th e
terial bureaucracy) will resist those measures with enough success that th e
actual level of investment, while lower than it would have been had planner s
not acted, will be higher than planners would have liked . The matri x
summarizes, then, the links between controls and states in light of th e
behavior of other elements in the system . And in that system, all one ca n
say is that planners can influence the level of the states with their controls, but they cannot determine the value of the states .
Note that in the matrix E the number of states and controls ar e

no t

equal, although for successful control of the system, there must be at leas t
as many control variables as there are state and exogenous variables . Th e
column
of of
E the
expresses
thevariables;
relationship
between
the " mth"
"mth"
control
each
"n" state
and in
any interesting
mode presumably

fluence on several states . For example, a decrease in profits taxes whic h

increases total investment will also increase machinery imports, which wil l
cause the balance of trade in one or both currency areas to fall .
There are limits on the values of the various control variables eithe r
for technical reasons (for example the investment share of national incom e
cannot exceed 100 percent) or for less easily quantified social reasons (fo r
example, if consumer prices grow too rapidly, there may be demonstrations i n
the street) . Some of these constraints are interconnected ; for example ,
nominal wages and the level of consumer prices are really constrained vis-a vis each other . For simplicity, the interconnections will be ignored and i t
will be assumed that constraints on control variables take the following for m

(2 .2)

ct < c t

< ct

u
where c 1t and c t are, respectively, vectors depicting lower and upper bound s
on the control variables in time " t" . The bounds will presumably chang e
over time, but such refinements need not be pursued here .
The state variables are also subject to constraints, the most obviou s
one being that C+I+G+X-N = GNP . Although planners may not think in terms of ,
or even keep accounts concerning, GNP, that constraint is still present an d
will, somehow, make itself felt . In addition, there may be constraints o n
the balance of payments deficit in either currency area (due to fund
s av ilable to fina ce them), or ther may be constraints flowing from the po ulation's general notion of an acceptable minimum growth rate for per capit a
consumption . These and similar constraints can be depicted in the inequality :

(2 .3)

kt

> Ks t

where "k " is a c-element vector of maximum values for state variables ,
t
individually and in combination ; and where the constraining equations ar e
included in the qxn matrix, " K"

For example, one row of " K" will designat e

the Keynesian equality, and the same row of "k" will be GNP .
Equations (2 .1) - (2 . 3) depict the influence planners and exogenou s
forces exert

the state variables, and the constraints planners face i n

exercising their influence . These equations in themselves outline a mode l

which would be complete if it were being built For planners, since they coul d
fill in the controls based on their own preferences over the states . Like wise, it would be sufficient to stop here if one wished to " play planne r " by ,
for example, adding a quadratic loss function in the control and state variables, and then applying optimal control techniques to solve for values o f
the controls over time . While both of those approaches have their merits i n
particular circumstances, for reasons discussed earlier, a third path shall

be pursued here in which planners " reveal " their preferences through empiri cally-estimated planners' react ion functions .
Planners have preferences over the states and, based on those, as wel l
as p ast performance of the system, assume they set targets for the states i n
the following way :
P
a .A(s
i=l

t- :i,

- s

t-i )

where s t is desired values for the state variables in time " t " and s t is th e
actual values ; A is a nxm element matrix where the on-diagonal elements re late deviations of the ith state variable from its desired value in "t-i" t o
the desired value in " t " , and the off-diagonal elements link deviations fro m
desired performance in the ith state variable to changes in the desired performance for the jth state variables (e .g . deviations in the balance o f
trade from desired last year affects the investment plan for this year) ; th e
a' s are weights planners place on each of the t-i previous years' performances ; and "p ' " is the number of time periods back that influence planners'
targets in year t . Planners set targets for investment, consumption, nationa l
income, and so on, based on actual performance and plans in past years, presumably placing the highest weight (highest a's) on what has happened mos t
recently . Equation (2 .4) suggests that planners learn from their system wha t
is possible, with a lag .

The " A " matrix embodies their preferences ; in a com -

ilex fashion . If, for example, they have, strong preferences for a . particula r
level or growth rate of consumption, they will " learn " very slowly if the y
are wrong in either direction .
Given these targets, assume that planners behave in the following wa y
when they determine the controls :

11 .

( 2 .5)

Ac

(2 .6)

ct

= B(st-1 - s

= ct-1 t Ac

t- l )

The matrix " B " is dimensioned mxn, and it denotes the link between deviation s
.of state variables from their desired levels and changes in controls . It i s
assumed that the planners only look one period back, which is probably clos e
to the truth . Notice that because there are more control than state variables, then the deviation of a particular state variable from its desire d
value last year will usually trigger changes in more than one control thi s
period .
These e q uations (2 .1)-(2 .6) form a complete system which, given th e
dynamics of the exogenous variables, describes the dynamics of the CPE . Th e
dynamic behavior of the system is determined by the coefficients in the equations, including the planners' reaction functions . One could specify a particular version of this system, solve it analytically, and specify stabilit y
conditions ; or one could estimate a particular version, and examine it s
dynamic properties through simulation experiments . The latter approach i s
used here, using the model discussed in Section III . Before turning to tha t
though, it is necessary to discuss how data limitations common to all CPE' s
force significant compromises in actually estimating the system outlined i n
equations (2 .1) - (2 .6) .

The Intact of Data Availability on Estimating a CP E

Macroeconometric Mode l
There are several difficult problems which arise whenever one attempt s
to build a macroeconometric model of a CPE : (1) Most of the control variable s
planners use are either difficult to quantify or, even if quantifiable, ar e
not reported in regular statistical publications ; (2) data on planned (i .e .

desired) values for the states are not available in all years, or in the sam e
and consistent format over the years ; and (3) frequently the published dat a
do not permit a full reconstruction of the macroeconomic accounts .
The first problem is the most serious one, and it is probably bes t
illustrated with an example from market econom ies . In the U .S . economy on e
of the key states is the rate of unemployment, and one of the key controls i s
the federal budget deficit . If unemployment deviates substantially from th e
level policy-makers have targeted for it, they may move to change governmen t
expenditures or taxes, which changes aggregate demand, production, income ,
consumption, and so on . One can clearly see the control that the polic y
makers use . In CPU's there are almost no controls of that sort . Many controls are administrative and typically difficult, if not impossible, t o
quantify . In the case of investment, for example, the problem in most CPU' s
today is that it can grow too rapidly, which causes some combination of short ages at home and a deteriorating balance of trade . The way central planner s
usually control investment is to freeze (or at least slow down considerably )
the initiation of new projects, in order to take pressure off of aggregat e
demand and the balance of trade, and then push for the completion of project s
in process, in order to speed up the addition of new capacity . The freeze i s
an administrative measure which, while it may be implemented in part throug h
the banking system, is im plemented primarily through the ministries .

