Working PaperPDF Available

Microfoundations of Evolutionary Economics

Authors:

Abstract and Figures

This is the first draft of the first chapter of the book we are planning to publish in 2018, which has the same title of the chapter. The second chapter will be titled as "A large economic system with minimally rational agents." Chapters from 3 to 8 will be written by Masashi Morioka and Kazuhisa Taniguchi. In this first chapter I have tried to elucidate the basic structure of human intensive behaviors. It is not maximization of an objective function as the neoclassical economics assumes but a short chain of C-D transformations. A C-D transformation is a couple of Cognitive meaning and Directive meaning. If a person receives C-meaning, he or she tries to do the D-directive. A human agent is a holder of many of these chains of C-D transformations. An economy is a huge network of interactions of human behaviors. It may count trillions of people and tens of millions of firms. They transact hundreds of millions of commodities (not necessarily goods but services), although each agents has knowledge of limited number of commodities. The main objective of economics is to know how this network works. The chapter 1 is a general introduction to the book. No specific economic theory is developed. The chapter 2 gives the first rudiments of a price theory and the quantity adjustment process. Chapters 3 to 8 are more detailed studies of quantity adjustment processes. The book as a whole shows an image of economy that is totally different from the ordinary neoclassical economics. The latter is obliged to assume that economic agents are infinitely rational and a fictitious auctioneer. Without these assumptions, the general equilibrium theory has no meaning. The new result gives for the first time in economics how the huge network as big as world economy works by the actions of each human beings who are limited in three capabilities: sight, rationality and execution. Thus the book presents a revolution on how the distributed controlled system with no central planning works.
Content may be subject to copyright.
1
Microfoundations of Evolutionary Economics
Yoshinori Shiozawa
Started on Aug. 10, 2015
Completed on April 28, 2016
This is a first draft before English revision. Revised final version will be the Chapter 1
of a new book with the same title as present paper. All comments on this paper are
welcome. Comments are requested before the end of May, 2016.
Abstract
Evolutionary economics lacked theoretical foundations: no theory of value, no theory on
behavior, no proper tool of analysis. Although we had the work like Nelson and Winter
(1982), later development was quite poor. Lacking microfoundations of its own,
evolutionary economics could not be an independent economics. It pretends to criticize
neoclassical mainstream economics, but in many of its arguments it imported implicitly
or explicitly neoclassical economics’ reasoning and results. This paper intends to change
this state of evolutionary economics.
The paper is the first chapter of the book with the same title. The main part of the
present paper is to clarify the structure of the economic behavior. Before entering the
examination of the structure itself, I was obliged to argue long how our rational
capability is limited, or how widely intractable problems exist in our life, and what it
implies to economics. Bounded rationality is the basis of all evolutions of economic
categories. They include behavior, commodity, (production) technique, institutions,
organizations, systems, knowledge, and others. Because of bounded rationality, any
existing entities are not optimal at any time. This is the main reason why evolution
takes place successively and incessantly.
The core structure of human behavior is If-Then behavior. We examine in detail this
structure and show how the skill of an experienced worker is build. Any behavior is a
time transition from the detection of a sign-mark of the world to the execution of a
directive. This transition occurs in time. Consequently, analytical framework must be
process analysis. A detailed study of process analysis reveals an existence of
micro-macro loop. This explains why evolutionary economics is the unique method to
2
understand economic processes that are going on everyday. It explains also why both
methodological individualism and holism are defective. Evolutionary economics stands
on a different methodology and thus escapes from old dichotomy of individualism and
holism.
The second chapter of the book treats more classical contents such as price
determination, quantity adjustment, stationarity of the process, and others. This gives
an alternative vision how a large scale network as large as global economy can function
by the actions of men who are limited by bounded rationality, myopic sight and locally
restricted actions. Perfect rationality and information are not the cogwheels that make
economy work. The following chapters are mathematical and computational
demonstrations of the above ideas.
Table of Contents
1. Introduction
2. Ubiquity of intractable problems
2.1 Bounded rationality
2.2 Solving a problem and computing complexity
2.3 NP-hard problems or really intractable problems to solve
2.4 Some economic consequences of the ubiquity of NP-hard problems
3. Myopic agents and the structure of human behavior
3.1 Myopic nature of our perception
3.2 Üxküll's biosemiotics and human behavior
3.3 The structure of animal and human behaviors
3.4 The nature of human skilled work
4. Environment of economic activities
4.1 Importance of stationarity of economic process
4.2 What determines effectiveness of human behavior?
4.3 Loosely connected nature of the system
4.4 Subsistence Conditions
5. Methodology of analysis
5.1 Some notes on process analysis
5.2 Hierarchy of time spans and controls
5.3 Micro-macro loop and a new methodology
References
3
Section 1. Introduction
Evolutionary economics lacked theoretical foundations: no theory of value, no theory on
behavior, no proper tool of analysis, and no prove how economy works. There was only a
belief that market economy works and evolves. Although we had the work like Nelson
and Winter (1982), later development was quite poor. Lacking microfoundations of its
own, evolutionary economics could not be an independent economics. It pretends to
criticize neoclassical mainstream economics, but in many of its arguments it imported
implicitly or explicitly neoclassical economics’ reasoning and results. This chapter and
the book as a whole intend to change this state of evolutionary economics.
Evolutionary point-of-view is the best way to understand the economy and its
development. This is
the central dogma
of evolutional economics. In this chapter on
foundations of evolutionary economics, we discuss (1) why this dogma is supportable, (2)
why most of economic entities evolve, (3) what are the defects of standard (or
neoclassical) economic theories and (4) ideas to reconstruct economics in an
evolutionary way.
The central dogma of evolutionary economics can be justified in various ways. Most
conspicuous and apparent facts are that many of important entities of the economy
evolve. They can be well understood when we see them as objects that evolve. We can
cite at least seven categories of such entities: (economic) behavior, commodities,
technology (including production and design techniques), institutions, organizations,
systems (e.g. various kinds of artificial systems, including market system), and
knowledge.1
An economic entity is very complex in itself. Although it is a result of human selection,
its complexity exceeds our capacity to understand and we cannot control it completely.
Here is the possibility of evolution. A simple commodity such as a drinking cup is a fruit
of a huge set of human knowledge: knowledge on clay soil, potter's wheel, techniques of
treating clay, glaze-making, design, baking oven or kiln, knowhow of temperature
keeping and so on. On many points of production process, there are some uncontrollable
factors. The present process of cup production is a crystallization of unaccountable trials
and errors.
1 I cited four of seven categories in Shiozawa (2003). I added three others in the General
Introduction to a handbook edited by Japan Association for Evolutionary Economics
(2006). Seven categories are not listed for classification purpose. They are not exclusive
or comprehensive.
4
The seven categories show major aspects of economic entities that have a different mode
of evolutions. Economic behavior can be changed by a decision of an individual, whereas
an institution is not changed by an individual. Even if it is a simple custom, it is socially
supported or inherited. Technology is a huge network of scientific and non-scientific
knowledge. It is transmitted by apprenticeship, schools, organizations and experience.
It is partially supported by workers' skill but develops through scientific researches.
Although internet is a new system and its basic concepts are a result of human design,
the present network grew evolutionally and nobody can control it completely.
Organization is a new kind of human group that works as a purposeful entity. Evolution
of a person to organization can be compared to the transition from unicellular to
multicellular organisms. Knowledge may be created by a person but a new creation is
only possible with the support of long accumulated knowledge. It forms a common
domain different from objective and subjective world2. Openness is one of key factors for
the development of human knowledge.
Evolution of economic entities takes a widely varied form. Despite of this variety of
evolutions, we can detect three moments that are observed in any evolutions. They are
retention, mutation, and selection
. In evolutionary biology, the same moments are
termed replication, mutation, and selection. The reason why we don't use term
replication is that many economic entities are not easily replicated or copied. Retention
is more fundamental concept than replication, because some essential features must be
retained when something is replicated. However, analogy between two sciences is not
important. Economic evolution has its characteristics proper to it. Our task is to clarify
how economic entities evolve and elucidate why they evolve.
As we have hinted above, the ubiquitous nature of evolution in economy comes from the
subtle relation between
complexity
and our
capability
. In Section 2, we explain how our
capabilities are bounded and how widely intractable problems are percolating into our
life. Neoclassical economics, based on maximization principle, ignores this fact, because
maximization generally requires extremely high rationality as we will show in the
subsection 2.1. Many economists are aware of this fact, but they cannot reformulate
their framework because they cannot abandon the maximization principle. Neoclassical
economists do not know how to formulate human intentional behavior without
maximization.
2 Karl Popper called this the World Three.
5
Section 3 starts from a simple common-sense observation that we human beings are
myopic in the sense that we are short-sighted with regard to future events. We are also
myopic in the sense that we know little about the present states of different industries,
areas and activities. The third limit of our capability is the limited range of execution.
How can an animal with these three limits (bounded rationality, myopic sight and
limited execution) behave and survive in a complex world? This is the main question of
Section 3. We present a different framework of human behavior as patterns of actions or
routine behaviors. Routine behaviors occupy ninety nine percents of our behaviors but
they function only in a specific environment. It will be clarified that human behavior is
extremely different from its conception of neoclassical economics.
Section 4 gives an overview on environment of economic activities. Three important
conditions are discussed. They are stationarity of the economic process, loosely
connectedness of the system and minimum subsistence conditions. Economy is
generated by a network of human actions.
Section 5 discusses a proper method of economic analysis. In subsection 5.1, some
special features of process analysis are discussed. Economy is generated by a network of
human actions. Economy changes when behaviors change. If we state more precisely,
the macro economic process forms an environment of human actions. Then we can
observe a kind of co-evolution of macro economic process and micro behaviors of
individuals. This is the
micro-macro loop
. We give two instances of the micro-macro loop
and consider on methodological questions that it engenders.
Section 5 is a preparatory section for Chapter 2. Economy is a network of routine
behaviors conducted by myopic agents who see very small part of total economy. A great
enigma in economics is why these myopic agents can generate roughly stable economy
and adapt to the changes of the economy. To solve some parts of this enigma is the main
object of our book.
Readers who are not interested in methodological aspect of evolutionary economics can
go to Chapter 2 directly. They can read independently from this chapter. Main target of
Chapter 2 is the presentation of a new framework of value theory. As market economy is
a series of exchanges that are concluded by mutual agreement, the theory of prices or
exchange value is crucial for any concrete understanding economic process. The value
6
theory we present in Chapter 2 is one in the tradition of classical theory of value,
especially that of Ricardo. Readers will see how this classical theory of value can be
rejuvenated into modern economics which can compete with modern mathematical
version of general equilibrium theory. Chapter 2 is an introduction to all researches
which will be deployed in the subsequent chapters.
Section 2. Ubiquity of intractable problems
Humans gained a wide capacity of voluntary motions and can control their actions by
intelligence. Most of our economic actions are a result of our decision making and the
decision is based on our intelligence. Why should we think evolution of our behavior
instead of rational decision making? Here comes the question of our capacity in relation
to the difficulty of the problem we want to solve.
2.1 Bounded rationality
Take an example of the utility maximization, which is the most common situation that
many economists suppose. Let N be the number of commodities and u be the utility
function. If a positive price vector p = (p1, p2, ..., pN) and a positive budget B are given,
then the problem is formulated as
maximize u(x1,. x2, ... , xN)
under the condition that (2-1)
p1 x1 + p2 x2 + ... + pN xN B and x1, x2, ... , xN 0.
When a solution or maximizer (x1*, x2*, ... , xN*) exists, it is usually assumed that
consumers choose a basket of goods x = (x1*, x2*, ... , xN). Then we can define the
demand function by
D(p1, p2, , pN) = (x1*, x2*, ... , xN).
There exits no problem, at first glance. Few people ask how this solution is obtained. Of
course, a solution exists if utility function f has some good property such as continuity
(Weierstrass theory on bounded closed set). However, the mathematical existence and
the obtainability of a solution are quite different. As Neumann and Morgenstern (1953)
stated, a wide range of alternating-move games such as chess and game of go have the
7
property that either the first player or the second player has a winning strategy3. In
that case, the theorem can be modified to assert that the first has a strategy to win or
the second player has a strategy by which he or she does not loose (can gain the game or
lead the game to draw). This theorem can be proved as a simple exercise of symbolic
logics}. If that strategy is easily obtained, these games have no fun, because the game is
determined before we play. Mathematically a winning strategy exists but there is no
way to find it (even by using a computer). This fact makes these games most intellectual
games and gives computer scientists a challenging task to beat professional players.
We are in the same situation as above games when we want to maximize utility function
under a budget constraint. Commodities are ordinarily sold by units. If a maximal
solution (i.e. a combination of commodities) contains quantities that are not integer,
that solution is not realizable as a basket of purchase. If we restrict all solution
variables to be integer, the maximizing problem (1) with a most simple linear function u
is equivalent to a famous problem called (unbounded) knapsack problem. It is known
that this problem is NP-hard. This means that there is no algorism that can compute
the solution in a polynomial time with regard to the size N (unless P = NP).4
A simple (but not perfect) explanation why the problem requires such a long computing
time is given by restricting xi to be either 0 or 1. Then the problem (1) reduces to know
the subset of set {1, 2, ... , N} that has the maximal value satisfying the budget condition.
The set of all subsets counts 2^N. If we are to check all possibilities, it is normal that
the computer requires a computing time proportional to 2^N.
In a worst case, the computing time may require a time that is proportional to 2 raised
to power N. This is a very serious problem. For example, if the problem for less than 10
commodities is solved by a computer at one thousandth of a second (or a millisecond), a
problem which counts 80 commodities requires a computing time about 36 billion years,
which is almost the double of the time that elapsed since the Big Bang of the universe to
our time(Shiozawa, 1990, §9 and 10 or Shiozawa, 1999, Table I.). However, 80 as a
number of commodities are comparatively small if we assume to make a purchase in a
3 The theorem can be stated as follows: If G is a two-person, open, alternating game,
and determinable within a bounded number of moves, either the first or second player
has a strategy by which one can win the game whatever the other plays. Chess and go
have a possibility of a draw (no game, stalemate in the case of chess.
4 The class N and NP are defined in the next heading. The proposition P NP is the
most basic conjecture of computing complexity theory but not yet solved.
8
convenience store. A standard convenience store counts more than 1,500 items in a
shop.
It is necessary to correctly understand the meaning that the knapsack problem is
NP-hard. It does not exclude that many instances of the problem can be solved rapidly.
We have many algorithms which work for special subclasses of the knapsack problem.
For example, if all prices are the same, the maximal solution is the top M/p commodities
that have the highest utility. The combined meaning of the fundamental conjecture and
the theorem that knapsack problem is NP-hard is that there is no algorithm that solves
all instances of the problem within a polynomial time.
For a practical purpose, an approximate solution will do. Some approximation
algorithms are very rapid. George Dantzig, the founder of linear programming,
proposed an algorithm called a greedy algorithm. It is to find the most cost performing
commodities. This algorithm ends in a computing time that is proportional to the first
order of N. It is not difficult to solve the problem for an instance with N more than one
thousand. This algorithm is guaranteed to achieve at least the half of the theoretical
maximum for any given instance. We also know an approximation algorithm that has a
polynomial computing time and is guaranteed to attain the value (1-ε) m, where m is
the maximum and ε is any positive real number.
