Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Discussion Paper No.21
Preliminary (Not for quotation, for discussion only)
DYNAMIC STRUCTURALISM: AN APPLICATION
FOR POLITICO-ECONOMIC ANALYSIS
by
Office of Systems Analysis*
March 1979
Economic Research Institute
Economic Planning Agency
Japanese Government
Tokyo
* The paper is a part of the politico-economic project at the
office of Systems Analysis of the Institute. The project is
undertaken by Eisuke Sakakibara and Taizo Yakushiji in cooper-
ation with Y. Niimura, H. Inoue, T. Takahashi and Y. Yamamoto
of the office.
Table of Contents
Chapter Page I. Introduction: Dynamic Structuralism . . . . . . . . . . . . . . . . . . . . 1
II. Economic Context of Dynamic Structuralism . . . . . . . . . . . . . . . 6
III. Political Context of Dynamic Structuralism . . . . . . . . . . . . . . . . 15 IV. Dynamic Incrementalism: Politico-Economic "Ensemble" . . . . 28
IV-A Static Nature of the Wildavskian Incrementalism . . . . 28
IV-B Bureaucratic Structural Model: Two-Actor
Bureaucratic Politics Model . . . . . . . . . . . . . . . . 30
IV-C Empirical Results: Evolution of Relative Power . . . . . 36
V. Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Appendix
I. General Dynamic Incremental Model . . . . . . . . . . . . . . . . . . . . . 50
-1-
I. Introduction: Dynamic Structuralism: An Application for Politico-
economic Analysis
"Structuralism" as espoused by Levi-Strauss is a general methodological
position that can be applied to any kind of social study.1) The essense of
"structuralism" as we understand it is to describe observed social relations
with the help of a model of social structure which is not observable. It is
important here to note that "the term 'social structure' has nothing to do
with empirical reality, but with models which are built after it". Social
relations as defined by Levi-Strauss are raw data on which the theorist must
build his models. These raw data are, of course, observable and their
correlations can be calculated. On the other hand, social structure belongs
to different epistemological dimension and can never be "reduced to the
ensemble of the social relations to be described in a given society." This
notion of "structure" corresponds to the concept of unobserved "state" in
control theory or state space models. The "state" in control theory is a
distributed lag function of inputs and white noise and uniquely determine
outputs with some stochastic errors. The state here completely defines the
system but because of the system noise and observation errors, it is usually
not observable. What we observe are outputs which are generated by the system
or by state variables. Social relations in Levi-Strauss definition can be
interpreted in the control theory context as inputs, outputs, and their
correlations. Also, "structure" or system can be defined by a vector of state
variables which are normally unobservable. In the politico-economic framework,
the structure that we want to model is, for example, a specific organization
or a combination of organizations where a certain politico-economic decision
1) Levi-Strauss (1949)
-2-
is made. In this particular study, we are interested in the structure of a
certain sector of government where budgetary decisions on public road construction
are made. The structure which relates to the particular decision making process
can be defined by a vector of unobserved state variables. What determine these
states are laws, regulations or a set of explicit rules and routines, ("conscious"
elements) and the deep rooted psychology of the members of the organization or
associated culture ("unconscious" elements).
In ordinary economic models, these "structure" or state are assumed to be
invariant over time and to be exactly known. This assumption allows the economist
to interpret the causality of events or social relations in terms of inputs and
outputs alone. In "classical" micro models, the relation of inputs to outputs
is always reduced to the level of individuals or individual firms and technical
relations and utility or profit maximization "mechanically" link the two.
The feedbacks or interactions among individuals which would give rise to
the dynamism of the "structure" or state of the social system is ruled out by
assumption. In macro models, inputs and outputs are simply related with each
other with some ad-hoc behavioral assumptions.
In the field of political science, concepts of state exist in abstract
terms. The notion such as power of structure of international relations are
exactly the state or structure that we are describing. However, the dynamism
of changes in structure has not been formally spelled out. Moreover, in many
existing models, rational behavior on the part of a single decision maker is
often assumed and this assumption has eliminated the necessity to describe the
structure or state of decision making process.
Herbert Simon1) and his behavioral school is one notable exception in economics
1) Simon (1959, 1965, 1969), McGuire and Radner (1972), Cyert and March (1963).
-3-
where the adaptive process of the organization has been a major focus. The
study of organizational behavior vis-a-vis such key concepts as factoring,
satisficing, search, uncertainty avoidence and repertoires replaced the
traditional rationality and optimization concepts by the detailed description
of the processes of decision making by the organizations. In Simon's own
words "Much of the activity of science is an application of that paradigm: given
the description of some natural phenomena, to find the differential equations
for processes that will produce the phenomenon."1) Although Simon does not
introduce the concept such as state or structure explicitly, this "differential
equations for processes" are nothing but the evolutions of states or structure
of the organization. Those decision making processes in organizations which
produce the observable phenomenon may not be observable and the aim of his
organizational theory is to describe the observable phenomenon in terms of the
dynamics of these unobservable decision making processes.
Graham T. Allison's2) organizational process paradigm and government (bureau-
cratic) politics paradigm can also be seen as a development in the same direction.
What he calls rational actor paradigm is nothing but the "mechanical" model
which connects inputs and outputs directly with the assumption of an invariant
and known "structure," which summarized by the concept of "rationality". His
second and third models,however, introduce the notion of state or structure
implicitly in terms of the rules and dynamism of organizational process and
bureaucratic politics. His emphasis on "standard operating procedure" as an
important concept in describing the organizational behavior is quite interest-
ing in this context. "Standard operation procedure" for an organization can
be interpreted in our framework as an important element of vector of state
1) Simon (1969). 2) Allison (1971).
-4-
variables that constitutes the state or structure of the government. The state
or structure of the government evolves as identified goals, standard operating
procedures and program changes. As in Simon, Cyert and March, the Allison
model can be interpreted as a description of the dynamic evolution of "structure"
or "state" that produce outputs given exogeneous inputs.
It is our contention, here, that one way to formally spell out this Levi-
Strauss-Simon-Allison "dynamic structuralism" is by the so-called state space
models or Kalman filter models in the engineering literature.1) Here, the
system is formally described by the following two sets of equations;
x(t) = f(x(t-1), u(t), v(t)) . . . . . . . . (1-1)
z(t) = g(x(t), u(t), w(t)) . . . . . . . . . . (1-2)
where x(t), u(t), z(t), v(t), w(t) are vector of state variables, vector of
inputs, vector of outputs, Gaussian system noise and Gaussian output measure-
ment errors respectively. Here the system or structure is completely defined by
a vector of unobservable state variables which evolves through time and reacts
to inputs and white noise. This set of unobservalbe state variables which
constitute the "structure" of the system determine the output stochastically
along with inputs. The Levi-Strauss notion of "social structure", Simon's
"differential equations for processes" and Allison"s "organizational and
bureaucratic politics processes" correspond to the differential equations for
state variables in equation (1-1). The "structure" or state which determines
the pattern of response of outputs to inputs are formally defined by vector of
unobservable state variables which evolve through time. The formal introduction
of this notion of dynamic structure or state which is unobservable to explain
1) See, for example, Athan (1974), Miyatake (1977), Yakushiji (1977).
-5-
observable social relations or phenomenon is the key to this "dynamic structur-
alism" approach.
In this paper, we formally adopt this "dynamic structuralism" or state
space models to describe the government budgetary decision making processes
for road construction in Japan. As explained in detail in the later section,
the Japanese budgetary decision making process for road construction can be
viewed as a "bureaucratic politics" model between Construction and Finance
Ministries with political and other intervention. The structure of bureaucratic
politics can be characterized by a set of unobservable state variables represent-
ing their relative power with regard to this particular decision. This structure
or state, evolving through time and reacting to various inputs, determines
the actual budgetary allocation for road construction (outputs). The model
which is built upon the incrementalist model is relatively crude but it is
hoped that it would fully illustrate the merit of "dynamic structuralism"
approach or the application of state space models for the description of
decision making processes of the dynamic structure of the government and other
organizations.
-6-
II. Economic Context of Dyanmic Structuralism
This "dynamic structural approach" to social systems could be effectively
employed to resolve some of the recent major controversies in economics.
First, we need to recognize that most existing economic models assume that
the "structure" or the "state" of the economic system is invariant over time
or at least within the period of analysis, and that the "structure" is known
to the analyst.
Let us take an example in the "classical" theory of the firm to illustrate
this point. There are three key assumptions in the classical theory of the
firm which define this invariant "structure" or "state" of the system; (1)
production funtion, (2) profit maximizing behavior of the firm and (3) the
"state" of the market summarized by the notion of perfect competition.1) Given
these three succinct assumptions about the "structure" and "state" of decision
making process of the firm "i", outputs of the firm "i" (supply of goods
produced (Y ) , derived demand for labor (L ) and capital (K )) can be expressed
as a function of relative prices (P
,P
) and the fixed parameters of the P P
production function alone.
L = f P
P
K = f P
(II-1) P
Y = f (f P
, fP
) P P
where f is the production function f(L , K ) and f and f are partial
derivative of f with respect to L and K .
1) See, for example, P.A. Samuelson (1948A), J.M. Henderson and R.E. Quant(1958). Indeed, there have been many researches in the field to overcome this classical nature of the theory, notable example of which, for example, is W.J. Baumol (1959).
