J Evol Econ (2011) 21:341–366
DOI 10.1007/s00191-010-0178-0
REGULAR ARTICLE
Evolutionary developmental economics: how
to generalize Darwinism fruitfully to help
comprehend economic change
Pavel Pelikan
Published online: 17 June 2010
© Springer-Verlag 2010
Abstract Darwinism is shown possible to generalize fruitfully to help comprehend
economic change by drawing on evolutionary developmental biology
(“evo–devo”)—its recent version, less concerned with replication of genes
than with genomic instructing of development of organisms. The result is a
conceptual model with multilevel applications, generalizing development as
instructed self-organizing with inputs from environments, and evolution as
experimental search for instructions making the development successful. Its
economic interpretation suggests to unite several existing fields into evolutionary
developmental economics, where economic change can be studied comprehensively
as development instructed by actual institutional rules, intertwined
with the evolution of these rules.
Keywords Instructed development · Evolution of instructions · Multilevel
evolution · Evolution of institutional rules · Development of economies
JEL Classification A10 · D02 · K10 · O10 · P50 · Z10
1 Introduction
There is a broad agreement that no version of Darwinian evolutionary biology
is directly applicable to the evolution of human economies and societies.
But evolutionary economists still disagree on whether or not some generalization
of some version of Darwinism might nevertheless be helpful. These
P. Pelikan (B)
Department of Institutional Economics, Prague University of Economics,
Prague, Czech Republic
e-mail: pelikanp@vse.cz
342 P. Pelikan
disagreements perhaps most clearly appear in the recent conflict between the
generalization of Darwinism elaborated by Hodgson (2002, 2004), Knudsen
(2004), and Hodgson and Knudsen (2006a), and the Continuity Hypothesis
advocated by Witt (2003, 2004) and Cordes (2006, 2007)—let me refer to the
two as HKGD and CH, respectively. While HKGD extracts from biological
Darwinism certain notions and principles that its authors claim also to apply to
socioeconomic evolution, CH considers Darwinism able to explain the origins
of the human cognitive abilities on which this evolution builds, but insists that
Darwinism has nothing to do with this evolution itself.
This paper can roughly be situated by its agreements and disagreements
with these two sides. It agrees with HKGD that biological evolution and
socioeconomic evolution do follow certain common principles that can be
regarded as Darwinian, but disagrees with it about what these principles are.
And it agrees with CH that socioeconomic evolution is a continuation of the
biological one, and that it is indeed not Darwinian in the sense of HKGD, but
shows that it does follow certain principles that can be considered Darwinian,
provided Darwinism is generalized more suitably than by HKGD.
Rather than criticizing existing generalizations of Darwinism, however, the
main purpose of this paper is to suggest one of its own, and illustrate how
useful to economists and other social scientists this might be. Its main novelty
is to generalize one of the latest versions of Darwinism–the evolutionary
developmental biology (“evo–devo”) following Carroll (2005) and Carroll et
al. (2008). Compared to the standard neo-Darwinian synthesis, the basis of
previous generalizations, this version extends attention in three directions:
(1) from genes to entire genomes, (2) from how genomes replicate to how they
instruct, with uses of inputs from environments, the forming and development
of organisms; and (3) from fully developed organisms to the entire developmental
process.
All of this makes Darwinism easier to generalize for the needs of social
scientists. Extension (1) makes it unnecessary to search for an exact socioeconomic
correspondent to a single gene. This is a great relief, as no such search
has been very successful, and one may doubt that it might ever be. Instead,
evolutionary social scientists only need to realize the prime importance of genomic
instructions, but may think of them in broader terms than those of genes
coding for proteins—now known not to be the only important parts of genomes
(see e.g. Mattick 2004).
Extension (2) allows principles of Darwinism to apply to socioeconomic
change without replicating. The change must only use some evolving instructions
that guide, at each stage of their evolution, some developmental process.
This also makes it clear that the main products of evolution are instructions,
whereas replicating is only one special way of their storing over time—
obligatory in biological evolution, where the storing must rely on short-lived
organisms. For evolutionary social scientists, this is another relief, as much of
socioeconomic change takes place without replicating. More often than not,
Evolutionary developmental economics 343
economies and societies form, develop, evolve, further develop, evolve, or
possibly fall, as childless singles.1
Extension (iii) makes it clear that development is an inseparable but distinct
companion of evolution, with plenty of room for all the phenomena that are
undeniably part of the forming and functioning of organisms, yet difficult to
accommodate within evolution itself–such as cooperation, altruism, and selforganization.
In economics, as shown below, the logic of this extension may
help to interconnect evolutionary, institutional, and development economics
into a single well-ordered field of evo–devo economics, where all the important
parts of economic change have their well-defined and clearly interrelated
places.
All three extensions converge on an important point: the central role of
instructions. These form the link between evolution and development, and
provide the main reference for distinguishing the two. In a first approximation,
the distinction may roughly be put as follows: Evolution produces instructions
for guiding development.
Attention to instructions also leads to a new promising direction for analysis,
starting with the logically obvious, but so far little exploited principle: all uses
of instructions require pre-existing instructions. It is from this principle that an
interesting constraint on what evolution can, and what it cannot, achieve will
be deduced.
Emphatically, however, the central role of genomic and other instructions
has nothing to do with genetic determinism. In addition to involving also other
parts of genomes than genes, it fully admits the importance of inputs from
environments. It will only lead to the conclusion, deduced from the basic logic
of information processing, that genomic instructions are primary: the inputs
must initially be prepared and are ultimately limited by them. Thus, without
determining what exact features an organism will actually develop, its genome
will nevertheless be found to set strict limits to what these features might, in
the most ideal environments, possibly become. In other words, each genome
will be seen to imply a certain developmental potential that environments can
more or less actualize, but never expand. But this may perhaps be described as
“genomic limitism,” not “genetic determinism.”
1The emphasis on replicating is an important reason why HKGD is of so little use in the
social sciences. In this respect, Campbell’s (1965) early generalization of Darwinism in terms of
“variation, selection, and retention” is wiser: the biology-loaded notions of “inheritance” and
“replication” are there carefully avoided. This makes it hard to understand why Hodgson and
Knudsen insist on building on these notions, and even misquote Campbell as if he also used them.
Hodgson (2004), for instance, contains two such misquotes: “variation, inheritance, and selection”
and “replication, selection, and variety-creation.” Both moreover ignore Campbell’s sound logic
with which he puts “selection” before “retention”—as, logically, whatever is retained must have
first been selected—and put instead “replication” and “inheritance” before “selection.”
