Professional Documents
Culture Documents
Simon to Economics
Author(s): Albert Ando
Source: The Scandinavian Journal of Economics, Vol. 81, No. 1 (1979), pp. 83-93
Published by: Wiley on behalf of The Scandinavian Journal of Economics
Stable URL: https://www.jstor.org/stable/3439460
Accessed: 11-05-2019 14:00 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
The Scandinavian Journal of Economics, Wiley are collaborating with JSTOR to digitize,
preserve and extend access to The Scandinavian Journal of Economics
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
ON THE CONTRIBUTIONS OF
HERBERT A. SIMON TO ECONOMICS
Albert Ando
I. Introduction
The contributions of Herbert A. Simon are extremely vast and diverse, ranging
from philosophy and methodology of science, applied mathematics, through
various aspects of economics, computer science, management science, political
science, cognitive psychology to the study of human problem-solving behavior.
What makes his work truly remarkable is that he not only has had something
fundamental and original to say in almost all of these areas, but he developed
almost all of his many ideas precisely and far enough to serve as a basis for
empirical work, which he himself has undertaken on many occasions. Today,
in many of these areas, we can recognize the characteristic stamp of Simon's
work imbedded in their basic body of knowledge.
Given this vastness and diversity of Simon's contribution, it is obviously
impossible to summarize it in a few pages. Throughout his writings, however,
there runs a consistent theme: to construct a comprehensive framework for
modelling and analyzing the behavior of man and his organizations faced with
a complex environment, recognizing the limitation of his ability to comprehend,
describe, analyze and to act, while allowing for his ability to learn and to
adopt. This limitation of human abilities is crucial not only in determining
the behavior of many and his society, but also in formulating the methodology
of science becaause the scientist is also a man with limited intellectual
capacity.'
In this review of Simon's work, I shall attempt to illustrate how the applica-
tion of this theme to four quite different areas has led Simon to critical
contributions in each of these topics. The first is the problem of describing
the goals and criteria of the decision maker in some approximate manner
which makes it possible to solve the decision problem. The second is a discus-
sion of the relationship between the dynamic adjustment processes in economics
and the study of the limits of man's capacity to adapt to complex changes in
1 In his Karl Taylor Compton Lectures at MIT, which was later published as The Science8
of the Artificial, Simon stated his basic view of human behavior as a hypothesis: "A man,
viewed as a behavioral system, is quite simple. The apparent complexity of his behavior
over time is largely a reflection of the complexity of the environment in which he finds
himself" (p. 25).
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
84 A. Ando
In the well-known work by Simon and his coauthors, a solution to the problem
of scheduling production, the work force, and inventories to meet uncertain
and fluctuating demand over time is decsribed in detail.2 For an abstract
theorist, the problem is a fairly simple one. Information about sales for future
periods can be summarized as a joint probability distribution. The cost
function is assumed to be known, however complex. Thus the problem becomes
one of dynamic programming, in which the total expected cost is minimized
subject to the probability distribution of sales, the relation between production
and the work force, and the identity among sales, inventory and production.
The theorist can then write down a set of conditions under which a solution
to this problem would exist, and perhaps characterize the nature of the solu-
tion.
But the manager who is faced with the problem of deciding how many
workers to hire and at what level production should be maintained for the
initial period must go beyond the mere existence and qualitative charac-
terization of the solution. He needs a concrete decision rule which translates
the initial conditions and information about future sales into employment
and production with reasonable effort within reasonable time. In general, the
dynamic programming problem arising from the situation of this sort involves,
at best, such an enormous amount of computation that it cannot easily be
solved for all practical purposes, even with the aid of today's most powerful
computer, or any computer that may become available in the foreseeable
future.
1 The work of Simon best known to economists, that on the logic of rational decis
and of heuristic decision making, sometimes referred to as the theory of bounded
rationality, is summarized in the companion review article by William Baumol. Probably
the most ambitious undertaking by Simon to model human problem solving behavior is
reported in his monumental book with Allen Newell entitled Human Problem Solving.
Although his findings reported in this book are of major potential importance to
economists, they are at such a micro level that most economists are likely to find them
somewhat remote from their everyday concern. See also Simon (1976), listed in the Biblio-
graphy.
2 Holt, Modigliani, Muth & Simon (1960) and (1956), listed in the Bibliography.
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
On the contributions of Herbert A. Simon to economics 85
In this situation, Simon saw that, if the cost function can be approximated
by a quadratic function, the probability distribution of the future sales can be
replaced by its first moments and the problem can be enormously simplified.
This proposition has become known as the "Certainty Equivalent Theorem".
