Documente Academic
Documente Profesional
Documente Cultură
Len H. Wallast
Evolvodynamics –
The Mathematical
Theory of
Economic Evolution
A Coherent Way of Interpreting Time,
Scarceness, Value and Economic Growth
Lecture Notes in Economics
and Mathematical Systems 665
Founding Editors:
M. Beckmann
H.P. Künzi
Managing Editors:
Editorial Board:
Evolvodynamics -
The Mathematical Theory of
Economic Evolution
A Coherent Way of Interpreting Time,
Scarceness, Value and Economic Growth
Len H. Wallast
Vlissingen
The Netherlands
ISSN 0075-8442
ISBN 978-3-642-34055-0 ISBN 978-3-642-34056-7 (eBook)
DOI 10.1007/978-3-642-34056-7
Springer Heidelberg New York Dordrecht London
Library of Congress Control Number: 2013934270
v
vi Contents
E.T. Jaynes’ life-work on Probability Theory bears the subtitle “The Logic of
Science” (Jaynes 2004). The subtitle expresses the notion that probability theory
is at the core of a wide (perhaps almost the complete?) range of what we observe
and how everything behaves within our universe. Originally developed to provide
modestly for the mathematics of chance games, probability theory has meanwhile
succeeded to provide the basic explanation (the logic) of more and more scientific
domains of inquiry. Simultaneously there has been a growing awareness that
probability theory is the basic logic behind the many things we observe and behind
the many ways we interact, behave and take our decisions. Boltzmann’s derivation
of the connection between probability and entropy may be regarded as the first
manifest demonstration that the macro-effect of what we experience and observe in
the foreground as real, fixed, rational and concrete is anchored in a background of a
dynamic micro-world of abstract probabilistic events. The probabilistic logic of that
micro-world has already been shown to hold good for many physical phenomena,
but it is certainly not restricted to the explanation of physical phenomena only.
Boltzmann was the first to recognize the connection between probability and
entropy. He went even further by conjecturing—far ahead of his time—a still
more profound relationship between entropy and the non-physical biological
order of Darwinian evolutionary systems.
It has since set many to explore Boltzmann’s principle further. The way we must
understand information about our world and interpret observations and behavior
within this world—it does not really matter whether of physical or non-physical
nature—was most profoundly reconsidered by Claude E. Shannon in the midst of
the twentieth century. Shannon formulated his path breaking views within the
framework of the transmittance of information over a communication channel
(Shannon 1948). Originally communication engineers were inclined to adopt the
most simple determinist idea of what people thought information to be: something
like countable atoms shot forward through the pipeline of the communication
channel. Just raise the frequency and/or the power of shooting and you get the
best of it.
Shannon subverted that picture completely. The ultimate idea is that anything of
information we find out about the things we observe and interact with is closely
connected with uncertainty and based on probabilistic events. According to
1.2 Generalized Darwinism and Shannon 3
the entropy factors that must ultimately control economic change and equilibrium,
just as they do in thermodynamics” (Jaynes 1991). That was 1991. The financial
crisis of 2008 is now going on for more than 4 years. That it was largely unforeseen
by the accepted models of economic analysis reveals that Jaynes’ remark of 1991 is
still applicable in 2012 and—if economists keep on cherishing their current models
and methods of analysis—it will remain so for many years to go.
Noticing the difficulties with progress in economic science the recent fundamen-
tal back-to-Darwin approach of Hodgson and Knudsen (2010) is therefore very
welcome because the progress of science must always first be sought in improving
our understanding, interpretation and application of the basic principles. H&K’s
agenda has been to disclose the basic principles of what they call “Generalized
Darwinism” with the yet unfulfilled purpose to work it out into a (more) complete
theory of evolution for the domain of social phenomena. Although there have
already been many attempts to formulate an evolutionary economics by other
economists, the strict and detailed emphasis on the three Darwinian principles of
evolution in which H&K state the basis of economic and social evolution stands
out. It is exactly what we need to work out in a general mathematical theory of
economic evolution. I shall therefore discuss their ideas extensively, comment on it
often where needed and use it to elaborate on it in a deductively based universal
theory of economic selection in the line of E.T. Jaynes suggestions and last but
not least Shannon’s ideas about information and entropy. More specifically, my
contribution is first and for all to be interpreted as a generalization of Shannon’s
stationary communication theory towards the non-stationary domain of economic
evolution.
H&K concentrate primarily on the descriptive Darwinian perspective of evolu-
tion, which—I think—is too restrictive to reach their goal. We need the purity of
deductive mathematical reasoning to make progress in the line of Shannon.
Hodgson and Knudsen mention Shannon’s mathematical conception of informa-
tion/entropy and its relationship with evolutionary selection only in passing,
although Shannon’s probabilistic contribution is a major building-brick of the
evolutionary edifice and can’t be missed to derive the quantitative relationships
between the economic variables.
Nevertheless there is considerable correspondence between my Shannon-
inspired approach and the main starting points of H&K with respect to evolutionary
selection. In fact I endorse much of their views. But there are certainly a number of
issues in which I take a slightly different and occasionally a diametrically opposed
stand. These issues concern reductionism, the acceptance/avoidance of parametric
models of explanation (because of their restricted capability to describe processes
of dynamic change), the necessity of a unit of selection that does not vary in the
course of time, the nature of group selection and the multi-level layers of social
evolution.
Throughout this treatment I adhere unconditionally to the method of applied
mathematics demanding deductive reasoning and inference upon the basis of the
postulate of selection on which the Darwinian model rests, as well as on the
principles of continuity and evolutionary homogeneity, and absolute exclusion of
1.2 Generalized Darwinism and Shannon 5
My message is not that mathematical rigor should not be pursued but that
conceptual clarity and mathematical imagination is here more important than
utmost mathematical rigor. I am aiming at an economic audience that is unaccus-
tomed to the methods of analysis I apply. My emphasis is on explaining the very
close connection between the great ideas of two outstanding scientists: the non-
mathematical world of Darwin and the mathematical world of Claude E. Shannon.
However I cannot avoid putting concepts in a mathematical setting that involves the
application of some elementary principles of set theory and probability theory, but I
hope to do that in a way understandable to the majority of my audience.
The mathematical way of reasoning, to which I adhere, conforms to the basic
mathematical methods of probability theory and resembles the way of reasoning
practiced by Shannon in his mathematical exposition of communication theory. In a
sense Shannon-like mathematical reasoning will be somewhat demanding for most
economists, unfamiliar as they are with that world and accustomed as they are to
1
The principle of evolutionary homogeneity should not be confused with economic homogeneity,
involving several disputable principles necessary to explain the mainstream theories of economic
orthodoxy.
2
Cited from (Jaynes 2004, page 676).
6 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
A point that should be clarified first here is that my contribution has another angle
of incidence than H&K’s. Hodgson and Knudsen stress the general features of
Darwinian selection as providing a universal overarching description of biologic
and social evolution. They consider a different class of social phenomena in order to
clarify subsequently how these phenomena fit in the general Darwinian scheme of
selection. They discuss also at length the insufficiency of Lamarckism to explain
social phenomena, the properties of the unit of selection (the replicator), the role
and function of interactors. They criticize Dawkin’s concept of memes. The leading
line is that their contribution is primarily descriptive like Darwin’s original contri-
bution (Darwin 1859). It is only occasionally and fragmentarily that H&K consider
how a system of equations can be mathematically framed such that it will accord to
the basic Darwinian selection schedule and that it can be used to explain observed
and observable social phenomena quantitatively.
3
The contours of a very precursory version of this novel theory have been sketched by the author
in 2009 and in revised form in 2012 (Wallast 2009a, b, c).
4
The terminology of inflow and outflow is chosen deliberately for clarity of exposition. Inflow is
the collective term for input and influx, each exercising its own special form of inflow. Likewise
outflow is the collective term for output and outflux, each also exercising its own special form of
outflow. Accurate definitions of these terms will be given as we proceed.
1.3 The Prerequisite of Reliable Inflow and Outflow Data 7
length of the forelimbs of mice, the evolution of the usage of tools by hominids, etc.
Fortunately, the outlook for economic theory testing is better.
We shall concentrate on the mathematics of inflow and outflow selection. The
first thing to do is to discuss and develop the mathematical concepts. They need
to be consistent in all respects. After that has been achieved, the more appealing
challenge is to derive and justify the equations that govern the universal
relationships between inflows, outflows and stocks of value of the macro-economy
and their (relative and absolute) change in a logical deductive manner from the
Darwinian principles of selection.
Hodgson and Knudsen state that Darwinian evolution and hence economic
relationships cannot be stated fully in derivable universal equations akin to those
of physics (Hodgson and Knudsen 2010, page 3). They consider any scientific
description of Darwinian evolution as incomplete and quite different from physics
(Hodgson and Knudsen 2010, page 3 and 4). This opinion is shared by many
economists, but it lacks substance. A motivation for this point of view has never
been given. It is merely one of those common statements one is inclined to believe
as self-evident so that further amplification is judged unnecessary. However
economic theory is concerned with the relationships between non-stationary time-
dependent statistical averages like there are many: inflows and outflows with their
various propensities, circulation rates, inflation rates, growth-rates etcetera. These
variables are statistical averages of non-stationary stochastic outcomes of decisions
that reflect the micro-economic selective behavior of individual agents. Thus the
economy can only be studied correctly as a framework of interacting selection
processes while employing the advanced tools of probability theory. Chances
govern what happens at the micro-levels of individual selection. Orthodox eco-
nomic theory is often concerned with explaining why an individual economic agent
makes a particular selective decision. However why an agent makes an individual
selection is irrelevant, only the probabilities by which a particular class of agents
takes such decisions are relevant. The associated probabilities determine the state of
the economy and the statistical averages, i.e. the economic macro-variables. The
dynamic relationships between these macro-variables are all what can be derived
using the tools of probability theory. Together the derived equations reflect the
universal theory of economics we seek. In essence the approach is completely akin
to the methods applied in statistical physics. Indeed the resulting economic nonlin-
ear system of equations is quite complicated, but the universal method of scientific
analysis does not warrant that the derived system of equations is always simple.
Let us in this respect recall how E.T. Jaynes addresses the challenge of scientific
economic explanation (Jaynes 1991). Instead of excluding the application of the
analytical mathematical method of physics to the explanation of economic behav-
ior, he wonders first why it is that the orthodox methods of economic analysis fail
1.4 The Method of Science 9
and in the line of that he suggests the application of a physical analytical method
instead, i.e. to deal with the economic process as a derivable probabilistic process
that resides in a time-dependent macroeconomic state to which a probability and
entropy is to be assigned in a similar manner as in statistical physics.
Despite that many economists are still unconvinced by this argument because
they think that there is more in the individual micro-behavior of agents than a
statistical theory that discloses only the statistical averages and the mutual
relationships this offers. In fact this objection relies on the assumption that there
is more information in the details of the micro-economic selective decisions than
the entropies and mutual relationships of the macro-economic statistical averages
can disclose. This incorrect idea rests upon a misunderstanding of the concept of
information. To clarify that, we need to explain what information is. It is here where
the entire mathematically based philosophy of Shannon’s conception of informa-
tion comes to our rescue. The matter will become clear after the complete
subsequent exposition of Shannon-inspired analysis on economic phenomena. We
cannot deal with it here at the introductory level of the present chapter.
The distrust and even disbelief in the universality of the logical deductive
method of science is widespread among social scientists. It is due to unfamiliarity
with the method of analysis. Perhaps it is also a misinterpretation of the nature of
explanations that a statistical theory of selection is capable of offering. The dictum
that good science must ultimately offer complete and universally valid descriptions
of evolution is then misunderstood as an assertion that science must ultimately
exactly predict and explain the aggregates of the evolutionary process. However
statistical theories do not provide exact answers like Newtonian mechanics under
certain idealized conditions provides (Note there are also many statistical
applications of Newtonian mechanics where it is different). The forecasts of
statistical selection theories are theoretically complete but have nevertheless always
a limited range of accuracy. If they would give determinist answers, there would be
no room and opportunity for evolving populations to behave freely. Such restriction
of freedom would of course be an absurd consequence.
But there is perhaps another cause that helped to create the distrust of
practitioners of the social sciences with respect to the application of the exclusive
method of logical deductive reasoning. We have already mentioned that cause
before within another context. What we observe and measure in many biologic
processes of evolution and in social evolution other than economic are secondary
matters such as the color change of the moth in industrial areas, but it is not what a
derived system of equations of evolutionary selection describes. The latter system
presents us the relationships between the various aggregate inflows, outflows,
growth rates and average unit price. However, that is exactly what is immeasurable
in most other spheres of evolution than the economic so that the derived equations
cannot be tested within the context of the latter domains. Science is there in a
deadlock. It appears that social scientists have already anticipated this by losing
their faith in the exclusive role of the logical deductive method of science. Never-
theless and fortunately inflows, outflows, growth rates and average unit price are
measurable within the sphere of economic evolution. Once established (as will be
10 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
achieved in this treatise) the relations between the various inflows, outflows and the
other aggregates can directly be tested against quantitatively based observations of
a money-economy in the course of time.
5
To cite E.T. Jaynes: “In a probabilistic model of the economy, we ought to include in the
probability of any macroeconomic state an entropy factor to take its multiplicity into account”
(Jaynes 1991, page 2).
1.6 The Three Principles of Evolutionary Selection, Boltzmann and Shannon 11
and the relationships of all this with the average price-level of consumption and
investment and with the relative rates of change of this price-level, the liquidity
flows and the rates of circulation.
Hodgson and Knudsen assert that any theory of social development should at least
pay tribute to three Darwinian principles of evolutionary dynamic change: varia-
tion, selection and inheritance. I will come back to discuss these basic principles of
selection in subsequent chapters comprehensively because they are core issues of
Darwinian selection, but as said before there is more. There is also the physicist
Boltzmann and the communication engineer Claude E. Shannon without whose
contributions the formulation of an evolutionary selection theory is unthinkable.
The nineteenth century physicist Boltzmann was deeply impressed by Darwin’s
theory and immediately understood that a well formulated mathematical selection
theory was required to work it out coherently for quantitative testing. Well,
Boltzmann had been involved in contemporary mathematical physical research
that ultimately led to a complete theory of statistical mechanics and thermodynam-
ics in the dawn of the twentieth century founded on Newtonian principles of
mechanical motion and on statistical selection theory. He used his expertise to
derive a relation that connected the entropy H to its state probability π:
H ¼ –log(π).6
Boltzmann immediately realized that the entropy H—stripped off as it was from
its thermodynamic background with its sole relationship with state-probability—
was a measure for the order of an evolutionary system. The more order it had, the
less probable its state. Boltzmann hinted that Darwin’s selection theory was
6
One is free to choose the base of the logarithm. As a rule I shall use the base 2, so that the entropy
is given by H ¼ –2log(π). The original form in which Boltzmann stated his H-theorem is H ¼
k · log(π), i.e. without the minus sign. Herein k is Boltzmann’s constant, a positive constant which
can be set to 1 if we adjust the dimensional unit in which entropy is measured accordingly. The
difference of sign with the formula in the main text is a matter of entropy definition. If we adhere to
the Clausius/Boltzmann measure of entropy, entropy is always negative and the second law of
thermodynamics must be stated as the law that in thermodynamically isolated systems entropy
will always increase. This is the common way of physicists to deal with the entropy concept.
I adopt the much more preferable Shannon measure of entropy in which entropy is related to state
probability as in the main text. The Shannon measure of entropy will have it that the entropy of
thermodynamically isolated systems will always decrease. Don’t worry: social systems are not
thermodynamically isolated. Thus the Shannon entropy of a social system can increase, i.e. create
more order. We can formulate the latter statement of increasing Shannon entropy also in the
following manner: that as a rule the selection of an economy will become more and more
improbable in the course of time. (Note this is not a necessity. Sometimes an economy is recessive
and during the recession phase its selection becomes more probable in the course of time).
12 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
responsible for creating more and more order in evolutionary systems in the course
of time.
In the decennia covering the period between 1920 and 1950 the transference of
messages by means of electric, electromagnetic and acoustic signals (telegraph,
telephone, radio, light, radar and sonar) gained rapidly terrain. Communication
engineers became increasingly involved in the design of methods by which the
messages had to be coded and decoded for optimal transference and optimal
reception of information. The Second World War accelerated the demand for
these techniques. It was this background that stimulated a young engineer, Claude
E. Shannon, to launch an exciting theory of information and communication
handling, not easy to grasp at first glance, but extremely illuminating in the way
it treated choice, uncertainty, information, complexity, emergentism under the
common heading of the concept of entropy (Shannon 1948).
Shannon realized that the coding and decoding of signals over a communication
channel was a typical problem of statistical selection, which involved the assign-
ment of a state inflow probability PrfX0 g to the transmitted inflow sequence hX0 i at
the transmitter side and a state outflow probability PrfY0 g to the received outflow
sequence hY0 i at the receiver side of the channel.7 He considered sequences
(signals) of individual samples to be of finite time-length T and statistically
stationary (or more accurately: ergodic). The stationarity condition did not really
seem to affect the generality of his approach with respect to the communication
channel although radar and sonar signals and echoes are all but stationary. For a
discrete signal that could only assume a discrete number of different states Shannon
defined and justified the information content H ðX0 Þ per trial/sample of the inflow
sequence and the information content H ðY0 Þ per trial/sample of the outflow
sequence to equal the respective entropies:
X X
H ð X0 Þ ¼ λi log λi and H ðY0 Þ ¼ μj log μj (1.1)
i j
7
See for the notational conventions Appendix I.
1.6 The Three Principles of Evolutionary Selection, Boltzmann and Shannon 13
Transmitter Receiver
X0 Communication Y0
channel
Fig. 1.1 The transition from communication channel selection towards Darwinian evolutionary
selection.
Each sample of the inflow sequence hX0 i and outflow sequence hY0 i is either in state 1 or in state 2. For the communication channel the
distinction in states may be achieved by two different volt levels or two different levels of acoustic pressure. For two-sector economic
evolution the distinction in states is marked by having a sample to get selected either in the consumption sector (state 1) or in the investment
sector (state 2).
H(Xi) is the entropy inflow per sample in state i (i ¼ 0, 1 or 2). H(Yi) is the entropy outflow per sample in state i (i ¼ 0, 1 or 2). Sequences of
samples extend over (t,t + dt). Entropy flows emerge as outcomes of the selection of samples. The entropy inflow of all the Zdt samples of
hX0 i in state i is Xi ¼ Zdt · H(Xi) bits. The entropy outflow of all the Zdt samples of hY0 i in state i is Yi ¼ Zdt · H(Yi) bits:
H(X0) ¼ H(X1) + H(X2), H(Y0) ¼ H(Y1) + H(Y2)
Let us next consider Fig. 1.2. This is the situation after the Shannon-Darwin time-
compression operation has been performed for the two-sector model of selection. In
this schedule we keep approaching Shannon’s communication channel from the
Darwinian perspective of evolutionary selection. After the transition, selection is no
longer on a finite time-interval of length T but on the infinitesimally small time-
interval (t,t þ dt) of length dt. Thus all the samples of the inflow signal in Fig. 1.1
will be parallel-like selected on (t,t þ dt) to form a vertical inflow sequence in
Fig. 1.2. This has been represented in Fig. 1.2 by rotation of the inflow sequence of
Fig. 1.1 over 90 on the drawing-paper.8 Likewise all the samples of the outflow
signal in Fig. 1.1 will be parallel-like selected on (t,t þ dt) to form a vertical
outflow sequence in Fig. 1.2.
Each sample of hX0 i selected to annihilate in state i during the time-interval
ðt; t þ dtÞ is in the same state i at time t with probability 1. Likewise each sample of
hY0 i selected to originate in state j during the time-interval ðt; t þ dtÞ is in the same
state j at time t þ dt with probability 1.
One trial of the selection experiment involves the selection of one sample of
inflow allocated to annihilate on ðt; t þ dtÞ together with another sample of outflow
allocated to originate on ðt; t þ dtÞ.
A count of the number of samples in the particular example of Figs. 1.1 and 1.2
delivers 21 samples. This is called the sequence-length. More generally, with the
8
Mark the inflow and outflow sequences in Fig. 1.2 should actually have a small inclination such
that all of the selection events of their trials occur between t and t þ dt.
16 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
state 1 2 1 2
dt
t t+dt
continuous time
Assembled Reassembled
Zdt ¼ 21 Zdt · H(X0) ¼ Zdt · H(Y0) ¼ 19.2842
λ1Zdt ¼ μ1Zdt ¼ 14 Zdt · H(X1) ¼ Zdt · H(Y1) ¼ 12.8561
λ2Zdt ¼ μ2Zdt ¼ 7 Zdt · H(X2) ¼ Zdt · H(Y2) ¼ 6.4281
In the general evolutionary context hY0 i is another sample sequence of the same
sequence-length as of hX0 i. However for the noiseless communication channel hY0 i
¼ hX0 i. I have left the identity of hY0 i and hX0 i unaffected by the Shannon-Darwin
transition T ! dt ! 0 in Fig. 1.2. However this does not correspond with the
general evolutionary situation, which requires hY0 i to differ from hX0 i. We shall not
consider the details of hY0 i of the example of Fig. 1.2 further in this section. We
restrict the discussion here mainly to the inflow selection process of Fig. 1.2 but it is
not insuperable to imagine a quite related outflow selection process for which
generally hY0 i 6¼ hX0 i.
Preparing for selection requires—in the very first instance—the gathering of
samples (here Zdt ¼ 21 in number). The gathered samples happen to be in different
states (here 14 samples in state 1 and 7 samples in state 2). These are the outcomes
of 21 trials. For clarity we shall call the initial gathering phase of the selection
process assembling and the subsequent reordering phase of the selection process
into the sequences hX0 i and hY0 i reassembling.
Shannon claims that the inflowP sequence of length 21 must be given information
content Zdt HðX0 Þ ¼ Zdt i λi logλi in accordance with (1.1). Recall that λi is
the probability of a sample of the inflow signal to be in state i. Note that, in the
particular example of Figs. 1.1 and 1.2, hX0 i has only 14 samples in state 1 and 7
samples in state 2 so that actually λ1 ¼ 14=21 ¼ 0:6667 and λ2 ¼ 7=21 ¼ 0:3333 in
this particular binary case of selection. It follows that
Well, this is not the complete story. The problem is that Shannon’s formula is
only applicable if the number Zdt of samples is very large and 21 samples is
unfortunately far from enough to apply the approximation on which Shannon’s
formula (1.1) is based. For 21 samples the right answer is that only 16.827 u.ov.
instead of the 19.284 u.o.v. calculated by Shannon’s formula are being
reassembled. Unfortunately, the correct outcome of 16.827 u.ov. is incompatible
with the principle of evolutionary homogeneity. Thus for small sequence-length Zdt
the principle of evolutionary homogeneity will not be preserved.
However, the calculation error in Shannon’s approximation formula (1.1) can be
easily avoided by choosing a much larger number of samples. Let us choose a new
unit of value (the milli-u.o.v.) that is 1,000 times as small together with 1,000 times
as many samples. Then there are 21,000 samples of which 14,000 will turn out to be
in state 1 and 7,000 in state 2. The correct answer is now that Zdt H ðX0 Þ ¼ 19; 277
milli-u.o.v. will be reassembled, which is the equivalent of 19.277 u.o.v. This is
very close to the outcome of 19.284 u.o.v. calculated by Shannon’s formula (1.1). If
we choose a still smaller unit of value, the error between the two outcomes gets
smaller and smaller. In fact, if the number of samples is chosen infinitely large, the
error vanishes completely. See Appendix A for a complete explanation. We must
thus adopt the rule that the number Zdt of samples is always very large. For
illustration and explanation purposes, it is however sometimes desirable to keep
the number of samples at a finite limit (as in Figs. 1.1 and 1.2).
At a much later stage of analysis we shall see why it is justified to look at
this process of selection in the Shannon way (See Chap. 4 and Sect. 4.1). We shall
here just accept it. It follows that the reassembled X0 ¼ Zdt H ðX0 Þ units of
value in state 0 consist of X1 ¼ Zdt H ðX1 Þ ¼ λ1 Zdt HðX0 Þ u.o.v. in state 1 and
X2 ¼ Zdt H ðX2 Þ ¼ λ2 Zdt H ðX0 Þ u.o.v. in state 2.
In the example of Figs. 1.1 and 1.2 with Zdt ¼ 21, λ1 ¼ 14=21 ¼ 0:6667 and
λ2 ¼ 7=21 ¼ 0:3333 the numerical answers we get are Zdt HðX1 Þ ¼ 12:8561
reassembled u.o.v. in state 1 and Zdt HðX2 Þ ¼ 6:4281 reassembled u.o.v. in state 2.
In the above we have restricted the discussion to the statistical experiment of
inflow selection. For the statistical experiment of outflow selection (the other part of
the selection process) the same considerations with respect to assembling and
reassembling apply.
Anticipating further explanation in the sequel, we remark here that economically
inflow selection is the activity of economic agents (entrepreneurs and laborers) by
which they sacrifice (assemble) production resources (financial, material as well as
labor) and form (reassemble) novel combinations of the sacrificed value thereof that
effectuates the reduction of the sacrificed value. Similarly, economically outflow
selection is the activity of economic agents (investors and consumers) by which
they spend (assemble) on investment opportunities and consumption and form
(reassemble) novel combinations of the spent value thereof that effectuates the
reduction of the spent value. It is another way of looking at economic activity, but it
is also the only consistent way that accords with a Shannon inspired approach of the
evolutionary process.
1.8 The Unit of Selection and the Unit of Exchange 19
In the previous section we have considered selection on the time interval ðt; t þ dtÞ.
Selection is a continuous dynamic process. Hence it is not restricted to just one
particular time interval ðt; t þ dtÞ, but it goes on for continuously increasing time t.
Thus selection on ðt; t þ dtÞ is prosecuted by subsequent selection on consecutive
time-intervals ðt þ dt; t þ 2dtÞ, ðt þ 2dt; t þ 3dtÞ, ðt þ 3dt; t þ 4dtÞ, , each of the
same infinitesimally small time-length dt. Each interval of this indefinite series of
contiguous differential time-intervals may have its own unit of value.
But there are some serious limitations with respect to these choices of the units
of value. Whereas we argued in the previous section that it did not matter in which
u.o.v. reassembled entropy was to be expressed because that choice could not affect
the outcomes of selection on a single time-interval ðt; t þ dtÞ, we must now admit
that it does affect the comparison of reassembled entropy over consecutively
infinitesimally small intervals of time. For instance if we choose the Euro to be
our common u.o.v. for the time-intervals ðt; t þ dtÞ and ðτ; τ þ dtÞ with τ t finite
and τ t > 0, we obtain for the joint entropy inflow at time t:
ZðtÞdt H ðX0 ðtÞÞ ¼ ZðtÞdt ½λ1 ðtÞ log λ1 ðtÞ þ λ2 ðtÞ log λ2 ðtÞ Euros
Z ðτÞdt H ðX0 ðτÞÞ ¼ ZðτÞdt ½λ1 ðτÞ log λ1 ðτÞ þ λ2 ðτÞ log λ2 ðτÞ Euros
However, one Euro at time τ has another real content than one Euro at time t so
that, to keep the real content of the two outcomes comparable, we are induced to
correct the expression at time τ for the loss of real value of the Euro over the time
interval ðt; τÞ.
Likewise, if we had chosen still another u.o.v., e.g. the US dollar, then the joint
entropy inflow at time t would have been:
ZðtÞdt H ðX0 ðtÞÞ ¼ ZðtÞdt ½λ1 ðtÞ log λ1 ðtÞ þ λ2 ðtÞ log λ2 ðtÞ dollars
ZðτÞdt H ðX0 ðτÞÞ ¼ Z ðτÞdt ½λ1 ðτÞ log λ1 ðτÞ þ λ2 ðτÞ log λ2 ðτÞ dollars
We are now also induced to correct the latter expression at time τ for the loss of
real value of the dollar over the time interval ðt; τÞ. Well this is no problem if the loss
20 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
of real value of the dollar over the time-interval ðt; τÞ is exactly the same as the loss of
the real value of the Euro over ðt; τÞ. In that case we can always reduce the values in
dollars to Euros in a consistent manner. However, a dollar’s change of real value
differs from a Euro’s change of real value. The latter inconsistency can only be
avoided if we stick to the same u.o.v over the contiguous differential intervals of
selection. This is what actually happens in a particular money economy: the agents of
that economy stick to the same money unit for exchange in the course of time.
However, exchange is another process than the process of entropy selection. It
suffices to use a money unit (e.g. the Euro) as our unit of exchange in the course of
time, but to compare the inflow, respectively the outflow, of selection during the
infinitesimally small time-interval ðt; t þ dtÞ with the inflow, respectively the out-
flow, of selection during another small time-interval ðt þ τ; t þ τ þ dtÞ, it is necessary
that inflows and outflows are expressed in a unit of value that is evenly scarce at time
t and at time τ. The unit of money (like the Euro) is inappropriate for that purpose
because it does not remain evenly scarce in the course of time due to price inflation.
Instead selection demands that the units of selection to be allocated and reallocated
are evenly scarce in the time-course of the process of selection. That is, they must
share the same state probability. In consequence of Boltzmann’s theorem equivalent
portions of entropy share the same state probability. Thus the unit of selection must
be a unit of entropy and preferably a unit of entropy that is as small as possible.
A natural candidate is the unit of entropy, i.e. the bit, because it holds everywhere the
same state probability equal to 12 at any time of selection because log 12 ¼ 1 bit.
However any constant quantity of bits can serve equally well as a unit of
selection. We will leave the choice of that constant quantity open and tentatively
appoint as our unit of selection a sample, however now with the restriction that each
sample stocks the same constant quantity of entropy everywhere and always.
The conclusion is then that a sample is the common article of selection to
allocate and reallocate economic flows of entropy in the course of time. That is,
the unit of selection is a fixed quantity of bits named the sample.
On the other hand within an economy economic value is not exchanged for units
of entropy but for units of money. At first glance one may think that the choice of
the money unit in which to exchange two entities A and B of the same unit price
does not affect the economy. The argument behind this false impression is that the
two entities A and B will remain equally expensive irrespective of the unit of value
in which their price is stated. However, the choice of the unit of exchange does
affect the economy because, to facilitate exchange within an economy, money must
be issued by a monetary authority. This is in itself a source of creating additional
entropy. Dependent on the price inflation induced by monetary policy the effects on
the economy will be different and this is why the choice of the unit of exchange
matters9 in the course of time.
In summary, selection is a process concentrating on assembling and
reassembling quantities divided up in bits of entropy. Exchange is a process
9
We might perhaps also call this the unit of transmission.
1.9 Production Functions, Marginalism and the Interaction of Inflow and Outflow 21
I feel that it is worthwhile to introduce and explain the superior Shannon inspired
conception of inflow/outflow against the particular background of a neoclassical
model. Moreover, this discussion will also contribute to understand how Darwinian
evolutionary selection must be mathematically described in a general dynamic manner.
Marginalist interpreters of neoclassicism have a lot to explain. Is it permissible
to define marginal products in the manner as is actually done in neoclassicism? Let
me discuss the subject by restricting the analysis to the two-sector production
function to which neoclassicists often adhere to explain the basic idea behind
their theory: the production function F ðK,LÞ of capital K and labor L. In this, K
is a capital stock expressed in units of real value, L is the stock of employed labor
expressed in units of labor (often the number of employed laborers). Noteworthy is
that F ðK,LÞ is hypothesized to equal the (yearly) aggregate output Ye of the
economy, which is here a flow variable expressed in units of real value10:
Ye ¼ F ðK,LÞ (1.2)
Ye ¼ ζ K (1.3)
It is well known that the circulation rate ζ is changing remarkably in the course of
the business cycle. It is decreasing when the recession phase sets in and it will
increase when recovery sets in reaching its maximum during the booming phase of
the business cycle. Thus—like Y, e K and L—ζ is a dynamic function of time t:
ζ ¼ ζðtÞ
10
Mark that the definitions of the neoclassical Ye and the evolutionary Y0 ¼ Zdt H ðY0 Þ, in the
present treatise applied, differ markedly. The most conspicuous difference is that Ye is a real value
of output usually accumulating over one year, whereas Y0 is entropy outflow of the executed
statistical experiment over the time-interval (t,t þ dt).
11
In case Ye is NDP, ζ is a net circulation rate per year, which neoclassicists consider to represent the
marginal product of capital under perfect competition. The problems to explain the business cycle
fluctuations of a net circulation rate ζ are not less great than the problems facing the explanation of
the fluctuations of the gross circulation rate considered in the text.
1.9 Production Functions, Marginalism and the Interaction of Inflow and Outflow 23
It is also quite well known that the relative time-dependent fluctuations of ζ are of
the same order of magnitude as the relative variations of output. The problem then
arises how the business cycle fluctuations of ζ affect the change of output Ye and the
change of K . The possibility that Ye is not very much affected and hence that—in
accordance with (1.3)—a decrease of ζ is largely accompanied by an increase of K is
quite beside the mark since it is well known that K is also apt to decline in the
recession phase.
Thus on account of expression (1.3) we expect that—as the recession sets in—
the decrease of ζ will affect a relatively larger fall of Ye. However it is also well
known that the decrease of ζ will not immediately be followed by a decrease of L.
The decline of L will follow with a delay of half to 1 year. Furthermore, after that
delay the decline of Ye will usually diminish or even stop and then expression (1.2)
deprives L to fluctuate very much.
Expression (1.2) does not reflect and may even contradict these observations,
because—if Ye declines relatively much more than K does—it is necessary on behalf
of (1.2) that the decline of Ye is accompanied by a simultaneous decline of L of
comparable magnitude.
[In fact the explanatory deadlock is even worse if we take into account the
role of the parametric exponents a and 1 a with Ye ¼ F ðK,LÞ ¼ b K 1a La and
0 < a < 1]. Therefore expression (1.2) does not reflect an instantaneous dynamic
universal relationship.
To save the neoclassical paradigm, the neoclassical functional relationships are
sometimes advocated as describing economic behavior and trends in the long run.
But this is quite an inadequate manner to deal with the problems. Indeed within the
context of the parametric method of scientific investigation one can always put the
long run defense strategy on the stage to save a theory. If the long run is long run
enough, the differences between the fluctuations of the left side and the right side of
(1.2) level off, but unfortunately this will also immobilize the dynamic import of the
theory. The more the fluctuations are averaging out, the more linear correspondence
there is between the left side variable Ye and the right side variables K and L of the
equation, but—despite that—the less these long run relationships explain. Long run
averaging cannot save a theory.
(Another but related way to save the neoclassical paradigm is to assume that it
holds only under conditions of static growth or by assuming stable or nearly stable
net growth. In that case Y,e K and L are assumed to share the same or nearly the same
net rate of growth. Fluctuations about that rate are completely ignored. Clearly such
assumption is totally unrealistic. It is just as much objectionable as the idea that the
theory has only long run significance.)
That the neoclassical production theory can only be given (weak) economic
significance under the unrealistic condition of long run averaging and/or under the
rigid restriction of static growth, is very much connected with the parametric nature
in which production functions must be stated. For instance we need two parameters
a and b for the Cobb Douglas production function:
Ye ¼ F ðK,LÞ ¼ b K 1a La with 0 < a < 1
24 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
the consumption sector in respective order. As there are two types of outflow,
consumption and investment, a satisfactory theory of inflow/outflow must take into
account that also two types of inflow exist: one that uses up the stock of entropy that
consumption builds up and another that uses up the stock of entropy that investment
builds up. The first stock I shall call labor capacity, the second stock I shall call
entrepreneurial capacity. Thus consumption is labor outflow, the flow of entropy
that increments labor capacity and labor inflow is the flow of entropy that
decrements labor capacity. Similarly, investment is entrepreneurial outflow, the
flow of entropy that increments entrepreneurial capacity and entrepreneurial depre-
ciation is entrepreneurial inflow, the flow of entropy that decrements entrepreneur-
ial capacity.
I shall elaborate on this topic in the only possible manner it can be done (i.e., in
an evolutionary context of economic development) in Sect. 2.1 of Chap. 2 and in
increasingly more detail in subsequent sections and chapters to come.
Labor and entrepreneurial capacity are stock quantities. If outflow is considered
to represent a flow variable, consistent analysis demands that inflow is dealt with on
the same footing as outflow, i.e. as a flow variable as well. Hence a more accurate
understanding of inflow/outflow analysis involves to discern the two inflows of the
two-sector economy: labor inflow (which includes wages and other inflows of the
consumption sector as we will demonstrate in due time) and entrepreneurial inflow
(which includes entrepreneurial depreciation as a measure of the wear and tear of
production resources as well as other inflows of the investment sector as we will
also demonstrate in due time).
In summary then the neoclassical production function does not provide for a
correct and consistent theory of interaction between the two flows of labor inflow
and entrepreneurial inflow on the one hand and the two flows of entrepreneurial
outflow (investment) and labor outflow (consumption) on the other. Let me further
clarify this as follows with the help of the inflow/outflow scheme of Fig. 1.2.
I will hereby apply the following notational conventions in accordance with
what has already been stated in Sects. 1.6 and 1.7:
The respective entropy volumes (per sample) of labor inflow, entrepreneurial
inflow and joint inflow per trial of the two-sector economy, P selected during
(t,t þ dt), are denoted by H ðX1 Þ , H ðX2 Þ and H ðX0 Þ ¼ i H ðXi Þ in accordance
with the Shannon-inspired scheme of Fig. 1.2.
The respective entropy volumes per sample of labor outflow, entrepreneurial
outflow and joint outflow per trial of the two-sector economy, selected during
P
(t,t þ dt), are denoted by H ðY1 Þ, H ðY2 Þ and HðY0 Þ ¼ j H Yj .
Agents select by using up and decommissioning entropy/value and by spending
and commissioning entropy/value. Mark that the subscripts (0, 1 or 2) decide about
to which state (i.e. to which sector) the inflow, respectively the outflow, belongs.
To avoid conceptual vagueness hX0 i and hY0 i are sequences each of Zdt samples,
with each sample consecutively selected over the time-interval ðt; t þ dtÞ,resulting
in the reassembled entropies/real values Zdt H ðXi Þ, respectively Zdt H Yj over
ðt; t þ dtÞ . The time-length of these sequences is always dt. HðXi Þ and H Yj
26 1 Darwin- and Shannon-Inspired Dynamic Economic Selection
CONSUMPTION SECTOR
Net growth of
labor capacity
Production
Y1 = Zdt ⋅ H(Y1) X1 = Zdt ⋅ H(X1)
Outflow Inflow
sequence Y0 sequence X0
Net growth of
entrepreneurial
capacity
INVESTMENT SECTOR
Fig. 1.3 The universal two-sector (binary) model of evolutionary inflow/outflow selection.
The system of equations is non-parametric and derivable from first principles of evolutionary economics.
H(Yi) is the reassembled entropy per sample in state i of the outflow sequence hY0 i. H(Xi) is the reassembled entropy per sample in state i of
the inflow sequence hX0 i
It from bit:
Every it—every particle, every field of force, even the space-
time continuum itself—derives its function, its meaning, its
very existence entirely—even if in some contexts indirectly—
from the apparatus-elicited answers to yes-or-no questions,
binary choices, bits.
John A. Wheeler
[
N
S0 ¼ Si ¼ S1 [ S2 [ S3 [ [ S N (2.1)
i¼1
S0 ¼ S1 [ S2 (2.2)
contributes to the differential growth of the entropy stock of Si over (t,t þ dt). I have
called this process entropy outflow selection.
Likewise let economic agents select inflow HðX0 Þ as being used up in the various
sectors Si ði ¼ 1; 2, 3, N Þ. The probability of the joint occurrence of the event
that a sample is used up in state i on (t,t + dt) and the event that it has been called in
existence after time instant t but prior to its moment of annihilation on (t,t þ dt) is
negligible relative to its singular chance of being used up in state i on (t,t þ dt).
Therefore we may also conclude here that—to first order of accuracy in dt—all the
inflow that is used up on (t,t þ dt) already existed in Si at time instant t. This implies
that the selected inflow H ðXi Þ reassembles entropy in state i on (t,t þ dt) that was
already in that same state i at time t. The reassembled entropy inflow so annihilating
over all Zdt samples of the statistical experiment executed during (t,t + dt) is
Zdt H ðXi Þ. Zdt H ðXi Þ contributes to the differential decrease of the entropy
stock of Si over (t,t þ dt). I have called this process entropy inflow selection.
What is crucial with respect to these statistical experiments of selection is that
not the entire content of entropy of Si is assembled/selected during (t,t + dt). Only
the assembled portion Zdt of the total entropy stock of S0 at time (t + dt) is object/
subject in the course of the statistical experiment of outflow selection during
(t,t + dt). Further only the assembled portion Zdt of the total entropy stock of S0
at time t is object/subject in the course of the statistical experiment of inflow
selection during (t,t + dt).
For the entropy that is reassembled from the assembled entropy it is helpful to
introduce the formal concept of differential sets of entropy. We define the differen-
tial subset dSþj to represent the subset of entropy of Sj stocked at time (t + dt) that
has originated during the infinitesimally small time-interval (t,t + dt) of selection.
Further we define the differential subset dS i to represent the subset of entropy of Si
stocked at time t that will annihilate during the infinitesimally small time-interval
(t,t + dt). This motivates us to say that the entropy “stocked” in dSþ j is Zdt H Yj
and that the entropy “stocked” in dS i is Zdt H ðXi Þ.
With the concepts of differential sets of entropy well defined, inflow selection is
between N alternative outcomes S of a trial/sample: entropy is in dS1 , in dS2 , or
in dSN . In that case dS0 ¼ i dSi is the playground of all the subsets that can be
selected. We call that playground the differential sample space of inflow selection.
Likewise outflow selection is between N alternative outcomes of a trial/sample:
S þ
entropy is in dSþ þ þ þ
1 , in dS2 , or in dSN . In that case dS0 ¼ i dSi is the differential
sample space of outflow selection.
The use of a set S0 (or dSþ
0 or dS0 ) of entropy stock from which entropy is selected
may be clarified with the help of so-called Venn diagrams. This is a handsome tool
to get a more concrete idea of the relationships between the various sets and
variables of entropy and the various probabilities that govern the evolution of
economic systems.
The N sets Si of entropy stock may be presented in a so-called non-differential
Venn diagram. This has been done in Fig. 2.1 for the two-sector economy with
N ¼ 2. Each of the N subsets Si holds a finite entropy stock H ðSi Þ. We may regard
32 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
S1 S2
the surface area of the set Si to represent an entropy stock H ðSi Þ of Si. We shall call
H ðSi Þ the capacity Ci of Si :
C i ¼ H ð Si Þ (2.3)
Ci is all the entropy stocked in Si .1 The outflow Yi ¼ Zdt H ðYi Þ is all reass-
embled entropy that goes into Si during the time-interval (t,t + dt) and the inflow
Xi ¼ Zdt H ðXi Þ is all reassembled entropy that is withdrawn from Si during the
time-interval (t,t + dt). It follows that the increase dCi of capacity Ci is the
entropy of the accumulating surplus of Yi over Xi . That is,
Thus
ðt
Ci ðtÞ ¼ Z ðθÞ ½Hθ ðYi Þ Hθ ðXi Þdθ (2.5)
1
1
Money holds a content of entropy equal to the quotient of the number of money units it represents
and the price level P.
2.1 Non-differential and Differential Sets of Entropy, Venn Diagrams, Sample. . . 33
dSþ þ
1 and dS2 do not overlap.
dS1+
dSþ þ þ
1 [ dS2 ¼ dS0 dS0−
Bottom configuration:
dS
1 and dS2 do not overlap. dS2+
dS
1 [ dS2 ¼ dS0
dS1−
dS0+
dS2−
For the two-sector economy C1 is labor capacity, the measure of the labor force
expressed in bits of entropy. C2 is entrepreneurial capacity, the measure of the
productive potential of the investment sector expressed in bits of entropy. C0 is
overall capacity.
The differential sets dSþ
j and dSi may also be presented in a Venn diagram. This
has been done so in Fig. 2.2 by means of a differential Venn diagram for the two-
sector economy with N ¼ 2 . Here the different sub-areas denote the various
differential subsets. Each differential subset holds the entropy reassembled in it
by the execution of the associated statistical experiment of selection during
(t,t + dt). We may regard the surface area of the subset to represent the entropy
of it, which is Zdt times the entropy per sample of this subset. Note that these
differential sets contain portions of entropy that vanish completely relative to the
infinitesimally greater entropy stocked in S0 . To focus attention on the borders
between the various differential sets, the same diagram has been configured three
times in Fig. 2.2. Whereas the non-differential sets S1 and S2 do not overlap as in
Fig. 2.1 (entropy is here either in state 1 or in state 2 to first order of accuracy in dt),
we notice that this is not the general rule for the differential sets of Fig. 2.2. The
entropy that originates in state j on (t,t + dt) (See the sets dSþ j of the center
configuration) has emerged from entropy that has annihilated both in state 1 and
in state 2 on (t,t + dt). Likewise the entropy that annihilates in state i on (t,t + dt)
(See the sets dS i of the bottom configuration) is bringing forth entropy both
originating in state 1 and in state 2 on (t,t + dt). Thus generally dSþ j and dSi
þ
overlap. In set-theoretical notation: dSj \ dSi 6¼Ø. Herein the symbol \ denotes
the intersection operation and Ø is the null set, which does not contain entropy.
34 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
This raises an important question. In the preceding we have asserted that the
probability of the joint occurrence of the event of a sample to originate on (t,t + dt)
and the consecutive event that this sample ceases existence on (t,t + dt) is negligi-
ble relative to the probability of the separate singular events of origination or of
annihilation of that sample. Does that not imply that there is virtually no chance that
dSþ
j and dSi overlap?
Well it does not really imply that. On the contrary, the joint occurrence of an
event of a sample both to originate on (t,t + dt) and to annihilate on (t,t + dt) is
concerned with the same sample first originating on (t,t + dt) and then afterwards
annihilating on (t,t + dt) or in reversed order. However, in Fig. 2.2 we are dealing
with the concurrent occurrence of a sample originating and a different sample
annihilating.2 The content of the overlapping area of dSþ
0 and dS0 is composed of
þ
samples of assembled entropy of dS 0 and dS0 that are sacrificed on (t,t + dt) to
effectuate (i.e. to reassemble) new samples of entropy within the entire differential
sample space. The chances of that production selection process are not negligible.
The deeper cause of overlapping is the statistical dependence between the elemen-
tary selection events of inflow and outflow. We will explain that in due time. It is
here where Shannon’s selection probabilities enter the scene as we will clarify at a
later stage of this exposition. The situation has been visualized from a slightly
different perspective in the Venn diagram of Fig. 2.3. The total surface area is called
þ
the union. This is the differential set denoted by dS0 ¼ dS 0 [ dS0 . It is also called
the overall differential sample space of selection because all statistical experiments
of selection take place within this sample space.
2
Note that the two samples—despite they are different samples—hold each the same content of
entropy.
2.2 Overall Selection, Transmission and Exchange, Conditional Entropy 35
þ
The differential entropy stocked in overall sample space per trial is H dS 0 [ dS0 ,
but—for closer connection to Shannon’s notational conventions—I will also denote
this differential entropy by H ðX0 [ Y0 Þ.3 The operator [ is the mathematical symbol
for the union operation that we met before in (2.1) and (2.2). We will use the new
notational convention interchangeably with the one with arguments X and/or Y closely
resembling Shannon’s notational convention for entropy. Thus HðdS i Þ ¼ HðXi Þ is
þ
the entropy content of dS i per sample, HðdS j Þ ¼ HðYj Þ is the entropy content of dSþj
þ þ
per sample and HðdS 0 [ dS0 Þ ¼ HðX 0 [ Y 0 Þ is the entropy content of dS
0 [ dS0 per
þ
sample. The new convention with arguments dS i and/or dSj is the more systematic
one, the other one is closer to existing practice. Thus the notational convention of
expressing the entropy content of a set S by HðSÞ as in (2.3) will also be followed for
the entropy content of differential sets. To clarify the production and using up of (or
the feeding on) entropy within the subsets of the Venn diagram of differential
selection further, Fig. 2.4 may be of additional aid for the two-sector economy
and Fig. 2.5 for the three-sector economy. Venn diagrams for any multi-sector
economy (N > 3) can be drawn in a similar way. It does not matter whether the
borderlines between the various subsets are curved or straight or just whimsical.