344 P. Pelikan
To be as concise as possible, the present generalization of Darwinism will
be given the form of a conceptual agent-based model. This will provide it
with a solid micro-basis and establish a well-defined link to methodological
individualism in the social sciences. That HKGD has no clear micro-basis
appears to be another reason, in addition to its misplaced focus on replicating,
why it is so short of meaningful socioeconomic applications.
Concerning such applications, the present generalization has an inbuilt
advantage. In contrast to HKGD, which stems from the highly abstract spheres
of ontology and philosophy of biology, this one was taking shape during
inquiries into the more mundane issues of comparative economics, economic
reforms, and transformation policies (Pelikan 1988, 1992, 2003a). While it is
thus guaranteed applicable at least to these issues, this paper will illustrate that
its economic applicability is much broader.
Emphatically, generalizing Darwinism is not claimed to be the only way to
understand economic change. That this understanding may also be sought with
purely economic tools is fully admitted. The choice is free, depending on what
one hopes to find with these tools, and how much interested one is to learn
new things from other disciplines. This paper is addressed to those economists
who are interested in such learning.
The rest of the paper is organized as follows. The general agent-based
model is described in Part 2, its possible economic applications, including a
rough outline of the new field of evolutionary developmental economics, are
discussed in Part 3, and its possible usefulness for other social sciences is briefly
noted in a concluding comment.
2 The origins of successful complex agents: a general evo–devo model
2.1 The overall plan
The main task of the model is to depict the processes by which given basic
agents (b-agents) in given environments can form, develop and operate a
complex agent (C-agent) successful in these environments. While dealing with
only one level of such processes, the model is recursively applicable to many
of their levels, both biological and socioeconomic—including cells that form
and operate organisms, and individuals who form and operate economies and
societies.
With one important exception, what needs to be kept in mind about the
agents and the environments may be summarized as follows:
The given b-agents may be heterogeneous, of different abilities and
behaviors.
The environments of each b-agent include parts of the common environments
and some of the other agents.
Each C-agent is a network of b-agents, which play there possibly different
roles, and thus interact with each other and with the environments in possibly
different ways.
Evolutionary developmental economics 345
The environments test the performance (“fitness”) of the C-agents according
to non-specified success criteria. These may, but need not, be those of natural
selection, and the environments may, but need not, include competition
and/or cooperation with other agents.
The b-agents’ behaviors are of two dimensions: (1) associative, which they
use to self-organize into, and find their roles within, a certain network, and
thereby form their C-agent; and (2) operational, which they use to operate
in the roles found, and thus enable the C-agent to operate and perform. The
network may also aggregate their associative behaviors, and thus enable the
entire C-agent to behave associatively in relation to other agents.2
An agent’s behaviors may be more or less goal-oriented (purposeful, intentional),
more or less responsive to inputs from the environments and other
agents, and more or less learning from these inputs and adapting to them.
That the model is reductionist, but not naively so, deserves emphasis. No Cagent
is reduced to a “simple sum” of its b-agents. What a C-agent is and does
also depends on the network into which its b-agents have self-organized. It is
thanks to this network that the C-agent’s behaviors may be more complex than
the behaviors of any of them. For instance, the C-agent may pursue objectives,
conduct trial-and-error searches, learn, and develop self-awareness—while
none of its b-agents may be able to do any of that.
Moreover, the network may in return influence the b-agents’ behaviors—
although, emphatically, only within the limits of their intrinsic abilities to allow
it. It is only that no C-agent is allowed to fall as a holistic mystery from the
sky, but each is required to originate with some b-agents forming its initial
network. Only then can the network start to influence, under the constraint of
their intrinsic malleability, the b-agents’ behaviors, while they may respond by
helping the network, some possibly more than others, develop and evolve.
2.2 Instructions and the ultimate constraint on their uses
The important exception that needs more than a brief mention in a summary,
is that all behaviors of all agents are guided by instructions, arranged into
more or less complex programs. This also means that all features of an agent’s
behaviors—including the above-mentioned intentionality, responsiveness to
inputs, and learning—need corresponding instructions for their realization.
2Distinguishing the two dimensions of agents’ behaviors, and considering both, appears to be the
only way to clarity about the link between self-organization and Darwinism. That most authors
deal with only one of these dimensions, while leaving aside the other, may well be why this link
has not yet been fully clarified. For instance, most economists limit attention to how economic
agents operate within an already organized economy, while most of the authors studying selforganizing—such
as Camazine et al. (2001), or Doursat (2008)—limit attention to the architecture
of the networks that given self-organizing agents will form, without studying how these agents and
the entire network will then operate.
346 P. Pelikan
With this link between agents and instructions, it is possible to make the
above distinction between evolution and development more precise: Evolution
produces instructions for the development of a C-agent by guiding the selforganizing
and operating of its b-agents.
But to get it right, the term “instruction” must be understood more broadly
than in computer programming. An instruction may prescribe a certain operation,
but it may also mean a constraint (“negative rule,” “rule-of-thegame”)
that only specifies a more or less broad variety of permissible actions.
Instructions-constraints may thus be combined with each other by logical
intersection, each excluding a more or less different subset of an agent’s
feasible actions, and then combined with instructions-for-operations, which
lead the agent to take a specific action within the remaining subset.
Illustrations can be found in games and chemistry. In a game, both its general
rules and the specific ways of playing it are seen to consist of instructions.
In chemical reactions, instructions are seen in the valences that guide atoms to
form molecules with certain atoms, and not others.
It is also important to realize that programmed behaviors need not be mechanistic,
purposeless, deterministic and uncreative—as many social scientists
used to believe. For this, it suffices to consider advances in two disciplines:
in computer programming, which can make computers behave purposefully,
search, learn, innovate and be “aware” of their activities; and in neurophysiology,
which makes it increasingly clear that all human thinking and behaviors
are governed by what may be understood as programs in human brains.
An important kind of instructions that are commonly used in computer
programming and appear widespread in brains are those for generating and
using what may be called “random signals.” These need not be random in the
absolute sense that they would lack definite causes, but only in the sense that
they are uncorrelated with the problems to be solved. They are important for
avoiding deadlocks in decision-making with incomplete information, and for
conducting trial-and-error searches, both in brains and in nature.
The main result of studying the agents’ instructions is the finding of an
important ultimate constraint. This follows from the principle that all uses
of instructions require pre-existing instructions, applied to the fact that an
agent’s actual instructions can only be of three origins: (1) initial endowment;
(2) received from environments; (3) elaborated by own learning. The principle
implies that acquiring new instructions—be it from environments or by own
learning—also requires pre-existing instructions. While these may also be
partly received and/or learned, also this required pre-existing instructions. The
constraint, easy to deduce by simple recursive reasoning, is: For each agent,
the development of its actual instructions must start with, and is ultimately
constrained by, its initial instructions.