It i s

in short, unquantifiable ; at best, our might be able to capture some of th e

effects through the use of a binary variable in an investment equation . Th e
same problems exist for controls designed to affect the level of consumption ,
most of which is handled through the enterprise wage funds, and through

taxe s

on consumer goods . Again, these controls are primarily administrative, an d

difficult to quantify . In short, much of the vector "c" is unobservable as

13 .

far as an outsider is concerned, and the question is how to proceed withou t

that crucial vector .
There are two ways to resolve this issue . One is to comb the literatur e
discussing economic policy in a particular CPE in hopes one could construc t
ct of binary variables describing changes in controls . This would be a n
enormous task, and the results could well be disappointing since some administrative measures are not announced, and many are much more subtle tha n
either being " on " or " off . "

The second way to resolve the issue, which i s

followed here, is to use changes in targeted-values of the state variable s

as indications that controls have changed . Thus a significant change in th e
planned growth rate of investment (relative to past years) is taken to indicate that controls are being changed in order to affect the desired change i n
investment . In terms of the equations above, the system reduces to :

(2 .1')

s' + Oe
t
t
:: , u

is 1

St

St

< S t
..

and

equations (2 .3) and (2 .4), where s

and

u
t

are, respectively, bound s

on the targeted values for the states . For example, one line in the inequal ities in (2 .2) might state that planned aggregate demand . must equal planne d
GNP .
Equations (2 .1 1 ), (2 .2'), (2 .3), and (2 .4) form a semi-reduced form o f
equations (2 .1)-(2 .6) because the controls have disappeared . They are a
less satisfactory version of the model because one can no longer explicitl y
follow the links between performance and controls, and between controls an d
performance . Still, as the empirical model in the next section illustrates ,
they do allow at least a partial depiction of planners' role in the system b y
tracking the interrelationship between plans and performance .

1'i .

In either case, it is also obviously important that planners publish goo d

data on their plans, namely, on the desired state variables . In all CPC' s
the annual plan is the operational plan, hence the appropriate source for

p la n

values, and in many CPE's the published information on the annual plans i s
extremely scanty . Some years g rowth rates are given other years levels an d
in yet other years nothing is available for certain variables . In thi s
spect Hungarian data are relatively good, although not without problems . Thu s
it has been possible to construct time series for s , and I will discuss thos e
in the next section . Not all of the important elements of

are published i n

Hungary, and in some other CPE's so few are published, that the approach use d
here might be impossible to implement empirically .
Finally, in many CPE's the published income accounts are inadequate i n
order to construct time series for all the elements of aggregate demand . Fo r
many CPE's this means it is impossible to quantify all of the constraints (an d
some of the structural equations) discussed above . Again, the Hungarian cas e
is unusual here, and it is possible to construct a full set of accounts on th e
product side .

15 .

III . The Hungarian Econometric Model No . 1 (HEM-1 )

An Overvie w
Hungary is a small country of ten million people at a level of economi c
development roughly equivalent in terms of per capita GDP to that of souther n
Europe . It is highly sensitive to developments in the world economy--export s
comprise over 40 percent of GDP--and it is truly a "price-taker" on worl d
markets . The economy has for over thirty years been centrally planned and ,
although much has changed in the way microeconomic decisions are handled i n
the system, macroeconomic decision-making has always been primarily under th e
control of central planners . The Planning Office, the Ministry of Finance ,
the State Bank and the Development Bank together wield considerable contro l
over the macroeconomic aggregates today, as they have for over a quarte r
century . 2
The Hungarian Econometric Model No . 1 (HEM-1) is a twenty-one equatio n
annual model (seventeen stochastic equations and four definitional equations )
estimated using data for the 1966-77 period.3 The model focuses almos t
exclusively on the interrelationships between the foreign sector and macro economic activity, interrelationships which are of fundamental, importance i n
the Hungarian economy, as in all East European economies . As this mode l
shows quite clearly, it is the dollar balance of trade (an
tible currency reserves) which disciplines planners
dlimteconvr
to search fo
balanced p ]

rmacoenily

All the data on actual economic activity (with the exception of a serie s
on GDP in the European Community) are from official Hungarian statistica l
sources ; and all income flows are in constant (1968) forints . Data on plan s
have been collected from Hungarian newspapers, and from official documents

16

published by the Planning Office . The data are discussed in somewhat mor e
detail in the notes to Table 7 .
Figure I presents a schematic view of the major interrelationships i n
HEM-1 ; Table 1 lists the variables, and their definitions, and the dat a
sources ; and Table 2 presents the model e q uations and the diagnostics associated with their estimation . I will first discuss the model using Figure 1
in order to provide an overview of its structure, and then turn in the nex t
subsection to a discussion of the individual equations .
Figure 1 contains all of the model variables and their major interconnections . Arrows indicate the direction of causality ; dotted lines mean tha t
the coefficient is negative ; and solid lines indicate the coefficient i s
positive : Variables enclosed in hexagons are planners' variables ; variable s
enclosed in circles are exogenous to the model ; and variables in rectangle s
are for the most part activity variables .

18 .

Table I

Variables
HEM-1

in

(Note : Exogenous variables are indicated by a The planned value s

of variables a r e indicated by the suffix " -PL" , for example, XPL is th e
planned value of X .)

Endogenous Variable s
(Numbers in parentheses indicate data sources, given in notes

below )

Balance in all trade for dollars (and othe rconvetibl

currencies), millions of dollars
)
(1
BOTR - Balance in all trade for rubles (and other inconvertibl
currencies), millions of rubles (1 )

- Total consumption, millions of 1968 forints (2,4 )

CPL - Planned total consumption, millions of 1958 forints

E$ Quantum

index of dollar. exports , 1970=100

(3 )

(1 )

Quantum index. of ruble exports, 1970 = 100 )(1

Government expenditures, of 1988 forints

(2,4 )

GDPDEV Ratio of actual GDP to GDP predicted from a second degre

polynomial time trend (8 )
I

Total investment, millions of 1968 for i nts (2, 4 )

IDEV -- Ratio of actual 1 to I predicted from a second degre
polynomial time trend (9 )

- Planners' estimate of I, millions of 1968 forints

IINV

IPL

- Planned value of I, millions of 1068 forints (3 )

(3 )

Inv e ntory investm e nt, millions of 1968 forints (2,4 )

K- ValueofthcpialsoknJury1,depciat
millions of 1968 forints

,
(4, 5 )

KNET - Depreciated value. of the capital stock on January 1, million

of' 1958 forints (7 )
M$

- Quantum index of dollar imports,

19 70= 100

(1)

MR

- Quantum index of ruble im p orts, 1970= 100 (1 )

NI

- National income (Marxian), millions of 1968 forints (2 , 4 )

NIPL

- Planned NI, millions of 198 forints (3 )

Net e x p orts (in the national inco m e accounts), millions o f
1968 forints (2,4 )

NETEX

Exogenous Variable s
GDPECDEV - Ratio of actual GDP in the European Community to GDP predicte d
from a second degree polynomial time trend (6 )
N

- Total employment, thousands of persons (4 )

-

Price index for dollar exports, 1970= 100 (1 )

PE P

Price index for ruble exports, 1970= 100 (1 )

PMR.