However, this does not change the point very much. We solve the maximization problem
(2-1) in view of defining demand function and for that purpose what we need is the
solution i.e. the maximizer (x1*, x2*, , xN*) and not the maximal value u(x1*, x2*, ,
xN*). Let a solution be given by an approximate computation and let it be (x1a, x2a, ,
xNa). If approximation is good enough, this may a approximate the utility value u(x1a,
x2a, , xNa) to the maximum utility value u(x1*, x2*, , xN*), but we cannot say that
the solution (x1a, x2a, , xNa) is close to (x1*, x2*, , xN*). (See Shiozawa 1999 and
2016b}.
At the very basic core of neoclassical economics, there is this problem. It ignores the fact
that human agents have a limited capacity of calculation. When it assumes that
consumers calculate, it assumes an infinite capacity for a consumer. Human being
obtained an intelligence that is incomparably greater than other animals. However it
may be great, human intelligence and other capacities are bounded and not perfect.
9
Neoclassical economists ignore this basic fact. They ignore this, either because they are
simply thinking that human capacity of computing is infinite, or because they do not
think that this raises a grave problem for their formulation. A prominent Japanese
economist once declared that he continues to assume the maximization hypothesis,
because in his opinion, economics looses all effective formulation for the behavior of
consumers, if once he abandons this hypothesis. This is a severe neglect as a scientist,
because he prefers mathematical formulation even though he knows well that it is
impossible that a consumer behaves like his formulation.
A general problem arises. H.A. Simon named it the problem of
bounded rationality
. In
the above, we examined consumers. Simon thinks that similar problem exists for
business firms. He once declared in this way: "If there is no limit to human rationality,
administrative theory would be barren. It would consist of the single precept: Always
select that alternative, among those available, which will lead to the most complete
achievement of your goals." (Simon, 1997, p.322) Simon contributed enormously to the
recognition of universal importance of bounded rationality. It really deserves a Nobel
prize for economics. However, he made two small mistakes. First, he compared
economics and management science as parallel sciences and admitted that each has its
own characteristics. As a fact of long duration, he was right. Nevertheless, by this
unnecessary concession, he renounced the chance to reconstruct (or at least propose to
reconstruct) economics on the basis of bounded rationality. Secondly, his focus on
rationality was too narrow to open a way towards formulation of a general theory of
human purposeful behaviors.
We give such a formulation in Section 4. Before attacking this problem, let us make a
detour about complex nature of our world.
2.2 Solving a problem and computing complexity
Evolution of economic behavior depends much more on intelligence than on hereditary
characteristics. One of major forces which drive to change our behavior is rational
computation. Of course, economic behavior remains within a wide range of hereditary
characteristics however they evolve enormously. In other words, evolution occurs by
economic reasons and is not determined by hereditary characteristics, although new
behaviors remain within the range of physical possibilities. Then, what are the reasons
that make evolution inevitable for almost all economic entities? To understand the true
nature of economic entities' evolution, it is necessary to take into account two conditions.
10
One is our limits of capabilities. The other is the complexity of the decision making.
There is no absolute criterion that determines something is complex or not. It depends
on our capacity. When we get computers, many of once unsolvable problems have
become solvable. Mathematics of optimization is developing everyday. Computing
capacity is expanding rapidly. Despite of all these manifest facts, it is ironical that
mathematics is also revealing that a class of "unsolvable" or "intractable" problems
exists in every corner of optimization. The class is called NP-hard. This is very
important concept in understanding the nature of complexity we encounter in the real
world. Before entering to the concept of NP-hard, we need some preparations.
A problem is a set of infinitely many instances with an integer called
size
of the instance
(there may be many different ways to measure the size of an instance). For example, a
linear equation of N unknown variables is
a11 x1 + a12 x2 + ... + a1N xN = b1
a21 x1 + a22 x2 + ... + a2N xN = b1
(2-2)
aN1 x1 + aN2 x2 + ... + aNN xN = b1.
An instance of the problem (2-2) is given, when we specifies all aij and bi. We know that
(2-2) is solvable when det
A
0, if
A
is the matrix of coefficients aij. The size of this
instance is, for example, N.
Consider an algorithm of solving (2-2). An
algorithm
is a predetermined procedure of
calculation to solve the problem. How much time does it take before we get a solution?
The computing time depends naturally on computing speed. In the computing
complexity theory, we normally count the number of elementary procedures. For
example, in the case of linear equations, we count the necessary number of four
operations (plus, minus, multiplication, and division). This number depends of course
on the algorithms and varies depending on goodness of algorithms. Take an example of
Gaussian elimination method. A standard procedure requires
{4 N^3 + 9 N^2 - 8 N}/6
operations. In this case, the computing time is given by a polynomial of the size N.
11
When we are interested to the growing speed of computing time, the highest order term
of the polynomial is only relevant. In that case, we often say that the computing time is
of order N^3 or by a mathematical abbreviation O(N^3).
Some problem can be calculated very rapidly (if we use a computer and a good
algorithm). A sorting problem is to sort any set of integers in the increasing order. This
sorting process ends by steps that are proportional to N log_2 N. This means that to sort
an instance of 10 thousand numbers requires about 23 times more steps than sorting 1
thousand numbers. Many effectively soluble problems can be solved at the order 2 or 4.
For example, multiplication of two matrices, or a system of linear equations, can be
solved in O(N^3).
Another example of rapid algorithm is linear programming, or LP. LP covers wide range
of practical problems and we can say that it is the most useful mathematical tool that is
applicable to problems of large scale.5 Classical simplex method runs normally in
polynomial time, e.g. O(N^3), but in some cases computation enters into an eternal
cycle and in some other cases it requires exponential order of time (or O(2^N)).
Karmarkar method (a variation of interior method) eliminated these troubles and it is
assured that the program runs in O(N) for any LP problem, where α is a constant
between 3 and 4. In some cases, seemingly difficult problem can be reduced to an LP
problem, which can be solved rapidly. The reduction is drastic. The classical assignment
problem is an example. With an enumeration method, computation requires N! steps of
a simple routine. Kuhn (1955), based on the works of Birkhoff and von Neumann,
proved that it can be solved as LP problem and the computation time was reduced to
O(N^3).6
However, the lesson we should learn here is not that some problems can be solved
rapidly by computers. The lesson we should learn is that there are many intractable
problems. They are
intractable
, not because there is no algorithm that solves the
problem, but because it takes too long time for the computation (many years or many
thousands of years). With the arrival of computers, study on the "goodness" of
algorithms became urgent and important. Such needs of research led to the
establishment of computational complexity theory.
5 In some cases we can solve problems with 1,000 unknowns or more.
6 Pak(2000) is a good illustration how LP works in the case of the classical assignment
problem.
12
2.3 NP-hard problems or really intractable problems to solve
Computational complexity theory
is a part of mathematics that studies questions how
complex a problem is.
Complexity
is measured in two major ways: time complexity and
space complexity. The first gives an estimate of the necessary number of operations. The
second gives an estimate of the necessary memory space, or the number of places for
arguments. We have seen that the time complexity of problem (2-2) is O(
N
3). To the
astonishment of many mathematicians, computational complexity theory revealed that
there are many intractable problems among the problems that we encounter in economy
and industry. NP-hard problem is one of them. To define this concept requires some
preparations.
A decision problem, in computation theory, is a problem that can be answered yes or no.
The class of problems
P
is the class of decision problems that has an algorithm whose
computing time is bounded by a polynomial function of the size N. In a rough
description, a problem in
P
is somehow "tractable" because we can solve it in a
polynomial time. Of course, even if a problem is soluble in polynomial time, it does not
assure that we can effectively solve the problem. If the degree of the polynomial is as
large as 6 or 7, an instance of a large size becomes difficult to solve. However, we are
here concerned with those problems which are far more difficult. Majority of computer
scientists believe that an NP-hard problem necessitates more computing time than any
polynomial order O(
NM
).
A verification problem of a decision problem is the problem to verify, if the problem is
affirmative and a solution is given (by chance for example), if it is really a solution. The
class of decision problems
NP
(meaning non-deterministic polynomial) is the one whose
verification problem can be solved in polynomial time. Note that
P
is a subclass of
NP
,
because an instance of
P
has an algorithm by which we can determine if the problem is
“yes” or noin polynomial time.
An interesting subclass of decision problems is
NP-complete
problems. A decision
problem H is NP-complete when any instance of a
NP
problem can be reduced to an
instance of H within polynomial time. It is astonishing to know that there are such
problems. In 1971 Stephen Cock proved that a problem called 3-SAT has such a property.
3-SAT is a special case of problems when we want to know if there is a truth values
which make a logical formula true. Cock's result opened a new era of computational
13
complexity theory.
After one NP-complete problem was discovered, many problems came to be known as
NP-complete. An easy way to prove it was to show that we can reduce a problem to
3-SAT problem.7 An example of NP-complete problem is the
subset sum problem
.
Suppose we are given a set of integers of
N
elements. The problem is to determine if
there exists a non-empty subset
T
such that elements of
T
sum up to zero. For example,
if
S
= {-13, -8, -4, 2, 5, 7, 19}, there exists a subset
T
= {-8, -4, 5, 7} which sums to zero.
Then the decision problem is affirmative. Evidently this is NP problem, because it is
easy to verify (in polynomial time) that -8-4+5+7 = 0. If such a subset
T
is given, the
verification ends by at most N-1 times of additions and subtractions. However, it is not
easy to determine if there is a subset whose elements sum up to zero. To answer this
problem by checking all possible subsets requires the computing time proportional to
2
N
.
When NP-complete problems were known, a new problem arose:
P
=
NP
? Since 1971
this problem has been the most challenging problem for mathematicians and computer
scientists. Many challenged the problem but no one has ever succeeded. The Clay
Mathematics Institute selected this problem as one of seven Millennium Prize Problems
(Cook 2000). It is promised that US$ 1,000,000 will be given for a person who has first
found a correct solution (i.e. to prove
P
=
NP
or show
P
NP
). Although this decision
problem is not yet solved and nobody knows how to approach the problem, majority of
researchers of this field believe that
P
NP
. Thousands of NP-complete problems
were found since 1970's, but there is no known algorithm which runs in polynomial time.
This is one of reasons why majority of researchers of this field believe that
P
NP
.
A problem is called NP-hard, when it has an associated NP-complete decision problem.
An optimization problem has usually its associated decision problem. For example,
Knapsack problem we have above examined is a maximization problem. Associated
decision problem of knapsack problem is the question: is there a 0-1 vector
x
= (
xi
) which
satisfies the constraint condition and whose total utility is higher than a given value?
We said that Knapsack problem is NP-hard. It is, because its associated decision
problem is NP-complete. In the same way, there are as many NP-hard optimization
problems as there are NP-complete decision problems which are associated to an
7 In an exact expression, this means that an instance of problem H can be reduced to an
instance of 3-SAT problem in polynomial time. We use this abbreviation from now on.
14
optimization problem. Recall that NP-complete problem is a decision problem by
definition and NP-hard problems are not necessarily decision problems. This is the
main difference between NP-complete and NP-hard problems.
One of most famous NP-hard problems is the travelling salesman problem. It is to find a
travelling route that passes all cities of a given list and requires the least cost. We
cannot say travelling salesman problem is important in real life. It is intuitively
understandable and this is the reason why it is presented so often. However, there are
many problems which we often face in real life. They are the scheduling problems.
Scheduling problems appear frequently in business and industry. A
schedule
is an
assignment of a set of personnel, machines, and other resources to a specific task or
duty on a specific interval of time. Making a schedule is a part of everyday work for a
manager.
As they appear in most varied situations, they have many variations and have different
names. For example, they are called job-shop scheduling problem, nurse scheduling
problem (or nurse rostering problem), optimal staffing problem, weighted assignment
problem, general assignment problem and others.
A job-shop scheduling problem is an optimization problem when we are given N jobs of
varying time lengths, which need to be scheduled on M identical or different machines.
Jobs may have sequence-order constraints. For example, job J2 should be placed after
the job J1 is finished. We can take as optimizing objectives various target functions: the
time span in finishing all jobs, the total cost of operating machines, the number of
machines used, the time of delivery of the finished goods, and so on. We do not enter in
the details of problem, but many problems we want to solve in a most common situation
turn out to be NP-hard.8
Although they are a common task to plan for managers, majority of scheduling
problems are NP-hard and intractable if we really want an optimal solution.
Before ending this long detour to NP-hard problems, it is necessary to add one more
remarks. It is important to know that NP-hard problem has many instances that can be
8 o discern if a given problem is really NP-hard or not is a delicate mathematical
problem. It is hard for non-specialists to tell that this problem is NP-hard and that
problem is not NP-hard. A minor modification of the problem may change NP-hard
problem to a problem which can be solved in polynomial time.
15
solved in a convenient lapse of time. As I have noted above, when I first introduced
knapsack problem, NP-hard problem does not mean that all instances cannot be solved
rapidly. On the contrary, it is known that many (or even majority) instances of a
NP-hard problem can be solved quite rapidly, even if they are of a large size. It is not
well known how computing time is dispersed. A possibility is that the computing time of
instances of the same size makes a landscape similar to the absolute value of a function
of a complex variable. Imagine a rational function defined on a complex plane. They are
finite for all points except for several poles. If the points approach to a pole the
computing time increases without limit and exceeds any predetermined one. Instances
whose computing time is less than a predetermined time will be a large area with some
holes. For a fixed maximum computing time, the holes become bigger and may cover
almost all area when the size of instances becomes bigger.
This fact has a serious consequence for the neoclassical economics. It is based on the
basic assumption that demand and supply functions exist and represents human
economic behavior. The above result implies that demand function defined on
maximization assumption cannot represent people’s demand behavior. As I have
pointed it, the computing time exceeds easily any practical scale of time when the
maximum computing time is proportional to 2 raised to N the number of commodities. A
demand function can represent economic agents’ behavior only for an extremely small
economy that counts at most a few tens of commodities.
Ubiquitous character of NP-hard problems signifies that formulating economic behavior
by a maximization principle is a bad characterization, be it a personal or organizational
one. Then, how is our intellectual behavior organized? This is the question we must pose
and solve. We will do it in the next section.
2.4 Some economic consequences of the ubiquity of NP-hard problems
NP-hard problems appear everywhere. They are really ubiquitous. Does this mean that
we should abandon rational pursuit of better solutions? By no means! In economic
situations, no exactness is required. You may not attain an optimum by computation.
Except in a very fortunate situation, you are obliged to satisfy by a non-optimal feasible
solution (a solution which satisfies all constraint conditions)9.
What matter in an economic situation are feasible solutions that you can obtain. They
9 Taking this fact more positively, H.A. Simon named it
satisficing principle
.
16
may have different values for the objective function. You can compare their values and if
you find that a solution is the best of all, it is sure you will choose this solution.10
The best solution you get is the best among feasible solutions you can compare. That
best solution may have a value which is far from the optimal value. You may not know
the optimal value. You cannot compare the solutions you obtained with the optimal
solution. Theoretically speaking, or in the eyes of god, the value of your solution may be
very bad. Your solution may give you a value that is one half of the optimal value. You
can inquire in what situation you are theoretically, but it will be a difficult
mathematical problem.
You c an continue the search for better solutions, for example by consuming more
computing time. However, you may loose a chance to get your profit by postponing your
decisions. Because of bounded rationality, any existing entities are not optimal at any
time. This is the reason why evolution takes place successively and incessantly.
Firms are always in competition. What matter for a firm are the set of solutions you
have and the sets of solutions of your competitors. Even if your firm has a solution
which attains only 51 % of the theoretical optimum, but if your competitors have
solutions which attain 49 % of the optimum, your management must be satisfied with
the present situation. If a firm finds a solution 53 % of the optimum, managers of your
firm and other competitors will become dissatisfied and will try to find a new solution.