A minor complication is that “initial” is relative. In part, it is relative to the
agent considered. Thus, many initial instructions of a C-agent may be younger
than those of its b-agents. The b-agents may have obtained them later, for the
specific forming of that C-agent, with the help and under the constraint of their
older initial instructions.
Evolutionary developmental economics 347
In part, what is “initial” may also depend on our choice of t = 0, the beginning
of our inquiry. The agent considered may have started its development
earlier, from some truly initial instructions, so that its initial instructions at the
chosen t = 0 also include all the other instructions that it has received and/or
learnt in the meantime, with the help and under the constraint of the truly
initial instructions. For instance, when considering a young person starting a
school, the initial instructions that limit what the school might possibly teach
her include both her inborn talents and the ways in which these talents have
been developed, or inhibited, by her education and experiences up to that time.
Interestingly, the ultimate constraint suffices to refute the not so long ago
widespread belief that a human individual is a blank slate (“tabula rasa”),
on which anything can be written by society.3
It also helps to settle the still
lasting debate whether it is a person’s genome (“nature”), or her experiences
and education (“nurture”), that determine her cognitive abilities (intelligence,
rationality). While both are now widely recognized to matter, how exactly
they cooperate is still disputed. The constraint makes it clear that “nature”
determines a certain development potential, which “nurture” can more or
less actualize, but not expand. The cognitive abilities are thus no simple sums
of “talents plus experiences and education,” where more of the latter could
compensate for less of the former. Quite to the contrary, more of the latter, to
be properly interpreted and used, requires more of the former.4
The key implication is that a certain minimum of an agent’s instructions
must always be initial, embedded in its internal set-up (“hardware”). For some
agents, these instructions may be so simple that they are all the instructions
that the agents may ever possess. Abilities to obtain additional instructions—
from other agents, or by own learning, or both—require more complex initial
instructions. The greater these abilities are to be, the more complex initial
instructions are required.
All this implies that b-agents can be classified into two categories: programmable,
able to receive and follow additional instructions, and nonprogrammable,
limited to their initial instructions, plus possibly those that the
initial ones are programmed to develop with time. This classification will be
important for understanding the possibilities of evolution.
2.3 How the given b-agents may form, develop and operate a successful
C-agent
The given environments are assumed moderately hospitable, to allow some,
but not all, of the feasible C-agents that the given b-agents might possibly form
3A more extensive refutation of this belief is in Pinker (2001). What may be helping today’s social
scientists to see its absurdity is their growing use of computers. This makes it indeed crystal-clear
even to the most non-technical of them that there are no ways of instructing and informing an
empty box.
4This largely agrees with Ridley (2003), with the exception of the title: instead of “Nature via
Nurture,” more logical here appears to be “Nurture via Nature.”
348 P. Pelikan
to be successful (“fit”). Clearly, if the environments were less hospitable, no
evolution and development could take place, and if they were more hospitable,
so that all of the feasible C-agents could succeed, there would be no interesting
problem to study.
In such moderately hospitable environments, the given b-agents need suitable
instructions, able to guide them to form, develop and operate one of the
successful C-agents, and steer clear of all of the unsuccessful ones. To visualize
how they may proceed, recall the two dimensions of their behaviors, for which
they need the corresponding two dimensions of instructions. The associative
ones guide their self-organizing to form a certain network, where each finds a
certain position, while the operational ones guide their acting and interacting
in the positions found. But the two dimensions need not be sharply separated
in time. Stages of self-organizing and operating may overlap or alternate with
each other.
The instructed self-organizing, by which the b-agents form and develop their
C-agent, proceeds by their selectively establishing, following their associative
instructions, certain connections with certain neighbors, while refusing other
connections and other neighbors. Their instructions may not be fully informative,
but may require uses of random signals for trial-and-error searches, during
which connections can be made, unmade and re-made. In some cases no stable
C-agent may ever form. For a successful C-agent, however, the development
must converge to a stable network, or at least to a stable envelope of networks,
within which the actual one may vary in function of states of environments.
The self-organizing may be classified into different types, in which different
b-agents play more or less different and more or less differently important
roles. The types may roughly be divided into “homogenous” and “diversified,”
these into “decentralized” and “hierarchical,” and these into “multicenter” or
“centralized.”
The common feature of all hierarchical self-organizing is that some b-agents
are there leading organizers, associatively more active and more selective than
others—such as enzymes in the development of organisms, or entrepreneurs
in the development of economies. Different mixes of cooperation with competition
may be used for their selection. But this is only one special kind of
competition that may take place as part of the development of a C-agent. A
suitable label for it is therefore “developmental.”
Which type of self-organizing will actually take place depends again on
the b-agents’ instructions. Thus, how cooperative or competitive the selforganizing
will be depends on their instructions for choosing between cooperative
and competitive behaviors, possibly in response to inputs from
environments and other agents. Whether or not the self-organizing will be hierarchical
depends on whether the agents are homogenous or heterogeneous.
Hierarchies most likely arise if they heterogeneous, but may also arise if they
are homogeneous and highly competitive. Then, however, it is possible to put
in doubt their homogeneity, and admit that their abilities may differ in hidden,
difficult to observe ways that only competition can reveal.
Evolutionary developmental economics 349
Whether or not the resulting C-agent will succeed at its environmental
performance tests most directly depends on its operating, which is an aggregate
of its b-agents’ operating. But, as the ways of the aggregating are determined
by the network into which the b-agents have self-organized, the C-agent’s
success also depends, less directly but not less importantly, on their associating.
To clearly see that the C-agent’s success depends indeed on both, consider
the two reasons why it might fail: (1) its b-agents’ operating could make
it successful, but only if they were put into a suitable network, into which
they are unable to self-organize themselves; (2) they are able to self-organize
into a network where other, more suitably operating b-agents could make it
successful, but are unable to do so there themselves.
The crucial problem of development can be put as follows: How to make
the b-agents adjust their associating to their operating abilities, to prevent
their network both from overtaxing these abilities and from leaving them too
underexploited? Clearly, if overtaxed, the C-agent would fail because of the
b-agents’ operating errors, and if too underexploited, it could only succeed in
easy environments, but fail in more difficult ones—although in many of these
it might still succeed, had only its b-agents self-organized into a more suitable
network that would exploit their operating abilities more efficiently.
