Granick D. the Ministry as Maximising Unit in Soviet Industry
Transcript of Granick D. the Ministry as Maximising Unit in Soviet Industry
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
1/19
JOURNAL OF COMPA RATIV E ECONOMICS 4, 255-273 (1980)
The Ministry as the Maximizing Unit in Soviet Industry1
DAVID GRANICK
Departme nt of Economics, University of Wisconsin, Madison, Wisconsin 53706
Received Decembe r 1978; revised July/November 1979
Granick, D.-The Ministry as the Maximizing Uni t in Soviet Industry
Microanalysis of Soviet industry has customar i ly chosen the enterpr ise as
uni t of analysis. This paper recommends that choice of the unit be made according
to the criterion of the lowe st hierarchical unit custo mari ly receiving plans that
hold unchanged for the entire planning period. It expe riments with use of the
ministry as unit of ana lysis. Conc lusions from statistica l data are (1) absen ce
of taut planning during normal years ofthe 1% 9- 1977 period, contrasting sharply
with the situation during 1949- 19.56; (2) absence of any ratchet effe ct in planning
during either period, with greater statistica l confidence in this result for the 1%9-
1977 period. J. Comp. Econ. Sep t. 1980, 4(3), pp. 255-273. Unive rsity of
Wisconsin, Madison, Wisconsin.
Journal of Econom ic Literature
Classi f icat ion Numb ers: 052, 113.
1. INTRODUCTION
In the use of maximization-under-constraints models of an economy,
a key choice for the economic
analyst
concerns the decision-making unit
that is held to do the maximizing. Conventionally, the analyst of capitalist
economies takes the firm as his unit; he does not take a subunit of the
firm (a division, a wholly owned subsidiary) because he is content to
accept the firm as a black box. Nor does he mount higher than the firm
to the common-ownership group or banking group; this is justified
partly by the nebulous nature of such groups, and partly by the degree
of managerial control expressed in individual large firms. The firm is
chosen as the relevant unit on the grounds that it is the highest-level body
that clearly makes direct supply choices.
This justification, however, wil l simply not do in making a similar choice
for Soviet industry. The links leading from the Council of Ministers
I Support for the writing of this article wa s provided through a fel lowship by the Kennan
Inst i tute for Advanced Russian Studies of the Woodrow Wilson Center and by the Graduate
School of the Universi ty of Wisconsin-Madison. I am indebted to Francis Seton, Daniel R.
Kazme r, and El izabeth Clayton for their com men ts.
255
0147.5967180/030255-19 02.00/O
Copy nght 0 198 0 by A c ademic P res s, I nc .
A l l r i gh t s o f reprod uc t i on i n any f a rm res e rv ed .
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
2/19
256 DAVID GRANICK
to the individual shop level are administrative rather than market-
determined. Although each hierarchical level could be treated as a maxi-
mizing unit, it is doubtful (see below) that such an analytic route would
yield substantive results; yet the choice of a specific level on which to
concentrate inevitably has a somewhat arbitrary character.
This article explores the implications of choosing the ministry, rather
than the enterprise, as the maximizing unit in Soviet industry. What is
significant here is that this alternative choice yields a rather different
picture of Soviet planning and industrial operation than does the con-
ventional choice.
2. THE CRITERION OF CHOICE OF UNIT
Symbols Frequently Used
performance of the unit in period (t),
expectation of performance held by superiors in higher units,
expectation of performance determined prior to period (t), with any
change of its value during (r) being independent of Pt,
G,t
constraints upon the unit that are imposed by higher units prior to
period (t), with any change of their values during (t) being inde-
pendent of P,.
Within any hierarchical system (including but not limited to Soviet
industry), the performance of any unit below the peak of the hierarchy
will be judged against the expectations of the higher units rather than
against some absolute standard. One might think of the units manage-
ment as maximizing I&, (Pt - EJ(1 + r)l, where Pt represents the unitss
performance in period (t), E, the expectation of performance held by
superiors in higher units, and r the rate of time discount used by the
management of the unit. Constraints placed upon the unit are partially
predetermined by events in earlier periods (e.g., capacity of plant and
equipment), and are partially imposed by the higher units (e.g., allocation
of materials). These constraints are labeled C1 and CZ, respectively.
Both Et and Cz,t may be determined prior to period (t), with any
changes in their values during (t) being independent of P,. This form of
expectation and constraints (& and CZ,l) represents a firm plan
established for the unit. The unit managements problem is to maximize
C& (Pt - . ?&/(I + r) subject to C,,, and C,,,. This still leaves the units
management with the gaming problem of estimating fi, fi, and f3 in the
simultaneous equations Z?t+i = fi(Pt,~I,Cl,t,Cl,t_+i,62,f,C2,tfi): d_z,t+i =
f2(Pt,~z,t,~t,~l+i,Cl,t,Cl,l+i) and C
I,t+i = f3(Pt,Cl,t,Et,~*.t,E(El+i)rEo)t
where E = the value expected during the previous period by the planners.2
* The functions fi J; are written, solely for simpli fication , witho ut inclu ding as
arguments the performance, plans and constraints of units other than the one under
examination.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
3/19
THE MINISTRY AS UNIT OF ANALYS IS 257
But this gaming problem (labeled in the literature of the Soviet firm as
describing the response to the ratchet effect in planning) is not more
serious than the gaming problem imposed upon the firm operating under
monopolistic competition in capitalist economies: the maximizing analytic
approach is no more and no less applicable.
What can be said, however, when Et is determined at the end rather than
at the beginning of period (t), and when C,,l consists of constraints that
change during period (t)? This form of expectation and of constraints (Et
and C,,,) represents a changing plan for the unit, since it can be
cksribed as .h = f4(Pt,~:I,Cl,t,~.2.t,E2,t) and e2,, = fS(Pt,Cl,t,~2,t),
where & and C2,1 epresent the plan and the constraints provisionally
imposed at the beginning of the year. The definition of changing plan is
that a(&, -C,,JIM, # 0, where the vector (E,, -C& represents Et as an
implicit function of Cz,t. If the plan for the unit is changing in the above
sense, particularly if a(E,, -C,,,)/I%, varies between periods and is
frequently a substantial fraction, then one may argue that the units
management is concerned almost exclusively with the gaming problem,
thatP, enters only very weakly as an argument in its objective function, and
that such a unit thus represents a poor choice for the analyst as the basic
maximizing unit.