This imaginary situation clarifies why evolution is ubiquitous in every economics
categories. The solution we have examined was formulated as decision problem. If the
solution is adopted, it gives an action of an agent. We have already seen that utility
maximization problem is NP-hard. Consumers do not behave by finding optimal
solution for their utility maximization problem. It is simply impossible. They must
behave according to some other principles, perhaps a rule of thumb and others.
Productivity of a production process is influenced by many factors. In every part of the
process, there are many planning problems. One of such problems is scheduling of
various kinds. Majority of them are NP-hard if formulated as optimization problem.
10 Solutions may have different effects on other aspects that are not taken in
consideration. If you are a manger of a firm, you cannot ignore these points. In the
above, we assumed that these side effects are all indifferent. The same remarks apply to
many later discussions, but we do not repeat the same caution.
17
Managers of the factory cannot wait until the optimal solution is obtained. They have to
continue their operations with the best knowledge they have. If they abandon
optimization, a feasible solution can often be found quite easily. Every factory manager
uses Gantt chart. Visitors to a factory can see two or three Gantt charts on a wall. They
show solutions of scheduling problems. Gantt chart has continued to be used more than
a century. It was used far before any electric computers are invented. We can construct a
Gantt chart by hands (or more exactly by hands and a brain). It does not require a
computer. Of course, a solution give by a Gantt chart is not optimal but is normally a
good and feasible solution.
Recall also that the mangers of a factory have to make more than one thousand small
decisions. This is one of the most impressive reports in the now classical Mintzberg's
book (Mintzberg 1973). Time to make decision is mangers' most demanding resource.
Recall again the mangers have always many different questions to decide. They might
be connected with each other, but normally managers have to solve them one by one.
After Goldratt's book
The Goal
(Goldratt 1984) became a best seller, many industrial
consultants preached that we should seek a global optimum, not partial optimums. In
this lies a misunderstanding, because global optimum is in most cases cannot be
attained. We should seek a global or total optimum if possible but we should also think
that how to approach to this final goal.
Complexity also intervenes in the designing of products. Don't imagine an artistic
design. Take an example from most common machines, that of a passenger car. Think of
a designing problem to incase all necessary parts in an engine room. This is a kind of
knapsack problem but much more complicated one, because there are many
supplementary constraints. In a case of knapsack problem, an item is specified only by a
weight or volume and the unique constraint was to satisfy that the total weight or
volume does not exceed a predetermined value. In the problem of encasing parts in an
engine room, the parts have 3-dimensional shape and pack them as dense as possible is
no easy problem. In addition, some parts should be kept separated, because one part
becomes too hot and the other should be kept cool. Designers must satisfy all these
complicated exigencies and find a solution. It is a difficult work even to find out a
feasible solution.
Engineers of all fields are working in a similar situation. Making something requires all
sorts of knowledge and skill. Designers of a consumer product should keep in mind all
18
physical and chemical properties of major parts and components. They should know
how the products are produced, because a design which can be easily machined
increases the productivity and by consequence lowers the cost of production. Product
engineers should also know how the product is used in the household (or in a production
site if the product is industrial one). A product should be a safe one when it was used by
children or others. It should not be too difficult to manipulate for a common person. A
good selection of various functions is important part of product concept, because some
consumers want a function and others want some others. Forms and colors must be
beautiful. Product design requires the knowledge how the used products are disposed of.
Compliance requires knowledge of laws and stipulations. All these knowledge should be
combined in order to make a good, useful, low-priced product.
Engineers talk often about optimum designing. It expresses their desire but what they
really do is improvement. Product design often starts from examining the actual model
or design. Engineers collect users' opinions or claims on it. They listen to sales people.
They care about specialists' opinions, including production engineers. Of course, they
study new possibilities that were opened by new materials and so on. Then they make a
rough concept: a new concept and new targets to achieve. They may solve many
optimization problems. They also care about balances of various parts. An optimal
solution may be replaced by a suboptimal solution, because the optimal solution of a
problem does not fit solutions to other problems. This is how evolution occurs in
products.
Many engineer-designers know that a global optimization is impossible and better
strategy for a good designing is to make good use of evolutionary techniques. A
handbook in three volumes was compiled by a special committee in The Institute of
Electrical Engineers of Japan. It is titled
Handbook of Evolutionary Technology
:
Computation and Applications
. It covers various techniques such as genetic algorithm,
machine learning and evolutionary multi-purpose optimization, and contains may
applications in various industries. As it is written in Japanese, I do not introduce it in
more detail but it represents eloquently the real nature of engineering. Evolutionary
technology is becoming indispensable tool in robotics and others.
Another important lesson that we can derive from ubiquity of complex problems is the
theoretical difficulty to know what will happen in the future. Predictability of the future
depends on a theory of the world and the capacity of computation. Even if we have a
19
perfect theory of the world, if we cannot compute the outcome, we cannot predict what
will happen. This is just the very question that Laplace posed. In the time of Laplace,
we knew only Newtonian dynamics. World movement is described in principle by a
(huge) system of differential equations. The system is normally well posed and has a
unique solution if initial conditions were given. As this is a completely deterministic
world, if the system of differential equations is solved, we can know the future without
any limit. However, as Laplace argued, there are two insurmountable obstacles that
prevent us to know the future: (1) we cannot collect all the initial conditions and (2) we
cannot solve such a big system of equations. Laplace believed that this proves the
necessity of probability theory. We cannot predict the future. We can only guess what
will happen.
However, mainstream economics totally ignores this fact and assumes extremely strong
hypothesis that we can plan what we do in the long future. At the base of mainstream
macroeconomics lies the assumption that human agents are farsighted in time.
Dynamic stochastic general equilibrium (DSGE) model is an example. It is the core of
present-day macroeconomic models either for New Classical (Real Business Cycle) or for
New Keynesian economics. DSGE models contain Ramsey model as a part of its
standard formulation. We may say that Ramsey model is one of the basic workhorse
models in macroeconomics. In this model, the representative household decides how to
distribute current income between consumption and saving. The model supposes that
the household has an intertemporal preference function with a constant rate of time
preference and maximizes its utility through time. If the situation is in a steady state
(where there is a growth but major variable remain constant), maximization may not
require perfect foresight, as the maximization problem can be solved by assuming an
"invariant" solution. This assumption reduces the problem to a simple fixed point
problem. However, if the household is once out of steady state growth path, the problem
becomes much more difficult. Ramsey model's asymptotic behaviors form a saddle point
and the convergence to a steady growth relies on the capability to know the converging
path (See for example Solow 1990). Without assuming perfect foresight for infinite long
future, stability cannot be warranted.
Section 3. Myopic agents and the structure of human behavior
We have talked much (maybe too much) about the limits of our rationality. As for limits
20
of our capacity, another problem as important as bounded rationality imposes. It is the
problem that our capacity to know what are happening now is very limited11. Our
knowledge of the world expanded tremendously after the Scientific Revolution of the
second half of the 16th and 17th century. It is enlarging rapidly even today. We may say
that the speed of gaining new knowledge is accelerating. Even though, the range we
know about the actual world is very small and narrow. We know about the beginning of
the universe but very little about what other people or firms are doing. In an economic
decision making, what matters is not the knowledge of the universe. We know very little
what is relevant to our decision making. We may say that our ignorance is much greater
than our knowledge.
3.1 Myopic nature of our perception
Development of information and communication technology (ICT) does not reduce the
degree of ignorance very much. What is necessary for a firm is the knowledge on what
competitors are doing or trying to do. Some information may be made public but most
important part is kept undercover by a wall of corporate secret. Even if there is no such
barrier, our capability to know is also very limited in space and in time. We are myopic
animals who know only a small part of the world close to our existence.
Mainstream macroeconomics assumes farsightedness in time. This is conspicuous. See,
for example, Dynamical Statistical General Equilibrium (DSGE) model. Essential core
of DSGE model is a rational agent who maximizes his or her expected utility. At the
back of this formulation lies the rational expectation. It assumes that an economics
agent knows the economic theory, can predict the far future, and make a decision after
taking in consideration all of what happens in the future. Mainstream macroeconomics
assumes farsightedness in space, too. This fact is not as apparent as farsightedness in
time, because macroeconomic is based essentially on one-good models with one
representative agent. Even when a model deals with different goods, the variety is only
an appearance. For example, Dixit-Stiglitz utility function assumes a strong
symmetricity. This makes it possible to treat different goods as if there is only a good in
the economy. If a model assumes different agents, they do not really intervene mutually.
Assuming one-good model is to assume that all agents have a complete far-sight or
11 We think that rationality and far-sightedness, i.e. the capacity to reason correctly
and the capacity to collect necessary information, are very different and it is better treat
them distinctly. H.A. Simon did not make a clear distinction between bounded
rationality and bounded sight and included two of them in a single concept of bounded
rationality.
21
capacity to gather all relevant information in the economy.
When we reflect on real life, all goods are different and hardly substitutable. Managers
of a firm can know the past series of demands for each of their product, but it is hard for
them to know the competitors' exact series of demands. At the base of mainstream
macroeconomics lies the assumption that human agents are farsighted in time and in
space. This is of course impossible. In order to make economics based on reality, it is
necessary to pose ignorance and short-sightedness at the base of our conditions.
Short-sightedness and bounded rationality are a kind of twins. No human being can
escape from these twin limits. In the next subsection, we add a third limit for our
capability. It is the limited capability to execute something. Even if we know what we
should do, our ability to do something in a certain lapse of time is limited. This third
limitation was rather well incorporated in all economics including classical and
neoclassical economics, because they assumed that there is a necessary amount of
man-hours for any production.
These three limitations are understandable if we once think that humans have evolved
from more primitive animal whose capabilities were very limited by any criterion. The
modern economics started to formulate human behavior as maximization of an objective
function and was conducted to assume unlimited rationality. There is no basis to
assume so, except that it was necessary for the neoclassical formulation. Evolutionary
economics should not be start on such absurd basis. Instead, it should start from the
opposite side. Our capacity is very limited, but we obtained step by step more elaborated
behaviors and ways of thinking. There is continuity between animal and human
behaviors. We can learn much by observing less developed animal behaviors.
3.2 Üxküll's biosemiotics and human behavior
By assuming infinite rationality and farsightedness, neoclassical economics represented
human being as an omnipotent and omniscient entity. In contrast, evolutionary
economics takes animals as an exemplary model of our behaviors. We have evolved from
animals and not from deity. Even if we have gained high capability compared to that of
animals, the gap between humans and animals is small and leaps occurred only
gradually. If we cannot observe any qualitative change, it is more natural to deem that
our capability is more close to animals than god.
22
With this regards, it is good to fix our starting point on von Üxkül's notions of "Umwelt"
and his idea of functional cycle. Jakob von Üxkül is known to have been critical of
Darwinism, but was a good animal observer. He inaugurated a theoretical biology by
asking how an animal percepts the world12. Animals have their own Umwelt, or a
surrounding world, specific to a species. For example, a dog is strongly myopic but has a
very good sense of smell. It is also partially color-blind and cannot distinguish yellow
and green. Then the world of a dog is very different from that of a human.
Von Üxkül studied lower animals such as ticks and sea urchins. They have only
undeveloped sense organs but they succeeded to survive. Egg laying behavior of a field
tick is astonishing. Ticks are blind and can feel if the world is bright or dark. They
cannot jump as fleas do. A flea can jump hundred times as high as his size. Ticks cannot
run as rapidly as spiders. This weak animal has to suck blood of a mammal before it
lays eggs. How can it succeed in this difficult task? At this point ticks are ingenious.
A female tick climbs to the tip of a tree twig with the help of her skin's sensitivity to
light. The place becomes her watch post. She waits there for a long time, even years.
She knows by the smell of butyric acid that a mammal is approaching. Butyric acid
emanates from all mammals, because sweat contains it. She blindly falls when the
smell reaches certain strength and in a very fortunate case arrives on her prey. She
knows if she was lucky enough to have caught a mammal by the temperature because
her organ is precisely sensitive to temperature. Then she searches a less hairy spot and
embeds her head in the cutaneous tissue of her prey. She can now suck a warm stream
of blood until she slowly swells many times heavier of her original weight. If she did not
succeed in catching a mammal, she has to restart her watch from the beginnings.13
The contrast between limited capabilities and the difficulty of task is impressive. As an
economic agent we are in a similar situation. Our capability is very limited, but
combining simple operations we can achieve an astonishingly complex and difficult
task.
The secret lies in the constant relation between animals and its environment. If
mammals suddenly changes to poikilotherms by some unkown reasons, or if they
suddenly stop to secrete butyric acid, ticks cannot catch mammals and lay eggs. As far
12 Now Jakob von Üxkül is thought to be the starter and pioneerof biosemiotics.
13 All this story appears in Uexküll (1992).
23
as they keep their egg-laying strategy, they are destined to extinct. This kind of
extinctions occurred many time for many species. The only fortunate animals have
succeeded to survive.
All species have specific relations with their environment and their survival depends on
these relations. Üxkül studied these relations by the concept of Umwelt. Each species
has its own world, perceptible through various senses that are proper to the species and
meaningful for its survival. Life is an eternal process of interaction between the organic
body and its environment.
Üxkül thought that an animal grasps the
world by two hands: one is a
receptor
and
the other is an
effector
. It receives mark
signs from the world, treats them in the
central nervous system and orders how to
behave. We may distinguish in this series
three functions of the world grasping:
perception, judgment and execution. As
these functions makes a cycle starting from
mark carrier to mark carrier through three function, Üxkül called the total system
functional cycle
.
It is important that these three functions are all limited in a strong way. Then we can
make a following list:
function representative fn. general scheme
(1) perception myopic sight limited information gathering
(2) judgment bounded rationality simple reflective thinking
(3) execution local effect effects in limited space and time
Table 1.1 Three functions of a functional cycle
Each species has its specific functions of different capacities and they are limited in a
different way. It is notable that Üxkül thought the information flow simply not as
physical signals but as signs which have some meaningto the animal. So the
functional cycle is not only a feedback loop in which a single-valued quantity flows. For
c, the objective world is not simple set of qualities and quantities but each object carries
Fig. 4. Functional cycle with reafferent cycle.
(Uexküll 1920 : 117.) Reproduced from Rüting
(2004.
küll 1920 : 117.)
24
a sign and animals percept and react to these signs. He was interested how the pattern
recognition works in the receptor, but it is not necessary for us to enter into such details.
Üxkül’s idea of the receptor is somewhat similar to the garbage can model for
organizational decision making (Cohen, March and Olsen, 1972). Of course, the tasks
are very different and the organizational decision making is a highly rational procedure
that requires spending various resources including information gathering and
deliberation. However, the total pattern of organizational decision making is to reduce
most complicated and diversified set of information into a predetermined set of
conclusions. In a very primitive way, ticks and sea urchins perceive the world and
classify the object for example into food, predator, sexual partner, and others.
3.3. The structure of animal and human behaviors
Now let us return to our main problem. Recall how difficult task a tick had to achieve
before she lays eggs. A tick is almost blind, cannot jump, or run fast. How can this badly
conditioned animal achieve a difficult task as catching a mammal? We now know how a
tick ingeniously solved this difficult question: by a series of patterned actions. All
animals including humans achieve a difficult task by a series of actions that are
patterned as a couple of stimulus and a response. Sociologist Tamito Yoshida (1990)
formulated this pattered behavior as C-D transformation. Here, C stands for cognitive
meaning and D directive meaning. Yoshida arrived to this formula after studying C. S.