To help interpreting the Venn diagram, a differential set with a smaller index
will be standard positioned vertically higher in the Venn diagram than a differential
set with a larger index. Thus in Fig. 2.4 the differential subsets of the consumption
sector S1 are at the top of the drawing and the differential subsets of the investment
sector are at the bottom.
More remarks about some conceptual notions of probability theory have been
listed in Appendix C. They may be helpful to understand Figs. 2.4 and 2.5 better.
Thus far we have considered the experiments of inflow selection and outflow
selection. We ascertained that the playground of the inflow experiment is differen-
tial sample space dS 0 and that the playground of the outflow experiment is
differential sample space dSþ þ
0 . Since dS0 and dS0 overlap and together form
þ
differential sample space dS0 ¼ dS0 [ dS0 , the two separate statistical experiments
of selection seem together to reflect more events than the selection events going on
þ
in dS0 alone. To oversee the whole of events within dS 0 and dS0 we must take into
account all the events that occur in dS0 and consider dS0 as the playground of an
overall statistical experiment of selection.
þ
The applicable rule here is that any single trial in sample space dS0 ¼ dS 0 [ dS0
will always involve the concurrent (simultaneous) selection of a sample annihilating
in state i (i ¼ 1; 2) and another sample originating in state j (j ¼ 1; 2). Thus a trial of
3
Shannon uses the notation H(X,Y), which is H(X0 [ Y0) in the notation to which I adhere.
36 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
dS 0
+
B G dS 0−
D
F
A U
Fig. 2.4 The Venn diagram of evolution of the two-sector economy S0 with the number N of
sectors equal to 2.
The non-differential sectors, S1 and S2 , have not been illustrated. The differential set dSþ
0 of outflow interacts with the differential set dS0 of
inflow.
As a standard convention a differential set with a smaller index will be positioned vertically higher in the Venn diagram than a differential
set with a larger index.
A, B, C, D, E, F, G and U are subsets of these differential sets. Their respective entropy contents per trial are H(A), H(B), H(C), H(D), H(E),
H(F), H(G) and H(U). Note that
þ þ þ
dS
0 ¼ dS1 [ dS2 ¼ E [ F [ G [ D [ C [ U, dS0 ¼ dS1 [ dS2 ¼ B [ D [ E [ A [ C [ F
þ þ
dS
1 ¼ E [ F [ G, dS2 ¼ D [ C [ U, dS1 ¼ B [ D [ E, dS2 ¼ A [ C [ F
þ
dS
0 \ dS0 ¼ D [ C [ E [ F
H ðY1 Þ ¼ H(B) + H(D) + H(E), H ðY2 Þ ¼ H(A) + H(C) + H(F), H ðY0 Þ ¼ H ðY1 Þ þ H ðY2 Þ
H ðX1 Þ ¼ H(E) + H(F) + H(G), HðX2 Þ ¼ H(D) + H(C) + H(U), H ðX0 Þ ¼ H ðX1 Þ þ H ðX2 Þ
H ðY2 jX0 Þ ¼ H(A), H ðY1 jX0 Þ ¼ H(B), HðX2 jY0 Þ ¼ H(U), H ðX1 jY0 Þ ¼ H(G)
M
L
C N
D
+
dS 0
K
O
P dS 0−
E
J
B F
H
A G
Fig. 2.5 The Venn diagram of evolution of the three-sector economy S0 with N ¼ 3 sectors. The
three non-differential sectors, S1, S2 and S3, have not been illustrated. The differential set dSþ
0 of
þ
outflow interacts with the differential set dS
0 of inflow . Notice that there are 3 subsets dSi \ dSj
2
dSþ
0 ¼ A[ F[ J[O [ B[ E [K [ N[C [D[ L [M
dS
1 ¼ M [ N [ O [ P, dS2 ¼ L [ K [ J [ H, dS3 ¼ D [ E [ F [ G
dSþ þ þ
1 ¼ C [ D [ L [ M, dS2 ¼ B [ E [ K [ N, dS3 ¼ A [ F [ J [ O
þ
dS
0 \ dS0 ¼ D [ E [ F [ L [ K [ J [ M [ N [ O
H ðY0 jX0 Þ ¼ H(A) + H(B) + H(C), H ðX0 jY0 Þ ¼ H(G) + H(H) + H(P)
H ðY1 Þ ¼ H(C) + H(D) + H(L) + H(M), H ðY2 Þ ¼ H(B) + H(E) + H(K) + H(N),
H ðY3 Þ ¼ H(A) + H(F) + H(J) + H(O), H ðX1 Þ ¼ H(M) + H(N) + H(O) + H(P),
H ðX2 Þ ¼ H(H) + H(J) + H(K) + H(L), H ðX3 Þ ¼ H(D) + H(E) + H(F) + H(G)
There are 15 subsets of differential sample space dS0 ¼ dSþ
0 [ dS0 . For the N-sector economy that number would be N ðN þ 2Þ
Each trial of the overall selection experiment may have therefore N 2 different
outcomes, each occurring with probability qi j ði; j ¼ 1, 2, 3, ,N Þ. On the analogy
of Shannon’s formula (1.1) it must follow that all entropy per trial “stocked” in
þ
overall differential sample space dS0 [ dS0 must be equal to
X
H ð X 0 [ Y0 Þ ¼ qi j log qij (2.7)
i;j
4
That formula (2.7) has a structure equivalent to that of formulas (1.1) need not surprise. Its
justification is based on the same argument that justifies (1.1). We will discuss that justification in
Chap. 4.
38 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
paper (Shannon and Weaver 1963, page 21). Note that Zdt H ðX0 [ Y0 Þ is the
surface area (entropy) of differential sample space dS0 in the Venn diagrams of
Figs. 2.2, 2.3, 2.4 and 2.5.
The process of overall selection and reassembly of Zdt H ðX0 [ Y0 Þ is attended
by a concurrent portion Zdt H ðX0 \ Y0 Þ of assembled inflow and assembled outflow
þ þ
of dS
0 \ dS0 , the intersection of dS0 and dS0 . The assembly of Zdt H ðX0 \ Y0 Þ is
the intersection process by which the entropy of that common portion is transferred
from the past into the future. We shall call the latter process entropy transmission or
exchange. The area that dSþ
0 and dS0 have in common in Fig. 2.3 “stocks” an
entropy flow Zdt H ðX0 \ Y0 Þ. This common area is the intersection of the differ-
ential set of inflow events and the differential set of outflow events. The entropy
Zdt HðX0 \ Y0 Þ of this set is also called transmission (after Shannon). The termi-
nology clarifies exactly what it does: transferring information/entropy from the past
into the future. We shall define the rate of transmission γ as the quotient of
transmission per unit of time and per unit of joint capacity C0 :
Z H ðX0 \ Y0 Þ
γ¼ (2.8)
C0
þ
H dS þ þ
0 [ dS0 ¼ H dS0 þ H dS0 H dS0 \ dS0 (2.9)
5
The concept “virtual sample” has been deliberately chosen here as it appears to play a similar role
as a virtual particle in physics: it does not have an independent existence of itself.
2.2 Overall Selection, Transmission and Exchange, Conditional Entropy 39
þ
HðdS 0 [ dS 0 Þ can also be stated as the sum of the entropy þ transmission
þ
H dS0 \ dS0 per trial, the conditional entropy inflow HðdS
0 dS0 Þ ¼ HðX0 jY0 Þ
þ
per trial of X0 given Y0 and the conditional entropy outflow HðdS0 dS0 Þ ¼ HðY0 jX0 Þ
per trial of Y0 given X0 :
þ þ
H dS þ þ
0 [ dS0 ¼ H dS0 dS0 þ H dS0 dS0 þ H dS0 \ dS0 (2.10)
We will encounter other (conditional) forms of inflow and outflow. We list here
four different forms of inflow and outflow of an economic sector Si :
• Unconditional inflow of the sector Si : Xi ¼ Zdt H ðXi Þ with HðXi Þ ¼ H dS i
þ inflow of the sector Si : Xi ¼ Zdt H ðXi jY0 Þ with H ðXi jY0 Þ ¼
• Conditional
H dSi dS0
• Unconditional outflow of the sector Si : Yi ¼ Zdt H ðYi Þ with HðYi Þ ¼ H dSþ i
outflow of the sector Si : Y i ¼ Zdt HðYi jX0 Þ with H ðYi jX0 Þ ¼
• Conditional
H dSþ dS
i 0
þ
These conditional entropies are formally defined in Appendix B. dS i dSj is that
portion of the subset dSi in the Venn diagram that it does not have in common with
þ
dSþ
j . Likewise dS j dS þ
i is that portion of the subset dSj that it does not have in
þ þ þ
common with dS
i . Clearly dSi dSj , dSj dSi : and dSi \ dSj do not overlap one
another.
þ
The financial consequences of selection manifest themselves within dS0 \ dS0 .
þ
As an example of conditional differential subsets, consider dS1 dS2 in the Venn
diagram of Fig. 2.4. We claim that dSþ þ
1 dS2 ¼ B [ E. To see this, note that dS1 and
dS2 have the subset D in common. Hence since the condition is that dS2 is certain,
the probability of selecting any subset of dS2 is equal
to 1 implying that the subset D
þ
of dSþ contributes the entropy log 1 ¼ 0 to dS 1 dS2 . That is, D is a null subset of
1
þ
þ
dS1 dS2 . What rests is dS1 dS2 ¼ B [ E.□
We are dealing with economic activity as the process of executing statistical
experiments of selection.
Clearly, the probability of an inflow sample to annihilate in state i within sample
space dS 0 on ðt; t þ dtÞ is λi ¼ H ðXi Þ=H ðX0 Þ and the probability of an outflow
sample to originate in state j within sample space dS 0 on ðt; t þ dtÞ is
μj ¼ H ðYi Þ=HðY0 Þ. That is,
H ð Xi Þ H Yj
λi ðtÞ ¼ ; μj ðtÞ ¼ for i; j ¼ 0; 1, 2, 3, ; N (2.11)
H ð X0 Þ H ðY0 Þ
þ
With the introduction of conditional entropy inflows H dS dS and condi-
þ i 0
tional entropy outflows H dSi dS0 we must be aware that these flows offer a range
40 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
þ
of candidates, additional to H dS
i and H dSi , for the definition of the economic
variables such as consumption, investment, wages and entrepreneurial depreciation.
Our task is to find out what the definitional assignments of these flows to the
considered conditional and unconditional entropy inflows and outflows should be.
We cannot solve this puzzle immediately. We need much more mathematical
reasoning to come to grips with it.
What exactly does it mean if uncertainty is at its maximum log N with λi ¼ 1=N?
Well, this reflects the situation that economic agents will not gather more than the
very minimum of knowledge from the selection decisions they take all the time on
the selection interval (t,t + dt). Such minimum reflects the complete lack of
knowledge of the selecting agents. With this in mind recall that inflow selection
is a process of selecting (i.e. assembling) Zdt samples from non-differential sample
space S0 randomly during ðt; t þ dtÞ with the aim to rearrange and reorder (i.e. to
reassemble) them for purposeful reallocation into differential sample space dS 0 . If
there is maximum uncertainty with respect to this achievement, this implies that the
surplus of assembled entropy Zdt P Hm of the Zdt unordered samples from S0 over
reassembled entropy Zdt i λi log λi of the reordered samples within dS 0
vanishes. That is, if there is maximum uncertainty,
X X
Hm þ λ log λi ¼ 0 resulting in Hm ¼
i i
1
iN
log N with λi ¼ N1
2.3 Selection in Non-differential Sample Space of the Multi-sector Economy 41
It follows that Hm ¼ log N. This implies that the assembled Zdt samples from S0
will always get selected each with equal probability 1=N during ðt; t þ dtÞ
irrespective of the state in which they reside. Else the surplus of assembled over
reassembled entropy can never become zero.
We are thus led to the conclusion that samples from non-differential sample
space S0 are each selected (i.e. assembled) with elementary probability 1=N
irrespective of state in case we are dealing with the statistical experiment of inflow
selection.
This was as far as inflow selection is concerned. For outflow selection equivalent
reasoning applies. For sake of completeness and because of the importance of the
argument we will deal with this in the same detail of argumentation as above for
inflow selection.
Outflow selection in differential sample space dSþ 0 is a process of selecting (i.e.
assembling) Zdt samples into non-differential sample space S0 randomly during
ðt; t þ dtÞ with the aim to rearrange and reorder (i.e. to reassemble) them for
purposeful reallocation of entropy into dSþ 0 . Let the assembled entropy of the Zdt
outflow-samples drawn from non-differential sample space S0 be Hmþ per sample.
þ
The reassembly of these samplesP in differential sample space dS0 effectuates the
entropy outflow HðY0 Þ ¼ i μi log μi per sample so that agents exploit a surplus
per sample of assembled entropy Hmþ over reassembled entropy HðY0 Þ on the
selection interval (t,t + dt).
P
Zdt HðY0 Þ ¼ Zdt i μi log μi is a measure of value of the outflow. It is also a
measure of the uncertainty economic agents face when they decide commissioning
economic equipment,
P sources and resources. In this respect greatest uncertainty is
attained if i μi log μi is at its maximum. This maximum is log N and is attained
for μi ¼ 1=N . That is,
X
H ð Y0 Þ ¼ i
μi log μi log N
What exactly does it mean if uncertainty is at its maximum log N with μi ¼ 1=N?
Well, this reflects the situation that economic agents cannot gather more than the
very minimum of knowledge from the selection decisions they take all the time on
the selection interval (t,t + dt). Such minimum reflects the complete lack of
knowledge of the selecting agents. With this in mind recall that outflow selection
is a process of selecting (i.e. assembling) Zdt samples into non-differential sample
space S0 randomly during ðt; t þ dtÞ with the aim to rearrange and reorder (i.e. to
reassemble) them for purposeful reallocation into differential sample space dSþ 0 . If
there is maximum uncertainty with respect to this achievement, this implies also
that the surplus of assembled entropy Zdt Hmþ of the unordered samples from S0
P
over reassembled entropy Zdt i μi log μi of the reordered samples within dSþ 0
vanishes. That is, if there is maximum uncertainty,
X X
Hmþ þ μ log μi ¼ 0 resulting in Hmþ ¼
i i
1
iN
log N with μi ¼ N1
42 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
and it follows that Hmþ ¼ log N. This implies that the assembled Zdt samples from S0
will always get selected each with probability 1=N during ðt; t þ dtÞ irrespective of
the state in which they reside. Else the surplus of assembled over reassembled
entropy can never become zero.
We are thus led to the conclusion that samples into non-differential sample space
S0 are each selected (i.e. assembled) with elementary probability 1=N irrespective of
state in case we are dealing with the statistical experiment of outflow selection.
Furthermore the assembled entropy per sample from/into S0 during ðt; t þ dtÞ is the
same for inflow selection and outflow selection Hm ¼ Hmþ ¼ Hm :
These findings have the following interpretation. Samples were or get stocked in
S0 before/after they are selected. Their assembled entropy during ðt; t þ dtÞ is then
log N per sample. After they have been reassembled the selected samples have
entropy H ðX0 Þ, respectively HðY0 Þ, in S0 .
The surplus of assembled entropy over reassembled entropy is
!
X
Zdt log N Zdt H ðX0 Þ ¼ Zdt log N þ λi log λi in case of inflow selection
i
and
!
X
Zdt log N Zdt H ðY0 Þ ¼ Zdt log N þ μi log μi in case of outflow selection
i
The difference of Zdt ½log N H ðX0 Þ and Zdt ½log N H ðY0 Þ is the net
growth dC0 of the entropy stock C0 ¼ HðS0 Þ of S0 :
Clearly there are H ðSi Þ ¼ Ci bits of entropy stored in Si and H ðSi Þ=log N samples
stocked in Si . Each bit of the sector Si is assembled with probability 1=2.
However this does not imply that a statistical experiment in non-differential
sample space S0 can be executed involving the selection of all the samples stocked
in S0 because such experiment cannot be executed in real time. Agents will actually
only select Zdt samples of S0 in real time during the time-interval ðt; t þ dtÞ, i.e. Z
samples per unit of time, which is Z log N bits of S0 per unit of time. There is no way
for agents to select all the samples stored in S0 during the available time on ðt; t þ dtÞ.
This implies that generally
Ci 1
6¼
C0 N
As asserted in Sect. 1.6 of Chap. 1, I endorse the view expressed by Hodgson and
Knudsen that “an adequate explanation of economic evolution must involve the
three Darwinian principles of variation, inheritance and selection.” (Hodgson and
Knudsen 2006, pages 2, 5; Hodgson and Knudsen 2010, page 45). These three
principles form the cornerstone of biologic, social and economic evolution. “At a
high level of abstraction social and biologic evolution share the same general
principles” (Hodgson and Knudsen 2006, page 14). Thus, accepting that economic
evolution is based on the same general basic principles we must search for the way
of formalizing and interpreting these principles of variation, inheritance and selec-
tion in economic and social evolution.
This is the goal that H&K have set themselves and that is also the starting point
for the formulation of a genuine universal theory of economics in accordance with
the Shannon inspired line of argument I have put forward in the foregoing sections.
The ideas of Hodgson and Knudsen demonstrate a continual but consequent
transition of thought in the course of time. In a sense it resembles part of the related
process I went through since the late 1970s, when I got more and more dissatisfied
with the tenets of orthodox and mainstream economics. It is instructive to follow
this process. I shall therefore reflect and comment on it from my particular
background.
In his work on the evolution of institutional economics (Hodgson 2004) that
anticipates the jointly written papers of Hodgson and Knudsen, which deal with the
basic Darwinian principles of economic evolution, Hodgson affiliates himself
closely with the Veblian institutional account of the evolutionary process. He
makes a distinction between institutions on the one hand: the dynamic and complex
emergent survival structures that continually adapt in the course of time as a result
of evolutionary selection, and on the other hand: individuals, i.e. the agents that act/
select under the formal and informal rules, ideas and conceptions of the institutions
with which they are involved. In this Veblian conception of the economic evolu-
tionary process, replication (the copying process of the carriers of traits, habits,
routines, thoughts and complexity) and selection (of combinations of the carriers
that survive) are conceived as multiple-level processes governing both the devel-
opment and evolution of human institutions and human individuals (Hodgson 2004,
page 193). From the Veblian perspective the role of individuals on the one hand and
institutions on the other are irreducible to one another, but the evolutionary process
manifests itself in a variegated interaction of individuals and institutions. Hodgson
stresses the Veblian notion that the evolutionary process of replication and selection
is a process full of the continual unfolding of emergent properties. That is, it is a
“force” that continuously adapts to changing conditions and so creates ever new
combinations, ever novel forms of complexity and variety, ever changing
properties, causal chains, rules, order and conceptions of institutional nature.
Hodgson remarks that for a successful universal theory of selection it is required
2.4 Variation, Inheritance, Selection, Interactors/Agents, Replicators. . . 45
6
The term replicator was first introduced by Dawkins as a synonym for the gene: the genetic
replicator (Dawkins 1976, 1984, 1988, 1995).
46 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
the loss of the reassembled sacrificed Zdt H ðX0 Þ ¼ Zdt ½HðX1 Þ þ H ðX2 Þ bits in
the course of production during ðt; t þ dtÞ.
The used Zdt H ðX1 Þ bits reduce the stock of labor capacity of S1 . The created
Zdt HðY1 Þ bits replenish/feed that labor capacity to provide for its needs. Labor
capacity is composed of C1 ¼ HðS1 Þ bits in state 1 that have accumulated up till
current time t. [See (2.3)]. This labor capacity is an entropy stock. Its selection,
whether of inflow or outflow character, is decided and executed by the laborers/
consumers. It is at the cost of the labor entropy stock when agents/laborers select to
execute labor and so use up labor capacity. It is at the benefit of the labor entropy
stock when agents/consumers select to consume and so contribute to creating labor
capacity.
The used Zdt H ðX2 Þ bits reduce the stock of entropy representing entrepreneurial
capacity C2 ¼ H ðS2 Þ. The invested Zdt H ðY2 Þ bits replenish/feed that entrepreneurial
capacity to provide for its needs. In this respect entrepreneurial capacity must
be considered to be the aggregate stock of bits in state 2 that has accumulated up till
current time t. This entrepreneurial capacity is an entropy stock. Its selection, whether
of inflow or outflow character, is decided and executed by the entrepreneurs/investors.
It is at the cost of the entrepreneurial entropy stock when agents/entrepreneurs select
to use up their entrepreneurial capacity for production, which is the cause of its wear
and tear. It is at the benefit of the entrepreneurial entropy stock when agents/investors
select to invest.7
I prefer to make here a more explicit semantic distinction between the samples of
selection and the agents who take the decisions to select those samples. Agents are
defined in accordance with the kind of decisions they take or with the kind of tasks
they undertake. That is,
• Investors are the agents who create entropy/value by investing.
• Consumers are the agents who create (labor) entropy/value by consumption.
• Entrepreneurs are the agents who annihilate entropy/value by the depreciation
of assets and by using up production inventories and supplies.
• Laborers are the agents who annihilate (labor) entropy/value by delivering labor
effort.
By taking selective decisions the agents add reordered entropy to or withdraw
reordered entropy from the unordered stocks of entropy they possess or represent
7
It is quite inevitable to make these connections. In fact: if agents/entrepreneurs select to use up
their entrepreneurial entropy, elementary logic and symmetry requirements demand that laborers
select to use up their labor entropy simply due to a related process of wear and tear of the labor
force. The ratio of this scheme of inflow selection is that the selective inflow of labor capacity
requires simultaneously the selective inflow of entrepreneurial capacity. The selected labor
entropy is what laborers wish to use up and the use of entrepreneurial entropy is what
entrepreneurs wish to be done when handling work. Clearly, the events of selecting, whether in
state 1 or in state 2, occur always at the same production-locations as well as simultaneously for
both selection parties (entrepreneurs and laborers). It has therefore all the requirements of a
statistical selection experiment of inflow.
2.4 Variation, Inheritance, Selection, Interactors/Agents, Replicators. . . 47
(the reordering is what we called reassembly). For investors and entrepreneurs this
is the stock of entrepreneurial entropy (capacity) of the investment sector. For
consumers and laborers this is the stock of labor entropy (capacity) of the consump-
tion sector (See Fig. 1.3 and the discussion in Sect. 1.9).
I consider the above division in classes of different agents as the most system-
atic. It provides for a satisfactory definition of agents. However, it is not essential to
allocate tasks to agents in this manner. One may just as well consider firms,
governments and institutions as decision taking agents. This is not the real problem
that we are facing here. The problem is how and by what mechanism the agents take
their selective decisions and how this can be described mathematically.
What follows from the above considerations is that the environment of biologic
evolution corresponds to the set S2 storing the entropy stock of entrepreneurial
capacity in the case of economic evolution. Where biologic evolution cherishes the
two-sector selective interaction between population and biotope (the environment),
the economic equivalent of that type of interaction is the interaction between labor
entropy and entrepreneurial entropy in respective order.
Although this correspondence is what makes Darwinism universally valid for all
kinds of biologic and social evolution, it is also a rather confusing classification
noticing that biologists and economists do not share a common vocabulary to
denote corresponding entities.
Let us now revert to the discussion of the model of evolutionary statistical
selection that Hodgson and Knudsen advocate. Does it fit in the above scheme
and in the affirmative how does it fit?
An important point that H&K claim is that the object of selection is a total of
available replicators (the units of selection) stocked in the interactors, because the
authors are aware that interactors host replicators (Hodgson and Knudsen 2010,
page 123). This corresponds with my observation in Sect. 2.3 of this chapter that
samples are stocked in the non-differential economy of S0 . Together with the
property that samples/replicators have a lifetime (the time they are hosted within
their interactor) this is part of the satisfaction of the third principle of Darwinian
evolution: the transmission (inheritance) of the information that the replicators
carry through time. It is at the end of their life that samples/replicators are selected
for inflow and cease existence. Then the information (entropy) they carry is used up
to serve producing new entropy (i.e. creating other samples/replicators that in turn
carry the information further). The interactor-inflow is formed by the samples/
replicators that are selected to be used up. The interactor-outflow is formed by the
samples/replicators that are selected to be produced.
It is however not completely clear whether H&K treat the environment as
another interactor, i.e. as a collective of economic agencies that participates in
selecting like the labor force does. I think they do not. I will demonstrate that the
denial of the environment as an interactor leads to a stalemate in realizing the
Shannon-Darwin time compression transition.
In accordance with the disputable idea that the environment is not an interactor,
Hodgson and Knudsen regard an interactor as an entity with boundaries between
itself and the environment (Hodgson and Knudsen 2010, pages 166, 240). Thus they
48 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
8
Note that any statistical selection theory demands that there is more than one state of selection,
for if there were only one, there is nothing to select. Hence it is required that there are at least two
different environmental states if interactors lack states.
2.4 Variation, Inheritance, Selection, Interactors/Agents, Replicators. . . 49
• the unconditional joint probability qi j that a sample expires in state i on (t,t + dt)
and another sample originates concurrently in state j on (t,t + dt).
The following general rules hold:
1. The event of expiration of a sample in state i and the event of expiration of
another sample in state j are statistically independent. Likewise the event of
origination of a sample in state i and the event of origination of another sample
in state j are statistically independent. Such independent events are selected
sequentially, i.e. in the process of executing two successive singular trials, one
with the outcome in state i, the other with the outcome in state j.
2. The event of expiration of a sample in state i and the event of concurrent
origination of another sample in state j are statistically dependent. This implies
þ
that the subsets of dSi and dSj overlap (See the Venn diagrams of Figs. 2.2, 2.4
and 2.5) for any i; j ¼ 1; 2, . . . . , N. See the discussion in Sect. 2.2.
Rule 2, which states statistical dependence between the event of expiration of a
sample and the concurrent event of origination of another sample assembled from/
þ
into sample space dS 0 \ dS0 and reassembled from/into overall sample space dS0 is
a very crucial one. If the events of origination on (t,t + dt) were not statistically
dependent on the events of expiration on (t,t + dt) and conversely, evolution will
not be possible as explained in Appendix F.9
Rule 1 states the absence of statistical interaction between the different succes-
sively ordered trials (and micro-events) of inflow selection on (t,t + dt). This is the
bottleneck in H&K’s model of inflow selection. It follows from rules 1 and 2 that
pi;j ¼ λi if it were not that H&K define the selection of the two replicators to be
executed conditional to the state of the environment. What this environmental
condition exactly means is not clear. It does not affect the probability of selection
if the condition is the event of expiration of another replicator in one of the states
the environment may possess. On the other hand it does affect the probability of
selection if the condition is the event of origination of another replicator in one of
the states the environment possesses. The measure of interaction depends on the
environmental state of replicator origination.
Presumably Hodgson and Knudsen had another idea in mind. They might have
assumed that their model of inflow selection reassembles the entropy outflow of the
collective of interactors (including the environment?), in the manner of Shannon’s
formula for the joint entropy H ðX0 [ Y0 Þ per trial of both inflow sequence hX0 i and
outflow sequence hY0 i of the communication channel:
X
H ð X 0 [ Y0 Þ ¼ pi;j log pi;j ?
i;j
9
It is recommended to read Appendix F after Sect. 4.1 has been dealt with.
50 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
Unfortunately this formula does not reassemble entropy outflow if pi;j refers
to the occurrence of the expiration of one replicator in state i conditional to
the expiration of another in state j. The correct expression for the joint entropy
H ðX0 [ Y0 Þ is (2.7), which we have discussed previously in Sect. 2.2.
Another critique on H&K’s inflow selection model is: Why should the stock of
replicators that form the non-environmental interactors demonstrate evolutionary
growth conditional to the state(s) of an external environment/non-interactor? Of
course a surplus of selected outflow over selected inflow will cause the selecting
interactors to grow. However, net growth can only be maintained if there is a
continuous update of the environmental conditions to ever more favorable environ-
mental conditions in the course of time. This requires the entropy of the environ-
ment to change/evolve as well. Well, why should that happen spontaneously? The
entropy of the environment can only evolve if it is also object of evolutionary
selection. This implies that the environment is a replicating stock of replicators
which is also object of inflow and outflow selection. But such a stock cannot evolve
relative to its environmental conditions only. Its outflow can only evolve in
interaction with inflow of all interactors outside and inside the environment. This
requires that we must deal with the evolution of environment and non-
environmental interactors in a symmetrical co-evolutionary manner. In summary
the environment must be an interactor as well. In this connection it is also necessary
to grant states of selection to the collective of non-environmental interactors just as
H&K furnished the environment with states. This implies that we have again
arrived at the Shannon-inspired interpretation of Darwinian selection as suggested
by Figs. 1.1, 1.2 and 1.3.
In conclusion, Hodgson and Knudsen make an untenable separation between
interactors and the environment by treating the environment as an entity that is
something different than an interactor. In the way H&K suggest to apply Shannon’s
theory to economic evolution, it cannot be done.
Replicators are stocked within interactors. If replicators are selected they must
be created or used up in one of the available particular states. For the two-sector
model of evolutionary selection we have the minimum number of states, which is 2.
For the multiple sector model of evolutionary selection there are more than two
states and this implies that a replicator will be selected while coming to reside or
ceasing to reside in one of those multiple states. Replicators are similar but they
may vary in state from one other. Else selection would be impossible. If selected for
outflow in state j, they are created in that state and selected to join the stock of
replicators that reside in state j. This stock is nothing else than the entropy H Sj
of the interactor Sj in state j. Likewise if on inflow selection a replicator is selected
in state i, it is used up and ceases existence in the interactor Si that stocks the
replicators residing in state i.
It follows that we have just as many states as there are different sets of interactors.
It is not clear whether H&K realize that mathematical necessity. They consider
indeed groups of different interactors (group selection) and the possible selection of
multiple “component” replicators existing on different levels, but what they intend to
state with this is not very clear (Hodgson and Knudsen 2010, page 151). Do they
2.5 Bitpulses and the Definition of the Unit of Selection 51
suggest that a replicator may change state during its lifetime and so assume different
levels of state during its lifetime? This cannot be the case because a replicator must
maintain the same state all the time during its existence as we have discussed before.
In Sect. 2.6 we will give other arguments of mathematical necessity why replicators
must adhere to the same state during the time they exist.
The subject of defining the units of selection is perhaps the most difficult topic to
bring to a successful end. It must have a solution that links the Darwinian world of
evolution with the rigorous world of the mathematics of selection. The subject is
surrounded by many pitfalls on both sides of the coin. The problem is not a typical
economic (or social) problem. Biologists have also been heavily involved in a fierce
dispute about the units of biologic selection: genes, genotype, phenotype, a combi-
nation of genes, individuals, even the complete population or whatever else that has
been proposed to be the subject or object of selection. The question is still unsettled
and cause of much confusion among evolutionary biologists. For the greater part
the contrasting views can be reduced to unclear ideas of what the selection process
is all about. Once the selection process has been delimited mathematically accu-
rately and consistently, there is only one way of defining the unit of selection: a bit
of entropy or, which boils down to the same, a sample of assembled entropy log N
stored or to be stored in non-differential sample space S0 and reassembled in the
applicable differential sample space in the course of selection on the infinitesimally
small time-interval ðt; t þ dtÞ.
In the papers of 2006 and 2007 written jointly by Hodgson and Knudsen (2006,
2007), the authors contend that for biologic evolution the Darwinian mechanism of
replication by genetic inheritance warrants the transfer of complexity and variety
from the past into the future. We have already argued in the two preceding sections
that a similar process of replication explains the transfer of complexity and variety
in the case of economic (social) evolution. The conceptualization of the samples/
replicators/bits involved is a key issue in understanding economic evolution.
Hodgson and Knudsen (2007, page 53; 2010, pages 77, 122 and following) opt for a
quartet of features that are required for a suitable definition of the replicator concept:
1. Causal implication, which implies the ability to produce copies.
2. Similarity, which implies that the replicator copy is similar to the copied
replicator.
3. Information transfer, which implies that the generated copy carries and transfers
the same generative information as the copied replicator.
4. (Conditional) generative replication, which implies the generative capacity10 of
each replicator to replicate under initializing conditions.
10
Not to confuse with the conception of capacity as defined in Sect. 2.1.
52 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
Entropy
content
logN
Sample/replicator
in state i
0
time
t–t t+q
random random
instant of instant of
origination expiration
t t+dt
In this and the following section we shall first discuss in more detail to what
mathematical interpretations this leads with regard to the samples/replicators of
evolution. In the previous section we noticed that samples/replicators are or get
stocked in interactors during the time they exist. The interactor within which they
are stocked determines their state of selection.
This results in a universal time-dependent representation of a sample/replicator
as depicted in Fig 2.6. A sample/replicator exists when it carries information, order,
complexity, emergent properties for future transference to comply with the third
feature listed above. The latter process can only occur if a sample/replicator
possesses entropy content. Prior to its time instant t τ of origination the sample/
replicator does not exist and we may thus also contend that its entropy is zero then.
Likewise posterior to its time instant t þ θ of expiration the sample/replicator does
not exist anymore and its entropy is zero as well. However from its moment of
origination until its moment of expiration the sample/replicator exists and possesses
positive entropy content. Let me now list the following three properties of a sample/
replicator that is the subject/object of selection of a particular experiment of
statistical selection:
(a) The entropy content is the same for all samples/replicators during their exis-
tence, irrespective of the state in which they are selected.
(b) The entropy content is constant during the time a sample/replicator exists. It
does not vary with the passage of time.
(c) A sample/replicator remains in the same state during its existence.
2.5 Bitpulses and the Definition of the Unit of Selection 53
Entropy
content
logN
sample/replicator?
0
time
t–t t+q
random random
instant of instant of
origination expiration
t t+dt
the current
time-interval
of observation
Property (c) holds generally for any economy. It has already been discussed and
motivated by requirements of selection consistency in the previous sections. We will
bring forward another forceful argument in favor of this property in Sect. 2.6.
Properties (a) and (b) are perhaps less self-evident but they are a necessary conse-
quence of the second and third feature listed above. The argument is as follows.
Suppose that property (a) does not hold good. As a consequence a copy originating
within any particular state will generally not carry and transfer the same entropy as
the copied sample/replicator. This is incompatible with the second and especially the
third feature listed above. Let us then suppose that property (a) holds true but that
property (b) does not hold true. Since property (a) holds good and the third feature
listed above must be satisfied, all samples/replicators will hold the same entropy at
the moments they are reassembled, i.e. at the time instant t τ they originate and at
the time instant t þ θ they expire, irrespective of the state they get in, respectively
were in. This implies that the entropy of a sample/replicator is only fixed at a
universal constant value at t τ and at t þ θ but varies (or may vary) with the
passage of time for the other time instants in between. I have depicted this situation in
Fig. 2.7. In this figure the entropy of any sample/replicator is assumed to differ from
the constant value it has at the instants of origination and expiration. Well this
dynamic pattern of varying entropy cannot be maintained for two conclusive reasons:
• Mark that only the entropy content at the selection moments t τ and t þ θ of the
various samples/replicators determines the inflow and outflow of the economy.
Hence, the variation of entropy content of a sample/replicator at other time
instants than its initial and final moments of existence is totally irrelevant.
54 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
Content of entropy at intermediate time instants is not selected and hence there
are no economic events that result from it or that have caused it. Moreover the
assumed variation of entropy content does not affect the transference of infor-
mation and complexity in the course of time. Consequently, if the character of
the variation of entropy content of a sample/replicator originating at t τ and
expiring at t þ θ is irrelevant, why not subscribe to the rectangular course it
might take?
• Another even more powerful reason is that the varying course of entropy content
in Fig. 2.7 demonstrates that also portions of entropy of the sample/replicator
originate and are used up at intermediate moments during the time it exists. This
implies that we can split up our original sample/replicator in sub-samples/
replicators that originate and expire during the time the original sample/
replicator exists. Consequently, there are events of entropy origination and
expiration at moments where they were formally mathematically absent.
Clearly, this is not a mathematically consistent way of describing evolutionary
selection. In conclusion, samples/replicators can only possess the same constant
entropy content during their entire life-time.
Having established the rectangular pattern depicted in Fig. 2.6 for any sample/
replicator, we must now conclude that each sample/replicator is the extension in
time of a constant content of entropy as long as it exists in accordance with our
contention in Sect. 2.3. We have established there that a sample has entropy log N.
However we are free to decide whether we wish samples each of log N bits as our
unit of selection or wish bits each stocking a single bit of entropy as our unit of
selection. It does not matter. We will often adhere to the bit as our unit of selection
in the sequel, but sometimes it is more convenient to use a sample of entropy log N
as our unit of selection.
If we choose for a bit as our unit of selection, each unit of selection carries a
single bit during its lifetime. Since the unit of selection has a lifetime it must be
stocked in the state i it is and during the lifetime it exists. This implies that non-
differential sample space S0 hosts all bits with a lifetime. As such these bits are not
yet selected but may be selected during ðt; t þ dtÞ. Pending selection this stock of
entropy represents assembled entropy rather than reassembled entropy, because as
long as bits are not selected on ðt; t þ dtÞ they remain hosted in S0 . Reassembled
entropy is the total of those bits that get reassembled in differential sample space.
However the number of bits in differential sample space is infinitesimally small in
comparison with the number of bits in non-differential sample space S0 . Hence the
bits of S0 that hold reassembled entropy form only a negligible differential portion
of the total non-differential entropy stocked in S0 . Reassembled entropy in S0 can
therefore be neglected relative to the rest of the stock in S0 . In conclusion the
economy S0 hosts all the bits currently existing and generally the sector Si hosts in
fact all Ci samples/replicators that are in state i.
In sum bits must have the dynamic form of rectangular pulses extending in time,
each pulse carrying the constant entropy content of one bit during its existence and
0 bits when it does not exist. We may just as well then replace the term bit as object
2.6 The Uniqueness of State of a Bitpulse 55
of selection by the term bitpulse carrying one bit of entropy in the course of time.
Thus the unit of selection is the equivalent of a single bitpulse carrying one bit of
entropy during its existence.
It is important to realize that the instants t τ of origination and the instants t þ θ
of expiration of sets of bitpulses are continuous random variables. At observation
time t it is of course impossible to know these instants in an exact determinist manner
for any of the bitpulses. Statistical selection theory can only establish relationships
between the aggregates of an economy. It does not predict how individual constitu-
ent entities at the micro-level behave deterministically. On the basis of the
relationships between the aggregates we can however derive the probability
distributions of the random variable τ of current lifetime and the random variable
θ of excess lifetime for each of the aggregate inflow and outflow variables. We shall
consider this subject in Chap. 5. It will deliver new starting-points for analysis of the
money economy and of the relationships between transmission and price inflation.
It is clear that the rectangular bitpulse of Fig. 2.6 shares the four features that
H&K discern in line with the three principles of Darwinian selection.
Moreover interactors can be conceived as holding stocks of information
(or stocks of many bitpulses). Interactors (otherwise than bitpulses) hold different
quantities of entropy varying and often growing in the course of time. On the other
hand bitpulses vary with respect to their instants of origination and expiration and
each of them stocks just a single bit of entropy during the time it exists.
Entropy
content B
in small
bits (B–A) sub-bitpulses in state j
one small bit
A
sub-bitpulses in state i
0
time
t–t t+q
random random
instant of instant of
origination expiration
t t+dt
choose the unit of entropy infinitesimally small so that the approximation of the old
bitpulses by infinitely many of B sub-bitpulses can be made as accurately as we
wish. In case the argument of the Shannon-Darwin transition operation has not
convinced to rule out the possibility that still bitpulses may exist that reside in more
than one particular state, the above consideration demonstrates that the number of
such bitpulses can always be shown to be relatively vanishing provided the unit of
entropy is chosen infinitesimally small. This infinitesimally small nature of the unit
of entropy is necessary for other reasons, a subject that has been discussed before
and that will return again and again.
We have noticed in Sects. 2.1 and 2.3 that we can discern between three different
kinds of statistical experiments of selection; one is the inflow selection experiment,
another the outflow selection experiment and the third is the overall selection
experiment. A fourth process playing a role for successful selection is exchange.
The successive Zdt mutually independent trials of the inflow experiment effec-
tuate Zdt draws of samples each expiring in a particular state within sample space
dS
0 . We noticed that according
P to Shannon the reassembled entropy inflow of this
experiment is H ðX0 Þ ¼ i λi log λi per trial [See Sect. 1.6 and (1.1)]. Likewise
2.7 Elementary Probability, State Probability, Some Formulas of Conditional Entropy 57
the successive Zdt mutually independent trials of the outflow experiment effectuate
Zdt draws of samples each originating in a particular state within
P sample space dSþ 0.
This delivers the reassembled entropy outflow H ðY0 Þ ¼ j μj log μj per trial [See
also Sect. 1.6 and (1.1)]. Finally the successive Zdt trials of the overall experiment
effectuate Zdt concurrent draws of a sample to annihilate in any particular state and
another sample to originate concurrently in any particular state within sample space
þ
dS0 \ dS0 . This delivers,P also in accordance with Shannon, the reassembled
entropy H ðX0 [ Y0 Þ ¼ i;j qi j log qi j per trial [See Sect. 2.2 and (2.7)].
Formulas (1.1) and (2.7) are asymptotically valid for dt ! 0 and require the
number Zdt of trials to tend to infinity on (t,t + dt), conditions that we encountered
before as necessarily to be satisfied.
We will come to see in Sect. 4.1 of Chap. 4 at a later stage of analysis why these
formulas (1.1) and (2.7) are justified to hold generally. Shannon justifies these
equations in his seminal paper by an argument subjected to the typical ergodic
condition of the communication channel (Shannon 1948). I will produce a related
and certainly not less convincing general justification of these Shannon formulas
with necessary allowance for the typical non-stationary condition of evolutionary
selection.
In a micro-world of events where there are no other events than events of
elementary statistical selection, the probabilities λi (respectively μj and qi j) associated
with the probability of executing a single trial must be equal to the proportion of
þ þ
entropy residing within dS
i (respectively within dSj and within dSi \ dSj ) to the
þ
total entropy residing within sample space dS 0 (respectively within sample space dS0
þ
and within sample space dS0 \ dS0 ). Thus, in accordance with what has been
remarked11 in Sect. 2.2:
H ð Xi Þ H Yj
λi ðtÞ ¼ ; μj ðtÞ ¼ for i; j ¼ 0; 1, 2, 3, ; N (2.11)
H ð X0 Þ H ðY0 Þ
and
H X i \ Yj
qi j ¼ for i; j ¼ 0; 1, 2, 3, , N (2.6)
H ð X 0 \ Y0 Þ
11
The relationships (2.11) and (2.6) have also been listed in Appendix B. Equation (2.6) has been
listed as (B.9) in Appendix B.
58 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
it follows that
X X
Zdt λ log λi Zdt μj log μj
Pr dS ¼2 i i ; Pr dSþ ¼2 j
0 0
and
X
Zdt qij log qij
PrfdS0 g ¼ Pr dS
0 [ dSþ
0 ¼2 i;j
2.7 Elementary Probability, State Probability, Some Formulas of Conditional Entropy 59
P P
Since H ðXi Þ ¼ λi j λj log λj and HðYi Þ ¼ μi l μj log μj on behalf of (1.1)
and (2.11) we have
X X
Zdt λi λ log λj Zdt μi μj log μj
Pr dS ¼2 j j ; Pr dSþ ¼2 j
i i
H dS
i HðXi Þ log Pr dSi
λi ¼ Prob dSi ¼ ¼ ¼
for i ¼ 0, 1, 2, , N
H dS0 H ðX0 Þ log Pr dS
0
and
h n oi
n o H dSþ
j H Yj log Pr dSþ
j
μj ¼ Prob dSþ
j ¼ þ ¼ ¼
for j ¼ 0, 1, 2, , N
H dS0 HðY0 Þ log Pr dSþ0
Let us further recall here Shannon’s definition of the conditional entropy HðY0 jX0 Þ
of the outflow sequence hY0 i . Shannon defined this conditional entropy as the
average of the entropy of Y0 for each value of X0 , weighted according to the
probability of getting that particular X0 .12 Similarly, the conditional entropy HðX0 jY0 Þ
of the inflow sequence hX0 i must be defined as the average of the entropy of
a sample of hX0 i for each value of the associated sample of hY0 i, weighted accord-
ing to the probability of getting that particular sample value of hY0 i . (See for
more elaborations Appendix B). It follows from the probability definitions in
Appendix B that
X X
H ðY0 jX0 Þ ¼ qi j log μ ðjjiÞ and H ðX0 jY0 Þ ¼ qi j log λ ðij jÞ (2.14)
i;j i;j
12
Mathematically more sound is: “Conditional entropy H ðY0 jX0 Þ is the average of the entropy of a
sample of hY0 i for each value of the associated sample of hX0 i , weighted according to the
probability of getting that particular sample value of hX0 i.”
60 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
and
X X
H ðX0 jY0 Þ ¼ qi j log qi j þ qi j log μj
i;j i;j
P P
Since j qi j ¼ λi and i qi j ¼ μ j ,
X X
H ðY0 jX0 Þ ¼ qi j log qi j þ λi log λi
i;j i
X X
H ðX0 jY0 Þ ¼ qi j log qi j þ μj log μj
i;j j
X
H ð X 0 \ Y0 Þ ¼ H Xi \ Y j
i;j
The entropy Zdt H Xi \ Yj is a transmission. It is an “outflow” of entropy from
the sector Si into the sector Sj . We noticed in Sect. 2.2 that transmission/exchange
involves the creation and annihilation of a pair of balancing samples, a “pure” entropy
sample in exchange for a virtual sample of money. On balance it seems that nothing
has been created in the economy of S0 , because per trial we gain one sample of
entropy in dSþ
j while losing another in dSi during ðt; t þ dtÞ or conversely. Thus it
appears more appropriate to call this transmission financial entropy, involved with
the exchange of entropy in state j for money in state i or conversely. Formulated more
2.8 Transmission, “Pure” and Financial Value, Inflow and Outflow, the Unit Price 61
þ
accurately, in dSi \ dSj a sample at time t þ dt, created in state j on ðt; t þ dtÞ
exchanges for an equal entropy content of money sacrificed in state i on ðt; t þ dtÞ. Or
þ
alternatively, in dS
i \ dSj a sample at time t, annihilating in state i on ðt; t þ dtÞ
exchanges for an equal entropy content of money created in state j on ðt; t þ dtÞ.
Which of the two possibilities is at stake will be subject of a later discussion. That
discussion requires a lengthy mathematical argumentation.
þ
These considerations allow us to call the nature of entropy H dS
i \ dSj of
dS
i \ dSþ
j financial.
On the other hand H dSþ i can be considered to reflect a gross flow of “pure”
entropy production. Therefore we associate H dSþ i with the complete collection of
bitpulses originating or with the (unconditional) entropy outflow of economic
production within Si . By the same token H dS i is a gross flow of “pure” entropy
loss. We are therefore led to associate H dSi with the complete collection of
bitpulses sacrificed or with the (unconditional) entropy inflow of economic produc-
tion within Si .
The difference of outflow Yi ¼ Zdt H ðYi Þ and the conditional outflow
Yi ¼ Zdt H ðYi jX0 Þ is equal to the financial outflow Zdt H ðX0 \ Yi Þ. The difference
of inflow Xi ¼ Zdt HðXi Þ and the conditional inflow Xi ¼ Zdt H ðXi jY0 Þ is
equal to the financial inflow Zdt H ðXi \ Y0 Þ.
Conditional entropy outflow is that portion of entropy outflow that it has not in
common with inflow and conditional entropy inflow is that portion of entropy
inflow that it has not in common with outflow.
Equations (2.16) and (2.17) can easily be verified with the assistance of a Venn
diagram.
þ
Selected entropy outflow balances selected entropy inflow within dS 0 \ dS0 .
Thus there is no net entropy gain or loss involved in the transmission from Si to Sj.
It is just a matter of financially exchanging entropy within Si for entropy within Sj
during the time-interval (t,t + dt) (See Fig. 2.9). That scheme is completely
symmetric with respect to an interchange of X and Y.