In the above case, a simultaneous solution based upon maximizing
behavior on the part both of the plan-making body (e.g., the ministry)
and of the unit receiving changing plans (e.g., the enterprise) also
appears to be a poor analytic strategy. Since the formulation of the maxi-
mizing strategy of the plan-receiving unit is primarily a problem in
gaming, the results of a simultaneous solution will depend heavily upon
the assumed gaming strategy of the plan-making body. The introduction
of alternative gaming assumptions with regard to the plan-receiving unit
simply muddies the analytic waters, vitiating the possibility of testing
hypotheses. In any case, if one wishes to adopt the simultaneous-solution
strategy, it should be recognized that the maximizing behavior by plan-
receiving units must differ fundamentally from that posited in treat-
ments of the Soviet enterprise which assume-explicitly or implicitly-
unchanging plans (for recent examples of such treatments see Berliner,
1976, and Granick, 1978).
The suggested criteria for the choice of a unit of microanalysis within
Soviet industry is the following:
constraint: the unit should be subject to firm rather than to
changing plans;
element to minimize:
the level of the unit in the planning-administra-
tive hierarchy.
The minimization criterion here arises from the desire to deal with units
that are as microeconomic as is feasible.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
4/19
258 DAVID GRANICK
How should we reconcile the fact that microeconomic theory applied
to capitalist economies takes as its unit of analysis the highest hierarchical
level making supply decisions within an organization (i.e., the firms top
management rather than the management of any of the firms subdivisions),
while my criterion for Soviet industry is to choose the lowest such unit?
The choice of neoclassical microeconomics is dictated by its concern
with market relationships and lack of interest in administrative relations;
its choice permits administrative relations within a firms hierarchy to be
capsulated in a black box. Put in our terms, this choice can be restated
as resting on the implicit assumption that-to the degree that planning
for the subunits internal to the firm exists at all-such planning should
be characterized as changing rather than firm. In this sense, the
enterprise is the lowest unit in capitalist economies that meets our con-
straint, and thus the choice rule in traditional microeconomic theory is
the sole choice permitted by our proposed criterion.
3. DISCUSSION OF THE RELEVANT UNIT
Traditionally, microanalysis of Soviet industry takes the enterprise (or
now, possibly, the association) as the unit of analysis; this applies to
analysis conducted by both Soviet writers and foreigners. Thus, relations
within the enterprise are treated as a black box, whileglavki, ministries
and other bodies up to and including the Council of Ministers are lumped
together as the Center. This choice of microunit can be justified on the
grounds that subunits of the khozraschet unit are controlled by methods
other than firm plans, but that the khozraschet units themselves are
objects of firm planning. My discussion will ignore the question of
subunits within the enterprise, and concentrate instead on the issue of
whether the hierarchical level of the unit of analysis should be raised.
Although my treatment is in terms of the enterprise vs the ministry,
the latter should really be categorized as gfavk-or-ministry; the discussion
is couched in terms of the ministry solely because the statistical data
available are ministerial.
It is perfectly clear that Soviet annual enterprise plans are neither
perfectly firm nor changing. A study of Romanian industry around
1970 indicated that their plans were
changing to such a degree that
the unit of microeconomic analysis should be taken as the ministry
(Granick, 1976, Chap. 4). The degree of plan change in Soviet industry,
particularly in the sense described above as relevant to our problem,
is unknown. Yet it is the quantitative dimension of such change that
should be decisive in our choice of unit.
The logic of making the assumption of firm plans is that the basic
structure of the Soviet incentive system of enterprises, both currently
and in the past, is predicated on it. Bonus rewards based on the difference
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
5/19
THE MINISTRY AS UNIT OF ANALYS IS
259
between enterprise performance during the year and planned performance
(as of the end of the year) make little sense if such plans for year (t)
are themselves a function of enterprise performance in year (t).3 If we
assume that the incentive system is in essence structured rationally as
judged from the standpoint of the Center,4 then we must also assume
that plans are essentially firm.
The author finds that the above argument presents a convincing case.
Nevertheless, there is nothing binding us to the assumption of rational
incentives. There are two pieces of evidence pointing us away from the
enterprise as the basic unit for analysis.
The first of these is a study of plan fulfillment within two groups, each
comprising about 100 Soviet industrial enterprises (SHaikin, 1974).
Comparison is made between the amount of actual sales of each enterprise
during 1970 and both the enterprises original sales plan for the year (which
might, however, include changes up to February 3, 1970) and its final sales
plan. The results are shown in Table 1. However, there is nothing to
indicate the representativeness of SHaikins sample; he himself makes no
assertions as to representativeness, stating only that he is dealing with a
question that has not been studied by others in the Soviet Union.
In presenting these data, SHaikin points out that, using a x2 test, we
cannot reject the hypothesis that deviations of plan fulfillment by individual
enterprises around the average plan fulfillment of their group are
distributed normally-when the plan is defined as the original plan.
But the hypothesis can be rejected for the final plan. Normality in
both cases is what SHaikin would expect if plans were firm; the present
author would expect in this case the same degree of normality (whatever
this might be) with respect to both types of plans, but with a smaller
standard deviation in the case of the final plan because of plan alterations
that take account of environmental changes from the planners original
expectations.
In any event, SHaikins sample displays changing plans at the
enterprise level; this is sharply the case for the nonfulfilling enterprises of
both groups, but for the high-overfulfilling enterprises it holds only for
group 2. Some slight slack presumably existed between the sum of the
original plans of the enterprises within each group and the total plan of the
Ofcourse, there is noth ing in this argumen t to suggest that plans are not altered at all, Pla n
alterations that result from changes i n the environme nt surrounding the enterprise, rather than
from the enterprises own performance, should, if anything, affect the incentive system
positively.