Much like self-organizing, networks can also be classified into different
types, where different b-agents occupy more or less different and more or less
differently important operating positions. The types may also be divided into
“homogenous” and “diversified,” these into “decentralized” and “hierarchical,”
and these into “multicenter” and “centralized.” Hierarchical networks
also make some b-agents leading—such as brains in organisms and managers
in firms. But these are leading operators, which may differ from leading
organizers, while the type of the network may also differ from the type of
the self-organizing that produced it. For instance, a multicenter ontogeny
may produce an organism with a central nervous system, while centralized
entrepreneurship may produce a decentralized firm.
The operating of networks may also use different mixes of cooperation with
competition, but this is another kind of competition, suitably denoted as “operational,”
different from the earlier considered developmental one. A clear
economic example is the price competition among established firms, different
from the Schumpeterian developmental competition for the establishment,
growth, and survival of firms.
2.4 Evolution as the search for, and producer of, needed instructions
Two points should be now reasonably clear: (1) what the b-agents must do to
form, develop and operate a successful C-agent, and (2) that for doing all this
they need suitable instructions. The next question is: how can such suitable
instructions be obtained?
In principle, as noted, the answer is: by evolution. But this cannot help
all b-agents. If they are non-programmable, they have no room for any addi-
350 P. Pelikan
tional instructions that evolution might produce. They may form, develop and
operate a C-agent only as pre-programmed by their initial instructions, and
may thus succeed only if these instructions happen to be suitable. Let me say
that such agents constitute a “self-assembling puzzle.”
Importantly, the puzzle need not be limited to a single solution. The
instructions may admit certain inputs from environment, and thus lead to
different solutions in function of these inputs. But they must then pre-program
the variants, the admissible inputs, and the function.
Evolution may only help b-agents that are programmable, with sufficiently
complex initial instructions that admit a certain variety of additional
instructions—somewhat like a complex hardware that admits a certain variety
of software. Let me say that such agents constitute a “programmable construction
set.”
The admissible variety of additional instructions is the window within which
evolution can work and to which it is limited. Success is possible only if this
variety includes some suitable instructions, and if evolution proceeds in ways
that can sooner or later find them.
As the admissible variety of additional instruction is determined by the bagents’
initial instructions, their relation to evolution can be summarized as
follows: Given b-agents in need of suitable additional instructions, the possibilities
of evolution to help them are constrained by their initial instructions.
Note well that in this evolution, the b-agents are given and only their Cagents
are in question. It will be the task of the following section to consider
that the b-agents may be C-agents of a lower level with initial instructions
produced by a lower-level evolution. As a first step to understanding this
complication, think of ants and humans. Biological evolution has produced
the initial instructions for both, but socioeconomic evolution may only work
with humans. Ants constitute socially non-programmable self-assembling puzzles,
which may form, develop and operate successful societies only if their
biologically evolved initial instructions prove suitable even for this purpose. In
contrast, thanks to their more complex initial instructions, humans constitute
programmable constructions sets with both a large room and an urgent need
for additional instructions from socioeconomic evolution.
All types of evolution must use random signals, random in the abovementioned
sense: they may have deterministic causes, but they lack information
on what instructions may be suitable. Without this information, the only
way to success is a trial-and-error search, in which random signals are needed
for triggering the trials.
Such a search may vividly be described in Campbell’s (1965) classical terms
of “variety, selection and retention.” Random signals may be said to choose
instructions from a variety of possible instructions; the b-agents use them to
try to form, develop and operate corresponding C-agents, and these are tested
by the environments. If the C-agents succeed, the instructions are selected and
retained, if not, they are rejected.
That in the present model, evolution select and retains instructions, and
not C-agents, deserves emphasis. The performance (“fitness”) of C-agents is
Evolutionary developmental economics 351
important, but only as a proxy through which their instructions take part in
evolutionary competition. This is the third kind of competition, reserved to instructions,
which must be distinguished from the earlier noted developmental
and operational kinds among the b-agents within their C-agent.
The present model thus disagrees with the old but still widespread view
of evolution as selection of organisms, but is fully in line with the modern
view that the main products of evolution are genes and genomes. It also
agrees, although with a qualification, with Dawkins’s (1976) argument that in
Darwinian evolution the units of selection are genes and not organisms. Since
genes are special cases of instructions and organisms are special cases of Cagents,
the agreement is with the exclusion from evolutionary selection of the
latter. The qualification is that the units are not limited to single genes, but
admitted to include unions of several genes and sequences of non-genic DNA.
For evolution to proceed, at least some of the b-agents must possess, in
addition to their instructions for associating and operating, also instructions
for making and spreading, with the help of random signals, trials with new
instructions—such as a genomic mutation, or an institutional change. Each new
trial typically originates in a single b-agent, from which it may spread by lateral
communication and/or by inter-generational transmission (inheritance).
Biological evolution mostly uses the latter way. The b-agents (cells) of a
C-agent (organism) are there usually offspring of the cell in which a new
trial originated. But lateral communication is not entirely absent: in some
populations of bacteria, a gene that one of them happens to find to provide
defense against an antibioticum may rapidly spread to many of its neighbors.
In socioeconomic evolution, new instructions spread both laterally, among the
actually present individuals, and from one generation to the next.
The present model admits that the evolutionary trial-and-error searches
may proceed both in parallel, involving many different short-lived C-agents,
as in the evolution of life, and in a time series using a few long-lived C-agents,
as in the evolution of economies and societies. Moreover, to be applicable to
socioeconomic evolution, it also admits that past trials may produce relevant
information that affects the probabilities of future trials. As long as the
relevant information is not complete, however, some random signals are still
needed, although the trials themselves then become probabilistic, rather than
entirely random. That this means to admit elements of Lamarckism does
not seem to be a problem. Following Nelson and Winter (1982)—with the
notable, but for me difficult to understand exception of Hodgson and Knudsen
(2006b)—most of evolutionary economists appear to agree that in this sense,
socioeconomic evolution is indeed partly Lamarckian.
2.5 Levels of evolution vs. levels of development
The basic difference between evolution and development should now be clear,
but a few points may need additional clarifications. The reason is that several
key terms—in particular randomness, competition and selection—make sense
in the context of both, but in different meanings. Much confusion has been
352 P. Pelikan
caused by not distinguishing the two. The distinctions that are most important
to realize may be summarized as follows.
Trials generated with the help of random signals play important roles in
all evolution and in some development. But the former consist of tentative
instructions, and the latter of tentative self-organizing of b-agents during the
development of their C-agent guided by its actual instructions. In biology, this
is the difference between mutations of genomes, and the possibly tentative
attempts of cells to find their places in their organism during the genomically
guided ontogeny. In economics, as considered in more detail below, this is the
difference between changes of institutional rules, and attempts of entrepreneurs
to form firms and introduce technological innovations under the actual
institutional rules.