So far all variables introduced have been stated in the dimensional unit of
entropy: bits. This is the unit of selection we need for analyzing the processes of
selection. On the other hand economic exchange is in money units. To convert flow-
variables stated in bits into their corresponding flow-variables stated in the ruling
money units we must multiply entropy by the unit-price of entropy currently valid
on the selection interval ðt; t þ dtÞ. For that purpose we define P ¼ PðtÞ to represent
the unit price of a bit of entropy at current time t. The similarity feature that the unit
of selection must necessarily satisfy forces P to be the same for any bit of entropy
irrespective of the state in which it may reside (See Sect. 2.5, feature 2).
62 2 Sets of Entropy, Selection, Venn Diagrams and Bitpulses
1
state i
0
time
1
state j
0
time
t t+dt
It follows that
money outflow ¼ PYi ¼ PZdt H ðYi Þ; money inflow ¼ PXi ¼ PZdt H ðXi Þ
Anticipating more detailed analysis in Chaps. 5 and 6 we mention here that the
following interpretations will ultimately take form:
money wages ¼ PZdt HðX1 jY0 Þ; money depreciation ¼ PZdt H ðX2 jY0 Þ
These relationships hold for flow variables of which the action is restricted to
ðt; t þ dtÞ, certainly not for flow variables covering one year as in the annual reports
of economic statistical agencies.
2.9 Capital and Liquidity Surplus 63
Let us next define the net growth of capital of Si as the difference of the conditional
outflow Z H ðYi jX0 Þ per unit of time and the conditional inflow Z H ðXi jY0 Þ per unit
of time. We shall denote capital by Ci . Then the net change dCi of capital is
In accordance with (2.4) dCi ¼ ci Ci dt is equal to Zdt ½HðYi Þ HðXi Þ. This is the
sum of two contributions: the non-financial component Zdt ½HðYi jX0 Þ H ðXi jY0 Þ
of net entropy growth and the financial component Zdt ½HðX0 \ Yi Þ H ðXi \ Y0 Þ
of entropy growth. That is,
The difference between dCi and dCi is a difference between the entropies of two
subsets of the transmission domain:
Whereas dCi is a net cash flow of entropy capital entering the sector Si during the
time-interval (t,t + dt), the difference d Ci dCi is an additional surplus of money
resources .I shall call this surplus the liquidity flux of the sector Si. It has the features
that are usually assigned to liquid resources.
Recall [See (2.5)] that the accumulated capacity Ci is equal to
ðt
Ci ðtÞ ¼ ðYi Xi Þdθ
1
Capacity growth dCi is the “pure” component of net entropy growth. dCi includes
also the net contribution of a virtual money component to entropy growth. dCi is the
surplus of outflow Yi ¼ Zdt H ðYi jX0 Þ over inflow Xi ¼ Zdt HðXi jY0 Þ of the
economic sector Si during (t,t + dt). Integration of d Ci dt over the time domain
yields
ðt
Ci ðtÞ ¼ ZðθÞ ½Hθ ðYi jX0 Þ Hθ ðXi jY0 Þdθ
1
Since dCi and dCi are flows of entropy-change of capacity and capital on the
selection interval ðt; t þ dtÞ, it follows that PdCi and PdCi are the corresponding
flows of money capacity change and money capital change. PdCi is the surplus of
money outflow PZdt H ðYi Þ over money inflow PZdt H ðXi Þ:
PdCi is the surplus of money outflow PZdt H ðYi jX0 Þ over money inflow
PZdt H ðXi jY0 Þ:
Thus
Since d C0 and dC0 balance, the economy S0 is without an additional liquidity
surplus. For the sectors Si it is different. What is lent by a sector Si is borrowed by the
other sectors of the economy. And what is borrowed by a sector Si is lent by the
other sectors of the economy.
For the two-sector economy (2.18) and (2.4) state that the financial rewards/costs
of the consumption sector are counterbalanced by financial costs/rewards of the
investment sector. It is not difficult to see, by consulting a Venn diagram, why dCi
and dCi do not balance for i 6¼ 0:
þ
H dS H dS þ
i i dS0 ¼ H dSi \ dS0 and
þ þ
H dSi H dSi dS0 ¼ H dS0 \ dSi þ
þ
þ
Since H dS
i \ dS0 6¼ H dS0 \ dSi it follows immediately that
þ
H dSþ H dS 6¼ H dSþ
i i i dS0 H dSi dS0
Hodgson and Knudsen attach a specific meaning to complexity. The definition they
give (Hodgson and Knudsen 2010, formula (61) on page 128) is not so much a
definition of complexity but has another relevant interpretation after some
inaccuracies, that the authors did overlook, have been fixed. In the unadjusted
form proposed by H&K it cannot be fitted in a consequent mathematical theory
of evolutionary selection. I shall sum up what is wrong with it just to clarify how
Shannon’s theory should not be interpreted and how it should be properly
interpreted when making the transition from Shannon’s ergodic selection to
nonergodic Darwinian selection.
Hodgson and Knudsen begin their exposition on complexity with describing a
notional generative replicator as a binary sequence (string) of bits with a length of
Zdt bits.1 (Hodgson and Knudsen 2010, page 128). This description results in an
immediate conflict with what we have established in the foregoing with respect to
the properties of replicators. We have seen in Fig. 1.2 that the inflow and outflow
sequences of the two-sector economy are binary strings of samples of infinitesi-
mally small time-length dt. Such sequences consist of Zdt samples and each sample
represents a replicator, which is either in state 1 or in state 2 in case of the two-
sector economy. Samples/replicators on the other hand extend over finite time (their
lifetime) as we have argued in Sect. 2.5 (See also Figs. 2.6 and 2.9) but they remain
in the same state during their lifetime. Clearly, an inflow sequence hX0 i is not the
equivalent of a replicator, but consists of very many replicators (Zdt in number) that
annihilate during the time-interval ðt; t þ dtÞ of selection. We should also bear in
mind that the total lifetime of a replicator is finite, but that the time-length of the
sequences is T ¼ dt after the Shannon-Darwin transition operation. However H&K
mistake a replicator for a binary sequence (of bits or of samples; that does not
1
H&K use the symbol L to denote the sequence length and the symbol C to denote complexity. In
accordance with the notational conventions to which I adhere L is labor, Zdt is the sequence length
and C will be reserved to denote capacity. I will therefore not use the symbol L for sequence length
and replace it by Zdt in order to avoid confusion. Likewise I will not use the symbol C to denote
complexity but replace it by the symbol K in this section.
3.1 The Reduction of Uncertainty by Selection 67
They point out that they are here concerned with binary sequences of informa-
tion. Thus there are only two alternative states 1 and 2 to select. This implies that
there are only two alternative interactors: population (state 1) and environment
(state 2). Thus, if p is the probability of selecting a replicator in state 1, ð1 pÞ is
the probability of selecting a replicator in state 2. Then, following the definition of
HðXÞ that H&K propose, we arrive at HðXÞ ¼ ð1 pÞ logð1 pÞ p log p. This
is Shannon’s formula for the joint entropy inflow H ðX0 Þ per trial of the binary
selection model with chance λ1 ¼ p that state 1 is selected and chance λ2 ¼ 1 p
that state 2 is selected. If there is noise on the channel, the probability μ1 that state 1
is received at the outflow side is different from p. However, if the channel is
noiseless, μ1 ¼ λ1 ¼ p. This is the context in which Shannon proposed his theory
for the discrete channel (i.e., the channel with a finite number of states).
On the other hand Hodgson and Knudsen associate HðXÞ with the entropy of the
population only (while it actually should be the joint entropy inflow per trial of
population and environment together). Moreover the entropy of the population is a
stock quantity rather than a flow quantity like entropy inflow is. Further H&K state
that—by maximizing interactor fitness in the environment—the notional generative
replicator realizes the theoretical entropy value Hmax ¼ Zdt bits. The “maximization
of fitness” is the typical jargon of evolutionary biologists. Usually the term “fitness”
lacks universally applicable concreteness and preciseness, certainly within
an economic Shannon-inspired context. In terms of Shannon’s notational
preferences HðXÞ cannot be anything else than the entropy inflow per sample.
2
It may be that H&K understand by locus i the i-th sample of the sequence of Zdt samples.
However that would imply that there are Zdt different probabilities pi, which is incompatible with
the fact that we are here dealing with a binary sequence. Hence I must assume that “probability at
locus i” must be interpreted as the probability of being selected in state i of which there are only
two: state 1 and state 2.
68 3 The Road from Generalized Darwinism to Evolvodynamics
P
Observe that HðXÞ ¼ i λi log λi is a maximum if HðXÞ ¼ 1 . It follows that
the maximum of HðXÞ to which H&K refer is 1. If HðXÞ acquires that maximum,
all the pi ¼ λi share the same probability. This probability is ½ in case of the two-
sector economy. However generally 0 HðXÞ 1. Thus if Hmax is to represent the
maximum attainable value of HðXÞ, it must follow that Hmax ¼ 1. Pushing this
further by substitution in H&K’s equation (3.1) above, we get
X
K ¼ 1 HðXÞ ¼ Zdt þ i
pi log pi
P
Since HðXÞ ¼ i pi log pi in this binary case, it follows that Zdt ¼ 1, which
implies that the experiment of selection consists of one single trial only.
To avoid this inconsistency it is necessary to assume that H&K actually intend to
measure the entropy content over all Zdt samples of statistical selection.
P Thus it
appears that we must replace HðXÞ by Zdt HðXÞ ¼ Zdt i pi log pi and
the formula for Hmax by Zdt, the maximum that Zdt HðXÞ can attain. In that case
H&K’s formula (3.1) for complexity K must be replaced by
X
K ¼ Zdt Zdt HðXÞ ¼ Zdt þ Zdt i
pi logðpi Þ
P
Recall our assertion in Sect. 2.3 that Zdt HðXÞ ¼ Zdt i pi log pi is the uncer-
tainty that agents face knowing the probabilities pi only and that Zdt Hm ¼ Zdt log N
is the maximum uncertainty there is if the selecting agents lack all information about the
probabilities of the states [See expression (2.12)]. This implies that Hmax ¼ Hm in case
of the two-sector economy ðN ¼ 2Þ. Thus while selecting on ðt; t þ dtÞ, the selecting
agents succeed to reduce their uncertainty by
X
Zdt log N Zdt HðXÞ ¼ Zdt log N þ Zdt i
pi log pi
Let us reconsider the example as illustrated in Fig. 1.2 and as dealt with in Sect. 1.7
with Zdt ¼ 21, λ1 ¼ μ1 ¼ 14=21, λ2 ¼ μ2 ¼ 7=21.
We shall first restrict the actual selection process to Zdt ¼ 21 samples so that—in
accordance with our exposition in Appendix A—we cannot apply Shannon’s formula
(1.1) because of the insufficient sequence-length. Instead formula (A.1) of Appendix A
is applicable and we establish that Zdt HðX0 Þ ¼ 16:827243 bits, Zdt HðX1 Þ ¼ λ1
16:827243 ¼ 11:218162 bits and Zdt HðX2 Þ ¼ λ2 16:827243 ¼ 5:609081 bits.
The sequential order in which the states are arranged in the inflow sequence hX0 i
does not affect the entropy inflow Zdt HðX0 Þ ¼ 16:827243, reassembled
by the
agents on ðt; t þ dtÞ. Neither does it affect the state probability Pr dS0 , which is
Pr dS
0 ¼ 2ZdtHðX0 Þ ¼ 216:827243 ¼ 8:59993 106
Hence we may consider all combinations of hX0 i that can be reassembled from
14 samples in state 1 and 7 samples in state 2 as equally likely. Below I have listed a
very few of these combinations. Among them also the combination
121121211212111211112 of samples (the fifth one) that is actually selected in
Fig. 1.2.
222222211111111111111
222222121111111111111
222222112111111111111
222221122111111111111
121121211212111211112
122121211212111211111
etcetera
70 3 The Road from Generalized Darwinism to Evolvodynamics
All of these combinations have the same state- probability 8:59993 106 and
share the same information content logð8:59993 106 Þ ¼ 16:827243 bits
within sample space dS 0 , because each of them has the same elementary inflow
probabilities: λ1 ¼ 14=21 and λ2 ¼ 7=21.
There are quite many different inflow combinations that can be realized given
the elementary inflow probabilities λ1 ¼ 14=21 and λ2 ¼ 7=21. We shall call them
variations. All of them share the same set of elementary probabilities. Let their total
number be Ω. That one of them will be selected from sample space dS 0 by the
selecting agents during ðt; t þ dtÞ is certain. Hence, since each of these variations has
the same state probability 8:59993 106, we must have that Ω 8:59993 106 ¼ 1.
!
It follows that Ω ¼ 7 !2114 ! ¼ 2
16:827243
¼ 116; 280 different inflow variations exist
that can actually be selected.
This implies that the assembly of 21 samples by agents on ðt; t þ dtÞ effectuates a
reassembly of economic resources such that 116,280 reassembled combinations of
inflow can be formed. Only a single combination of those 116,280 combinations is
actually selected. Its entropy is logð8:59993 106 Þ ¼ log 116; 280 ¼ 16:827243
bits. This entropy content reflects the economic significance/value of inflow. It
reflects also the uncertainty the selecting economic agents face because the more
variations that can be reassembled, the less firm they are in their decision to choose
the combination actually selected.3 Therefore Shannon uses the terms entropy,
information and uncertainty interchangeably.
Well, there are 116,280 equally likely combinations that can actually be selected
by the agents, given the 21 samples of the inflow selection experiment and given
λ1 ¼ 14=21 and λ2 ¼ 7=21. However, what about this number of combinations if the
elementary probabilities λ1 and λ2 are not given, i.e., are actually unknown? If
nothing is known about the probabilities λ1 and λ2 , we can only guess that the
following combinations of inflow can be formed in case of the two-sector economy:
111111111111111111111
111111111111111111112
111111111111111111121
111111111111111111122
111111111111111111211
111111111111111111212
111111111111111111221
etcetera
In this particular case 221 ¼ 2; 097; 152 different combinations can be formed.
This is much more than the 116,280 combinations that can actually be selected.
I call the 2,097,152 combinations the potential variations and the 116,280
combinations the typical variations. A potential variation requires 21 bits of
3
We may also argue that the more improbable the selected combination is, the more uncertain the
selecting agents are.
3.2 Combinations, Variations, Uncertainty and Information 71
entropy, which is nothing else than the number Zdt ¼ 21 of samples. It is the
uncertainty the economic agents face if they lack all the knowledge they need in
order to take purposeful selection decisions.
So far we have assumed in the example of this section that selection is
restricted to 21 samples in sample space. This had the advantage that the numbers
of typical and potential variations can be overseen and easily calculated respec-
tively from formula (A.1) and from 2 raised to the power of 21. However,
the actual selection process is concerned with a very large number Zdt of samples.
In that case Shannon’s formula (1.1) is applicable. If there are one million as many
samples we have Zdt ¼ 21;000;000 whereas λ1 ¼ μ1 ¼ 14=21 , λ2 ¼ μ2 ¼ 7=21
remain unchanged. It follows now from (1.1) that Zdt HðX0 Þ ¼ 19;284;212, Zdt
H ðX1 Þ ¼ λ1 19;284;212 ¼ 12;856;141 and Zdt HðX2 Þ ¼ λ2 19;284;212 ¼
6;428;071.
The sequence length of 21,000,000 is too large to write out the sequence of a
particular typical variation here. The total number Ω of typical variations is also
enormous: Ω ¼ 219;284;212 . This is a number of the order of 10 to the power
5,805,126,255. The number of potential variations is still much greater:
Ω ¼ 221;000;000 , which is a number of the order of 10 to the power 6,321,629,909.
Hence the number of typical variations is 10 to the power 516,503,654 as small
as the number of potential variations.
Shannon’s formulas (1.1) and (2.7) of asymptotic approximation are determined
by the elementary probabilities λi , μj , respectively qi j . The basic Shannon claim is
that—since the number of samples is always very very large—all the knowledge/
þ þ
information of differential sample space, dS
0 , dS0 , respectively dS0 \ dS0 , is
contained in the elementary probabilities. No more information/entropy per sample
can be gathered than what can be extracted from the knowledge structure of the
elementary probabilities. Thus, given the elementary probabilities, there is no
additional information per sample to gain and/or to lose by the selecting agents if
they choose another sequential order in which all the samples in state 1 and in state
2 are reassembled over the available sequence-length.
Another important observation in favor of Shannon’s fundamental claim is that
all samples are executed on the time-interval ðt; t þ dtÞ of infinitesimally small
length dt. That time-length is so very small that it is inconceivable that the order in
which the samples of different state are being selected by a great variety of agents
can affect the complexity in which the economy is being organized. The length of
the time-interval dt is so small that, economically, all Zdt trials take place
simultaneously. Thus in fact differences in the sequential order by which the states
are distributed over the samples, given the elementary probabilities of selection,
cannot have an economic effect. It is only changes of the elementary probabilities
of λ1 and λ2 that matter.
Note further that the maximum
P uncertainty Hm of 21,000,000 bits is attained by
maximizing HðX0 Þ ¼ i λi log λi ¼ λ1 log λ1 ð1 λ1 Þ logð1 λ1 Þ with
respect to λ1 ¼ 1 λ2 . It is easy to see that the maximum is attained for λ1 ¼ λ2 ¼ 12.
72 3 The Road from Generalized Darwinism to Evolvodynamics
Xi
¼ H ðXi jY0 Þ ¼ λi H ðX0 jY0 Þ (3.2)
Zdt
and
Yj
¼ H Yj jX0 ¼ μj H ðY0 jX0 Þ (3.3)
Zdt
Hence
μj H ðY0 Þ ¼ H Yj jX0 þ μj HðX0 \ Y0 Þ
so that
H Yj jX0 ¼ μj ½H ðY0 Þ H ðX0 \ Y0 Þ ¼ μj HðY0 jX0 Þ
It follows that μj is the quotient of H Yj jX0 and H ðY0 jX0 Þ. Further, in virtue of
(2.11),
H Y j j X0 H Yj Yj
μj ¼ ¼ ¼ (3.6)
H ðY0 jX0 Þ H ðY0 Þ Y0
Hodgson and Knudsen hang on to the four features listed in Sect. 2.5 as a general
definitional description of the replicator. They pay much attention to discussing the
matter (Hodgson and Knudsen 2010, Chap. 6). They contend that the first three
(causality, similarity and information transfer) of these four features are necessary
but not enough (Hodgson and Knudsen 2010, page 119). To complete the definition
they add the mechanism of generative replication as a fourth definitional prerequi-
site for the replicator. They assert that this generative property is a necessary feature
to exclude the possibility that e.g., a paper copying machine can also be called a
replicator (Hodgson and Knudsen 2010, page 121).
74 3 The Road from Generalized Darwinism to Evolvodynamics
Well the latter statement is not without some contradiction because the action of
a paper copier may also contribute to evolutionary growth as far as it invites users to
exploit increasingly innovative usage of the copier.
More particularly, evolutionary progress is not the sole result of outflow (i.e.,
generative) selection, but also the result of inflow (i.e., degenerative) selection.
Evolutionary systems do not only increase in complexity, they may also
decrease in complexity. Both processes occur simultaneously and one may
dominate the other for some time. Hence statistical selection may also result
in the decrease of aggregate complexity. That decrease may predominate in the
end. This is the phase when species become extinct, often after a period of
prosper (In fact this has been the overwhelming rule with respect to the
evolution of life on earth since many millions of years. The species that live
now on earth form a tiny fraction of all the species that have ever lived on
earth). Likewise institutions, economic enterprises and social organizations will
follow a similar pattern of initial evolutionary prosper and ultimate decline in
the end. It is often argued that in the end extinction of one particular species is
followed up by more complex entities, but this observation is typically anthro-
pocentrically colored by the eye of the human researcher of the evolutionary
process. Recall entropy is a relative concept. Different species (i.e., of different
evolutionary systems) do not share similar preferences about the quantity of
entropy of their environment. There is no connection. In summary we might just
as well purport that a replicator must have the feature of degenerative replica-
tion. This is what makes a paper-copying machine certainly also a collection of
replicators, since the machine will ultimately stop working and become replaced
by a better model.
The fact is that there is inflow selection (degeneration) and outflow selection
(generation). If we accept the mechanism of generative replication as a fourth
definitional feature of the replicator on behalf of outflow selection, we must
accept as well the mechanism of degenerative replication as a fifth definitional
feature of the replicator. This looks like a contradiction, although it is not if we
accept that replicators originate (are being generated) as well as annihilate (are
being degenerated). Well, I think that the concept of replicator can better be
defined within the context of the complete mathematical theory of evolutionary
statistical selection, i.e., it must be incorporated in the mathematical details of
that theory. It is therefore preferable to replace the term replicator by the term
“bitpulse.” This will clarify firstly that all replicators carry an even amount of
entropy during their lifetime, secondly that all of them have a variegated
random lifetime during which they remain stocked within an interactor of
their particular state, thirdly that as selected inflow bitpulses they are
sacrificed/used up to produce new outflow bitpulses, fourthly that as selected
outflow bitpulses they contribute to replenishing the stock of bitpulses within
the various interactors and fifthly and finally that they transport the information
they carry from the past into the future over their lifetime by processes of
exchange.
3.5 Habits, Traits, Customs, Routines and the Limits of Informational Knowledge 75
But we can also settle the definitional question of the units of selection in a
satisfactory manner with the following extraordinary simple definition:
A sample/replicator/bitpulse is a unit of selection that holds, carries and transfers the same
amount of information irrespective of time and state during the time it exists
In their 2007 paper Hodgson and Knudsen candidate entities like thoughts and ideas
as the replicators of social evolution (Hodgson and Knudsen 2007). However later
they opt for habits and more specifically traits, customs and routines as the
replicators of social selection because “they satisfy the conditions of longevity,
fidelity and fecundity” (Hodgson and Knudsen 2010, page 80). In comparison with
their position in 2007 they have decided that “the replication of unobservable habits
of thought is less straightforward.”
The disadvantage of the direct identification of replicators with “gene-like
replicators” as habits, traits, customs and routines as with thoughts and ideas is
that it may make us think that replicators differ from one another and are no longer
similar with respect to their entropy content. Similarity, the second feature of
replicators (see Sect. 2.5) is then explained as applying to many different replicators
(like complete genes in biological evolution) each satisfying the similarity condi-
tion for itself by copying only. Unless we cut them up into small abstract pieces of
similar constant entropy content (the “bit-like replicators” or bitpulses), they do not
satisfy all similarity conditions required for statistical selection. Thus we should
keep in mind that there is always the common unit of entropy in which to express
the measure of evolutionary complexity of habits, traits, customs, routines,
thoughts, ideas and so on.
One may think to reject the latter conception of the bit-like replicator as being too
abstract and instead opt for the gene-like replicator (varying in entropy content), but
the question then is: what is so concrete about unspecified habits, traits, customs,
routines, thoughts and ideas of which we don’t know and—worse still—of which we
will never come to know their individual entropy content by comparative measure-
ment? Remember entropy is a statistical average of a random variable. It is physically
and economically impossible to measure the entropy content of each replicator
individually. Moreover, what makes us think that the entropy content of a copy of
76 3 The Road from Generalized Darwinism to Evolvodynamics
a gene-like replicator remains equal to the entropy content of the original in the
course of time? If it doesn’t remain constant, the copy process of a gene-like
replicator will always produce the “same” gene-like replicator with an entropy
content that is continually changing. This implies that the feature of similarity
irrespective of time and state has not been complied with satisfactorily. It is therefore
untenable to consider habits, traits, routines, thoughts etcetera as being similar if they
differ in entropy content and if they are not mathematically defined to contain
each the same entropy content and if they are not to fall apart in units of selection
with the same constant entropy content.
A consistent mathematical contextual delimitation of the replicator concept is far
to be preferred. Recall that all statistical selection theories (including evolutionary
selection theory) are ignorant about what in detail is going on at the level of each
individual trial of the statistical experiment. Remind the outcome is always a
random occurrence. For the study of a selection experiment only the state and the
probability of the state in which the replicator is selected is relevant (and note that
this probability is an average time-dependent variable). So as regards the selection
process, why bother about the details of what the composition and the features of
the selected item are (other than its state and its average probability of occurrence)
at the level of micro-economic selection?
As a result of the selection processes a stock of replicators settles within the
selecting interactors. During the time replicators do not originate or annihilate they
are stocked in entrepreneurial capacity and labor capacity (in the case of the two-
sector economy) and in their storage position we might assert that they form a
complex of ideas, traits, habits, customs, routines and perhaps even thoughts
stocked in S0 , i.e., in the interactors. The entropy stock of all of these replicators
can be regarded as the aggregate entropy of habits, thoughts, traits, routines etcetera
of the economy. Thus an indirect identification of replicators with “gene-like
replicators” as habits, traits, customs and routines, thoughts and ideas is possible
as long as they are not subject/object of selection and remain stocked in the
interactors. What we can then perhaps say in this respect is that thoughts, ideas,
habits, customs etcetera are principal storage factors that serve as an ordered set of
recipes, a collective of equipment and crystallized prescripts for production and
directions for use for all sorts of items of value during the selection-interval (t,t+dt)
and for enabling production-workers/wage-earners to do their work during (t,t+dt).
They remain stocked as ideas, traits, customs, prescripts, routines, intuitions in non-
differential sample space S0 . By assembling samples of them in the form of
bitpulses during ðt; t þ dtÞ (in the manner described in Sect. 2.3) they contribute
to reassembled inflow and outflow after reordering the samples as part of the
selection processes. As such ideas, traits, customs, routines, intuitions, stocked as
part of the entropy of S0 , contribute to further economic evolution after differen-
tially selected very small portions of them, cut up in bitpulses each carrying one bit,
have been reordered by reassembly of their entropy. This is the connection we must
keep in mind.
However, by cutting these gene-like replicators up, the interactors are just as well
stocks of bit-like replicators. Only bit-like replicators are subject/object of selection.
3.5 Habits, Traits, Customs, Routines and the Limits of Informational Knowledge 77
of one bitpulse? We will return to dealing with those questions later in Chap. 4. It is
all about what information actually is. H&K seek it in the micro-leveled details of
habits, traits, customs, routines, thoughts, but do these entities hide any knowable
information if we cannot observe the variable extent of their individual role
quantitatively? Is information rather not an average statistical concept, dependent
on the elementary probabilities of selection?
Hodgson and Knudsen repeat stressing that replicators are similar, but we have
already argued that we cannot be certain about their similarity if there is no way to
observe and measure the individual content of each of the replicators. Moreover
there are several passages in their work that may need some further amplification to
substantiate that similarity as a general feature. One of these passages is concerned
with multiple level selection.
H&K contend that there are multiple levels of replication from which is selected
(Hodgson and Knudsen 2010, page 80). If multiple selection is nothing else then
multi-sector selection, this is ok. It is then merely a generalization of two-sector (or
binary) selection. However if they understand something else by multiple level
selection, it may be confusing.
In their table on page 173 (Hodgson and Knudsen 2010) Hodgson and Knudsen
discern between multiple ontological levels of selection from high to low: organi-
zational level, group level, individual level. They contend that the composition of
the replicators differs dependent on the applicable level. “The major information
transition that laid down the habitual foundation of culture involved the develop-
ment of a new type of generative replicator” (Hodgson and Knudsen 2010, page
188). At the individual level and group level genes and habits are the replicators, at
the organizational level routines, habits and genes are the replicators. H&K suggest
that the layers of evolution succeed each other from low to high. Well this is all
right if we strictly adhere to the point of view that the replicators of the different
layers of evolution keep the same constant content of entropy (of one bit each)
indefinitely. Else there will be a transition from one layer to its successive higher
layer with replicators that possess a time-dependent content of entropy during the
time they exist as in Fig. 2.7, which is mathematically inconsistent as we have
argued before. Thus we may allow successive phases of multi-level evolution to
exist if the same evolving system (e.g., a particular evolving population with its
evolving environment) remains subject and object of evolution. In that case the
system evolves from a lower entropy content to a higher entropy content that is so
much different from its original content that we prefer to conclude that it has
reached a higher ontological level of evolution.
The study of evolution concentrates always on a particular closed system S0
consisting of N different sectors. We may regard and study an infinite variety of
parallel systems of evolutionary development within our universe, each involving
3.7 Reductionism, Emergentism and Homogeneity 79
have its own methodological standards because the methods of science are univer-
sal. The differential geometrical method by which Newton derived the three laws of
Kepler is a perfect illustration of the fact that the method of science is reductionist.
In a similar manner emergent Darwinian evolution can only be effectively
described as a differential selection process during an infinitesimally small time-
interval (t,t+dt). This is the necessary reductionist step to be taken for analyzing
Darwinian evolution in order to remove the clouds that obstruct our view on the
matter. As long as that step is not considered, the weather will remain bad and
science is more or less in a deadlock. The history of the much older science of
physics has demonstrated that time after time.4
In the present exposition it is the mathematical reduction of time and value to
infinitesimally small lengths of time and infinitesimally small units of entropy that
led to the disclosure and exploration of differential sample spaces of selection in
which all the events occurring in an economy during the time-interval (t,t+dt) are
brought together. Thereupon, it was this reductionist differential approach that
made it possible to employ the effectiveness of Venn diagrams, which turn out to
be indispensable for understanding economic relationships. Without the tool of
mathematical reduction it is unthinkable that so much progress and so much insight
can be attained.
Thus I strongly oppose the idea that there is a potential conflict between the
emergent character of the evolutionary process and the reductionist approach.
Reductionism is the method of science. The idea that reductionism is the wrong
approach usually pops up when there are phenomena for which an adequate
(reductionist) scientific explanation is still lacking. Besides what does the obscure
phrase that the whole is not the sum of its parts mean? Perhaps 3 þ 4 ¼ 9? or 3 þ 4
¼ 10? or 3 þ 4 ¼ 11?. It would be very welcome to know the answer or do we mean
that there are at least several answers or that there is no answer? Science cannot be
built upon such talk. Or do we mean that HðXÞ þ HðYÞ does not equal H ðX [ Y Þ
but rather H ðX [ Y Þ HðX \ Y Þ ? If the latter is what we intend, we are back
at Shannon-inspired evolution. However, within that Shannon-inspired framework
3 þ 4 equals 7 as it always did.
The variety of emergent properties does not imply that emerging properties are
elusive, impossible for science to evaluate quantitatively. If evolution would
produce emergent properties that have no measure, we cannot grapple with evolu-
tion quantitatively. If there is no measure to express the extent of it, there can be no
variables to represent the extent of evolutionary complexity and order such as
inflows, outflows and stocks of entropy. Neither can there be a manner to order
evolutionary processes and can there be tools to analyze evolution like a carpenter
cannot use his tools if there is no way to measure his piece of work. Irrespective of
the many terms, such as e.g., uncertainty, information, knowledge, complexity,
4
No more oppressive has been the deadlock of Ptolemaic astronomy lasting from the hey-days of
the Roman empire, extending over the middle ages until the epoch of Kepler, Gallilei and Newton
finally arrived.
3.7 Reductionism, Emergentism and Homogeneity 81
trace and locate the correct differential space of reduction. Reductionism has no
mathematical limits in the small, just like Newtonian differentiation meets no
mathematical limits in the small. The Newton/Leibniz/Fermat theory of differenti-
ation is the crown-jewel of reductionist reasoning. It can be applied in many
different ways. Unfortunately it can also be maltreated as illustrated by the tenets
of economic marginalism (see Sect. 1.9). But maltreatment can never disqualify the
application of reductionist reasoning.
Chapter 4
Blind and Purposeful Selection
We have discussed in Sect. 3.2 the basic Shannon claim that all the information is
contained in the elementary probabilities of selection. That claim was also the basis
for the statement in Sect. 2.3 that the selection (the assembly) of any sample from or
into non-differential sample space S0 occurs with elementary probability 1=N.
However the arguments in Sects. 2.3 and 3.2 rested on the prior adoption of
Shannon’s conception of information. We took it for granted in Sects. 2.3 and 3.2
that the formulas (1.1) and (2.7) were the correct expressions for the content of
information:
X X
H ð X0 Þ ¼ λi log λi and H ðY0 Þ ¼ μj log μj (1.1)
i j
X
H ð X 0 [ Y0 Þ ¼ qi j log qi j (2.7)
i; j
1
There are some mathematical subtleties that we shall not consider in much detail here. One of
them is that the number k of samples depends on time. This implies that for the inflow selection
process the number of k samples in the basket depends on the time t the inflow selection
experiment starts. That is k ¼ k(t). On the other hand for the outflow selection process the number
of k samples in the basket depends on the time t þ dt the outflow selection experiment starts.
88 4 Blind and Purposeful Selection
the process of selection there are ki ¼ ωi k samples within the “basket” in the i-th
state such that
X
N X
N
k¼ ki ¼ ωi k in which N is the number of states
i¼1 i¼1
k!
Ω¼ (4.1)
k1 ! k2 ! kN !
P implies that there are Ω different equally likely ways (variations) to allocate
This
k ¼ i ki samples over the various sectors (states). The probability W is the same
for each of these variations/sequences:
k1 ! k2 ! kN !
W¼ (4.2)
k!
These equally likely variations are the only variations that can be chosen subject
to the knowledge of ωi . It reflects the information-situation of the agents if the
selecting agents manage to use all the knowledge contained in the probabilities ωi .
Let us then first see whether and how agents succeed to get access to that informa-
tion by their selection behavior. The way agents manage that, is in fact extraordi-
narily simple.
Inflow reassembly is a process of drawing samples from a basket dS 0 ðtÞ originally
filled with k samples at time t. In the end at time t þ dt all samples have been drawn
and the box dS 0 ðt þ dtÞ is empty. This is typically a statistical selection experiment
without replacement. Likewise outflow reassembly is a similar selection experiment
without replacement with the arrow of time reversed. A similar remark applies to the
þ
case of overall reassembly within differential sample space dS 0 [ dS0 . Having
concluded that, we can return to the mathematics of selection without replacement
within differential sample space because this is how agents actually select. The
chances of selecting a sample in a particular differential subset will now differ
dependent on the consecutive order in which the samples happen to become selected
in the various states within the applicable differential sample space.
That is k ¼ k(t þ dt). Clearly k(t) 6¼ k(t þ dt). However since k(t) is proportional to dt, the
difference between k(t) and k(t þ dt) will only have a second order effect in dt so that this
difference will vanish as dt ! 0.
4.1 The Significance of Shannon’s Information Concept 89
k! k1 ! k2 ! kN !
Ω¼ and W ¼
k1 ! k2 ! kN ! k!
pffiffiffiffiffiffiffiffiffiffi
ki χ
ki ! ¼ ðki Þ 2π ki exp ki þ with 0 < χ < 1
12ki
2
Provided k and Ak are very large. Recall the discussion in Appendix A.
4.2 Blind Selection 91
Let us for completeness sake consider the selection scheme of Fig. 4.1. The rule
is that—if there is only choice between two states with ω ¼ ω1 ¼ 1 ω2—entropy/
information is given by Hω ¼ ω log ω ð1 ωÞ logð1 ωÞ . And if the two-
sector selection splits into four states as in Fig. 4.1, entropy/information is
That is,
H ¼ Hω þ ω Hq þ ð1 ωÞHs
in Sect. 4.1 that the information/uncertainty acquired byP the selecting agents on
ðt; t þ dtÞ is equal to the reassembled entropy kH ¼ k i ωi log ωi . This is the
maximum amount of information/uncertainty that the selecting agents can
acquire. We have also argued in Sect. 2.3 that the reduction
P of uncertainty realized
by agents on ðt; t þ dtÞ is k log N kH ¼ k log N þ k i ωi log ωi. Herein k log N ¼
Zdt log N is the entropy randomly assembled from non-differential sample space S0
during ðt; t þ dtÞ necessary for reassembly in differential sample space. This implies
that the elementary probability ProbfSi g of selecting a sample from Si is equal to
1=N (i 6¼ 0).
Mark that ProbfSi g differs from the elementary probability ωi (See Sect. 2.7).
In the hypothetical case that agents would not reassemble their samples in
differential sample space with elementary probability ωi, but with even probability
1=N for all states i, agents would not realize a reduction of uncertainty by their
selection behavior. The latter selection behavior amounts to a case of completely
random reassembly of samples, which we shall call blind selection.
Blind selection does not result in any acquirement of information (or knowledge)
by the selecting agents during the selection interval ðt; t þ dtÞ. Clearly, agents do not
select blindly. Blind decision making is not the way by which selecting agents are
led while investing, consuming, putting equipment and other production capacity
out of use and employing labor effort. The elementary probability by which they
reassemble their samples is ωi, which will generally differ from 1=N. Their selection
behavior will bring about a reduction of uncertainty. By doing so agents select with
purpose rather than blindly and pure randomly.
Blind selection implies that the knowledge of the probabilities ωi is completely
ignored in the process of selection. Then the only thing we can say is that N states
can get allocated in N k ¼ 2k log N different ways over k samples. This number N k is
the maximum possible number of variations the agents would then make while
selecting on ðt; t þ dtÞ. We have called these variations the potential variations in
Sect. 3.2, where we only considered the case of the two-sector economy, in order to
distinguish this set of potential variations from the set of typical variations, which is
a subset of the set of potential variations. By ignoring the knowledge of ωi it would
remain hidden that the typical variations are the only variations that can actually be
selected. We have seen in Sect. 2.3 that—in the absence of any knowledge about
the probabilities ωi —each sample will be assembled with probability 1=N and that
the assembled entropy is then log N per sample. Subject to this ignorance of the
knowledge of ωi , each of the variations will then be assigned the probability
N k ¼ 2k log N of occurrence.
It is easy to see that the reassembled entropy H per sample is smaller or equal to
the assembled entropy log N per sample, i.e. H log N. Thus typical variations that
in fact have a probability W ¼ 2kH of occurrence, are assigned a much smaller
probability 2k log N were the knowledge of ωi completely ignored in the process of
selection, i.e. were selection blind.
Equality of reassembled entropy H and assembled entropy log N will only be
attained if all the ωi equal 1=N. Generally this is not the case. It cannot be the case if
4.2 Blind Selection 93
evolution is to occur. E.g. for the modern two-sector economy the outflow proba-
bility of consumption μ1 (the propensity to consume) is of the order of 0.7.
The inflow probability λ1 of labor is of the same order. This implies that generally
H < log N so that typical variations have a much greater probability than N k to
occur because (with N 2).
N k N kH ! 0 for k ! 1
Moreover the knowledge of ωi implies the information that only the typical
variations with equal probability W ¼ 2kH can be selected and that all the other
variations (in number 2k log N 2kH ¼ N k 2kH ) have vanishing zero probability to
occur.
It follows that, if the knowledge of ωi were ignored, the chance of selecting an
atypical variation is so small that it will factually never occur relative to the
occurrence of a typical variation. However the selecting agents do in fact not ignore
the knowledge of ωi but unwittingly exploit it by selecting without replacement and
so they succeed to acquire the maximum attainable knowledge to their own benefit.
It is sometimes debated by practitioners of the social sciences whether the
Shannon measure of Zdt H reflects all the information that agents can gather
over the time-interval (t,t þ dt). Hodgson and Knudsen accept the Shannon infor-
mation concept to express the information content, complexity and emergent
properties of social evolution quantitatively. Nevertheless they adopt the view
that this Shannon measure “omits key features of information such as ideas and
knowledge in the human domain, particularly meanings and interpretations”
(Hodgson and Knudsen 2010, pages 123, 124). Perhaps this explains their predilec-
tion for identifying habits, traits, customs, ideas and routines with replicators.
Nevertheless they remark that information cannot be defined more narrowly than
Shannon information because of the high level of generality of their conceptual
discussion (Hodgson and Knudsen 2010, page 124).
Well the idea that Shannon’s information concept misses key features as ideas,
knowledge, meanings and interpretations in the economic human domain is a
curious one. The first thing to stress here again is that Shannon entropy/information
is a statistical average (an aggregate) over all the events occurring at the micro-level
of human behavior during (t,t þ dt). The critique on this statistical average concept
of information often enlarges on the richness and variety of micro-behavior of
individual decision-making agents with the unproven suggestion that the mathe-
matics of Shannon can of course never catch up with that variety. We are here
however concerned with different categories: one (Shannon entropy) at the aggre-
gate average level at a particular instant t of time, the other unknowable multifari-
ous phenomena at the micro-level at the same instant t of time.3 This is not a correct
comparison. Furthermore were the Shannon concept of information to miss those
3
Mark that the Shannon average conception of information is not a time-average, but an average
over all micro-economic events at a particular time.
94 4 Blind and Purposeful Selection
key features of variety, the argument can only be taken seriously if those key
features are knowable and affect the selection by the economic agents on (t,t þ dt).
This can only be the case if there are typical variations that hold portions of
replicators (representing such a key feature) with larger information content per
replicator than the average. This is quite impossible. If some replicators hold more
information than other replicators, this is at variance with the second and third
feature defining the replicator of evolution (See Sect. 2.5). Samples, bitpulses,
replicators must be similar. Therefore the contention that Shannon’s information
concept misses key features of information in the human domain is untenable unless
we reject the complete idea of generalized Darwinian evolutionary selection,
a consequence that, I assume, Hodgson and Knudsen do not desire to accept.
Moreover, entropy is fluent. It is impossible to trace afterwards that a particular
sample spent in the i-th sector is actually residing in the i-th sector. Replicators,
samples have no individual label, neither an identity paper nor some individual
certificate. The sample may just as well be anywhere else and consequently
meanwhile have been substituted in the i-th sector by any other sample of entropy
such that the elementary probabilities ωj of entropy over the various states j remain
unaffected. Thus the only certain thing we can say about economic allocation and
reallocation is that the joint number of samples spent (or used up) in a particular
sector i during the time-interval of selection is always ki . There is no way to avoid
that conclusion given the selection probabilities ωi .
Is evolution the result of meaningful and purposeful selection? Or is it just the result
of mere blind chance? This has been a subject of much dispute among evolutionists.
The majority of evolutionary biologists has adopted the view that evolution is
without purpose and meaningless. E.g. Jacques Monod (1971) and in particular
Richard Dawkins (1976, 1984, 1988, 1995) have incessantly described evolution as
a blind selection process. Their point is that there are very, very many variations,
the potential variations, that can be selected. The chance to get selected is assumed
to be the same and very small for each variation (This is the common but incorrect
assumption made by almost all evolutionary biologists). It follows then from the
previous assumption that it is a matter of mere blind chance for a single variation to
get selected. And hence, it is mere blind chance that we are the ones that have
survived. The viewpoint expressed by Monod and Dawkins is not very welcome to
economists and has been a formidable obstacle for economists to adopt the view
that economics is evolutionary. Economists do not like to consider the allocation/
reallocation of economic entities as an activity subjected to the whims and fancies
of blind random forces only. Economic behavior is considered to be fairly purpose-
ful, not in all respects rational and optimal, but just reasonable at an average level
and meaningful, certainly not irrational and not without any purpose. This is also
4.3 Purpose and Shannon’s Existence Theorem 95
the opinion that Hodgson and Knudsen express (Hodgson and Knudsen 2010,
page 48) and I endorse that view.
However, H&K postulate that intentionality is an evolved property. Thus, so
they reason, humans select purposefully. It is the easiest way to refute the point of
view of the proponents of blind evolutionary selection. But there is a serious
difficulty involved with this counter argument. It does not really refute the idea
that all variations (inflow-variations as well as outflow-variations, superior as well
as inferior) are selected with equal chance of selection. If all potential variations are
selected with equal chance, it does not really make a difference for the selecting
agents what they actually select. If it would make a difference, the individual
selection probabilities of the potential variations would differ. If the selection
probabilities do not differ, then indeed selection is blind with respect to what is
selected. That is the point the blind selection proponents make and it is a good point
were it not that all potential variations do not possess an equal probability of
selection as we will stress here again in this section. But this is not the stand
Hodgson and Knudsen take. The reply they give does not really enter into the
matter.
The flaw in the argument of Monod and Dawkins is not that they rule out the
goal-directedness of the evolutionary process beforehand. Their point is that if
agents were to select purposefully they would not select all the variations, bad or
good, with the same probability of selection. Instead they would select the purpose-
ful with a much greater chance than the purposeless variations. Since Monod and
Dawkins take it for granted—on the basis of their inadequate interpretation of
Darwinian selection—that all potential variations are equiprobable, they conse-
quently conclude that agents do not select purposefully, but blindly without any
preference for what they select.
The crucial point in the argumentation is therefore whether Monod and Dawkins
are right in assuming that all variations, the atypical as well as the typical variations
(See Sects. 3.2 and 4.1), are selected with equal chance. Clearly that assumption is
wrong. Monod and Dawkins neglect the role of the selecting agents and they also
neglect the effect of that role on the value that these agents attach to the various
variations between which they must select. The agents value and devalue variations
in the course of executing the selection of inflow and outflow.
In the previous sections it has been demonstrated that, if selecting agents act
without using any knowledge, the total number of variations/sequences that the
selecting agents can select on the time-interval (t,t þ dt) of selection is equal to
2klog N . We have also demonstrated that, quite the contrary, the selecting agents use
all the knowledge that comes available during the process of selection to their
maximum advantage. As a result of that exploitation of knowledge only 2kH typical
variations, in number a vanishing portion of the number 2klog N of potential
variations, are favored by the selecting agents. That is,
2klog N 2kH ! 0 for k ! 1
The fact then is that there are an overwhelming number of variations (the
atypical ones) that will never get selected. These rejected variations have a
96 4 Blind and Purposeful Selection
2− k log N
0
Sequence number of variations
arranged in order of increasing
probability
vanishing probability zero to get selected. On the other hand the typical variations
are the only sequences that can be selected, each with the same finite chance 2kH of
selection.
The above argumentation implies that the agents use all the available applicable
knowledge while selecting. Therefore selection is not blind; instead it is meaningful
and purposeful. And hence the selected typical variations are meaningful and
purposeful. The rest of the variations is meaningless. That rest is junk with
vanishing zero probability to get selected.
The thesis that actually agents can only select meaningful variations and actually
never select one of the very many purposeless and unprevailing variations I shall
call Shannon’s existence theorem. In fact that theorem states that only purposeful
variations can exist. The above argumentation by which the theorem is established
is closely related to the argument by which Claude Shannon justified his definition
of information transmitted over the communication channel (Shannon 1948; See
also Jaynes 2004; Papoulis 1985).4 Figures 4.2 and 4.3 clarify the argument
visually. Figure 4.2 represents the Monod/Dawkins conception of evolutionary
selection. It is based on the wrong perception that each variation is selected with
equal probability of selection. Figure 4.3 illuminates the correct Shannon-inspired
conception. P
In Sect. 4.1 it was argued that Zdt H ¼ Zdt i ωi log ωi is the maximum
amount of information that the selecting agents can derive from the knowledge
that they gather gradually during the selection process. No more information than
Zdt H and no less.
4
For good order’s sake, Claude Shannon listed his related theorem as one of the theorems in
Shannon (1948). The proof requires specific adaptations in the mathematical argumentation within
the context of the transmittance and reception of stationary sequences with a finite time-length T
rather than with a vanishing time-length dt.
4.3 Purpose and Shannon’s Existence Theorem 97
2−k⋅H
From the point of view of evolutionary selection Zdt H ðY0 Þ can be interpreted
as a maximum level of attainable aggregate outflow information that economic
agents on outflow selection acquire. By their selection behavior they make the best
of it they can by choosing a strategy and deal with the situation as purposefully and
as best as possible given the skills, knowledge and opportunities they have. In this
way they manage to reach the maximum attainable level Zdt H ðY0 Þ of outflow
information during the selection interval (t,t þ dt).