-7-
If the functional form and the parameters of the production function is known,
outputs L, K and Y are obtained once inputsP
andP
are given. P P
Assumptions (1), (2) and (3) are nothing but the conditions which allow
the form and parameters of the production function, alone, to determine the
invariant "structure" or "state" of the system. One often mistakes the notion
of production function as being a neutral technical relationship, the form and
parameters of which can be determined mechanically. However, we need to re-
cognize here that it presupposes technical efficiency and denotes maximum
obtainable output from every combination of inputs. The second assumption of
profit maximization compels the firm to use the existing techniques most
efficiently to produce maximum output as expressed by the production function.
The beauty of the "classical" theory is that it allows economists to
describe the invariant "structure" of the process in a very neat and simple way.
In this classical world, the "structure" or the "state" of the system can be
completely defined by the functional form and the parameters of the production
function. This relationship among inputs, states and outputs is shown below;
-8-
The "structure" or the "state" of the system, above, is known and invariant as
long as the functional form and the parameters of the production function are
known and invariant. Ordinarily, it is presumed that the firm completely
knows the functional form and the parameters of the production function which
are fixed. This fixity assumption, along with other assumptions mentioned
above, renders the model "mechanical" in the Levi-Strauss sense. That is, the
model allows for "mechanical" correspondence between outputs and inputs without
introducing the concept of "structure" or "state" explicitly.
The classical theory of consumer behavior can be summarized in a similar
fashion. With the assumption of mutually independent utility maximization and
the prevalence of perfect competition, the "structure" or "state" of consumer
behavior is completely summarized by the notion of the utility function. If
the forms and parameters of the utility function are known, outputs of consumer
behavior (demand for goods, say, Y , and Y ) would be given once inputs (relative
price P / P and income Y) are given. This situation is depicted below;
It seems quite obvious that the "structure" or "state" in the above two
systems is unobservable while inputs and outputs are observable. Of course,
-9-
production function or utility function can be estimated using the raw data
for inputs and outputs. Estimation of Cobb-Douglas, CES and other forms of
production function1) or the theory of revealed preference2) are nothing but
attempt to calculate these unobservable states visualized in the theoretical
model by the use of observable data. However, it is often overlooked that
the calculation of the "structure" or "state" represented by the production
(or the utility) functions does not only involves technical or mechanical
relationship stated by the function alone, but also involve all the other
assumptions such as perfect competition and profit maximization in the model.
It is only with all those other assumptions that the simple concept of production
function can summarize the "structure" or "state" and can mechanically link
inputs with outputs. Despite the unobservability of the structure, it is usually
assumed in the theoretical and empirical discussions that the true structure
of the system is invariant and that the actors in the system as well as the
analyst completely knows this invariant structure or state. The "classicists"
as well as "neo-classicists" usually take such position.
Let us now turn to the macro or Keynesian model. In this type of model,
the "structure" (or "state") is usually a black box although some plausible
descriptive explanations are often given.3) However, these explanations do not
usually provide complete logical statements of the "structure" and on this
score, they are often termed ad-hoc models. Let us illustrate this point in
a simplistic Keynesian consumption function, the relationship among inputs,
states and outputs are as follows;
1) M. Brown (1967). 2) P.A. Samuelson (1948 B) 3) R. Dornbusch and S. Fischer (1978) is an excellent recent textbook in
this area. See, also, A. Leijonhuvud (1968) for a critical account of "Keynesian" models.
-10-
Since the form and the values of parameters of consumption function which
constitutes the "structure" or "states" of this model cannot be explained
logically from any basic assumptions of human or organizational behavior, the
model is called ad-hoc. One might claim that the life-cycle hypothesis or
permanent income hypothesis is exempt from this criticism of ad-hockery.1)
However, to the extent that forecasting procedure of consumers assumed in these
models is adhoc, the model is still adhoc. If one tries to make macro-consump-
tion models consistent with classical micro models of consumer behavior, the
notion of "rational" expectation needs to be introduced in the former.2)
Since total "structure" or "state" could never be completely spelled out,
some ad-hockery is a necessary evil if one is interested in the total explanation
of the system in question. Policy makers, who want to forecast and control,
are usually in this position.
The problem, viewed in this context, lies not much in the ad-hockery but
more in the assumption of perfect knowledge and invariance of the "structure"
or "state", Keynesians are similar to "classicists" in this respect. For
example, they take the form and parameters of the consumption function which
are estimated as fixed. That is, they mechanically connect input and output
1) J.S. Deusenberry (1949), M. Friedman (1957), A. Ando and F. Modigliani (1963). H.S. Honthakker and L.D. Taylor (1966).
2) T.J. Sargent (1978).
-11-
with these fixed parameters. In other words, Keynesian models are also "mechani-
cal" in the Levi-Strauss definition. In terms of policy analysis, both of these
two mechanical models, "classical" and Keynesian, calls for deterministic optimal
control.
It is this framework of deterministic optimal control, which is a result
of the assumption of known and invariant "structure" or "state, that sometimes
makes policy controversies between Keynesians and non-Keynesian (monetarists,
classics and etc.) unnecessarily metaphysical. The issue is often the subjective
choice of who is and should be the optimizer of this deterministic control
system where the real issue should be supplementary learning process or "adaptive
control".
In "Keynesian" models, government expenditures, taxes, money supply and
others are instruments of control and the government controls the output re-
presented by economic variables such as GNP and consumption through these inputs
given the "structure" which is assumed to be fixed and known. It is true that
frequent reestimation of econometric models used to identify the "structure"
implies some learning process or "structural" change but such adjustment is
ad-hoc and is not formally incorporated in the model.
In "classical" economic models, the same type of behavior, that is,
optimization of well defined objectives on the basis of perfect information,
is assumed for the private sector and the concept of government as a single
controlling agent is non-existent. The major difference between the two types
of model is whether it is the government or the private sector that optimizes.1)
1) The symmetry is not perfect in that the private sector is not a single entity in "classical" models. In "classical" models the private sector consists of infinite numbers of individuals and firms and it is the market which co- ordinates the infinite numbers of optimizing agents. Thus, the optimality achieved in these models which is called Pareto optimality is relative as against absolute or normative optimality accomplished in Keynesian framework with social welfare function. Also note that in these "classical" models, the problem of interdependence of expectations and controls which is probably the essense of multiple agents model is assumed away.
-12-
As a matter of fact, markedly different policy implications of the two models
which have been sources of constant debate and controversy1) in the economic
profession derives mostly from this alternative assumption about who actually
optimizes.
The "Keynesian revolution" challenged the classical belief that
it is the private sector that optimizes and should optimize by pointing out
rigidities in the private market. The rigidities such as inflexible wages
and unemployment renders the optimization mechanism of the private sector
imperfect and makes the government intervention necessary. Not only that, but
statistical macro economic relationships derived by Keynes and others presumes
the passivity on the part of private sector. The "Keynesian revolution", inter-
preted in this manner, is nothing but the complete shift of optimization visualiz-
ed in the theoretical framework from the private to the public sector. Thus,
it is only natural that government (fiscal and monetary) policies are considered
highly effective in controlling the private sector (real values of GNP for
example) in Keynesian models.
Recent debates on rational expectations focus again, on this issue.2)
If the expectation of the private sector is rational, then it follows that it
is the private sector which attempts to control. Note the important duality
relationship between forecast (expectation) and control. The assumption of
perfect forecast on the part of private sector is logically equivalent to its
1) See for example, A.S. Blinder and R. Solow (1973), M. Friedman (1977), F. Modigliani (1977).
2) R. Barro (1976, 1978), R. Lucas (1972, 1973), T.J. Sargent (1973, 1976),
B.M. Friedman, (1977, 1978), S. Fischer (1977, 1978), O.J. Blanchard (1978), R.J. Shiller (1978).
-13-
complete control. Thus, a major conclusion of rational expectation models
that only the nonsystematic parts of government policies can affect the private
sector follows naturally from the assumption that their expectation is "rational",
since "rationality" is defined in their models as the ability to correctly
forecast the economy except for random disturbances.
Forecasting here is only necessary for the white noise part of the
system. However, since white noise is zero on the average, forecasting or
control is complete on the average and errs only in the short-run. The long-
run classical property and short-run non-neutrality of rational expectation
models is nothing but another way of expressing the above definition of
"rationality".
Thus, as long as one maintains the assumption of invariance of the
"structure" of the system as in "Keynesian", "classical" or rational expecta-
tion models, the important policy question such as effectiveness of fiscal and
monetary policies always boils down to the subjective choice of the sector
which should or does actually optimally control the economy. If people believe
that it is the government which should optimize the economy and behave according
to this belief (thati is, the private sector behaves passively and government
optimizes), government policies are effective in controlling the economy.
On the other hand, if the private sector believes that it is the private sector
which optimizes and if both the government and the private sector behave
accordingly, classical neutrality results follow.1)
1) The situation where both sectors try to forecast and control the economy is inconsistent with the assumption of invariance and perfect knowledge of the "structure" of the system.
-14-
The reality probably is not in neither of these extreme situations. The
people do not know the true "structure" of the economic system and learn only
through experience. Equivalently, the true "structure" changes over time. As
pointed out by B. Friedman,1) rational expectation, in this situation, is actually
adaptive, and the innovation in information, that is, the difference in actual
and expected value of the observations become the key factor in describing the
forecast and control mechanism. The notion of adaptive rather than deterministic
optimal control is essential, here.