The main difference between the evolutionary and the developmental kinds
of competition and selection is, as noted, that the former take place among
instructions and the latter among b-agents. This difference is particularly
important for the issue of cooperation. As clarified below, developmental competition
may be replaced by cooperation to a larger extent than evolutionary
competition.
A tricky issue that may appear to challenge the present model is the
one of epigenetics, concerning inherited changes in features of organisms
caused by mechanisms other than changes in DNA. As this issue is of little
interest to economists, let me just mention the key to coping with it. This is
to further extend the biological meaning of the term “instructions” beyond
genes and genomes to the non-DNA factors that may also help to govern
the gene expression. An organism’s instructions can then be seen to form
hierarchies, where higher-level instructions help govern the uses of lowerlevel
instructions. While many of these levels may consist of DNA itself, what
epigenetics points out is that they may also include some non-DNA factors.
But in the present model, these are also instructions.
A particularly tricky issue is the one of multilevel evolution and multilevel
selection. This one evolutionary economists need to understand, while not
even biologists appear to be entirely clear about it. What allows the present
one-level model to help is that it can embrace several levels by recursive
applications. Except for what is defined as the lowest and the highest levels,
it allows b-agents to be modeled as C-agents of a lower level and their initial
instructions as products of some lower-level evolution, and C-agents as bagents
of a higher level which, thanks to their initial instructions, are more
or less able to form, operate and develop C-agents of a higher level. The help
comes from its four elementary, but often neglected conceptual distinctions:
(1) between evolution and development; (2) between instructions and agents,
(3) between initial and additional instructions; and (4) between self-assembling
puzzles and programmable construction sets. What these distinctions help to
clarify can be summarized as follows.
There may be more levels of development than of evolution. This is because
evolution needs programmable construction sets, whereas development may
also take place with self-assembling puzzles. Some levels of development may
Evolutionary developmental economics 353
therefore lack evolution, while one level of evolution may produce instructions
for several levels of development.
As an example, think again of ants and humans. In both cases, their biologically
evolved genomic instructions suffice for several levels of development—
from proteins through cells to individuals. Each of these levels can be understood
as a self-assembling puzzle with possibly several solutions, as directly or
indirectly programmed by the genomes. Ultimately, however, ants are genomically
programmed to constitute socially non-programmable self-assembling
puzzles, so that no higher-level evolution can take place. In contrast, humans
are programmed to constitute socially programmable constructions sets, for
which a higher-level evolution is both possible and necessary. Perhaps the
schematic Fig. 1 may help to visualize this apparently complicated situation.
For clarity on multilevel selection, the main help comes from the distinction
between the evolutionary selection of instructions and the developmental
selection of b-agents for positions within their C-agent. This makes it possible
to see the key dependence of the latter on the outcomes of the former. As
the instructions selected by evolution guide the development of the C-agent,
they also determine what competition and selection of its b-agents, if any, this
development will employ. As development may have more levels than evolution,
developmental selection may also have more levels than evolutionary
selection. In nearly all biology, the former may have many levels, but the latter
only one: the evolutionary selection of genomic instructions.
atoms
programmable
molecules
cells
individual organisms
societies with genome-implied rules
human groups
large human societies
evolution with development:
search for additional instructions
for same-level self-organizing
development only: selforganizing
guided by already
found lower-level instructions“Big Bang“
non-programmable
molecules
one-cell organisms multicellular organisms
social organisms
programmable
(mainly humans)
non-programmable
(e.g., ants)
Fig. 1 A schematic survey of levels of evolution and levels of development
354 P. Pelikan
Evolutionary selection that has more levels, as in the case of humans, raises a
new question: which instructions will be rejected, if the highest-level C-agent—
such as a national economy—fails its performance tests? The first victims will
always be the highest-level additional instructions. Their fault is not to lead the
highest-level b-agents (= the next-to-the-highest-level C-agents) to make the
highest-level C-agent successful. But this is may not be the end of the story. It
may also be the fault of the b-agents’ initial instructions not to make enough
room for evolution that could find in time suitable highest-level additional
instructions. Next victims may therefore be the b-agents’ initial instructions.
Another new question is: how do the different levels of evolutionary
selection influence each other? Important examples of such influences can
be found in the literature on gene–culture co-evolution concerned with links
between socioeconomic evolution and the biological evolution of human
genomes (for a recent account, see Laland et al. 2010). The general principle
can be summarized as follows. To recall, the instructions of the b-agents
forming a C-agent consist of two parts: (1) their initial instructions, and
(2) the additional instructions for this forming. Each part has its own level of
evolution, which may at first be independent. But once the C-agent is formed,
it may internally modify the success criteria for its b-agents, and thus modify
the evolution of their individual instructions. These in turn may influence its
further development and performance, and through it the evolution of the
additional instructions. For more on this question, see Section 3.2 below.
2.6 Cooperation, altruism, and group selection
These three often debated notions have two things in common: their roles
in evolution are still far from clear, and Darwinism has been claimed unable
to accommodate them. To show that the present evo–devo model can both
accommodate them and clarify a few points about them is therefore a strong
argument in its favor—and a nice way to wrap up Part I.
The main problem with these notions is that they are used in different
meanings, not always properly distinguished from each other. The model need
not do much more than bring the relevant distinctions to light, and most of the
missing clarity immediately appears.
The two possible meanings of cooperation and altruism are: (1) among
instructions, (2) among the b-agents of C-agent. The conceivable third meaning,
among free C-agents, need not be considered separately. Cooperating or
altruistically interacting C-agents may formally be defined as b-agents of a
higher-level C-agent, and thus included in meaning (2).
Cooperation is admitted in both meanings, but in meaning (1) only among
instructions specialized in different types of b-agents’ activities. Such instructions
must indeed cooperate to guide all the different actions that bagents
must take to develop and operate a successful C-agent. But alternative
instructions specialized in the same type of b-agents’ activities may only
compete: in each C-agent, only one can rule. As noted, their competition uses
Evolutionary developmental economics 355
their C-agents as interposed heroes, whose performance success imply their
evolutionary success.
Cooperation (2) is accommodated without restrictions, and admitted to mix
with competition in many ways and degrees. How much, or how little, the bagents
will cooperate depends on their instructions of both dimensions. Their
operational instructions are responsible for how cooperative each of them
will be in response to the conditions they face within the C-agent’s network,
such as the incentives for cooperating vs. those for defecting. Their associative
instructions are responsible for the network, and thereby for the conditions.