Likewise from the point of view of evolutionary selection Zdt H ðX0 Þ can be
interpreted as a maximum level of attainable aggregate inflow information that
economic agents on inflow selection acquire. By their selection behavior they make
the best of it they can by choosing a strategy and deal with the situation as
purposefully and as best as possible given the skill, knowledge and opportunities
they have. In this way they manage to reach the maximum attainable level
Zdt HðX0 Þ of inflow information during the selection interval (t,t þ dt). The
economic agents actually select only a single inflow variation and only a single
outflow variation out of a gigantic supply of inflow, respectively outflow, variations
during (t,t þ dt). The vast majority of the variations (the atypical ones) will be
immediately rejected, they will never be selected because the selecting agents know
or feel that these variations don’t work or the agents just don’t like them. Thus
agents will not select a butcher to get employed to bake bread in a bakery. Nor will
they send a ship on the highway to transport cargo to a foreign country. Neither will
a consumer/agent purchase chocolates for his love if she hates sweets. What
98 4 Blind and Purposeful Selection
remains for potential selection are the relatively very few meaningful typical
variations among the enormous supply of (potential) variations. According to
Shannon’s existence theorem these favored meaningful typical variations have
each equal probability of being realized. Thus a baker has a chance to get employed
in a bakery, a ship has a chance to sail cargo abroad over the sea, and a lady which
hates sweets but is fond of flowers has the chance to get surprised by a bouquet of
flowers.
It follows that Shannon’s justification of his conception of information and
uncertainty was also in the notion that agents can factually only select typical
variations each occurring with even probability W , from which he concluded—
like in Sect. 4.1—that more information is gathered the smaller W is. This is more
or less also the argument we used in Sect. 3.2 to motivate the Shannon definition of
information.
The additional argumentation we gave in Sect. 4.1 with respect to the argument
given in Sect. 3.2 is the observation that agents get unwittingly full access to the
source of hidden knowledge that is in the elementary probabilities because they
select without replacement. This observation was still wanting in the argument
developed in Sect. 3.2. I think the latter observation completes the best argumenta-
tion there is to justify Shannon’s information concept.
Let us now return to considering the arithmetic and the mathematics of the selection
of samples in another way. We shall deal with the two-sector economy only so that
each of the samples has the possibility to reside in state 1 (the consumption sector)
or, alternatively, to reside in state 2 (the investment sector). The economy can then
be considered as a stock of entropy consisting of an endless number of samples
stored in that economy either in the consumption sector or in the investment sector.
The joint number of samples stored in the investment sector is entrepreneurial
capacity, the joint number of samples stored in the consumption sector is labor
capacity.
Each sample of the economy observed at current time t has originated at its
specific (random) initial time t τ ðτ > 0Þ in the past and will annihilate at its
specific (random) final time t þ θ ðθ > 0Þ in the future. We have visualized the joint
stock of samples resident in the economy in Fig. 4.4. This has been done quite
schematically just to clarify that there are many individual samples with limited
lifetime that together form the joint entropy stocked in the economy, i.e. in
investment sector and consumption sector. Clearly, the initial rise of the joint
content of entropy from the non-existence level of 0 bits, followed by a later decline
to the non-existence level of 0 bits again, as illustrated in Fig. 4.4, is generally not
representative for the entropy of an economy. Here it has only been done to
4.4 Blind Selection, Purposeful Selection and Economic Inheritance. Some More. . . 99
Number of
samples in Approximation of a set by
a set various samples each with
its own initial time and
final time
0
time
Future
Past
Single sample lifted
Entropy out of the stock of
in bits samples that
represents the
population at
logN current time t
current time t
demonstrate the various initial times and final times of the constituting samples in a
rather specific manner.
The samples stocked in the economy of Fig. 4.4 await their moment of differen-
tial selection. This is clarified in Fig. 4.5. Here we have been drawing the selected
(inflow) samples that cease existence on (t,t þ dt) and the selected (outflow)
samples that originate on (t,t þ dt).
Each inflow sample has had its specific finite time-instant t τ of origination in
the past. Each outflow sample will have its specific finite time-instant t þ θ of
100 4 Blind and Purposeful Selection
Fig. 4.5 The creation and annihilation of samples of value on the time interval (t,t þ dt) of
selection.
The transmission of inflow and outflow manifests itself in the concurrent selection of the pairs Aa, Bb and Cc of outflow/inflow samples.
þ þ þ
Outside dS
0 \ dS0 overall selection reassembles inflow samples to annihilate in dS0 |dS0 and outflow samples to originate in dS0 |dS0
samples marked with A, B and C combine with the inflow samples marked with a, b
and c in respective order. Note that as a consequence of their simultaneous selection
A and a are selected as a concurrent Aa pair of samples, likewise B and b are
selected as a concurrent Bb pair of samples and C and c are selected as a concurrent
Cc pair of samples. Thus the interaction between inflow and outflow manifests itself
in the concurrent selection of these pairs Aa, Bb, Cc of “pure” and virtual entropy5
þ
within the intersection area dS
0 \ dS0 . The result is that equal quantities of entropy
will annihilate and originate in the intersection area as required. This is exactly the
requirement to be satisfied to identify transmission with the volume of financial
exchange.
On the other hand unequal quantities of entropy will be reassembled in
dSþ dS and dS dSþ .
0 0 0 0
Let us—as in Sect. 3.2—for illustrative and explanatory purposes consider in
more detail the hypothetical situation depicted in Fig. 4.5 for which only a very
limited number of samples are selected on ðt,t þ dtÞ. In that case we can do the
numerical calculations of numbers and probabilities without using the asymptotic
approximations of Sect. 4.1, but instead use the exact formula as given by (A.1) of
Appendix A and by (4.1). For instance: if all that is selected during the time-interval
(t,t þ dt) are Zdt ¼ 10 samples either within state 1 or within state 2, the number of
potential variations is 210 ¼ 1024: Part of these variations has been listed in
Table 4.1. There are already too many to list all of them on one page. Only one
of the 1024 potential variations is actually selected during (t,t þ dt). If there are 7
of the 10 samples in differential outflow sample space in state 1 as in Fig. 4.6,
agents will always select a purposeful typical variation that has 7 samples in state 1
and 3 samples in state 2. The number of typical variations that can be selected by the
agents in this way is 10!=ð7! 3!Þ ¼ 120. [See e.g. (4.1) and Sect. 4.1]. The typical
variations have been bold printed in Table 4.1, but note that we cannot list all 120 of
them in the table. Only one of the typical variations is actually selected by the
agents (as a result of selection without replacement). As we have argued this is
achieved in a purposeful manner. The selected typical variation has a chance 1=120
to occur. Since all other typical variations have the same chance of selection, the
probability to select any purposeful typical combination is 120 1=120 ¼ 1. Hence
the probability of selecting a purposeless atypical variation is zero. No chance left
for that.
We have now the opportunity to detail the property that bitpulses possess finite
lifetime! One single sample drags on the same content of entropy during its
lifetime. The economic replicator is the equivalent of the sample with a lifetime.
þ
The intersection domain dS 0 \ dS0 is the domain of exchange. Here the various
pairs of bitpulses exchange for money or conversely: A for a, B for b and C for c.
The phenomenon of exchange is typical for the processes of selection within the
transmission domain. It can be explained as follows:
þ
We must discern between “pure” entropy bitpulses within dS 0 \ dS0 that leave
a trace of entropy and the counterbalancing virtual bits of entropy (money/
5
The difference between “pure” and virtual entropy will be defined more comprehensively in
Sect. 5.2 and within the context of Chaps. 5 and 6.
102 4 Blind and Purposeful Selection
Table 4.1 Evolutionary variation and selection in the hypothetical case that there are only 10
samples that get selected either within state 1 or within state 2 and if the elementary probability ω
of selecting a sample within state 1 is equal to ω ¼ 0.7. In this case the total number of possible
(atypical) combinations (the number of variations) is 210 ¼ 1024. The total number of actually
selected purposeful typical variations (bold printed) is much smaller and equal to:
Ω ¼ k1 !kk
!
2!
¼ 7 !103! ! ¼ 120 with k1 ¼ ω k and k2 ¼ ð1 ωÞk
List of potential variations
1111111111 1111111112 1111111121 1111111122 1111111211 1111111212
1111111221 1111111222 1111112111 1111112112 1111112121 1111112122
1111112211 1111112212 1111112221 1111112222 1111121111 1111121112
1111121121 1111121122 1111121211 1111121212 1111121221 1111121222
1111122111 1111122112 1111122121 1111122122 1111122211 1111122212
1111122221 1111122222 1111211111 1111211112 1111211121 1111211122
1111211211 1111211212 1111211221 1111211222 1111212111 1111212112
1111212121 1111212122 1111212211 1111212212 1111212222 1111221111
. . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . .
. . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . .
. . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . .
1212111212 1212111211 1212111122 1212111121 1212111112 1212111111
1212112121 1212112112 1212112111 1212111222 1212111212 1212111221
1212121111 1212112222 1212112221 1212112212 1212112211 1212112122
1212121212 1212121211 1212121211 1212121122 1212121121 1212121112
1212122122 1212122121 1212122112 1212122111 1212121222 1212121221
1212211112 1212211111 1212122222 1212122221 1212122212 1212122211
1212211222 1212211221 1212211212 1212211211 1212211122 1212211121
1212212212 1212212211 1212212122 1212212121 1212212112 1212212111
. . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . .
. . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . .
. . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . . . . .. . .. . .. . .. . .
2221221111 2221221112 2221221121 2221221122 2221221211 2221221212
2221221221 2221221222 2221222111 2221222112 2221222121 2221222122
2221222211 2221222212 2221222221 2221222222 2222111111 2222111112
2222111121 2222111122 2222111211 2222111212 2222111221 2222111222
2222112111 2222112112 2222112121 2222112122 2222112211 2222112212
2222112221 2222112222 2222121111 2222121112 2222121121 2222121122
2222121211 2222121212 2222121221 2222121222 2222122111 2222122112
2222122121 2222122122 2222122211 2222122212 2222122221 2222122222
2222211111 2222211112 2222211121 2222211122 2222211211 2222211212
2222211221 2222211222 2222212111 2222212112 2222212121 2222212122
2222212211 2222212212 2222212221 2222212222 2222221111 2222221112
2222221121 2222221122 2222221211 2222221212 2222221221 2222221222
2222222111 2222222112 2222222211 2222222212 2222222221 2222222222
Note that neither all of the potential variations nor all of the purposeful typical variations have
been listed in the table for practical reasons. The sixth typical variation (1212111211) in the
second column of the list is the one actually selected in the example illustrated in Fig. 4.6.
4.4 Blind Selection, Purposeful Selection and Economic Inheritance. Some More. . . 103
Time
t t + dt
Time interval
(t,t + dt) of
current selection
Fig. 4.6 The statistical experiment of outflow selection of samples during the time interval
(t,t þ dt) of selection.
Only a single one of the purposeful typical sequences of the outflow samples is actually selected, but all the purposeful sequences have equal
probability to get selected. The remaining possible combinations (the variations) have zero probability to get selected. Because of that they
are purposeless
þ
liquidity/accounts receivable etcetera) within dS 0 \ dS0 that leave no trace of
entropy except for an eventual surplus of virtual bits. “Pure” entropy (or just the
collection of bitpulses) has lifetimes and leaves as such a trace of entropy in the
course of time. Virtual bits of entropy have no lifetimes and therefore they cannot
be bitpulses. The pair wise selection of samples will manifest itself in two
possible ways:
þ
1. In the intersection domain dS 0 \ dS0 a bit a of virtual entropy sacrifices its
imaginary content of a single bit of entropy at its manifestation time θ0
ðt θ0 t þ dtÞ in exchange for a concurrently created bitpulse A at initial
þ
time τ ðt τ t þ dtÞ within dS0 \ dS0 .
þ
2. In the intersection domain dS0 \ dS0 a bitpulse b is sacrificed at its final time θ
ðwith t θ t þ dtÞ to create concurrently a virtual bit B of entropy at its
manifestation time τ0 ðt τ0 t þ dtÞ.
104 4 Blind and Purposeful Selection
þ
There are (infinitely) many of these pair wise occurrences within dS 0 \ dS0 (we
have only depicted three of those pairs in Fig. 4.5). Since the various virtual bits
cancel as far as they can, they leave only a surplus of created virtual bits over
annihilating virtual bits or a surplus of annihilating virtual bits over created virtual
þ
bits within dS
0 \ dS0 .
Thus if there is a surplus Δ of annihilating virtual bits over created virtual bits
þ
within dS0 \ dS0 during ðt; t þ dtÞ, then:
þ
• the number of created “pure” bitpulses is Zdt H dS 0 \ dS0
þ
• the number of annihilating “pure” bitpulses is Zdt H dS 0 \ dS0 Δ
• the number of annihilating virtual bits is Δ
On the other hand if there is a surplus Δ of created virtual bits over annihilating
þ
virtual bits within dS
0 \ dS0 during ðt; t þ dtÞ, then:
þ
• the number of created “pure” bitpulses is Zdt H dS 0 \ dS0 Δ
• the number of created virtual bits is Δ
þ
• the number of annihilating “pure” bitpulses is Zdt H dS 0 \ dS0
The contended cancellation of a pair of virtual bits, of which one manifests itself at
another time τ0 on ðt; t þ dtÞ than the manifestation time θ0 of the other on ðt; t þ dtÞ,
may meet some reserves as it appears that complete cancellation can only be realized
if τ0 ¼ θ0. Likewise the creation (annihilation) of a “pure” bitpulse at a particular time
on ðt; t þ dtÞ in exchange for a virtual bit at another manifestation time on ðt; t þ dtÞ
may call forth similar reserves. However, the distribution of initial times, final times
and manifestation times over the selection interval ðt; t þ dtÞ is of no significance. As
we already remarked in Sect. 3.2 “the time-length dt of the selection interval is so
small that, economically, all Zdt trials take place simultaneously. Thus in fact
differences in the sequential order by which the states are distributed over the samples,
given the elementary probabilities of selection, cannot have an economic effect”.
Selections are carried out individually by very many economic agents. How we order
their individual selection decisions sequentially over the time-interval ðt; t þ dtÞ in the
course of time is quite irrelevant noticing that dt ! 0 . The entropy flows remain
unaffected. There is no information in the way initial times and final times of bitpulses
and manifestation times of virtual bits are ordered on the selection interval ðt; t þ dtÞ.
Exchange or its equivalent, transmission, facilitates information to transfer
continually from the past into the future so that the evolutionary process can go
on. Clearly, simultaneously with the exchange process entropy will be reordered
(reassembled) in response to outflow selection and inflow selection, resulting in a
þ þ
gain of entropy in dS0 dS0 , respectively a loss of entropy in dS0 dS0 and similar
þ
gains and losses of entropy in the transmission domain dS 0 \ dS0 due to overall
selection (See Sect. 4.5).
Hodgson and Knudsen assert the realization of information transfer from the past
into the future (the Darwinian principle of inheritance) to depend on the mechanism
of generative replication by which they understand the copying of replicators at the
4.4 Blind Selection, Purposeful Selection and Economic Inheritance. Some More. . . 105
end of their lifetime into new replicators at the beginning of their lifetime. Well,
replication is certainly quite necessary to warrant the effect of economic inheri-
tance. However for completeness sake, the generation (creation) of new replicators
is only one side of the coin. There is also the degeneration (sacrifice) of old
repliclators. So why add the adjective generative? I do not think that the terms
generative and degenerative contribute to catching the essence of conveying
information from the past into the future. Instead exchange is the right term. To
warrant that information is conveyed by information carriers (the replicators/
bitpulses) in the course of time, these carriers must possess lifetime. Else entropy
does not circulate and then the samples are incapable of replication and cannot pass
the information of the past into the future. However, we all know that the economy
produces entropy that possesses lifetime: Production equipment is used for lengthy
periods of time; laborers apply their skills and experience during many laborious
years. Thus Darwinian inheritance (otherwise a typical mechanism of evolutionary
genetically based biology which is less appropriate for describing economic evolu-
tion) meets its economic counterpart in the circulating properties of entropy with a
lifetime.6 Methods of production and skills are stored and expressed as entropy with
a lifetime in entrepreneurial capacity (state 2) as well as in labor capacity (state 1).
Let us revert to the hypothetical selection situation of Table 4.1. Recall that only
the 120 purposeful typical variations have each a chance 1=120 to get actually
selected. The remaining number of 1024 120 ¼ 904 atypical variations has zero
chance to get selected and these variations are therefore purposeless (on behalf of
Shannon’s existence theorem).
Actually only one out of the purposeful typical variations will be selected on
ðt,t þ dtÞ. In Fig. 4.6 I have assumed that the typical outflow sequence (1212111211),
composed of the 10 samples, is the specific typical outflow variation actually
selected. It could of course—with equal probability—have been one of the other
bold printed purposeful typical variations listed in Table 4.1. The one selected in
Fig. 4.6, allocates the 2nd, 4th and 8th sample to state 2 and the other samples to state 1.
Note this is only the outcome of the outflow selection experiment on ðt,t þ dtÞ. It must
yet be completed with the outcome of the inflow selection experiment.
As stressed before the number of Zdt ¼ 10 samples is a hypothetical case. That
number is far too small to represent the actual situation of evolutionary selection.
To improve on the scheme we must have the number of samples to tend to infinity.
In Sect. 1.7 we discussed the effects of increasing Zdt on the Shannon formula (1.1).
It was further analyzed in Appendix A and Sect. 3.2. Let us now see what the effects
of increasing Zdt are on the number of potential variations and the number of typical
variations here. To see what happens let us successively multiply the number of
samples by increasingly greater powers of 10, i.e. from 10 to 100 to 1,000 to 10,000
6
There is a remarkable correspondence in the aim expressed in the sub-title Bringing Life Back
into Economics of Hodgson’s 1993 book (Hodgson 1993) and the finding that lifetime is the
necessary requisite to reach that goal. Indeed evolution is about the development of life and life
needs lifetime to exist.
106 4 Blind and Purposeful Selection
Table 4.2 Numbers of potential variations and numbers of actually selected purposeful sequences
for different numbers of samples
Probability of
selecting a
single Quotient of number of
Number Number of Number of purposeful typical purposeful potential variations
Zdt of potential variationsa variation and number of
samples variations ðω ¼ 0:7Þ ðω ¼ 0:7Þ purposeful variations
10 2 ¼ 1024
10 120 1 =120 8.533
100 2100 100!=ð70! 30!Þ 288 288 212
1,000 21000 1000!=ð700! 300!Þ 2881 2881 2119
10,000 210000 10000!=ð7000! 3000!Þ 28812 28812 21188
a
The numbers in the 3rd, 4th and 5th row of this column have been calculated by applying the
following very accurate asymptotic approximation:
log[k!/(k1!k2!)] ¼ –[(1–ω)log(1–ω) þ ωlogω]k – ½log[(1–ω)ωk] – ½log(2π)
while keeping the quotient ω ¼ ω1 of the number of samples in state 1 and the total
number of samples within sample space (the “basket” of selection) at exactly 70 %.
The larger the number of samples in differential sample space, the better the
asymptotic approximation, as dealt with in Sect. 4.1, must be. For instance if the
number of samples is 100, the number of potential variations is a very huge number:
þ
Let us consider the intersection dSi \ dSj ði; j ¼ 1 or 2Þ. This is the space in
which even samples of inflow annihilating in state i and outflow originating in state j
combine concurrently to reassemble entropy inflow and entropy outflow. The
probability by which an inflow sample annihilating in state i and an outflow sample
originating in state j are assembled concurrently is qi j. Recall that the sampling rate
in overall sample space dS0 is Z. Hence there are Zdt qi j different trials within
þ
dSi \ dSj , which bring forth reassembled entropy annihilating in state i together
with reassembled entropy originating in state j . The reassembled probability of
selecting all the samples during ðt; t þ dtÞ is
Zdtqi j
qi j
Thus the total reassembled entropy annihilating in state i and originating in state j
during ðt; t þ dtÞ is
Zdtqi j
log qi j ¼ Zdt qi j log qi j
0
time
0
time
t Number of samples t + dt
is Zdt ⋅ qij
That common portion is the result of statistical dependence between the events
of origination in state j and the events of annihilation in state i. This is the entropy
of the event that the assembly/selection of the annihilating sample in state i of
non-differential space S0 would occur concurrently with the assembly/selection
of the originating sample in state j of non-differential
space S0 during the
selection interval (t,t þ dt). This portion H Xi \ Yj of concurrent entropy is yet
to be deducted from the above total H Yj =N þ H ðXi Þ=N to avoid a double count
because dSþ
j and dSi have it in common.
In summary,
1 1
qi j log qi j ¼ H ðXi Þ þ H Yj qij H ðX0 \ Y0 Þ ði; j ¼ 1,2,3, , NÞ (4.8)
N N
4.6 The Three Degrees of Freedom of Evolution 109
P P
Moreover, on behalf of (B.4) of Appendix B, qi j ¼ 1 and qi j log qi j ¼
H ðX0 [ Y0 Þ. After substitution in (4.9) it follows that
N H ð X 0 [ Y0 Þ ¼ N H ð X 0 Þ þ N H ð Y 0 Þ N H ð X 0 \ Y 0 Þ
With the help of the Venn diagram of Fig. 2.4 and the auxiliary subsets A, B, C,
D, E, F, G and U as given in this diagram we establish for the two-sector economy:
q11 log q11 þ q12 log q12 q21 log q21 þ q22 log q22 ¼ HðBÞ HðAÞ
¼ ð2μ 1Þ HðY0 jX0 Þ
so that
q11 log q11 þ q12 log q12 q21 log q21 þ q22 log q22
H ðY0 jX0 Þ ¼ (4.12)
ð2μ 1Þ
110 4 Blind and Purposeful Selection
q11 log q11 þ q21 log q21 q12 log q12 þ q22 log q22
H ðX0 jY0 Þ ¼ (4.13)
ð2λ 1Þ
and
q11 log q11 q22 log q22 þ q12 log q12 þ q21 log q21 ¼
¼ HðDÞ þ HðFÞ HðCÞ HðEÞ ¼ 2 12 q11 q22 H ðX0 \ Y0 Þ
so that
q11 log q11 q22 log q22 þ q12 log q12 þ q21 log q21
H ð X 0 \ Y0 Þ ¼ (4.14)
2 12 q11 q22
X q11 log q11 þ q12 log q12 q21 log q21 þ q22 log q22
ðμi log μi λi log λi Þ ¼ þ
i
ð2μ 1Þ
q11 log q11 q12 log q12 þ q21 log q21 þ q22 log q22
ð2λ 1Þ
(4.15)
With the help of the Venn diagram of Fig. 2.5 this results in the following
equations for the three-sector economy:
4.6 The Three Degrees of Freedom of Evolution 111
We will not work this out any further. To solve this system on the basis of
available data of some of its time-series is a matter of numerical analysis. We
shall suffice with a calculation of the number of degrees of freedom to get an idea
of the numerical job. Clearly, the above system of equations is a system of 9
equations in the elementary probabilities μj , λi and qi j . That is, the number of
different elementary probabilities is 3 þ 3 þ 9 ¼ 15. The number of equations is
5 more than the four equations (4.10) of the two-sector economy. The number of
elementary probabilities is 7 more than the 8 elementary probabilities μj, λi and qi j
of the two-sector economy. Thus at first glance it appears that for the three-sector
economy there are 4 [i.e. (15 8) (9 4) þ 2] more elementary probabilities
than there are equations describing the relationships between these elementary
probabilities, i.e. two more than for the two-sector
P economy.P However in the two-
sector economy there are only four equations q
i ij ¼ μj , j qi j ¼ λi instead of
six of them for the three-sector economy. Hence it follows, with the additional
time-series of Z, that the number of degrees of the three-sector economy is also
three (6 4 þ 1), the same as for the two-sector economy.
The above reasoning can be extended to the N-sector economy as follows. The
number of equations of the type (4.8) is N 2. Relative to the ðN 1Þ-sector economy
this number has increased by N 2 ðN 1Þ2 ¼ 2N 1. The number of variables of
the type μj , λi and qi j is N 2 þ 2N ¼ N ðN þ 2Þ: Relative to the ðN 1Þ-sector
economy this number has increased by N ðN þ 2P Þ ðN 1ÞðP N þ 1Þ ¼ 2NP þ 1.
The number of independent equations of the type i q i j ¼ μ j , j q i j ¼ λ i , j μj
P
¼ 1 and j μj ¼ 1 is 2ðN þ 1Þ: Relative to the ðN 1Þ-sector economy this number
has increased by 2. Hence the number of degrees of freedom of the N -sector
economy is ð2N þ 1Þ ð2N 1Þ 2 ¼ 0 more than the number of degrees of
112 4 Blind and Purposeful Selection
7
If the timeseries of the unit price PðtÞ are considered to be unknown as well, the number of
degrees of freedom is 4. I conjecture that this fourdimensional behavioral freedom of decision
within economic value-space-time is connected with the behavioral freedom of loco-motion
within four-dimensional physical space-time. It is an attractive line of explanation, but a complete
convincing demonstration is still lacking.
Chapter 5
Exchange and the Circulation of Entropy
Cþ
i ðtÞ ¼ Ci ðtÞ ¼ Ci ðtÞ
because the total historic money expenditures for which C0 ðtÞ has been built up by
processes of entropy origination must eventually be liberated/exchanged in the
future by processes of entropy annihilation within S0 .
Consider the Venn diagram of the multi-sector economy. The transmission domain
þ
dS0 \ dS0 is the domain of events of exchange, the location where selection events
of entropy are settled by money exchange. Each entropy bit is worth the same
common average unit price PðtÞ.
On the other hand dSþ 0 and dS0 are
domains where entropy is created and
þ
annihilated. The entropy flow Zdt H dSi dS0 I shall call the outflux of Si .
þ
The entropy flow Zdt H dS dS I shall call the influx of Si . The entropy flows
þ i 0
þ
Zdt H dSi \ dS0 and Zdt H dS
i \ dS0 are financial flows of entropy of Si
expressed in the dimension of bits. It is what we have also called transmissions after
Shannon. I shall also call them financial flows just to stress their economic meaning
with respect to the exchange process of entropy for money or for money-related
certificates of value like accounts receivable and accounts payable.
These financial flows may reflect the origination of bitpulses in exchange for
þ
money within dS 0 \ dSi . They may also reflect the sacrifice of bitpulses in
þ
exchange for money within dS i \ dS0 , dependent on the circumstances. The
money value of these financial flows is PZdt H dS0 \ dSþ i , respectively PZdt H
dSi \ dSþ0 .
The entropy flow Zdt H dSþ i I shall call the output of Si . The entropy flow
Zdt H dSi I shall call the input of Si . From the perspective of selection all the
entropy within dSþ 0 , thus including the financial entropy output originating within
þ
dSi \ dS0 , does contribute to the procreation of purposeful output variations and
likewise all the entropy within dS 0 , thus including the financial entropy input
þ
annihilating within dS i \ dS 0 , does contribute to the sacrifice of purposeful input
variations.
In accordance with the above terminology the bitpulses of dSþ dS and of
þ
i 0
dSi dS0 will be called outflux bitpulses, respectively influx bitpulses of Si. Likewise
the bitpulses of dSþ
i and of dSi will be called output bitpulses, respectively input
bitpulses of Si : Output bitpulses include the outflux bitpulses. Input bitpulses
include the influx bitpulses.
The creation of bitpulses and the annihilation of bitpulses on ðt;t þ dtÞ within the
þ
transmission domain dS 0 \ dS0 do of course depend on the availability of events
116 5 Exchange and the Circulation of Entropy
that create value, respectively that annihilate value. But there is something special
þ
with the creation and annihilation of financial value within dS 0 \ dS0 .
We must distinguish between “pure” entropy, which consists of bitpulses with a
þ
lifetime within dS 0 \ dS0 . This is the entropy of bitpulses that possess probability
distributions of current lifetime and excess lifetime. This is also the entropy that
þ
exchanges for units of exchange (e.g. money) in dS 0 \ dS0 : These units of
exchange do not constitute bitpulses, but they have entropy content without having
lifetime (to calculate that entropy content just divide their money content by P).
They form a collection of bits of entropy without probability distributions of current
lifetime and of excess lifetime (money and money-like accounts have no lifetime).
This collection of bits I will call virtual entropy. Like a virtual particle in physics
manifests itself in interaction with the creation and annihilation of real particles,
virtual economic entropy manifests itself in interaction with the origination and
annihilation of “pure” economic entropy.
The origination of “pure” bitpulses in exchange for money and, alternatively, the
annihilation of “pure” bitpulses in exchange for money are mutually exclusive.
þ
That is, either “pure” bitpulses are created in exchange for money within dS 0 \ dS0
during ðt;t þ dtÞ or “pure” bitpulses are annihilating in exchange for money within
þ
dS0 \ dS0 during ðt;t þ dtÞ. The first situation develops subject to the constraint of
average price inflation. The second, alternative, situation develops subject to the
constraint of average price deflation. Each in a very specific but related manner.
Let me first explain the constraint of average price inflation. Note that in the
money economy the gross increase of capacity C0 is provided for by money output,
i.e. the money value of entropy output. In the course of time the unit price P of a bit
of entropy changes. Nevertheless exchange is in money units and the entropy output
at time t is exchanged for another unit price than the unit price of entropy output at
an earlier time t τ. Thus the total C0 ðtÞ of entropy output accumulating in the
capacity stock over the distant past until current time t, i.e. over the time-interval
ð1; tÞ, as far as it has not been used up in the past before time t, will have been
acquired for an average historic unit price per bit of entropy different from the
þ
current unit price PðtÞ per bit of entropy. Let, for convenience of explanation,Pi ðtÞ
denote the average historic unit price of the entropy Ci ðtÞ stocked in Si at time t.
þ
Now, if on an average prices in S0 have risen in the past such thatP0 ðtÞ < PðtÞ at
time t, then the exchange of entropy selected from the entropy stock C0 ðtÞ during
þ
ðt; t þ dtÞ will be accompanied by a financial gain PðtÞ P0 ðtÞ per bit of entropy
exchanged. Clearly, this implies that subject to the constraint of average historic
price inflation there is a financial output on top of the outflux PZdt H dSþ
0 dS0 :
þ
The total of this financial output is PZdt H dS0 \ dS0 ; the entropy flow
þ
originating within the transmission domain dS 0 \ dS0 .
Well the reader might now object that we have neglected the eventual simulta-
þ
neous occurrence of loss of financial value within dS 0 \ dS0 during ðt;t þ dtÞ .
However that objection is ill-founded. If there is average price inflation in S0 such
þ
that PðtÞ P0 ðtÞ > 0, the annihilation of bitpulses in exchange for money in S0 on
5.2 The Different Sorts of Bitpulses: Outflux, Output, Influx, Input and. . . 117
ðt; t þ dtÞ would imply the absurdity that bitpulses each with a negative load of one
bit annihilate in the differential subset dS
0 . However the entropy content transferred
by a bitpulse is always a positive single bit as we have argued in the foregoing.
There is no way to escape the h conclusion that subject i to the constraint of
þ
average historic price inflation i:e: for PðtÞ P0 ðtÞ > 0 the complete entropy
þ
flow PZdt H dS 0 \ dS0 represents financial output of bitpulses originating in S0
in exchange for money during ðt;t þ dtÞ.
The argumentation is almost analogous if the constraint of average price defla-
tion applies. Note that in the money economy the gross decline of capacity C0 is
provided for by money input, i.e. the money value of entropy input. In the course of
time the unit price P of a bit of entropy changes. Nevertheless exchange is in money
units and the entropy input at time t is exchanged/liberated for another unit price
than the unit price of the entropy input at a later time t þ θ. Thus the total C0 ðtÞ of
entropy input to be sacrificed onward from current time t over the future, i.e. over
the time-interval ðt; 1Þ, as far as it will not be produced in the future after time t,
will be sacrificed for an average future unit price per bit of entropy different from
the current unit price PðtÞ per bit of entropy. Let, for convenience of explanation,
Pi ðtÞ denote that average future unit price of the entropy Ci ðtÞ stocked in Si at time t.
Now, if on an average prices in S0 decline in the future (this is the condition of
average price deflation) such thatP0 ðtÞ < PðtÞ, then the exchange of entropy selected
for input from the entropy stock C0 ðtÞ during ðt;t þ dtÞ implies that one sacrifices
PðtÞ P0 ðtÞ more money per bit of sacrificed entropy at current time t than the
average future unit price for which it will be sacrificed in the future. Clearly, this
entails a financial loss subject to the constraint of average price
þdeflation.
This
financial loss is financial input on top of the influx PZdt H dS dS . The total of
þ
0 0
this financial input is PZdt H dS 0 \ dS0 , the entropy flow annihilating within the
þ
transmission domain dS 0 \ dS0 in exchange for money.
Note that, if there is average price deflation such that P0 ðtÞ PðtÞ < 0, we
þ
cannot have entropy originate in exchange for money within dS 0 \ dS0 on
ðt; t þ dtÞ because that will have the absurd consequence that bitpulses each with
a negative load of one bit originate in the differential subset dSþ 0 . Thus there is no
way to escape
the conclusion that subject to the constraint of average future price
þ
deflation P0 ðtÞ PðtÞ < 0 the complete entropy flow PZdt H dS 0 \ dS0
represents financial input of bitpulses annihilating within S0 during ðt;t þ dtÞ.
Let us then summarize what we have so far established:
þ
We have claimed that to understand the meaning of transmission H dS 0 \ dS0
we must discern sharply between two mutually exclusive cases:
þ
• The case of average price inflation in S0 with PðtÞ P0 ðtÞ > 0. Subject to this
constraint not only outflux bitpulses are being created within dSþ
0 dS0 on the
selection interval ðt;t þ dtÞ, but on top of that financial entropy is created within
þ
dS0 \ dS0 in exchange for virtual money on the time-interval ðt;t þ dtÞ of
selection. The total of bitpulses thus created on ðt;t þ dtÞ forms the complete
118 5 Exchange and the Circulation of Entropy
collection of output bitpulses of S0. We will call this case the inflationary mode
of evolution.
• The case of average price deflation in S0 with P0 ðtÞ PðtÞ <þ 0. Subject to this
constraint not only influx bitpulses annihilate within dS
0 dS0 on the selection
interval ðt; t þ dtÞ, but on top of that financial entropy is annihilating within
þ
dS0 \ dS0 in exchange for virtual money on the time-interval ðt; t þ dtÞ of
selection. The total of bitpulses thus annihilating on ðt; t þ dtÞ forms the complete
collection of input bitpulses of S0 . We will call this case the deflationary mode of
evolution.
þ
The transmission domain dS 0 \ dS0 is the intersection (the common domain) of
events of origination and events of annihilation. If the constraint of average price
S0 ,þwe
inflation applies within conclude from our explanations in Sect. 4.4 that
the output Zdt H dS0 \ dS0 of originating bitpulses is counterbalanced by a flow
þ
Zdt H dS 0 \ dS0 Δ of “pure” annihilating bitpulses and a flow of Δ
“annihilating” virtual bits with the number Δ yet unsettled. However, since we
have argued in the above that the constraint of average price inflation rules out the
þ
annihilation of “pure” bitpulses within dS 0 \ dS0 , we must now further conclude
that Zdt H dS0 \ dSþ 0 Δ ¼ 0 in the inflationary mode of evolution. That is, in
the inflationary mode of evolution we have only a flow Zdt H dS \ dSþ of
0
þ
0
created “pure” bitpulses counterbalanced by a flow Zdt H dS0 \ dS0 of
“annihilating” virtual bits.
If on the other hand the constraint of average price deflation applies in S0 , we
þ
conclude from our explanations in Sect. 4.4 that the input Zdt H dS 0 dS0 of
\
annihilating bitpulses is counterbalanced by a flow Zdt H dS0 \ dSþ 0 Δ of
“pure” originating bitpulses and a flow of Δ “originating” virtual bits with the
number Δ yet unsettled. However, since we have argued in the above that the
constraint of average price deflation rules out the origination of “pure” bitpulses
þ þ
within dS 0 \ dS 0 , we must now further conclude that Zdt H dS 0 \ dS 0 Δ¼ 0
in the deflationary mode of evolution.
That is,in the deflationary mode of evolution
þ
we have only a flow Zdt H dS \ dS of annihilating “pure” bitpulses
0 0
þ
counterbalanced by a flow Zdt H dS0 \ dS0 of “created” virtual bits.
In summary, there is either a surplus of “annihilating” virtual bits (in the
inflationary mode of evolution) or a surplus of “originating” virtual bits (in the
deflationary mode of evolution).
As the inflationary and deflationary mode of evolution rule out one another,
þ
we may extend similar conclusions to the transmission domains dS 0 \ dSi and
þ
dSi \ dS0 in which all money exchange of the sector Si is being handled. In this
respect we must discern sharply between two mutually exclusive cases of money
exchange within the sector Si :
þ
• In the inflationary mode of evolution with PðtÞ Pi ðtÞ > 0 not only outflux
bitpulses are being created within dSþ
i dS0 on the selection interval ðt; t þ dtÞ,
5.3 The Circulation of Bitpulses 119
Number of
bitpulses
0
t–t t t+q
time t
current excess
lifetime t lifetime q
lifetime t +q
Fig. 5.1 Any bitpulse can be attributed a total lifetime, a current lifetime τ (or age) and an excess
lifetime θ
þ
but on top of that a financial entropy Zdt H dS 0 \ dSi of “pure” bitpulses
þ
is created within dS \ dS in exchange for a counterbalancing flow of
þ
00 i
Zdt H dSi \ dS0 Δ “annihilating” virtual bits on the time-interval ðt; t þ dtÞ
of selection.
Clearly Δ0 must equal the surplus of Zdt H dS þ
i \ dS0 over Zdt H dS0 \ dSi
þ
To set the stage for the treatment of the probability distributions of entropy lifetime
let us reconsider in Fig. 5.1 the various bitpulses that constitute the stock of the set Si
of entropy in the course of time as e.g. earlier depicted in Fig. 4.4. Each bitpulse of
the set Si, observed at time t, has an initial time and a final time, which determines its
total lifetime, current lifetime (or age) and excess lifetime. It has originated at an
120 5 Exchange and the Circulation of Entropy
initial time ðt τÞ prior to observation time t and will become annihilated at a final
time ðt þ θÞ posterior to observation time t. τ is the current lifetime; θ is the excess
lifetime and ðτ þ θÞ is the (total) lifetime of the bitpulse.
We shall be concerned with the probability distribution of current lifetime of the
bitpulses of Si and also with the probability distribution of excess lifetime of the
bitpulses of Si , as observed at time t. At first we shall deal with the probability
distribution of current lifetime of the outflux bitpulses (Sects. 5.3.2 and 5.3.3). Next
we will consider the probability distribution of excess lifetime of the influx
bitpulses (Sect. 5.3.4). Furthermore we will consider in Sect. 5.3.5 the probability
distribution of current lifetime of the output bitpulses. This probability distribution
can only exist in the inflationary mode of evolution. It does not exist in the
deflationary mode of evolution because output consists partly of virtually “created”
entropy in case the deflationary mode of evolution applies. As the variable of
current lifetime of (partly) virtual entropy cannot be defined (it does not exist),
we cannot define its probability distribution either.
Finally we will consider in Sect. 5.3.6 the probability distribution of excess
lifetime of the input bitpulses. This probability distribution can only exist in the
deflationary mode of evolution. It does not exist in the inflationary mode of
evolution because input consists partly of virtually “annihilating” entropy in case
the inflationary mode of evolution applies. As the variable of excess lifetime of
(partly) virtual entropy cannot be defined (it does not exist), we cannot define its
probability distribution either.
Φi ðt,τÞ ¼ Prob y τ dS
0 at time t (5.1)
1 1– F i (t,t )
1
0 t-axis
t-axis
Note further that Ci ðtÞΦi ðt,τÞ is the total number of outflux bitpulses of the set Si,
observed at time t, that have originated since ðt τÞ.
In order to conceive Φi ðt,τÞ as a continuous differentiable function of τ , it is
necessary for Ci to tend to infinity. This requirement will be realized since
theoretically the bit is chosen infinitesimally small (See our previous discussions).
As a matter of fact
Generally, φi ðt,τÞ andΦi ðt,τÞ are functions of observation time t as well and we
have referred to this time-dependency by explicitly quoting the dependence on t.
We must after all draw a sharp distinction between the functional dependence on
t and the functional dependence on τ. Figure 5.2 illustrates the difference between
either dependence for the probability 1 Φi ðt,τÞ in solid geometric perspective.
We refer to current lifetime τ as it is observed at time t. Note, we are not allowed to
identify Φi ðt θ,τ θÞ with Φi ðt,τÞ. That is, the number of outflux bitpulses of Si
having a current lifetime not larger than ðτ θÞ and observed at a time-distance θ
before time t, need not equal the number of outflux bitpulses of Si with a current
lifetime not in excess of τ that are perceived at time t. Similarly Φi ðt θ,τÞ 6¼Φi ðt,τÞ.
122 5 Exchange and the Circulation of Entropy
But, if θ τ; Prob y θ; y τ dS
0 ¼Φi ðt,θÞ Φi ðt,τÞ so that
and if θ < τ; Φi ðθ j y τÞ ¼ 0
After differentiating to θ we arrive at the associate conditional density
φi ðθ j y τÞ:
φi ðt,θÞ
φi ðθ j y τÞ ¼ for θ τ
1 Φi ðt,τÞ
and
φi ðθ j y τÞ ¼ 0 for θ < τ
ζ i ðt,τÞ ¼ φi ðτ j y τÞ
ζ i ðt,τÞdτ is the probability that an outflux bitpulse of Si, observed at time t, has its
current lifetime between τ and τ þ dτ assuming that it is not younger than τ.
Stated differently, ζ i ðt,τÞdτ is the probability that an outflux bitpulse that exists
at observation time t, will have originated at an initial time between t ðτ þ dτÞ and
t τ; assuming we extend our inquiry only to those outflux bitpulses, which
originate not later than time t τ.
Connected with this interpretationζ i ðt,τÞ is called a circulation rate of the outflux
bitpulses that are not younger than τ.
Proceeding further we obtain,
φi ðt,τÞ @ Φi ðt,τÞ=@τ
ζ i ðt,τÞ ¼ φi ðτ j y τÞ ¼ 1
Ð ¼ (5.2)
1 Φi ðt,τÞ
φi ðt,ξÞdξ
τ
5.3 The Circulation of Bitpulses 123
ζ i ðt,τÞCi ðtÞdτ
This number does not include the outflux bitpulses originated between
t ðτ þ dτÞ and t τ that have ceased existence before time t, because they cannot
be observed at time t anymore.
If we choose τ ¼ 0 , we find the number of outflux bitpulses that come into
existence on the infinitesimally small time interval ðt dτ; tÞ of duration dτ. With
dτ ¼ dt this number (in bits) is:
This is what I have called the outflux of Si. ζ i ðtÞ is a rate of outflux circulation of
the sector Si . As PðtÞ is the unit price of a bit of entropy on the selection interval
ðt; t þ dtÞ it follows that the money value of the outflux of Si is equal toζ i PCi dt. This
is what I call the money outflux of Si .
In like manner as we have set out the basic theory of the probability distribution of
current lifetime of outflux bitpulses, we can deal with the theory of the probability
distribution of excess lifetime of influx bitpulses. We will sketch the main lines of
the theory of the distribution of excess lifetime, which is associated with the events
of annihilation of the influx bitpulses only.
LetΨ i ðt,θÞ denote the probability of selecting, from the set Si of Ci ðtÞ bitpulses at
observation time t; an influx bitpulse with excess lifetime not in excess of θ
conditional to the certainty of events within dSþ 0 over the complete time-domain:
Ψ i ðt,θÞ ¼ Prob x θ dSþ
0 at time t (5.4)
Finally note that Ci ðtÞΨ i ðt,θÞ is the total number of influx bitpulses of the set Si ,
observed at time t, which will annihilate before ðt þ θÞ. Clearly,
Generally, ψ i ðt,θÞ and Ψ i ðt,θÞ are functions of observation time t as well and we
have referred to this dependency by explicitly quoting the dependence on t.
The conditional probability distribution Ψ i ðτ j x θÞ is the probability of x τ
assuming x θ:
x θ dSþ
Ψ i ðτ j x θÞ ¼ Prob x τ j 0
Prob x τ; x θ dSþ
¼ 0
by definition
Prob x θ dSþ0
But, if τ θ; Prob x τ; x θ dSþ
0 ¼ Ψ i ðt,τÞ Ψ i ðt,θÞ so that
Ψ i ðt,τÞ Ψ i ðt,θÞ
Ψ i ðτ j x θÞ ¼
1 Ψ i ðt,θÞ
and if τ < θ; Ψ i ðτ j x θÞ ¼ 0
Differentiating to τ to arrive at the associate conditional density ψ i ðτ j x θÞ,
ψ i ðt,τÞ
ψ i ðτ j x θÞ ¼ for; τ θ
1 Ψ i ðt,θÞ
and
ψ i ðτ j x θÞ ¼ 0 for τ < θ
ρi ðt,θÞ ¼ ψ i ðθ j x θÞ
5.3 The Circulation of Bitpulses 125
ρi ðt,θÞdθ is the probability that an influx bitpulse of Si, observed at time t, has its
excess lifetime between θ and θ þ dθ assuming that it will still live longer than θ.
Stated differently, ρi ðt,θÞdθ is the probability that an influx bitpulse that exists at
observation time t, will annihilate at a final time between t þ θ and t þ ðθ þ dθÞ,
assuming we extend our inquiry to influx bitpulses that do not annihilate anterior to
time t þ θ.
Connected with this interpretation ρi ðt,θÞ is a circulation rate of influx bitpulses
that live longer than θ.
Proceeding further we obtain,
ψ i ðt,θÞ @ Ψ i ðt,θÞ=@θ
ρi ðt,θÞ ¼ ψ i ðθ j x θÞ ¼ 1
Ð ¼ (5.5)
1 Ψ i ðt,θÞ
ψ i ðt,ξÞdξ
θ
ρi ðt,θÞ is a characteristic rate of influx bitpulses that will annihilate at final time
t þ θ.
Since Ci ðtÞ is the number of bitpulses of Si , it follows that the number of influx
bitpulses observed in Si (observed at time t) that has an excess lifetime between t þ θ
and t þ ðθ þ dθÞ must be equal to:
ρi ðt,θÞCi ðtÞdθ
This number does not include the influx bitpulses annihilating between t þ θ and
t þ ðθ þ dθÞ that will come into existence after time t , because they cannot be
observed at time t when they don’t yet exist.
If we choose θ ¼ 0, we find the number of influx bitpulses that cease existence on
the infinitesimally small time interval ðt,t þ dθÞ of duration dθ. With dθ ¼ dt this
number (in bits) is:
ρi ðt,0ÞCi ðtÞdt ¼
ρi ðtÞCi ðtÞdt (5.6)
Like the collection of outflux bitpulses and the collection of influx bitpulses of Si
have their characteristic time-dependent probability distributionsΦi ðt,τÞ andΨ i ðt,θÞ
of current lifetime τ and excess lifetime θ, the collection of output bitpulses has its
126 5 Exchange and the Circulation of Entropy
In this, y is now the random variable of current lifetime τ of an output bitpulse of Si.
Furthermore note the condition that the events within dS 0 are certain to occur, as
stated in Sect. 5.3.2 for the outflux bitpulses, has been dismissed for the output
bitpulses. The condition that the events dS 0 are certain to occur defines only the
outflux events that occur in Si , but we are here concerned with all the output
bitpulses.
The partial derivative φi ðt,τÞ of Φi ðt,τÞ to current lifetime τ is the probability
density function of current lifetime of the output bitpulses:
@ Φi ðt,τÞ
φi ðt,τÞ ¼
@τ
granted that Φi ðt,τÞ is differentiable to all τ.
The conditional probability distribution Φi ðθ j y τÞ is the probability of y θ
assuming y τ:
Probfy θ; y τg
Φi ðθ j y τÞ ¼ Probfy θ jy τg ¼ by definition
Probfy τg
φi ðt,θÞ
φ i ðθ j y τ Þ ¼ for θ τ
1 Φi ðt,τÞ
and
φi ðθ j y τÞ ¼ 0 for θ < τ
The circulation rate ζ i ðt,τÞ of the output bitpulses not younger than τ is defined as
ζ i ðt,τÞ ¼ φi ðτ j y τÞ
5.3 The Circulation of Bitpulses 127
φi ðt,τÞ @ Φi ðt,τÞ=@τ
ζ i ðt,τÞ ¼ φi ðτ j y τÞ ¼ 1
Ð ¼
1 Φi ðt,τÞ
φi ðt,ξÞdξ
τ
ζ i ðt,τÞCi ðtÞdτ
This number does not include the output bitpulses originated between
t ðτ þ dτÞ and t τ that have ceased existence before time t, because they cannot
be observed at time t anymore.