- Price index for ruble imports, 1970- 100 (1 )

PM$

- Price index for dollar imports, 1970= 100 (1 )

TIME

- Time, 1960 = 1

Data Sources :
(1) Kzponti Statisztikai Pivotal (KSH : Central Statistical Office) ,
Klkerereskedelmi statisztikai vknyv (Foreign Trade Statistica l
Yearbook), various issues .
(2) KSH
Fbb, npgazdasgi folyamatok
1976 and 1977

folyamatok (Main National Economic Processe s

(3) Information
on planned values was obtained from Tervgazdasg i
ntesit
(Economic Planning Gazette), official journal of the 'Planig Hung-Tan
Office ; and from Npszabads g, the Hungarian Socialist Workers' Percy
daily newspaper . Planned values for investments are typicall ygiven
forints, and have been deflated here using a price index for investment .
Plans for consumption and national income are given in growth rates, an d
the levels here are obtained by multiplying the planned growth rat e
times last year's consumption or national income (in 1968 prices) .
(4)

KSH , Statisztikai vknyv (Statisti ca l Yearbook), various issues .

(5)

KSH, Beruhzsi adattr 1950-1977 (Investment Data Series, 1950-1977) .

(6)

The GDP series is constructed from data in the United Nations Statistical Yearbook, 1977 and OECD, Main Economic Indicators (May 1980) . Th e
estimated time trend used to define the deviations is given in equatio n
(3 .19) .

(7)

Constructed from K, assuming a depreciation rate of 4 percent (see

equation (3 .21)) .

(8) Constructed
using an estimated time trend whose coefficients ar e
given
in equation (3 .19) .
(9)

Constructed using an estimated time trend whose coefficients ar egivn

in equation (3 .20) .

Table 2
1*
E q uations in HEM -

Planners' Reaction Functions

Table 2 (Continued)

Table 2 (Continued )

*Notes :

Mean values for the dependent variables are in parentheses below those variNumbers in parentheses below the coefficients are the standard errors .
ables .
For each equation, the right-hand side of the table indicates (corrected )
the standard error of the estimate, the Durbin--Watson statistic, and what (i f
any) estimation procedures were used to correct for serial correlation in th e
in the last three column s
error terms (see the text for a discussion) . A "
of this table indicates that corrections for serial correlation we r e unnecessary .

24

Planners forecast

the supply

of national income

available

base don

1:
recent trends in factor
inputs . That influences plans for
aggregate deman d

of consumer and investment goods, and those plans (along with GDP and ne t
ex p orts) affect
spending

is

the

actual level of consumption and investment . Governmen t

directly affected by planned national income (there are no pla n

data available for that aggregate demand category) . Net exports (for whic h
there is no published p lan information) are affected by exports and imports ;
and they in turn are negatively related to the planned levels of consumptio n
and investment and to each of the final demand categories .4

Inventories ar e

a residual, balancing the difference between GDP and other final deman d
categories .
Actual GDP produced is determined by capital and labor inputs, capita l
being a function of previous investment levels . Export and imports for dol lars (convertible currency) and rubles (non-vertible currency) ar e
mainly determined by macroeconomic activity variables and lagged balances

of

trade (although GDP in the EEC does influence dollar exports) . The balance s
of trade are determined by exports and imports, and the prices of exports an d
imports .
The structure

of

the model is best illustrated by following through a

somewhat simplified version of the results of a change in an exogenous variable, for example an increase in the dollar price of imports in year t . Thi s
causes a deterioration in the dollar balance of trade in year t, and nothin g
else happens that period. The following year dollar imports fall as planner s
seek to improve the trade balance, which increases net exports, decreasin g
planned consumption and investment, and all actual final demand categories .
The fall in investment cuts dollar and ruble imports somewhat, : t . The third yea r
dollar exports increase to further improve the trade balance . The consequences

2 5.

continue to work through the system, but they are too complicated t oflw
through
change of
and make simple qualitative statements on the direction o f
particular variables . Further discussion of the effects of terms of t r ad e
changes in the model will be deferred to Section IV which traces through th e
actual impact of the drop in Hungary's terms of trade steming from th eOPEC
price changes .

Estimation of the Equation s

Table 1 lists the variables in to model . Table

lists the equation s

and the diagnostics associated with their estimation . The model has bee n
estimated and simulated using the TROLL system . The coefficients have bee n
estimated by Two-Stage Least Squares (2SLS) using Principal Components . Th e
2SLS technique is not only appropriate in a model . such as this where endoge nous variables appear on the right--hand side of other equations, but als o
early ex p eriments with a preliminary version of HEM-1 indicated that th e
2SLS coefficients provided better within-sample simulation results tha n
coefficients estimated using Ordinary Least Squares .
In many of the equations early estimates suggested rather complicate d
autocorrelation or the residuals . Further attempts to respecify the equation s
in order to identify the source of the unexplained pattern were only partiall y
successful, and therefore it was necessary to utilize Cochrane-Orcutt-typ e
iterative procedures to simultaneously estimate the coefficients characterizing the p attern of serial correlation, and the coefficients of the equations .
In some of the equations the error process seemed to follow a two perio d
autoregressive scheme (called here AUT02) . 6

In other cases, the erro r

process was best described by a one or two period moving average (respectively ,
MAV1 and 51V2) . 7

In some cases it would appear that the error processes were

26 .

even more complex than these three possibilities allowed, however the limite d
number of degrees of freedom precluded explorations into more complicate d
error processes .
Table 2 lists the estimated coefficient :: for the stochastic equations ,
and the coefficients in the definitional equations . For each equation, th e
number in parentheses under the left-hand side variable is the mean value o f
that variable ; numbers in parentheses under the coefficients are the standar d
errors of the estimated coefficients ; and the far right-hand side of th e
table records, respectively, the corrected R for each equation, the Standar d
Error of the Estimate, the Durbin-Watson statistic, the technique used (i f
any) to correct for autocorrelation of the residuals, and--where correction s
were made--the resulting coefficients .