But, how cooperative the b-agents will be made by their instructions is only
one side of the story. The other side is, how cooperative they must be to
allow their C-agent to succeed. This depends in part on the distribution of
their abilities, and in part on the severity of their environments. If these are
severe and they are heterogeneous, the C-agent will need to develop efficient
hierarchies, which will require a certain minimum of competition among
them for leading positions. But if they are homogenous, no competition is
needed: all of their permutations are equally good. As all competition is costly,
success in severe environments will on the contrary require a maximum of
cooperation.
The problem for evolution is how to make the two sides meet—that is, how
to find suitable instructions that would make the b-agents combine cooperation
with competition as required for their C-agent’s success. As explained,
a solution may exist only if the b-agents constitute either a self-assembling
puzzle, with all the needed instructions already produced by their lower-level
evolution, as in the case of ants, or a programmable construction set with
instructions that would allow a higher-level evolution to find in time some
suitable instructions of the higher level, as we must hope is the case of humans.
While there are many features that altruism shares with cooperation, there
is also an important difference: cooperation is admitted in both meanings,
but altruism only in meaning (2). In other words, both agents and some
instructions may cooperate, but only agents may be altruistic to each other.
The only relation to altruism instructions may have is that, to make their Cagent
successful, they may have to prescribe much of altruism (2) for its bagents.
But they are always selfish in the sense in which Dawkins (1976) shows
genes to be selfish.
Altruism (2) is not only admitted, but recognized to be often necessary for
a C-agent’s success—and thereby for its b-agents’ success, to the extent to
which this depends on the C-agent’s success. The instructions may prescribe
altruism directly, as for ants, or conditionally, as for humans, but no agent can
be altruistic without instructions for it.
Altruism (2) thus raises similar problem for evolution as cooperation (2):
how to endow the b-agents with instructions that would make them as altruistic
as required for their C-agent’s success. There are also similar conditions for a
solution to exist: the b-agents must constitute either a self-assembling puzzle,
with all the needed instructions already evolved, or a suitably programmable
construction set, allowing a higher-level evolution to produce them.
356 P. Pelikan
A particularly important factor, on which the actual level of altruism
strongly depends, is the C-agent’s defense against selfish b-agents—such as
cancer cells in an organism, or free-riders in a society—realized by the bagents’
self-organizing. Obviously, it will be easier to attain the required level
of altruism if selfishness is punished and eliminated rather than tolerated
and rewarded. Taking such a defense into account qualifies and extends
the propositions by which Wilson and Wilson (2007) summarize their argument:
“Selfishness beats altruism within groups. Altruistic groups beat selfish
groups.” The qualification is that selfishness beats altruism only in groups
without an effective anti-selfishness defense. The extension is that altruistic
groups with such a defense beat altruistic groups without it.
To clarify the issue of group selection, the first step is to understand
“groups” as C-agents and specify their b-agents. The key distinctions are:
(1) between free C-agents and subordinate C-agents that are b-agents of a
larger C-agent; and (2) between fully instructed b-agents, constituting a selfassembling
puzzle, and b-agents in need of additional instructions, constituting
a programmable construction set.
In general, as the model excludes evolutionary selection of agents, it also
excludes evolutionary selection of groups as such. Evolution can only select
instructions that enable given individuals (b-agents) to form, develop and
operate successful groups (C-agents), but not the groups themselves. In the
two-level case, however, the model admits developmental selection of lowerlevel
groups (subordinate C-agents) for positions within the higher-level group
(larger C-agent)—such as selection of firms for positions within the national
economy.
The evolutionary selection concerning groups is thus limited to their instructions,
but admitted to have several levels. There is always the level selecting
the initial instructions of the group members (the lowest-level b-agents). For
groups of organisms, this is the level of phylogeny, selecting the organisms’
genomes. These genomes determine what groups, if any, the organisms will be
able to form, develop and operate. For fully instructed organisms, like ants,
this is the only level. Only for organisms that have genomically provided room
for additional instructions may evolutionary selection concerning groups (but
not of groups!) have more than one level. It is such organisms that the rest of
this paper is about.
3 Evolutionary developmental economics
3.1 A simple evo–devo model of economic change
To build an evo–devo model specialized in economic change, the notions of
the general model must first be given a clear economic meaning. To convey the
main ideas, it suffices to include agents of three levels: individuals, economic
organizations, and economies. Individuals are b-agents of the lowest level.
The organizations—understood very broadly to include households, firms,
Evolutionary developmental economics 357
governments agencies, and markets5
—are C-agents for individuals and bagents
for economies. These are the C-agents of the highest level: their bagents
may be individuals, or organizations, or both.
The corresponding instructions are of four levels, starting with human
genomes. These provide a large, but not unlimited room for three levels of
additional instructions with the corresponding three levels of socioeconomic
evolution. In the short run, human genomes may be viewed as stabilized
outcomes of purely biological evolution and natural selection. In the long run,
however, they must be admitted to continue to evolve, and their selection to
be increasingly artificial rather than natural, shaped in a feedback fashion by
the very economies and societies that they have made possible to form and
develop.
The three levels of additional instructions are termed “cultural,” “organizational,”
and “national.” The cultural level contains the first additional instructions
that individuals receive during their education, before they enter into
what may be defined as their “economic life.” In economic analysis, the cultural
and the genomic instructions may often be considered together as sources
of individual preferences, values and cognitive abilities (“rationality”)—in part
for individual decisions, and in part for the understanding and respecting of
social rules.
The instructions of the organizational and national levels, the domain
proper of economic analysis, are termed “institutional rules.” They may be
visualized as “the rules of the game” that constrain and shape the acting and
interacting of individuals within the corresponding organization or economy.
Such rules may be formal, produced by specified rule-makers, and/or informal,
consisting of widely imitated contributions of anonymous institutional
innovators.6
In no economic C-agent—neither in a firm, nor in an economy—do its
institutional rules fully determine how it will develop and perform. There
nearly always leave a large room for their different interpretations by different
individuals in different environments. Yet they do specify many of the Cagent’s
features—such as the form of corporate governance for a firm, or the
extent of government policies for an economy. They thus also indicate where,
in the popular classification of economies into different types of capitalism,
welfarism, etatism, and socialism, the economy may best be placed. Last but
5Why markets are counted as organizations is that they, too, may be viewed as C-agents containing
networks of b-agents. They may also be ascribed objectives—such as seeking equilibrium prices—
even if these may not be the objectives of any of their participants, and may never be fully attained.