If we choose τ ¼ 0; we find the number of output bitpulses that come into
existence on the infinitesimally small time interval ðt dτ; tÞ of duration dτ. With
dτ ¼ dt this number (in bits) is:
ζ i ðtÞ is a rate of output circulation of the sector Si. This is what I have called the
output of Si . As PðtÞ is the unit price of a bit of entropy on the selection interval
ðt; t þ dtÞ it follows that the money value of the output of Si is equal to ζ i PCi dt. This
is what I call the money output of Si .
arbitrary bitpulse from the set Si of input bitpulses constituting its stock that has an
excess lifetime smaller than or equal to θ:
In this, x is now the random variable of excess lifetime θ of an input bitpulse of Si.
Furthermore note the condition that the events within dSþ 0 are certain to occur, as
stated in Sect. 5.3.4 for influx bitpulses, has been dismissed. That condition defines
the influx events only. However we are here dealing with all the input events of
annihilation within Si . This represents the unconditional complete set of input
events.
The partial derivative ψ i ðt,θÞ of Ψ i ðt,θÞ to excess lifetime θ is the probability
density function of excess lifetime of the input bitpulses:
@ Ψ i ðt,θÞ
ψ i ðt,θÞ ¼
@θ
Probfx τ; x θg
Ψ i ðτ j x θÞ ¼ Probfx τ jx θg ¼ by definition
Probfx θg
and
ψ i ðτ j x θÞ ¼ 0 for τ < θ
The circulation rate ρi ðt,θÞ of the input bitpulses not younger than θ is defined as
ρi ðt,θÞ ¼ ψ i ðθ j x θÞ
ψ i ðt,θÞ @ Ψ i ðt,θÞ=@θ
ρi ðt,θÞ ¼ ψ i ðθ j x θÞ ¼ 1
Ð ¼
1 Ψ i ðt,θÞ
ψ i ðt; ξÞdξ
θ
5.4 The Probability Distribution of Outflux and Influx and the Markov Property 129
ρi ðt,θÞ is a characteristic rate of input bitpulses that will cease to exist at final
time t þ θ ðθ > 0Þ. Since Ci ðtÞ is the number of bitpulses of Si , it follows that the
number of input bitpulses observed in Si that has an excess lifetime between t þ θ
and t þ ðθ þ dθÞ must be equal to:
ρi ðt,θÞCi ðtÞdθ
This number does not include the input bitpulses annihilating between t þ θ and
t þ ðθ þ dθÞ that will come into existence after time t , because they cannot be
observed at time t when they don’t yet exist.
If we choose θ ¼ 0, we find the number of input bitpulses that cease existence on
the infinitesimally small time interval ðt,t þ dθÞ of duration dθ. With dθ ¼ dt this
number (in bits) is:
This is what I have called the input of Si . ρi ðtÞ is a rate of input circulation of
the sector Si . As PðtÞ is the unit price of a bit of entropy on the selection interval
ðt; t þ dtÞ it follows that the money value of the input of Si is equal to ρi PCi dt. This is
what I call the money input of Si .
Mark dSþ þ
0 dS0 does not overlap with dS0 dS0 (See the Venn diagram). If we let
dS0 represent the differential sample space dS0 of an economy, it must follow
that for this economy dS0 ¼ d S0 and hence that this must be an economy for which
þ
dS0 \ dS0 is the null set. That is
dS þ
0 \ dS0 ¼
In accordance with what has been derived in Appendices F and G this implies
that for this economy all the outflow events are statistically independent
of all the
þ
inflow events. As only outflux events are involved within dS0 dS0 and only influx
þ
events are involved in dS
0 dS0 , we conclude that generally outflux events are
statistically independent of influx events.
130 5 Exchange and the Circulation of Entropy
observed to be present at time t θ (with θ < τ) and for that reason have not yet (at
time t θ) ceased to exist on ðt τ,t θÞ.
It is obvious that S1 is a subset of S2 and that in turn S2 is a subset of S3 so that
S1 S2 S3
The set S1 is a subset of the set S2 , the latter containing all the original outflux
bitpulses that were observed at time t θ > t τ and that came into existence
between t τ dτ and t τ. Additional to the subset S1; S2 contains also the outflux
bitpulses of S3 that will annihilate on the time-interval ðt θ; tÞ.
Recall that the circulation rate of S1 on the time-interval ðt τ dτ; t τÞ is
ζ i ðt; τÞ at observation time t. Since the outflux bitpulses of S2 are observed at time
t θ (with θ < τ), their circulation rate on the time-interval ðt τ dτ,t τÞ is
ζ i ðt θ,τ θÞ.
S1 is the remaining subset of bitpulses of S2 only after a process of random
annihilation has removed these bitpulses from S2 during the time-interval of
ðt θ; tÞ. However, since the random selection of annihilating bitpulses at time-
instances that differ from the time-instances of creating bitpulses are mutually
independent as a consequence of the Markov property, it must follow that the
statistical averages of outflux of S2 on ðt τ dτ,t τÞ cannot be affected by
the process of annihilation. Therefore, it follows that the outflux bitpulses in S1 that
remain—after the outflux bitpulses from S2 have been removed by a process of
random annihilation—share the same circulation rate with the outflux bitpulses of
S2 on the time-interval between t τ dτ and t τ. That is,
ζ i ðt,τÞ ¼ ζ i ðt τÞ (5.8)
With this in mind, let us integrate both sides of (5.8) over the τ- domain. Then,
with the help of (5.2):
ðτ
ln 1 Φi ðt,τÞ ¼ ζ i ðt ξÞ dξ (5.9)
0
This results in
8 τ 9
< ð =
Φi ðt,τÞ ¼ 1 exp ζ i ðt ξÞ dξ (5.10)
: ;
0
132 5 Exchange and the Circulation of Entropy
and
8 τ 9
< ð =
φi ðt,τÞ ¼ ζ i ðt τÞ exp ζ i ðt ξÞ dξ (5.11)
: ;
0
ρi ðt,θÞ ¼
ρi ð t þ θ Þ (5.12)
ðθ
ln 1 Ψ i ðt,θÞ ¼
ρi ðt þ ξÞ dξ (5.13)
0
8 θ 9
< ð =
Ψ i ðt,θÞ ¼ 1 exp ρi ðt þ ξÞ dξ (5.14)
: ;
0
8 θ 9
< ð =
ψ i ðt,θÞ ¼
ρi ðt þ θÞ exp ρi ðt þ ξÞ dξ (5.15)
: ;
0
Equations (5.8) and (5.12) emphasize that the outflux and influx circulation rates
depend on current time t only.
We are here dealing with the inflationary mode of evolution. This implies that the
probability distribution Φi ðt,τÞ of current lifetime τ of output bitpulses exists and
that the probability distribution Ψ i ðt,θÞ of excess lifetime θ of input bitpulses does
not exist.
5.5 The Circulation of Output in the Inflationary Mode of Evolution 133
As Cþ þ
i ðt,ξÞ ¼ Ci ðtÞΦi ðt,ξÞ, the gross number @ξ Ci ðt,ξÞ of bitpulses originating
between t ξ dξ and t ξ is
@ξ Cþ
i ðt,ξÞ ¼ Ci ðtÞ φi ðt,ξÞdξ (5.16)
On the other hand, as Viþ ðt,ξÞ ¼ Viþ ðtÞ Φi ðt,ξÞ, the money @ξ Viþ ðt,ξÞ exchanged
in dSþ
i for the acquirement of these bitpulses between t ξ dξ and t ξ is
because φi ðt; ξÞdt is the gross chance that an output bitpulse is created between
t ðξ þ dξÞ and t ξ . The bitpulses that have emerged in the sector Si are
exchanged for unit price Pðt ξÞ at time t ξ. The money value exchanged is then:
and
Viþ Viþ ðt,ξÞ ¼ Viþ ½1 Φi ðt,ξÞ (5.20)
Herein Ci 1 Φi ðt,ξÞ is the entropy stock of the bitpulses that have originated
before time t ξ and Viþ ½1 Φi ðt,ξÞ is the money exchange value of the bitpulses
that have originated before time t ξ.
It follows further from (5.18) and (5.16) that
@ þ
V ðt,ξÞ ¼ Pðt ξÞCi ðtÞφi ðt,ξÞ
@ξ i
If τ tends to infinity, we get the formula for the money exchange value Viþ ðtÞ of
all the bitpulses that reside in S0 at time t and that have originated in state i in the past.
That is,
ð
1
þ þ
Vi ðtÞ ¼ Vi ðt,1Þ ¼ Ci ðtÞ Pðt ξÞφi ðt,ξÞdξ (5.21)
0
5.5 The Circulation of Output in the Inflationary Mode of Evolution 135
þ
The quotient of Viþ ðtÞ and Ci ðtÞ is the average bit price Pi per bit spent within Si
in the past as observed at current time t. It follows that
þ Viþ ðtÞ
Pi ðtÞ ¼ (5.22)
Ci ðtÞ
þ
Of course Pi ðtÞ is also a random function of time.
It follows further that
ð
1
þ
Pi ¼ Pðt ξÞφi ðt,ξÞdξ (5.23)
0
þ
Pi ðtÞ is the average historic unit price of the outflux bitpulses stocked in S0 at
time t that have originated in state i in the past. This is the average historic unit price
for which these bits have been acquired in the past. It does not match the current
unit bit price PðtÞ. Mark that Viþ ðtÞ is not equal to the product of Ci ðtÞ and PðtÞ,
because the average historical unit price spent on outflux bits in the past (before
observation time t) is another than the current unit price PðtÞ at time t.
I shall now derive two important theorems that relate the circulation of the output
bitpulses to the circulation of the outflux bitpulses. The first theorem to be derived is
þ
ζ i ðtÞ ¼ ζ i ðtÞ þ pþ
i ðtÞ conditional to PðtÞ Pi ðtÞ 0 (5.24)
þ
Mark that we claim (5.24) to be valid only if PðtÞ Pi ðtÞ 0, i.e. if there is
average price inflation within Si at current time t . This marks the inflationary
constraint for the specific sector Si .
Proof of theorem (5.24): The right side Viþ ½1 Φi ðt,ξÞ of expression (5.20) renders
the money exchange value of output bitpulses that have originated before time t ξ.
þ
Now recall from (5.22) that Pi ðt ξÞ is the average price spent per bit within Si in
the past at time t ξ. Further, the number of outflux bitpulses
that have originated
before time t ξ and are still present at time t is Ci 1 Φi ðt,ξÞ . It follows that
þ
Pi ðt ξÞ must also be the quotient of Viþ ½1 Φi ðt,ξÞ and Ci 1 Φi ðt,ξÞ . That is,
þ
Viþ ½1 Φi ðt,ξÞ ¼ Pi ðt ξÞ Ci 1 Φi ðt; ξÞ
136 5 Exchange and the Circulation of Entropy
þ
Integration of Pi ðt τÞ over the time-domain τ yields
8 ξ 9
< ð =
þ þ
Pi ðt ξÞ ¼ Pi ðtÞ exp pþ
i ðt τ Þdτ
: ;
0
φi ðt; ξÞ φi ðt; ξÞ
¼ þ pþ
i ðt ξÞ
1 Φi ðt; ξÞ 1 Φi ðt; ξÞ
or
ζ i ðt,ξÞ ¼ ζ i ðt; ξÞ þ pþ
i ðt ξÞ
ζ i ðt,ξÞ ¼ ζ i ðt ξÞ þ pþ
i ðt ξÞ (5.25)
The right side of this expression is a function of t ξ only. Hence the left side of
the expression must be a function of t ξ only. That is,
ζ i ðt ξÞ ¼ ζ i ðt ξÞ þ pþ
i ðt ξ Þ
there would then be average price deflation rather than average price inflation. Thus
to satisfy the constraint for which (5.24) holds true, it is also required that
þ
pþ
i ðtÞ 0 conditional to PðtÞ Pi ðtÞ 0 (5.26)
ζ i ðt,τÞ ¼ ζ i ðt τÞ (5.27)
dPi þ þ
¼ pþP
i i ¼ ζ i P P i for pþ
i 0 (5.29)
dt
þ ð
1 ð
1
dPi @ Pð t ξ Þ @ φi ðt,ξÞ
¼ φi ðt,ξÞdξ þ P ðt ξÞ dξ
dt @t @t
0 0
138 5 Exchange and the Circulation of Entropy
@ Pð t ξ Þ @ Pðt ξÞ
With ¼ and after partial integration of the first integral
@t @ξ
over the ξ- domain,
þ ð
1
dPi @ φi ðt,ξÞ @ φi ðt,ξÞ
¼ ζi P þ Pðt ξÞ þ dξ (5.30)
dt dξ dt
0
Let us first deal with the factor of the integrand between braces. In virtue of (5.11):
8 ξ 9
< ð =
φi ðt,ξÞ ¼ ζ i ðt ξÞ f in which the factor f ¼ exp ζ i ðt τÞdτ
: ;
0
Also
@ ζ i ðt ξÞ @ ζ ðt ξÞ
¼ i
@t @ξ
So,
@ φi ðt,ξÞ @ζ ðt ξÞ 2
¼ i f ζ i ðt ξÞ f
@ξ @t
and
@ φi ðt,ξÞ @ ζ i ðt ξÞ
¼ f þ ζ i ðt ξÞ ζ i ðtÞ ζ i ðt ξÞf
@t @t
These expressions may be inserted in (5.30). This yields after some elaboration:
þ ð
1
dPi
¼ ζi P Pðt ξÞζ i ðtÞφi ðt,ξÞdξ
dt
0
þ
The second term of the right side is ζ i Pi on behalf of (5.23), so that theorem
(5.29) is established. □
With (5.29) proven, the observation expressed in (5.26) has been verified to hold
þ
analytically. This implies that the constraint PðtÞ Pi ðtÞ 0, which marks and
defines the inflationary mode of evolution within Si (See Sect. 5.2), may just as well
be replaced by the constraint pþ i ðtÞ 0 . Thus the theory derived in the present
section holds under the constraint that pþ
i ðtÞ 0. □
5.6 The Circulation of Input in the Deflationary Mode of Evolution 139
To prove theorem (5.28) we will rely on theorems (5.24) and (5.29). It follows
from (5.24) and (5.29) that
þ þ
ζ i ðtÞPi Ci ¼ ζ i ðtÞ þ pþ
i ðtÞ Pi Ci
or
þ þ þ
ζ i Pi Ci ¼ ζ i Pi Ci þ ζ i P Pi Ci ¼ ζ i PCi
and
ζ i Viþ dt ¼ ζ i PCi dt
We are here dealing with the deflationary mode of evolution. This implies that the
probability distribution Ψ i ðt,θÞ of excess lifetime θ of the input bitpulses exists and
that the probability distribution Φi ðt,τÞ of current lifetime τ of output bitpulses does
not exist.
In foregoing sections we have explained how the probability distributions
Ψ i ðt,θÞ and Ψ i ðt,θÞ of excess lifetime relate respectively to the rate of influx
circulation and to the rate of input circulation in this particular case. We shall
now bridge the connection between Ψ i ðt,θÞ and Ψ i ðt,θÞ. The relationship between
Ψ i ðt,θÞ and Ψ i ðt,θÞ is also closely related with the time-course of the unit price PðtÞ
of a bit of entropy. As we are here dealing with the annihilation of bitpulses within
þ
dS
i (i.e. including the transmission domain dS0 \ dSi ), for analysis to make sense
the deflationary mode of evolution can only apply.
Note that Ψ i ðt,θÞ ¼ Ψ i ðt,θÞ if P is a constant over the entire time-domain. Recall
also our discussion with respect to the deflationary mode of evolution in Sect. 5.2.
In this section we demonstrated that if the constraint Pi ðtÞ PðtÞ < 0 of average
price deflation applies, all entropy annihilation (entropy input) is in dS i (i.e. in
dS dSþ as well as in dS \ dSþ ). Financial entropy annihilation in dS \ dSþ
i 0 i 0 i 0
occurs in exchange for virtual entropy (money).
140 5 Exchange and the Circulation of Entropy
We shall now consider the structure of the probability distribution of the input
bitpulses subject to the above constraint of average price deflation. Meanwhile we
will solve how the average future unit price Pi ðtÞ of Si relates to the current unit
price PðtÞ, a question that we haven’t yet considered in Sect. 5.2.
We must follow the money value involved with the exchange of entropy. We can
look at that money exchange value from two perspectives: that of the money
liberated with the annihilation of influx bitpulses and that of the money liberated
with the sacrifice of input bitpulses. Irrespective of the perspective chosen the
money value liberated/exchanged cannot be different.
Let Ci ðt,ξÞ be the number of bitpulses residing in S0 at observation time t that will
annihilate in state i prior to time t þ ξ. Let further Vi ðt,ξÞ be the aggregate total of
the money value exchanged at time t in consequence of the annihilation of bitpulses
within Si before time t þ ξ.
þ
C
i ðt,1Þ ¼ Ci ðtÞ ¼ Ci ðtÞ ¼ Ci ðtÞ is capacity, the total of the entropy of
bitpulses, residing in S0 at time t and that will cease existence within Si in the
future, i.e. after time t (See Sect. 5.1). Mark also that Vi ðt,1Þ is the future money
exchange value of bitpulses, residing in S0 at time t and that will cease existence in Si
in the future, i.e. after time t. Clearly, Vi ðt,1Þ like Ci ðtÞ depends on time t only and
for this reason we may write
Recall that Vi ðtÞ6¼ Viþ ðtÞ for i 6¼ 0 and that V0 ðtÞ ¼ V0þ ðtÞ (See Sect. 5.1).
Let us next consider how C
i ðt,ξÞ and Vi ðt,ξÞ contribute to the gross expiration of
bitpulses and respectively to the money value for which they are being exchanged
on this infinitesimally small time-interval ðt þ ξ,t þ ξ þ dξÞ.
As C
i ðt,ξÞ ¼ Ci ðtÞΨ i ðt,ξÞ, the gross number @ξ Ci ðt,ξÞ of bitpulses annihilating
between t þ ξ and t þ ξ þ dξ is
@ξ C
i ðt,ξÞ ¼ Ci ðtÞψ i ðt,ξÞdξ (5.31)
On the other hand, as Vi ðt,ξÞ ¼ Vi ðtÞ Ψ i ðt,ξÞ, the money @ξ Vi ðt,ξÞ liberated in
dS
i with the sacrifice of these bitpulses between t þ ξ and t þ ξ þ dξ is
because ψ i ðt; ξÞdt is the gross chance that an input bitpulse is annihilating between
t þ ξ and t þ ξ þ dξ.
The bitpulses that cease existence in the sector Si are being exchanged for unit
price Pðt þ ξÞ at time t þ ξ. The money value exchanged is then:
Ci C
i ðt,ξÞ ¼ Ci 1 Ψ i ðt,ξÞ (5.34)
and
Vi Vi ðt,ξÞ ¼ Vi ½1 Ψ i ðt,ξÞ (5.35)
Herein Ci 1 Ψ i ðt,ξÞ is the entropy stock of the bitpulses that will cease
existence after time t þ ξ and Vi ½1 Φi ðt,ξÞ is the money exchange value of the
bitpulses that will annihilate after time t þ ξ.
It follows further from (5.33) and (5.31) that
@
V ðt,ξÞ ¼ Pðt þ ξÞCi ðtÞψ i ðt,ξÞ
@ξ i
If τ tends to infinity, we get the formula for the money exchange value Vi ðtÞ of
all the bitpulses that reside in S0 at time t and that cease existence in state i in the
future. That is,
ð
1
The quotient of Vi ðtÞ and Ci ðtÞ is the average bit price Pi per bit to be liberated/
exchanged within Si in the future as observed at current time t. It follows that
Vi ðtÞ
Pi ðtÞ ¼ (5.37)
Ci ðtÞ
Of course Pi ðtÞ is also a random function of time.
It follows further that
ð
1
Pi ¼ Pðt þ ξÞψ i ðt,ξÞdξ (5.38)
0
Pi ðtÞ is the average future unit price of the influx bitpulses stocked in S0 at time t
that will cease existence in state i in the future. This is the average unit price for
which these bits will be sacrificed in the future. Pi ðtÞ does not match the current
unit bit price PðtÞ. Mark Vi ðtÞ is not equal to the product of Ci ðtÞ and PðtÞ, because
142 5 Exchange and the Circulation of Entropy
the average future unit price to be sacrificed on influx bits in the future (after
observation time t) is another than the current unit price PðtÞ at time t.
As in Sect. 5.5.2 I shall derive two important theorems that relate the circulation of
the input bitpulses to the circulation of the influx bitpulses. The first theorem to be
derived is
ρi ðtÞ ¼ ρi ðtÞ þ p
i ðtÞ conditional to Pi ðtÞ PðtÞ 0 (5.39)
Mark that we claim (5.39) to be valid only if Pi ðtÞ PðtÞ 0, i.e. if there is
average price deflation within Si at current time t . This marks the deflationary
constraint for the specific sector Si .
Proof of theorem (5.39): The right side Vi ½1 Ψ i ðt,ξÞ of expression (5.35) renders
the money exchange value of input bitpulses that will cease to exist after time t þ ξ.
Now recall from (5.37) that Pi ðt þ ξÞ is the average price sacrificed per bit in the
future as observed at time t þ ξ. Further, the number of influx bitpulses that will
annihilate after time t þ ξ and that are still present at time t is Ci 1 Ψ i ðt,ξÞ . It
follows that Pi ðt þ ξÞ must also be the quotient of Vi ½1 Ψ i ðt,ξÞ and
Ci 1 Ψ i ðt,ξÞ . That is,
Vi ½1 Ψ i ðt,ξÞ ¼ Pi ðt þ ξÞCi 1 Ψ i ðt; ξÞ
The integration of Pi ðt þ θÞ over the time-domain θ yields
8ξ 9
<ð =
P i ðt þ ξÞ ¼ Pi ðtÞ exp p
i ðt þ θ Þdθ
: ;
0
ψ i ðt,ξÞ ψ i ðt,ξÞ
¼ p
i ðt þ ξÞ
1 Ψ i ðt,ξÞ 1 Ψ i ðt,ξÞ
or
ρi ðt,ξÞ ¼ ρi ðt,ξÞ p
i ðt þ ξ Þ
ρi ðt,ξÞ ¼ ρi ðt þ ξÞ p
i ðt þ ξÞ (5.41)
The right side of this expression is a function of t þ ξ only. Hence the left side of
the expression must be a function of t þ ξ only. That is,
ρi ðt þ ξÞ ¼ ρi ðt þ ξÞ p
i ðt þ ξÞ
ρi ðt,θÞ ¼ ρi ðt þ θÞ (5.43)
ðθ
lnf1 Ψ i ðt,θÞg ¼ ρi ðt þ ξÞ dξ
0
8 θ 9
< ð =
Ψ i ðt,θÞ ¼ 1 exp ρi ðt þ ξÞ dξ
: ;
0
144 5 Exchange and the Circulation of Entropy
8 θ 9
< ð =
ψ i ðt,θÞ ¼ ρi ðt þ θÞ exp ρi ðt þ ξÞ dξ
: ;
0
Proof of theorem (5.45): Differentiate the expression (5.38) to time t. It follows that
ð
1 ð
1
dPi @ Pð t þ ξ Þ @ ψ i ðt,ξÞ
¼ ψ i ðt; ξÞ dξ þ Pð t þ ξ Þ dξ
dt @t @t
0 0
@ Pðt þ ξÞ @ Pðt þ ξÞ
With ¼ and after partial integration of the first integral
@t @ξ
over the ξ- domain,
ð
1
dPi @ ψ i ðt,ξÞ @ ψ i ðt,ξÞ
¼ ρi P þ Pðt þ ξÞ dξ (5.46)
dt dt dξ
0
Let us first deal with the factor of the integrand between braces. In virtue of
(5.15):
8 ξ 9
< ð =
ψ i ðt,ξÞ ¼ ρi ðt þ ξÞ f in which the factor f ¼ exp ρi ðt þ θÞdθ
: ;
0
Also
@ ρi ðt þ ξÞ @ ρi ðt þ ξÞ
¼
@t @ξ
5.6 The Circulation of Input in the Deflationary Mode of Evolution 145
So,
@ ψ i ðt,ξÞ @ ρi ðt þ ξÞ
¼ f fρi ðt þ ξÞg2 f
@ξ @t
and
@ ψ i ðt,ξÞ @ ρi ðt þ ξÞ
¼ f þ fρi ðtÞ ρi ðt þ ξÞg ρi ðt þ ξÞf
@t @t
These expressions may be inserted in (5.46). This yields after some elaboration:
ð
1
dPi
¼ ρi P þ Pðt þ ξÞρi ðtÞψ i ðt,ξÞdξ
dt
0
The second term of the right side is ρi Pi on behalf of (5.38), so that theorem
(5.45) is established. □
With (5.45) proven, the observation expressed in (5.42) has been verified to hold
analytically. This implies that the constraint Pi ðtÞ PðtÞ 0, which marks and
defines the deflationary mode of evolution within Si (See Sect. 5.2), may just as well
be replaced by the constraint p i ðtÞ 0 . Thus the theory derived in the present
section holds under the constraint that p
i ðtÞ 0. □
To prove theorem (5.44) we will rely on theorems (5.39) and (5.45). It follows
from (5.39) and (5.45) that
dPi
ρi ðtÞVi ðtÞ ¼ ρi ðtÞPi Ci ¼ ρi ðtÞ p
i ðtÞ P i Ci ¼ ρ i ðtÞP i Ci C i
dt
or
ρi Vi ¼ ρi Pi Ci þ ρi P Pi Ci ¼ ρi PCi
and
ρi Vi dt ¼ ρi PCi dt
Abstract Each differential subset of the Venn diagram is associated with a partic-
ular economic flow. The assignment differs dependent on whether the inflationary
or the deflationary mode of evolution applies. The inflationary and deflationary
modes of evolution are mutually exclusive. Only one of these two alternative modes
can apply within all the subsets of the economy simultaneously at a particular time.
It is demonstrated that the point of transition from the inflationary to the deflation-
ary mode (and conversely) is always in the turning point of evolution. In this point
there is absence of statistical dependence between input and output. Also HðX0 \ Y0 Þ
¼ 0 , H ðX0 Þ ¼ H ðY0 Þ ¼ 0 and the rate of economic growth is zero in this point.
Statistical dependence between input and output (which implies both that HðX0 \ Y0 Þ
> 0 and that there is a sufficient rate of price inflation) is crucial for evolution to
succeed and to go on. The absence of statistical dependence smothers evolution.
Positive evolutionary growth is impossible in the deflationary mode. This explains
that economies can only manifest themselves if there is a small but persistent positive
rate of price inflation on an average. Then a positive rate of capacity growth will be
maintained in case labor input probability exceeds consumption probability (labor
output probability). The general line is that to attain more positive economic growth,
investment must be stimulated relative to consumption.
It is further claimed that to maintain positive evolutionary growth all evolution-
ary systems must have a unit of exchange additional to the unit of selection (the bit
of entropy) that slowly revaluates relative to the unit of exchange.
ζ i PCi dt ¼ ζ i PCi dt þ pþ þ
i PCi dt with pi 0 (6.1)
and
P
þ
Financial output ¼ pþ
i PCi dt ¼ ζ i P Pi þ dt
Pi
and
þ
pþ H dSþ
i PCi dt ¼ PZdt H dSi i dS0 ¼ PZdt H ðYi \ X0 Þ
þ
ζ i P ¼ ζ i Pi (6.5)
150 6 The Interpretation of the Economic Variables
In accordance with (2.4) the evolutionary growth dCi of the sector Si is the net
result of two opposing flows: output Yi of Si and input Xi of Si :
λi
Zdt H dS þ þ
i \ dS0 ¼ λi Zdt H dS0 \ dS0 ¼ Zdt H dS þ
0 \ dSi
μi
λi þ
¼ p Ci dt
μi i
þ
Let us then define p
i and ρi as follows for pi 0, i.e., in case the inflationary
mode of evolution applies:
λi þ
p
i ¼ p for pþ
i 0 (6.7)
μi i
and
ρi ¼ ρi þ p
i for pþ
i 0 (6.8)
and
Xi ¼ ρi Ci dt ¼ Zdt H dS
i
þ
Finally recall from Sects. 5.2 and 5.3.4 that the entropy influx Zdt H dS
i dS0
of the sector Si is defined by
þ
Xi ¼ ρi Ci dt ¼ Zdt H dS
i dS0 (6.10)
Let us next consider the context of outflux, output, influx and input in the deflationary
mode of evolution, for which the theory set out in Sects. 5.3.6 and 5.6 and the
constraint p i 0 is applicable.
We have argued in the foregoing that in the deflationary mode of evolution ði:e:
þ
for P Pi Þ bitpulses annihilate within dS i \ dS0 in exchange for virtual entropy.
On behalf of (5.44) we may then ascertain that there is a money value of
ψ i ðt; 0ÞVi dt ¼ ρi ðtÞVi dt of influx bitpulses annihilating within Si during ðt; t þ dtÞ.
This is the money influx of Si .
Further, on behalf of (5.6) there is a number of bitpulses annihilating within Si
during ðt; t þ dtÞ and equal to ρi Ci dt. This is an entropy influx of Si . It corresponds
with a money value ρi PCi dt because the unit price of a bit is P during the selection
interval ðt; t þ dtÞ.
The link ρi PCi dt ¼ ρi Pi Ci dt ¼ ρi Vi dt between the entropy influx ρi Ci dt and its
money value ρi Vi dt was established by the proof of (5.44).
The number of all the bitpulses that annihilate in Si on ðt; t þ dtÞ is ρi Ci dt (see
Sect. 5.3.6). This is the entropy input of Si . These bitpulses will annihilate in the
þ
entire differential set dS i , either within the conditional differential set dSi dS0 as
þ
entropy influx ρi Ci dt or within the intersection dS i \ dS0 as financial input of Si in
exchange for money. The expression for the financial input follows from (5.39) for
p
i 0, which characterizes the deflationary mode of evolution:
ρi PCi dt ¼ ρi PCi dt p
i PCi dt with pi 0 (6.11)
P
p
i PCi dt ¼ ρi Pi P
with pi 0 (6.12)
Pi
and
P
Financial input ¼ p
i PCi dt ¼ ρi Pi P
Pi
152 6 The Interpretation of the Economic Variables
and
þ
p H dS
i PCi dt ¼ PZdt H dSi i dS0 ¼ PZdt H ðXi \ Y0 Þ
In the deflationary mode of evolution Yi ¼ Zdt H dSþ is the sum of entropy
þ i
outflux Y i ¼ Zdt H dSi dS0 and a virtual component Zdt H dS þ
0 \ dSi .
Note further that generally
μi
Zdt H dS þ
0 \ dSi ¼ μi Zdt H dS þ
0 \ dS0 ¼ Zdt H dS þ
i \ dS0
λi
μi
¼ pi Ci dt
λi
μi
pþ
i ¼ p for p
i 0 (6.16)
λi i
and
ζ i ¼ ζ i pþ
i for p
i 0 (6.17)
and
Yi ¼ ζ i Ci dt ¼ Zdt H dSþ
i
Y i ¼ ζ i Ci dt ¼ Zdt H dSþ
i dS0 (6.18)
Table 6.1 presents a summary and glossary of the various flows, both for the
inflationary and deflationary mode of evolution.
We have argued in Sect. 5.2 that the mode of inflationary evolution and the mode of
deflationary evolution in Si are mutually exclusive. That is, either pþ
i > 0 or pi < 0.
At first glance the dependence of these constraints on the state i might suggest
that the mode of evolution differs dependent on the sector Si we consider. However,
this is not the case.
To substantiate this claim, my task is here to demonstrate that the mode of
evolution is inflationary in the sector Si whenever it is inflationary in S0 and
conversely. Likewise my task is to demonstrate that the mode of evolution is
deflationary in the sector Si whenever it is deflationary in S0 and conversely.
Well, it is easy to see that if pþ þ
0 > 0 applies, it must follow that pi > 0 for any
state i.
For if pþ þ
0 > 0 , bitpulses can only originate in dS0 \ dS0 in exchange for
þ
“annihilating” virtual bits. Thus in any subset of dS0 \ dS0 , more specifically the
154
Table 6.1 The nomenclature of the flows and circulation rates of the sector Si of the economy
Inflationary mode of evolution pþ
i 0 Deflationary mode of evolution p
i 0
Virtual Virtual
Flow category Flow variable Yes or no Residence Equivalent Yes or no Residence Equivalent
Entropy output ζi Ci dt N dSþ
i ζi Ci dt ¼ ðζi þ pþ
i ÞCi dt Partly dSþ
i ζ i Ci dt ¼ ζ i μi p
i =λi Ci dt
Entropy outflux ζi Ci dt N dSþ ζi Ci dt N dSþ ζ i Ci dt
i dS0 i dS0
þ
Financial output p Ci dt
i
N dS0 \ dSþi Zdt H dS \ dSþ
i
Y dS0 \ dSþi μi p i Ci dt=λi
0
Entropy input ρi Ci dt Partly dS
i ρi Ci dt ¼ ρi þλi pþi =μi Ci dt
N dS
i ρi Ci dt ¼ ρi p i Ci dt
Entropy influx ρi Ci dt N dS dSþ
i 0
ρi Ci dt N dS dSþ
i 0
ρi Ci dt
þ þ
þ
Financial input p Ci dt
i
Y dS
i \ dS0 λi pþ
i Ci dt=μi N dS
i \ dS0 Zdt H dS i \ dS0
Money output ζi PCi dt N dSþ
i ζi PCi dt Partly dSþ
i ζ i PCi dt
Money outflux N N þ
ζi PCi dt dSþ
i dS0
ζi PCi dt dSþ
i dS0 P ζi Ci dt ¼ ζ i PCi dt
i
Money input ρi PCi dt Partly dS
i ρi PCi dt N dS
i ρi PCi dt
þ þ
Money influx ρi PCi dt N dS
i dS0
ρi PCi dt N dS
i dS0 Pi ρi Ci dt ¼ ρi PCi dt
Circulation rates
Input circulation rate dS
i ρi ¼ ρi þ p i dS
i ρi ¼ ρi p i
Output circulation rate dSþ
i ζi ¼ ζi þ pþ
i dSþ
i ζ i ¼ ζi pþ
i
Influx circulation rate ρi dS þ dS þ
i dS0 i dS0
Outflux circulation rate ζi dSþ dSþ
i dS0 i dS0
Capacity growth dCi Zdt H dSþ
i H dS
i Zdt H dSþ
i H dS
i
6 The Interpretation of the Economic Variables
6.3 The Impacts of the Inflationary and Deflationary Modes of Evolution 155
þ
subset dS 0 \ dSi , bitpulses can only originate in exchange for virtual money. And
hence pi > 0 if pþ
þ þ þ
0 > 0. That is, pi is positive whenever p0 is positive.
Likewise if p0 < 0, bitpulses can only annihilate in dS0 \ dSþ
0 in exchange for
þ þ
“originating” virtual bits. Thus in any subset of dS0 \ dS0 , e.g., the subset dS
i \ dS0 ,
bitpulses can only annihilate in exchange for virtual money. And hence pi < 0 if
p
0 < 0. That is, pi is negative whenever p0 is negative.
þ þ
Similarly if p0 gets zero, no longer bitpulses originate in dS 0 \ dS0 and no
longer bitpulses will originate in dS0 \ dSi . Hence it must follow that pþ
þ
i ¼ 0 if
pþ0 ¼ 0. Also if p
0 gets zero, no longer bitpulses annihilate in dS
0 \ dS þ
0 and no
þ
longer bitpulses will annihilate in dSi \ dS0 . Hence it must follow that p
i ¼ 0 if
p0 ¼ 0.
This implies that the evolutionary process can only alternate from inflationary
mode to deflationary mode of evolution or conversely after the rates pþ þ
i , p0 , pi and
p0 vanish altogether simultaneously. It follows that the mode of evolution is
inflationary in the sector Si whenever it is inflationary in S0 and conversely. And
it follows that the mode of evolution is deflationary in the sector Si whenever it is
deflationary in S0 . □
I shall call the time instance for which pþ
i and pi change sign the turning point of
evolution. Inflationary mode and deflationary mode can only exist together at a
moment of time t for which one of them vanishes for a particular index i. In that case
all pþ þ
i and pi vanish together, thus causing pi ðtÞ ¼ pi ðtÞ ¼ 0 for all i at that
particular time t.
And in case either pþ
i or pi vanish it is necessary that H ðX0 \ Y0 Þ ¼ 0 as one may
easily check.
Well, the circumstances for which H ðX0 \ Y0 Þ ¼ 0 are very special, but the
consequences are far-reaching. To get an idea of what is at stake when H ðX0 \Y0 Þ¼0,
I refer to the detailed expositions in Appendices F and G. It is demonstrated in Appendix
F that the case with H ðX0 \Y0 Þ¼0 and that the case of statistical independence of input
events and output events go hand in hand. That is, if there is statistical independence
between input and output, it must follow that H ðX0 \Y0 Þ¼0 and, if H ðX0 \Y0 Þ¼0,
it must follow that there is statistical independence between input and output.
Moreover, if HðX0 \ Y0 Þ ¼ 0, it is shown in Appendix G that H ðX0 Þ ¼ HðY0 Þ ¼ 0
as well and that labor output probability μ (consumption propensity) and labor input
probability λ must satisfy λ ¼ μ ¼ 1 and finally that economic growth is then totally
absent. Therefore H ðX0 \ Y0 Þ ¼ 0 marks the begin and the end of evolution as well
as the begin and end of evolutionary selection. It marks both the point of evolution-
ary take off and the point of ultimate evolutionary failure and extinction. This is
why it is a turning point of evolution. Thus, as the constraint pþ
i ¼ pi ¼ 0 is reached,
the economy has reached that very end of change or that very begin of change. Here
þ
the differential entropy H ðX0 \ Y0 Þ of dS 0 \ dS0 vanishes entirely and also all the
differential entropy H Xi \ Yj of the subsets dSi \ dSþ
j and the differential entropy
þ
of dSi and of dSi vanishes entirely.
156 6 The Interpretation of the Economic Variables
All in all it is not very likely for a modern economy to reach the turning point,
where in fact the economy lands in the middle of nowhere, in effect not a situation
of temporary economic stagnation but more likely a situation of definitive stagna-
tion and downfall.
Of course an economy may sometimes experience that the current price-level P
stops rising and P may even temporarily decline. This is a situation of temporary
deflation. That may reflect a very serious situation of economic stagnation, but the
economy need not yet be in the deflationary mode of evolution, which is much more
demanding because to reach the latter we have to get in the turning point of
evolution. Temporary deflation ðp < 0Þ does not immediately bring about the
average historic price inflation rate pþ i to get zero. On the contrary, due to price
increases in the past this will not happen lightly.
What these observations do reveal is that a deflationary fall of the current unit
price P will have serious devastating effects on the growth potential of any economy,
because it will then approach the turning point closer and closer. It is dangerous to
have the economy kept in a state of declining unit price P or a state of a very low
inflation rate. The longer it stays in such state, the closer it reaches a constant average
þ
historic price level P0 and the more dangerous it is for the economy to decline
further.
This is the main reason why our modern growth economies need a few percents
of average price inflation each year in order to enhance the realization of long term
growth.
Noticing the small probability of survival of an economy in the turning point of
evolution or in the deflationary mode of evolution, it is very likely that all modern
growth economies are in the inflationary mode of evolution. The deflationary mode
of evolution appears to be a very rare exception, a point that is also a point of
evolutionary take off. To analyze that further is interesting but rather irrelevant for
the study of the performance of our modern economies.
I shall therefore proceed in the sequel with analyzing the evolution of the
economy as staying in the inflationary mode of evolution.
The definition (6.7) of pi in the inflationary mode of evolution and the definition
(6.16) of pþ
i in the deflationary mode of evolution, can be condensed in a single
equation:
λ i pþ
i ¼ μ i pi irrespective whether pþ
i 0 or pi 0
whereas (5.21), (5.22), (5.23) and (5.29) are valid in the inflationary mode of
evolution and alternatively (5.36), (5.37), (5.38) and (5.45) are valid in the defla-
tionary mode of evolution.
6.4 Some Further Relationships of Price and Circulation Rates 157
X
p
0 C0 ¼ p
i Ci (6.19)
i
X X
ζ 0 C0 ¼ ζ i Ci ; ρ0 C0 ¼ ρi C i
i i
X X
ζ 0 C0 ¼ ζ i Ci ; ρ0 C0 ¼ ρi C i
i i
and
ζ 0 þ ρ0 C0 ¼ Z H dSþ þ
0 dS0 þ Z H dS0 ¼ Z H dS0 [ dS0 ¼ Z H ðdS0 Þ
Hence
ζ 0 þ ρ0 C0 ¼ ðζ 0 þ ρ0 ÞC0 ¼ ZH ðX0 [ Y0 Þ ¼ ZH ðdS0 Þ
and
ζ 0 þ ρ0 ¼ ζ 0 þ ρ0
ρi Ci ¼ λi ρ0 C0 , ρi Ci ¼ λi ρ0 C0 , ζ i Ci ¼ μi ζ 0 C0 and ζ i Ci ¼ μi ζ 0 C0
p þ þ
i Ci ¼ λi p0 C0 and pi Ci ¼ μi p0 C0 (6.20)
It follows that
þ
p Ci ¼ γ μi C0 ¼ Z HðX0 \ Yi Þ
i
and
p Ci ¼ γ λi C0 ¼ Z H ðXi \ Y0 Þ
i
158 6 The Interpretation of the Economic Variables
γ ¼ pþ
0 ¼ p0 (6.21)
Hence the transmission rate γ is the absolute value equivalent of the overall
average rate of historic price increase in the inflationary mode of evolution and it is
the absolute value equivalent of the overall average rate of future price decrease in
the deflationary mode of evolution.
On behalf of (1.1), (2.11), (3.2) and (3.3) we can reformulate the expressions for
ρi , ζ i , ρi and ζ i as
P P
λi Z λi log λi μi Z μi log μi
i i
ρi ¼ and ζ i ¼ (6.22)
Ci Ci
Like the variables Ci , Ci , Z, H ðXi jY0 Þ, H ðXi Þ, H ðYi jX0 Þ, H ðYi Þ, λi , μi and qij , the
transmission rate γ and all the circulation rates ρi, ρi, ζ i and ζ i are dynamic functions
of time t.
The sales volume of goods and services produced in the economy is summarized in the
economic concept of spending or output. To keep the production of output going input
is demanded. To that end agents deliver effort which involves wear and tear of their
capacity: physical in the form of labor, and rather more materially but not too strictly
so in the form of wear and tear of mechanical production capacity and production
facilities. From an evolutionary point of view the term spending is less appropriate to
reflect the production of goods and services. Perhaps the term procreation is more
appropriate to denote the productive output resulting from effort.
Procreation is then the output of the economy, effort is the input of the economy.
We can conceive input as the delivery of effort and output as procreation.
Procreation results by effort. In turn procreation feeds the effort delivering agents
in order to sustain the exertion of effort. Procreation is not restricted to apply only
to concrete material entities, neither is effort. Procreation and effort emerge in
physical goods, services, goodwill, patents and more generally in agents as well as
in thoughts, ideas, etcetera in the sense Hodgson and Knudsen describe the
properties of replicators. Financial flow is “stocked” within the transmission area
þ
dS0 \ dS0 .
6.5 The Two-Sector Economy 159
Let us next turn to what concrete meaning the flow variables dealt with in
Sects. 6.1, 6.2, 6.3, and 6.4 have for the two-sector economy of Fig. 2.4 consisting
of a consumption sector S1 and an investment sector S2 . We will here adopt the
division of the sample space dS0 in subsets A, B, C, D, E, F, G and U as given in
Fig. 2.4.
þ
Using this scheme we get: dSþ þ
1 ¼ B [ D [ E, dS1 dS0 ¼ B and dS1 \ dS0 ¼D [ E.
Here we shall use the terms GNP and gross national income (GNI) to denote the
gross national product and the GNI over the time-interval ðt; t þ dtÞ of selection
rather than over a year or quarter. A similar remark applies to all the other flow
variables of the economy.
Since the economy is closed it is necessary that the valueζ 0 PC0 dt ¼ PZdt H dSþ
0
of national output (GNP) balances the value of GNI.
Clearly GNP is total spending, the sum of consumption and investment:
consumption ¼ labor output ¼ Y1 ¼ ζ 1 C1 dt ¼ Zdt H dSþ
1
¼ Zdt H ðB [ D [ EÞ
investment ¼ entrepreneurial output ¼ Y2 ¼ ζ 2 C2 dt ¼ Zdt H dSþ
2
¼ Zdt HðA [ C [ FÞ
Herein the first term PZdt H ðA [ BÞ PZdt H ðG [ UÞ on the right side
is equal to the surplus of PZdt H dSþ dS over PZdt H dS dSþ . On
0 0 0 0
behalf of (2.18) this surplus is equal to the net
growth
dC 0 of overall
capital
of S0 .
It is also equal to the surplus of PZdt H dSþ 0 over PZdt H dS 0 , which is
the net growth dC0 of overall capacity C0 in accordance with (2.4) and (2.19).
The second
term on þ
the right, PZdt H ðC [ DÞ, is entrepreneurial financial input
PZdt H dS 2 \ dS 0 .
Producer profit
þ
¼ PZdt H dSþ þ
0 dS0 PZdt H dS0 dS0 þ PZdt H dS2 \ dS0
or
Herein PZdt H ðA [ BÞ is the money cost of total output at average historic unit
þ
production cost P0 rather than at market unit price P per bit. Further, PZdt H ðC [ DÞ
and PZdt H ðE [ FÞ are the respective entrepreneurial financial input and labor
financial input. Alternatively, PZdt H ðD [ EÞ and PZdt HðC [ FÞ are the respec-
tive labor financial output and entrepreneurial financial output.
Mark that on the input side the following terminology for the two-sector
economy has been employed:
labor input ¼ ρ1 C1 dt ¼ Zdt H dS
1 ¼ Zdt H ðE [ F [ GÞ
entrepreneurial input ¼ ρ2 C2 dt ¼ Zdt H dS
2 ¼ Zdt H ðD [ C [ UÞ
þ
labor influx ðwagesÞ ¼ ρ1 C1 dt ¼ Zdt H dS
1 dS0 ¼ Zdt HðGÞ
þ
entrepreneurial influx ðdepreciationÞ ¼ ρ2 C2 dt ¼ Zdt H dS
2 dS0 ¼ Zdt HðUÞ
6.6 The Growth of Capacity 161
þ
Labor input Zdt H dS sum of wages (labor influx) Zdt H dS dS
1 is here the
1 0
and labor financial input PZdt H dS1 \ dSþ 0 . The idea behind this choice of
terminology is that labor realizes its financial input from the return on savings that
ultimately proceed from past wages. Thus labor input is more than wages alone, just
like entrepreneurial input is more than entrepreneurial depreciation alone.
A summary of the inflow and outflow terminology for the two-sector economy
has been given in Table 6.2.
Since dCi ¼ Zdt ½H ðYi Þ H ðXi Þ ¼ ðζ i ρi ÞCi dt, let us first observe that
c i ¼ ζ i ρi (6.25)
ρ i H ð Yi Þ ¼ ζ i H ð X i Þ (6.26)
After substitution of these expressions for H ðYi Þ and H ðXi Þ in (6.26) we arrive at
X X
μ i ρi μj log μj ¼ λi ζ i λj log λj
j j
162 6 The Interpretation of the Economic Variables
and
X
μj log μj ¼ μ log μ ð1 μÞ logð1 μÞ ¼ QðμÞ
j
The function QðωÞ has been plotted in Fig. 6.1. QðωÞ has an absolute maximum 1
for ω ¼ 0:5 and two absolute minima 0 for ω ¼ 0 and 1.