See Granick (forthcoming, Sect. 1) for an argumen t as to why neoclassical economists are
obli ged to make this assumption if they are to remain with in the neoclassical rules-of-the-
game . Note that the Rom ania n incentive system in 1970 was constructed alon g lines radically
different from the Sovie t one (Granick, 1976, Chap. 4).
5 The source does not state how these groups were formed. One wou ld presume from the
logic of the authors analysis that each represented a different administrativ e unit.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
6/19
260 DAVID GRANICK
TABLE 1
FULFILLME NT OF THE ANNUAL SALES PLAN FOR 1970
BYTWO SAMPLE GROUPS OF ENTERPRISES
Group 1
Original Final plan
plan (%)
m
Group 2
Original Final plan
plan (%)
(%I
Plan fulfi l lment
Average percentage s for the group
(standard deviation)
Enterprises not fulfil l ing plan
Enterprises fulfil l ing 106% or
more of plan
(Enterprises in category)/(enterprises
predicted to be in catego ry under
assum ption that the distr ibution
is normal)
Enterprises not fulfil l ing plan
Enterprises fulfil ling 106% or
more of plan
102.2 102.8 102.5 103.5
(2.68)
(2.29) (7.45) (3.69)
17.7 2.4 33.1 5.6
7.1
9.4 26.6 14.5
86
16 90 33
91 116 83 58
body administering the group, since the final total plan was between 99.0
and 99.4% of the original total plan of each group.
The second piece of evidence pointing away from the enterprise as the
basic unit for analysis arises from the introduction, in the 1971- 1975 Plan,
of the system of 5-year planning of annual control figures for ministries and,
at least supposedly, for enterprises. Incentive systems for the enterprises
were supposed to be linked to their own control figures. In fact, however,
as annual plans for industrial ministries fell sharply below the ministerial
control figures, it was the ministries-not individual enterprises-that
were penalized in 1973 and again in 1975 with regard to their bonus funds
(Rzheshevskii, 1975, pp. 82, 85). It was the ministries that were here
regarded by the Council of Ministers as the basic units that had to allocate
these reductions (33% in 1975 for the Ministry of Forestry and
Woodworking, and 15 to 18% for the Ministries of Chemical Production
and Nonferrous Metals) among their enterprises. This was similar to what
had occurred earlier in the late 1960s.
Even if, on balance (as the author believes), enterprise plans should be
considered as sufficiently firm so that the enterprise should be taken as
the unit of microanalysis, clearly the level of firmness is a matter of
degree. The present article can be interpreted as examining the sensitivity
of various standard interpretations of Soviet industrial reality to the choice
of unit of microanalysis. It may be viewed as comparable to Solows
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
7/19
THE MINISTR Y AS UN IT OF ANALYSIS 261
explorations of models of both disembodied and capital-embodied
technological change, with Solows personal preference between the
models being a matter of no great significance to his readers.
4. RESULTS OF USING THE MINISTRY AS
THE UNIT OF ANALYSIS
Tautness of Annual Plans
Let us define plans as being taut when they are difficult to meet even if the
environment surrounding the planned unit is in fact the same as that
originally predicted by the planners. The term is devoid of meaning if plans
are changing. At the extreme, taut plans act as a set of constraints that
allow no room for maximizing behavior by the units. Under taut planning
we would expect that, even if the mean level of plan fulfillment should be
above lOO%, it will not be very much above that level. Second, and more
significant for our testing procedure, we would expect that the distribution
be reasonably normal in the area between such a mean and the point of
100% plan fulfillment-although the expected value of the integral of the
area between the mean and (100% - E) would be somewhat higher than the
one predicted under conditions of normality.
Data are available by industrial ministry for fulfillment of annual
f i There are three sources of deviation of performan ces from plan. (1) Planners ma y set their
targets with the expec tation that execu tant units wil l overfulfi ll them by an average of.r%. This
would presum ably be done for incentive reason s, and can explain the fac t that mean plan
fulfi llment is generally greater than 100% . (2) Planners m ay ma ke incorrect assu mption s as to
the environme nt within which the exec utant units wil l be operating. Wh ile i t is true that a
planning error w ith regard to one unit affects the plan-fulfil lment indices o f other un its,
neverthe less it does not seem unreasonable to consider that w ithin a given year the net
planning errors for different u nits wil l be random ly distr ibuted around a value
equal
to the
observed mean of uni t per formance plus some K that is constant over years close to one
another. (3) Performan ce by exec utant units may show a pattern of distr ibution around (-K)
plus the level that could be reached given their actua l environm ents. Due to consideration of
the ratchet effect (see the text below), we might expe ct the manage ments of such uni ts to avoid
exces sive overfulfi l lment; thus the performance deviations should not be normally
distr ibuted in the area at and above the mean of unit plan fulfil lment. But how about the
norma lity of performance deviations on the lower side of the mean?
It has been argued in the l i terature that a mana geme nt that know s that i t canno t fulfil l i ts plan
wil l permit sharp underfulfil lment in the hope that this behavior wil l lead to a lower plan in the
fol lowing period. Assu ming this to be true, i t sti l l does not affe ct the norma lity of the
distr ibution in the area of 100% or more plan fulfi l lme nt which is below the mean: i ts only
effe ct is upon the distr ibution of plan under-fulfil lment, and this distr ibution is integrated in our
stat ist ical treatment below.