6The term “institutional rules” means what North (1990) defined as “institutions.” While I like
North’s word-economizing terminology and also used it in Pelikan (2003a, b), it has proved to
have a major flaw: not even after twenty years has it been generally adopted. Many economists
continue to call “institutions” also large organizations, and some even such disparate things as
routines, money, and languages. It is this continuing ambiguity of the term “instructions” that
induced me to return to the longer term I used in Pelikan (1988, 1992).
358 P. Pelikan
not the least, they constitute a crucial constraint on the C-agent’s performance
and development: they imply a limit to what the firm or the economy might
ever achieve in the most favorable environments.
Like genomes in evo–devo biology, institutional rules constitute the interface
between economic evolution and economic development: they are the
products of the former and instructions for the latter. Each of their levels
has its own level of evolution, with its more or less random trials, criteria
of selection, and methods of retention. Why the trials, despite their origins
in human minds, are at least partly random—in the sense of being only
imperfectly correlated with their future success—is that humans, including the
most advanced institutional economists, are still far from able to foresee all
of their important effects. While some learning of these effects has been going
on—which is why socioeconomic evolution is partly Lamarckian—this learning
is still far from eliminating all randomness.
A novelty is that the selection criteria are double: external and internal.
To succeed in their evolution, the institutional rules of a C-agent must both
enable it to be efficient enough to pass its environmental performance test,
and gain sufficient approval of its b-agents not to be internally rejected. The
difficulty of achieving both may be illustrated by the still numerous national
economies where economically efficient institutional rules cannot obtain the
needed political support, whereas the politically popular ones are ruinous.
Retention typically uses written documents for formal institutional rules,
and states of minds, maintained by verbal and non-verbal communication, for
the informal ones.
As economic evolution only concerns institutional rules, the entry and exist
of firms together with changes in technologies and industries are parts of
development, shaped and constrained by the so far evolved institutional rules.
This disagrees with Nelson and Winter (1982), who label such changes as
“evolutionary,” but agrees with Schumpeter (1934/1912), who calls his theory
for dealing with them “of economic development.”
In this simplified version, the model includes two levels of developmental
competition and selection: (1) of individuals for positions within organizations
and/or the economy, such as those of managers, leading investors, and policymakers;
and (2) of organizations for positions within the economy, such as
those of firms on markets. The success criteria usually require efficiency in
the uses of scarce resources—including technologies, innovations, and human
capital. But they may even more importantly depend on the economy’s
institutional rules—such as the competition law, the bankruptcy law, and the
conditions for obtaining government subsidies. It depends on these rules how
the success of individuals and organizations will be correlated with the success
of the economy. If the rules allow the correlation to be weak or negative, they
will pay for it by failing in their own evolution.
Why so many studies of economic development are improperly labeled
“evolutionary” may be that the difference between the two is more difficult to
see in economics than in biology. First, economic development is usually more
experimental than ontogeny. Like evolution, it also extensively uses tentative
Evolutionary developmental economics 359
trials, elimination of errors, and selection of successes. It is indeed in these
terms that Schumpeter’s notion of “creative destruction” may be described.
Second, the speeds of the two differ much less in economics than in biology.
There, phylogeny is so much slower than ontogeny that hardly anyone may fail
to see the difference. In economics, in contrast, institutional rules sometimes
change as often as networks of firms and markets, which makes the two more
difficult to disentangle.
The possible similarity of speeds has two other consequences. First, institutional
rules that change after a relatively short period of economic development
may not have had enough time to demonstrate their full development
potential. Second, new institutional rules cannot start from scratch, but must
build on the possibly wrongly developed economy left to them by their
predecessors. Much of the previous development may have first to be undone
before the development under their guidance may start taking off.7
But there is a simple way to clarity: specify the C-agent and distinguish its
institutional rules from the network of its b-agents. The trial-and-error changes
of the rules are evolution, and those of the network, instructed by the so far
evolved rules, are development.
3.2 Levels of economic evolution vs. levels of economic development
If economic evolution has more than one level, the way to clarity is slightly
more complicated. More levels of evolution mean more levels of development,
and this creates a maze of effects through which the different levels may interrelate.
In principle, this maze should be possible to understand by interpreting
in economic terms Section 2.5 above. But an explicitly economic map may
nevertheless be useful.
To draw such a map, start from the top: consider an economy and identify
the effects of its institutional rules on its performance. Distinguish their agentcoordination
effects from agent-quality effects, and divide these into ex post
and ex ante branches.
The agent-coordination effects concern the economy’s actual agents—both
organizations and individuals. They shape the agents’ choice sets, incentives,
and information exchanges. It is these effects, and especially those on transaction
costs, that most of modern institutional analysis has been about.
The less explored, but not less important agent-quality effects are on the
agents’ internal instructions—that is, the institutional rules of organizations,
and the cultural-genomic instructions of individuals. Because of them, the
economy’s institutional rules also affect the agents’ cognitive abilities, preferences
and values. The ex-post branch do so through its effects on the suc-
7This consequence is clearly illustrated by the J-curve growth of the economies of Central and
Eastern Europe after the transformation of their institutional rules from the failed trials with forms
of socialism to new trials with forms of capitalism.
360 P. Pelikan
cess criteria of the developmental selection, which influences the qualities of
the b-agents promoted to leading positions. The institutional rules with such
effects include competition law, bankruptcy law, patent law, tax law, and the
conditions for obtaining government subsidies. The success of b-agents in
this developmental selection implies the success of their instructions in the
economy’s internal evolutionary selection.
The ex ante branch intervenes in the agents’ internal instructions directly.
It restricts the permissible variety of these instructions at the outset, and thus
also reduces the room for their evolutionary selection. Examples include the
corporate law that restricts the variety of forms of corporate governance of
firms, and the social and religious norms that restrict the variety of individual
choice procedures. If the restrictions happen to be targeted correctly, they help
the evolution succeed faster, but if not, they doom it to fail. If the restrictions
are so strong that they prescribe the agents’ internal instructions in all relevant
detail, they close this evolution. There is then only one level of evolution—the
one of the economy’s rules.
In general, the economy’s rules constrain all of its agents, whether these are
members of organizations or not. An organization’s members are moreover
constrained by the organization’s internal institutional rules. Like the economy’s
rules, these may also be formal and/or informal, and also responsible
for the organization’s performance via member-coordination and memberquality,
ex post and ex ante. But their responsibility is only partial. Much of the
responsibility stays with the economy’s rules: these constrain the permissible
variety of the organization’s rules, and possibly strongly influence its success
criteria.