6.6 The Growth of Capacity 163
0.9
0.8 Q( μ)
Q(ω) 0.7
Q( λ)
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ω
μ λ
Fig. 6.1 The two-sector economy. The function Q(ω) as a function of ω. The absolute maximum is
Q(ω) ¼ 1 for ω ¼ 0.5. Furthermore Q(ω) ¼ 0 for ω ¼ 0 and ω ¼ 1. Z · [Q(μ) Q(λ)] ¼ dC0/dt,
the volume of capacity growth. This is positive for λ > μ
1β δ2 1 δ3 1 2β
QðλÞ QðμÞ δ log þ þ 2
β 2 βð1 βÞ ln 2 6 β ð1 βÞ2 ln 2
with β ¼ ðμ þ λÞ=2. This third order expansion in δ offers very good approximations
for μ and λ in a 15 % vicinity around β between 0.65 and 0.80.
On behalf of (6.30) it can be shown that the common rate c0 ¼ c0 of capacity
growth and capital growth of the economy is mainly determined by the difference
ðλ μÞ. For λ > μ the rate of economic capacity growth is positive. For λ < μ it is
negative. The dance of that difference about its long run average in the course of
time explains the instantaneous fluctuations of the growth rate c0 of overall capacity.
In accordance with expressions (6.28) and (6.30) c0 is also equal to the growth rate
c0 of overall capital.
164 6 The Interpretation of the Economic Variables
μi λi þ
c i ¼ ζ i ρi ¼ ζ i ρi þ pi , pþ
i 0
μi
In the deflationary mode of evolution ci follows from (6.11), (6.16), (6.17) and
(6.25):
λi μi
c i ¼ ζ i ρi ¼ ζ i ρi þ pi , p
i 0
λi
The growth rate of output Yi ¼ Zdt H ðYi Þ is equal to the growth rate of Yi ¼
ζ i Ci dt:
Likewise the growth rate of input Xi ¼ Zdt HðXi Þ is equal to the growth rate of
Xi ¼ ρi Ci dt:
dY i =dt dζ i =dt
yi ¼ ¼ þ ci
Yi ζi
Likewise, on behalf of (6.10), the net growth rate of influx Xi ¼ Zdt H ðXi jY0 Þ ¼
ρi Ci dt is
The surplus of labor output over labor input is equal to the net growth dC1 of labor
capacity C1 :
We are interested in the surplus dC1 of labor output (consumption) over labor
input (wages + financial labor influx), how it depends on ðλ μÞ. The questions are:
Does the net growth rate c1 of labor capacity display a more or less similar behavior
like c0 as a function of ðλ μÞ? Is it necessary that labor input X1 (the income side of
the consumption sector) exceeds labor output Y1 (the spending side of the consump-
tion sector) to sustain a positive overall rate c0 of growth, i.e., if ðλ μÞ > 0 ?
Another question is: what happens with an eventual surplus of X1 over Y1 if c0 is
negative, i.e., if ðλ μÞ < 0?
To analyze this, the behavior of ωQðωÞ as a function of ω is decisive. The course
of ωQðωÞ as a function of ω has been plotted in Fig. 6.2.
Notice ωQðωÞ has a single maximum for ω ¼ 0:703506. The maximum value of
ωQðωÞ here attained is 0.616949. We shall discern between six different situations:
1. For μ < λ < 0:703506 we have that μQðμÞ < λQðλÞ: c0 > 0, consumption <
labor input and c1 < 0 moving in a direction opposite to c0 . Case of positive
overall economic growth, but negative growth in the consumption sector S1 .
2. For λ < μ < 0:703506 we have that μQðμÞ > λQðλÞ: c0 < 0, consumption >
labor input and c1 > 0 moving in a direction opposite to c0 . Case of negative
overall economic decline, but positive growth in the consumption sector S1 .
3. For 0:703506 < μ < λ we have that μQðμÞ > λQðλÞ: c0 > 0, consumption >
labor input and c1 > 0 more or less behaving like c0 . Case of positive overall
growth.
166 6 The Interpretation of the Economic Variables
0.9
0.8
ω Q(ω ) 0.7
0.6 0.616949
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ω
0.703506
Fig. 6.2 The two-sector economy. The function ωQ(ω) as a function of ω. The absolute maximum
ωQ(ω) ¼ 0.616949 is attained for ω ¼ 0.703506
4. For 0:703506 < λ < μ we have that μQðμÞ < λQðλÞ: c0 < 0, consumption <
labor input and c1 < 0 more or less behaving like c0 . Case of negative overall
decline.
5. For μ < 0:703506 < λ it is uncertain whether consumption is greater or smaller
than labor input: c0 > 0 but c1 will often tend to be rather small, either positive or
negative.
6. For λ < 0:703506 < μ it is uncertain whether consumption is greater or smaller
than labor input: c0 < 0 but c1 will often tend to be rather small, either positive or
negative.
The economic growth in the investment sector of the two-sector economy is
determined by
1. For μ < λ always ð1 μÞQðμÞ > ð1 λÞQðλÞ: c0 > 0, investment > entrepre-
neurial input and c1 > 0 more or less behaving like c0 . Case of positive overall
growth.
2. For μ > λ always ð1 μÞQðμÞ < ð1 λÞQðλÞ: c0 < 0, investment < entrepre-
neurial input and c1 < 0 more or less behaving like c0 . Case of negative overall
decline.
Thus we conclude that a substantial increase of investment is the best thing to do
in order to resume overall economic growth.
A very well developed system of money exchange distincts our modern growth
economies from evolutionary systems with much slower rates of entropy growth. It
has been shown in Appendices F and G that this must be attended with sufficient
statistical dependence between the output events and the input events. This in turn
demands that a moderate level of price inflation is sustained. In this manner
transmission Zdt H ðX0 \ Y0 Þ plays a substantial role and is of the same order as
output Zdt H ðY0 Þ and input Zdt H ðX0 Þ . The result is the attainment of the
considerable growth rates that our modern money economies characterize. On the
other hand an economy in a state with very low inflation rate is likely to demon-
strate poor positive or even negative growth. It will be in a state of economic
stagnation and decline.
Clearly then, money with a little but sufficient inflationary impetus is the
lubricant of economic performance. This raises the point how the primitive barter
economy has managed to realize economic growth. The conclusion we must draw
here is that economies of a (very) primitive state of development cannot have
survived without a sort of unit of exchange that slowly devalued in the course of
time. Well, the use of a unit of exchange (shells, beads, corn) in primitive
economies has a very old record. Due to the scarce availability of appropriate
units of exchange and the lack of ease of transportation and the limited acceptance
of their units of exchange, these barter economies achieved only small or very small
growth rates, at least much less than the rates of growth the modern economies have
realized and still manage to attain.
It must further be stressed that the prerequisite for any economy to possess a
moderately devaluating unit of exchange does not imply that always the same
unique entity should serve as the unit of exchange. The moment another suitable
unit of exchange appeared on the scene, that unit could take over the role of the
former unit of exchange. Moreover, more units of exchange have been used side by
side or alternately in one and the same economy. Nevertheless there were and there
are still important restrictions for the acceptance of a proper unit of exchange. That
unit ought to be sufficiently scarce and it must remain so in order to prevent the
supply of the exchange unit to rise abundantly beyond control. On the other hand it
168 6 The Interpretation of the Economic Variables
1
I use the term ecotope on the analogy of the term biotope in biologic evolution.
6.8 Economic Evolution and Biologic Evolution 169
Table 6.3 The core variables and core dimensional units of evolution
Equivalent evolutionary terminology
Core concepts of evolution Biologic evolution Economic evolution
Genotype Quantities of exchange Genes Entities
(article of of variable content each holding a each holding a variable
exchange) variable number number PHðSÞ of
of quanta money-units
Unit of exchange Quantum Money unit
Phenotype Quantities of variable Species/Biotope Subjects/Ecotope
(article of content and different each stocking a each stocking a variable
selection) state variable number number HðSÞ of
of bitpulses bitpulses
Unit of selection Suggested: Bitpulse
Bitpulse holding one bit of entropy
holding one bit of
entropy
Unit price of unit of P quantum units (P is P money units (P is a
selection a variable of time) variable of time)
Abstract The transmission Zdt H ðX0 \ Y0 Þ represents all the money-like entities
(expressed in bits) issued in the economy on the time-interval (t,t þ dt). This
justifies articulate mathematical definitions of money flow and money stock for
the economy as a whole and for its separate sectors.
Another line of investigation considered in this chapter is the role of time. We
can calculate outflow and inflow handling rates, the time duration of selective
attention involved per outflow bit respectively per inflow bit on the selection
interval (t,t þ dt). These handling rates assist to calculate the relative time gain
of input handling over output handling per unit of time per bit.
The product of this relative time gain and the circulation rate is a net rate of
capacity growth. Thus the relative time gain can be conceived as a total of work-
units per unit of time passage that goes into the growth of capacity during the
selection interval (t,t þ dt). This suggests a direct and unique relationship between
the stock Li ðtÞ of work-units of an economic sector Si that has accumulated in the
past until present time t and the capacity Ci ðtÞ, the entropy that has accumulated in
the past until present time t: Ci ðtÞ ¼ EðtÞ Li ðtÞ. Herein EðtÞ is the effectiveness of
production at present time t.
The basal condition for statistical dependence is the exchange of goods and services
for money. Money is the source of exchange required for facilitating selection. We
may consider money as the lubricant of economic selection. Without money, the
1
I use the term “earned” here just for comfort of description. Actually it is a net money amount
collected on input and a net money amount collected on output, which will be shown to balance in
the end.
7.1 The Role of Money and Transmission in Production 173
ζ i þ þ
ζ i P Pþ
i Ci dt ¼ p P Ci dt ¼ pþ
i PCi dt
ζi i i
Hence, the net money earned and received by producers on the selection interval
ðt; t þ dtÞ can also be restated as:
pþ þ
0 PC0 dt ¼ p þ
1 PC1 dt þ p 2 PC2 dt
From this input angle of incidence, the net money received by producers on the
selection interval ðt; t þ dtÞ is:
p
0 PC0 dt ¼ p
1 PC1 dt þ p 2 PC2 dt
of which the financial influx p 2 PC2 dt is the money received by entrepreneurs/
producers and of which the financial influx p 1 PC1 dt is the money received by
consumers for the disposal of financial resources for the benefit of production.
In conclusion, the total money received by producers on the selection interval
ðt; t þ dtÞ is pþ
0 PC0 dt from the angle of incidence of output. And it is p 0 PC0 dt from
the alternative
þ angle of incidence of input. These flows are in equilibrium because
p ¼ p ¼ γ in accordance with (6.21).
0 0
We are thus led to the conclusion that p 0 PC0 dt ¼ Zdt H ðX0 \ Y0 Þ ¼ γPC0 dt is
all the money issued by the banking system during the time-interval ðt; t þ dtÞ of
selection.
Traditional accounts of economic theory define the money stock of agents as the
resources that can be liberated for exchange immediately or within a very short time
period. This depends on the quantity of money volume issued by the banks against
long term loans. The terms under which the loans are to be redeemed, the depen-
dence of these terms on economic performance and also the availability of other
forms of liquidity than issued by the central bank have an impact on the available
means of liquidity in the near future. To account for that, traditional economic
theory employs various definitions of the money stock. However, it appears that the
transmission Zdt H ðX0 \ Y0 Þ provides for a clear theoretical definition of
the entropy increase dM0 of the money stock M0 (measured in bits of entropy) on
ðt; t þ dtÞ. In the dimension of money units this increase is PdM0 :
Let us next consider how the total money flow dM0 is divided between the sectors S1
and S2 of the two-sector economy. In this regard the Venn diagram of Fig. 2.4 will
again be very helpful.
Let dMi denote the increase of the money stock Mi of the sector Si during
ðt; t þ dtÞ. Thus
7.2 The Liquidity Flux and Money Flows and Stocks 175
X X
dM0 ¼ dMi and M0 ¼ Mi
i i
More specifically for the two-sector economy, dM1 is the increase of the money
stock M1 of the consumption sector S1 during ðt; t þ dtÞ and dM2 is the increase of the
money stock M2 of the investment sector S2 during ðt; t þ dtÞ. Of course,
Mark that it costs the money wage ρ1 PC1 dt to produce consumer and investment
goods before bringing this production to the market for the money value ρ1 PC1 dt
during the selection interval ðt; t þ dtÞ. The net money realized on the utilization
of labor effort by producers for economic production is then ρ1 PC1 dt ρ1 PC1 dt ¼
p
1 PC1 dt ¼ PZdt H ðFÞ (See the Venn diagram of Fig. 2.4). As we have set out
before, this can be regarded to represent the financial input generated by the
consumption sector only. On the other hand the surplus of money sales ζ 1 PC1 dt
of consumer good production over the money cost ζ 1 Pþ 1 C1 dt thereof is
ζ 1 PC1 dt ζ 1 Pþ 1 C1 dt ¼
p þ
1 PC 1 dt ¼ PZdt H ð D Þ. As we have also explained before,
this surplus is the financial output realized on consumer good production. Clearly
the financial input p 1 PC1 dt ¼ PZdt H ðFÞ realized on the utilization of labor effort
does not match the net financial output pþ 1 PC1 dt ¼ PZdt H ðDÞ realized on
consumer good production. This is comprehensible because labor effort serves
also to bring forth investment good production. Furthermore consumer good
production will also require the cost of wear and tear of capital equipment. Neither
does the cost of labor effort, required for investment good production, balance the
depreciation cost caused by consumer good production only.
In conclusion it must follow that generally p 1 PC1 dt p þ
1 PC1 dt 6¼ 0.
p
1 PC 1 dt is the amount by which the money in stock of the consumption sector S1
increments during ðt; t þ dtÞ after all input costing obligations of that sector have
been met at time t þ dt minus the amount pþ 1 PC0 dt by which the money in stock of
the consumption sector p decrements during ðt; t þ dtÞ after all output spending
obligations of that sector have been settled at time t þ dt . It follows that the
difference p 1 PC1 dt p þ
1 PC1 dt is the remaining net flow of money that enters the
consumption sector S1 during ðt; t þ dtÞ.
Likewise p 2 PC2 dt pþ2 PC2 dt is the remaining net flow of money that enters the
investment sector S2 during ðt; t þ dtÞ. However, recall that
p þ
1 PC1 dt p
1 PC1 dt þ p þ
2 PC2 dt p
2 PC2 dt ¼ p þ
0 PC0 dt p 0 PC0 dt ¼ 0
Hence the surplus of money flow of the consumption sector is the deficit of
money flow of the investment sector or conversely. We have called this surplus of
money flow the liquidity flux of the consumption sector S1 and refer to its entropy
level by the symbolic differential notation dB1 (See Sect. 2.9). That is, on behalf of
(6.4) and (6.19),
176 7 Money and Liquidity, Time, Work and Effectiveness
dB1 ¼ p þ
1 C1 dt p 1 C1 dt ¼ Zdt ½H ðFÞ H ðDÞ
¼ Zdt ½H ðX1 \ Y2 Þ H ðX2 \ Y1 Þ
and
dB1 ¼ p þ
1 C1 dt p 1 C1 dt ¼ ðq10 q01 ÞZdt H ðX0 \ Y0 Þ ¼ γ ðλ μÞC0 dt (7.3)
dM0 dB1
dM1 ¼
2
and
1þλμ 1λþμ
dM1 ¼ Zdt H ðX0 \ Y0 Þ and dM2 ¼ Zdt HðX0 \ Y0 Þ
2 2
1þλμ 1λþμ
dM1 ¼ γC0 dt and dM2 ¼ γC0 dt (7.5)
2 2
Mark that dB1 and dMi are entropy contents. They are expressed in bits. PdB1 and
PdMi are the corresponding money flows in money units.
For the money economy γ > 0. Thus we see, on behalf of (7.3), that for the
money economy the liquidity flux dB1 is only positive if labor input probability λ
exceeds consumption probability μ. Recall that under this condition also positive
entropy growth of the money economy will sustain [see (6.30)]. Clearly, in the very
first incipient phase of evolution, positive entropy growth is necessary for evolution
to take off. Thus μ should tend to decline below λ at incipience time of economic
7.2 The Liquidity Flux and Money Flows and Stocks 177
evolution. It follows from (7.3) that the liquidity flux of the consumption sector will
be positive when economic evolution takes a path of positive entropy growth and
also that dM1 and dM2 will remain positive as long as economic evolution has
taken off and the money economy endures to exist. On the other hand the liquidity
flux of the investment sector remains negative because it is opposite to the money
flow of dB1 .
For any particular sector Si of the multi-sector economy the liquidity flux dBi in
bits of entropy is defined by:
Ðt
We should be aware that the integral 1 PðξÞ dBi ðξÞ is the accumulated
money value of dBi over time. The money value of dBi at current time t is
ðt ðt ðt
PðξÞdBi ðξÞ ¼ p
i ðξÞPðξÞCi ðξÞdξ pþ
i ðξÞPðξÞCi ðτÞdξ
ξ¼1 ξ¼1 ξ¼1
or
ðt ðt
p
½i ðξÞ pþ
i ðξÞ PðξÞCi ðξÞdξ ¼ ½λi ðξÞ μi ðξÞ γ ðξÞ PðξÞC0 ðξÞdξ
ξ¼1 ξ¼1
Direct integration of the liquidity entropy flux dBi over the time-domain yields
the entropy stock Bi of the liquidity surplus in bits, which—as we have seen—may
be positive as well as negative.
The surplus dBi ¼ p i p þ
i Ci dt will only differ from 0 (for i 6¼ 0) in case
transactions are being settled in money units on the time interval ðt; t þ dtÞ of
selection. The following properties must always be awarded:
X X
dB0 ¼ dBi ¼ 0; B0 ¼ Bi ¼ 0
i i
178 7 Money and Liquidity, Time, Work and Effectiveness
Ðt
Furthermore the integral 1 PðξÞ dMi ðξÞ is the accumulated money value of
dMi over time. This differs from Mi . The money value of dM0 at current time t is
ðt ðt ðt
PðξÞdM0 ðξÞ ¼ p
0 ðξÞPðξÞCi ðξÞdξ þ pþ
0 ðξÞPðξÞCi ðτÞdξ
ξ¼1 ξ¼1 ξ¼1
or
ðt ðt
PðξÞdM0 ðξÞ ¼ 2γ ðξÞPðξÞC0 ðξÞdξ
ξ¼1 ξ¼1
dCi is the real growth of the economy. On behalf of (2.4) and (2.18) we have
It follows that
Further
1 1 1
σ
i ¼ ¼ ¼ P
Z H ðXi Þ ρi Ci λi Z λj log λj
j
1 1 1
and σ þ
i ¼ ¼ ¼ P
Z H ðYi Þ ζ i Ci μi Z μj log μj
j
so that
X X
σ þ
i ρi ¼ σ i ζ i and σ i λi λj log λj ¼ σ þ
i μi μj log μj (7.8)
j j
1 1
σ
i ¼ and σþ
i ¼
ρi Ci ζ i Ci
It follows that
σ ρi ¼ σþ
i i ζi (7.9)
For the two sector economy equation (7.8) assumes the form
λi QðλÞ σ þ
i ¼ μi QðμÞ σ i for N ¼ 2
Mark that the handling rate is the time required per unit of time to select a bit.
However, the total time to be distributed over inflow selection and outflow selection
is the same time length dt. Inflow selection and outflow selection occur simulta-
neously on ðt; t þ dtÞ. In fact it is all part of the overall selection experiment. Hence
when the selection of an input-bit is settled, the selection of an output-bit is
þ
concluded simultaneously. What then about the difference σ i σ i in handling
rates per bit between input-bits and output-bits. As the bit is our reference unit of
þ
real value, it is clear that there is a surplus σ i σ i of input-time per unit of time
over output-time per unit of time required to select a bit in the process of selection.
The relative time-gain realized per unit of time per bit during ðt; t þ dtÞ is
þ
σ
i σi
σ
i
It follows that
σi σþ
i σi σþ
i
ci ¼ ζ i ¼ ρi (7.10)
σ
i σþ
i
þ
þ
The expression on the right expresses time-gain σ i σi σ i relative to
output-bit handling time.
We see that the rate ci of capacity
growth isthe product of the output circulation
þ
rate ζ i and the relative time-gain σ
i σ i σ i per bit. Equation (7.10) can also be
stated as
σi σþ
i σi σþ
i
dCi ¼ ζ C
i i dt ¼ ρi Ci dt
σ
i σþ
i
or as
σi σþ
i σi σþ
i
dCi ¼ Yi ¼ Xi
σ
i σþ
i
Thus we conclude that the realized relative time-gain of each bit of output,
originating during ðt; t þ dtÞ, goes into the net growth dCi of capacity Ci during
ðt; t þ dtÞ. Here the unit of time acquires another meaning: the time of attention, the
7.4 The Handling Rate of Inflow and Outflow Selection and the Net Growth of. . . 181
work-time, spent to bring forth a bit of entropy per unit of time. In line with this we
must interpret capacity Ci as the stock in which all the surplus of work-time of the
sector Si has accumulated.
The explanation of the net growth rate ci of capacity Ci in (7.10) is concerned
with the net surplus of input-bit handling time over output-bit handling time
required per unit of time. We can explain the net growth d Ci of capital in quite a
similar manner, as a sort of net growth rate dCi =Ci based on expression (7.9) of
influx-bit handling rate and outflux-bit handling rate. Then the proper way of
explaining is to identify d Ci =Ci with
þ
dCi σi σþ
i ζ i dt ¼ σi σi ρi dt
¼ (7.11)
Ci σ
i σþ
i
þ
Herein σi σþ
i σi and σ
i σþi σi are the relative time-gains, respec-
tively measured relative to influx-bit handling time and relative to outflux-bit
handling time.
Equation (7.11) results in
þ
σi σþ ζ i Ci dt ¼ σi σi ρi Ci dt
dCi ¼ i
σ
i σþ
i
or as
σ
i σ þ σi σþ
dCi ¼ i
Yi ¼ i
Xi
σ
i σþ
i
Thus we conclude that the realized relative time-gain of each bit of outflux,
originating during ðt; t þ dtÞ, goes into the net growth dCi of the capital of Si during
ðt; t þ dtÞ.
I shall adhere to the following notation when denoting the various relative time-
gains.
þ þ
σ
i σi σ
i σi σ
i σ þ σ
i σ þ
η
i ¼ ; ηþ
i ¼ þ ; η
i ¼
i
; η
i ¼ þ
i
σi σi σi σi
þ
I shall call η ηi and η
i , ηi , i rates of relative time-saving. Clearly,
ci ¼ η þ η ηþ
i ζ i ¼ ηi ρi and ðd Ci =dtÞ=Ci ¼ i ζi ¼ i
ρi (7.12)
and
dCi ¼ η þ η ηþ
i Yi ¼ ηi Xi and d Ci ¼ i Yi ¼ i Xi (7.13)
182 7 Money and Liquidity, Time, Work and Effectiveness
þ
The pair of η η
i and ηi is very much related, like the pair of η
i and i :
η þ þ
i ηi ¼ ηi ηi and η ηþ
i η
i ¼ ηþ
i i
In this section I shall be concerned with the definitions of variables such as work
and effectiveness that play a role in circumscribing the role of labor and productiv-
ity of the production process. In economics employment and how employment can
be influenced by economic policy is a major issue. To deal with that issue I shall
initially define the concept of labor force in the following way:
L1 ðtÞ ¼ Labor force (also called labor), the number of employees actively
involved with delivering paid occupational labor at time t.
There are some doubts about the suitability of this definition. E.g. the definition
does not seem to weigh the role of part-time work properly. We shall not immedi-
ately go into that matter here, but later on we shall consider the issue from a more
general perspective.
The total time-effort delivered by the labor force during the production interval
(t,t þ dt) is of course proportional to the length dt of that interval. We shall call that
time-effort work or work-time:
L1 ðtÞdt ¼ Work.2
The labor force is part of the consumption sector, the conglomerate of agents of
S1 delivering work and building up reserves to maintain doing work. Thus so far the
above definitions apply to the sector S1 only. However for purposes of generaliza-
tion we will just as well consider cases with the index 1 of the consumption sector
replaced by the general index i denoting an arbitrary sector Si of the economy. That
is, Li ðtÞdt is defined as the total of work delivered by the sector Si during
the production interval (t,t þ dt). This implies there are Li ðtÞ work-producing
units in Si each delivering dt units of time-effort during (t,t þ dt). We wish to
stress here that we are primarily concerned with definitions here. The physical
nature of these work-delivering units is of no concern here at the moment. We will
discuss that subject later.
The quantity of work delivered is one aspect of economic activity; the other
aspect is the effectiveness, the productivity by which work is delivered. Thus we
must also provide for a definition of the concept of productivity.
2
I realize that the term work may call forth many different associations. Nevertheless I think it is
the best term to refer to the strict economic mathematical meaning that I have in view. It is not the
first time that science claims frequently used words of ordinary language to denote mathematical
strictly defined concepts. In physics the claims are on: energy, matter, force, power, charge, mass
and even work to mention only a few. Mathematically clear developed science needs such
restrictively defined concepts.
7.5 The Definition of Work and Effectiveness 183
ζ_ 0
Productivity growthrate ¼ þ c0 l1
ζ0
I think it is a minor flaw that this formula for the productivity growth rate contains
an additional term ζ_ 0 =ζ 0 , which is the rate of change of the output circulation rate.
We might of course reason that it should be that way. The larger the circulation rate,
the faster the economic cycle moves around and the more output we get per unit of
time so that ζ_ 0 =ζ 0 should be one of the components of the productivity growth rate.
However there are very good reasons to refute that and to pursue a more systematic
line of argumentation for the definition of productivity. We shall present that
argument here below. I think it wise to avoid confusion with the classical meaning
of productivity and that is why I shall instead use the term effectiveness with the
similar intention to reflect the efficiency of production.
Let then Ei denote the effectiveness of production of the sector Si. Let us further
þ
recall the expressions (7.12) that relate the rates of relative time-saving η i , ηi ,
ηi
and ηi and the circulation rates to the rates ci and ðdCi =dtÞ=Ci of economic growth:
þ
ci ¼ η þ η ηþ
i ζ i ¼ ηi ρi and ðd Ci =dtÞ=Ci ¼ i ζi ¼ i
ρi ð7:12Þ
In Sect. 7.4 it has been explained that the rates of relative time-saving go into the
growth of capacity Ci (respectively the growth of capital of Si). E.g. ηþ
i is the relative
decrease of handling time to effectuate dCi relative to input-bit handling rate σ i .
The understanding that work-time is the determinant realizing the growth of
entropy on the production interval ðt; t þ dtÞ immediately suggests the notion that
the work-time stored in the entropy stock of Ci consists of Ci bits of entropy, that
each carries some work-units of time. However the number of these work-units
stored per bit differs dependent on the moment of time they have been added to Ci.
The entropy content per work-unit of the total stock Ci is not the equivalent of the
entropy content per work-unit of the currently [on ðt; t þ dtÞ] created “stock” dCi of
work-units. The reason is that the manner entropy is produced currently is more (or
less) effective than it was done yesterday. Hence, by definition,
Ci ¼ Ei Li (7.14)
184 7 Money and Liquidity, Time, Work and Effectiveness
ci ¼ ei þ li (7.15)
Thus we see that economic growth can be made to increase in two alternative
ways:
1. By increasing effort through the rise of the work force Li effectuating a growth
rate li .
2. By increasing the efficiency of production through the rise of effectiveness Ei
effectuating a growth rate ei .
Note the differences between the general definition (7.14) of effectiveness and
the classical definition of productivity given by E1 ¼ ζ 0 C0 =L1. The factor ζ 0 has yet
to be omitted from the classical definition and C0 has to be replaced by C1 .
Well let us now revert to the question what actually Li stands for. The answer
begins with the critical aspects of the initially chosen definition L1 of the labor force:
the number of employees actively involved with delivering paid occupational labor
at time t.
The argumentation given in Sect. 7.4 clarifies that work-time per bit is the
equivalent of the time it takes per bit to effectuate the entropy growth as a result
of the overall selection process. This is the time during which the agents
involved attend the selection process. It appears then that the best measure of
work-time or work L1 dt is the number of hours actively worked by the labor
force during ðt; t þ dtÞ. Thus if we know the average number of hours employees
work per week, we must replace the calculation of work by the product of L1 dt and
the average number of hours per week employees are on duty, divided by the
constant number of 168 h that go in a week. Well instead of correcting the
calculation of work in that manner, I prefer to correct the calculation of L1 by
replacing it by the product of the labor force and the average numbers of hours per
week employees are on duty, divided by the constant number of 168 h that go in a
week. Thus we have a new definition of labor L1 rather than another definition of
work L1 dt.
We must next deal with providing a more concrete notion of the Li work-
delivering units of Si for i 6¼ 1, i.e. for effort delivered by sectors other than the
consumption sector S1 . The same question, restricted to the investment sector with
i ¼ 2 , is also an issue of classical economic theory. The neoclassical claim on
capital represents in fact the physical expression of all the production units of the
investment sector. For the investment sector S2 of the two-sector economy we might
identify L2 with an active stock of entrepreneurial production units capable of
delivering L2 dt time like units of effort such that C2 ¼ E2 L2 : Herein C2 is
entrepreneurial capacity and E2 is entrepreneurial effectiveness in accordance
7.5 The Definition of Work and Effectiveness 185
with (7.14). Well, despite the actual economic existence of factories, workshops,
retail shops, office buildings, assurance agencies, car dealers, building contractors,
showrooms, etcetera, etcetera, it is quite impossible to delimit the physical contours
of a uniform unit of entropy production within the investment sector. To approach
the problem from this side leads nowhere. The solution must be sought in what the
essence of entropy is. Before following that line of explanation, let me first give a
more classical answer.
Note that the sector S1 delivers L1 dt work-units during ðt; t þ dtÞ representing a
value of PL1 E1 dt money units. Similarly the sector S2 delivers L2 dt work-units
during ðt; t þ dtÞ representing a value of PL2 E2 dt money units.
However, if S2 would instead deliver the number L1 dt of work-units delivered
by S1 during ðt; t þ dtÞ, it would represent a value of PL1 E2 dt money units. If
PL1 E1 dt is smaller than PL1 E2 dt, there will be a natural tendency to replace labor
work-units immediately by entrepreneurial work-units. Or to put it in more
classical terms: then there will be a permanent natural tendency to substitute labor
for more advanced production technology. This implies that any disequilibrium be-
tween PL1 E2 dt and PL1 E1 dt cannot subsist. It is necessary that PL1 E2 dt ¼ PL1 E1 dt
and hence that E1 ¼ E2 or more generally that effectiveness cannot depend on
the state i:
Ei ¼ E (7.16)
equal as well. Efficiency is a stochastic function of time and keeps changing all the
time, just like Ci and Li .
Let us finally summarize the general definitions associated with work and
effectiveness of a sector Si :
Work-unit ¼ The effort delivered by an actively engaged unit of the work-force
per unit of time.
Li ¼ Work Force ¼ The number of work-units actively delivering work at time t.
Li dt ¼ Work, the work-time delivered by the work-force during ðt; t þ dtÞ.
E ¼ Ci =Li ¼ Effectiveness ¼ The productivity of delivering work.
The labor force L1 of an economy can be measured independently. Thus, if we
get an estimate of c1 from the study and analysis of the other variables of
the economy, we can calculate the rate of growth e of effectiveness. Then further,
if we get likewise knowledge of the various ci other than c1 from the study
and analysis of the other variables of the economy, we can calculate all the li
including l2 without bothering about the physical aspects of these entropy
producing units of sectors other than the consumption sector.
Additional to the variables of labor force L1 and effectiveness E, some other
variables play a role in the economics of employment and labor:
W1 ¼ Money Wage per work-unit, the average wage earned per work-unit of the
work force per unit of time elapsing.
_
L1 dt ¼ L1 W1 dt ¼ Money Wages, the total of money wages earned by the work
force during the production interval ðt; t þ dtÞ.
On the other hand money wages are also equal to PX1 ¼ PZdt H ðX1 jY0 Þ. Hence,
PX1 ¼
ρ1 C1 Pdt ¼ L1 W1 dt and
ρ1 C 1 P ¼ L 1 W 1
It follows that
_ ρ_ 1
_
p þ x1 ¼ l1 þ w1 ¼ l 1 and þ c1 þ p ¼ l1 þ w1 ¼ l 1 (7.17)
1
ρ
The first equation of (7.17) delivers a handsome way to calculate the rate of
_
growth x1 of labor influx (real wages) from the knowledge of the rate of growth l 1
of money wages and the rate of inflation p.
w1 p ¼ x1 l1 ¼ ρ_ 1 =
ρ1 þ c1 l1 is the rise of the wage per work-unit over
the rate of price inflation. In the long run the rate ρ_ 1 =ρ1 averages out.
From (7.15), (7.16) and (7.17) it follows that
ρ_ 1
w1 ¼ þeþp (7.18)
1
ρ
Chapter 8
Calculation
The question that concerns us here is the measurement of the unit prices and their
rates of change. Thus far we have become acquainted with the unit price P of a bit
of entropy and its relative rate of change pðtÞ ¼ ðdP=dtÞ=PðtÞ per unit of time,
which is also called a rate of (price) inflation.
As there is equivalence of both sides of the equation all the time, we might call P0
the average unit price of all the entities stored in warehouse. However, there
is ambiguity in fixing K0 and P0 separately if only the product K0 P0 is known.
E.g. K00 ¼ 10 K0 with P00 ¼ P0 =10 will do equally well. Well, this is solvable as the
choice of a unit of dimension is free. Thus we can fix P0 at a constant at a particular
time. However there is still another problem with the application of (8.1), because it
does not work for non-differential sectors Sj. Presuming that Sj is non-differential, the
difficulty is that the economic entities of category j in stock of Sj at current time t have
accumulated in the past for varying unit prices dependent on the variable times for
which they have been acquired. The reason is that Pj varies in the course of time.
Recall that almost all of the stock of non-differential sets remains unselected during
ðt; t þ dtÞ. Price formation at time t will only be affected at the moments agents select,
i.e. within the differential sets of entropy creation and entropy annihilation rather than
within non-differential sets. Hence (8.1) must be rejected if the subsets Sj are non-
differential.
We can still apply (8.1) for price index calculation provided the subsets Sj are
differential. Recall that the entropy in the economy is accumulating from the
surplus of entropy output over entropy input. PðtÞ is the unit price per bit of this
surplus and we can use that property to calculate the average unit price of output
and input from the unit prices of the constituents of the output flow and of the input
flow during a very small time-interval ðt; t þ dtÞ on which time comes to a standstill.
P is an average price that is constant to first order degree of accuracy in dt during the
time-interval ðt; t þ dtÞ of selection. Hence let us then calculate P as an average of
8.1 Divisia’s Index Formulas for Output Calculation 189
the unit prices of all equally priced constituent portions of economic entities on
which the individual agents spend their money in dSþ dS during ðt; t þ dtÞ.1 Let
i 0
then Yi be the number of equally priced entities of unit price Pj that agents spend on
entities of the category j in dSþ
i dS0 during (t,t þ dt). It follows in accordance with
(8.1) that
X
Yi P ¼ Yij Pj (8.2)
j
Herein Yi is the joint entropy outflux of Si. We shall denote the subsector of Si in
which only the economic entities of category j are stocked by Sij. The subdivision of
Si in Sij as well as the subdivision of dSþ þ
i dS0 in dSij dS0 we shall call
compartmentalization. Mark Yij is “stocked” in dSij dS0 .
þ
As regards (8.2) we encounter the same ambiguity problem in fixing Yi and P
separately if only the product Yi P is known. We can however proceed with
differentiating left and right side of (8.2) to time t as (8.2) holds for all time t.
This results in
X
ðyi þ pÞ Yi P ¼ yij þ pj Yij Pj (8.3)
j
Mark that Yi P is the joint money outflux of the sector Si of the economy. It
includes here the outflux of Si only. Also note that Yij Pj is the money spent on
entities of the category j within the sector Si of the economy. Let us here adhere to
_ _
the notation Yij Pj ¼ Y ij and Yi P ¼ Y i so that we can restate (8.3) as
_ _
X Y ij X Y ij
yi þ p ¼ yi j _
þ pj _
(8.4)
j Yi j Yi
_
._
The quotients Y ij Y i are measurable quantities. They are often called the budget
index factors. We will denote them by fþ : ij
1
Mark that financial entropy is not part of the collection of economic entities. E.g. interest on
mortgage has an entropy content but it lacks a quantity content like books and cars have. Thus only
the contents of conditional differential sets are suitable for price and quantity index calculation.
190 8 Calculation
_
Y ij
fijþ ¼ _ (8.5)
Yi
In the left side of (8.6) we encounter the sum of two terms: the relative rate of
quantity change yi and the relative rate of average price change p. Likewise we have
P
on the right side a quantity rate change component ij fijþ and a price rate
jy
P P
þ þ
change component j pj fij . It is tempting to equate p and j pj fij . Then
X X
yi ¼ yij fijþ and p ¼ pj fijþ (8.7)
j j
Equations (8.7) give us the means to calculate yi and p from the constituents yij
and pj. It presents two separate expressions for quantity index calculation and price
index calculation. These equations are known as Divisia’s quantity index and price
index formulas.
Divisia refrained from presenting a general proof of his quantity index and price
index formulas. When stating expression (8.7) the first expression of (8.7) was
justified by demanding the average relative price-level change p to vanish should all
pj vanish in (8.6). Similarly, the second expression of (8.7) was justified by
demanding the average relative quantity change yi to vanish should all yij vanish
in (8.6).
Indeed, under these restrictions (8.7) follows from (8.6). However, this argument
fails to establish a universal proof, because (8.7) has not been demonstrated to hold
also in situations were not all pj or all yij to vanish.
The correct argument for inferring (8.7) is as follows.
_
Let us express Y i , Yi and P in (8.2) as dependent variables of time:
_
Y i ðt þ dtÞ ¼ Yi ðt þ dtÞ Pðt þ dtÞ
Herein Yi ðt þ dtÞ is the number of outflux bitpulses originating in dSþ
i j dS0
during the time-interval (t,t þ dt). Pðt þ dtÞ is the average unit price of these
_
bitpulses. Y i ðt þ dtÞ is their money value. Likewise
_
Y i ðtÞ ¼ Yi ðtÞ PðtÞ
8.1 Divisia’s Index Formulas for Output Calculation 191
Yi ðtÞ is the number of outflux bitpulses originating during the time-interval
_
ðt dt; tÞ. PðtÞ is the average unit price of these bitpulses. Y i ðtÞ is their money value.
Clearly, the set of Yi ðt þ dtÞ outflux bitpulses originating on (t,t þ dt) is different
from the set of Yi ðtÞ outflux bitpulses originating on ðt dt; tÞ . Hence the
_
money value Y i ðt þ dtÞ, observed at time t þ dt, is the money value of a different
number Yi ðt þ dtÞ of bitpulses than the number of bitpulses associated with money
_
value Y i ðtÞ, observed at time t. The same number Yi ðtÞ would cost a money value of
Yi ðtÞ Pðt þ dtÞ at time t þ dt. The difference of Yi ðtÞ Pðt þ dtÞ and Yi ðtÞ PðtÞ is
then the rise of non-financial spending due to price change only. Thus
Yi ðtÞ Pðt þ dtÞ Yi ðtÞ PðtÞ ¼ Yi ðtÞ dPðtÞ
¼ The rise of non-financial spending due to price change only on (t,t þ dt).
_
The same argument applies to the money valueY ij ðt þ dtÞ ¼ Yij ðt þ dtÞ Pj ðt þ dtÞ,
spent on Yij ðt þ dtÞ outflux entities of the category j during (t,t þ dt) and the money
_
value Y ij ðtÞ ¼ Yij ðtÞ Pj ðtÞ, spent on Yij ðtÞ outflux entities of the same category j
during ðt dt; tÞ. That is
Yij ðtÞ Pij ðt þ dtÞ Yij ðtÞ Pj ðtÞ ¼ Yij ðtÞ dPj ðtÞ
resulting in
X
pðtÞ ¼ fij ðtÞ pj ðtÞ
j
2
The application of this formula is correct for consumer commodities because consumption goods
and services represent only finished goods of which the relative change of unit prices is generally
known. This is different for the complete range of goods and services produced by the investment
sector. This contains also intermediate firm to firm deliveries of which the relative change of the
unit prices is difficult to determine.
192 8 Calculation
This completes the proof of Divisia’s price index formula of (8.7). The proof of
Divisia’s quantity index formula of (8.7) follows from the latter and expression
(8.6).3 □
A disadvantage of the Divisia price-index formula (8.7) is that it requires
continual updating of the budget basket and for that reason it is difficult to
implement in statistical routine calculations. Collecting appropriate budget data is
a time consuming occupation. To make the effort controllable, more practical
schemes have been devised to implement (8.7). Among these the Laspeyres’
index formula, Paasche’s index method, Fisher’s index method and Törnqvist
index formula are quite common. They should be put to the test by comparing
them analytically with Divisia’s index formulas. This is a subject we shall not
consider here.
Since it is impossible to measure output, outflux, input and influx over time-
intervals of infinitesimally small time-length dt, the first practical adjustment is the
numerical approximation of instantaneous output, outflux, input and influx by
measurements over finite time-intervals, e.g. monthly or quarterly.4 This is the
first source of inaccuracy of price and quantity index calculations. The relative error
(and bias) will depend on the statistical properties of the time-series of outflux and
influx (among these properties the frequency spectrum is the most important).
Much research has still to be done in this direction.
For price and quantity calculations in the investment sector Divisia’s price and
quantity index formulas are less suited. The price and quantity index formulas can
best be applied in the consumption sector S1 . To that end we must approximate
Divisia’s formulas by practical schemes.
Most common for calculating the consumer price index (CPI) is Laspeyres’
method. This method treats the budget index factors fiþj in the Divisia price-index
formula as constants. In actual practice the budget basket is periodically determined
by budget research studies extending over one particular year of consumer spend-
ing. Once determined the budget basket remains fixed for several years to come and
will only be revised after consumption patterns and habits have changed so much
that it must be updated in order to remain applicable. The set of budget factors will
therefore be revised once in about 5–6 years. The current intention is to revise the
consumption basket more frequently since modern consumption patterns tend to
change faster and computer technology constantly opens more possibilities to attain
3
Divisia’s index formulas are the only equations of orthodox economic theory that hold exactly,
although the proof here stated for Divisia’s index formula is not part of the orthodox legacy. The
validity of definitional equations of orthodox economic theory such as Y ¼ C þ I does not rest on
a derivation, but on common consent.
4
This is a common practical problem facing all differential quotients in the sciences. E.g. in
physics the measurement of instantaneous velocity v ¼ ds/dt, in which s represents distance and
t time as usual, is confronted with the same difficulties for ds and dt tending to zero. Accuracy of
measurements and observation is always limited irrespective of the domain of investigation. To
tackle the matter we need the powerful techniques of numerical analysis, which is not the subject
of the present treatise.
8.2 Compartmentalization and Influx Calculation 193
that. Another intention is to adjust the set of consumption budget factors each year
in accordance with the consumption pattern over the last period of 2–3 years. The
assumed constancy of the budget-index factors introduces an error-term in the
calculation of the rate of inflation.
Laspeyres’ method is widely used to calculate the rate of inflation of consump-
tion. It appears that it warrants the least error and bias because it is closest to
Divisia’s theoretical index formulas.
We will here not dive into the details of the pros and cons of the various methods
of price index calculation. The problem is one of econometrics. It appears that still
much can be done to improve the accuracy.
In the preceding section we have dealt with deriving Divisia’s index formulas of
which approximations have historically been successfully implemented and applied
to labor output (consumption). In principle Divisia’s scheme of index calculation is
applicable to other processes of economic flow: influx, input and output. However,
there are some serious practical limitations. Like Divisia’s formulas are difficult to
implement for the outflux Y2 of the investment sector, the obstacles that beset
the
implementation of Divisia-like index formulas for influx Xi ¼ Zdt H dS þ
i dS0
are often not less difficult to overcome than for entrepreneurial outflux. Like for
outflux Y2 , the great difficulty is to establish a reliable and practical scheme of
compartmentalization.
Fortunately, compartmentalization for labor influx (wages) X1 appears to be
within reach. As it helps to understand the mechanism of index calculation,
þ I shall
therefore sketch how this can be done for the influx Xi ¼ Zdt H dSi dS0 of the
sector Si .
To that end we shall first devise a general scheme of compartmentalization. All
forms of compartmentalization have in common that they divide up the non-
differential sector Si in non-differential subsectors Sij which do not overlap one
another. We have
[
Si ¼ Sij
j
3. The division of the differential set dSþ
i dS0 in non-overlapping differential
P
subsets dSþ
ij dS0 . In this case Y ij is the outflux of Sij with Y i ¼ j Y ij .
þ
4. The division of the differential set dSi dS0 in non-overlapping differential
þ P
subsets dS
ij dS0 . In this case X ij is the influx of Sij with X i ¼ j X ij .
Pj ðtÞ is the average unit price of entropy within Si j at time t. This will generally
differ from the average unit price PðtÞ of entropy within Si. We introduce further Cij
and Cij to denote the capacity, respectively the capital, stocked in Sij .
The compartmentalization dealt with in Sect. 8.1 is a typical example of the third
kind of compartmentalization. þ
In the present section we shall deal with compartmentalization of dS dS in
þ
i 0
dSij dS0 such that Xi j is the influx of Sij. This is the fourth kind of compartmentali-
zation listed above and for i ¼ 1 it deals here with labor and wages. However,
although we shall do so in the sequel, we need not delimit the discussion to the case
i ¼ 1 only.
We have in the general case:
_ X_ X X
Li dt ¼ Li Wi dt ¼ Xi P ¼ Lij dt ¼ Lij Wij dt ¼ Xij Pj (8.8)
j j j
With i ¼ 1, PX1 is the total of money wages within the economy. Pj X1j is the total
of money wages within dS þ
1j dS0 . The basal idea is that generally the quotients, the
weighing factors
_
Pj Xij Lij
fij ¼ ¼ _ (8.9)
PX 1 Li
are measurable quantities and are fairly well to measure independently in the course
of time. Moreover it is assumed that the price inflation rates pj of the subsets Sij can
be measured with sufficient accuracy. After differentiating left and right side of
(8.8) to time t we obtain
X X
xi þ p ¼ xij fij þ pj fij
j j
8.2 Compartmentalization and Influx Calculation 195
Thus for i ¼ 1 this delivers a rather roundabout way to calculate xi from its
compartmentalized portions xij as well as another manner to determine the inflation
rate p. The relationship p þ x1 ¼ l1 þ w1 of (7.17) suggests to substitute l1 for x1, w1
for p in (8.10) and introduce the labor force L1j employed in S1j for X1j and the money
wage W1j dt earned per employee in S1j during (t,t + dt) for Pj in (7.17). Then for
i ¼ 1 (8.10) would transform into
X X
l1 ¼ l1j f1j and w1 ¼ w1j f1j (8.11)
j j
_ _
As PX1 ¼ W1 L1 dt ¼ L1 dt and Pj X1j ¼ W1j L1j dt ¼ L1j dt, we have
_ _
f1j ¼ Pj X1j =ðPX1 Þ ¼ W1j L1j =ðW1 L1 Þ ¼ L 1j L 1 (8.12)
confirming (8.9).