There is, however, one factor that would cause the expec ted distr ibution of plan fulfi llment
in the relevant area(where 100% plan fulfi llment 2 realized fulfi llment by a given unit < mean
fulfi llment by al l units) to be above that predicted through use of the normal curve. This is the
fac t that the personnel of a unit that, given its environm ent, would narrowly mis s 100% plan
ful fi llment wi th custom ary work e ffor t and degree of concentration on the measured goat,
ma y exert su ppleme ntary and unusual effort or concen tration so as to me et i ts plan (Bonin and
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
8/19
262 DAVID GRANICK
production plans during 1949-1956 and for annual sales plans during
1969-1977.7 The latter period covers 21 to 23 industrial ministries each
year; during this time, there were two years (1969 and 1972) in which 29%
of the ministries missed their targets
.R But in the remaining seven years,
there was only one ministerial underfulfillment out of 157 ministerial
observations. If we take these seven years as a sample, calculate a mean for
each year and a standard error for that mean, and assume that the area
between (100% - E) plan fulfillment and the mean plan fulfillment (> 100%
in all years) is normally distributed, then we would predict eleven
ministerial underfulfillments instead of the observed one. As indicated
above, the predicted underfulfillments under tautness should be less than
11; but I would not expect that their deviation from the integral of the
relevant portion of the normal curve would be anything approaching the
observed 11: 1 ratio. (Observed cases of plan fulfillment were: [< lOO%] - 1;
lOO%- 1; lOO.l- 100.3%-9; 100.4- 100.5%-4. Under the normality as-
sumption, predicted cases ~100.3% of plan fulfillment-22, and
5 100.5%-30.)
This record differs considerably from that of 1949- 1956, during which at
least one ministry underfulfilled its plan during all but one year (1949).
Excluding for comparison with 1969-1977 the two years in which the
proportion of ministries under-fulfilling was the greatest (1952 and 1953),
underfulfillments constituted 56% of the number predicted under the above
normality assumption as opposed to 9% in the later period. Furthermore,
the unweighted arithmetic mean of the standard error of the mean of
fulfillment declined sharply as between the normal years of the two
periods: from 2.1% of plan to 0.9%.
Marcus, 1979). Moreover, such supplementary effort migh t wel l slightly overshoot the mark,
and thus the plan fulfillment observed could be a minor amount above 100%.
Our concern is with the join t distribution of the plann ing errors and the performance
deviations in the relevant area. Assuming that these deviations are independent, the net
effect is that the expected value of the integr al ofthe distributio n in the relevant area should be
greater than would be predicted on the basis of the normality assumption, but only to the
degree that the effort/concentration factor is at work. T he less the importance of this factor
in the distributio n pattern-and, particularly, thegreater the relative importance ofpla nning
errors to performance deviations -the more the expected fina l distributi on in the area of
interest wil l approach normality.
i Data exist also for 1978. but these are not comparab le to those for earlier years. In 1978, for
the first time , counterplans were includ ed as part of the annu al sales plans (Rzheshevskii,
1978. pp. 7-9).
Ministries processing foodstuffs are excluded on the ground that their results are
exceptionally subject to acts of God.
o During 1949-195 6, 17 to 23 ministries were covered during seven years and 10 in the
rema ining year (1953). The average number o f ministries covered annua lly during the
normal years was 20.2 during 1949- 1956 compared with 22.4 during 1969-197 7. This
difference accounts for some part, but presumably not much, of the difference betwe en the
two periods standard errors of the mean .
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
9/19
THE MINISTRY AS UNIT OF ANALYS IS
263
In summary, there has been very loose planning as to sales volume
during seven of the past nine years, a situation differing substantially from
that of the previous period for which we have data. The central planners
twice retreated during the 1969- 1977 period after each interspersed year of
taut planning, when under-fulfillments reached 75% of the number predicted
under the normality assumption stated above. The statement as to the
degree of tautness is predicated on the assumption that plans set for the
ministries were firm, but without- this assumption no meaning can be
given to the concept of tautness as discussed here.
However, there is a fundamental difference between planning during the
two periods, which modifies the implication of the above summary. For
1949- 1956, one can realistically describe output as being the single plan
indicator that really mattered.
But this was not the case during
1969- 1977-in particular, the rate of profitability was of major importance
throughout, at least at the enterprise level. We have no data as to fulfillment
of annual ministerial plans for profitability; it is conceivable that the
combination of sales and profitability plans constitutes planning at the
ministerial level no looser than that of the earlier period. In any case, it is
of no litt le significance that there has been a radical change in the degree of
looseness of plans along the traditional Soviet output/sales dimension.
The Existence of a Ratchet Effect in Planning
The ratchet effect hypothesis holds that planners learn from
experience by increasing the level of their plans for a given unit as a
function of that units performance relative to plan in the preceding period.
It can be stated in two variants. The first states that the level of a units plan
is a positive function of the units performance in the immediately prior
period. The second states that the above is the case only when the plan for
the previous period was overfulfilled; in the case of under-fulfillment or
exact fulfillment, no functional relationship exists.
Variant (1): aEtiapt-l > 0.
Variant (2):
a&/apt-, > 0
if
Pt-,/El--l >
1,
a&tap,-, = 0 if P,+,/E~:t_l d 1.
Variant (1) and the first part of variant (2) yield the following result (see
the Appendix): lo
a(P,/~,)ia(P,_,/~,_,) < 0
iff aP,/aE, < P,E,.
I The reason for empl oying the transformation below , rather th an running a regression in
terms of the hypothesis itself, is that the transformation reduces the errors of measurem ent of
the indep enden t variables. In the transformation, all variables except CR,, which is used in
only some of the equation s of Table 2, are taken directly from reported data. If, instead, the
terms of the hypothesis had been place d in the regression, then the values of the indep ende nt
variables P,-, and P,-, woul d have had to be calculated-with resultant substantial roundin g
error.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
10/19
264
DAVID GRANICK
The second part of variant (2) yields
(aP,/~,>la(P,-,l~,_,) = 0.
Thus both variants, if we linearize them and assume that for at least one
period and one unit in the economy Pl-,/&, > 1, yield the OLS equation
PJE , = (Y(P~-~E-~) + /IX + et,
where the matrix X represents the constant and the other independent
variables, and where CY< 0 [since a(P,/~,) la(P,_,/~,_,) < 01.
If we are willing to define i? and P in terms of output and sales, this OLS
matrix equation is transformed into one using the variables of Table 2, with
AFt
defined as
P&:
AFt = aAFtml + PX + e,,
where (Y < 0.