Planners' Reaction Function s

At the macro level planners serve two functions : they projec t
aggregate
supply, and they balance aggregate demand against that projected aggregat e
supply . The state variables for which they have targets are national output ,
the components of aggregate demand for that output (consumption, investment ,
government spending, net exports, and inventory investment), and the balanc e
of trade . The latter is probably the most important state variable fo r
planners in a small economy highly dependent on foreign trade . In such a n
economy, the ultimate operative criterion for macroeconomic balance is whethe r
or not there are foreign exchange reserves sufficient to finance net import s
implied by the plan, and service existing debt . If not, then no matter ho w
"balanced" the plan may appear to be in domestic prices, it must be changed .
Hungarian planners do not publish a complete version of their annua l
plans, the most unfortunate omissions from the viewpoint of this model being

targets for net exports, inventory investment, government spending, and ex ports and imports by currency area . As a consequence, the planners' reactio n
functions discussed in this section only cover a portion of the state variables for which planners set targets . The remainder of the reaction function s
are left implicit in the other equations of the model, and will be discusse d
as we proceed .
Planners tend to work in terms of growth rates of, and increments to ,
the state variables ; and this section of the model is specified to reflec t
that . Where possible, more general forms or these equations were estimated ,
which included the first difference specification as a nested hypothesis, an d
generally those specifications confirmed the first-difference specifications .
However, those general specifications were so plagued by multicollinearit y
problems, that it was decided in some cases to accept the first-differenc e
specification as adequate based primarily on East European planners' well known propensity to work on the margin .
Dating is very important in these equations . In each case they ar e
explaining the variation in plans for the activity variables in year "t" .
Planners make those plans late in year "t-l" using preliminary estimates o n
activity variables for " t-1" and (usually final) estimates of activity vari ables for t-2 .
Equation (3 .1) models the planned increments to the supply of nationa l
income as a function of recent changes in factor supplies . Planners seem t o
know with a great deal of accuracy what (barring poor weather) national in come produced will be the following year, and they seem to base that primarily on their predictions of the growth of factor inputs .
Equation (3 .2) and (3 .3) model planned increments to consumption an d
investment as a function of the same variables or types of variables . The

planned increment to consumption this year is positively linked to the planne d

increment to national income . It is negatively linked to the differenc e
between actual and planned consumption, reflecting the fact that planner s
constantly readjust for surprises in previous years . In fact the coefficien t
suggests that planners "take

back" ( " give back " ) about 80 percent of the pre-

vious year's overfulfillment (underfulfillment) of the consumption plan .

Finally, the increment to planned consumption this year is negatively relate d
to net exports this year . Ideally this should be planned net exports thi s
year, reflecting the fact that an increase in planned net exports must caus e
a decrease in planned consumption, investment, and so on . Because there ar e
no data on planned net exports, actual net exports are used as an admittedl y
imperfect proxy variable .
Data problems concerning planned investment create special difficultie s
whose conse q uences can be seen in the several insignificant coefficient sfor
equation (3 .3) . The difficulties are related to the timing, and preliminar y
data discussed above . It turns out that the "t-l" estimates of investment i n
Hungary have traditionally been low, particularly in the 1960's where the y
might easily be 10 percent below what investment actually turned out to he .
Equation (3 .4) models planners '

estimates of actual investment as a functio n

of time, the actual increment to investment, and a correction factor equal

on e -third ofthe previous year ' s difference between actual and estimate d
investment .
Because of this problem, the difference between planned investment i n
year " t " and what investment finally was in the previous year is usuall y
negative, which would imply that planners constantly planned to decreas e
investment . Obviously that was not the case ; on the contrary they planne d
most years up until . very recently to increase investment over what they

to

29 .

thought it had been the previous year . It is the difference between planne d
investment and what planners estimated investment to be the previous yea r
which forms the dependent variable in equation (3 .3), because that reall y
explains how planners behaved in their investment plans in the context o f
economic performance as they saw it at that moment .
The investment equation is designed to operate identically to the consumption equation . Planned increments to national income are positivel y
linked to increments to planned investment, although the relationship i s
statistically insignificant . As with consumption, when estimated investmen t
rises above (falls below) plan in one year, planners cut (increase) planne d
investment the following year, but only by about one-half of the overfulfillment . Net exports are linked to planned investment with an unexpecte d
sign, and the coefficient is insignificant . The insignificant results her e
may in part be a reflection of the use of estimated investment, but they ma y
also have more substantive foundations . In reading the annual plan documents ,
and the discussions of what considerations go into the annual plan figures ,
one gets the impression that planners plan increments to investment not o n
the basis of aggregate supply, but mainly on the basis of whether the investment goods p roducers (particularly construction and construction material s
producers) can handle the demand--for which equation (3 .3) has a proxy i n
the overfulfillment of the investment plan--and whether there are balance o f
payments problems which require a reduction of investment . While I coul d
find no direct link between the planned level of investment and previou s
balance of payments experience, there is a strong link through net export s
in the investment equation, of which more below .
Equation (3 .5) models government spending as a function of planne d
national income and net exports (as a proxy for planned net exports) . The

im p licit assumption here is that the actual value of G is virtually th esam

its planned value, which in fact is probably the case judging from th e
stability of the data series and the fact that planners have absolute contro l
over the major portion of this category of aggregate demand .
As a group these planners '
nature

equations testify to the apparentl

of plans as a macroeconomic adjustment mechanism . A ystabilz- rise in tne

exports does "crowd out " planned consumption and government spending ; an d
while it has no significant effect on planned investment, we see below tha t
it does operate directly and negatively on actual investment . When actua l
aggregate demands deviate from planned, planners move to change planned ,
therefore actual, consumption and investment to keep the system on track .
Finally, there is a direct link between the planned level of aggregate suppl y
and all of the final demand categories, with planned consumption and government spending receiving (at zero net exports) 75 percent of the planne d
increment to national income, the rest being reserved for investment, ne t
exports, and inventory accumulation .

Foreign Trad e
Hungary trades in numerous currencies, which the statistics (and th e
planners) divide into dollars (convertible) currency trade and ruble (non con v ertible) currency trade . Until very recently, published data on plan s
for these four trade flows are very scanty, and as a consequence it prove d
necessary to include in the foreign trade equations reduced-form links whic h
implied planners' reaction functions . Thus, the dollar balance of trad e
appears in three out of the four equations, with the implication that whe n
that trade balance deteriorates, planners take measures to cut dollar imports ,
increase ruble imports, and increase dollar exports .