In terms of analyzing and describing the system, such methodology as a
"dynamic-structural" approach described in the previous section becomes
necessary. In the model, the concept of "structure" or state needs to be
introduced formally. Moreover, dynamics of "structural" change, or equivalently,
the mechanism of adaptive learning process needs to be explicitly spelled
out. The "structure" or the "state" of the system can be identified on
different levels according to the sophistication of the theory one develops.
One can only identify the "structure" given by visible institutions, laws and
norms or one can go deeper into the psychology of persons involved and the
culture associated.
In what follows, we would model the structure of bureaucratic and politi-
cal interaction in budgetary decision making on road construction in a most
crude way building upon the incrementalist model. It is attempted here to
describe the mechanism of evolution of the structure or state of interaction
between bureaucracy and polity and show how the "dynamic structural approach"
could be effectively employed to model the certain sector of politico-economic
system.
1) B. Friedman, (1977).
-15-
III. Political Context of Dynamic Structuralism
There appears the cyclical ups and downs of confidence of the "fuzzy"
theoretical advocates in social science against the Eaglet-like1) disdainful
criticism from natural scientists for the lack of the clarification of concepts.2)
Political science in particular, has suffered from a long agony since the
establishment of the Royal Society.3) When asked the clear definitions of
nomenclatures, political scientists have given almost no persuative answers.
In such a stormy cynicism, a clever sophistry is given to the declaration
of the end of "hyperfactualism" and the "conscious" borrowing of the matured
sciences.4) It is often true that reality is more cynical than optimistic
expectations: The amalgamation of political science with other disciplines
has often led to a dreadfully insensitive use of technical words and methods.
The frustrating dualism between the words-rich theoretical arguments on an
abstract level and empirical research with excessive foci on methodological
sophistication thus seems to be a prevailing dilemma for students in politics.
Whether it is a conscious borrowing or unconscious one is a difficult
issue. Easton wrote:
"In exploring theoretical alternatives in political science, we have usually succeeded in distinguishing clearly between methodology and the substantive treatment of theoretical issues, assigning to each its proper place and weight."5)
Our observation is, to regard this as simple political substance appears
in disguise of technical sophistication. Examples are numerous.6) What is
1) L. Carroll,(1960). 4) D. Easton ed., (1966). 2) J.G. Kemeny,(1959). 5) op. cit. 3) J. Salomon, (1973).
-16-
needed here is a counter argument within ordinary science.1) Given the
above examples, a strong motivation immediately arises in casting "conscious"
doubt against the above Easton's statement which looks too legitimate that
we "unconsciously" agree to. Our counterargument is that political theory
and methodology are inseparable. An improper weight on methodology is
misguided;
6) The Blalockian recipe for bit-by-bit figure calculation (H.M.Blalock, 1960.) was combined with fancy SPSS Kitchen tools(Statistical Package for Social Sciences), which has consequently inflated the budget bills as well as intellectual density in political research and eduction; A group of people were devoted to reconsideration of the 1930s Richardsonian differential equations (P. Smoker, 1969), which was reshaped into the sophisticated optimal control model; (J. Gillespie, et al., 1977), The Brunner and Brewer simulation of political develop- ment was based on Keynesian multiplier analysis (R.D.Burunner G.D. Brewer, 1971); In a similar fashion, Forrester's world dynamics (J.W. Forrester, 1971), what the economist Nordhaus called "measurement without data" (W.D. Nordhans, 1973), functioned as advocacy for the non-linear political system against the ubiquitous linearity in most political models; A new development in time-series analysis and the availability of computer packages (TROLL System) fostered a bold optimism for war-occurrence probability calculation (N.Choucri & R. C. North, 1974); A "search of the political curves" (A. Przeworski & G.A.D. Soares, 1971) turned out to be a deliberate application of classical control theory on the undergraduate level in engineering; An Anglo-American collaboration for the "noise filting" (G.E. Box & G.M.Jenkins, 1976), gave rise to the "ignorance-except-for-inter- ventions" model (D.A. Hibbs, 1974). 1) T. Kuhn,(1970)
-17-
and an improper weight on political theory is equally misleading. All problems
seem to originate from the fatal misunderstanding of the separability of political
theory and methodology. It is exactly this point to which our "dynamic structuralism"
addresses itself in the context of political science.
Consider, for example, the voting analysts' common conjecture on the causal
relationship between social atrributes and partisanship on voting.1) As sociologists
have consumed valuable energy, social attributes data can be collected in an
"observable" explaining variable vector. Gallop and other similar institutions
made partisanship data available. Then a political scientist (called the first
political scientist for convenience of discussion) who has read Blalock's (or
others') textbook and knows the existence of the SPSS (or similar statistical
packages) would immediately formulate the linear regression model:
y = xβ
where y: partisan voting variable
x: social attribute variable vector
β: regression coefficients.
Suppose that the second political scientist dislikes "methodology" so much as of
course knows neither Blalock or SPSS. He would maintain his own normative
political theory such that the some social attributes determine the political
partisanship through "political consciousness," "disposition to political
information," and the like2). These concepts are assumed to be "unobservable
time-invariant theoretical entities" but act as the linear transformation of
the associated social attributes. He might propose the same mathematical model
as follows:
1) A. Campbell,P.E. Converse, W.E. Miller and D.E. Stokes,(1967).
2) On this point, see, for example: S.M. Lipset,(1963).
-18-
y = xβ
y: partisan voting variable
x: social attributes variable
β: "political concepts" as the linear transformation of the associated social attributes.
Now the third political scientist comes along. He in virtual agreement with
the second political scientist, except that some of political concepts should
change cross-sectionally or over time. Then his theory is expressed equivalently
as:
Y = xβ ,
β = f (t),
where t denotes either time or an individual sample in
cross-sectional data.
As far as the mathematical expression is concerned, it is impossible to
identify which political scientist has which political theory. The difference
becomes clear only when we know how he "methodologically" treats empirical data
in his model. The first political scientist would have no strong political
theory. He is interested merely in feeding different social attribute data in
x and partisan voting data in y and "estimating the parameters." But he should
be certain that in applying the regression technique he adopted the "aggregated
political theory" such that individuals' social attributes make impacts on
partisan voting in a quite different magnitudes. What he estimates in β is the
aggregate magnitude in a sampled political group and can be no longer applied
to any individual.
On the other hand, since he has not acquired sophisticated methodological
-19-
tools, the-second political scientist can not but underutilize the available
data. He would pick up only a small sample data and see the relationship
between the social attributes and partisan voting trends. And, next, he
probably "simulates" a hypothetical situation by "subjectively" putting a
different value in β and calculates the resultant y in a weak attempt to check
his theory. Since he has no knowledge for estimating β, he can still maintain
his normative political theory, not on an aggregate level but on an individual
basis. He might, however, be condemned for a unscientific approach by other
political scientists who hold an excessive belief in regression technique. But
his methodology is quite appropriate to match his political theory. Political
scientists who criticize him meant to convert forcefully his normative politi-
cal epistemology to their aggregate world-looking. It is quite likely that the
theory of the second type political scientist might be strongly supported by
natural scientist who have extensive expertise in statistical methodology and
know the limits of regression. Thus, in place of the normative political scientist,
they spell out that the above normative political theory can only be tested by
a deterministic "maximum likelihood" simulation. They are the so-called "system
dynamics" advocates.
The political theory of the third political scientist best represents our
intuitive understanding of political phenomena. It releases the least likely
constraint such that the transformational political concepts are in eternal
quiescence. However, in political science methodology no one knows how to
handle empirical data for this theory except for the subjective "maximum likeli-
hood" simulation as in the second political theory. For,totally drowned in the
biased epistemology of linear regression, the first political scientist has no
-20-
idea that he can also admit the third mathematical model with the constraint:
β = β .
This idea comes only from the third political scientist for the above formula
is only a special case of :
β = f (N).
The implication drawn from the difference in political theory between the first
and the third political scientists is quite profound in regard to the methodologi-
cal choice. Let us see how the third political scientist treats the regression
methodology with a different viewpoint. First, reformulating the first model
gives,
β = β ,
y = x .β .
As discussed elsewhere than in political science,1) the recursive estimate
β would fluctuate as a new individual's data (y ,y ,・・・・; x ,x ,・・・・)
enter. Then, the change in β depicts somewhat one below:
("Recursive" Estimates of political concepts)
(Number of Individuals Counted)
1) D.G. Luenberger,(1969).
-21-
Until individuals are sampled up to N , the "aggregate" estimate of political
concepts remain unchanged. But, when the N+1-th individual is counted the
estimate drops sharply. This indicates the significant impact of one individual
data. As far as the ordinary regression technique is applied, this impact is
not conceivable. It is worth nothing that such an emphasis on the impact of
individuals shares the same epistimology as of the second political scientist.
More formally, the third political scientist's theory is in what we call
the "dynamic structural" approach. The excellent arguments by Przeworski and
Soares1) helps facilitate our further discussion. What they meant by the
Lockean interpretation of class-consciousness exactly coincides with the first
political theory, but in a concrete and simplified version such as:
y = a + px
where y: the proportion of left-party votes (i.e., partisan votes)
x: the proportion of workers in a constituency (i.e., social attributes of a class) a: y when x = 0.