We have here a testable form of the ratchet-effect hypothesis, so long as
we combine it with the assumption that
aPtIt& < Pt/&.
This auxiliary assumption says that the marginal effect of plan change on
performance (when the level of the plan increases between periods (t - 1)
and (t)) is less than the average relationship. Clearly, it always holds if we
further assume that
~P,/c~I?,
= 0; i.e., that the plan is used simply as a
prognosis and/or as a measure (for incentive purposes) against which to
evaluate performance. If, however, we make the more realistic assumption
that aPJa& # 0 for most likely values of &, and further assume that the
marginal effect of plan, on performancet is both continuous and subject to
diminishing returns (see, e.g., Hunter, 1961), then the auxiliary assumption
holds provided that ,?:t s notfar below that level that will maximize Pt (see
Fig. l). Thus the auxiliary assumption is fairly weak so long as we assume
both that Soviet planners are concerned with maximizing
Pt
and that over the
years they have gained even the roughest notion of the effect of 8, on Pt.
There are, however, two additional nonstatistical arguments that might
be advanced for rejecting the auxiliary assumption, and thus the sign of a in
the OLS equations above, as a proper test of the ratchet-effect hypothesis.
The first of these is that positive externalities exist; i.e., where (k) and
(k + h) are industrial ministries, that (P& is a positive function not only of
(PJkfh
but also of the degree of assurance that (k) will receive its allocated
supplies of the products of (k + h). Thus for some ministries (k + h), the
Center may wish to set & at a lower level than that which would maximize
expected Pt; at the extreme, the Center may wish to maximize that portion
of [(expected
Pt)/l?l]k+h
which lies in the second and third sections of Fig.
l-depending upon the maximum value of this fraction that lies in these
sections.
I The draw ing of the curves in Fig. 1 begins only slightly to the left of the boundary betwee n
sections 1 and 2, rather than at the P, axis, because the ratchet-effect hypothesis does not
define M,l&, for small values of the ratio .??,/P,.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
11/19
THE MINISTRY AS UNIT OF ANALYS IS
265
This position rests on the foundation that material allocations (cZJk are
determined simultaneously with (I? )k+h and that their magnitude is
determined on the assumption that (Pt)k+h will equal (I?.f)k+h. The latter
however, is an unnecessary restriction placed on the Center. There is no
reason why (Pt/I?t)k+h
cannot be set by the Center with a view to
maximization of (Pt)k+h
subject to the constraints placed upon (k + h),
while (c:2,f)k is determined as a function of the minimum value of (Pt)k+h
which the Center expects to be realized rather than in terms of (I?,),,,.
Such a planning procedure need not undercut the moral significance of (I?,)
to (k + h) if ministry (k + h) is instructed to sign normal contracts for a
volume of its products equal to the minimum (PJkfh expected by the
Center, and to retain the remaining portion of its planned production for
shipments during the year to undesignated customers.
The second ground for objection to the proposed test is more substantial:
the existence of trade-offs between output/sales performance by ministries
and performance along other planned dimensions (such as rate of
profitability) or unplanned ones (e.g., product mix at a disaggregated level).
Because the Center believes that such trade-offs exist, it will not wish to
maximize Pt as expressed in output/sales. Thus it may set the ministerial
plans to the left of where aP,/& = 0 in Fig. 1.
The author cannot evaluate this objection. Certainly it must be valid as
stated above, but it has force as related to our problem only if ministerial
plans have typically been set at or to the left of the point that maxi-
mizes P,lEt.
We turn now to the OLS regression equations of Table 2. Two periods
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
12/19
266
DAVID GRANICK
TABLE 2
INDEPENDENT VARIAB LES IN OLS REGRESSION EQUATIONS
Eq.
NO.
AFf.1
CR,
Dummy
variables
YRS MIN C D.F. ds
Test for serial correlation
1950S I 0.667***
2 0.2391
3 0.531** X
4 X X
5 X
6 X
1970s 7
8
0.530***
0.260
X X 87 0.47
X X X 69 0.55
X X 56 0.28
X X X 38 0.35
X X 57 0.12
x x x 39 0.36
X X 160 0.40
x xx 139 0.52
Reject at 5% contidence level
Test fails
Test fails
Test fails
Test fails
Test fads
Accept at 10% confidence level
Reject at 5% confidence level
9 0.581***
X X X
138 0.48 Accept at 20% confidence level
IO 0.32 6** X x xx
117
0.52 Test fails
II 0.420*** X X 139
0.27 Reject
12
X x xx II8 0.47 Reject
13 0.445*** 0.093*** x X 159 0.48 Reject at 5% confidence level
14 0.226** 0.214*** X X X 138 0.61 Accept at 20% confidence level
15 0.506 X 0.076 X X 137 0.53 Accept at 10% confidence level
16 0.282** X 0.203 X X X 116 0.56 Accept at 20% confidence level
1950s and 17 0.234
x xx 208 0.57 Test fail s
1970S 18 0.234* X x xx 156
0.47 Test fails
AF, is the depen dent variable.
Symbols. X-The indep ende nt varia ble indicate d at the head of the column is used as an argument in the regression; for
variables other tha n dummies and the constant, the r value o f the coefficient is less than 1.65 (two-tailed f). t-l.65
c f < l.% [IO% confidence level]. *-I.% < r c 2.575 (5% confidence level]. -2.575 < f < 5 [I% confidence level].
***-I > 5. D.F. -Degree> of freedom. AF,+-Percentage overfulfi llment of the annu al output (saks)plan in year (r - i).
CR,-Percentage growth rate of absolute gross sales in year (1). C-Constant of regression. YRS-Dummy
variable for each year include d in the regression equat ion. MN-Dummy variable for each ministry include d in the
regression equa tion. A dummy variable used for any given ministry in the 1950s is different from that used for the 1970s.