Equation (3 .6) models the first difference in dollar exports as

func-

tion of the increment to GDP (a supply variable), the change in the dolla r
balance of trade two periods back (as the balance of trade deteriorates ,
ex p orts increase with a two period lag), and the deviation of European Community GDP from a long-term time trend . This latter vari able captures th e
effect of changes in demand conditions for Hungarian exports on their majo r
dollar markets, and it is specified to reflect the fact that Hungary is

marginal su p p lier on those markets, a kind of " last in, first out" supplier .
In fact, while the coefficient is about three times its standard error, i t
is surprisingly small . It would be unusual for EEC GDP to deviate more tha n
4 percent from its trend (judging from the sample period), yet such a deviation would only change exports by

.12 index points, a little over 1 percen t

of the average increment to dollar exports in a given year . Thus this equation suggests that Hungary is not greatly affect e d by fluctuations in Wes t
European economic activity .
The two period lag in the dollar trade balance explained far more tha n
a one-, or one- and two-period lag, which may simply reflect the fact tha t
when the balance of trade deteriorates (improves) , it may take a while t o
seek out new customers in the West and make new contracts (finish ol d
contracts) .
Dollar imports are purely demand determined by GDP, last period' s
balance of trade (again a proxy for planners' intervention to cha n g e dolla r
imports as the balance of trade improves or deteriorates), and the deviatio n
of investment from its trend . The latter has a positive sign (as expected )
but an insignificant one . The coefficient is retained nevertheless becaus e
there is ample evidence that when investment surges in Hungary, dollar imports increase . It could be if the GDP term were replaced by' a more precise

32 .

term reflecting demand for non-investment importables, that the investmen t

coefficient would be significant .
The first difference in ruble exports are treated here simply as

lagged function of the increment to ruble imports, which is indeed a fai r

representation of the bilateral trade relationships in Hungary's trade wit h
other CPE's . The RHO coefficients sum up to 1 .4, reflecting in part th e
absence of the increments to imports in "t", and in part to the fact tha t
the net barter terms of trade have deteriorated over this

p eriod, thus s o

have the gross barter terms of trade .

Ruble imports are a function of all of the variables used to explai n
dollar imports, including the dollar balance of trade last period, and th e
ruble balance of trade . Again, GDP is a demand term, and the coefficien t
is identical to that on dollar im p orts . The ruble balance of trade las t
period has the expected sign, but is not significant, which is reasonabl e
since unplanned fluctuations in this variable are less likely than for th e
dollar balance of trade . 9

The dollar balance of trade enters negatively ,

and is two times its standard error, suggesting that a deteriorating dolla r
balance of trade will cause planners to increase ruble imports, making u p
for about half (compare this coefficient of

.02 with the .04 in the dolla r

balance of trade term for equation (3 .7)) of the decrease in dollar imports .
Ti e deviation of investments from trend also enter this equation with

positive, but insignificant, sign ; and the coefficients is retained here fo r

the same reason as above .
The export and import flows in equations (3 .6)-(3 .g) are quantit y
indices ; there is no published series in fixed domestic prices for export s
and imports by currency areas . Therefore it is necessary to combine th e
export and import quantity indices with price indices, and then link those

33 .

stochastically to the dollar and ruble balances of trade, which i

. .10) and (3 .11) .

sacomplihednquatos(3

Consumptio n
Equation (3 .12) models the change in consumption as a function

th e

deviation of GDP from trend (typically due to unusual weather', which affect s
agricultural production, hence real consumption), the change in net ex p orts ,
and the planned increment to consumption . The change in net exports a p pear s
here as well as in the planned consumption equation on the hypothesis tha t
when net exports change because of changes in exports and imports, planner s
take measures to directly affect real consumption (i .e . through price, tax ,
subsidy, and wage policy changes) which may not show up in planned consumption, even though they do comprise part of the plans for consumption . Equations (3 .2) and (3 .12) jointly suggest that a 1 Forint increase in net export s
will cause planned consumption to fall

.37 forints, which decreases actua l

consumption by .57 . .37 = .21 Forints, and in addition consumption fall s

another .26 Forint because of direct links between net exports and consumption, for a total of .47 . Consumption's share of GDP over the 1960-77 perio d
averaged

.60, which indicates that consumption has been somewhat protecte d

from net export fluctuations .

Production and Accumulatio n

Equation (3 .13) models the increment to the undepreciatcd capital stoc k
as a function of the previous two periods' investment . The actual lag i s
most assuredly longer than this, but equations with longer lags encountere d
severe multicollinearity problems . This equation does a fair job of trackin g
the sample period increments to capital stock, and it does quite well in th e
simulations .

Equation (3.1.4) uses a Cobb-Douglass production function to explain GD P

as a function of an exogenously determined labor force, and a capital stoc k
depreciated in equation (3 .21) . The equation fits well, and makes sense ,
suggesting a constant 3 percent

rd

e of increase in total factor

productivity .

The investment equation (3 .15) again raises the issue ofthe role o f
planners '

(biased) estimates of actual investment . This time the issu eis

what role those biased estimates play in determining actual investment . A s

planners discover early in year t that the "t- 1 " estimates were way too low ,
do they alter their plans and slow down the rate of investment? The hypothesi s
here is that they do not, that the inertia in the system is such that tha t
planned level of investment sticks, and more importantly that all the measure s
set into force for increasing the rate of investment work on the actua
meat to increase it. Therefore the first RHS variable of thisnlinvest equatio
last year's investment multiplied by the growth rate implied by planned in vestment over the (biased) estimate of last year's investment . If planners
plan in year t to increase investment 10 percent over what they thought i t
was in year

and if it turns out that their estimate of " t- l " invest .

was 10 percent too low, the actions they have taken (and cannot

sto p

soon enough) will act to increase investment 10 percent over the actua l
(higher) figure . The coefficient, a little under twice its standard error ,
suggests that there may be something to that hypothesis .
The second explanatory variable in the investment equation is ne texpors
the hypothesis here being the same as in the consumption equation : tha t
planners take measures to control . investment and make room fornetxpors,
aside from those they signal in the annual plan figures . In this case, a 1
Forint increase in net exports has no significant effect on planned
.33. Investment's
investm,budcreas tlinvmesdrctlyb
average

share in GDP over the sample period was

.27, suggesting that investment ,

unlike consumption, takes more than its share of the adjustment burden whe n
net exports change . Finally, GDP is included in the investment equation t o
reflect a necessary link between changes in national output and changes i n
investment necessary to increase the capital stock, hence p roductive capacity .

Income Accounts and Identitie s

Equation (3 .16) in the income accounts 1 . isle ; national income to GDP .
These series are virtually identical in their growth rates, hence the ver y
tight fit . The other equation, (3 .17), models net exports as a function o f
the four trade series . A stochastic equation is necessary here because th e
trade series are quantity indices, and, in any event, the national incom e
concept of net exports is not identical to exports and imports as counte d
in the foreign trade statistics .
Equation (3 .13) balances the system, treating inventories as a r e s idual . 1 0
The remaining equations in the system are identities defining endogenou s
variables used elsewhere in the model . The obvious identities (e .g . ,
C

= C -1 + AC) have been omitted .

Within-Sam p le Simulation Error s

Table 3 presents data on bias and Root Mean Squared Percent (RMSP )
Errors for all endogenous variables (excluding definitions) in the mode l
for the single equation forecasts and dynamic simulations over the 1964-7 7
period (the maximum possible given lags in several equations), and the 1970 77 period . For the inventory identity and the three equations which mode l
balances (BOT$, BOTR, NETEX) simple RMS errors are given, since the denominator can easily approach zero, leading to enormous RMSP errors .

3U .