In this model, they explain the Lockean theory of class-consciousness as:
"Intuitively, it seems that class-consciousness, as the term is used, connotes the extent to which the workers vote for the working class party, or move exactly, the proportion of the workers who vote this party . . . . . . this interpretation would imply that consciousness of a class is simply an aggregate of consciousness of individuals. Some individuals are class-conscious -- they vote for the party of their class -- and the class conscious of itself to the extent to which individuals who are conscious frequently occur within this class."2)
In contrast, the third political theory can more easily accomodate the Marxian
theory of class-consciousness3) such that:
1) A. Przeworski, and G.A.D. Soares,(1971).
2) op. cit., p.54.
3) R. Dahrendorf,(1959).
-22-
". . . . . . . . . . . the consciousness of an individual is determined by the consciousness of his class, not vice versa. . . . . . . it seems more appropriate to treat class consciousness as a nascent concept rather than as a characteristic of state. . . . . . . . . . It should be emphasized that "class- consciousness" is a theoretical construct and cannot be directly observed. Only the proportion of workers (x) and the proportion of the left vote (y) are directly observable."1)
Our dynamic structuralism accepts most of the above interpretation except
that we consider that class-consciousness is truly a characteristic of state
but it is a nascent state of a class. Implying the models of the first and
third political theory, we entertain the following block diagrams:
The First Political Scientist's Class-consciousness Theory (Lockean)
The Third Political Scientist's class- consiousness Theory (Dynamic Structuralism)
Rather than Przeworski and Soares' non-linear formulation of the dynamism of
P as the first derivative dy/dx, the dynamic structural political theory
smoothly maintains the Marxian definition such that class-consciousness is
the dynamic state of class structure and such a state completely governs the
partisan vote. Thus, we entertain the model:
1) op. cit.
-23-
P = f (p , p ,・・・・・・・; u , u ,・・・・・・・・)
y = P . x .
where u : causes to change class-consciousness (independent of x ).
It is worth noticing that class-consciousness has its own memory and is
perturbated by the real exogenous causes u 's. The methodology to handle
this theory has not been available until the 1960s.1) It is called the state-
space method, and which is still incomplete in the sense that f in p 's
equation must be linear.2) For a simpliest theory that class-consciousness has
only the first-order memory and is perturbated by the unknown random cause
like:
P = p + v
y = p . x ,
the afore-mentioned recursive regression is directly applied with a minor
change.3) It is quite cynical that the 19th century Marx's political theory
already implied the 1960's mathematical theory. This strongly suggests the
danger of the pseudo-expertization of methodology in political science, which
is exactly the same stance as structural functionalism takes.
Our dynamic structuralism can give further insight into what have been
called "synchrony" or "diachrony" by the system-analytic-minded political
scientists.4)
1) R.E. Kalman,(1960).
2) K. Furuta,(1967).
3 P.C. Young,(1971).
4) F. Cortes, A. Przeworski and J. Sprague,(1974).
-24-
By their definition, synchrony implies the time-invariant system structure
as in the previous reformulation of the first political theory by the third
political scientist such that:
β = β
y = x . β .
It should be noted that the above model a priori specifies dynamic causality.
In a similar manner, an example of diachronic model resembles our dynamic
structural formulation of class-consciousness discussed above:
β = f (β , β , ・・・・・・・; u , u ,・・・・・・)
y = x . β .
Thus, diachrony is the system property of the time-invariant structure, i.e.,
dynamic structure.
The fundamental difference between our epistimology of social dynamics
and Cortes, Przeworski and Sprague's system analysis is in regard to the
concept of the system "state". In their definition, the system states are
the "intermediate observable outputs", whereas by our definition they are the
"full information of the behavior of a system" or the "intermediate unobservable
concepts". This difference can be clarified in the example below.
Cortes, Przeworski and Sprague propose the dynamic process in which social
tension becomes open conflict through socialization, political repression and
political organization. Their theory is succinctly illuminated in the follow-
ing system diagram:
-25-
X and X are what have been called the "states" by Cortes, Przeworski and
Sprague. They are both observable intermediate outputs. On the other hand,
the application of our dynamic structuralism to the same model yields the
following flow diagram:*
-26-
*) The imput-output formula for Cortes, Przeworski and Sprague model is expressed with the z-function as:
y = PS/R
.u . z + [(1-R)/R]
Then,
Y(z) = PS/R .
u(z) z + [(1-R)/R]
Multiplying the both sides by [z .x(z)] becomes:
y(z) = (PS/R).[z .x(z)] . . . . . . . . . . . . . (1)
u(z) = x(z) + [(1-R)/R].z .x(z). . . . . . . (2)
(2) is rewritten as:
x(z) = -[(1-R)/R][z .x(z)] + u(z). . . . . (2)'
(1) and (2)' give the above flow diagram. And, their z-inverse yields the time-domain dynamic structural equations.
-27-
The above diagram is equivalent to the "dynamic structural" equations:
x = -[(1-R)/R].x + u
y = (PS/R).x .
The distinct feature of the dynamic structural formulation is the fact
that there exists only one state variable sufficient to govern open conflict.
This state is not observable. The first equation suggests that the state
evolves as the first order autoregressive scheme. The magnitude of the
autoregressive factor is (1 - 1 ) which is negative with the absolute value R
greater than 1. Thus, the state oscillates explosively. The coefficient of
the input (the level of tension) is unity, so that the dimension of the state
is equivalent with the level of tension. Therefore, the second equation
indicates that the state as the degree of tension of this system is multi -
plicatively accelerated by the effectivenesses of both political socialization
and organization of the system, and decelerated by political repression, in
becoming open conflict. If the initial state of the system tension is known,
the dynamic structural model suggests that the future behavior of open
conflict is completely known, while the two states in Cortes, Przeworski
and Sprague's synchronic model can not supply such complete information.
This is the implication of our dynamic structural epistemology of political
behavior.
-28-
IV. Dynamic Incrementalism: Politico-Economic "Ensemble"
IV-A Static Nature of the Wildavskian Incrementalism
The well known Wildavskian incrementalism model summarizes budgetary
decision making process of the government by a simple reduced form equation
of the following form.1)
G = gG + E (IV-1)
Although some modifications are made to allow for deviations from the above
rule, those are simply the exceptions and the basic hypothesis of the model
is that the state or structure of the system of budgetary decision making
process is invariant over time and can be represented by a single scalar. That
is,
g = g (IV-2)
Because of this static nature of the assumption, the system is closed
except for ad-hoc modifications inserted arbitrarily by the model builder and
a strange conclusion follows that bureaucracy cannot generally be influenced
by outside interventions. It is true that Allisonian standard operationing
procedure is quite rigid and, in that sense, bureaucratic response is
predictable. However, it is quite a different matter to assume the predictable
patterns of behavior, or the existence of structure or state, and to assume its
invariance.
This invariance assumption, as pointed out in previous sections, allows
the analyst to disregard the notion of structure or state completely and
jumps directly to reduced forms or input-output relations without explicitly
specifying interactions among observable inputs and outputs and unobservable
states. The Wildavskian incremental model is very typical, in this sense.
1) A. Wildavsky (1964), J. Wanat (1974), F.W. Hoole (1976), F.W. Hool, L.J. Brian and J.T. Harvey (1976). Y. Noguchi et al. (1976)
-29-
Behavioral models of budget a la Crecine1) attempt to spell out the
processes of the budget compilation in some details but they fail to avoid the
pitfalls of directly substituting the unobservable process by some observable
relation. Since there is no logical way to directly substitute unobservable
process by observable relation, their models usually suffer from some ad-
hockery or arbitrariness.
In the next section, we adopt the dynamic structural approach to the
analysis of the Japanese budget compilation where structure or state is assumed to
be time-variant and explicitly written down in terms of unobservable state
variables. Note that, we spell out the unobservable dynamics of the system
in terms of unobservable state variables, and not in terms of any observable
relation as in behavioral or SD models.
1) J.P. Crecine (1969)
-30-
IV-B Bureaucratic Structural Model: Two-Actor Bureaucratic Politics
Model
The Japanese budget compilation process is somewhat different from that
of the United States in that the role of Diet (Congress) is inconsequential.
Throughout the post-war years,the government draft, formally agreed upon by
all branches of bureaucracy, had gone through the Diet untouched except for
the 1977 budget where a very minor nominal change was made.1) The budget is
essentially completed through the interactions of three major participants,
Ministry of Finance, the Ministry or Agency in charge of particular budget
items and the ruling Liberal Democratic Party before it is submitted to the
Diet. Since the consensus of the ruling party which has had the absolute
majority in the Diet is achieved for the governmental draft, the revision of
the draft in the Diet is unlikely.2)
Thus, it is appropriate to model the Japanese budgetary process for road
construction as a two-actor (Ministry of Finance (MOF) and Ministry of Con-
struction (MOC)) Bureaucratic Politics Model with political interventions from
outside. There are some other minor participants in the proecess such as
1) For the details of Japanese budget compilation process, the reader is referred to J. Cambell (1977) and Y. Noguchi et al. (1977)
2) John Campbell gives a very good and concise overview of Japanese budget works in his book J. Cambell (1977), an revised except of which is given below: "Under the Constitution, the annual budget is prepared by the Cabinet, but provisions of the Finance Law assign the responsibility of actual preparation to the Ministry of Finance (Okurasho; MOF). The process begins shortly after the beginning of each fiscal year on April 1, when the various ministries and other agencies begin formulating their budget requests for the following fiscal year. Requests, are submitted to the Finance Ministry by August 31, after securing the general agreement of the Liberal Democratic Party (Jiyu Minshuto; LDP)--in particular, that of the appropriate division (bukai) of the Policy Affairs Research Council (Seimu Chosakai, or Seichokai: PARC) .