Given this particular feature of the problem, the remaining problem is whether
or not some quadratic function can serve as a satisfactory approximation of
the true cost function. Since the true cost function is not known with
certainty, the manager and researchers must answer this question through a
variety of informal analysis.
Once the certainty equivalent theorem was understood and the decision
was made that a particular quadratic function is a "satisfactory" approxima-
tion, then the minimization of this "approximate" cost subject to the con-
straints mentioned earlier can be carried out relatively easily. The desired
decision rule can then be formulated.
In this example, the final result appears as though it is an application of
the maximizing principle, but in a very critical way it is not. We do not
know how good the quadratic approximation to cost really is. We do not even
know that a particular quadratic function chosen by the team of researchers
is the "best" among all possible quadratic functions. Hence, in spite of a
maximization technique used to complete the solution, the solution described
by Holt, Modigliani, Muth and Simon is not a solution ot the maximization
problem in the sense of classical economics, but a "satisficing" solution.
Hence, Simon and his colleagues stressed the importance of continually
checking how satisfactorily the decision rule is performing in each practical
application of this procedure.
The crucial point of this example, however, is that the decision maker
could not have obtained a straight maximizing solution even if he tried.
First, he would not have been able to write out the exact cost function and
the exact description of constraints (probability distribution of all future sales).
Secondly, even if he succeeded in writing out the "true" problem, the
maximization of the "true" problem would have been beyond his computa-
tional capacity however augmented by, say, the most powerful computer that
may become available in the foreseeable future.
Hence, in order either to describe the behavior of a decision maker faced
with the problem like the one in the above example, or to assist him in im-
proving his performance, the researcher must understand, in addition to the
formal theory of maximization proper (the choice among known alternatives),
the process by which alternative decisions to be considered are generated as
well as the characteristics of efficient and feasible computational procedure
involved. This is one aspect of the basic message of Simon, one that most
economists, particularly those who have attempted to serve and advise
businessmen and government officials, can readily appreciate.
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
86 A. Ando
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
On the contributions of Herbert A. Simon to economics 87
1 See Simon (1952), Simon (1953), Simon & Ando (1961), and Simon (1962), listed in the
Bibliography. For another, somewhat related idea on causality among economic rela-
tionships, see Wold (1949).
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
88 A. Ando
group. Simon initially considered the limiting case in which weaker rela-
tionships among elementary variables are exactly zero's. This led him to a
definition of causality among variables in scientific discussion and a reinter-
pretation of identification conditions in simultaneous equations estimation
procedures.
A few years later, Simon took the critical step of replacing zero's for
relations between any pair of elementary variables belonging to two separate
subsystems by epsilon's (8's) representing a very small interaction between
them compared with interactions among variables belonging to the same
subsystem. He also understood more explicitly the system under considera-
tion as a dynamic one.
Let us call the earlier, exactly decomposable system A, and the second, only
approximately decomposable system B. Simon was then able to prove several
propositions, the most important of which are the following: (1) Provided that
the e's are small enough, if the subsystems of A are stable, then the subsystems
of B are also stable, and the subsystems of B will reach internal equilibria
which are approximately the same as those of A in the short run; and (2) over
a much longer period, subsystems of B will interact with each other and
gradually approach the overall equilibrium position of the whole System B,
while maintaining the internal equilibrium position of each subsystem.
These results serve as the justification for a particular way of approximately
describing the complex environment faced by human problem solvers. In the
short run (depending on the size of the 8's), we can approximate the more
complex system B by system A, and treat each subsystem as though it were
independent of the others. In the longer run (again depending on the size of
the e's), we can represent each subsystem of B by an aggregate variable and
consider relations among these aggregate variables without paying explicit
attention to the internal structure of each subsystem. We should note that this
hierarchical structure of the world need not be limited to two tiers, but can
be of many layers for a very complex system.
Hence, if man's environment is indeed capable of being described as nearly
decomposable, then, as the system becomes more and more complex, the
number of variables that the decision maker must deal with at any one
time need not necessarily increase. A greater complexity of the system can
be viewed as a larger number of layers, and hence would not necessarily
results in much greater complexity of the problem with which the decision
maker must be concerned at any one point in time.
Simon believes that man's environment does indeed have a nearly de-
composable representation most of the time, and speculates on reasons why
nature and society naturally tend to form such a structure.' While this is a
strong hypothesis, it does enable ordinary decision makers to cope with their
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
On the contributions of Herbert A. Simon to economics 89
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
90 A. Ando
1 Most of Simon's work on this subject is collected in Simon (1977); see in particular
Section I of this volume.
2 Simon (1968), listed in the Bibliography.
3Simon (1955) and Ijiri & Simon (1977), listed in the Bibliography.