The single equation forecast results are quite satisfactory, predictin g

the endogenous variables with only a minimal amount of bias . The investmen t
equation is one exception, where the RMSP error is over 5 percent, and th e
upward bias is almost that large . However the bias decreases during simulations, and in fact for the 1970-77 simulation there is only tiny bias and a
considerably lower RMSP error . The foreign trade equations are somewha t
more problematical, particularly dollar imports . These are never easy flow s
to model in any economy, and in addition, the reliance in this model on

few aggregate variables could exacerbate the problem . For example, the demand for dollar imports is basically composed of food, raw materials, an d
machinery, and variables reflecting those particular flows might do a bette r
job of explaining dollar imports . In the interest of maintaining the smal l
size of the model, these possibilities of dissaggregation were not pursued ,
and instead are left to a possible second version of the model .
Two sets of within-sample dynamic simulation results are presented ,
the first--1964-77--because it covers the entire sample, and the second- 1970-77--because it covers the period of most interest for further simulations .
Most of the interesting macroeconomic shocks hit Hungary in

the 1970's, an d

therefore that is where the most interesting possibilities exist fo

.

rcounte-falsimuton

The 1964-77 simulation results . are not terribly satisfactory, althoug h

they do an acceptable job, particularly for the major macroeconomic aggregates . Some of the RMSP errors are high, and most of the simulated variable s
are u p wardly biased by 1 to 5 percent .

The largest increase in error : ; ove r

the single-equation errors is in the three balances, particularly the BOT R

and =EX . All of those balances are the net result of the four foreig n
trade equations, which are the least satisfactory of the model, consequentl y
this result is not surprising .

37 .

Table 3
HEM-1 Properties o f
Error

Endogenous
Variable

Single Equation
Forecast

1964-77
Simulation Simulation

1970-7 7

33 .

39 .

The 1970-77 simulation results are much better, which is important sinc e
this

is

the period on which the counterfactual simulations are focused . Her e

the RMSP errors are much lower than for the 1964-77 simulation, and, in man y
cases, they are lower than the single equation RMSP errors . None of th e
equations show more than a 2 percent b i a s. in either direction, and many ar e
far less than 1 percent . The errors on the trade balances and net export s
remain high, but with the exception of net exports, there is no significan t
bias in the
. values of th evaribls
The model considerably underestimates inventories in the 1964-77 simula tions, but does a much better job in the 1970-77 simulations . In fact th e
downward bias in the 1970-77 simulations for inventories is virtually th e
same as the upward bias for net exports, suggesting that it would be bette r
to estimate these together . However, the important role of net exports i n
determining final demand aggregates would be muddled in a net exports plu s
inventory tern, and for purposes of simulation, it seems better to maintai n
the current treatment of these two variables .
In sum, the 1970-77 simulations suggest HEM- .1 is a solid representatio n
or the Hungarian macroeconomy, useful for counterfactual simulations coverin g
that period . Before discussing those simulations, I turn briefly to th e
results of applying the model to a prediction of economic performance in 1978 .

Predictions of. 1978 Economic Activity Using HEM-1.

After the hulk of the estimation was completed, data for 1 .978 becam e
available, making it possible to check the model by predicting the values fo r
t year, and comparing them with the actual values . The predicted an d
actual values are presented in Table 4 .
This year was one in which planners began to discuss major modification s
in the economic mechanism and the price system in order to dramatically

40 .

Table 4

Notes
a
aAbsolute error .

Predictonsf1973UigHEM-

increase enterpriSe efficiency and improve the dollar balance of trade (se e
Hewett [1980a] for details) . As a result, enterprises became uncertai n
about future prices and they accumulated inventories, primarily throug h
dollar imports . Thus, while the model accurately predicts the macroeconomi c
aggregates and the plans, it missed much of the unusual inventory accumulatio n
phenomenon, severely underestimating inventory accumulation and the dolla r
imports which resulted from it, and the balance of trade deterioration tha t
it caused, therefore overestimating net exports .

The Stability of HEM- 1

Even in a model of this modest magnitude, it is difficult to tell jus t
by "looking" at the equations whether they portray a stable system whic h
returns to equilibrium when it is perturbed . Obviously there are man y
negative feedbacks built into the model, so that instability could onl y
come through their complex interaction in an unanticipated fashion . Severa l
experiments with the model testified to its stability . In one case, fo r
example, the dollar terms of trade in one year (1973) were increased 10 per cent, and then the effects through 1973 were compared with those of a base line solution for 1970-78 (discussed in the next section) . There is no nee d
to go into the details of that exercise here . Briefly, what happens is tha t
the terms of trade improvement, hence balance of payments improvement set s
off damped oscillations in dollar imports (hence the dollar balance of trade ,
net exports, and the final demand elements) which are large but diminishin g
the first three years, and insignificant thereafter .

42.

IV .

Using HEM- 1 to Assess the Impact of Hungary' s

Declining Terms of Trade in 1974-7 8

One of the most important developments in Hungary's (and the rest o f

Eastern Europe's) economic environment in the 1970's was the precipitat e
decline in its terms of trade in hard currency, and then its torus of trad e
in Rubles . As Figure 2 illustrates, the dollar terms of trade dropped abou t
18 percent during 1974-78, most of that occurring in 1975-76 . The Ruble terms ,
of trade followed with a lag after the Soviet Union and its Lent Europea n
trade partners agreed to introduce the terms of trade changes into their ow n
bilateral trade through a lagged moving average . (Hewett [1980b] )
A decline on this scale in a country's terms of trade represents a
massive cut in the world value of its exportables, and hence creates th e
necessity for a combination of borrowing (assuming there are willing creditor s
available), and domestic austerity to increase the fixed-price value of ne t
exports . Naturally, borrowing only postpones the austerity to a later date ,
assuming the terms of trade changes remain . The question, then, for a country such as Hungary, is how much additional borrowing , they undertook in th e
197 1 1-present period in order to postpone adjustment to the new terms of trad e
shift, and accordingly, how much adjustment have they undertaken? HEM-1 wa s
constructed to provide an answer to such a question, and this section discusse s
the counterfactual simulation developed using HEM-l .
A base line simulation of HEM-1 for 1970-78 will be used to represen t
what actually transpired during those years . It is necessary to use thes e
simulated data, in lieu of the actual data, in order to separate out th e
deviations from actual performance which exist because the model cannot perfectly predict reality, from the deviations which occur because conditions

43 .

Figure 2
Hungarian Terms of Trade in Rubl e
and Dollar Trad e

Source : HEM-1 Data Bank .

in the counterfactual simulation are not the same as in reality . The implici t
assumption is that the errors in the baseline simulation and in the counter factual simulation are proportional, thus differences between the two simulations will accurately reflect what actually would have happened if the term s
of trade had been better . Even though 1978 data were not used in the estimation, 1978 is included in the baseline simulation because there are so fe w
years in which to study the terms of trade decrease .
To represent what could have happened, HEM-1 was simulated under th e
assumption that Hungary's terms of trade with the dollar and ruble area s
remained at their 1973 values through 1978 . 11

This introduces approximatel y

a 20 percent improvement in the dollar terms of trade (relative to baseline )

by 1976, and about a 15 percent improvement in ruble terms of trade by 1977 .
That in turn improves the trade balances in dollars and rubles, which subsequently causes changes in imports, exports, net exports, and therefore fina l
demands .