Peirces semiotics. In Üxkül’s functional cycle, C is a sign received by the receptor and D
is a sign directed to the effector. C-D transformation can be interpreted as a conditional
directive. For example, we may interpret it as a message: if condition C is satisfied, do
D.
Similar formulation is given in evolutionary computation. John Holland, the creator of
genetic algorithm, adopted if-then rules as a simple representation of behaviors and
called this representation
classifier system
. This became the tradition in almost all
agent based simulations. Holland adopted this formula, because he was thinking to use
it in his evolutionary computation. If partand then part, or conditional and directive
parts, were expressed by a couple of binary codes of predetermined length.
Hollands classifier system is highly universal in the sense that any optimization
problem can be in principle transformed into a genetic algorithm problem for a classifier
system. Indeed, it is a question of encoding. Recall that there are two parts in classifier
25
system: conditional and directive parts. If these parts are sets of sign-marks, as Üxkül
assumed, they are finite sets and you can encode each element into a different binary
code. Then, the “optimization”, transformed in an evolutionary computation, is to search
a conditional binary code whose resultant code gives you a good value of the objective
function. However, this universality does not assure that if-then rule behavior can be a
prototype of all animal or human behaviors, because the coding correspondence may be
extremely complicated and may not have any practical meanings. For an animal
receptor, sing-marks should be as simple as it can recognize them instinctively. A
directive must be also as simple as the animal can effectively execute it. For a human
being, his or her judgment may be more complicated and a directive can be more
sophisticated, but the difference is question of degrees. A sign-mark should be within
the three limits of human agent, i.e. myopic sight, bounded rationality and local action.
In this sense, Hollands if-then rule formula is not general enough to cover all animal
and human behaviors. However, it may constitute the atoms of behaviors. Indeed, we
can rightly believe that any behavior, be it animal or human, can be decomposed to a
series of if-then behaviors.
This is a strong contention and it is difficult to prove this highly universal thesis, but
there are some circumstantial evidences for this.
In stead of presenting a formal proof, let me talk about my own experience. How have I
arrived to this idea? The story is a bit long and sinuous. It was 1985 that I really
realized that a simple utility maximization problem comprises NP-hard problem in it
(i.e. if it is reformulated as integer problems). Before that, I knew that knapsack
problem is NP-hard but I was not sure if it can be applied to utility maximization
problem. In 1985 I made my mind and started to think how human behaviors are
organized. There were some keys. H.A. Simon was proposing the satisfising principle.
This gave me some hints, but as a formulation of proto-type behavior, this was
somewhat too ambiguous. Simon and March (1958) and Cyert and March (1963)
comprised the words like routine, routine behavior, and rule based action/routine, but
there was no precise expression how these routines were structured. Routine was also
the key concept for Nelson and Winter (1982). The word routinewas a big hint for me,
but it seemed too ambiguous and unstructured. Instead of routine or routine behavior, I
adopted the Japanese expression teikei kōdō which means rule-based behavior or
patterned behavior. With this key word in mind, I strolled in various fields from
ethology to psychology to philosophical anthropology. I did not know Hollands if-then
26
formula. The word routinewas doubly indicative. It signified a routine behavior, but
also meant a small package of computer program which served as a ready-made
operational function. This reminded me a formulation of the Turing machine that I have
read in my student days.
It was Martin Daviss book
Computability and Unsolvability
(1958). He defined a Turing
machine as a set of quadruples of the form
qi Sj Sk ql
that contains no two quadruple with the same first two symbols.
If I omit the details, the quadruple meant that if you are in an internal state
qi
, observe
if the external state is
Sj
and if if it is, do
Sk
to the external world and change your
internal state to
ql
. I thought this Turing machine parable is very good for two reasons.
First, quadruple indicates the most elementary form of behaviors. Second, the fact that
a set of quadruples expresses a Turing machine indicates that a set of quadruples can
express highly structured and complicated function. All computable functions on a
computer, or recursive functions mathematically formulated, can be computed by a
Turing machine. I knew this fact when I was high school student. I was once deeply
interested in mathematical foundations or metamathematics.
After I came to think humans as a kind of Turing machine, I searched stories which
exemplify the parable. There were many of them. However, Üxkül’s tick story was most
impressive. I first used it in the last chapter of my book
The Science of the Market Oder
(in Japanese) which was published in 1990. The book was subtitled From
Anti-Equilibrium to Complexity. This was the first book in Japanese which carried the
word Complexity in titles.
Encounter with Üxkül was lucky. I did not know that he was the father of biosemiotics.
Ticks egg laying story was not only impressive for me, but it told me many things.
When I stayed one year in Cambridge, UK, 1986-87, Roberto Scazzieri taught me the
existence of Heiner (1983). It was telling that a big C-D gap (or Competence-Difficulty
gap) conditions predictable, regular behavior. This paper was enlightening. In
economics, we normally assumed optimization. When we know that optimization is
impossible, the second best method was to approximate the optimization. However, as I
have told it above, this causes various problems for the equilibrium formulation,
especially for the definition of demand functions. We had to think from the opposite
direction. We have to search how an efficient behavior can be organized when we have
27
big gap between our competence in selecting alternatives and the difficulty of the
problem. This is the way that less competent animals were successful for their survival.
Humans are much more competent and capable of more complicated calculation, but in
view of the complexity of the real world we are also in the same situation as animals. We
are not as competent as to solve any maximization problem. With this regard, we must
be acting in the same ways as animals do. This was really a revelation. In the next year,
when I visited the USA, I went to Provo, Uta, to meet Heiner as he was working for
Brigham Young University at that time.
3.4 The nature of human skilled work
Hiners thesis, Üxkül’s tick and Turing machine parable all fitted in one idea.
Combining and arranging elementary patterns of behaviors we can achieve most
complicated tasks. It was great. From that time on, I continued to search other
examples and try finding exceptions to my formula. I found many fitting examples.
Yos h id a s C-D transformation was one of them. Hollands classifier or if-then behavior
was another. Psychologists framework of Stimulus-Response formula or reflective
behavior was showing the simplest cases. Skinners operant behavior was more
complicated, but at any rate they were too simple example to use as a proof of
universality of my thesis in real economic lives. However, I found various good and
persuasive examples in Nakaokas books. Nakaoka is one of my personal teachers and
was a colleague at Osaka City University. He is a historian and philosopher of
technology.
Nakaoka (1971) was a book which investigated how the workersskill is formed and
structured. I found in this book many examples of my thesis. Examples comprised
operations of a medical team, working operations in a steel making factory using
electric furnace, and clerks administrative processes in an office of a business house. In
another book, Nakaoka cited books from classic Greece and Chinese and illustrated how
the signs in the sky or in the nature were used to inform farmers to know the good time
for specific work like sowing and cultivating. In many places, he showed that work is
decomposed into a series of simple operations and workers skill consists in the
judgment of each operation. He pointed that a judgment has a form of a symptom -> an
action to take. This was just examples of Yoshidas C-D transformations and Daviss
quadruples in a simple form.
There were of course many auxiliary questions. If a behavior accompanies a judgment,
28
how do we detect a symptom? We are conditioned by many scarcities. We have only
limited thinking or computing time. We have to determine how much of time we spend
for an activity. The same kind of scarcity applies to our attention. I reflected on my own
mental activity and observed that the target of our attention is strictly limited to one or
a small number of things. I do not know why. At any rate, this must reflect the result of
our evolution. To focus our attention on one or a small number of things must be an only
possible way to survive for animals which have much more restricted ability to judge
what is happening around them. How do we select a target or a mark to which we pay
our attention? I recalled that Simon and March (1958) used the notion definition of the
situation.” Let me cite a paragraph from it. Everything was beautifully argued:
The theory of rational choice put forth here incorporates two fundamental
characteristics: (1) Choice is always exercised with respect to a limited,
approximate, simplified modelof the real situation. We call the choosers model
his definition of the situation.(2) The elements of the definition of the situation
are not given -- that is, we do not take these as data of our theory but are
themselves the outcome of psychological and sociological processes, including the
choosers own activities and the activities of others in his environment. (Simon and
March, 1993[1958], p.160)
We find an astonishing coincidence with my Turing machine parable of animal and
human behaviors. A quadruple is divided into two parts: conditional half and directive
half. The conditional half contains two symbols:
qi
and
Sj
. What role does
qi
play? It
defines the internal state. It is an outcomeof the previous action and the environment.
It defines the definition of the situation and suggests what kind of stimuli we have to
observe. The observed result is
Sj
. If we are in the state
Sj
, we do
Sk
to outer world and
transit to the internal state
ql
. If the observed state is not
Sj
, it is understood to transit
to the next quadruple
qi’Sj
in the set.
What seems to be very difficult can be achieved once we know each elementary behavior
and the order to follow. Üxkül’s egg laying behavior of the tick can be written in the
same way in a series of quadruples. Recall that all
Sj
and
Sk
are simple and restricted
observations and actions. Nakaoka gives us many other more elaborated examples. Now
I firmly believe that human behavior if it is a very difficult one can be also decomposed
in series of simple behaviors.
29
Where do our judgment and rationality work? We have to distinguish two levels. First
level works in a specific behavior. We must judge if we are in a state
Sj
. If yes,
Sk
may
require some calculation. To state it more precisely,
Sj
may contain some parameters
and
Sk
is a simple function of those parameters. These judgments must be
instantaneous and similar to instinct of animals. Even though, this is one of essential
skills of high ranked workers. Mintzberg (1973) reports that a factory head makes more
than a thousand decisions a day.
Second level of judgment works on the behaviors themselves. We have a repertoire of
behavior patterns. They are classified with respect to the situations. In each situation,
we have several candidates of possible behaviors. If a behavior does not produced an
average result as good as we have expected, we may choose another behavior in the
repertoire. In some occasions, we increase the repertoire, by a pure invention or
learning from others. This second level judgment works mainly on observations. No
complicated computation or consideration is required. What we do is to observe and
compare the results. Each judgment lies within our capacity of our sight and rationality.
This is essentially different from maximization by calculation. Except for an imaginary
problem setting, a pursuit of a better result by a calculation is in most occasions
impossible. Instead, we observe what happens if we behave this way or another. This is
closer to natural selection than rational choice. Very few calculation and rationality are
demanded.
I deny the maximization as a principle of economics behaviors, because in many
occasions it exceeds our capacity of calculation or judgment. This does not mean that I
deny the rationality when it works. This only means that we have to reconstruct from
the very base of economics the theory of value and the theory of production, exchange
and consumption within a framework that do not violate our capacity of sight,
rationality and execution.
The concept of repertoire of behaviors helps us much to understand what is skillfulness
of a worker. We sometimes confuse dexterity with skillfulness. Of course, dexterity is a
part of skillfulness but skillfulness does not stop to the fact that a works is dexterous.
Dexterity is concerned to the quality of a behavior. A skilled worker normally has a
dexterous action of behavior. He or she has a better judging capacity and exact way of
actions. However, the skillfulness is a capability much wider than dexterity. Normally, a
skilled worker has a larger repertoire than unskilled workers.
30
In a fortunate time, a factory work is a simple repetition of routines. If you have a few
patterns of behavior, you can do your work. However, various unexpected events may
happen: power breakdown, malfunction of a machine, repetition of defective products,
lack of parts, interaction of two independent machines, defecting of a worker (because of
sickness, injury or simple absence) and so on. Some troubles happen quite frequently,
for example, one a week or two. Even a young unskilled worker can soon learn how to
deal with the state, if the trouble happens frequently. We have on the other hand very
rare events. For example, a machine may fall in a trouble which rarely occurs, say every
10 years or so. An old experienced chief of workers has knowledge how to deal with the
trouble. After K. Koike (1995), this is the core of intellectual skill of workers. He
distinguishes usual and unusual operations.
Workshop jobs include usual and unusual operations. Work on a mass-production
assembly line does not appear to be dependent on skills and seem entirely
repetitive. Only speed seems to affect efficiency. This, however, is usual operation.
Observe the line closely, and you see frequent changes and problems. Dealing with
these situations constitutes unusual operations. (Koike, 1995, p.63)
New workers of short experience have not knowhow to deal with these unusual
operations. Of course, there are gradations between usual and unusual. One operation
may be required very two months. Another operation is required once or twice in ten
years. Imagine, for example, an introduction of a new machine system when the older
machines were used five years. Workers whose job career is less than five years have no
chance to experience the works and troubles when a new machine system is installed.
Koike argued that the major part of the intellectual skill of workers is based on this
wider experience and its contribution to the efficiency is comparable to the expertise of
high learned engineers.
We have also arrived to an important conclusion. Observing what we can do and
investigating how our behaviors are organized, we find without intention how our
behavior evolves. Normally we have a pool of behaviors and we choose them, not by
rational calculation, but by observing and comparing the average result of a behavior
with other comparable behaviors. This is an evolution of our behaviors. The selection
works on the second level that we have examined above. Although we use minimal
rationality, this selection, repeated many times, produces a result that was
31
unimaginable at the beginning. This is the core mechanism of the economic evolution.
We have elucidated a principle of evolutionary economics.
The main purpose of this book is to show that a worldwide network of economic
transactions can work with these limited assumptions. However, before we go to a
concrete discussion how economics process works, it is necessary to examine in what
kind of situation our behavior can be effective. This is the task of the next section.
Section 4. Environment of economic activities
If our behavior evolves by experience and comparison, instead of rational
maximization14, our economy must have various features that permit us to behave
effectively by a behavior that is a result of long series of evolutionary selective process.
There are three major conditions: stationarity, decomposability and subsistence. The
core of all conditions is the stationarity (or stationariness) of the economic process. This
expression may induce many possible misunderstanding and I will explain this concept
in detail in subsection 4.1. The second important and even vital feature is
decomposability or loosely connectedness of our economy, on which I explain in
subsection 4.3. Before I enter to this crucial feature of the economic system, let me make
a deviation in subsection 4.2. I will argue there a question on why and when our
behavior becomes effective and when our behavior becomes ineffective. The third and
least mentioned feature is concerned to our ability to survive, because human is a being
that is restricted by bounds in all aspects: myopic sight, bounded rationality and local
execution. In the last subsection 4.4, I will argue the importance of (ample) margin of
subsistence.
4.1 Importance of stationarity of the economic process
When we speak of economic process, it may indicate any of processes from a series of
transactions in a particular market to the whole network of transactions that spread
worldwide. Whichever process we imagine, stationarity must be the most important
feature of the economic process.
Stationarity is completely different from stability. In standard economics, two kinds of
14 This is not the claim that we are irrational or behave irrationally. As we have argued
in Section 2, our capacity of calculation is limited and we are obliged to behave
differently from what is assumed by maximization principle, which was long assumed
in neoclassical economics.
32
stability are argued. The first is the stability supposed in the general equilibrium
framework. In this case, the stability means the invariance of agents behavior. In
equilibrium, agents have no incentive to change their actions (e.g bids and offers of a
brand of security). The second meaning of stability concerns the behavior or movement
of temporal equilibrium. We say that the equilibrium is stable when the economic state
shifts to a fixed state when the state is out of the equilibrium.
Stationarity means only that the concerned process has some regularity or keeps
constancy in some sense. A process is stationary, when the state of the process repeats
itself essentially in the same way. The epithet essentiallyis crucial here. In a simple
process in which only a single variable changes, the process may take a variety of
movements. The adverb essentiallymeans the variable come near to the same value
repeatedly. In a process that comprises many variables, no same state is repeated in the
sense that all variables take the same value at two different point of time. Even in that
case, we say the process is essentially stationary, when some variables repeatedly come
near to the same combination of values.