Well the crux is: Is (8.11) correct? Let us analyze this in much more detail in the
_
way we derived (8.7) in Sect. 8.1. With L1 dt ¼ L1 W1 dt being the total of money
wages earned within dS þ
1 dS0 during (t,t þ dt) we notice that
_ X
L1 dt ¼ L1 W1 dt ¼ L1j W1j dt (8.13)
j
_
Let us further express L1, L1 and W1 in (8.13) as dependent variables of time. We
have
_ _
L1 ðtÞ ¼ L1 ðtÞW1 ðtÞ and L1 ðt þ dtÞ ¼ L1 ðt þ dtÞW1 ðt þ dtÞ
_
pj þ x1j ¼ l1j þ w1j ¼ l 1j
Herein pj is the inflation rate of the average unit price for which the influx X1j can
be produced. From (8.10) we obtain
X X
x1 ¼ l1j þ w1j pj f1j and p ¼ l1 þ w1 x1 ¼ pj f1j
j j
_
Mark that l 1 ¼ x1 þ p so that the first equation can be restated as
_ X X_
l1 ¼ l1j þ w1j f1j ¼ l 1j f1j (8.14)
j j
Note the subtle difference between (8.14) and the incorrect (8.11). The
coefficients f1j represent the relative weight of the wages by definition of (8.9)
and (8.12). Differentiation of (8.8) to time yields
_ _ X_ _
l 1 L1 ¼ l 1j L1j
j
8.3 Aggregation Properties of Capacity and Capital 197
_
Subsequent division by L1 yields (8.14). The relative rate e of effectiveness
growth is easy to calculate from the knowledge of w1 and p [consult (7.18) for help:
if there is additional knowledge about ρ_ 1 =
ρ1 ]. With the knowledge of the rate e of
effectiveness growth we can find x1 as the sum of l1 and e.
Equation (8.10) gives a correct scheme of index calculation of x1, but it is not the
most practical to calculate x1 . We have only used it to explain the theoretical
background of index calculation. Of course a more practical scheme for the
calculation of x1 is furnished by (7.17):
_
x1 ¼ l 1 p
Once we know x1 we have a basis to calculate X1 and, if we know Y1 as well, we
can also calculate dC1 .
The insertion of dCi for Yi Xi and dCij for Yij Xij delivers
X X
dCi dP þ P d2 Ci ¼ dCij dPj þ Pj d2 Cij
j j
Integration yields
X
PdCi ¼ Pj dCij (8.16)
j
It might help to calculate dCi assuming that the data of Yij Xij , Pj and P are
available.
In the above we have derived the aggregation formulas for capacity. We may
follow the same approach for capital Ci with Yij representing the outflux of the
subset Sij and Xij the influx of the subset Sij. Let further dCij be the growth of capital
Cij stocked in subset Sij : The aggregation formulas for money outflux and money
influx are:
X X
Yi P ¼ Y Pj and Xi P ¼
j ij j
Xij Pj
Differentiation yields
X X
Yi dP þ Pd Yi ¼ Yij dPj þ Pj dYij
j j
X X
and Xi dP þ PdXi ¼ Xij dPj þ Pj dXij
j j
8.4 The Calculation of the Macro Variables of the Economy 199
The insertion of d Ci for Yi Xi and d Cij for Yij Xij delivers
X X
dCi dP þ P d2 Ci ¼ dCij dPj þ Pj d2 Cij
j j
Let us recall the first equation of (7.13) dealing with the relationship between the
growth dCi of sector capacity and the relative saving of work-time:
dCi ¼ η
i Yi (8.17)
In (8.17) η0 s are rates of relative time-saving. This equation lends itself for
compartmentalization.
S That is, Si will be divided in separate small subsectors Sij
ð j Sij ¼ Si Þ that serve as entropy producing units with money input Pj Xij and
^
money output Pj Yij . Let us introduce the notation Xij ¼ Pj Xij for money input and
^
Y ij ¼ Pj Yij for money output of Sij . It follows that
^ X^ ^ X^
Xi ¼ Xij and Y i ¼ Y ij
j j
dCij ¼ η þ
ij ζ ij Cij dt ¼ ηij ρij Cij dt (8.18)
þ
η
ij and ηij are the rates of relative time-saving of the subsector Sij . Further ρij
denotes the input circulation rate of Sij and ζ ij denotes the output circulation rate of
Sij .
Then, with Xij ¼ ρij Cij dt and Yij ¼ ζ ij Cij dt, we can restate (8.18) as
dCij ¼ η þ
ij Yij ¼ ηij Xij (8.19)
P
Mark that dCi ¼
6 j dCij because the unit prices of entropy of the subsets Sij differ
from one another. To arrive at variables that can be aggregated directly, we must
first multiply by the unit prices. The multiplication of (8.17) by its average price
^
level P results in PdCi ¼ η
i Y i and the multiplication of (8.19) by its average price
^
level Pj results in Pj dCij ¼ η
ij Y ij . P
Then, on behalf of (8.16), PdCi ¼ Pj dCij . Hence it must follow that
j
^ X ^
η
i Yi ¼ η
ij Y ij (8.20)
j
Herein
σ
ij σ ij Yij Xij σ
ij σ ij Yij Xij
η
ij ¼ ¼ and η þ
ij ¼ þ ¼
σ
i Yij σ ij Xij
þ
σ
ij and σ ij represent the input-bit handling rate, respectively the output-bit
þ
handling rate of Sij . They satisfy the equation σ
ij ρij ¼ σ ij ζ ij .
Equation (8.20) can be restated in the form:
^
X Yij
η
i ¼ η þ
i j fij with fijþ ¼ ^
(8.21)
j Yi
^
.^
In this equation the weighing factors fijþ ¼ Y ij Y i should not be confused with
the budget index factors fþ of (8.5). The coefficients f þ represent measurable
ij ij
quotients of money output rather than quotients of outflux. The scheme of (8.21)
requires another compartmentalization scheme. The focus is on the calculation of
η
i for which we need to know the rates ηi j of relative time-saving in the course of
production within Si j during ðt; t þ dtÞ. To do that separately for the production
of consumption ði ¼ 1Þ and the production of investment ði ¼ 2Þ is difficult if not
8.4 The Calculation of the Macro Variables of the Economy 201
To calculate this coefficient we must know the total money value added Pj Y0j of
the subsector S0j as well as its net money result Pj dC0j, the surplus of money output
over money input, which is the most difficult to determine sufficiently accurate.
The index formula (8.21) is not intended to include the entire collection of
subsectors S0j that together form the economy S0 . But it should include a represen-
tative collection of sample subsets S0j of the economy S0 large enough to average out
the errors in the determination of the coefficients η0 j, especially those due to errors
and bias in the measurement of the money results Pj dC0j .
η
Instead of calculatingP 0 by employing
P the index formula (8.21), we could just as
well calculate the totals j Pj Y0j and j Pj dC0j directly from the available samples
of S0j :
P
j Pj Y0j
η ¼P
0
j Pj dC0j
QðμÞ QðλÞ
η
0 ¼
QðμÞ
QðμÞ QðλÞ
η
0 ¼
QðμÞ
202 8 Calculation
Now that we have calculated λ we can calculate PX0 ¼ PX1 =λ and PX2 as the
difference of PX0 and PX1 .
Next we find PY0 from PY0 ¼ PY1 =μ and further PY2 as the difference of PY0 and
PY1 .
As P has already been calculated we can get all the entropy flows Xi, Xi, Yi and Yi
by dividing the corresponding money flows by P.
Net capacity growths dC0 , dC1 and dC2 follow from dCi ¼ Yi Xi . This will
assist us to calculate C0 , C1 and C2 as well as the growth rates c0 , c1 and c2 in the
course of time.
Also the financial influxes p
1 C1 dt , p
2 C2 dt , and p 0 C0 dt and the financial
þ þ þ
outfluxes p1 C1 dt, p2 C2 dt, and p0 C0 dt can now be calculated with the help of
p þ
i Ci dt ¼ Xi X i and p i Ci dt ¼ Yi Y i
þ þ
This in the knowledge of p1 , p2 , p1 , p2 and in the knowledge of
þ results
γ¼ p ¼ p . Then dCi follows from
0 0
X0 Y0
Zdt ¼ ¼
H ð X 0 Þ H ð Y0 Þ
1 1
qij log qij ¼ H ðXi Þ þ H Yj qij HðX0 \ Y0 Þði; j ¼ 1,2,3, ,NÞ (4.8)
N N
2H ðX1 jY0 Þ þ HðY0 Þ ¼ 2q11 log q11 2ðλ q11 Þ logðλ q11 Þ
ζ_
y1 ¼ 1 þ c1 (8.24)
ζ1
We shall assume that p and y1 have been determined by price and quantity
index calculation in the way set out in Sect. 8.1. We need next to calculate ζ 1 ðtÞ and
ζ_ 1 ðtÞ ζ 1 ðtÞ sufficiently accurate at observation time t.
Well ζ 1 is the labor outflux circulation rate, which is connected with the
probability density function φ 1 ðt; τÞ of current lifetime of the labor outflux bitpulses
as given by (5.11):
204 8 Calculation
8 τ 9
< ð =
1 ðt,τÞ ¼ ζ 1 ðt τÞ exp ζ 1 ðt ξÞ dξ
φ
: ;
0
Here τ is the current lifetime of outflux bitpulses. This differs slightly from the
current lifetime τ0 of work-units.
Current lifetime τ0 is the time elapsed at observation time t since the work-unit
has been recruited in S1 . Thus we must collect the data of the various initial times
t τ0 at which the work-units of S1 have been recruited for delivering effort. This
can be done by collecting all the initial times at which the employees of S1 have
been recruited while weighing these data appropriately for the effect of part-time
labor.
This will be used to determine the aggregate statistical circulation rate χ 1 ðtÞ of
the current lifetime τ0 of the work-units of S1. We shall not go into a discussion of the
technical details of how this can be done. It appears that it can be done quite
accurately.
However we must take into account that χ 1 ðtÞ is the outflux circulation rate of the
work-units L1 of S1. To get the outflux circulation rate ζ 1 ðtÞ of the outflux bitpulses
we need compensate for the change of the effectiveness E of the work-force L1
during ðt; t þ dtÞ as follows:
ζ 1 ðtÞ ¼
χ 1 ðtÞ þ eðtÞ
The problem is now that we have been able to calculate χ 1 ðtÞ, but we don’t know
ζ 1 ðtÞ and eðtÞ yet. However, (7.15) comes to our rescue:
c1 ¼ e þ l1
ζ 1 ðtÞ ¼
χ 1 ðtÞ þ c1 ðtÞ l1 ðtÞ
Furthermore, c1 ðtÞ can be eliminated from the latter equation with the help of
(8.24). This results in
_
ζ 1 ðtÞ þ ζ 1 ðtÞ ¼
ζ 1 ðtÞ χ 1 ðtÞ þ y1 ðtÞ l1 ðtÞ
Knowing χ 1 , y1 and l1 this differential equation can be solved for ζ 1 and ζ_ 1 ζ 1 .
Subsequently the growth rate c1 of labor capacity and hence dC1 can be
calculated. Also the growth rate eðtÞ of effectiveness follows from e ¼ c1 l1 .
The calculated dC1 and Y1 may then be used to determine η 1 on behalf of (7.13).
It can be shown that
8.6 Other Sources of Data Collection 205
This opens a way to calculate λ from the knowledge of μ and all the other
variables of the economy as discussed in more detail in Sect. 8.4.
If our expectation is confirmed that
χ 1 can be calculated sufficiently accurate, this
alternative route of calculation may be a very attractive one to implement.
We have concluded the latter two sections with particular schedules for calculating
all the statistical dynamic averages of the economy. However there are more ways
of how this can be done and much of that is still to be investigated given the many
newly derived equations that determine economic evolution.
One major issue is whether we can measure the transmission Zdt HðXi \ Yi Þ and
more specifically Zdt H ðX0 \ Y0 Þ directly. On behalf of (7.1) the total money dM0
(in bits) issued by the banking system during ðt; t þ dtÞ is equal to the latter
transmission. That is
ð
1
Pþ
i ¼ Pðt ξÞ
φi ðt,ξÞdξ (5.23)
0
8 τ 9
< ð =
i ðt,τÞ ¼ ζ i ðt τÞ exp ζ i ðt ξÞ dξ
φ (5.11)
: ;
0
Thus if we have knowledge of the historic time-course of PðtÞ and ζ i ðtÞ for the
determination of φ i ðt; τÞ in accordance with (5.11), we can use that information to
calculate Pþ
i ðtÞ.
Chapter 9
Theory and Confirmation
Abstract The ins and outs of evidence and the interplay of evolvodynamics with
Keynesianism, monetary policy, Euro crisis and orthodoxy are discussed.
In science consistency and coherence of laws with explanations and observations
have top priority. Postulated principles and equations that derive from these
principles must be mutually consistent and coherent. If there is absence of evidence
for a theory or if our derivations from these principles lead to mutual conflicts and
contradictions, the premises of a theory are the first to be questioned and, only if
possible, reshaped and reformulated We must uncompromisingly stand up for these
scientific standards. Nevertheless it is possible to maintain a theory, in spite of its
failure in a general sense, if there is a well-defined region of its variables for which
the theory is still accurate. However no such orthodox economic theories are known
to exist.
A major obstacle to test for evidence, certainly for the discipline of economics, is
the limited accuracy by which we can do our calculations and measurements. We
must always take into account the inaccuracy of our measurements and insist on
improving the accuracy where we can, e.g. by improving our calculation schemes
or by introducing other schemes of calculation or by adjusting the way we collect
data. What we can’t tolerate is to adopt theories when measurements of sufficient
accuracy contradict what the theory asserts.
The equations of successful physical science are based on the principle of physical
homogeneity. This principle requires that the dimensions of each of the terms of the
equation on both sides of the equation are the same. The principle does allow
conversion of the scale of dimension in which a variable is expressed to another
scale of that dimension without affecting the validity of the equation.
The principle of physical homogeneity is in fact the equivalent of the principle of
evolutionary homogeneity that applies to the variables of economic evolution (See
Sect. 1.2).
The principle implies that the measure in which the content of a variable is
expressed is relative. Absolute variables do not exist, neither in physics nor in
economics. Without this principle physics is inconceivable and I claim that the
formulation of economic theory without this principle of homogeneity, focused on
time-dependent variables of entropy, money value and time, is just as
inconceivable.
The curious thing is that orthodox economics cherishes instead her definition of
homogeneity, which is quite different from the definition of homogeneity applicable
to physics and evolvodynamics. We shall state that definition here for completeness
sake with respect to an equation with one independent variable F and two dependent
variables x and y: The following equation involving the function F ¼ Fðx; yÞ is called
homogeneous of degree g if
The definition can be easily extended to cases with many independent variables
and many dependent variables.
To distinguish the evolutionary definition of homogeneity from the orthodox
definition of homogeneity we shall add the adjectives “evolutionary” and/or “physical”
to the first.
Core orthodox economic theories like neoclassical theory of production and the
Marshallian theory of demand, rely on the assumption that the equations are
homogeneous, e.g. of degree zero, degree one or degree 2.
I have great doubts about the admissibility of these orthodox homogeneity
assumptions. Neoclassical production functions are usually said to satisfy homoge-
neity of first or second degree. It appears to me that this is in direct conflict with the
principle of evolutionary homogeneity because output F is a flow (per unit of time)
and the inputs x (capital) and y (labor) are stock variables in accordance with the
neoclassical model.
This is a conflict of dimensions between the two sides of the equation that should
not be lightly accepted. I have already criticized the neoclassical theory of produc-
tion from another angle in Sect. 1.9 so that I shall not go into this matter further with
respect to neoclassical production theory. Instead I shall here focus on the signifi-
cance of orthodox homogeneity for mainstream demand theory.
9.1 Consistency, Yes or No? 209
The economist Keuzenkamp has written an interesting book on the latter matter
in which he confronts methodological considerations with the acceptance of ortho-
dox demand theory as an explanatory scientific device for explaining the demand
side of price-formation (Keuzenkamp 2004). As Keuzenkamp remarks, “The con-
dition of homogeneity of degree zero in prices and income of Marshallian demand
functions belongs to the core of micro-economic wisdom. Simply put, this condi-
tion says that if all prices and income change proportionally, expenditure in real
terms will remain unchanged.” (Keuzenkamp 2004, page 180).
The principle is more widely known as the principle of the absence of money
illusion. In this respect the state of mind by which economic agents might think to
effectuate a change in their real volume of expenditures consequent on a propor-
tional change of prices and income is considered to be an illusion.
Keuzenkamp investigates the validity of the absence of money illusion by
discussing the outcomes of the many tests that economists have carried out in
that respect. He begins with remarking that the homogeneity condition gives an
idealized picture of the world of demand. Whether it is valid, so he pursues,
depends on the validity of auxiliary assumptions [page 190].
Then he makes mention of quite a number of serious tests to verify demand
homogeneity of which almost all confirm that the empirical data are inconsistent
with the homogeneity property [pages 196–199]. Eventually Keuzenkamp had at
least to admit that “economists have tested these (homogeneity) conditions, and, in
many cases, have had to reject them statistically” [page 181]. Clearly then, the
results point in one conclusive direction: the hypothesis of homogeneous demand is
not more than a speculation on shaky ground.
Keuzenkamp is very detailed in enumerating the various testing attempts and
one should expect that he would have concluded with rejecting the homogeneity
property of functions of demand because of the lack of evidence. Well, this is not
what he does.
The principle of homogeneity of degree zero is definitely a pillar of orthodox
demand theory. To reject orthodox demand theory is manifestly a step too far to go
for Keuzenkamp. So, he devotes a considerable part of his book on discussing the
philosophy of scientific reasoning and, particularly, of econometric reasoning,
which he imputes a methodology distinctive of that of the exact sciences. In this
respect he pays much attention to contradict Popper’s realist philosophy of science
with its emphasis on consistency and falsification. What the connection of these
methodological considerations is with the subject is not very clear. It appears that
the oddity of ignoring the consequence of negative test results (i.e. to renounce
Marshallian demand theory) is concealed by sheltering behind the methodological
dust-cloud of a great many of partially conflicting views about the meaning and
significance of scientific theory formation. Notwithstanding Keuzenkamp’s
expositions on diverse schools of methodology, the matter is more likely an
exercise in propositional logic:
Let A be the statement that demand is homogeneous (of degree zero). Let B be
the statement that micro-economic demand theory is correct. For B to hold good it
is required that A is true. But evidently Keuzenkamp’s hope might be that this does
210 9 Theory and Confirmation
not exclude the possibility that B is true in some exceptional cases for which A is
false. However, it seems impossible to derive the properties of micro-economic
demand theory if A is false. Thus this is not a good escape. And we must then
conclude that if A is false, B must be false. The very last hope for Keuzenkamp is
then that there might still be exceptional cases for which A is true. However, this
reduces the validity of B to exceptional cases, not a very attractive prospect for a
scientific theory that is supposed to provide for explanations as widely universal as
they can be formulated.
Keuzenkamp avoids discussing the matter from this side. Instead he focuses on
what philosophers of science have to say about methodology and the truth status of
scientific laws.
Much what Keuzenkamp says looks reasonable in the passages devoted to
methodology: “I believe explanations serve a purpose: they are useful in
constructing analogies which may improve theoretical understanding.” [page
215]. He remarks that “explanations are not roads to truth”. Well that remark forces
an open door. What is truth? In the sense Keuzenkamp uses it, he seems to hint at
big truths, but why should scientific economic explanations have that status? Only a
convinced materialist or lop-sided scientist may identify an economic testresult or
theoretical derivation with a big truth. Let us remain modest in science but keep
demanding consistency and coherence and satisfying the principles of propositional
logic of our empirical findings and of our derived theoretical laws describing the
behavior and relationships of economic variables! No more and no less than that.
Keuzenkamp accepts the negative test results with respect to the homogeneity of
demand. In line with that he thinks also that money is an illusion: “In the form of a
penny or a dime money is pretty real, but its true value may be an illusion! And as
an aggregate, like M1, it is highly problematic.” [page 216]. He remarks further that
it has no real existence outside the context of a specific theory [page 216]. Does he
intend to suggest with the latter remark that—as far as money is illusive—it cannot
be described within a scientific framework? And that money is scientifically elusive
so that we can only analyze it scientifically if we render it reality and if we accept
the absence of money illusion contrary to what money actually is?
If this is what Keuzenkamp is hinting at, that does not make very much sense. It
implies that we do better to accept A as true although it is false because it is
otherwise impossible to formulate as good a theory as B is.
Or should we interpret his latter remark as an apology that economic science
cannot be done better than that?
According to Keuzenkamp econometrics is positivist. “It is characterized by an
emphasis on observation, measurement and verification. Furthermore, positivists do
not believe in the need to search for ‘true causes’ and tend to be sceptical about deep
parameters.” [page 258]. Well no problem, but why should that imply that we must
accept the negation of a core premise not to be of much significance for a theory that
is crucially based on that premise?
Keuzenkamp concludes that “the homogeneity condition survived thanks to its
strong prior support.” [Page 212]. After all falsification is not applicable to
9.2 Evolvodynamics and Keynes 211
Today it is a little more than three quarters of a century ago that the great economist
Keynes introduced the consumption function as an explanatory device to demon-
strate the multiplier effect in response to increments of investment (Keynes 1936).
The model was appropriate to elucidate the argument that investment stimulation
was first and for all required to accelerate the growth of the economy. The fact that
consumption probability μ is nearer to 1 than ð1 μÞ is to 1 played a vital role in
Keynes’ explanation. The model illustrated that the larger μ is, the greater the
multiplier effect and the more the investment impulses will bring about the growth
of the economy. These investment impulses will ultimately work out that μ gets
reduced.
The great merit of Keynes’ consumption function theory was that it foresaw in a
descriptive explanation of the requirement to invest in order to get economic
recovery and to regain growth rather than to effectuate that by a reduction of
those and other expenses. The bad thing was that the argumentation was mathemat-
ically incomplete, even unsound and impossible to translate into a mathematically
adequate model. As far as it has been done, it has very often led astray and to a crisis
in policy implementation. I shall not discuss that in detail further here. Cognizance
and interpretation of the alternative theory of evolvodynamics explained in the
foregoing is enough because this delivers the correct dynamic relationships.
What I wish to emphasize here is that Keynes himself held a much more
complete idea of the economics of growth stimulation than the many mathematical
212 9 Theory and Confirmation
Keynes’ plea for investment stimulation has often resulted in output stimulation
programs that also stimulate consumption. As the stimulation of consumption will
usually cause the output circulation rate ζ 0 to rise, it will often induce y0 ¼ c0 þ ζ_ 0 =ζ 0
to rise initially. As a result a policy of output stimulation that stimulates consumption
relative to investment will sometimes look successful in the short run.
E.g. if λ < μ < 0:703506, the situation discussed in Sect. 6.7 under number 2,
this will bring about growth in the consumption sector ðc1 > 0Þ, but nevertheless
overall economic decline ðc0 < 0Þ. Such policy may appear to work out reasonably
well in the short run, especially if ζ_ 0 =ζ 0 is positive such that y0 may remain slightly
positive for a while, which may convince policy-makers that they are on the right
track to reach recovery. However, the stimulation of consumption will likely cause
μ to rise so that the long-term condition μ < λ for economic growth will not be
reached. A prolonged increase of ζ 0 may extend the period of positive growth of Y0
and it may be very reasonable to achieve if ζ 0 were lagging much behind its normal
magnitude. But ζ 0 cannot rise forever and may even decrease again so that y0 will
not have increased in the end. Thus a policy that concentrates on consumption
stimulation rather than on investment stimulation will likely ultimately reach the
point where c0 is reduced and it will hence be unsuccessful in the end and often
postpone the recovery process.
However in case c0 is sufficiently large, y0 may nevertheless be too small or even
negative due to a negative growth rate ζ_ 0 =ζ 0 of output circulation. In that case
consumption stimulation is enough to eliminate the negative ζ_ 0 =ζ 0 and to restore the
growth of output Y0 again. Thus a policy of consumption stimulation may incidentally
or accidentally be effective to combat some stagnation. Such stagnations are usually
mild and relatively easy to overcome.
Keynes was of course unaware of (6.30), but he had a profound intuitive
understanding of the economic process. To endorse his views he used the consump-
tion function in combination with the multiplier effect. He did not intend to embed
it in a fabric of parametric mathematical equations like the following generation of
economists did. What happened afterwards was a mathematical consumption func-
tion fitted in a parametric model in which consumption probability was merely a
constant parameter so that the most sensitive component δ ¼ ðλ μÞ=2 of economic
growth could no longer theoretically change in response to changes of μ. Well, for
our quantitative engagements it can be done in that way, but it can also be done
much and much better.
The bottom line is that investment stimulation is the best policy to regain
economic recovery after an economic recession. However, how that must be
worked out and will work out for a particular economy depends strongly on the
state of growth the economy is in. The equations that determine the time-course of
economic growth demonstrate complicated relationships, not less complicated for
the two-sector economy, which in fact can only be properly analyzed with the help
of the calculating power of digital computers. However, some qualitative insight
can be gathered by considering four different growth-positions an economy can be
in and by analyzing the prospects of growth for each of these positions. This has
214 9 Theory and Confirmation
It appears that the Dutch economy reached growth position 4 in 2012, although
the economic growth rate is only slightly negative.
Current policy to combat the crisis, strongly and persistently advocated by the
Northern Eurozone-memberstates with a good budget record, stresses to restore
government budget equilibrium as the first and immediate thing to be done by each
of the memberstates. However, given the untenable critical state of budget
disequilibria combined with very high interest rates in some of the member states,
the fiscal union should be implemented without delay. Once the member states had
made the choice for a monetary union nearly two decennia ago, a political/fiscal
union was actually inevitable and it must therefore now be implemented without
delay to avoid persistent serious economic and political drawbacks within Europe.
This requires political wisdom and solidarity among the member states rather than
the play of nationalist populist sentiments, which has brought us only adversity in
the past and will bring us nothing else than that in the future if we don’t succeed to
dress up the political/fiscal union timely.
The absence of a political/fiscal union has not only led to a wide variation of
budget deficits among the memberstates but also to a wide variation in the interest
rates these memberstates pay. The good-record memberstates consider their low
interest rate as their own inalienable reward for their disciplinary budget behavior
of the past. However, that is only very partially so because they would certainly not
have borrowed money for such extreme low interest rates if each of the Euro
memberstates would have kept its own currency.
Moreover one should take into account that within a great economic zone like
the USA, like China or like the Eurozone there will always be regions performing
better and regions performing worse and some even much better or much worse.
This is inevitable and absolute solidarity within the system is a must to prevent that
the zone disintegrates, certainly if democracy is the basis of the cooperation.
The point is that a rigid policy to restore budget equilibrium by each country
separately will everywhere and immediately induce still greater differences in
interest rates between the countries, which does not help to solve the problem but
will only help to worsen it. It appears that a substantial part of the decline of the
interest rates of the good-record member states is connected with the sharp rise of
the interest rates for which the less-disciplinary memberstates borrow money.
All member states of the Eurozone, including the member states with good
budget record, must accept their common responsibility for the total of budget
deficits. As far as this acceptance will have a drawback on the economies of the
Northern member states, it is much more preferable than the serious drawback
caused by muddling on with our own national priorities. Moreover, all member
states should accept their individual responsibility for having contributed to imple-
ment a monetary union in the past without a backbone of sufficient political and
fiscal unification.
Once the Eurozone has established this political/fiscal union of common solidarity,
budget equilibrium can soon be reached by each of the member states. Presumably
interest rates for which the Northern member states borrow will rise, but it is rather
unlikely that it will rise very much because the overall budget state of the Euro zone is
more favourable than the American with its already low interest rate.
216 9 Theory and Confirmation
No doubt the political fiscal union has to be implemented with strict budget
rules. Solidarity between the member states implies that each of the member states
will strictly subscribe to the same budget rules.
The current credit crisis has primarily been caused by the banks and is also due to
insufficient control on the banks by the monetary authorities. The crisis reflects the
general pattern of the most severe recession an economy can go through. Therefore
I shall give it the necessary attention here. To begin let me first state where banks
are for.
The banking system is there to finance the economy. That is the only task banks
have. To that end banks have been granted the monopoly of money creation under
certain restrictions and under independent control of a central monetary authority.
The monopoly of money creation is not the exclusive property of banks. On the
contrary, it has been granted to the banking system by law with the mandate to
manage the financial needs of the economy as good as possible.1 Most important in
this respect are the financial needs of small and middle-sized firms. It is here where
new innovative ideas are brought to life within an economy, often with great
personal sacrifice of time and welfare of the entrepreneurs concerned. It is here
where the risks of financing may be greater than elsewhere (although I doubt it
whether this is really true), but where the economic rewards in the long run are the
greatest. In brief, this is pre-eminently the sphere of work for which the banking
system is first and for all needed and for which it was intended. The general rule
appears to be that economies without a middle class of entrepreneurs are always
stagnating. Nevertheless time and again banks have neglected this essential role of
their mandate and their curious reasons were always that small credit financing is
too costly for the banks and that the risks are too great relative to the risks in other
domains of financing, as if it were not a very special mandate they were endowed
with. But actually this is nothing else than the failure of banks to comply with their
mandate. Banks have been granted the monopoly of money creation to help finance,
not to withdraw from what they are supposed to do. And first and foremost to
sustain and bring forth economic growth is to provide for financing the middle class
of entrepreneurs.
Well banks often forget to stay with the special and very important task they
have been given literally as a gift. As they grow larger, their executives get to dream
of complementary financial activities and the banks begin to supply their customers
with all sorts of often superfluous financial (even usurious) products other than
lending money, just to increase their share of the national pie. Their executives get
the idea that they seriously contribute to (inter)national economic performance in
this manner, but what they achieve for the greater part is creating additional money
value, which is in fact merely a financial portion of the national pie snatched from
the non-banking sector. That is, the banking sector manages in effect to raise the
costs of financing and banking considerably beyond the level it should have in a
1
We should be aware that much of the current banking practice (such as investment banking and
the bonus-culture) does not accord with the intentions of a responsible banking mandate.
9.3 Eurocrisis and Monetary Expansion 217
normal competitive market. The financial share they take and redistribute among
their personnel and executives rises to unrealistic proportions.
As this process goes on and on, the capital actually accumulated by the banking
sector is primarily backed by paper claims on banks and third parties that the
banking/financial sector herself has issued. The accumulation of these paper claims
goes as far as it can go. The more it grows the relatively less the buffer capacity of
banks and the closer the limits are reached that the non-financial sector can cough
up. Thus the more vulnerable the banks get to hitches in the solvability of the non-
banking sector. If it keeps on growing in this manner, the entire system will
ultimately break down and—as a default of the banking system is a very inattractive
option—governments are forced to save the banks massively. In this manner banks
will not, but society will be confronted with the ultimate burden of their behavior.
In effect banks have misused their monopoly of money creation.
This is in a nutshell what happened since the turn of the century and even before
but then yet at a smaller scale.
To recover from this recession is not an easy matter. In the Netherlands mone-
tary and fiscal authorities have put rigid emphasis on reduction of government
expenses, paying off debts, raising bank buffers. This has effectuated credit restric-
tion in the private domain or at least affected businesses to delay their plans of
expansion. Well, clearly, government expenses must be reduced, debts must be paid
off, bank buffers must be raised all in the interest of recovery but such policy should
never accept credit financing to decline. Governments and banks have the bounden
duty to provide sufficiently for the financial needs of firms and businesses in order
to stay far enough off from the turning point of evolution with its very low inflation
rates and growth rates. Policy should have emphasized a different order: first and
foremost credit expansion and then the other policy objectives. Credit expansion
should have priority even if it leads to greater risks (Well, I doubt the risk is greater
because credit expansion will stimulate growth and diminish the risks of failure).
Actually rescuing the banks from default has costed and will cost a lot more,
certainly if we fail to change the rules the banks must accord to.
In the past the lack of sufficient financial means has often driven firms and
businesses into despair and time and again it has resulted in initiatives to found new
banks with the task to finance a specific sector of the economy that until than had
almost no access to the financial resources that banks control. Unfortunately, when
these newly founded banks grew more and more, they gradually withdrew from the
tasks they had been founded for and sooner or later the lack to provide for financing
the small and middle-sized business sector was again on the menu. This has been
too manifest in Europe. (In the Netherlands small banks are gradually disappearing
from the stage). The rules for the entry of newly founded banks have perhaps been
too restrictive and the market dominance of the large banks has outgrown far too
much.
Since the early begin of the current crisis, the business credit problem has popped
up with vehemence. For the investment sector credit limitation rather than credit
expansion became the current fate. Particularly starters’ businesses, small and
medium-sized enterprises were and still are the victim. These entrepreneurs began
218 9 Theory and Confirmation
1þλμ 1λþμ
dM1 ¼ γ C0 dt and dM2 ¼ γ C0 dt (7.5)
2 2
dB1 ¼ b1 B1 dt ¼ γ ðλ μÞC0 dt
dM1 is the increase of the money stock of the consumption sector during
ðt; t þ dtÞ. dB1 is the liquidity surplus of the consumption sector during ðt; t þ dtÞ.
Additional to its own money resources the investment sector employs the liquidity
surplus dB1 of the consumption sector S2 to realize its investment objectives during
ðt; t þ dtÞ. The larger the flow of dB1 the larger the availability of liquidity for S2 .
After differentiation of dM1 and dB1 to time we obtain
m_ 1 γ_ 2δ_ b_1 γ_ δ_
m1 þ ¼ þ þ c0 and b1 þ ¼ þ þ c0 (9.1)
m1 γ 1 þ 2δ b1 γ δ
Let us first discuss the stationary growth solution. Under stationary economic
growth of an economy we understand an economy for which all λi , μi , and the
transmission rate γ remain constant in the course of time as well as the growth rates
ci of capacity. On behalf of (6.27) and (6.30) the maintenance of stationary growth
warrants that ζ i , ρi , ζ i and
ρi are time-independent constants as well.
For the stationary growth two-sector economy it follows thatδ ¼ ðλ μÞ=2,m1 and
b1 are independent of time. In that case (9.1) result in m1 ¼ b1 ¼ c0 . This is indeed
what might be expected: the liquidity stock and the joint capacity stock must share the
same growth rate under the condition of stationary growth. Moreover H ðX0 Þ, H ðY0 Þ
and transmission H ðX0 \ Y0 Þ become constants for stationary growth, whereas the
number Z of samples per unit of time, Zdt HðX0 \ Y0 Þ ¼ γ C0 dt, X0 ¼ Zdt H ðX0 Þ
and Y0 ¼ Zdt H ðY0 Þ share the same stable growth rate z ¼ m1 ¼ b1 ¼ c0. Further γ, λ
and μ behave as constants independent of time under stationary growth.
However, the condition of stationary growth is exceptional and it is entirely
inadequate for economic forecasts to rely on. The economic process is concerned
with stochastic time-series. Much more relevant is the general case for which every
variable is a dynamic stochastic variable of time. Non-stationarity reflects the
general condition of the economy, e.g. the circumstances we face in the time course
from top to bottom of the economic cycle and conversely. We must therefore
consider (6.30) and (9.1) in full ornate in order to draw sensible conclusions.
These equations are appropriate for getting a tentative general qualitative idea of
the effect of monetary expansion and contraction in the case of the two-sector
economy. Let us then assume that the initial condition of a two-sector economy is
stationary growth and that from then onwards a policy of monetary expansion is
pursued discontinuing stationarity and resulting in the increment of the transmis-
sion rate γ and of the transmission HðX0 \ Y0 Þ ¼ γ C0 =Z proportional to γ . The
question is then: what happens next?
Let us first conclude that in case of accelerated monetary expansion the
growth rates m1 and b1 of the money stock and B1 will increase and in particular
m_ 1 =m1 and b_1 =b1 will increase, even more the greater the amount of money injected
into the economy. In virtue of (9.1) γ_ =γ þ 2δ_ =ð1 þ 2δÞ þ c0 and γ_ =γ þ δ_ =δ þ c0
must increase as well. An increase of c0 will not be effectuated immediately.
The immediate effect of a change of m_ 1 =m1 and b_1 =b1 will have an impact on
γ_ =γ þ 2δ_ =ð1 þ 2δÞ and on γ_ =γ þ δ_ =δ.
As far as the change of m_ 1 =m1 and b_1 =b1 affects γ_ =γ, this will have the secondary
effect of raising inflation (in case the inflationary mode of evolution applies)
because γ ¼ pþ 0 ¼ p 0 on behalf of (6.21). On the other hand as far as the change
of m_ 1 =m1 and b_1 =b1 affects 2δ_ =ð1 þ 2δÞ and δ_ =δ, this will have the secondary effect
on the economic growth rate c0 on behalf of (6.30). Whether that raises c0 in the
longer run depends on the sign of δ. In Appendix H several scenarios have been
listed.
220 9 Theory and Confirmation
Consequent on a rise of 2δ_ =ð1 þ 2δÞ and δ_ =δ, c0 will eventually also rise. Hence it
seems at first glance that the change of m_ 1 =m1 and b_1 =b1 will likely to have more
effect on c0 þ 2δ_ =ð1 þ 2δÞ and on c0 þ δ_ =δ than on γ_ =γ. However, this conclusion is
too rash. Note that the changes in m_ 1 =m1, b_1 =b1, γ_ =γ, 2δ_ =ð1 þ 2δÞ and δ_ =δ are short
run changes and that the affected secondary effect on c0 manifests itself in the
longer run, when the short run changes of m_ 1 =m1, b_1 =b1, γ_ =γ, δ_ =ð1 þ δÞ and δ_ =δ may
have smoothed out. Moreover, as a secondary response to the initial rise of γ_ =γ, the
transmission rate γ will eventually rise as well in the longer run while γ_ =γ will then
smooth out. Thus,
In the short run : m_ 1 =m1 and b_1 =b1 effects γ_ =γ þ 2δ_ =ð1 þ 2δÞ and γ_ =γ þ δ_ =δ
In the longer run : m1 ; b1 ; pþ 0 ; δ and c0 rise
The success of realizing more economic growth in the longer run by an impul-
sive rise of m_ 1 =m1 and b_1 =b1 depends on how much λ_ =λ, 2δ_ =ð1 þ 2δÞ and δ_ =δ will be
affected separately. The growth-position the economy is in at the moment a
liquidity flux impulse is injected into the economy, has considerable influence on
that matter (See Appendix H). We should further be aware that the effect of the
impulse of m_ 1 =m1 and b_1 =b1 on λ_ =λ and μ_ =μ will be of the same order of magnitude
as the effect of m_ 1 =m1 and b_1 =b1 on γ_ =γ , 2δ_ =ð1 þ 2δÞ and δ_ =δ.
For a moderately growing economy, a monetary spending impulse will bring
about a rise of the rate of inflation as well as a rise of the rate of economic growth.
How the increase is divided between inflation and real growth depends on the state
the economy is in.
If the growth rate c0 is not affected simultaneously in the longer run, the price
inflation rate p must increase. Then accelerating inflation is our fate and eventually
this will lead to hyperinflation if we keep accelerating monetary expansion condi-
tional to an immovable c0. Hence accelerated monetary expansion is only effective
if it is accompanied by an accelerating growth rate c0 .
Observations on modern growth economies have taught us that usually a part of
m_ 1 =m1 and b_1 =b1 goes eventually into c0 and the remainder into γ_ =γ. The higher the
attainable growth rate c0 , the higher the inflation rate as well.
If monetary expansion is directed at economic recovery or at more growth, it is
necessary that it boosts an initial rise of δ ¼ ðλ μÞ=2 and ultimately a rise of c0 as
we have argued in Sect. 9.2. For that matter monetary policy must be accompanied
by stimulating investment relative to consumption (which decreases the propensity
μ to consume) as well as by keeping depreciation low relative to wages (which
increases λ). These incentives can best be realized if economic policy concentrates
on expansion investing rather than on replacement investing. The more monetary
9.5 Evolvodynamics and Orthodoxy 221
expansion is accompanied by such policy incentives, the less chance that money
expansion will leak away in the growth of the rate of price inflation p and in the
transmission term γ_ =γ of the right side of (9.1).
A policy of monetary restriction can be pursued to reduce inflation. However, its
effectiveness depends on how we succeed to avoid that δ ¼ ðλ μÞ=2 will not
decrease as a result of that policy. If we do not succeed so and δ will only decrease,
as a result, monetary contraction will have little effect on the rate of inflation.
Success can only be realized if investment is not reduced relative to consumption.
The nature of the above expositions of the present and preceding sections is
qualitative. The expositions demonstrate the feasibility and power of the theory of
evolvodynamics. It is of course possible to postulate an alternative system of
parametric equations that may more or less also provide for a qualitative explana-
tion of monetary expansion and economic growth.
However, these alternative systems of equations do not consistently provide for
a general explanation covering the entire time-domain and every economy. They
must be “repaired” continuously in the course of time by adjusting the parameters,
which demonstrates that the parameters are not time-independent despite the oppo-
site assertion by the proponents of those systems. Because of their lack of universal-
ity and their maltreatment of time it is very questionable whether these systems
provide for as good a dynamic explanation as the theory of evolvodynamics does.
The great merit of the equations of evolvodynamics is that they are universally
valid and fully dynamic and that they have been derived. Moreover these univer-
sally valid equations demonstrate that economic constants do not exist (as all
economists are actually aware of) and hence constant parameters cannot exist
either. The application of evolvodynamics is not restricted to qualitative
expositions but is appropriate for quantitative calculations to forecast with optimum
attainable accuracy. If performed without blemish, these calculations are reliable as
long as the few underlying principles of Darwinian selection keep their validity.
Our above qualitative exposition applies to the two-sector economy only. For the
N- sector economy ( N 3 ) the equations become so complicated that useful
qualitative expositions appear to be very difficult. We must then resort to quantita-
tive calculations. To analyze quantitatively rather than qualitatively, it is of course
necessary to oversee the complete system of evolvodynamic equations in detail and
to apply sophisticated tools of numerical analysis that need to be handled by digital
computers. Since the number of the degrees of freedom of the system of equations
is 3 (See Sect. 4.6), it is sufficient to collect the data of three different time-series for
the computation of complete forecasts of the economy.
several new ones have been fitted in a general definitional context and satisfy quite
many exact universally valid relationships, so far unheard of and unexploited in
orthodox economic theory.2 These relationships have not been postulated, but have
been derived from a very minimum of basic principles of Darwinian evolutionary
selection. Moreover, Shannon’s conception of entropy has been shown to represent
the inevitable proper measure of uncertainty, information and value: uncertainty
that agents face, information and value that agents gather and get rid of while
selecting.
It has also been demonstrated that economic behavior of agents is purposeful
rather than purposeless. Nonetheless the way economic agents behave is very
selective. Purpose manifests itself in the way economic agents avoid to select the
enormous bulk of potential variations that can possibly be chosen but that they
consider being useless. Instead they prefer to choose from a much and much smaller
set of typical variations deliberately. The latter set contains an infinite number of
equally meaningful opportunities to choose from. Clearly, this corresponds with our
human experience that as individual decision makers we have very great freedom to
choose between the things we consider meaningful to consume, and between the
things in which to invest purposefully and how to use them up, as well as to decide
between the jobs that we choose to do and to create.
This implies that orthodox economic models in which agents are programmed to
seek equilibrium and to optimize some postulated object function (production
function, output, input, profit, utility) are far from realistic. The purported equilib-
rium and/or optimization always results in a single or a very limited number of
possible solutions to choose from. Moreover it clings together very many
variegated individual decisions of individual consumers, investors, employees
and entrepreneurs as if there is only one decision maker per category that makes
the same consideration for that entire category.
These robot-like agents weigh the aggregate of consumption, the aggregate of
investment, the aggregate of economic resources and the aggregate of jobs as if it
were only a single grand item, or a single grand job. Time is often not considered to
change and if time is considered to change, it is always a poor model of deficient
dynamics.
Growth theories will usually reach not much further than describing theorems of
stable growth and the listing of some golden rules of accumulation.
This slavish conception of economic reality does not correspond with the
unrestricted dynamic freedom of behavior economic agents have and demonstrate
with respect to the way they spend on and choose between multifarious goods and
the way they select between the great variety of jobs under dynamically ever
changing economic conditions.
2
With the exception of Divisia’s index formulas.
9.6 Science, Premise and Prejudice 223
Notwithstanding its relevance and despite its anchorage in the logic of proper and
consistent scientific reasoning I foresee that there is a long way to go before the new
theory will meet some acceptance among economists. This is nothing to worry
much about. Let me in this respect reproduce a famous quote from James Lovelock:
“If you start any large theory, it takes about 40 years for mainstream science to
come around”. So if you may have stumbled on this quote while glancing through
this book, you may lay the book aside for 40 years before reading it.
A major obstacle is the prejudice with which generations of economists have
been educated as to what to understand by good and sound economic reasoning.
Ask any economist, accomplished in orthodox economic theory, what he thinks
about the existence and bearing of economic laws. He will certainly respond by
remarking that economic laws like the derivable type of laws in the physical
sciences do not exist (whether he has any real notion of the struggle by which
these physical laws have evolved, often from originally primitively formulated
parametric equations,3 is—to say the least—questionable). And he will certainly
stress that you cannot describe economic (human) behavior fully by equations
(since he seems to think human behavior is of some other and higher order exalting
far above the material laws that can only describe the dead world of physics. Why
he thinks so is a riddle. It has never been explained nor demonstrated. We are told to
believe it.). The idea that the methodology of scientific reasoning is a common one
for all sciences, he will reject as absurd and qualify as sheer nonsense, because it
conflicts with the prejudices he cherishes.
Well, in accordance with the standards of rational science, any conception,
premise or prejudice that is at the basis of the human pretention of scientific
explanation should be subjected to meticulous scrutiny. There have always been
two procedures to achieve that. Both of them must be well observed to warrant good
science in accordance with good methodology of science. The first is that the
forecasts of a scientific theory must be tested and confirmed by field
measurements.4 This is consistency of facts. The second is to subject the consecu-
tive steps of deductive reasoning to meticulous scrutiny and to compare whether all
the derived laws are consistent with one another as well as universally valid.5 This
is consistency of laws.
3
Ptolemaic astronomy is an example. Another instance is Planck’s derivation of the law of black
body radiation.
4
Mark that confirmation is always tentative, never definitive.
5
One might mention Occam’s razor here, which states that the less the number of basic postulates,
the better and more general the theory. But I think Occam’s razor is just as much a natural result of
the two procedures of scientific scrutiny. The greater the applicable domains of space (read here
also: economy) and time for which there is mutual consistency between the various laws derived
from the first postulates (the principles) of a theory and for which these laws remain sustained by
measurements, the more universal the theory is and the more probable it is that the number of basic
postulates will be reduced to a minimum. For if there are more postulates on which a theory is
224 9 Theory and Confirmation
based, there is greater chance that a law derived from one or a few of the postulates is inconsistent
(or incompatible) with a law derived from another postulate.
6
The following consideration clarifies the shortcomings from a Popperian perspective. While
neglecting the importance of the second procedure (the consistency of laws), practitioners of
orthodox science can propose infinite many parametric models as providing for a universal
economic explanation, because one can always add and adjust the parameters so that
measurements are always in harmony with what the model predicts. Thus there is no way to
refute the “laws” of economic orthodoxy.
9.6 Science, Premise and Prejudice 225
observations of the orbit and the calculated orbit predicted by the theory. The
Ptolemaic system was of course very accurate in explaining the motion of the
distant stars without the need to add any epicycle. Further to explain the orbit of
the sun, epicycles were also unnecessary. An eccentric circle was sufficient to
account for the sun’s orbit around the earth within reasonable limits of accuracy.
For the motion of the planets and moon more was required and here Ptolemy was
lavish in applying epicycles.
Like the parametric models of orthodox economics, it was necessary to adjust
the parameters regularly, those for explaining the orbits of the moon and the planet
Mars often, those for some of the other planets only after several decades. The
appraisal of the success of the Ptolemaic model depends considerably on how
accurate astronomical observations can be made. Science is of course concerned
with explaining the smallest differences that we can observe. Well, it has been
demonstrated that by introducing still more epicycles Ptolemaic astronomy can
explain away any smallest deviation between theory and observation, certainly with
the help of the calculating power of our computer technology age. However, we
must keep on adjusting the many parameters regularly and continually. Every
second of the world-time-clock has in fact its own system of Ptolemaic parameters
and we must recalculate these parameters every second so that unprecedented
accuracy of prediction will be realized for the next second. Well, does that affect
the extent of our scientific appreciation of the suitability of the Ptolemaic model?
Yes or no? Well, if we keep believing and preaching that we can’t replace the old
dogmas by deductive reasoning (like the dominating authority, the Church, did at
least until the year 1633 when Galileo was compelled to repeal his heliocentric
world view), it appears that we may definitely answer that question in the negative.
Indeed “no” is the answer of the orthodoxy! And it was the answer of all the
astronomers that practiced the science of astronomy for more than 1300 years after
Ptolemy, upon whose work the Church based her calendar and the improvements
thereof confirming that the idea of the earth as the immovable center of the universe
was a fruitful one. All well in accordance with the prejudices of that time.