-31-
During the month of September, ministry officials appear at the MOF Budget Bureau (Shukeikyoku) to explain these requests to the budget examiners (shukeikan), who ask questions about programs but do not actually negotiate over budget figures. In October and November, examiners and their staffs go over the requests in detail, discussing possible options with the respon- sible bureau vice-director, the director and other officials. This process is what we term "microbudgeting"; at the same time higher-level MOF officials are discussing the total budget figure, the size of the yearly tax cut and other "macrobudgeting" matters. For about two weeks in early December, the MOF ministerial budget conference (yosan shogi) meets to ratify, and some- times modify, the draft budget prepared within the Budget Bureau; at time of release (naiji) it is called the "Finance Ministry draft" (Okurasho gen'an). At about the same time, a brief and abstract Budget Compilation Policy (Yosan Hensei Hoshin) is drawn up within the Budget Bureau, ratified at the MOF top level, and passed by the Cabinet.
For the most of this period the regular organs of the majority party do not intervene in the compilation process going on within the Finance Ministry, although individual Dietmen will often let their preferences be known (for example, by accompanying groups of petitioners to the Finance Ministry). However, in December the Policy Affairs Research Council's Deliberation Commission (Seisaku Shingikai or Seicho Shingikai) draws up the annual LDP Budget Compilation Program (Yosan Hensei Taiko). This program is then passed by the party's Executive Council (Somukai) and referred to the Cabinet and MOF before the release of the Finance Ministry draft.
The release of the MOF draft--scheduled for mid-December, although more often than not it is postponed into January or later--begins the "revival negotiations" (fukkatsu sessho) period of about a week. During this rather hectic period ministry appeals are heard by progressively higher levels of Finance Ministry officialdom, and small amounts of supplementary funds which are usually hidden unspecified in the MOF draft are doled out. The Liberal Democrats also formulate "political" revival requests though a series of brief divisional hearings followed by reviews at higher levels, which are presented during the penultimate Cabinet-level negotiations or at the final session, one which often has lasted all night, between top officials of the party and government. The resulting budget is then ratified by a meeting of the Cabinet to become the government draft (Seifu-an), and is sent on to the Diet for passage."
-32-
Economic Planing Agency (EPA), the press, pressure groups and etc. However,
the leverage of EPA and the press is usually concentrated in the overall size
of the budget and how that leverage is transformed to specific items of budget
depends upon LDP. The leverage of pressures groups on specific items also
filters through LDP. Thus, for the current study they can be clustered with
the LDP in a "political" sector. A major input to MOC each year in drafting
their original request is their "objective" estimation of demand for new road
construction and their repair which are based upon the successive Five Year
Road Construction Plans. The plan is a yardstick but it is informally revised
each year as original estimation of demand is corrected. MOC does have a
formal macro model to calculate new demand, the output of which is compared
to the cumulative micro figures collected from local agencies and revised to
obtain their original request.3) In what follows it is assumed that their
estimate of the growth rate of future road demand depends upon past growth
rates of changes in number of cars. For Ministry of Finance, what is important
is, of course, their estimate of revenue. For road construction, particular
taxes (Gasoline tax, oil and gas tax, local road concession tax, oil and gas
concession tax, light oil tax, automobile tax and automobile tonnage concession
tax) are earmarked but general tax revenue is also used. Moreover, funds from
Fiscal Investment and Loan Program (so called FILP, bulk of which come from
postal savings) are used mainly for the construction of toll road. Thus, the
MOF makes formal estimates of these revenues at the time of the budget compila-
tion which are released with MOF draft of the budget.
3) The reader is reffered to E. Sakakibara, T. Yakushiji et al. (1979) for a detailed institutional description of this and other processes described in this section.
-33-
First, corresponding to the two actors, we assume one state "RP" which is
an unobservable theoretical concept and which represents the relative power
of the two ministries. The basic assumption, here, is that the "relative power"
of the two ministries changes over time depending upon the magnitudes of demand
pressure, "political" interventions and other interventions. Specifically
RP = RP x x x e . . . . . . . . . . . . . . . . . . . (IV-3)
or logRP = αlogRP + βlogx + γlogx + δlogx + ε
where x : factors affecting estimated demand for road construction by
MOC (purely "rational" calculations without "political" considerations).
x : factors directly affecting "political" pressure x : interventions other than x RP: relative power of the two ministries ε: white noise
The relationship among various sectors are depicted in Figure 1. Given
this "relative power", the expenditure for road construction is given by,
R = RP ・T e . . . . . . . . . . . . . . . . . . . . . . . (IV-4)
or logR = logRP + logT + u
R = total amount of budget allocated for road construction
(National and FILP)
T = tax revenue for road construction as estimated by MOF
Note, here, that the state variable RP, relative power, is defined as the
proportion of road budgeting to MOF's estimates of revenue with an observation
error.
-34-
If the "relative power" of MOC is very strong, the road budgeting far
exceeds estimated revenue, while if MOF's relative power prevails, the budget
falls within the estimates of revenue. Our task, then, is to estimate the
following reduced form equation which is derived from equations (IV-3) and
(IV-4),
logR(t) = logT + β
logx +γ
logx 1-αL 1-αL
+ δ
logx +
ε + u . . . . . . . . . . . . . (IV-5) 1-αL 1-αL
where L is the lag operator.
-35-
Figure I
Bureaucratic Politics Model of
Japanese Budgeting Process for Road Construction
-36-
(IV-C) Empirical Results: Evolution of Relative Power
The two-actor bureaucratic politics model specified in the previous
section is estimated using two sets of data, original and revised budget figures
for road construction. The original budget is submitted by MOF in mid-December,
although it is often postponed into January and later, and approved by the Diet
during its regular session (150 days starting in January). As explained earlier,
original budget is virtually completed when MOF draft is released in mid-
December although minor changes are made in the "revival negotiation" that takes
place in its subsequent week. Thus, exogeneous factors x , x and x that affect
relative power, RP, in reduced form equation (IV-5) should be lagged one period.
Of course, longer delays and lags for x , x and x are theoretically possible.
It has been a rule in post-war Japan that the original budget is revised
around the second quarter of that fiscal year, considering the developments of
the economy since the compilation of the original budget. First of all, tax
estimates released by MOF at the time of submission of MOF draft in the previous
year is revised. The original MOF estimate is quite conservative and the re-
vision is ordinally upward. During the post-war period, downward revision has
been made only twice, in 1965 and 1975. Accordingly, budgets were revised up-
ward by the additional floatation of national bonds. Even in those years when
tax estimates were lower than actual (1965 and 1975), the budgets were revised
upward. Also since there is an interval of around nine months between the
compilation of original and revised budget, factors affecting internal and
external pressure (cyclical and external conditions of the economy) change
substantially. In the following estimation, such developments in taxes and
exogeneous variables are taken into account.
-37-
Road construction projects in Japan can be roughly divided into three
categories, general road projects (national projects and nationally subsidized
projects), non-subsidized local projects, and toll road projects (mainly
financed by Fiscal Investment and Loans Program). Since the leverage of both
ministries of Finance and Construction is relatively small for non-subsidized
local projects, only national and toll road projects are analyzed in this paper.
For the fiscal year 1978, the sum of national and toll road projects in the
original budget amounted to 3,301.9 billion yen (70.0%) out of total road
budget of 4,719.9 billion yen.
Since the reduced form equation (IV-5) contains autoregressive lag operator
term, (1-αL), the equation is estimated using the maximum likelihood estimation
algorithm a la Davidon-Fletcher-Powell. The algorithm allows for the estimation
of an ARIMAX model which is equivalent to rational distributed lag model where
rational lag distribution is allowed both for exogeneous and error terms. Although
there are severe restrictions of degrees of freedom because of the annual nature
of fiscal data, some experiments are conducted for various lag structure of
error terms and exogeneous variables.
Standard results for both original and revised road budget figures are
presented below.
(i-1) Original budget
(1 - 0.2732*L) [log R - logT ] = 0.0854logx - 0.6381logSTA
(].53) (1.45) (2.59)
- 0.0768EXP - 0.1394I68 - 0.1410Ioil . . . . . . . . . . . . . (IV-6)
(1.26) (3.74) (5.14)
-38-
R = 0.7575
D.W. = 1.88
S = 0.0356
Period = 1959 - 1978
where R = original road budget for national and toll road projects
t = national revenue (national taxes and financial sources of Fiscal Investment and Loan Program--FILP) as estimated by MOF at the time of release of MOF budget draft. For specified taxes, original estimates are used while for general account and for FILP revenues, total general account and FILP estimates are multiplied respectively by the previous year's shares of road financing out of the total.
x = c
c
c = five year average of total numbers of automobiles owned
STA = this year's estimate of real GNP growth over five year average of real GNP growth lagged one period.