The word stationaryis used in the stochastic process theory. The term stationarity
here does not have such a specific meaning. It has much wider or looser meaning. A
stationary process in the stochastic process theory is stationary in my meaning, but we
must admit many other stochastic processes, those that are not stationary in the
stochastic process theory, are also stationary in our sense. Remind Koikes unusual
operations in the previous section. Our concept of stationarity includes unusual states
as a part of stationary process. Economics process always comprises various degrees of
unusualness.
Stationarity of this broad sense is the vital condition that a human intentional behavior
is effective15. We have argued in the previous two sections that our capacity of judgment
is strongly restricted either by information collecting and rational calculation capacities.
The effectiveness of our behavior depends much upon the evolutionary selective process.
If economic process changes much, the present behavior may not be the best one even
among the acquired repertoire of our behaviors. Our actual behavior is chosen only
because it was effective, in the past experience, for obtaining higher value of an
objective function than others. This fact remains effective only when the concerned
15 For more details of this argument and its implications to economics, see Shiozawa
(1989).
33
process did not changed much in an essential manner.
It is important to recognize that our knowledge and behavior are deeply dependent on
the stationarity of our world, or constancy of the time pattern of everyday life. A day
starts by sunrise and a night comes by sunset. Years are a repetition of spring, summer,
fall and winter. Mankind has invented to make many other rhythms for the convenience
of life: a week of seven days, a month, hours and minutes, decades and centuries, and
others. All these customs or institutions help to make rhythms and punctuation in the
everyday life. We eat breakfast, lunch and dinner in a day. Working hours starts at nine
a.m. and ends at five p.m. Firms pay wages once every week or once every month. Shops
are open six days a week except for bank holidays. You can buy your baguette at a
bakery, your macaroni and paste at a grocery, papers, notes and ball pens at a
stationary shop, and books at a book shop. An order on a web site arrives in a day or two.
You can draw your money from your bank if you have enough deposit, or if you have a
credit account. In the end of a summer, you can buy an overcoat and in spring summer
shirts. Almost all things necessary for your life are repeated constantly if they are not
exactly the same as one year ago. These are the basis of our life and without these
constancies it is very difficult for us to live. However, we easily forget this fact and
believe that you are organizing your life by your plan and calculation. This is a very
special mindset that did not exist in pre-modern worlds.
Modern economics conceived our economy through the looking glass of modern science.
Galileo Galilei succeeded to predict on calculation how a mass drops in a free fall.
Johannes Kepler succeeded to describe how planets move around in their orbits.
Pierre-Simon Laplace imagined that an omnipotent being can calculate the future state
of the world by knowing the present state. If the world is governed by Newton dynamics,
this is in principle possible, because the movement of the world can be described by a
(huge) system of differential equations and because it is well posed and has a unique
solution.
Economists, in particular after the neoclassical revolution, imagined that a human
agent behaves on a similar calculation. They supposed that an agent predicts what will
happen in the future, calculates his or her profit or utility and decides what he or she
does. A typical example is the utility maximization under a budget constraint of a
consumer. We have proved that this simple calculation requires exorbitant computing
time and it is practically impossible except for an extremely simple case of two or three
34
commodities. We should abandon this mode of thinking. Even in the case where we
really calculate or contemplate, our decision makings are helped enormously by
constant patterns of the process of events. If there are calculations, it is the objective
world that calculates and human calculation is only a small part of them. We must not
mistake this fact and believe too much on our ability to calculate and predict.
In relation to this point, it is opportune to give a few comments on G.L.S. Shackle’s
kaleidics. He was right to emphasize the uncertainty and the ignorance of the future. It
may serve as a good criticism of the rational expectation hypothesis and contribute to
refute what Davidson named ergodic axiom16. However, I have to say that Shackle and
Davidson still remain in the
problématique
of the future calculation, or
Galilei-Descartes-Newton-Laplaces world view. Galilei, Descartes, Newton and Laplace
all imagined a mechanical world if it was more dynamic than that of people in the
medieval period: complex clockwork, turbulent cosmic flow, a system of differential
equations and probability theory. They were thinking in common to predict the future
by calculation or rational inference. This is the spirit of the modern science. But, in a
complex system, it is not possible to predict what will happen in the future by
calculation or any rational inference. If we can do it, it is the very small part of the
world, which is isolated from others and composes a simple system. Computer
simulation of a world requires a computer with the same weight of the universe, if we
want to calculate the movements or interactions of all elementary particles. The
question does not change much if you think of a stochastic prediction.
Keynes and Knight were right when they argued that uncertainty excludes even the
calculation of probabilities. We are in a world of
non-stochastic
randomness (Alvarez
and Ehnts 2014). In this regard, we can say that Shackle and Davidson follow Keynes
16 Paul Davidson argues many times (at least 13 times in Davidson 1991, 1999 and)
that Samuelson postulated what Davidson named ergodic axiom. However, in all
times, he cites the same Samuelsons paper which is a reprint of Samuelson (1968).
Samuelson nowhere claimed that ergodic hypothesisis a sine qua non of economics as
scienceas Davidson argues (Davidson 1999, p.154, p.382). Samuelson only pointed that
ergodicity is necessary if the classical dichotomy works independent of initial
distribution of money. Ergodic axiom is not an axiom of neoclassical economics but
rather a scarecrow invented by Davidson. It is not exact to say that the ergodic axiom is
one of three axioms that Keynes rejected. It may be implied from his idea but Keynes
had no clear idea of the axiom. In addition, Davidson’s concept of ergodicity does not
exactly correspond to physics concept of ergodicity. Alvarez and Ehnts (2014) reasonably
propose to use terms stochastic and non-stochastic randomness instead of ergodicity
which has an ambiguous meaning in economics.
35
closely and loyally. However, we have to say that Keynes was not free from the future
calculation, or world calculation problématique. If we really acknowledge that our
capabilities are extremely limited, we must think from the opposite end. Let us imagine
a lower animal with little reasoning power, Üxkül’s tick for example. The tick does not
calculate or predict what will happen. She waits until the world changes to a state that
the inner state dictates. It is not the tick which calculates. It is the world which evolves
by itself. The tick at the tip of a branch waits until she smells butyric acid. She catches
the sign-mark of the world. A sign-mark is a symptom of the world and it is usually a
special feature of a small part of the world. Even a lowest animal has some power to
detect a sign-mark and deploys its series of C-D transformations. In the Turing machine
parable, if the state is in Sj, we try to realize Sk. Both Sj and Sk are but two small
marks of the world. The effectiveness of behavior does not depend on our rational power
of prediction. It depends on the sequential constancy of the result that follows a
combination (Sj, Sk). Through a long history of evolution, a tick has discovered an
ingenious tactics to catch a mammal. A man or a woman is not very different from a tick,
a flea, or a spider. He or she mainly behaves just like the lower animals do: detect a
sign-mark of the world and add a small effort to change it.
The most important target of economics is to explain how the economy that spreads
world-wide works. It is not our capacity of calculation and prediction that warrants the
well functioning of an economy. It is the mode of interactions that warrant it. A
fundamental change of paradigm is required. We need a new paradigm on how the
complex world work and what we are competing in this difficult environment. Keynes
and Shackle were not very insightful to deal with this difficult task.
This is not to deny the modern sciences. Physical science from Galilei and Descartes to
Newton and Laplace enlightened our understanding of our world. What is required in
economics and human science in general is to acknowledge how our behavior is
organized and why they are in general effective in a way or other. Analytical mechanics
was once called rational mechanics. Newton contrasted it to practical mechanics. The
latter referred to all manual arts that people used to practice from old times. It was
based mainly on experiences and not on theory and experiments. Modern science
clarified how the physical world works. This was indeed a tremendous achievement.
However, it did not made clear how our behaviors are organized. Social scientists and
economists in particular followed the track of rational mechanics. They imagined that
human agents calculate their behaviors. The only difference was that material things
36
had no intentions or purposes while human agents are stimulated by motivations.
Fortunately for those social scientists, and unfortunately for social sciences, analytical
mechanics provided the principle of virtual displacement or virtual work. A movement
of a system could be described by the variational principle. Variational method employs
minimization principle. It describes the movement of a system in such a way that the
system optimizes something (e.g. minimize the virtual work). Why is it impossible to
use this method for human systems? Modern economics after Walras was all based on
this optimization principle. If we believe in this system, it is inevitable to assume that a
human being has a sufficient rational capacity. This was the main reason why the
optimization principle was believed to be the essential factors that warrant the
efficiency of economics system. This explains why optimization principle preserved the
prime status in economics long after the discovery of bounded rationality.
We have to change our rational computationist paradigm to the paradigm of Üxkül. He
made a real revolution not only in ethology but also for a theory of human behavior.
Semiotics presupposes this giant revolution. Without it, we cannot understand why we
are semiotic rater than rational animals.
4.2 What determines effectiveness of human behavior?
Now let us return to our question. Why is our behavior effective while our rationality
plays only a minimal function? When is it effective and when does it loose effectiveness?
What is the mechanism that gives us a good performance of a behavior? The answer is
not easy as far as we continue to think in a rationalist computationist paradigm.
However, if we change our paradigm, the answer is almost already given above.
First, reformulate the question straighter. Our behavior is a series of C-D
transformations. It is deployed in time. The behavior is a process in its essence. Then
our examination is to be organized along the time line. The simplest component of a
behavior is a C-D transformation, but for the present objective, we need an action of
collecting the gain or the result of the action. Consequently, our simplest series will be
composed of at least two C-D transformations.
A simple scheme of interactions between agent (Me) and the World is shown in Figure 2.
To clarify the sequence of events, the quadruple expression is more convenient. We start
from inner state q0. It dictates me to observe the world. The world is in state W0. I find
37
the mark M0 which is a small part of the world W0. If the sign I receive is S0, I do an
action S1 and transit to internal state q1. The action S1 makes a small effect E1 to the
world and the world will change to state W1. Then the world continues to change by
itself and may arrive to the state W2. In the means time I continue to observe the world
and waits until I receive the sign S2. This waiting process is described by a quadruple
qSSq where q is an order to observe the mark M2. If I do not receive the sign S2, I do
nothing to the outer world but wait for 10 seconds or so (this time lapse can change
conveniently) and return to internal state q. With this quadruple or program, I continue
to observe M2 every 10 second until I receive the sign S2. S2 is the sign of cropping. I
make an action to collect the yield G. Then I will probably estimate the quantity of my
gain and then transit to other behavior.
During this process, the world proceeds by itself. The effect I add may changes the
course of the world a bit but the effect is most of the case very small. Normally I do not
know what happens between W0 and W2. I expect that a mark will appear in the world,
but I do not know why and how it does happen. I only know that if the case is q0 and the
observed sign is S1 and if I act S2, then I have a good chance of getting the result R.
Objective World
Inner World (Me)
t
0
t
1
t
2
t
3
W0W1W2W3
M
0
S
0
S
1
E
1
M
2
S
2
S
3
R
q
0
q
1
q
2
q
2
time
Internal
state
Marks
Sings
World
State
Figure 2. A Scheme how we behave in a complex world
The knowledge I know about the world is not a very scientific one. I only know how to do.
38
The reason why I act like this is based on the past experience. This experience may be
my own one or may be that of other persons. I learn from what other people do and my
neighbors will learn from me.
When Gilbert Ryle (1949) talked about the difference between knowing thatand
knowing how, he must be thinking a process close to our question. Arguing knowing
how, Ryle mainly wanted to refute the intellectualist legend,which hides in most of
our thinking. He defines this legend as a belief that a good performance is to do a bit of
theory and then to do a bit of practice.
Intellectualist legend reveals our anchored tradition of our way of thinking. To obtain a
good performance, people think, it is necessary to have a good theory of the world.
However, if we reflect on our behavior as it is formulated in Figure 2, it is not the
knowledge of laws of the world that gives us a good result. Even if we do not know how
the world develops, if the action S1 when S0 is observed gives a good result R with a good
probability, our performance is good. The knowledge of laws of the world may contribute
to improve our behavior, but the effectiveness of a behavior depends in large part on the
combinations of C-D transformations.
Although his main purpose was different, Ryles comparison between knowing that and
knowing how was extremely valuable. In the classical Greece, mathematics and
astronomy were models of our intellectual accomplishments. Philosophers thought that
it was in the capacity for rigorous theory that lay the superiority of men over animals,
of civilized men over barbarians and divine mind over human minds(Ryle, 2009, p.15).
Then, as Ryle put it, the following understanding of rationality naturally emerged,
To be rational was to be able to recognise truths and the connections between them.
To act rationally was, therefore, to have one’s non-theoretical propensities
controlled by one’s apprehension of truths about the conduct of life. ((Ryle, 2009,
p.15))
History of modern physics strengthened this belief. The great success of Newtonian
physics made us believe that the world is governed by laws and if we know these laws
better, our capacity to govern the world will extend. This was indeed true and modern
world changed much owing to this world conception. Despite of this enormous
significance, mathematics and modern sciences can only be a small part of our
39
intelligence. This is the sphere of know that. Another part lies in the sphere of
knowing-how. However great the sphere of knowing-that is, majority of our knowledge
still lies in the sphere of knowing-how. Ryle did not emphasized this fact, but this is his
greatest contribution to the under standing of the human behavior.
Human behavior is organized as knowing-how. Mathematical statements and scientific
laws are described by propositions. The value of a proposition is true and false.
Knowing-how is described by directives. The value of a directive is not true or false, but
good or bad. The mode of knowledge is fundamentally different. Even though, we have
no good theory of this sphere of knowing how. In schools we are taught both of them, but
teachers have a deep tendency to underestimate knowledge of knowing howand
preach that knowhow has no general applicability as true knowledge. They mean by
true knowledge the knowledge of knowing how.” They are right in their statement, but
they do not the real variety of knowledge and the weight of each type.
While Ryle talked long on what it means to act intelligently, he did not explained how a
good performance is obtained. As I have mentioned above, a good result of a behavior
(and of a decision) does not depend much on rationality or calculation, but on the
knowledge of the pattern of the world development. In a few fortunate cases,
optimization gives a better result, but we cannot think them a typical case.
The performance of a behavior depends on many factors: the definition of the state, the
accuracy of observation, the exactness and the timing of the execution, and others. A
good behavior is sometimes difficult to learn. Even if we know the rough pattern of
behavior, the mark we have to observe may not be well defined, the sign we catch may
depend on a delicate difference of something not well defined, and so on. Scientific
research of the behavior of a skilled laborer may reveal the secret of his or her good
performance, but it requires a long specialized study. Even the skilled workers
themselves cannot tell the delicate nuance of their judgments. So, the possibility of
improvement always remains and labor productivity increases with the accumulation of
experience and trials and errors.
Experience and efforts improve the performance in general and the improvements may
be enormous. However, it is important to know that in some cases the structure of the
process limits the best level of performance. A best example would be given by an
investor who tries to outperform the stock market by technical analysis. Let the
40
investor be a professional day-trader. He has a repertoire of deciding buying and selling
moments. One of such decision rule is golden cross.It is the moment that two different
average curves crosses. If he observes a golden cross of a brand of stock, he buys the
brand. However, if we believe the weak version of the efficient market hypothesis (i.e.
the irrelevance of technical analysis)17, he cannot expect to get a profit constantly from
his strategy. Stated more precisely, he cannot expect that his average return is positive.