Comments. The test used for serial correlation is the Durbin large-sample test (Johnston, 1972, pp. 312-313). Reject
(unless otherwise specified) means the rejection of the hypothesis of zero autocorrelation of error terms at the 1% con-
fidence level. Accept means that, at the specified confidence level, the hypothesis of such zero autocorrelation cannot
be rejected. Both p lans and performance refer to gross o utput durin g 1949-19 56 and to gross sales during 1%9-1977.
are studied: 1949- 1956 and 1969- 1977. The source consists of the annual
Soviet reports of plan fulfillment by ministry; these directly provide the
figures for AF during the first period, and for AF and GR (growth rate of
sales) during the second period (TsSU). Sample sizes differ among
alternative specifications of the equations because of the need to omit one
or niore years for the dependent variable when a two-period-lagged
independent variable is introduced; for the same reason, the number of
dummy variables (described below) also varies even though this is not
indicated in the table by the dummy-variable matrices.
The 1949- 1956 period was subject to considerable ministerial reorgani-
zation; thus most ministries show gaps for at least one year, and several
that were abolished and then recreated had a slightly different coverage in
varying years. Although nineteen ministries were covered over 8 years, the
largest samples here (having a l-year lag embodied in the regression)
consist of only 95 observations.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
13/19
THE MINISTRY AS UNIT OF ANALYS IS 267
The 1969- 1976 period is much better. Data are available for each year for
twenty ministries; for 6 years for another ministry and for 5 years for
the last.
Turning now to Table 2, the basic regression equation takes the form of
AF, = C, + alAFt-, + &YRS + e,,t,
where the matrix of YRS is a set of dummy variables for the years treated.
The use of these dummy variables seems essential for testing the ratchet-
effect hypothesis since, as seen in the preceding subsection, each period
,contained highly atypical years with regard to Pt/l?I.
The result here is very strong. Each period (see Eqs. (1) and (7) of Table
2) as well as the two combined, shows a positive sign for (Ye, with a
confidence level substantially greater than 99%. This leads to rejection of
the ratchet-effect hypothesis.
Estimation of the remaining equations of Table 2 can be properly
regarded as a fishing expedition. Such fishing is normally considered bad
practice from a methodological standpoint, and that would be the case here
as well if the different equations yielded varying results. What is striking,
however, is that the same hypothesis rejection follows from all the equation
specifications and for both of the periods studied. To continue with the
fishing simile: our first cast suggested that the pool is empty of the
hypothesis; further casts were then made in different parts of the pool, and
yielded the same results; thus the additional casts reinforce our original
view as to emptiness.
Justification for introducing the additional independent variables is as
follows: AFI-* is a two-period lag of plan fulfillment; one might argue that,
since 8, is determined at a time when
AF,-,
can only be estimated by the
Center, this body also (or only) uses the latest plan-fulfillment figure that
can be reliably established. In fact, however, Table 2 shows that AF,-, has
no effect except for what it may pick up as a proxy variable forAF,-, (as in
Eq. (11)).12
GRt
is the growth rate of sales; one might suspect that
high-growth ministries-since mastery of new plant and equipment is
particularly important in these cases-might be subject to only a very
attenuated form, if any, of the ratchet effect. Indeed, GR, does have a fairly
strong influence in the most properly specified equations (Eqs. (13) and
i* An F test has not been used here because of the different size n in equations including and
excludingAF ,-,. It could b e appl ied, of course, for samples restricted to the ones used in the
equation s including AF,-,, but the resulting coefficients for equation s excludingAF,-, wou ld
then be less reliable than those reported. Comparison of fi2 in paired equations strongly
suggests the results indica ted as to the absence of effect ofAF,-,. Add itio n ofAF,-, increases
A* only i n the two cases of Eqs. (9) vs (7), and (15) vs (13)-both cases in which the
dummy-variable matrix MIN is omitted.
If one of the lagged-A F variables serves as a proxy for the other, it woul d seem thatAF,+, is
a proxy for AF,+,. This is strongly suggested by comparison of the I? for all of the paired
equations containing either AF,-, or AF,+,, but not both.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
14/19
268
DAVID GRANICK
(14)), and this is strongest in the one (Eq. (14)) that has both the highest 2,
and does best in a test for serial correlation.13 But the introduction of GRt
has no effect on the sign of the coefficient ofAF,-,. Dummy-variable matrix
MN (the individual ministries in the sample) is introduced on the ground
that the Center may treat different units in a systematically different
fashion, thus reflecting the Centers views as to their differential effort and
efficiency; if such views changed only slowly, then these dummy variables
would represent an important addition to the regression equations.
Turning now to statistical matters, we cannot use the F test for
homogeneity between the two periods with regard to the coefficient of
AFteI because of the existence of different dummy variables in the two
periods, since a dummy is introduced for (n-l) individual years in each
period. The Durbin- Watson statistic is inappropriate due to the use of
lagged values of the dependent variable as independent variables, but the
Durbin large-sample test for serial correlation of the error terms can be
used. Unfortunately, the test breaks down for the 1950s (Eq. (2) here
seems to be the one best specified); but for the 1970s it leads to the absence
of rejection of the hypothesis of zero autocorrelation at the 10% level for
the basic equation, Eq. (7), and at the 20% level for the best-specified
equation, Eq. (14).
The apparent absence of autocorrelation of error terms in Eqs. (7) and
(14) is particularly gratifying; it implies that OLS here yields consistent and
asymptotically efficient estimators (Johnston, 1972, p. 305), since the bias
existing in small samples is irrelevant here. On the other hand, our
confidence in the results of the hypothesis test for the first period is
considerably weaker (Johnston, pp. 308-309), a conclusion that we had
previously reached on other grounds.
While measurement error of the dependent variable introduces no bias,
measurement error of the independent variables depends on the
variance-covariance matrix of these measurement errors. (In the case of
AFt-, and AFt-2, these measurement errors are not only due to inadequate
knowledge by the Central Statistical Administration and to differing degrees of
rounding that have been employed in reporting plan fulfillment, but are also
due to changes made during the year in the ministries plans. Such possible
changes should be considered as measurement errors since the hypothesis
is stated in terms of fixed plans. GR, suffers from the first two types of
measurement error plus error due to rounding introduced by my
calculations from the reported data.) For at least the two basic equations,
Eqs. (1) and (7), the measured coefficient of
AFt-,
is biased toward zero.