EXP =ΔFR
- ΔEX HM EX
FR = Foreign reserves (average for the third fiscal quarter)
HM = high powered money (average for the third fiscal quarter)
EX = exchange rate (average for the third fiscal quarter)
I68 = a dummy for fiscal restraint campaign 1 for 1968 0 otherwise
Ioil = a dummy for "oil shock" 1 for 1973 and 1974 0 otherwise
(ii-1) Revised budget
(1 - 0.6001*L) [logR - logT ] = 0.3273logx - 0.6432log STA
(4.44) (2.75) (2.03)
+ 0.2711EXP - 0.1147I68 - 0.0726Ioil . . . . . . . . . . . . (IV-7)
(3.05) (2.01) (1.57)
-39-
R = 0.5517
D.W. = 1.51
S = 0.0515
period = 1959 - 1977
where R = revised road budget for national and toll road projects
T = average of T and T*
T* = realized national revenues for road construction. That is, realized specified taxes plus realized total of general taxes and FILP revenues multiplied respectively by the previous year's shares of road financing out of the total.
STA = this year's realized growth rate of real GNP over that of previous five year average
EXP =ΔFR
- ΔEX HM EX
FR = foreign reserves (average for the second fiscal quarter)
HM = high powered money (average for the second fiscal quarter)
EX = exchange rate (average for the second fiscal quarter)
The fits for both equations is quite good for the dependent variable
which is essentially a percentage change. In view of the marginal significance
of some of the coefficients in (IV-6) and (IV-7), minor alternations are
introduced in the following equations,
(i-2) Original budget
(1 - 0.2551*L) [log R - logT ] = 0.1337logxl - 0.6229logSTA
(2.49) (2.44) (2.83)
+ 0.3094EXP *I7678 - 0.1481I68 - 0.1465Ioil . . . . . . . . . . (IV-6)'
(2.24) (4.29) (5.79)
-40-
R = 0.7943
D.W. = 1.85
S = 0.0328
period = 1959 - 1978
where I7678 = a dummy for U.S. pressure for inflationary policy 1 for 1976, 1977, 1978 0 otherwise
(ii-2) Revised budget
(1 - 0.5031*L) [log R - logT ] = 0.4280logx - 0.4813logSTA
(3.98) (3.83) (1.66)
+ 0.2356EXP + 0.9574EXP * I7678 - 0.1247I68
(3.01) (2.34) (2.64)
- 0.0613Ioil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (IV-7)'
(1.48)
R = 0.6137
D.W. = 1.88
S = 0.0446
period = 1959 - 1977
For other versions the two equations, readers are referred to Tables I and II.
The estimation results seems to verify our general framework for budgetary
decision making. In particular, it is confirmed that exogeneous interventions
or changes in the environment have significant impacts on the compilation of
the road budget. Internal stabilization needs as well as external pressure
have considerable effects on the relative power of the two ministries. Their
combined effects are usually much greater than the autoregressive parts.
-41-
Also, discontinuous interventions or disturbances such as "the fiscal
restraint compaign" of 1968 and the "oil shock" of 1974 have considerable
impact. The "fiscal restraint campaign" was conducted under the leadership
of Kotaro Murakami, then vice minister of Finance, which was the reflection
of the general mood of the Finance Ministry, at that time, on the rigidity
and the secular increase of fiscal deficits. Murakami had intended to abolish
the "standard operating procedure" for regular revision of the budget but was
unsuccessful in persuading politicians. He resigned in the summer of 1968
just before the compilation of the revised budget for that year. Estimation
results that the value of coefficient on I68 is reduced somewhat between
original and revised budget is consistent with this episode, although formal
statistical significance of the difference is somewhat marginal.
In the Figures II and III, relative power of the two ministries, MOC and
MOF, are plotted for original and revised budgets, respectively. It could
easily be seen that both "fiscal restraint campaign of 1968" and "oil shock"
have substantially increased the relative power of Ministry of Finance while
the combination of "Nixon shock" and "American pressure of 76-78" with stabiliza-
tion needs raised the MOC power appreciably. Relatively strong power of MOC
in early sixties is related to mushrooming ownership for cars in the extremely
high growth environment. The instance of MOC power accerelation in 1971 and
1972 is generally associated with ex-Prime Minister Kakuei Tanaka who effectively
used these internal and external pressure to push his plan of "Reconstruction of
Japanese Achipelago".
Other implications from the estimation results are as follows;
(1) Autoregressive coefficient is increased substantially in the revised budget.
This is quite an interesting and plausible result. It implies that a part of
incremental nature of the budget, or the sluggishness of bureaucracy is hidden
-42
-
Figure II. Estimated Relative Power of the Two Ministries (Original Budget)
-43
-
Figure III. Estimated Relative Power of the Two Ministries (Revised Budget)
-44-
in the original compilation of the budget. Equivalently, the effects of
exogeneous shocks or interventions decline as the cooling off period is prolonged.
Since the revised budget is a sum of the original budget and the subsequent
modifications, the hypothesis of longer cooling off period is consistent with
the result.
Hidden sluggishness seems particularly notable when the interventions or
disturbances are negative. That is, MOF takes full advantage of such
interventions or disturbances at the compilation of original budget but as
it becomes apparent that revenues are more abundant than expected, it distributes
financial resources according to the incremental rule plus the demand of MOC.
This is reflected in the smaller coefficients for I68 and Ioil and larger
coefficient for logx , for revised over original budgets. The extra leverage
"oil shock" on the original budget is partially explained by the timing since
the "shock" started right in the middle of the compilation of 1974 original
budget.
(2) The constant term appears significant in the revised budget and not in
the original budget. Although the magnitude is rather small, this is consistent
with the hidden sluggishness or longer cooling off period hypothesis. We
might interpret this significance of constant term as an indication of hidden
stability as against hidden sluggishness of bureaucracy. Of course, the
constant term in linear regression is a catch-all item and the above inter-
pretation should be considered as a possible explanation.
(3) For positive interventions, the picture is somewhat different. Of course,
it is not easy to distinguish internal pressure (stabilization needs) from
external pressure since the two are correlated. Domestic economic slump which
generates domestic pressure normally leads to the accumulation of foreign
reserves and/or revaluation of yen, thus, causing external pressure to mount.
However, estimation results roughly indicate the following.
-45-
( i ) Despite the general hidden sluggishness, the budget revision seems to
take full account of mounting external pressure. This is reflected in the
result that the value of coefficient on EXP for the revised budget is sub-
stantially larger than that on EXP for the original budget. The budget
revision of 1971 and 1972 are typical of such examples. It is interesting
to note that both of those years are associated with ex-Prime Minister Kakuei
Tanaka and ex-Vice Minister of Finance Hideyuki Aizawa.
(ii) During the 1976-78 period, the U.S. pressure on the exchange rate and
trade liberalization mounted. Since the domestic economy was also in a slump,
the two pressure were combined to produce extra leverage on the budget.
(Significance of variables, EXP *I76-78 and EXP *I76-78)
To summarize, the impacts of positive interventions, particularly those
associated with external pressure tends to be accentuated in the budget revision.
The extra leverage in the revision seems to be stronger when internal and
external pressure are combined with shrewd political maneuvering.
(4) In the original budget, the lag of demand variable becomes somewhat
significant, implying that it is not only the growth rate of numbers of cars
owned but also the rate of acceralation that increase the relative power of
MOC.
All of the above interpretations needs to be interpreted cautiously since
there are only twenty or nineteen data points, and since the degrees of freedom
is relatively small. Also, some of the exogeneous variables are correlated
with each other rendering the estimates of some of the coefficients unstable.