I turn now to a discussion of the results, which are presented i n

Table 5 . In all cases the terns "fall" or "rise," and so on, are relativ e
to the baseline simulated values .
Ruble imports fall, as the improved dollar balance

F trade allows a

shift to dollar imports, and ruble exports, which depend solely on rubl e
imports, fall by approximately the same amount . The net effect is to leav e
the ruble trade balance in surplus and substantially improved due to th e
improved ruble terms of trade over 1974-78 . 12

Dollar imports ruse consider-

ably, all of the increase coming in 1975-78 (due to the one year lag), i s
planners respond to the much improved foreign exchange reserve situation .
Dollar exports fall, but by a smaller amount, and with a two year lag . Th e
result leaves Hungary with a 1974-78 trade balance $1 .52 billion more tha n
baseline .

During the 1974-78 period, Hungary's net hard currency debt

Table 5

Baseline and Hypothetical Economic Performanc e

(Assuming
1974-78
No Changes in the Terms of Trade) :

Selected Endogenous Variables

Baseline d

Hypo theti cal a

Income Accounts (Planned an d

Notes :
`aAssumes dollar and ruble terms of trade remained at their 1973 value s
.
during 1974-78 .
bColumn (2)

- Column

(1 )

c Sum of trade balances over the 1974-78 period .

1970-78 simulated values based on the actual values of all exogenou s
variables .

Rati o

46 .

actually rose $3 .962 billion, of which $1 .552 billion was increased short -

term debt . (Hungarian National Bank) Thus this simulation suggests tha t
Hungary could have avoided all of that increased short-term debt if th e
terms of trade changes had not transpired . The increased long-term debt ,
a good deal of which had gone to financing new investments, might well hav e
been incurred in any event .

13

GDP itself changes only slightly due to the terms of trade changes, an d
that because investment goes up somewhat, which increases the capital stoc k
and therefore GDP. GNIPL is included here only for its growth rate to con firm that planners do not noticably increase GNIPL, which makes sense sinc e
all that has changed is the relative price of exportables .
But the final demand elements do change because net exports fall consid erably--due mainly to large increases in dollar imports--and the additiona l
g oods are distributed among final demand categories . This happens both in directly through the visible links between net exports, plans, and realize d
values ; and it happens directly (although planners '

actions are definitel y

behind it) in the link between net exports and the final demand categories .
Consumption benefits the most from the terms of trade improvement, lying

percent above the baseline simulation over the five years . Government spend4
ing,, and investment increase somewhat less . 1
in sum, the simulation suggests that Hungary has paid a considerabl e
price because of the terms of trade decline . Part: of the cost was in rea l
goods to the population, as planners adjusted the utilization of GDP downwar d
in order to free up net exports, hence reduce what otherwise would have bee n
a considerable hard currency deficit resulting from the terms of trad e
changes . Shaving 2 percent off consumption, and over 1 percent off investment and government spending on average over a four period is' considerably

47 .

more of an adjustment than Hungary (or other East European countries) ar e

usually given credit for .
Nevertheless the adjustment was not immediate, and the government borrowed $1 .S billion more than it would have in the absence of the terms o f
trade changes (which is all of the short-term borrowing actually undertake n
during 1974-78) . That debt, amounting to about one-half of 1978 dollar ex ports, can only be repaid through further austerity measures, which is precisely what the government introduced this year . (Hewett [1980a])

V.

Future Research Possibilities.

The fact that it was possible to identify significant links betwee n

economic performance and clans, and plans and economic performance, suggest s
t hat this is a promising approach to constructing macroeconometric models o f
centrally planned economies . There are several areas in which this mode l
could be improved, and several other possible applications for this, or a n
expanded version of this model .
it would seem useful and necessary to enla r g e this model at least enoug h
to include an agricultural sector, which is important in Hungary, into th e
supply side and into the foreign trade equations . This would make it possible to trace the effects of weather on the system, an important source o f
macroeconomic disturbances . Secondly, the importance of energy to Hungary' s
future economic perfo r mance suggests the importance of building energy int o
the supply side of the model, possibly along with other raw materials in, fo r
example, a KLEM production function . The crucial importance of this latte r
change is that it would link imports to aggregate supply, an important an d
real link in the Hungarian economy .
In 1980, as part of its austerity program designed to improve the he r
currency trade balance, the Hungarian government introduced stiff controls o n
imports, which appear to be exerting an effect on production . With a mode l
modified as discussed here, it would be possible to explore the consequence s
of such a move for aggregate supply, and therefore for the aggregate deman d
categories . In addition, by adding explicitly an agricultural sector, i t
would be possible to explore the potentially important contribution of tha t
to easing the hard currency balance of payments .
The planners' reaction function in the model might be improved throug h
expe r imenting with asymmetric responses to deteriorating and improving

economic performance . One suspects that planners are less worried abou t
redistributing unforeseen windfall gains, than they are and must be) abou t
distributing losses . No such hypotheses were explored in HEM-1, which assumes ,
symmetric responses, and such a line of inquiry might produce interestin g
results .
Finally, an expanded version of HEM-1 would allow one to explore wha t
might transpire in the early 1980's as Soviet energy supplies to Hungar y
stagnate, as the hard currency trade balance deteriorates, and therefore a s
planners must make difficult choices concerning adjustment or further in creases in hard currency debt .

Footnote s

1SOVMOD does include some endogenous planners' reactions, e .g . in th e

foreign sector, but in the main planners' behavior is left outside the model .
2Hungarian economic institutions are discussed by Balassa, Portes ,
Hare, and Hewett (1980a) .
but they wer e
3 F or many variables, data for the 1950's were
not used because : (1) planners were learning a good deal about their jo b
in those years, which shows up in the erratic nature of some of the series ;
(2) the data are .less reliable and somewhat more limited in scope than fro m
1960 on ; and (3) political events in the Soviet Union (the transition fro m
Stalin to Khrushchev) and in Hungary (the 1956 uprising) affects much of th e
data for those years .
The direct negative relationships between net exports and actual consumption, investment and government spending imply that when net export s
rise, planners take measures to cut final demands to "make room" for ne t
exports . I cannot quantify the measures they take (direct control s
on investment, various controls on wage funds, and so on), because there ar e
no planned values for net exports, it is necessary to reduce these links t o
a simple one between net exports and aggregate demand .
5
Marxian national
income (value added in material sectors) is used i
n
the planner's reaction functions because that is the state variable the y
use in making their plans . On the other hand GDP is the appropriate variable for tracking macroeconomic balance . Because Hungarian statistical publications provide data for both income measures, each are used where appropriate in the model, then linked through a stochastic equation .
6The AUTO2 scheme assume
s
li t

p l~I t-1 + p 2 u t 2

t `~ t

where e t is an independently distributed random variable, and u t is the disturbance in period t . The procedure consists of assuming values for pi and P
and then running the regression, choosing new values which will reduce th e
residual sum of squares, rerunning the regression and continuing to do s o
until the change in p 1 and p 2 is below a set criterion . For further detail s
see the National Bureau of Economic Research, Section 7 .11 .
/ Here the processes are assumed for MAV1 to b e
Et

ut

E
1. t-1.

and for MAV2 to b e

Po t

E t - p l t-1 - p

25
t- 2

The iterative procedure works as described in the previous footnote . ,

i.