Another strategy is to place the same amount of bid and ask order at the beginning of
the day. If two orders are executed and the spread of the two orders is larger than the
commission of brokerage, our trader makes a profit. This strategy has a risk. If only one
of two orders is executed, he has to close his account out by buying or selling contrarily,
even at a loss. Taken this risk in consideration, if the volatility of a brand is high enough,
the trader can make a profit with this strategy. However, if this strategy is really
profitable and many day traders employ a similar strategy, the normal volatility of the
market becomes small and the strategy would loose the possibility of making profits
(Shiozawa, 2008, §6.4; 2016, §1.4.5).
The lesson to draw here is that the performance we can expect from a behavior, a
decision rule or a strategy depends on the development of the economic process.
4.3 Loosely connected nature of the system
Stationarity of the economic process enables human agents to behave in a rule-based
way. The process gives a cue for the action and we draw benefits from some constancy of
the process. However, as we have observed, human agents are under the yoke of three
limits: myopic sight, bounded rationality and locality of execution. If we compare the
bigness of an economy and the narrowness of the range of human actions, it is a natural
question to ask by what mechanism we can influence the economy.
We mankind live in the interface of land and atmosphere. We learned to stand up and
walk vertically. This enabled us to have two free hands by which we work and
manipulate everything. This must really be basic conditions that make us possible to do
almost everything we can do. However, human economy has a dimension that is far
beyond the range that a man or woman can manipulate. Indeed no body can manipulate
17 Fama’s efficient market hypothesis proves information efficiency of stock markets
but should not be interpreted as proving that market system is efficient in a normal
sense of efficiency. Stock markets are full of bubbles and crashes.
41
or control the total economy even if it is a small economy with one million inhabitants.
In order that an agent with three limits can behave in a suitable way, the economy itself
must be equipped of special characteristics. In a word, the world must be nearly
decomposable (Simon, 1962). It must be loosely connected and the small part can be
changed independently form other parts of the economy. H. A. Simon (1962, 1979) called
this feature (almost) empty world assumption.” In his expression, “most things are only
weakly connected with most other things(Simon, 1962, p.478) and most things are not
related directly to most other things(Simon, 1962, p.74). Near decomposability is really
the very basis of all economics activities but I prefer the expression loosely
connectedness” because the economy is a connected system after all.
Components of a system are loosely connected when each component has some range of
independence, or free move. They are connected because they cannot take values beyond
the range of independence. In the most abstract way, loosely connected system LC is
defined to be a set that satisfies the following conditions:
(1) LC is a set of entities E1, E2, , E
N
for a large integer
N
.
(2) For all
i
, a vector
v
(
i
) in a fixed vector space is attached for each E
i
.
(3) For any pair of indices (
i
,
j
) a scalar
R
(
i
,
j
) is affected.
R
(
i
,
j
) is normally a positive
real number but may take nominal value infinity .
(4) A vector v(
i
) can take arbitrary values provided that for any pair (
i
,
j
), vectors
satisfies the constraints:
v
(
i
)
v
(
j
)∥≦
R
(
i
,
j
).
Simon gives the similar definition for his nearly decomposability by assuming that
almost all entries except a few in a relation matrix are near to zero. (Simon, 1962. p.475)
The trouble with Simons nearly decomposable system is that it assumes (almost) linear
relations. Such an assumption is necessary when we want to analyze a large scale
system. However, all variables must move simultaneously in a nearly linear
decomposable system. A human agent with three limits cannot engage in influencing
such a system. What we can do is to interact with a small part of the system which is
relatively independent from the rest. This is possible, but when two or several
components are connected tightly by the constraints like (4), we are sometimes
incapable to control even the very small part of the system. These constraints are in
general non-linear. This is one of reasons we prefer the definition above rather than
Simons nearly decomposable system. Non-linearity is an essential feature of loosely
42
connected system. Of course, this is not an easy option, because analysis necessarily
becomes complex and complicated.
Our main intention is to study a dynamics of a loosely connected system and we present
some concrete examples in the chapters after the second. In order to understand what is
really happening, we need a linear analysis of large scale but we are obliged to exclude
the cases when inventories are depleted. In such cases, we are obliged to make
non-linear operation such taking the maximal of two variables. Even in such cases, we
can use computer simulations and grasp a general feature of the process we investigate.
Of course, we cannot establish a theorem by such simulations. As a consequence, we
think both linear analysis and computer simulations are necessary and inevitable
methods of analyzing loosely connected systems.
In order that an economic system is a loosely connected system, the system must be
provided with specific instruments or material bases that make each part independent
even within a small rage. One of such universal instruments is inventory. Inventories
exist everywhere: material inventories, work-in-process inventories, product inventories,
inventories in transit (distribution), in-home stockpiles, and others. Ubiquity of
inventories shows how they are important. In fact, every parts of an economy is
disconnected by the existence of inventories. Imagine a world with no inventories. It is
like a railway system where all trains are connected rigidly between them. Such a
railway system does not function at all. In the similar way, economy without inventories
does not work at all.
Another important instrument of disconnection is money. Money disconnects buy and
sell. It is quite evident that a modern large scale economy does not work without money.
Money has many functions that we know in the textbooks. Few textbooks point that
money works as an instrument which makes economy loosely connected system.
Closely related to money, credit plays the similar role as money. Credit permits some
one to procure a commodity without having enough money for the moment. Deferred
payment is now very common in the transactions between firms. It is astonishing that
selling on credit for consumers was common and popular in Edo period Japan. These
facts may show also the importance of disconnecting function of the credit system.
A different kind of loosely connectedness is operational in any kind of organizations. For
43
example, organizations are structured in a hierarchy. A director at any hierarchy level is
delegated a power to decide by him- or herself within a certain assigned range.
Delegation of authority is the
sine qua non
principle which makes an organization
works.
4.4 Subsistence Conditions
This is the most often forgotten conditions for the well functioning of an economics
system. Imagine that a majority of members of an economy perish by some reason, for
example by an invasion of creatures from outer space. (Of course, I do not believe such
nonsense.) It is easy to imagine that all economic networks will be broken down and we
have to search from the start who can afford this and that in which price and quantity.
We will be put in a rude market situation that a neoclassical economics presupposes.
After Josef Schumpeter advanced the concept of creative destruction, it became very
popular among a wide range of people. Cox and Alm (2008) appreciated in their
encyclopedic article that creative destruction has become the centerpiece for modern
thinking on how economies evolve.We can now find many books that comprise
creative destructionin their titles. Creative destruction was accepted as a necessary
cost of efficient market economy. Although Schumpeters vision that innovative entry of
entrepreneurs is necessary for creative capitalism is correct, the appropriate level of
destruction is crucial for it to be creative. If it is one percent of firms per year that exit
by bankruptcy or close down, the economy can be active and prosperous. If more than 10
percents of firms go bankrupt, it is disastrous for an economy. The gale of creative
destruction must not be too strong and lethal.
A sudden, wide and strong destruction changes the economy too rapidly and it disrupts
the vital stationarity of the economy. People or firms cannot adapt to the new economic
situation, but it takes time for them to adapt themselves. They loose the very basis of
their behavior and will be lost what to do.
Considerable part of economic knowhow is supported by a team of workers. When a
factory is closed, each worker may retain his or her knowhow and that may be useful in
a new workplace if he or she is employed, but if the team is dissolved major part of
teamwork knowhow will be lost perhaps for ever. Innovation is necessary, but we should
not forget that creative destruction has its two faces. If the destructive face is too strong,
44
the gale of creative destruction kills the creativity of people itself. For a healthy
economy, a measure to moderate destruction should not be excluded.
The term subsistencemay remind us the classical economistsconcept of subsistence
wage, but this subsections remark has little connection to the theory that wages must
remain at the subsistence level (iron law of wages). It is doubtful if there is a sharp line
that divides the level of life where the population grows and that of population declining.
This subsection does not imply that a society is in a so-called Malthusian trap. It only
claims that an economic state that brings too many households and firms to bankruptcy
or physical destruction in a short time is not sustainable as normal economy. Sufficient
room or buffer for the survival of agents is a necessary condition for the good working of
an economy.
Section 5. Methodology of Analysis
Human behavior as well as animal behavior has a special time structure: observation,
sign-mark, action, and transition to next internal state. These are deployed in time.
Consequently, the core of our analysis must be sequential changes in the time axis. This
kind of analysis has various names: sequential analysis, sequence analysis, period
analysis, step by step method, process analysis, and others. We adopt here term
“process analysis as the unified name. In time series analysis period analysis and
continuous analysis are compared ad contrasted. Our concern lies in the different
opposition. Our process analysis is contrasted with equilibrium analysis. In economics,
as a result of long dominance of equilibrium analysis, many analyses are contaminated
by equilibrium analysis. In the subsection 5.1 we will see the minimal characteristics
that a process analysis. For a practical purpose, differences of time spans important.
Subsection 5.2 discusses briefly how to conciliate different time spans and decision
hierarchies. Human agents learn by experience and creation. As this learning occurs
inside of the economics process, a special cycle emerges between individualsbehavior
and total economic process. We call this cycle micro-macro loop. Micro-macro loop is not
only important for understanding various features of economic processes but raise a
necessity of a new type of methodology. Subsection 5.3 is devoted to this topic.
5.1 Some notes on process analysis
If we admit that human behavior is a pattern that follows events in time, the stage of
analysis cannot be equilibrium. The analytical framework must comprise time variable
45
in an essential way. Consequently, our framework of study must be process analysis.
This forces us a big problem. Major method of economic study has been equilibrium
analysis. This notion existed in the days of the classical political economy. Neoclassical
economics polished vague ideas of the classical period and refined them to mathematical
formulations. Equilibrium framework was at first adopted as it was more tractable than
other methods. Even today, it is not easy to abandon equilibrium analysis and adopt
another framework. This explains the conservative attitude of many economists to
abandon equilibrium analysis. As I have noted above, there are economists who believe
that we loose all analytical tools if we oust equilibrium and maximization.
Discussing two methods employed by Keynes, Meir Kohn (1986) pointed that the switch
to equilibrium framework was one of reasons of the success of
The General Theory
. In
his opinion, Keynes employed process or sequence analysis in
The Treatise of Money
but
switched to equilibrium analysis in
The General Theory.
Yoshida (1994) expresses the
same observation. Equilibrium analysis is easier to understand and made it more
acceptable to wider range of economists. However, this concession accompanied the
abandoning of true monetary analysis. Equilibrium framework is not consistent with
true monetary analysis. For example, hoarding and forced saving were in conflict with
the static nature of liquidity preference theory (Kohn, 1986, p.1218). The principle of
effective demand would be another example, because it cannot be defined coherently in
an equilibrium framework. As we all know, a commodity named money in Walrasian or
Arrow-Debreu systems is not money at all and its plays no role of money. Although the
General Theory contains the word money in its title, Keynes could not treat real
functions of money. The Keynesian revolution was destined to remain in real analysis.
Then, was it better that Keynes continued to be attached to sequence analysis of
The
Treatise of Money
? Kohn does not simply believe so. Sequence analysis, or step by step
method in Dennis H. Robertson’s phraseology, is much more difficult and Keynes could
not have succeeded to develop and formulate his new ideas and principles that became
the core of The General Theory. Process analysis was a new method of analysis among
the Young Turks of economic thinking including R.G. Hawtrey, D.H. Robertson, B. Ohlin
and Keynes himself (Keynes 1979, p.270, cited in Kohn, 1986, p.1201 ).18 This new
method was a criterion for Keynes when he wanted to distinguish real-exchange
economic(meaning barter economy analysis) and true monetary analysis. Thus,
18 Keynes might have named Ohlin as representative of Stockholm school economists.
46
according to Kohn, a real revolution of The General Theory should be a revolution not in
contents but in method19. Despite of all, Keynes finally abandoned this revolution and
returned to more classical method of equilibrium analysis.
This episode illustrates well the difficulty and problems of the process analysis. Keynes
had enough reason to abandon sequence analysis in favor of equilibrium analysis.
Ambiguity remained whatever Keynes tried to define a period, which was to give the
unit of analysis and failed to do it in a logically coherently way. Another difficulty with
process analysis was the vagueness of the obtained results. Keynes wanted more
clearer-cut conclusions than those obtainable by process analysis. In a word, process
analysis was intractable. We should keep this episode in mind and get aboard a ship of
process analysis. This is a difficult way. And yet this is the way we have to take in order
to make economics really monetary and evolutionary one.
Is there any prospectus of success? I would say yes. In the time of Keynes, Robertson,
Hawtrey, and Stockholm School, they had practically no tools to analyze a bit
complicated process. We have now many tools. The most important and universal tool
for process analysis must be computer simulations. Many varieties of agent-based
simulations are now developed (Shiozawa, 2016). Other tools comprise bang-bang
control theory, dynamical systems theory, inventory control theory, stationary and
non-stationary stochastic theory and non-linear complexity sciences and mathematics
in general.
The fact that we have many tools of analysis does not imply that our study will be
organized in a good framework. We must be specially aware of risks that equilibrium
framework infiltrate into our analysis and contaminated it. A typical example may be J.
R. Hickss notion of shifting equilibrium. This notion exists in Keyness General Theory
but it was Hick who gave a precise concept of temporary equilibrium and its shift.
Hicks himself re-evaluated temporary equilibrium concept in
Value and Capital
in his
later writings. Reservations Hicks made were three types: conception of uncertainty,
assumption of perfect competition and that of flexible prices (De Vroey, 1999, p.33).
However, in view of building true process analysis, these are no crucial problems. The
main trouble of Hickss temporary equilibrium is that it is a mixture of decision making
19 Kohn (1986) guesses that Keynes meant by the epithet General monetary theory
which he deemed more general than special real exchange economics.
47
and negotiation without explicit description of the process. Typical example is the
determination of price by demand and supply. Hicks himself worries about flexible price
assumption but does not inquire how this prices are determined. Are these prices
natural phenomena? If they are determined by some agents, it is necessary to clarify
how this process of price determination proceeds.
The spirit of process analysis is to clarify the time structure of all decision making and
information transfers. In other word, it is to clarify how and in what order relevant
variables are determined. To make this spirit effective, we have to keep two principles.
(1) Never use the variables of future dates in the determination of present variables.
Time order is the most important imperative that we must not violate. (2) All variables
are either determined by physical relations from other variables or determined by some
agent.
The assumption that prices are determined by law of demand and supply violates above
two principles. First, how are the time orders of demand, supply, and price of a
commodity? How are they determined? Standard formulation assumes that a price is
announced by an auctioneer and consumers and producers react to the price. Who is an
auctioneer? Except in the case of an organized market such as stock market, no such
agents exist in the economy. How can we know the total sums of demand and supply?
Who and by what means calculates them? What happens when the demand and supply
are not equal? Standard formulation assumes that the auctioneer tries again to
announce a second price and consumers and producers respond to this announcement.
When does this process come to an end? Process analysis is not the method to follow a
virtual time series. It follows what actually happens. Every determination must be
made within a predetermined lapse of time. Of course, some decisions are postponed
until some convenient chance arrives. Even in that case, the decision to postpone a
decision is made. In the concept of temporary equilibrium important process of price
determination remains in a black box. A time process that requires infinite length of
time is inserted in a temporary equilibrium and no one questions this absurdity. We are
too much accustomed to the mythology of Walrasian groping. Process analysis is a way
to demolish this firmly established custom of economic thinking.