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
On the contributions of Herbert A. Simon to economics 91
the observations line up very closely on the linear line with the slope of
negative unity.
That the relationship is linear and has extremely close fit in many cases is
striking enough, but the additional fact that the slope is very close to negative
unity makes this observation an obvious and striking candidate for an
empirical generalization. Yet it is not an exact relationship, and hence it
also calls for specifying some limiting, ideal set of conditions under which the
deviation of observed data from this distribution would tend to vanish. For
this purpose, Simon sought to construct an "explanation" of why the size
distribution of so many different phenomena should take this particular
form. Because the same regularity of patterns appears among so diverse
phenomena having no obvious common mechanism, Simon suggested chance
operating through the laws of probability as a plausible candidate for explain-
ing this regularity.
It was known for some time, due to the work of Yule and others, that the
simplest kind of stochastic process that will yield skew distributions described
above as the steady state distribution is based on the following assumption,
known as the strong Gibrat assumption (taking the case of the distribution
of the size of firms as our example):
Year to year changes in firm sizes are goverend by a simple Markoff process in
which the probabilities of the size changes of any specified percentage magnitudes
are independent of a firm's present absolute size.
In this strong form, it is fairly easy to point out instances in which the
actual growth behavior of firms contradict this assumption. In a series of
articles beginning in 1955, Simon was able, for purposes of deriving the Yule
distribution as the steady state distribution, to replace the above assumption
by a much weaker one in which the expected percentage change in the size
of the totality of firms in each size stratum is assumed to be independent of
stratum. This "explanation" or theory enabled Simon and his later collabora-
tor, Injiri, to look for situations in which the basic distribution would fit
particularly well, and to predict specific departures of data from the basic
distributions when the underlying assumptions are known to be violated in a
certain way.'
Simon began his work on the size distribution as that of identifying a
striking regularity of empirical observations and providing scientific explana-
1 In the case of word counts, Benoit Mandelbrot proposed an entirely different explana-
tion for its observed skew frequency. It involved the assumption that the frequencies
are determined so as to maximize the number of bits of information, in the sense of
Shannon, transmitted per symbol. It seems that Mandelbrot's assumption is open to a
number of serious doubts and in any case it cannot be generalized to phenomena other
than word counts, and therefore of no interest to economists. I mention Mendelbrot's
work here only because the controversy between Simon and Mandelbrot on the subject of
word counts was rather famous in the 1950's. See Mandelbrot (1959) and Simon (1960),
listed in the Bibliography.
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
92 A. Ando
VI. Conclusions
Simon's scientific contribution, from the point of view of economics, has been
to work out the implications of recognizing the limitation of man to
comprehend, to adapt to, and to affect his environment completely and
instantaneously, as largely assumed in classical economics.' In this sense,
Simon has not tried to replace the existing economic theory with something
alien to it, but rather he has attempted to modify and to generalize it so
that it fits better the real behavior of man and his organizations. In the
process, Simon has exhibited almost unparalleled originality, profound ana-
lytical insight, and an extraordinary ability to extract information from a vast
quantity of data. Among hundreds of items in his bibliography, there are
necessarily some duplications, but there is no item that is either trivial or
otherwise unworthy of a true scientist.
Before him, J. M. Keynes introduced into economics an explicit recogni-
tion that equilibration of markets does not take place instantaneously, that
some price and quantity responses are slow and imperfect, and thus reformu-
lated economic theory-with a macroeconomic emphasis. We can recognize
1 His name is often associated with the notion of "bounded rationality" but his contribu-
tion is far broader than this term indicates, as I have tried to indicate in this review.
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms
On the contributions of Herbert A. Simon to economics 93
some parallel between the works of these two men. Simon, however, starts his
reformulation as well as his empirical work at a much more micro level of
individuals, small groups, and organizations such as business firms. Because
at these micro levels theoretical structure is almost necessarily more tedious,
Simon's influence has been slower to penetrate the works of others than
Keynes' was. Also, Simon has not been as active in the formulation of
economic and social policy as Keynes was, and hence he is less well known
publicly.
On the other hand, Simon has paid much closer attention and devoted
a great deal more time and energy to the development of methodology and
to actual empirical work, and therefore it is much easier for serious economists
to take advantage of Simon's initiative and build on it than it was in the
case of Keynes. Simon has not shown any sign of slowing down in his serious
work, and it will be many years before we can evaluate his work in its
entirety. But even now it is already clear that his contribution has been and
will be, directly or indirectly, an important ingredient in every future
economist's knowledge. It will become more and more important and more
and more explicitly recognized as time goes on.
References
This content downloaded from 103.38.89.106 on Sat, 11 May 2019 14:00:26 UTC
All use subject to https://about.jstor.org/terms