-46
-
Table I
Alternative Versions for Original Budget (1959-78)
constant Autoregressive
part logx logx logSTA logSTA EXP EXP *I76-78 I68 Ioil R D.W. S
(IV-6)
(IV-6)'
(TI-1)
(TI-2)
(TI-3)
(TI-4)
(TI-5)
(TI-6)
(TI-7)
-
-
-
-
-
-
-
-
0.0067 (0.43)
0.2732
(2.53)
0.2551 (2.49)
0.2535
(2.39)
0.3769 (2.56)
0.3093
(3.28)
0.3799 (3.43)
0.3627
(3.06)
0.3722 (3.58)
0.2790
(2.57)
0.0854
(1.45) 0.1337
(2.44) 0.1233
(2.12) 0.1165
(2.31) 0.3472
(3.12) 0.2675
(2.08) 0.3057
(2.29) 0.2736
(2.23) 0.0600
(0.73)
-
-
-
- -0.2423
(2.13) -0.1669
(1.26) -0.2555
(1.93) -0.1836
(1.45)
-
-0.6381(2.59)
-0.6229
(2.83) -0.5830
(1.78) -0.7178
(3.32) -0.7208
(3.29) -0.7647
(3.68) -0.7262
(2.66) -0.7019
(3.01) -0.6082
(2.35)
-
-
- 0.4933
(2.20)
- 0.3540
(1.24) 0.1426
(0.51) 0.3162
(1.16)
-
0.0768(1.26) 0.324
(0.53) 0.0902
(1.57) 0.0381
(0.65) 0.0644
(0.95)
- 0.3094 (2.24) 0.2778 (1.84) 0.3699 (2.90) 0.2818 (2.23) 0.3324 (2.57) 0.2921 (2.12) -
-0.1394(3.74)
-0.1481
(4.29) -0.1469
(3.86) -0.1446
(4.61) -0.1566
(4.90) -0.1512
(4.97) -0.1481
(4.37) -0.1513
(4.90) -0.1426
(3.76)
-0.1410(5.14)
-0.1481
(5.79) -0.1461
(4.79) -0.1492
(6.53) -0.1326
(5.39) -0.1387
(6.04) -0.1279
(5.02) -0.1375
(5.74) -0.1494
(4.48)
0.7575 0.7943 0.7811 0.8281 0.8254 0.8348 0.8016 0.8228 0.8354
1.88 1.85 1.99 2.10 1.53 1.83 1.91 2.02 1.97
0.0356 0.0328 0.0325 0.0297 0.0300 0.028 0.0313 0.0277 0.0354
-47
-
Table II
Alternative Versions for Revised Budget (1959-77)
constant
Autoregressive
part logx logx logSTA logSTA EXP EXP *I76-78 I68 Ioil R D.W. S
(IV-7)
(IV-7)'
(TII-1)
(TII-2)
(TII-3)
(TII-4)
(TII-5)
(TII-6)
(TII-7)
-
-
-
-
-
-
-
-
0.0503(2.32)
0.6001 (4.44)
0.5031
(3.98)
0.6505 (3.62)
0.7159
(5.00)
0.7596 (4.33)
0.5209
(3.06)
0.6199 (5.11)
0.6302
(4.03)
0.4575 (3.44)
0.3273
(2.75)
0.4280 (3.83)
0.3813
(1.92)
0.1761 (1.26)
0.2359
(1.19)
0.4352 (2.46)
0.2753
(2.33)
0.2879 (1.62)
0.2053
(1.75)
-
- -0.0996
(0.38)
- -0.1000
(0.41) -0.0227
(0.10)
- -0.0221
(0.10)
-
-0.6432(2.03)
-0.4813
(1.66) -0.5738
(1.65) -0.5788
(1.86) -0.5554
(1.68) -0.4256
(1.29) -0.4091
(1.54) -0.4049
(1.32) -0.5874
(2.14)
-
-
- 0.6448
(1.80) 0.6450
(1.87)
- 0.6564
(2.33) 0.6564
(2.21)
-
0.2711
(3.05) 0.2356
(3.01) 0.2884
(3.01) 0.2797
(3.32) 0.2885
(3.23) 0.2470
(2.84) 0.2444
(3.44) 0.2466
(3.24) 0.2185
(2.72)
- 0.9574
(2.34) -
-
-
0.9553
(2.16) 0.9739
(2.72) 0.9680
(2.61) -
-0.1147(2.01)
-0.1247
(2.64) -0.1215
(2.12) -0.1147
(2.28) -0.1199
(2.28) -0.1277
(2.58) -0.1250
(2.95) -0.1261
(2.84) -0.1285
(2.73)
-0.0726(1.57)
-0.0614
(1.48) -0.0792
(1.59) -0.0764
(1.72) -0.0810
(1.70) -0.0647
(1.46) -0.0653
(1.81) -0.0664
(1.69) -0.0938
(2.28)
0.5517 0.6137 0.5255 0.6043 0.5733 0.5827 0.6873 0.6531 0.6089
1.51 1.88 1.51 1.67 1.71 1.85 2.06 2.07 1.81
0.0515 0.0446 0.0526 0.0477 0.0475 0.0456 0.0394 0.0394 0.0447
-48-
V. Concluding Remarks
In this paper, we have argued that a dynamic decision making process of
the government can be formalized by a "state space" frameword. What is im-
portant here is the distinction between observables and unobservables. The
"structure" is always a theoretical construct which is unobservable. Also, if
we allow for innovation or learning processes, such an unobservable structure
needs to be cast in a dynamic context. It is here, that the concept of dynamic
and unobservable "state" in the engineering literature can be usefully applied
to social sciences.
In both economics and political science, the "structure" or "state" has
traditionally been assumed to be either invariant or left unspecified. In the
analysis of interaction between economy and polity, however, we feel it is
necessary to introduce the concept of dynamic structure since political inter-
ventions in the economy often involve structural changes that remain in the
memory of the system. Enactment of laws and regulations that affect the
functioning of the economic system is an example of such interventions.
We have modeled a budgetary decision making process for public road con-
struction as a two actors bureaucratic politics model a la Allison. The pro-
cess is viewed as a "relative power" game between the Ministries of Construction
and Finance with political and other interventions from outside. Although the
model reduces to a simple reduced form equation, it seems to capture the essense
of bureaucratic politics in the Japanese context. Stabilization needs and
external pressure (though exchange market) are assumed to be two major leverage
for outside interventions.
-49-
The empirical results generally confirm the model specification although
the correlation among exogeneous variables and limited numbers of observations
require a cautious interpretation. The results indicate the effectiveness of
outside intervention, or openness of the bureaucratic decision making process.
However, effectiveness of interventions is generally reduced as the original
budget is revised in the fall, implying the hidden stability and sluggishness
of bureaucratic standard operating procedure, except for positive interventions
associated with external pressure. Both the "Nixon" shock and the American
pressure of 1976-78 had substantial leverage in increasing public road con-
struction particularly through the revision of budget.
Although the model presented in this paper is extremely simple, even this
level of sophistication may be too excessive for the numbers of observations
available. The empirical results of this paper, however, seem to warrant a
promising future for the application of dynamic structural model, or state-
space model, for the analysis of politico-economic decision making.
-50-
Appendix I General Dynamic Incremental Model
a) Structural State Model and Corresponding Reduced Form
Given the discussions on the linkage between dynamic structuralism and the
engineering state-space concept in the text, we treat the incremental factor to
be a structural state to govern the social system output G . In so doing, we
first hypothesize the linear discrete dynamic model for the log-natural value
of g in the following way:
g = φ (L) g + φ (L) v (A-1)
where φ (L), φ (L) : the lag polynomial transfer functions,1)
v : the random distarbance to perturbate the log-natural incremental factor (assumed: v ~ N(O, Q)).
Note that since (IV-1) is of a multiplicative form, the structural equation
was conveniently expressed in terms of log g rather than g itself in avoiding
the expected non-linearity of the reduced system output equation. In other
words, our dynamic assumption on g is of the form:
g = ∏ ∏exp(β v ). (A-2)
If G in (IV-1) is an observable variable, the unobservable structural
state g in either (A-l) or (A-2) can be derived given its initial state
1) L is the lag operator (or known as the inverse of the "z" operator, the discrete counterpart of the Laplace operator) to be defined as: L x x .
i
Δ =
-51-
which consequently means that (IV-1) is the definitional process to calculate
g . Taking the same reason employed for proposition of the multiplicative
structural state model (A-2), we should not take the additive observational
model like (IV-1) but adopt the following multiplicative (or log-linear)one:
log G =log G + log g + e . (A-3)
Substituting (A-l) into (A-3), we now obtain the reduced form as:
log G =log G + φ (L)log g + φ (L)v + e
Letting log G Y , log g G , the above reduced form becomes a typical
univariate polynomial structural form (PSF)2)3)as below:
Y =φ (L)
G + φ (L)
v +1
e . (A-4) 1-L 1-L 1-L
It is worth nothing that in the reduced form (A-4) the unobservable
incremental factor or state variable g is hidden below an observational level.
2) On an PSF model, see Wall, K., The Rational Distributed Lag Structural Form- A General Economic Model, NBER Working Paper No. 65, 1974.
3) By definition, G = e , in which both terms are equally functions on time. Hence, G = e , whose long-natural form is G = Y .
Δ =
Δ=
-52-
b) Exogenous Intervention to Influence the Incremental Factor
As shown in (A-2), our dynamic incremental model is a log-linear
autoregressive and moving average (ARMA) model which involves no exogenous
input. It is quite difficult to conceive G in (A-4) as the exogenous input.
It is a constant that appears in autoregressive transfer function (i.e., homo-
geneous different equation). Explicit incorporation of exogenous intervention
to change dynamic evolution of the incrementalism requires the following system
equations:
x = φ (L)x + φ (L)v + φ (L)p
(A-5)
y = y + x + e
where x g
p : a set of exogenous interventions.
The reduced form for this system is readily obtained as:
y = φ (L)
x +φ (L)
p +φ (L)
v +1
e . (A-6) 1-L 1-L 1-L 1-L
Fig. IV-1 illustrates a conceptual scheme of the above argument.
Δ =
-53-
Fig. IV-1
-54-
References
1. G. Allison, Essence of Decision, Little, Brown and Company, Bostom, Mass., 1971.
2. A. Ando and F. Modigliani, "The 'Life Cycle' Hypotheses of Saving: Ag-
gregate Implications and Tests," American Economic Review, March, 1963. 3. M. Athans, "The Importance of Kalman Filtering Methods for Economic Systems,"
Annals of Economic and Social Management, Vol. 3, No. 1, 1974, pp. 49-64. 4. R. J. Barro, "Unanticipated Money, Output, and the Price Level in the United
States," Journal of Political Economy, Vol. 86, No. 4, 1978. 5. R. J. Barro, "Rational Expectations and the Role of Monetary Policy,"
Journal of Monetary Economics, Vol. 2, 1976. 6. W. J. Banmol, Business Beharior, Valne and Growth, Macmillan, 1959. 7. H. M. Blalock, Social Statistics, McGraw-Hill, N. Y., 1960. 8. O. J. Blanchard, "Monetary and Fiscal Policy Under Rational Expectations,"
June 1978 (Preliminary), revised August 1978. 9. A. S. Blinder and R. M. Solow, "Does Fiscal Policy Matter?" Journal of
Public Economics, Vol. 2, 1973.