Expectation is the topic which appears in almost all economic arguments of
recent days. Some economists talk about the necessity to act on the expectation
so that expected inflation rate and by consequence expected interest rates will
48
go down. Recent macroeconomic models have explicit variables that represent
peoples’ expectation and those variables play an important role in the
determination of real variables like investments and productions. From the
evolutionary point of view, expectation cannot play such an important role. All
economic agents are adaptive actors who change their expectation adaptively. In
other words, people adjust their expectation each time they experience
disappointment. In this adjustment, the reliability of expectation is included. If
this adjustment works, the reliability must not be very high, because we are
very often disappointed when we make any expectations. Present
macroeconomics ignores this fact and put too big weight on expectations20.
Over reliance on expectations reveals the rationalist world view which lies in all
neoclassical economics. It sees that economic process is governed by rational calculation
of human agents. As we have seen in Section 2, it is an apparent misunderstanding. It is
rather the outer objective world which calculates. Section 3 revealed that human agents
with three capability limits behave just like animals do. They calculate but
parsimoniously. Confusion exits on the role of expectation and what might be named
anticipated preparedness
. We prepare for the future events, but normally we do not
calculate the probability distribution of what happen in the future. There are so bi
uncertainties and it is not wise to act on the calculation of expected returns. In a real
life, we anticipate various cases that will happen and prepare for the time when a case
occurs. This is anticipated preparedness. If we prepare for more cases, we are safer
because the chance that an anticipated case happens will be bigger. This is another form
of the repertoire of behaviors. Anticipated preparedness means we possess an action
plan when an anticipated case happens.
I have talked long about expectations. It is because some economists contend that
expectation makes it difficult to follow the first principle of process analysis and in view
of importance of expectation this is a fatal weakness of process analysis. This contention
is based on two misunderstandings. First, expectation of time t+F is a variable set in
mind of someone who is at time t. Then, expectation is a variable at the time point t. It
tells the state of mind at time t about the occurrence of some events at time t+F. This
expectation is formed from the past experience and obtained information. It has no real
20 Keynes is partly responsible for the actual state of macroeconomics because of his
observations on expectation in the Chapter 12 of
The General Theory
. The present
arguments are forgetting the Keyness theory on the weight of an inference which is an
innovative core of his
Treatise on Probability
.
49
relationship what will happen at time t+F. When the expectation is betrayed, we are
disappointed. If disappointment continues, we are urged to change (1) expectation
formation formula and (2) the reliability or weight of expectation. It is only for the
equilibrium analysis that a difficulty occurs. In equilibrium, expectation e(t, t+F) at
time t must be identical what happens at time t+F. This is the reason why rational
expectation hypothesis is necessary in equilibrium method21.
5.2 Hierarchy of time spans and controls
Time unit plays important function in process analysis. In practice, we have to use
various scales of time unit: a second, a minute, an hour, a day, a week, a quarter, and a
yea r. The choice of a time unit depends on what process we want to analyze. If we do a
motion study, second would be a good unit. If we are concerned with investment,
perhaps a quarter or a year would suit well. In theory, the time proceeds each time an
event occurs. In this sense, steps are not necessarily equal length. For example,
customers come randomly and buy this and that items. It may make a Poisson point
process if you take a short duration like an hour but the frequency may swing from
morning to afternoon and to night. If we are concerned with investment in new factories,
the time span between each investment may change from this and that. Essential point
is to keep the time order of events.
Economy is a complex system and comprises too many features. We cannot contain
everything in one analysis. Each analysis has its purpose. We should take a time unit
that is appropriate for the purpose. If we are examining a production process in a
passenger car factory, second or minute will be a good unit. If we are examining a
supplying process of an independent small shop, day or week would be a good unit. In
every process, a variety of events of different time scales are running. For a convenience
of analysis, we condense a series of events as it is an event at a point of time. If a shop
owner is calculating if it is necessary to supply an item next day, we may condense the
series of sales of the day as if the total sale was made at once, because what matters for
the owner is the past series of sales volume of each day and not the detail of the time of
each sale. The owner calculate an average sale and judge and check the amount of
inventory left and judge if a new supply is to be made or not. This procedure of
condensing time is necessary if we want to make our analysis tractable. Physicists call
this procedure
coarse graining
. In process analysis, we are always doing coarse graining
21 For the concept of theoretical necessity of a theory,see Shiozawa (2016) Section 1.
50
if we do not know what we are doing. Appropriate coarse graining is as necessary as
taking an appropriate time units.
Selection of an appropriate time unit often corresponds to the time scale of decision
making. As an organization is structured in hierarchy, decision makings are arranged in
a hierarchy of time spans. Workers in a production site make judgment at each tact time
if a piece in process is finished as it is required. Production manager decide how many
pieces the factory makes for a given day. The factory manager decides each quarter if
the factory increases the production capacity of a product or not. The top management
decides perhaps each year if the firm builds a new factory or not. These are only a very
rough description of judgments and decision makings which take place in a firm and in
an economy as a whole. Time unit of analysis must be taken in such a way that it
corresponds to the time span of the concerned decisions. For example, if we want to
examine if the total process of quantity adjustment can follow the slow movement of the
final demand (as we will do in Sections 3 to 6), a day or a week may be a good unit
because productions and inventories are adjusted everyday or every week. If we are
concerned in the investments, a year will be a good unit. Thus the time span hierarchy
of decision making gives us an objective base in making an appropriate coarse graining
and selecting a time unit.
Characteristic time span changes according to a class of behaviors. If the time span is
short, the decision making is fast and almost automatic. The behavior seems like a
simple couple of stimulus- response. When the stake of a decision is bigger, decision
making becomes more important and we can afford and must spend more time and
resources. Inevitably, the interval between two decision makings becomes longer. A
manager at the higher hierarchy deals with the problems of wider variety and bigger
stakes and the time interval of two consecutive decisions is longer. Thus, we can observe
the following tendency in the different levels of the hierarchy. .The lower the level of the
hierarchy, the decision making becomes more instantaneous and automatic and the
variety of decisions narrower. The higher the level, the decision making becomes more
complicated and difficult, the stake bigger, and the variety of decisions larger.
Beer (1981) described that a hierarchical firm functions when each level of hierarchy
has autonomy and the higher level intervenes to lower levels by exceptions. This image
helps much in building a model of process analysis, because in analyzing a level of
decision making we can often assume that the process in the lower level works as an
51
autonomic system.
Economy is a large scale complex system. In final analysis, everything is dependent to
everything. We can interpret Walras that he wanted to analyze these relations. He was
in part right in this attempt. However, he was (or more correctly economists after him
were) wrong that these dependences are simultaneous relations. An economic agent can
observe only a small part of the economy and can act on small number of variables.
Influence of this act is transferred step by step to other variables and in the end it
propagates to the whole economy. General equilibrium theory neglects all these process
and assumes that the final possible state is the real one. If all production techniques,
consumers preference, the states of natural resources, climates and others do not
change for a long time, maybe we will arrive at such a state where nobody wants to
change his or her actions and quantities and prices are repeated day by day. Our
economy is much more dynamic and full of changes. It is an ever changing world.
Without taking this in account, economics loose all relevance to the reality.
After general equilibrium theory became exclusive framework of economics, people
began to forget that there is no instantaneous adjustment. All unrealistic fantasies like
no involuntary unemployment, no trade conflict and no financial instability come from
this instantaneous adjustment myth. Process analysis provides more realistic method of
examination. Although it is a big challenge, process analysis has a duty to change this
state of mind among economists.
Because process analysis is a new framework, it requires new methodological concepts.
As an economic agent (a person or a firm) sees and acts on a tiny part of an economy,
there is always big gap between the small world that each agent possesses and the
whole economy which exists objectively. Time span of human actions is not very long,
whereas an economic structure changes most of time very slowly. This also makes a gap
of perceptions. These two gaps pose a necessity of a new concept which I named
micro-macro loop. This is the subject matter of the next section.
5.3 Micro-macro loops and a new methodology
Micro-macro loop extends in two dimensions: one in time and one in space. In both cases,
micro-macro loop stands for a loop composed by double causal links from micro to macro
and macro to micro. The link from micro to macro is easy to understand. Many social
52
sciences and economics in particular suppose that social process is composed of
individuals acts. This is the stance of methodological individualism. If we stand on
methodological individualism, all we have to study is to examine how individuals
behave and compose the total process from these actions. This methodological stance is
quite right so far as we confine ourselves in the study of short time duration where we
can suppose that all our behaviors are given and remain constant. However, our
behavior changes in the ling run, and this change is an evolutionary process.
Suppose our behaviors are selected just like natural selections of animal species.
Suppose a situation where two subspecies that have similar behavior patterns and one
is better adapted to environment than the other. It is normal to think that a better
adapted subspecies survives and in the end dominates the other subspecies. However,
this selection depends on the environment. If the two subspecies have a different
environment, it is possible that another subspecies is better adapted and becomes to
dominate the species.
Methodological individualism is constructed by ignoring this simple fact. This
methodology continued in economics for a long time because it believed that human
agents are rational enough that their behavior is objectively best and does not depend
on the environment. In reality, human beings rationality is bounded and its sight is
myopic. As we have discussed in Section 3, our behavior is a result of long process of
selection. Selection may be intentional and conscious but it is often an unconscious
process. That is why we are not well aware of the fact that our behavior is a result of
long time selection.
Let us cite an example. When the Japanese economic miracle was still impressive,
Japanese management was praised as one of best practices in the management science.
Japanese management comprises three established customs: (1) life-long employment,
(2) seniority-based wage and promotion system, and (3) labor-management cooperation.
When Japanese economy was growing rapidly (more than 3 % per year in real terms),
all went well and many commentators argued that the three customs explains the high
performance of Japanese economy. They were right at least in the sense that the three
customs contributed positively to the Japanese economy. However, this lucky
combination did not continue eternally. When, in and after 1990s, Japanese economy
came to stagnate long, it became clear that three customs are supported by the high
growth rate. For example, many enterprises could not continue seniority-based wage
53
and promotion system and were obliged to modify the system in adaptation to low
growth regime. In the high growth age, a fortunate loop exited between individual firms
behavior and the high growth rate. Firms behavior represented by Japanese
management contributed to the high performance of Japanese economy and the high
performance made Japanese management possible and rational. There was a
micro-macro loop which helped high performance of Japanese economy. If we borrow
two words from cybernetics, the micro-macro loop was self-enforcing. When the bubble
burst, the micro-macro loop became self-destructive and Japanese management was
forced to change a lot.
Similar relations between macro-features and individual behaviors can be found in
various fields in economy. Another example of micro-macro loop is more universal and
explains an important feature of modern economy. Economics talked much about
higgling and haggling in price determination. In reality, haggling and haggling is an
episode which is seldom observed in everyday life of a developed economy. In everyday
life, prices are fixed by sellers and we buy this and that at given prices. This one-price
policy was declared publicly in Japan in the late 17th century (c. 1673) by Mitsui
Takatohsi, the founder of Mitsui group, at his shop in Edo (now Tokyo). I do not know
the detailed history of fixed price system but people in Edo welcomed this new policy
and other shops stated to imitate this fixed price system. Now this system spreads
almost everywhere except for, for example, some carpet shops in some part of South
Asia and elsewhere. This fixed price system is also common in trade between firms.
Why did this system spread widely? No laws stipulated to do so. Firms have right to
negotiate with customers and to fix different prices to different customers. One reason
to adopt this policy may be the sense of fairness. For a merchant who wants to keep
shop for a long time, unreasonable differentiation of customers may engender rage
among customers. Another reason for one-price policy is efficiency. Shop owner had to
pay the time cost for negotiation. If there are enough customers, it would be more
profitable to sell at the fixed price than to aim windfall profits. The policy was welcomed
if the fixed price was as low as other shops prices after negotiation. For busy customers,
negotiation meant time cost. So, both sides had merits to one-price policy and this must
be the reason why one-price policy spread all over the world.
If we stop here, this is only a case of evolutionary stable strategy in the economy. Let us
ask more deeply the reasons why one-price policy spread widely and ask at the same
time why in some cases higgling and haggling remains. One-price policy is profitable
54
when the commodity has some special characteristics. First of all, the commodity must
be reproducible. Second, the stability of supply is assured. Third, the procurement price
is stable. If these three conditions are satisfied, and if large demand is expected,
on-price policy was a good selling strategy. These conditions became common after
industrial revolution and cheep and fast transportation. Thus one-price policy was
dependent to the general change of economic conditions. It is noteworthy that
widespread one-policy system provides the basis for other merchants and producers to
adopt one-policy system, because one-policy system guarantees the stability of prices
and supplies. If we dig into the reason, we find a (self-enforcing) micro-macro loop in
this case too.
Existence of micro-macro loop undermines methodological individualism, because
actual individual behavior is a result of long process of selection and is conditioned by
the characteristics of the economic process as a whole. At the same time, micro-macro
loop destroys methodological holism. Without examining behaviors and interactions of
individuals (both persons and firms), we cannot analyze what happens in the economic
process. Thus micro-macro loop destroys both the methodological individualism and
holism. These two methodologies were two alternative philosophies when we want to
study social phenomena. Process analysis makes possible a totally new method of social
investigation.
Micro-macro loop presents a deep reason why we need evolutionary economics.22 It
explains why evolutionary economics is the unique method to understand everyday
economic processes. It explains why both methodological individualism and holism are
defective. Evolutionary economics stands on a different methodology and thus escapes
from old dichotomy of individualism and holism.
As Kohn (1986) emphasized it, true monetary analysis is only possible by process
analysis. Other topics, which are possible by process analysis but not in equilibrium
analysis, include circular and cumulative causation (Argyrous, 1996), quantity
adjustment process by means of inventories (Chapter 3-6), effective demand constraint
(Chapter 2), and economy as dissipative structure (Chapter 2). Now it is time to depart
from general methodological arguments and enter into more concrete economic analysis
which is the main theme of Chapter 2.
22 Gordon (1963) emphasized a similar loop between economic behaviors and
institutional conditions.
55
References
Alvarez, M. C, and D. Ehnts (2014) Samuelson and Davidson on ergodicty: a
reformulation. To appear in
Journal of Post Keynesian Economics
.
Aoyama, Hideo 1941 A Critical note on D. H. Robertson's theory of saving and
investment (I),
Kyoto University Economic Review
,
16
(1):49-73.
Argyrous, G. Cumulative causation and industrial evolution: Kaldors four stages of
industrialization as an evolutionary model. Journal of Economic Issues 30(1): 97-119.
Arthur, W. Brian 2013 Complexity economics: a different framework for economic
thought. Oxford University Press, 2013. SFI WORKING PAPER: 2013-04-012.
Beer, S. 1981
Brain of the firm
, 2nd edition, Chichester, John Wiley and sons. Paperback
edition 1984.
Calcott, B., and K. Sterelny (Eds) 2011 The Major Transitions in Evolution Revisited,
MIT Press.
Cohen, M.D., J. G. March, and J.P. Olsen 1972 A Garbage Can Model of Organizational
Choice,
Administrative Science Quarterly
17
(1): 1-25.
Cook, S. 2000 The P versus NP problem. Official problem description of the Millennium
Problem of the Clay Mathematical Institute.
http://www.claymath.org/sites/default/files/pvsnp.pdf
Cox, W.M., and R. Alm 2008 Creative Destruction, an article of The Concise
Encyclopedia of Economics.
http://www.econlib.org/library/Enc/CreativeDestruction.html11
Cyert, R.M., and J.G. March 1963
A Behavioral Theory of the Firm
, Prentice-Hall.
Davidson, P. (1991) Inflation, Open Economy and Resources. Vol. 2, Collected Writings
of Paul Davidson, Macmillan Press.