This makes the observed positive sign of this coefficient in the two
equations even more significant than it might otherwise be; on the other
hand, it implies that if the true coefficient of AFteI were significantly
13 Data for GR, (whether in the form of sales or output growth) are not avail able for the
1950s period.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
15/19
THE MINISTRY AS UNIT OF ANALYS IS 269
negative, its measured value might be indistinguishable from zero because
of measurement error.
The regression models tested are all single-equation models. A further
step would be to build in simultaneity by making
(AF&
a function of
(AFdk+h,
where (k) and (k + h) are different ministries. Such simultaneity
might be expected as a result of input-output relationships. The problem is
that we do not know how a shortfall or excess of actual over planned sales
by (k + h) is allocated by the Center or otherwise to individual user
ministries. This extension of my single-equation estimates must be left to
others.
How can we explain our rejection of the ratchet-effect principle? Unless
it is due to the absence of firm plans at the ministry level, to the auxiliary
assumption of our test being incorrect, to statistical problems, or to
specification errors, we may conclude that in forming (Z?,),,, the Center
either assumes that its earlier (Z?t-l)li+h was a justified plan in terms of the
capacity of ministry (k + h) in year (t - 1) if (k + h) were working at
normal efficiency, or that the Center makes a de novo evaluation each year
of the current potential of ministry (k
+ h). The fact that the introduction of
MZN into the regression equations has no effect on the sign of (Y einforces
this either/or conclusion.
However, not only does (Y not have the negative sign called for by the
hypothesis, but the sign is consistently and highly significantly positive. A
story explaining this is that the Center interprets (AFf-l)k+h as a true
indicator of the efficiency of ministry (k + h ), and that it follows the policy
of capturing (positively or negatively) a portion of such efficiency and
leaving the remainder to the ministry as a quasi rent. A glance at the
coefficients of
AFT-,
in Table 2 suggests that the Center captures some
one-half to three-quarters of such efficiency advantages. An alternative
story within this framework is that the Center does not capture any of this
quasi rent, and that (Ydiffers from + 1 simply because of regression toward
the mean in period (I) either of the efficiency of ministry (k +
h)
or of
special circumstances and chance factors which in (t - 1) caused the
ministry to appear to be particularly efficient or inefficient.
5. SUMMARY AND CONCLUSIONS
This article has been predicated on the assumption that the unit of
microanalysis within a centrally planned economy should be chosen
according to the criterion of being the lowest-hierarchical-level body
whose annual plans are firm; i.e., for whom a(E,, -C,,JaP, = 0.
Starting from the fact that the plans of Soviet enterprises are not perfectly
firm, we have explored the implications of choosing the industrial
ministry as the appropriate unit.
On the basis of the choice of the ministry as the lowest unit making
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
16/19
210 DAVID GRANICK
decisions by maximizing under constraints, two conclusions were drawn
through testing the null hypothesis. Both conclusions are based on the
fundamental assumption that annual plans at the ministerial level
are firm.
(1) Unit annual plans for sales have been loose rather than taut
during the normal years of the period 1969- 1977. This contrasts sharply
with the situation during 1949- 1956 for a close proxy of sales (production).
(2) No ratchet effect existed in annual plans of units during either of the
above periods. Both the data and statistical tests are considerably stronger
for the more recent period.
The author believes that the well-informed reader will find these two
conclusions surprising. But what is significant in terms of the theme of this
article is a comparison between them and their counterparts drawn from
microanalysis of the Soviet economy which takes the enterprise as the unit
of analysis. Unfortunately, the data underlying the latter type of analysis
are nonstatistical in nature.
With regard to conclusion (l), the conventional Western wisdom
concerning Soviet enterprises is that their plans are taut; this statement
is made in terms of production (sales) plans, and so is directly comparable
to our data. However, such conventional wisdom arises from analysis of
pre-1969 Soviet literature and here it is confirmed by our 1949-1956
ministerial data. Furthermore, there seems to be an impression that such
tautness has been reduced in recent years. Thus the treatment of the
ministry as the unit of microanalysis substantially amplifies, but does not
really contradict, the conventional Western wisdom based on the
enterprise.
Conclusion (2), however, sharply contradicts both Soviet and American
conventional wisdom applied specifically to the late as well as to the early
period. A major goal of the Soviet reforms of the 1960s was to weaken the
force of the much-deplored ratchet effect on planning, and this has been the
justification for the Governments effort to bring 5-year plans down to the
enterprise level along with dividing them into annual periods. If we assume
that Soviet writers and administrators are correct in their description of
planning as it reaches the enterprise, then we find that our understanding of
the working of the Soviet economy is significantly different depending upon
the unit of microanalysis that we select.
Three additional conclusions can be elaborated which do not come from
the body of this article, but which also depend heavily on the unit of
micro-analysis chosen.
(3) What happened during the reform period of 1968- 1973 to the degree
of detail in which goods were allocated to the microeconomic units? If the
enterprise is taken as the unit, then the number of separately allocated
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
17/19
THE MINISTRY AS UNIT OF ANALYS IS 271
products increased by 197%; if the ministry is taken as the unit, it declined
by 42% (Ivanov, 1969, p. 340; Kurotchenko, 1975. p. 82).- Depending upon
our definition of microeconomic unit, we obtain a radically different picture
as to whether the reform has been accompanied by decentralization in the
critical realm of supply, and thus as to whether the unit is more or less
constrained. Furthermore one would suspect, although we do not know,
that the production plans of the units producing these products are
disaggregated (and thus constrained) to much the same degree of detail as
exists in the distribution plans of the products.
(4) What is the significance of precontract negotiations and arbitration
among Soviet enterprises? Such negotiations occur annually on the basis of
the producing enterprises output plan and the purchasing enterprises
materials-allocation plan: they relate to such matters as packaging,
delivery dates, and, of particular importance, to details of product mix.