The estimated value of investment which planners use is apparentl y

derived from data for the first three quarters, and is reported as p art o f
the discussion of the plan for the following year .
- There is always a plan for balance, exclusive of credits, in trad e
flows among Soviet and East European trade partners, and if there are an y
problems in fulfilling its export obligations, then further negotiation s
resolve how balance will be restored, that year or the next . The wea k
positive coefficient here would seem to indicate that most of the tim e
Hungary's deficits or surpluses with the Ruble area either were planned o r
else resolved within the same year, and therefore did not require furthe r
exports the following year .
10 There is one inventory series available for Hungary, which lump s
together finished goods and intermediate goods inventories, consequently ,
this is a difficult series to model . In any event it makes sense to sup pose that inventories, which are very large in Hungary, can act as a buffe r
whenever there is strong divergence between other final demand elements an d
aggregate supply .
11All the price indices were set t their 1973 values for 1974-78 ,
hence setting the terms of trade at their 1973 value . Because HRH-1 ha s
foreign trade price levels in it, the results are somewhat differen t
depending on how one freezes the terms of trade ; however experiments wit h
various assumptions suggested that the effects on the results were minor .
12Actually such a surplus would he unlikely, and indeed unwise, sinc e
the planners would have trouble buying anything with their accumulate d
Rubles . What would be m o p e likely is that the potential Ruble extorts wou ld
have been diverted to the domestic market, thus increasing final deman d
categories somewhat more than is indicated here .
13 0f course it is hard to say with any certainty, since the increase d
borrowing for new investments was, in part, stimulated by hard currenc y
balance of payments problems which intensified East European (includin g
Hungarian) planners ' interests in improving the supply of hard currenc y
exportables through new investments .
14ln the earlier discussion of the consumption and investment equations ,
it was stated that consumption takes less than its share, and investmen t
more than its share, of any change in net exports ; yet this simulation re sult suggests the opposite . What in fact happens it that because net ex ports play virtually no role in determining planned investment, planne d
investment does not "catch on " to the lower net exports, but actual invest ment does, and those two combine in the investment equation to slow dow n
actual investment somewhat as planned investment divided by estimate d
investment in the previous year is somewhat below the base line value .

Dz .

Reference s

Bela Balassa, The Hungarian Experience in Economic Plannlng

Ya le University Press, 1959) .

(New Haven

Jnos Gcs and Mria L cko, " A npgazdasgi szint tervezsi magtarta s
vizsgalata," ( " A Study of Pinning Behavior at the National Econom y
Level"), Kzgazdasagi Szemle, Vol . 21 (March 1974), 257-74 .
Donald W . Green, Gene D . Guill ., Lawrence R . Klein, Herbert S . Levine, Pete r
Miovic, and Ross S . Preston, The SRI-WEFA Soviet Econometric Model ,
Phase Three Documentation, Two Volumes, Mimeo, October 1976 .
Donald W . Green and Christopher I . Higgins, A M acroeconometric Model of th e
Soviet Union (New York : Crane-Russak, 1977) .
Josef Goldman and Karel Kouba, Economic Growth in Czechoslovakia
International Arts and Sciences Press, 1969) .

(New York :

Laszlo Halabuk, Katalin Hulyk, Zsigmond Nyri, and Mrs . Gyul Kotsz ,
A Magyar npgazdasg konometriai, modellje (An Econometric Model o f
the Hungarian Economy) (Budapest : Akadmiai
Kiad, l_973) .
Paul G . Hare, "Economic Reform in Hungary : Problems and Prospects, "
Cambridge Journal of Economics, 1(1977), 317-33 .
Edward A . Hewett, "The Hungarian Economy : Lessons of the 1970's and Pr
' s, " forthcoming in U .S . Joint Economic Committeeopectsfrh1980 ,
East European Assessment (Washington, D .C . : GPO 1980a) .
"The Impact of the World Economic Crisis on Intra-CMEA Trade, "
in Egon Neuberger and Laura Tyson (eds .), The Impact of Internationa l
Economic Disturbances on the Soviet Union and Eastern Europe (Pergamon ,
forthcoming 1980b) .
rvidtvu Judit Nemnyi, Pter Subicz, and Andrs Biala, A
Lszl Hunyadi,
tervezs konometriai modellje (A Short-Term Planning Econometric Model) ,
Two Volumes, mimeo, Budapest, 1979 .
Michael Kalecki, Ocherk teorii rosta sotsialisticheskoi
ekonomiki (Essays o n
the Theory of Growth of th e Socialist Economy (translated from Polish ;
Moscow
: Progress, 1970)
Kornai, " A gazsasgi viselkeds
.Janos
normai s a normi szeriniti szablyzas , "
( " The Norms of Economic Behavior and Norm-Related Regulation") ,
ki Kzgazdasgi Szemle, Vol . 23 (January 1976), 1-14 .
Anti-Equilibrium (Amsterdam: North-Holland, 1971).
National Bank of Hungary, "Information Bulletin" (concerning a $250 millio n
medium-term loan offering), September 28, 1979 .
.

National Bureau of Economic Research, Inc ., TROLL/1 User's Guide, mimeo ,

June 1972 .
Richard Fortes, "Hungary : Economic Performance, Policy and Prospects," i n
U Post-Helsinki
.S . Joint Economic Committee, East European Economie s
GPO, 1977), 766-815 .
(Washington, D .C . :
Harold T . Shapiro, "Macroeconometric Models of the Soviet Union and Eas t
European Economies : A Tabular Survey," Econometrica (November 1577) ,
No . 8, 1747-66 .
Harold T . Shapiro and Laszlo Halabuk, "Macroeconometric Model Building i n
Socialist and Non-Socialist Countries : A Comparative Study," Inter national Economic Review XVII, No . 3 (October 1976), 529-65 .
Andrs Simon, An Econometric Model of the Hungarian Economy," mimeo
September 1978 .

U .S . Central Intelligence Agency, "SOVSIM : A Model of the Soviet Economy, "

ER 79-10001 (February 1979) .