56
Davidson, P. (1999) Uncertainty, International Money, Employment and Theory, Vol. 3,
Collected Writings of Paul Davidson, Macmillan Press in UK and Saint Martin Press in
USA.
Davidson, P. (2007) Interpreting Keynes for the 21st Century: Volume 4, Collected
Writings of Paul Davidson, Palgrave Macmillan.
Davis, M. 1958
Computability and Solvability
. Dover Publications.
Dixit, A. J and J. E. Stiglitz 1977 Monopolistic competition and optimum product
diversity,
American Economic Review
67
: 297-308.
Elsner, W. 2012 Theory of Institutional Change Revisited: The Institutional Dichotomy,
Its Dynamic, and Its Policy Implications in a More Formal Analysis. Journal of
Economics Issues 46(1): 1-44. Wofram Elsner uses the word "set of patterns of
behavior."
(不要かも) Gary, M.R, and D.S. Johnson 1979
Computers and IntractabilityA Guide to
the Theory of NP-completeness
, W. H. Freeman.
Goldratt, E. M. and J. Cox 1984
The Goal
:
A Process of Ongoing Improvement
. North
River Press.
Gordon, R.A. 1963 Institutional Elements in Contemporary Economics, in
Joseph Dorfman, C.E. Ayres, Neil W. Chamberlein, Simon Kuznets and R.A. Gordon
(eds),
Institutional Economics: Veblen, Commons, and Mitchell Reconsidered
, Berkely:
University of California Press, pp.123-147.
Heiner, R. A. 1983 The Origin of Predictable Behavior. American Economic Review
73(4): 560-595.
Hicks, J.R. 1991 The Swedish influence on
Value and Capital
. In Lars Jonung (Ed.)
The
Stockholm School of Economics Revisited
, Chapter 15, pp.369-76.
Hodgson, G.M. 2002 Reconstitutive Downward Causation: Social Structure and the
57
Development of Individual Agency. in E. Fullbrook (Ed.) Intersubjectivity in Economics:
Agents and Structure, Routledge, 2002, pp.59-180.
IEEJ 2010 Shinka Gijutsu Handobukk (in Japanese; Handbook of Engineering
Technology: Computation and Applications), three volumes. Kindai Kagakusha.
Kohn, M. 1986 Monetary Analysis, the Equilibrium Method, and Keynes's "General
Theory,"
Journal of Political Economy
94
(6): 1191-1224.
Koike, K. 1995 The Economics of Work in Japan. LTCB International Library
Foundation.
Koopmans, T.C., and M. Beckmann 1957 Assignment Problems and the Location of
Economic Activities.
Econometrica
25
(1): 53-76.
Maynard Smith, J., and E. Szatha'ry 1998 The Major Transitions in Evolution. Oxford
University Press, Reprint Edition. (The original edition was published in 1995)
March, J.G., and H. A. Simon 1958
Organizzations
, Wiley.
Mintzberg, H. 1973
The Nature of Managerial Work
. Paperback: Harper Collins College
Div.
Nakaoka, T. 1971 Kōjō no Tetsugaku (in Janaese, Philosophy of Factories), Heibonsha.
(The title is an adaptation from Ure’s
Philosophy of Manufactures
)
Nelson and Winter 1982
An Evolutionary Theory of Economic Change
. The Belknap
Press of Harvard University Press.
Neumann, J. von and Morgenstern, O. 1953 Theory of Games and Economic Behavior.
Princeton University Press.
Pak, Igor 2000 Four Questions on Birkhoff Polytope.
Annals of Combinatorics
4
(2000)
83-90.
Birkhoff polytope の特性を説明、しかしそれで simplex method がなぜうまくいかを確率
58
的に説明
Riolo, R.L., M.D. Cohen, and R. Axelrod 2001 Evolution of cooperation without
reciprocity,
Nature
414:
441-443 (November).
Rüting, T. (2004). Jakob von Uexküll: Theoretical biology, biocybernetics and
biosemiotics.
European Communications in Mathematical and Theoretical Biology
,
6
:
11-16.
Ryle, G. 2009[1949]
Concept of Mind
. 60th anniversary version, Routledge.
Samuelson, P.A. 1969 Classical and neoclassical theory, in R.W. Clower (Ed.) Monetary
Theory, London: Penguin. Pp.1-15.
Shiozawa, Y. 1989 The Primacy of stationarity: a case against general equilibrium
theory.
Osaka City University Economic Review
24
(1): 85-110.
Shiozawa, Y. 1990 Shijō no Chitsujogaku / Han-kinkō kara Fukuzatuskei e (
The Science
of market Order: From Anti-Equilibrium to Complexity
), Chikuma Shobō.
Shiozawa, Y. 1999 Economics and Accounting: A comparison between philosophical
backgrounds of the two disciplines in view of complexity theory. Accounting Auditing
and Accountability Journal 12(1):19-38.
Shiozawa, Y. 2001 Economic theory and the complexity of capitalism, in Itoh, M., N.
Yok o kawa, G. M. Hodgson (Eds.) Capitalism in Evolution: Global Contentions--East and
West, Edward Elgar, pp.36-47.
Shiozawa, Y. 2004 Evolutionary Economics in the 21st Century: a Manifest.
Evolutionary and Institutional Economics Review 1(1): 5-47.
Shiozawa, Y. 2006 General Introduction, in Japan Association for Evolutinary
Economics (Ed.), Handbook of Evolutionary Economics (In Japanese: Shinka
Keizaigaku Handobukku), Kyoritsu Shuppan, pp.4-134.
Shiozawa, Y. 2016a The Revival of Classical theory of Values, in Yokokawa et als. (Eds.)
59
The Rejuvenation of Political Economy, Routledge: Oxon and New York, Chapter 8, pp.
Chapter 8, pp.151-172.
Shiozawa, Y. 2016b A Guided Tour of the Backside of Agent-based Simulations, in H.
Kita and K. Tanigichi (Eds.) Realistic Simulation of Financial Markets, Springer. Tokyo,
2016. Chapter 1: pp.
Simon, H. A. 1962 The Architecture of Complexity,
Proceedings of the American
Philosophical Society
,
106
(6): 467-482.
Simon, H. A. 1979 The Meaning of Causal Ordering. In : R.K. Merton, J.S. Coleman,
and P.H. Rossi (Ed.) 1979
Qualitative and Quantitative Social Research: Papers in
Honor of Paul F. Lazarsfeld
, Free Press. Chap. 8, pp.65-81.
Simon, H. A. 1997 [1945] Administrative Behavior. Fourth Edition, The Free Press.
Solow, R.M. 1990 Reactions to Conference Papers, Chapter 12 of Diamond, P.A. (Ed.)
1990 Growth, Productivity, Unemployment: Essays to Celebrate Bob Solow's Birthday,
MIT Press, pp. 221-229.
Uexküll, Jakob von, 1992 [1934 ] A Stroll through the Worlds of Animals and Men.
Semiotics 89(4): 317-377. Japanese translation was published in 1973 and now
available in Iwanami Bunko.
Vroey, M. de, 1999 J. R. Hicks on equilibrium and disequilibrium /
Value and Capital
revisited,
History of Economics Review
29
: 31-44.
Yoshida, M. 1997 Keynes: Rekishiteki jikan kara fukizatsukei e (in Jpanese, Keyens :
from historical time to complexity systems), Nihon Keizai Hyoronsha.
Yoshida, T. 1990 Jiko-soshikika no Joho Kagaku (in Japanese, Information science of
self-organization), Shinyosha.
Article
Masahiko Aoki is known in the English-speaking world as a creator and founder of comparative institutional analysis. However, Masahiko Aoki’s interests changed several times. As he himself recalled in his autobiography, Aoki made in his life seven intellectual ventures in five different fields. In this paper, I would like to address a more little-known aspect and period of Aoki’s research agenda. Although Aoki’s interests changed greatly throughout his career, there was under the surface a consistent and coherent stream: the study of organizations. His interest in organizations started from a sizable one, i.e., the centrally planned economy, to a smaller one: firms. Between these interests, there was a period where Aoki was deeply interested in the dynamic process of the market economy. During this period, I had a chance to work as his research assistant at the Kyoto Institute of Economic Research (KIER), an institute at Kyoto University. This paper is a recollection of my days in KIER with Professor Aoki.
Research
Full-text available
Agent-based simulation brings a host of possibilities for the future of economics. It provides a new analytical tool for both economics and mathematics. For a century and a half, mathematics has been the major tool of theoretical analysis in economics. It has provided economics with logic and precision, but economics is now suffering; economics in the 20th century made this clear. Theorists know that the theoretical framework of economics is not sound and its foundations are fragile. Many have tried to sidestep this theoretical quagmire and failed. Limits of mathematical analysis force theorists to adopt mathematically tractable formulations, though they know these formulations contradict reality. This demonstrates how economics lacks a tool of analysis that is well suited to analyzing the economy’s complexity. Agent-based simulation has the potential to save economics from this dead end and can contribute to reconstructing economics from its very foundations. Achieving this mission requires those engaging in agent-based simulation to have an in-depth understanding of economics based on its critical examinations. This guided tour leads readers around the backside of economics, tells what is wrong with economics and what is needed for its reconstruction, and provides hints for a new direction open to incorporation of agent-based simulation. This is the first draft for Chapter 1 of H. Kita, K. Taniguchi, and Y. Nakajima (Eds.) Simulation of Financial Markets with Agent-Based Model, to appear in 2016, by Springer (Vol. 9 in Evolutionary Economics and Social Complexity Science Series)
Research
Full-text available
Agent-based simulation brings a host of possibilities for the future of economics. It provides a new analytical tool for both economics and mathematics. For a century and a half, mathematics has been the major tool of theoretical analysis in economics. It has provided economics with logic and precision, but economics is now suffering; economics in the 20th century made this clear. Theorists know that the theoretical framework of economics is not sound and its foundations are fragile. Many have tried to sidestep this theoretical quagmire and failed. Limits of mathematical analysis force theorists to adopt mathematically tractable formulations, though they know these formulations contradict reality. This demonstrates how economics lacks a tool of analysis that is well suited to analyzing the economy’s complexity. Agent-based simulation has the potential to save economics from this dead end and can contribute to reconstructing economics from its very foundations. Achieving this mission requires those engaging in agent-based simulation to have an in-depth understanding of economics based on its critical examinations. This guided tour leads readers around the backside of economics, tells what is wrong with economics and what is needed for its reconstruction, and provides hints for a new direction open to incorporation of agent-based simulation. This is the first draft for Chapter 1 of H. Kita, K. Taniguchi, and Y. Nakajima (Eds.) Simulation of Financial Markets with Agent-Based Model, to appear in 2016, by Springer (Vol. 9 in Evolutionary Economics and Social Complexity Science Series)
Book
Part 1 Income distribution and inflation: a clarification of the Ricardian rent share increasing employment, diminishing returns, relative shares and Ricardo inequality, and the double bluff a discussion of Leijonhufvud's social consequences of inflation is there a shortage of savings in the United States? the role of financial institutions monetary and fiscal policy in capital accumulation during periods of stagflation incomes policy as a social institution policies for prices and incomes can effective demand and the movement towards income equality be maintained in the face of robotics? only in America - neither the homeless nor the yachtless are economic problems markets and governments - the comparison of means and objectives under different economic systems. Part 2 Macroeconomics and expectations: causality in economics - a review rational expectations - a fallacious foundation for studying crucial decision-making processes user cost Shackle and Keynes vs rational expectations therory on the role of time, liquidity, and financial markets a technical definition of uncertainty and the long run non-neutrality of money is probablity theory relevant for uncertainty - a different perspective. Part 3 Open economies: monetary policy, regulation & international adjustments international money and internatiional economic relations liquidity proposals for a new Bretton Woods plan is free trade always the right policy? - monetary theory and policy in a global context with a large international debt. Part 4 National resources - energy: public policy problems of the domestic crude oil industry public policy problems of the domestic crude oil industry - rejoinder the depletion alowance revisited oil - its time allocation and project independence testimony before the US Senate subcommittee on antiturst and monopoly the relations of economic rent and price incentives to oil and gas supplies the price system, conglomerate energy companies and OPEC divestiture and the economics of energy supplies the economics of natural resources the Carter energy proposal the United States internal revenue service - the fourteenth member of OPEC? oil conservation - theory vs policy. Part 5 Natural resources - water and recreation: the social value of water recreational facilities resulting from an improvement in water quality in an estuary an exploratory study to identify and measure the benefits derived from the scenic enhancement of federal-aid highways the valuation of public goods an analysis of recreation use of TVA lakes the economic benefits accruing from the scenic enhancement of highways. Part 6 Other testimonies: testimony before the house of representatives committee on ways and means, March 6, 1975 federal communications commission - US government the future of the three R'S - regulations, reserves and Reagonomics.
Article
This article aims at critically assessing Hicks’ conception of equilibrium and disequilibrium, as developed in Value and Capital. In a first part, I argue that Hicks cannot be regarded as belonging fully to the Walrasian research programme in spite of his significant contribution to its development. I claim that this is due to his ill-fated will to introduce categories specific to the Marshallian approach in a Walrasian framework. In a second part, his “tripartite classification” interpretation of Marshall’s equilibrium conception will be evaluated It will be contrasted with an arguably more coherent interpretation giving pride of place to a binary opposition between market and normal equilibrium. Hicks’ posterior reservations about his temporary equilibrium construct are then critically assessed.
Article
The concept of ergodicity in economics seems to have the qualities of a shibboleth—a word or saying used by adherents of a party, sect, or belief, and usually regarded by others as empty of real meaning. It is in use by both neoclassical economics—after Samuelson (1965, p. 43), who used the term in his paper on what later became a foundation of the efficient market hypothesis—and post Keynesian economics—after Davidson, who picked up the term in order to highlight methodological differences. Considering the origin of the concept in statistical physics and its use in the topology of dynamical systems, which most economists are not conversant with, the importance ascribed to ergodicity in economic debate seems mystifying. We deconstruct the meaning of the term in the major contributions of Samuelson and Davidson. We suggest an alternative to (non)ergodicity to discuss the nature of randomness in the real world. While neoclassical theory assumes stochastic randomness, post Keynesians assume nonstochastic randomness, a term developed by the mathematician Kolmogorov (1986, p. 467). We argue that even in an ergodic world there is a problem with the idea that stochastic randomness can be dealt with by the financial system.
Article
Kaldor, who built on the work of Allyn Young. 1 Kaldor argued that industrialization is a cumulative process in which the development of industries producing consumer goods precedes the development of those producing capital goods, and where production for sale precedes production for export. Kaldor's theory has had an important influence on other writers in the Cambridge, England, tradition [Eatwell 1982; Skott 1985]. The notion of cumulative causation, through its origins in the writings of Thorstein Veblen [1898] and in the subsequent work of Gunnar Myrdal [1944, 1968], has also permeated the work of institutionalist authors.2 Despite the use of a common theoretical term, however, important differences between the Young/Kaldor and the Veblen/Myrdal versions exist that have not been fully explored. In particular, Kaldor provides a taxonomy for investigating the industrial development of individual countries. However, without a discussion of the process by which a country moves from one stage to another, this taxonomy retains mechanistic overtones of inexorable, linear development. This is contrary to the evolutionary spirit with which the Veblen/Myrdal stream develops the notion of cumulative causation. The task of this paper is to reconcile these two streams of thinking. Kaldor's four-stage model of industrial development is used and supplemented with other work that provides a more complete explanation of the industrialization process. Drawing on literature that emphasizes learning-by-using and the historical development of consumption patterns, Kaldor's four-stage model can be given evolutionary