That prejudices have a long lifetime is illustrated by the time it took to adopt the
heliocentric system. Copernicus published his revolutionary heliocentric world
view in the year 1543, also the year he died. It was immediately rejected almost
by anyone who took note of it with the exception of a few. True enough, there was
the difficulty that the Copernican system was still based on circular motion and
epicycles, which did not really contribute to improved accuracy of its predictions
and that did neither contribute much to the acceptance of the Copernican system.
Thus, when the leading Danish astronomer Tycho Brahe (1546–1601) made his
pre-telescopic celestial observations of unprecedented accuracy, he took the stand
that the Copernican system was definitely wrong. Instead Brahe kept working on his
own geocentric system of explanation without succeeding to complete it. It was the
more mathematically oriented Johannes Kepler who—since 1600 in possession of
the observations of Tycho Brahe—discovered that the planetary orbits around the
226 9 Theory and Confirmation
sun were elliptic7 (Crowe 1990; Dreyer 1953). His eight years older contemporary
Galileo was immediately convinced (Galileo 1954), but the latter was compelled by
the Church to recall his position in 1633 and to admit that a moving earth “is
contrary to Holy Scripture and therefore may not be defended or held” (Stillman
Drake 1980).
With Kepler’s publication of 1609 there was overwhelming evidence for
adopting the heliocentric world view. Kepler’s discovery of elliptical orbits implied
also the expulsion of time-dependent parameters from the model. This, I believe,
was the greatest scientific innovation of the new theory. No dynamic parametric
adjustments anymore in the course of time. Nevertheless it took still many years
before the Ptolemaic system with its parametric burden was definitely overthrown
by Isaac Newton and replaced by a general deductive theory of celestial mechanical
motion.8 The lesson is that prejudices have a long survival time. That our genera-
tion is less conservative than the generation of the seventeenth century is improba-
ble. Each generation cherishes its prejudices. The prejudice of orthodox economics
is the clincher that human behavior cannot fully be caught in a deductive theory. If
our belief in this principle is unshakable it places orthodoxy on the thrown of
authoritive inaccessibility like the Church in the dispute with Galileo in 1633.
In this respect we should realize that the idea of Darwinian selection is a
fundamental premise as well. It must permanently be subjected to meticulous
scrutiny just as well as other dogmas. We can’t prove the principles of Darwinian
selection by deductive reasoning. What we can do is to establish evolutionary laws
that derive from it and collect measurements to test the predictions of the laws and
phenomena following from it. If there is correspondence between measurements
and predictions this does not imply that we have proved the basic premise. We will
only be strengthened in our conviction—how much depends on the extent of the
evidence—that the premise of Darwinian selection offers a correct and effective
tentative description of the basic fabric of our domain of investigation as long as it
is not overthrown by a still more general explanatory theory.
That orthodox theory is offering a better testing record than the theory of evolvo-
dynamics does would be a false impression. In the practical situation of collecting
and predicting macro-economic statistics, orthodox economics offers only a frame-
work of unproved parametric equations. Sufficient correlation with test results can
only be sustained by adjusting the parameters all the time, like for the Ptolemaic
system. And even then economic predictions are not better than the predictions a
7
Kepler published his findings in the year 1609.
8
Newton devised his theory of mechanical and gravitational motion in the years 1665–1967 and
kept silent on the matter for many years. It was finally published in 1687.
9.7 Theory and Measurements 227
The task of the scientific experimenter remains to verify and test the theoretical
forecasts against observations. This is a role that has to be fulfilled with respect to
the theory of evolvodynamics. Gathering the required data about macro-economic
time-series is a major enterprise. Many variables of the theory of evolvodynamics
are differential quotients. Their measurement demands an indirect approach.
Measurements always extend in time so that numerical procedures and filtering
techniques are required to prepare them for comparison with the theoretical
predictions. It is a classic problem of time-series analysis, also well known in the
physical sciences, communication and signal processing sciences. This is where
numerical analysis and the expertise of econometrists are badly needed. Thus
instead of composing a set of parametric equations, econometrists must take up
the classic role of the translation between the instantaneous variables of
evolvodynamics and the measurements of those variables by the experimenters
involved in data measurement.
In this respect the present practice of seasonal detrending by economists needs
critical reconsideration. Economic time-series contain information. One of the main
attainments of time series analysis is that data mutilation is fundamentally wrong,
e.g. by the application of the wrong filtering techniques. By filtering out seasonal
fluctuations information gets lost. As E.T. Jaynes remarks in the preface of his book
on probability theory, “this is happening constantly from orthodox methods of
detrending or seasonal adjustment in econometrics” (Jaynes 2004, page xxvii).9
The subject deserves the utmost attention in the process of implementing the
equations of evolvodynamics in numerical computational routines.
The seasonal fluctuations are of considerable importance and optimum use of
them in computational routines improves the interpolations and extrapolations of
the variables over longer periods of time. Economists should therefore be on their
guard with applying the right techniques to link the momentary flow variables, such
as money input PXi money influx PXi , money output PYi , money outflux PYi , the
price levels and circulation rates, to the observed averages over a fixed period T of
time. E.g. Xi ðtÞ respectively Yi ðtÞ is something different than
ð0 ð0
1 1
Xi ðt þ θÞdθ and Yi ðt þ θÞdθ
T T
T T
Ð0 Ð0
We cannot simply replace Xi ðtÞ by T1 T Xi ðt þ θÞdθ and Yi ðtÞ by T1 T
Yi ðt þ θÞdθ. That will cause a loss of information, the more so the larger T is. On the
other hand, some smoothing out over time is inevitable to assure that the zero-mean
errors of measurement smooth out. This is however not a typical economic prob-
lem. It is a common problem of time-series analysis with respect to all sorts of
quantitative scientific disciplines.
9
Neoclassicism is an example of an approach with a very substantial waste of information because
the tenets of neoclassicism do not hold in the short run as we have argued in Sect. 1.7.
Appendix A
Let us consider two-sector input selection for the case that there are only Zdt samples
of which two thirds turn out to be in state 1 and one third in state 2 (See Fig. 1.2).
Then the probability of selecting a sample in state 1 is λ1 ¼ 23 and the probability of
selecting a sample in state 2 is λ2 ¼ 13. In accordance with Shannon’s formula (1.1) the
reassembled entropy input is Zdt HðX0 Þ ¼ Zdt ½ð1 λ1 Þ logð1 λ1 Þ þ λ1 log λ1
and it follows in this special case that Zdt H ðX0 Þ equals 0:9183 Zdt units of
value (u.o.v.).
Well there is a snake in the grass. Shannon’s formula (1.1) is an asymptotic
approximation, which is the better the more samples there are, i.e. the larger Zdt is.
In fact the formula is very accurate for Zdt sufficiently large but it is not if Zdt
remains quite small. The exact general expression for HðX0 Þ is
k!
Zdt H ðX0 Þ ¼ log with k ¼ k1 þ k2 (A.1)
k1 ! k2 !
in which k ¼ Zdt is the total number of samples (in state 0), k1 ¼ λ1 Zdt the number
of samples that turn out to be in state 1 and k2 ¼ λ2 Zdt the number of samples that
turn out to be in state 2.
Suppose then that the statistical experiment is chosen so that Zdt ¼ k is finite and
equal to 21 as for example is illustrated by the input sequence in Fig. 1.2. Calcula-
tion with the help of the exact formula (A.1) results in
Table A.1 The reassembled information content of input selection for different numbers k ¼ Zdt
of samples
Number of Number of ReassembledP information
samples in state 1 samples in state 2 content Zdt λi logλi
Unit of value relative to i
standard u.o.v. λ1 Zdt λ2 Zdt in standard u.o.v.
Standard 14 7 16.827243
10 times as small 140 70 18.874421a
100 times as small 1,400 700 19.226624a
1,000 times as small 14,000 7,000 19.276793a
10,000 times as small 140,000 70,000 19.283304a
100,000 times as small 1,400,000 700,000 19.284105a
1,000,000 times as small 14,000,000 7,000,000 19.284200a
10,000,000 times as small 140,000,000 70,000,000 19.284211a
100,000,000 times as 1,400,000,000 700,000,000 19.284212a
small
a
To calculate these numbers I have used the following very accurate asymptotic approximation:
log[k!/(k1!k2!)] ¼ –[(1 – λ)log(1 – λ) + λlogλ]k – ½log[(1 – λ)λk] – ½log(2π) with λ ¼ λ1
Shannon calculation due to the circumstance that the number of samples is chosen
too small. The point is that Shannon’s theory is only applicable for Zdt sufficiently
large, i.e. in the limit Zdt should tend to 1. On the other hand Shannon’s formula is
compatible with the principle of evolutionary homogeneity, but the calculation of
16.827243 u.o.v. on behalf of (A.1) is not. The problem is however easy to manage.
We must choose the number of samples much larger and simultaneously adjust the
choice of the u.o.v. so that the outcome of Zdt H ðX0 Þ on behalf of (A.1) tends to the
outcome on behalf of Shannon’s formula (1.1). Thus all statistical experiments will
be so executed that the number Zdt of samples is increased by a sufficiently large
factor while simultaneously choosing the new unit of value inversely proportion-
ately smaller. The error between the outcome of the Shannon formula and the
outcome of (A.1) will vanish then. This is illustrated in Table A.1. In this table each
row reflects a particular choice of Zdt ¼ k and the u.o.v., starting with k ¼ 21 in the
second row. Each following row has 10 times as many samples and 10 times as
small units of value as the preceding row. E.g. in the third row of Table A.1 the
number Zdt ¼ k samples is fixed at k ¼ 10 21 ¼ 210 so that k1 ¼ 140 and k2 ¼ 70
with a new u.o.v. that is 10 times as small as the old one in the second row. A
calculation with the help of (A.1) yields Zdt H ðX0 Þ ¼ 188:74421 units in the new
u.o.v. This is 18.874421 units in the old unadjusted u.o.v. The error between the
Shannon measure (19.2842) and the correct outcome (18.874421) has declined
considerably. If the number of samples is multiplied once more by a factor 10
and the u.o.v. is reduced once more by a factor 10 we have k ¼ 2; 100, k1 ¼ 1; 400
and k2 ¼ 700 (See fourth row in Table A.1) and the new outcome is 19.226624 units
in the old unadjusted u.o.v. This is now very close to the Shannon measure 19.2842.
Appendix A 231
Note that the larger the number of samples and the proportionately smaller
the choice of the unit of value, the smaller the error in the calculation of the
reassembled information content. This illustrates that it is necessary to choose
the number of samples as large as possible together with the u.o.v. as small as
possible to attain evolutionary homogeneity.
There is of course still the question why economic selection should, by neces-
sity, take place against the background of a sufficiently (very) small u.o.v. and an
infinite sequence length.
I can think of two reasons for that:
• It is a necessity effectuated by the requirements of economic exchange. If the
unit of value is too large, exchange will be hampered and even obstructed,
because transactions between agents can then only be rounded off in a way
that injures one of the agents and benefits the other unnecessarily and against the
will of both. Hence noticing the exchangeability, fluidity and divisibility of
money, agents will seek and find ways to settle transactions in a smaller u.o.v.
• Agents will seek to reassemble as much entropy as they can. They will not leave
attainable knowledge (by selection) unexploited, thus maximizing the informa-
tion that they gather about input and output separately. As the data in the fourth
column of Table A.1 teach us: the smaller the unit of value, the larger the input
and output in terms of an invariable u.o.v. Hence agents will apply the smallest
unit of value in which to reassemble their samples.
Appendix B
The probabilities of input and output selection in sample space dS0 ¼ dS
0 [ dS0
based on the Shannon-inspired conception of Darwinian evolution are the
following:
• The probability λi that a sample annihilates in state i on (t,t + dt).
• The probability μj that a sample originates in state j on (t,t + dt).
• The probability λðij jÞ that a sample annihilates in state i on (t,t + dt) conditional
to another sample originating in state j on (t,t + dt).
• The probability μðjjiÞ that a sample originates in state j on (t,t + dt) conditional
to another sample expiring in state i on (t,t + dt).
• The unconditional joint probability qij that a sample annihilates in state i on
(t,t + dt) and a sample originates in state j on (t,t + dt).
The following general rules of probability calculus hold with respect to these
probabilities:
1. Any trial of a sample expiring in state i ði ¼ 1; 2, 3, ; N Þ and another
sample concurrently originating in state j ðj ¼ 1; 2, 3, ; N Þ will induce
the emergence/cancellation of entropy in overall sample space dS0 ¼ dS
0 [ dS0 .
See the discussion in Sect. 2.2. The event of expiration of a sample in state i is
statistically dependent on the event of origination of a sample in state j. This
þ
implies that the subsets of dS i and dSj overlap (See the Venn diagrams of
Figs. 2.2, 2.4 and 2.5) for any i; j ¼ 1; 2, 3, ; N.
2. The event of expiration of a sample in state i and the event of expiration of
another sample in state j are mutually independent. Likewise the event of
origination of a sample in state i and the event of origination of another sample
in state j are mutually independent. Such independent events are selected
sequentially, i.e. in the process of executing two successive singular trials, one
with the outcome in state i, the other in state j.
q10 þ q20 ¼ 1
q01 þ q02 ¼ 1
Mark that
q21 q12 ¼ μ1 λ1 ¼ λ2 μ2 q11
q21
q22 q11 ¼ μ2 λ1 ¼ λ2 μ1 q12
q22
Y0 X0
The Venn diagram of Fig. B.1 gives a geometrical insight into the various
relationships between the entropy variables and the probabilities for the two-sector
economy.
I have extended the range of i and j to include also the state 0, the union of all the
other states of the multi-sector economy. That is,
X X X X
μj ¼ q0j ¼ qij ; λi ¼ qi0 ¼ qij ; q00 ¼ qi0 ¼ q0j ¼ 1 (B.1)
i j i j
The probabilities λi and μj and the conditional probabilities λðij jÞ and μðjjiÞ satisfy
the necessary partition requirements (Papoulis 1985, Chap. 15):
Appendix B 235
X X X
λi ¼ λ0 ¼ 1; μj ¼ μ0 ¼ 1; λðij jÞ ¼ λð0j jÞ ¼ 1;
i j i
X
μðjjiÞ ¼ μð0jiÞ ¼ 1
j
qij qij
μðjjiÞ ¼ and λðij jÞ ¼ (B.3)
λi μj
and
X X
H ðX 0 j Y 0 Þ ¼ qij log qij þ qij logμj
i;j i;j
We have
X
H ðX 0 [ Y 0 Þ ¼ qij log qij (B.4)
i;j
and
X
H ðX0 jY 0 Þ ¼ H ðX0 [ Y 0 Þ þ q0j logμj
j
236 Appendix B
Herein H ðX0 Þ is the aggregate joint entropy input. H ðY 0 Þ is the aggregate joint
entropy output. It follows from the two expressions (B.5) and (B.6) that
It follows that
X X
H ðXi Þ ¼ H ðX0 Þ and H Y j ¼ H ðY 0 Þ
i j
Appendix B 237
Let us next consider the equalities following from the associated Venn diagram:
H X i [ Y j ¼ H ðX i Þ þ H Y j H X i \ Y j (B.8)
Herein H Xi [ Y j is the entropy of the union Xi [ Y j of Xi and Y j . H Xi \ Y j
is the entropy of the intersection Xi \ Y j of Xi and Y j . H Xi \ Y j is related to
H ðX0 \ Y 0 Þ by H Xi \ Y j ¼ qij H ðX0 \ Y 0 Þ. The easiest way to check that is with
the help of the Venn diagram of Fig. B.1. Mark that the probability of selecting both
a sample originating in state j and another sample annihilating in state P i is equal to
μj λðij jÞ ¼ λi μðjjiÞ. On behalf of (B.3) this is equal to qij. Note q00 ¼ i;j qij ¼ 1.
Hence
H Xi \ Y j
qij ¼ for i; j ¼ 0; 1, 2, 3 (B.9)
H ðX 0 \ Y 0 Þ
H ðX 0 \ Y 0 Þ ¼ H ðX 0 Þ þ H ðY 0 Þ H ðX 0 [ Y 0 Þ
Substitution of H ðXi Þ ¼ λi H ðX0 Þ and H Y j ¼ μj H ðY 0 Þ in (B.8) yields
H X i [ Y j ¼ λ i H ðX 0 Þ þ μ j H ðY 0 Þ H X i \ Y j
H Xi \ Y j can be replaced by qij H ðX0 \ Y 0 Þ on behalf of (B.9). It follows that
H Xi [ Y j ¼ λi H ðX0 Þ þ μj H ðY 0 Þ qij ½HðX0 Þ þ H ðY 0 Þ HðX0 [ Y 0 Þ
That general situation has been depicted in Figs. 2.2, 2.3, 2.4 and 2.5. In the
particular case that i ¼ j ¼ 0, we have also
dS þ
0 \ dS0 6¼
The surface area that a set occupies in the Venn diagram is proportional to the
entropy content of that set.
On behalf of Boltzmann’s principle it follows from (C.3) that for any two sets
SA and SB
HðSA [SB Þ HðSA Þ HðSB Þ HðSA \SB Þ
1 1 1 1
¼
2 2 2 2
PrfSA g: PrfSB g
PrfSA [ SB g ¼ (C.4)
PrfSA \ SB g
We use the notation ProbfεA \ εB g with the operator \ here, but we should
be aware that ProbfSA \ SB g with the same operator \ has a completely different
interpretation. In ProbfSA \ SB g we are concerned with the chance that a sample
is drawn from (or into) the common area of the subset SA and the subset SB . In
ProbfεA \ εB g we are concerned with the chance that both the event εA and the
event εB occur.
By definition two events εA and εB are statistically independent if their
elementary probabilities satisfy
Hence it follows from (C.5) that ProbfεA jεB g ¼ ProbfεA g and ProbfεB jεA g ¼
ProbfεB g are properties of two independent events εA and εB . If (C.6) is not satisfied
the events are statistically dependent.
From the above expositions it follows that the general concepts of mutual
exclusiveness of selection and statistical independence of selection differ markedly
in definitional respects. They should therefore not be confused with one another.
Appendix D
P
Derivation of H ¼ ωi logωi as the Measure of Information
i
Per Sample
It follows that
dFðgA Þ d ðgA Þ
ln g ¼ A
Fð g Þ g
and
dðA ln gÞ
d ln F gA ¼ ¼ d lnðA ln gÞ
A ln g
This results in
ln F gA ¼ lnðA ln gÞ þ ln η and F gA ¼ Aη ln g
Z0 ¼ Z1 [ Z2 and Z1 \ Z2 ¼
Let us first sum up the various variables and elementary events that we shall need
for the derivation of the probabilities of the sequences that may be formed as
outcomes of the selection process:
• k1 ¼ The number of bitpulses of Z 0 in state 1.
• k2 ¼ The number of bitpulses of Z 0 in state 2.
• k ¼ The total number k ¼ k1 þ k2 of bitpulses in state 0.
• Λi The event that the selected bitpulse is in state 1 on the i-th draw.
• Γ i The event that the selected bitpulse is in state 2 on the i-th draw.
• u The number of bitpulses in state 1 that have been drawn.
Let us start the experiment by drawing the first bitpulse of Z 0 . The probability
ProbfΛ1 g that the first bitpulse being drawn is in state 1 is
k1
ProbfΛ1 g ¼
k
and the probability ProbfΓ1 g that the first bitpulse being drawn is in state 2 is
k2 k1
ProbfΓ 1 g ¼ ¼1
k k
The interpretation is crucial here. When the first bitpulse is drawn, all of the k
bitpulses that are currently present in the population are engaged in the selection of
this bitpulse. In a world of many uncertainties this engagement is properly
expressed in the two expressions above for ProbfΛ1 g and ProbfΓ 1 g.
Next we consider the probability ProbfΛ1 \ Λ2 g that the first two bitpulses are
selected in state 1.1 This requires the second draw to be performed conditional to
the event that a bitpulse in state 1 was selected with the first draw. That is,
k1 1 k1 k1 1
ProbfΛ2 j Λ1 g ¼ and ProbfΛ1 \ Λ2 g ¼
k1 k k1
The interpretation is again crucial here. Since one of the bitpulses in state 1 has
been taken into account as having been selected on the first draw, it is removed from
1
Mark that ProbfΛ1 \ Λ2 g is the elementary probability that both the event Λ1 and the event Λ2
occur.
Appendix E 247
the current population and does not play a role anymore in the selection of a bitpulse
on the second trial and on successive draws thereafter.
In this manner we may proceed with the consecutive selection of bitpulses in
state 1 on the third, fourth etc. trial.
In the end we arrive at the probability ProbfΛ1 \ Λ2 \ Λ3 \ . . . \ Λu g of u
¼ k1 consecutive draws of bitpulses in state 1:
k1 ðk1 1Þðk1 2Þ . . . . . 3 2 1
ProbfΛ1 \ Λ2 \ Λ3 \ . . . \ Λu g ¼
kðk 1Þðk 2Þ . . . . . ðk2 þ 1Þ
k1 ! k2 !
¼ (D.1)
k!
Each time a bitpulse in state 1 is selected and taken into account, the number of
the remaining bitpulses in state 1 reduces by one and the probabilities of forthcom-
ing selections must be adjusted accordingly because, since the moment a bitpulse is
selected, it does not contribute to the forthcoming selection of bitpulses anymore.
After all u ¼ k1 draws of bitpulses in state 1 have been executed, all the k2
bitpulses in state 2 form the remainder population of Z 0 . From then on, the
continuation of the selection process has no secrets anymore.
Each time we select a bitpulse, we know it will be in state 2, i.e. it will be
selected with probability 1.
Hence, (D.1) gives the complete expression for Π ¼ ProbfΛ1 \ Λ2 \ Λ3 \ . . .
\Λu g:
k1 ! k2 !
Π¼
k!
Appendix F
We shall here discuss the crucial role of statistical dependence in evolution and its
relationship with the overlapping of input and output. To this end we shall prove
that both the transmission H ðX0 \ Y 0 Þ will vanish relative if the elementary events
of input and output are statistically independent. This implies that the overlapping
of input and output will vanish if there is statistical independence, but that
overlapping will be more the more statistical dependence between the events of
input and output there is. Moreover we will see that the complete absence of
overlapping, with H ðX0 \ Y 0 Þ ¼ 0, is the point where evolution begins or ends. It
need not imply that selection is totally absent at that turning point of evolution so
that always evolutionary development can return as a result of random processes of
selection. However if the transmission H ðX0 \ Y 0 Þ vanishes relative to input HðX0 Þ
and output HðY 0 Þ such that
k H ðX 0 \ Y 0 Þ
lim ¼ 0 with k representing the number of samples
k!1 log k
selection reduces to a determinist choice between only one single state and so
ultimately rules out selection to occur at all because we need at least two states to
select between. We are then leaving the free world of random occurrences to
replace it by a determinist world view.
For good order’s sake note that we are here concerned with the interaction
between the events of entropy origination in dSþ 0 , which are the output events,
and with the events of entropy annihilation in dS 0 , which are the input events.
Thus we are not considering the output events of dSþ0 relative to the inflow events of
dS dSþ . Neither do we consider the input events of dS relative to the outflow
0 0 0
events of dSþ dS .
0 0
To demonstrate the above assertions with respect to the input and output events
we must take full account of the asymptotic bias error in (1.1) and (2.7). These
Shannon equations are based on expression (4.4), which is the first order asymptotic
approximation of (4.4) for k ¼ Zdt ! 1. However a more accurate second order
asymptotic approximation of (4.3) takes also in account the terms that are propor-
tional to log k in (4.3), so acknowledging that also log k is tending to 1. This implies
that the Shannon equations (1.1) and (2.7) should be replaced by
X
k H ðX0 Þ ¼ k λi log λi 12ðN 1Þ log k (F.1)
i
X
k H ðY 0 Þ ¼ k μj log μi 12ðN 1Þ log k (F.2)
j
X
k H ðX0 [ Y 0 Þ ¼ k qij log qij 12 N 2 1 log k (F.3)
i;j
in order to make them applicable to the exceptional situations in which the influence
of log k plays a role.
Note that k ¼ Zdt ! 1 and log k ! 1 as well. Equations (F.1), (F.2) and (F.3)
are the second order asymptotic expressions for the contents of information of
þ þ
respectively the sample spaces dS
0 , dS0 and dS0 ¼ dS0 [ dS0 . They can be
justified in the same manner as the expressions (1.1) and (2.7) have been justified
to represent contents of information in Sect. 4.1. This is permissible because
ðlog kÞ=k ! 0 for k ! 1.
In accordance with these more accurate expressions for H ðX0 [ Y 0 Þ, H ðX0 Þ and
H ðY 0 Þ we have on behalf of (B.3) of Appendix B,
X X
k H ðX0 [ Y 0 Þ ¼ k qij log μðjjiÞ k qij log λi 12 N 2 1 log k
i;j i;j
(F.4)
Mark that generally μðjjiÞ μj and that equality of μðjjiÞ and μj is applicable if the
events of output and input are statistically independent. We have
X X
k qij log μðjjiÞ k qij log μj
i;j i;j
It follows that
X X
k H ðX0 [ Y 0 Þ k qij log μj k qij log λi 12 N 2 1 log k
i;j i;j
Appendix F 251
Hence
We can now immediately apply (B.8) and conclude that the left side can be
replaced by the negative transmission k H ðX0 \ Y 0 Þ . However, a point of
concern here is whether (B.8) is only a first order asymptotic approximation itself
and should not be replaced by a second order asymptotic expression of the kind of
equations (F.1), (F.2) and (F.3) to maintain the required precision. To rule out the
latter possibility we will rely on the following fundamental rule of probability
theory (See Appendix C):
H ðX 0 [ Y 0 Þ ¼ H ðX 0 Þ þ H ðY 0 Þ H ðX 0 \ Y 0 Þ (F.6)
Mark that the equality of left and right side of (F.7) applies only for the case that
þ
the input events of dS 0 and the output events of dS0 are statistically independent.
Thus we conclude from (F.1), (F.2) and (F.7) that
for statistical independence of input and output:
H ðX 0 \ Y 0 Þ ðN 1Þ2 log k
lim ¼ lim P ¼0 (F.8)
k!1 H ðX 0 Þ k!1 2k λi log λi ðN 1Þ log k
i
252 Appendix F
and
This completes the demonstration that H ðX0 \ Y 0 Þ as well as H Xi \ Y j will
vanish relative to H ðX0 Þ and H ðY 0 Þ if the elementary events of input and output are
the vanishing of H ðX0 \ Y 0 Þ must
statistically independent. Note
þ
also imply the
vanishing of the entropy H Xi \ Y j of the subsets of dS 0 \ dS 0 .
On the other hand (F.7) indicates that H ðX0 \ Y 0 Þ need not vanish if the input and
output events are statistically dependent. P We conclude P from (F.4) that
k H ðX0 \ Y 0 Þ will then be equal to k i;j qij log μj þ k i;j qij log μðjjiÞ ,
which is proportional to k. Hence H ðX0 \ Y 0 Þ will seize a positive finite portion of
input and output of the same order as input and output seize in case there is
statistical dependence between the events of input and output. Statistical depen-
dence of the input and output events thus implies that there is overlapping between
input and output and this marks annihilation and origination of entropy to have
impact on each other. This is the evolutionary world of interaction and actual
exchange of information.
Can H ðX0 \ Y 0 Þ 6¼ 0 attend with statististical independence of input and output?
Definitely not because, if there is statistical independence, H ðX0 \ Y 0 Þ must satisfy
the limiting behavior stipulated by expressions (F.8) and (F.9). Overlapping and
statistical dependence must go hand in hand.
What if HðX0 \ Y 0 Þ ¼ 0 ? Does that imply that input and output must be
statistically independent? Indeed it does. The applicable proposition is the
following:
It follows that all the logarithmic factors in the sum of equation (F.12) must share
the same sign. However, since also qij 0, that sum can only equal 0 if all the terms
of the sum vanish together. That is,
" #
λðij jÞ μðjjiÞ
log ¼0 for all i and j
qij
λi ¼ λðij jÞ
By the same token, with the exclusion of the degenerate solution qij ¼ 0,
μj ¼ μðjjiÞ
Consequently,
254 Appendix F
dS0 ¼ dSþ
0 dS0 [ dS0 dS0
þ
This implies that, in compliance with that selection condition, dS 0 \ dS0 is the
þ
null-set and that H dS0 \ dS0 ¼ 0, which in turn implies statistical independence
between the events of output and the events of input on behalf of (F.10). We may
thus treat an experiment of selection restricted to these two conditional domains as
one in which the events of origination of bitpulses are statistically independent of
the events of annihilation of bitpulses.
þ
However there is even more involved in the event that H dS 0 \ dS0 ¼ 0. It is
demonstrated in Appendix G that H dS0 \ dSþ 0 ¼ 0 implies also that for the two-
sector economy both consumption probability μ and labor inputprobability þ λ must
get to equal 1. Moreover it is shown in Appendix G that H dS ¼ H dS0 ¼ 0 if
þ
0
H dS0 \ dS0 ¼ 0. These coincidences mark the begin and end of evolution. That
is ðλ ¼ μ ¼ 1Þ is the point in the ðλ; μÞ-plane where evolution takes off and where it
ends in the case of the two-sector economy.
Another consequence of (F.7) is that the number N of states between which is
selected must equal 1 if k H ðX0 \ Y 0 Þ vanishes entirely and if it does not even keep
up with log k such that
k H ðX0 \ Y 0 Þ
lim ¼0
k!1 log k
In that particular case it must follow from (F.7) that N ¼ 1. Selection is reduced
to a determinist choice for only one single state ruling out selection to occur at all
because we need at least two states to select between. This implies that evolutionary
selection and hence evolution itself is impossible because there is nothing to select
if N ¼ 1. It is the borderline where we exchange our worldview of free random
choice for a determinist worldview.
Appendix G
q11 log q11 q22 log q22 þ q12 log q12 þ q21 log q21
H ðX 0 \ Y 0 Þ ¼ (4.14)
2 12 q11 q22
With (F.10) it follows that the numerator of the right side is a function of λ and μ
only. Let us then use the shorthandnotation R0 ðλ;μÞ for that numerator. Mark also
that on behalf of (F.10) the factor 12 q11 q22 in the denominator of the right
side of this expression equals
ð2λ 1Þ ð2μ 1Þ
1
2 q11 q22 ¼ 12 λμ ð1 λÞð1 μÞ ¼
2
R0 ðλ; μÞ
H ðX 0 \ Y 0 Þ ¼ ¼0
ð2λ 1Þ ð2μ 1Þ
Further we conclude that the denominator of the term in the center can only
1
vanish (with λ ¼ or with μ ¼ 1 0
2) if the numerator vanishes, i.e. if R ðλ; μÞ ¼ 0. We
2
will soon see that R0 ðλ; μÞ cannot vanish if λ ¼ 12 or with μ ¼ 12.
With (F.10) R0 ðλ; μÞ can be restated as a function of λ and μ. That is,
0.9
0.8
R(ω) 0.7
0.6
0.5
0.442695
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
R0 ðλ; μÞ ¼μð1 2λÞ logðμÞ þ λð1 2μÞ logðλÞ þ ð2λ 1Þð1 μÞ logð1 μÞ
þ ð2μ 1Þð1 λÞ logð1 λÞ ¼ 0
This results in
μ logðμÞ ð1 μÞ logð1 μÞ
H ðX 0 \ Y 0 Þ ¼
2μ 1
λ logðλÞ ð1 λÞ logð1 λÞ
þ ¼0
2λ 1
and
dRðωÞ 2
lim ¼ þ lim logð1 ωÞ ¼ 1
ω"1 dω ln 2 ω"1
and
d 2 R ð ωÞ 8 1
lim ¼ lim lim 4 logð1 ωÞ ¼ 1
ω"1 dω2 ln 2 ω"1 ωð1 ωÞ ω"1
dH ðX0 \ Y 0 Þ dH ðX0 \ Y 0 Þ
lim ¼ lim ¼ 1
λ"1 dλ μ"1 dμ
and
d 2 H ðX 0 \ Y 0 Þ d 2 H ðX 0 \ Y 0 Þ
lim ¼ lim ¼ 1
λ"1 dλ 2 μ"1 dμ2
This implies that in the turning point μ ¼ λ ¼ 1 of evolution any small first
and second order perturbation of λ and or μ will be amplified in a manifold change
of HðX0 \ Y 0 Þ. This amplification of perturbations of λ and μ in the change of
H ðX0 \ Y 0 Þ at the turning point of evolution appears to go on for third order and
even n-th order perturbations. At the incipience point of evolution there is extreme
sensitiveness to boost development, as well as extreme sensitiveness to smother it.
Any perturbation dλ dμ > 0 may be the beginning of a new evolutionary
development. Any perturbation dλ dμ < 0 may be the end of evolutionary
development.
258 Appendix G
We shall consider the following four different growth positions that an economy
can be in:
1. δ ¼ ðλ μÞ=2 > 0, λ is raised relatively more than μ, i.e. λ_ =λ > μ_ =μ
2. δ ¼ ðλ μÞ=2 > 0, λ is raised relatively less than μ, i.e. λ_ =λ < μ_ =μ
3. δ ¼ ðλ μÞ=2 < 0, λ is raised relatively more than μ, i.e. λ_ =λ > μ_ =μ
4. δ ¼ ðλ μÞ=2 < 0, λ is raised relatively less than μ, i.e. λ_ =λ < μ_ =μ
Position 1:
It follows from λ_ =λ > μ_ =μ that μ λ_ =λ > μ_ and
μλ_ μ_ λ_ δ_ λ_
μ_ λ_ < λ and ¼ >
λ μλ δ λ
λ_ μ_ δ_ λ_ μ_
¼ > >
λμ δ λ μ
H ðX 1 Þ H ðY 1 Þ
>0
H ðX 0 Þ H ðY 0 Þ
Position 2:
It follows from λ_ =λ < μ_ =μ that μ λ_ =λ < μ_ and
μλ_ μ_ λ_ δ_ λ_
μ_ λ_ > λ and ¼ <
λ μλ δ λ
λ_ μ_ δ_ λ_ μ_
¼ < <
λμ δ λ μ
λμ λ_ μ_ δ_ μ_
λ_ μ_ > μ_ and ¼ <
μ λμ δ μ
λ_ μ_ δ_ μ_ λ_
¼ < <
λμ δ μ λ
Appendix H 261
Note that in this particular position our starting position of analysis is a recessive
economy with δ being negative. Our policy aim is to stop negative growth. This
demands δ_ =δ to become positive.
λ_ < 0 will effectuate δ to decrease and μ_ < 0 will limit the possibility to increase δ
sufficiently. An even less economic growth rate is then likely to ensue. Hence only
the combined stimulation of consumption with μ_ > 0 and wages with λ_ > 0 will
create the circumstances to avoid this.
Furthermore with μ_ < λ, _ δ will keep increasing and eventually this will bring
about that λ and μ will balance again and that δ ¼ ðλ μÞ=2 gets zero so that
eventually positive growth can be regained.
However we should keep in mind that with restoring the balance of λ and μ the
starting position of analysis has shifted to that of another position.
Position 4:
It follows from λ_ =λ < μ_ =μ that λ_ < λμ_ =μ and
λμ λ_ μ_ δ_ μ_
λ_ μ_ < μ_ and ¼ >
μ λμ δ μ
λ_ μ_ δ_ μ_ λ_
¼ > >
λμ δ μ λ
A bit is the unit in which entropy is expressed. A bitpulse is the unit of selection
of entropy. The money unit is the unit of exchange.
The capital letter S is reserved to denote a non-differential set, i.e. a set holding a
stock of entropy. Interactor is a synonym for a non-differential set. dS is reserved
to denote a differential set, a subset holding the entropy that annihilates on an
infinitesimally small time-interval ðt; t þ dtÞ of selection. dSþ is reserved to denote a
differential set, a subset holding the entropy that originates on an infinitesimally
small time-interval ðt; t þ dtÞ of selection.
The numeral (or variable) subscripts i and j represent the state of selection of a
variable. They refer to the i-th (respectively j-th) subset or to a variable of that
subset. For the multi-sector economy (or N- sector economy) the total number of
states is N. For the two-sector economy ðN ¼ 2Þ the convention is:
Superscripts
The union of several sets holds all the entropy that resides in all of these sets. E.g.
S0 is the union of S1 and S2 for the two-sector economy. Notation: S0 ¼ S1 [ S2 .
266 Appendix I
Further dS
0 is the union of dS1 and dS2 for the two-sector economy. Notation:
dS0 ¼ dS1 [ dS2 . Also dS0 is the union of dSþ
þ þ
1 and dS2 for the two-sector economy.
þ þ þ
Notation: dS0 ¼ dS1 [ dS2 . Note that overall differential sample space dS0 is the
þ þ
union of dS
0 and dS0 : dS0 ¼ dS0 [ dS0
Furthermore, an intersection of two sets is the set of entropy that the two sets
þ þ
hold in common. Notation: dS
i \ dSj is the intersection of dSi and dSj . Only
intersections of differential sets will be considered. Intersections of non-differential
sets do not play a role in evolvodynamics except in cases of compartmentalization.
þ þ þ
dS
i dSj is the subset of dSi that it has not in common with dSj . Likewise dSi dSj
H ðÞ is the entropy operator. H dS i returns the entropy content per sample of
þ
the differential set dSi . H dSi returns the entropy content per sample of the
differential set dSþ
i . H dSi \ dSj
þ
returns the entropy content per sample of the
þ þ þ
intersection dS
i \ dSj . H dSi \ dSj is called the transmission of dS
i and dSj .
þ þ
H dS i [ dSj returns the entropy content per sample of dS i [ dSj .
þ
H dS i dSj returns the entropy content per sample of the conditional subset
þ
dSi dSj .
As regards non-differential sets, H ðSÞ returns the total entropy content of the set
S. HðSÞ is called the capacity of S. It is usual to denote capacity by the symbol C
(only applicable for the entropy of non-differential sets). Thus Ci ¼ H ðSi Þ.
C is the notation of the capital of S (only applicable for non-differential sets).
Bi is the liquidity surplus of the sector Si .
The H symbol will sometimes also be used as a general variable of entropy.
Appendix I 267
P is the unit price of a bit of entropy. It is the number of money units that goes
in a bit. P is time-dependent. P is also called the price-level.
u.o.v. is the abbreviation to denote an arbitrary unit of value.
Inflow is the entropy content of a differential set in which entropy annihilates.
Outflow is the entropy content of a differential set in which entropy originates.
The symbols hX0 i and hY 0 i are reserved to denote sequences of entropy input,
respectively sequences of entropy output. The inflow Xi denotes entropy input, the
content of dSi . The inflow Xi denotes entropy influx, the entropy content
entropy
þ
of
dSi dS0 . The outflow Y j denotes entropy output, the entropy content of dSþ j . The
outflow Yj denotes entropy outflux, the entropy content of dSþ
j dS0 .
Zdt H dS i \ dSþ is financial input, the entropy content of dS þ
i \ dS0 .
0
Zdt H dS0 \ dSþ i is financial output, the entropy content of dS þ
0 \ dSi .
A sequence is the result of the execution of a differential selection experiment. It
is an ordered random string of Zdt selected bitpulses of varying states. The string is
of infinitesimally small time-length dt. Input sequences reside in dS 0 and output
sequences reside in dSþ 0 .
formed, also the ones that will never be selected as a result of executing a
differential selection experiment. A typical or purposeful variation is a variation
that has a chance to be selected. One out of all typical variations will be selected
with probability 1. An atypical or purposeless variation is a sequence that will
never be selected as its chance of selection vanishes relative to the infinitesimally
much greater chance of selection of a typical variation.
H dS
i H ðX i Þ H ðX i j Y 0 Þ X i Xi
λi ¼ Prob dS
i ¼ ¼ ¼ ¼ ¼
H dS0 H ðX0 Þ H ðX0 jY 0 Þ X0 X0
H dSþ
i H ðY i Þ H ðY i jX0 Þ Y i Yi
μi ¼ Prob dSþ
i ¼ þ ¼ ¼ ¼ ¼
H dS0 H ðY 0 Þ H ðY 0 j X 0 Þ Y 0 Y 0
H Xi \ Y j
qij ¼
H ðX 0 \ Y 0 Þ
The following convention is in force for the growth rates of average price-levels
and the growth rates of the various flow- and stock-variables: Let F (capital Roman
letter) be an economic variable, then its lower case letter f is defined as the growth
rate of F:
dF=dt
f ¼
F
Thus, the growth rate ci of sector capacity Ci, the growth rate yj of entropy output
Y j, the growth rate yj of entropy outflux Yj and the inflation rate p of price-level P are:
ρi is the (entropy) input circulation rate. This is the circulation rate of entropy within
dS i .
ζ i is the (entropy) output circulation rate. This is the circulation rate of entropy
within dSþ i .
ρi is the(entropy) influx circulation rate. This is the circulation rate of entropy
within dS þ
i dS0 .
ζ i is the (entropy) outflux circulation rate. This is the circulation rate of entropy
within dSþ
i dS0 .
γ is the transmission rate.
Work-unit is the effort delivered by an actively engaged unit of the work-force
per unit of time.
Li is the work-force, the number of work-units actively delivering work at a
particular time.
Li dt is work, the work-time delivered by the work-force of Si during ðt; t þ dtÞ.
W 1 is the average money wage per work-unit of L1 .
E is effectiveness, the productivity of delivering work. e ¼ ðdE=dtÞ=E is the rate
of growth of effectiveness.
270 Appendix I
Crowe MJ (1990) Theories of the world from antiquity to the Copernican revolution. Dover
Publications, Mineola, NY
Darwin CR (1859) The origin of species by means of natural selection or the preservation of
favoured races in the struggle for life. Reissued 1968, Penguin Books, Harmondsworth,
Middlesex, UK
Dawkins R (1976) The selfish gene. Oxford University Press, Oxford, UK
Dawkins R (1984) The extended phenotype, the long reach of the gene. Oxford University Press,
Oxford, UK
Dawkins R (1988) The blind watchmaker. Penguin Books, London
Dawkins R (1995) River out of Eden, a Darwinian view of life. BasicBooks HarperCollins, New
York
Drake S (1980) Galileo. Oxford University Press, Oxford
Dreyer JLE (1953) A history of astronomy from Thales to Kepler, revised with a foreword by
W.H. Stahl. Dover Publications, New York
Einstein, Lorenz, Weyl, Minkowski (1952) The principle of relativity, a collection of original
memoirs on the special and general theory of relativity. Dover Publications, New York
Galileo G (1954) Dialogues concerning two new sciences. Translated by Henry Crew and Alfonso
de Salvio. Dover Publications, New York
Hodgson GM (1993) Economics and evolution – bringing life back into economics. Polity Press,
Cambridge, UK
Hodgson GM (2004) The evolution of institutional economics – agency, structure and Darwinism
in American institutionalism. Routledge, London
Hodgson GM, Knudsen T (2006) Why we need a generalized Darwinism, and why generalized
Darwinism is not enough. J Econ Behav Org 61:1–19
Hodgson GM, Knudsen T (2007) Information, complexity and generative replication. Published
online, Springer Science + Business Media B.V.
Hodgson GM, Knudsen T (2010) Darwin’s Conjecture, the search for general principles of social
and economic evolution. The University of Chicago Press, Chicago
Jaynes ET (1991) How should we use entropy in economics. St. John’s College, Cambridge CB2
1TP, England, published on the Internet
Jaynes ET (2004) (edited by G.L. Bretthorst) Probability theory: the logic of science. Cambridge
University Press, Cambridge, UK
Keuzenkamp HA (2004) Probability, econometrics and truth. The methodology of econometrics.
Cambridge University Press, Cambridge
Keynes JM (1936) The general theory of employment, interest and money. MacMillan, London
Monod J (1971) Chance and necessity: an essay on the natural philosophy of modern biology.
Alfred A. Knopf, New York
Papoulis A (1985) Probability, random variables, and stochastic processes, 2nd edn. McGraw-Hill,
New York
Shannon CE (1948) A mathematical theory of communication. Bell System Technical Journal 27,
pp. 279–423 and 623–656
Shannon CE and Weaver W (1963) The Mathematical Theory of Communication, reissued by The
University of Illinois Press, Urbana, USA
Wallast LH (2009a) Value, probability and economic growth. SSRN 2009: (revised in 2012) http://
ssrn.com/abstract=1427825
Wallast LH (2009b) Dealing with the economic recession. SSRN 2009: (revised in 2012) http://
ssrn.com/abstract=1427813
Wallast LH (2009c) The understanding of time. SSRN 2009, http://ssrn.com/abstract=1427780
Index
A Combinations, 69, 70
Absence of money illusion, 209 Compartmentalization, 189, 194
Agents defined, 46 Concurrent events of annihilation and
Annihilation, 31, 34 origination, 31, 49
Assembly of selected entropy, 19, 30, 58, 68 Concurrent selection. See Overall selection
Average future bit price. See Average future Conditional entropy inflow, 39, 91
unit price Conditional entropy outflow, 39, 91
Average future unit price, 117, 140–142 Consistency of facts, 223
Average historic bit price. See Average historic Consistency of laws, 223
unit price Consumer price index, 192
Average historic unit price, 116, 133, 135 Consumers, 46
Consumption. See Labor output probability
Consumption function, 211, 212
B Consumption probability, 211, 213. See also
Bit (binary digit), 20, 30, 33, 43, 78 Labor output probability
Bit of entropy. See Bit (binary digit) Consumption sector, 30
Bitpulse, 29, 55, 113 Consumption stimulation, 213
Blind selection, 91, 92, 106 Copernican system, 225
Boltzmann, Ludwig, 2, 11 Copernicus, 225
Boltzmann’s H-theorem (see Boltzmann, Creation. See Origination
Ludwig) Current lifetime of bitpulse, 55, 113, 119
Boltzmann’s principle, 2, 43
Boltzmann’s theorem, 58 (see also
Boltzmann, Ludwig) D
Brahe, Tycho, 225 Darwin, 5–7, 10, 16
Budget index factors, 189 Darwinian principles of evolutionary selection,
3–6, 10, 44, 221
Dawkins, Richard, 94, 95
C Deflation, 214
Capacity, 30, 32, 63 Deflationary mode of evolution, 118, 119, 139,
Capacity, net growth, 27 145, 151, 152
Capital, 63, 64 Degenerative selection. See Inflow selection
Capital, net growth, 63 Degrees of evolutionary freedom, 110
Case of average price deflation. See Differential sample space, 40
Deflationary mode of evolution Differential selection, 80
Case of average price inflation. See Inflationary Differential sets of entropy, 30, 31
mode of evolution Divisia’s index formulas, 187, 190, 192, 193
O
L Origination, 34, 49
Labor capacity, 25, 33 Outflow, 10, 13
Laborer, 46 Outflow selection, 18
Labor financial input, 159 Outflow sequence, 12, 66
Labor financial output, 160 Outflux, 115
Labor force, 182, 184 Outflux (inflationary mode), 148
Labor inflow, 25 Outflux-bit handling rate, 139
Labor influx, 160 Outflux bitpulse, 115
Labor input probability, 73, 176, 212 Outflux circulation rate, 122, 123
Labor outflow, 25 Output, 115
Labor output probability, 73, 155 Output (inflationary mode), 148
Lifetime, 98, 101, 103, 105 Output-bit handling rate, 133
Lifetime of bitpulse, 47, 51, 54, 66 Output bitpulse, 115
Liquidity flux, 63, 175, 220 Output circulation rate, 126
Overall capacity, 33
Overall differential sample space, 34
M Overall selection, 35, 100
Markov property, 130, 131 Overlapping, 33–35, 49
Method of science, 4, 7, 9, 80, 81 Overlapping sets, 34
Methodology of science, 223
Monetary contraction, 221
Monetary expansion, 220 P
Money Price of one bit. See Unit price of entropy
money flow, 148 Probability distribution of current lifetime,
money influx, 125, 145 120, 125
money input, 129 of outflux bitpulses, 115, 120
money outflux, 123, 148 of output bitpulses, 115, 125
278 Index