10. G. E. Box and G. M. Jenkins, Time Series Analysis: Forecasting and Con- trol, Holden-Day, San Francisco, Calif., 1976 (Revised Edition).
11. G. E. Box and G. C. Tiao, "Intervention Analysis with Application to Eco-
nomic and Environmental Problems," Journal of the American Statistical Association, March 1975, pp. 70-79.
12. M. Brown ed., The Theory and Empirical Analysis of Production, National
Bureau of Economic Research, Columbia University Press, 1967.
13. R. D. Brunner and G. D. Brewer, Organized Complexity, The Free Press, N. Y., 1971.
14. J. C. Cambell, Contemporary Japanese Budget Politics, University of Cali-
fornia Press, California, 1977.
15. A. Campbell, P. E. Converse, W. E. Miller and D. E. Stokes, The American Voter, John Wiley & Sons, Inc., N. Y., 1967.
16. L. Carroll, Alice's Adventures in Wonderland & Through the Looking-Glass,
The New American Library of World Literature, Inc., N. Y., 1960.
17. N. Choucri and R. C. North, Nations in Conflict: National Growth and International Violence, W. H. Freeman and Company, San Francisco, 1974.
18. F. Cortes, A. Przeworski and J. Sprague, Systems Analysis for Social
Scientists, John-Wiley & Sons, Inc., N. Y., 1974.
-55-
19. J. P. Crecine, Governmental Problem-Sloving, A Computer Simulation of Municipal Budgeting, Rand McNally & Company, Chicago, 1969.
20. R. M. Cyert and J. G. March, A Behavioral Theory of the Firm, Prentice-
Hall Inc., Englewood Cliffs, 1963.
21. R. Dahrendorf, Class and Class Conflict in Industrial Society, Stanford University Press, Stanford Calif., 1959.
22. R. Dornbusch and S. Fisher, Macroeconomics, McGraw-Hill, Inc., 1978.
23. J. S. Duesenberry, Income, Saving and the Theory of Consumer Behavior, Harvard University Press, 1959.
24. D. Easton ed., Varieties of Political Theory, Prentice-Hall, Inc., N. J.,
1966.
25. S. Fischer, "Long-Term Contracts, Rational Expectations, and the Optimal Money Supply Rule," Journal of Political Economy, Vol. 85, No. 1, 1977.
26. S. Fischer, "On Activist Monetary Policy with Rational Expectations,"
September 1978. (Prepared for the NBER Conference on Rational Expectations and Economic Policy, Bald Peak, New Hampshire, October 13-14, 1978.)
27. J. W. Forrester, World Dynamics, Wright-Allen Press, Camb., Mass., 1971.
28. M. Friedman, A Theory of the Consumption Function, Princeton University Press, 1957.
29. M. Friedman, "Nobel Lecture: Inflation and Unemployment," Journal of
Political Economy, Vol. 85, No. 3, 1977.
30. B. M. Friedman, "Optimal Expectations and the Extreme Information Assump- tions of 'Rational Expectations' Macromodels," (Mimeo), May,1977.
31. B. M. Friedman, "Stability and Rationality in Models of Hyperinflation,"
International Economic Review, Vol. 19, No. 1, Feb. 1978.
32. K. Furuta, Senkei-shistemu-no-Kansoku-to-dotei, (Measurement and Identi- fication of Linear Dynamical System), Korona-sha, Ltd., Tokyo, 1976. (Japanese)
33. J. Gillespie, D. Zinnes, G. S. Tahim, P. A. Schrodt and R. M. Robinson,
"An Optimal Control Model of Arms Races," American Political Science Review, Vol. LXXI, No. 1, March 1977, pp. 226-244.
34. E. N. Hayes and Tanya Hayes, eds., Claude Levi-Strauss; The Anthropologist
as Hero, M. I. T. Press, Camb., Mass., 1970.
35. James M. Henderson and Richard E. Quandt, Microeconomic Theory, A Mathe- matical Approach, McGraw-Hill, Inc., 1958.
36. D. A. Hibbs, Jr., "On Analyzing the Effects of Policy Interventions: Box-
Jenkins and Box-Tiao vs. Structural Equation Models," Working Paper, Center for International Studies, M. I. T., Cambl., Mass., December 1974.
-56-
37. H. S. Honthakker and L. D. Taylor, Consumer Demand in the United States, 1929-1970 Analysis and Projections, Harvard Unitersity Press, 1966.
38. F. W. Hoole, Politics and Budgeting in the World Health Organization,
Indiana University Press, 1976.
39. F. W. Hoole, L. J. Brian and J. T. Harvey, "Incremental Budgeting and International Organizations, "American Journal of Political Science, 1976.
40. R. E. Kalman, "A New Approach to Linear Filtering and Prediction Problems,"
Journal of Basic Engineering, March 1960.
41. R. E. Kalman and R. S. Bucy, "New Results in Linear Filtering and Pre- diction Theory," Journal of Basic Engineering, March 1961.
42. J. G. Kemeny, A Philosopher Looks at Science, Van Nostrand Reinhold, Compa-
ny, N. Y., 1959, pp. 244-258.
43. T. Kuhn, Sturcture of Scientiffic Revolutions, The University of Chicago Press, 2nd edition, 1970.
44. A. Leijonhufvud, On Keynesian Economics and the Economics of Keynes, Oxford
Press, 1968.
45. Claude Levi-Strauss, Les Structures Elementaries de la Parente, University of Paris Press, Paris, 1949. (The Elementary Structures of Kinship, trans- lated by J. M. Bill et al., Beacon Press, Boston, Mass., 1969)
46. S. M. Lipset, Political Man, Anchor Brocks, N. Y., 1963.
47. R. E. Lucas, "Expectations and the Neutrality on Money," Journal of Eco- nomic Theory, Vol. 4, 1972.
48. R. E. Lucas, "Some International Evidence on Output-Inflation Tradeoffs,"
The American Economic Review, Vol. 63, No. 3, June 1973.
49. D. G. Luenberger, Optimization by Vector Space Method, John Wiley & Sons, Inc., N. Y., 1969, chapter 4.
50. C. B. McGuire and Roy Radner, Decision and Organization, North-Holland pub.
Co., 1972.
51. N. Miyatake, "An Application of Kalman Filtering for Economic Forecasting," Operations Research, Nov. 1977. (Japanese)
52. F. Modigliani, "The Monetarist Controversy or, Should We Forsake Stabili-
zation Policies," The American Economic Review, Vol. 67, No. 2, March 1977.
53. Y. Noguchi, Y. Nhimura, H. Uchimura and K. Habamura, "Analysis of Budgetary Decision Making," Economic Analysis 66, Economic Planning Agency, March 1977.
54. W. D. Nordhaus, "World Dynamics: Measurement Without Data," The Economic
Journal Vol. 83, No. 332, December 1973.
-57-
55. A. Przeworski, and G. A. D. Soares, "Theories in Search of a Curve: A Contextual Interpretation of Left Vote," American Political Science Review, Vol. 65, 1971, pp. 51-68.
56. E. Sakakibara, T. Yakushiji et al., Politico-Economic Analysis of Road
Budget, Economic Analysis 74, Economic Planning Agency, April 1979.
57. J. Salomon, Science and Politics, M. I. T. Press, Mass., 1973, p. 21.
58. Paul A. Samuelson, Foundations of Economic Analysis, Harvard University Press, Cambridge, Mass., 1948(A).
59. Paul A. Samuelson, "Consumption Theory in Terms of Revealed Preference,"
Economica, Vol. 15, No. 5, 1948(B).
60. T. J. Sargent, "Rational Expectations; the Real Rate of Interest, and Natural Rate of Unemployment," Brookings Papers on Economic Activity 2, 1973.
61. T. J. Sargent, "The Observational Equivalence of Natural and Unnatural
Rate Theories of Macroeconomics," Journal of Political Economy, Vol. 84, No. 3, 1976.
62. T. J. Sargent, "Rational Expectations, Econometric Exogeneity, and Con-
sumption," Journal of Political Economy, Vol. 86, No. 4, 1978.
63. R. J. Shiller, "Rational Expectations and the Dynamic Structure of Macro- economic Models," Journal of Monetary Economics 4, 1978, pp. 1-44.
64. H. A. Simon, "Decision-Making in Economics," American Economic Review,
1959.
65. H. A. Simon, Administrative Behavior, The Free Press, N. Y., 1965.
66. H. A. Simon, The Sciences of the Artificial, M. I. T. Press, Camb., Mass., 1969.
67. P. Smoker, "Fear in the Arms Race: A Mathematical Study," J. N. Rosenau
ed., International Politics and Foreign Policy, The Free Press, 1969, pp. 573-580.
68. Statistical Package for the Social Sciences, Stanford University Press.
69. TROLL System, Center for Computational Research in Economics and Manage- ment, M. I. T. and The National Bureau of Economic Research.
70. J. Wanat, "Bases of Budgetary Incrementalism," American Political Science
Review, 1974.
71. A. Wildavaky, The politics of the Budgetary Process, Little Brown and Company, Inc., Boston, 1964.
72. T. Yakushiji, Dynamics of Policy Interventions, Unpublished Ph. D. Disser-
tation, M. I. T., 1977.
73. P. C. Young, "Discussion on Dr. Bray's Paper," Journal of the Royal Sta- tistical Society, Series A (General), Vol. 134, No. 2, 1971, pp. 220-223.