Precontract arbitration can also occur (Kurotchenko, 1975, pp. 164- 165).
While it is certainly true that, regardless of whether the relevant
maximizing unit is the enterprise or the ministry, it is the producing unit
that has the upper hand both in negotiations and arbitration, our choice of
unit would seem to make a significant difference. If it is the enterprise, then
it is usually far from being in a long-run monopoly position. Its future
negotiating power will be greatest when its reputation with purchasers,
relative to the reputation of possible alternative suppliers, is high. Thus it is
market forces combined with a concern for the future which act as a
constraint on the unit in its negotiations. Here is a market explanation for
why neither the product mix described in contracts nor the actually
delivered detailed product mix varies so severely from the needs of the
purchasing units that the physical planning system breaks down (Granick,
forthcoming, Section III).
However, if it is the ministry that is the basic maximizing unit, then
virtual-monopoly conditions with only limited entry frequently prevail and
the market explanation collapses. Such choice of unit would force us to
turn to quite different types of explanations.
(5) Is it substantially easier, compared to the standard of practice in
developed capitalist economies,
for R&D innovations of products
(especially for minor products) and of major process changes to be
I The 42% reduction is calculated by excluding those goods whose distribut ion was
plan ned by ministries and departments (the latter bein g organs at the same level as ministries
in other sectors). This is done on the basis that the ministries distributin g these products are
their prime users (Kurotchenko, 1975, p. 83: Berline r, 1976, p. 68). This meth od ofcalcula tion
exaggerates the difference in results betwee n the two defini tions of unit, since some or most of
the ministe rially distributed goods are also used in enterprises of other m inistries. But it seems
to capture the fundamental issue.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
18/19
272 DAVID GRAN ICK
successfully implemented in Soviet enterprises that are newly formed than
in established enterprises? This author has argued this case positively
elsewhere (Granick, 1978) on the basis of an analysis of the incentive
system facing existing enterprises. But this argument was predicated on the
assumption that the enterprise is the maximizing unit.
If it is the ministry that is the microanalytic unit, the argument fails
completely and there is no reason for a distinction in behavior between old
and new enterprises. Unfortunately, the facts themselves have not been
systematically investigated by anyone; thus we cannot currently use the
differing factual predictions as a means of evaluating the two conflicting
assumptions as to the proper unit of analysis.
APPENDIX
To prove: a(P,lB,)la(P,_,/~,-,) < 0, given either variant (1) of the
ratchet-effect hypothesis or variant (2) for the case Pt-,/bt-, > 1, if and
only if r3Pt/d~t < PtlEt. a(P,/~,)la(P,-,l~,_,) = 0 in variant (2) for the
case PI--ll&--l 5 1.
Proof. Differentiating a(P,/~,)la(P,-,l~,_,) % 0, and using 1.h.s.
= Z[13(P,/~,)lc?l?,], where Z = t&l~3(P,-,&,), we get (Z/l?,)(8Pt18&
-
P,/E,)
I 0. Differentiating Z in turn, since
~8,-,l~Pt-, = 0, Z = (a~??,/
c~P,-,)(~,-,).
In the case of variant (2) of the ratchet-effect hypothesis,
when Pt-Ji?-, 5 1 (and thus L3k,/W-, = 0 by hypothesis), Z = 0 and
thus the 1.h.s. of the original equation = 0. In the case both of variant (I),
and of variant (2) when Pt-,/Etml > 1, Z > 0. This implies that for these
cases the 1.h.s. < 0 iff aP,/& < P,l&.
REFERENCES
Berl iner, Joseph S., The Innovation Decision in Soviet Industry. Cambridge, Mass./London:
MIT Press, 1976.
Bonin, John P., and Ma rcus, Alan J., Information, Mo tivation, and Control in Decen tral ized
Planning: The Case of Discretionary Manager ial Behavior. J. Cam p. &on. 3,
3:235-253, Sep t. 1979.
Granick, David,
Enterpr ise Guidance in Eastern Europe.
Princeton, N.J .: Pr inceton U niv.
Press, 1976.
Granick, David,
Soviet Research and Development Implementat ion in Products: A
Compar ison wi th the G. D. R., In F. Levcik, ed., Znfernational Economies-
Compa rison and Znterdependencies, Festsch r i f t f ii r F. Nem schak . Vienna/New York:
Springer-Verlag, 1978.
Granick, David, Soviet Use of Fixed P r ices: Hypothesis of a Job-Right Constraint,
for thcoming.
Hunter , Hol land, Optimum Tautness in Development Planning, Econ. Developm. Cult .
Change 9:561-572, July 1961.
-
8/11/2019 Granick D. the Ministry as Maximising Unit in Soviet Industry
19/19
THE MIN ISTRY AS UNIT O F ANALYSIS
213
Ivanov, N. V., Lokshin, E. lu., and Dem ichev,
G.
M. ,
Ekonotnika i p lanirovanie
materialno-tekhnicheskogo snabzhe niia promyshlennosti. Moscow: Ekonomika,
1969.
Johnston, J.,
Econometr ic Methods,
2nd ed. New York: McG raw-Hi ll , 1972.
Kurotchenko, V. S.,
Materialno-tekhnicheskoe snabzhenie v novykh usloviiukh
khoziaistvovaniia. Moscow : Ekonomika, 1975.
Rzhesh evski i , V. A., Ekonomichesk ie eksperimenty v promyshelennosti. Moscow:
Ekonomika, 1975.
Rzhesh evski i , V. A.,
Sotsidisticheskii Trud
no. 6:3- 13, 1978.
SHaikin, V. P., Ekonomika i Mutemat icheskie Metody 10, 1:98- 109, 1974.
TSentral noe Statistiche skoe Upravlenie SSS R (TsS U), Ob itogakh vypolneniia
Gosu darstvennogo plana razviti ia narodnogo khoziaistva SSS R v . . godu. Annual
reports published in Pravda, Izvestiia, and elsewhere.