Finally, the map must also show how all these levels are constrained by
the individual instructions produced by the gene-culture co-evolution. This
makes it necessary also to put into it instructions for what may be termed
“rule-making” and “rule-following.” The former constrain the variety of the
institutional rules that the individuals may potentially imagine and economic
evolution might consequently try. The latter constrain the sub-variety of these
trials that the individuals may actually follow. The larger the potential variety,
the more likely it contains some actual successes, but also the longer the
expected time for finding any of them.
For a deeper insight, it is necessary to leave pure economics and try to
separate cultural instructions from the genomic ones. It may be that human
genomes endow people with too much fantasy, implying a too broad potential
variety of institutional rules that people could imagine and be willing to try,
where successes are too rare and therefore too difficult to find. Cultural
instructions may then be understood as helping economic evolution by ex ante
narrowing this enormous potential variety. But they carry the same risk as the
institutional rules with ex-ante agent-quality effects: they may help only if they
happen to be targeted correctly—that is, if the narrowed potential variety is not
entirely outside the successful sub-variety. Otherwise far from helping, they
are the main obstacle to the success of economic evolution, and hence an error
of cultural evolution.
Evolutionary developmental economics 361
Economists may play an important role in cultures where knowledge is
valued more than blind faith and superstition. They may reduce the above
risk by building and teaching reasonable theories of institutional rules that can
effectively predict, if not which rules can succeed, at least which ones are bound
to fail. While much of the risk then reappears in the evolution of theories,
which also must proceed by trials and errors—or, in Popper’s (1963) words, by
conjectures and refutations—the advantage is that such errors are usually less
costly.
3.3 Two rough examples
A map cannot solve problems, it may only help to organize the search for their
solutions. But this may still be useful. If the problems are complex, with many
different parts where their solution may hide, a good map of all the parts may
even be necessary not to overlook it.
Details must be left to another occasion. For a rough idea, consider some of
the often overlooked points that the map brings to light in two hotly debated
issues: (1) the growth and development of poor economies; (2) multiculturalism
in developed economies.
In issue (1), the map turns more attention than usual from availability of
resources to institutional rules. These were for a long time entirely neglected,
as if the growth of an economy only depended on how much resources it
possessed or was given in foreign aid. Since the beginning of the 1990’s,
development economists have been increasingly recognizing that institutional
rules also matter, but not much more than other growth factors. The map
points out that they do matter much more. They imply a certain development
potential that a lack of resources may leave unfulfilled, but no surplus of
resources can exceed—much like a lack of nutrition may prevent a mouse
embryo from developing, but no nutrition, however abundant, can make it
grow into an elephant.
More specifically, the map calls attention to some of the effects and limitations
of institutional rules that their usual analysis has been leaving aside.
While this analysis has been preoccupied with their coordination effects at the
economy level, the map also brings to light their agent-quality effects and the
limitations stemming from the cultural-genomic instructions of individuals. It
is these effects and limitations that appear to hold the keys to the old puzzle
of why some economies continue to perform poorly even when endowed with
the same formal institutional rules that make other economies prosper.
For issue (2), it must first be realized that cultures are much more than
instructions relevant to economic behaviors. In Dawkins’s (1976) terms, a
culture may be described as a set of “memes”—meaning “units of cultural
ideas, symbols or practices, which can be transmitted from one mind to
another.” The memes with instructions relevant to economic behaviors—such
as those concerning respect for property rights, sense of fairness, business
ethics, propensities for lying and corruption, ways of solving conflicts, and
trust—constitute only a subset of this set. The memes from the remaining
362 P. Pelikan
subset—such as those concerning costumes, food, music, dances, and rituals—
may suitably be denoted as “ornamental”.
Admittedly, the precise divide between the two subsets may not be entirely
clear, and may have to be shifted in the light of new discoveries. For instance,
some memes that at first appear purely ornamental may turn out to have
significant effect on economic outcomes—such as food diets that may turn out
to affect health and longevity, or religious rituals that may turn out to harm
production. Where precisely to draw the boundary is therefore to some extent
an open question that calls for further inquiry.
But once the boundary is at least roughly identified, the answer to issue
(2) becomes clear. A developed economy may unlimitedly accommodate ornamental
multiculturalism, but must set strict limits to the economically relevant
one—under the threat of undermining its own development. It must protect
the institutional rules and values on which its development fundamentally
depends against contamination by any of the economically relevant memes
that have been hindering the development of other economies.8
All this modifies the usual valuation of the labor imported into a rich
country from a poor country. In addition to the usually considered labor
quantity, it calls attention to the risks of import of economically relevant
memes which may be the prime causes of the poor country’s poverty—such
as low respect for property rights, rules of honor that foster wasteful conflicts,
soft constraints on corruption and hard constraints on education opportunities
for women, which both waste their talents and lower the quality of education
of their children.
The difficult question is, what civilized policies can defend developed
economies against imports of economically detrimental memes. To steer clear
of all forms of racism and fascism, a minimum requirement is that they
must welcome ornamental multiculturalism and adopt strictly individualistic
approach to immigrants, not to discriminate against the many of them who
are keen to learn and adopt the economically successful memes of the host
country. But the exact form of such civilized defense policies is far from clear.
Interestingly, the division of a culture between ornamental memes and
memes with economically relevant instructions formally corresponds to the
division of a genome between non-instructing DNA and DNA with ontogenetically
relevant instructions, namely genes and the non-genic DNA coding
for RNA regulators. Whereas all memes and all DNA segments may replicate,
only some carry instructions for the corresponding b-agents on how to form,
develop and operate a successful C-agent. In both biology and economics,
much still remains to be learned about the exact position of the divide,
which may be found somewhat different than initially believed. To recall, the
biological divide was first believed neatly to separate genes from non-genic
8Some of the best explanations of what rules and values lead to economic development can be
found in North and Thomas (1973) and Rosenberg and Birdzell (1986).
Evolutionary developmental economics 363
DNA, which was called “junk”–before parts of the “junk” were discovered to
contain important instructions for the regulation of gene expression.
The division between the two kinds of memes also calls for a correction
of Dawkins’s argument. By itself, as just demonstrated, the term “meme” is
useful, but the argument that memes logically correspond to genes is faulty.
According to their definition, they only share with genes the abilities to
replicate, but not necessarily those to instruct, which Dawkins is leaving aside.
Memes may therefore correspond to any segments of DNA, of which only
some are genes. But, as genes are now known not to be the only relevant parts
of genomes, it is no longer very interesting to look for exact socioeconomic
correspondents just to them.
3.4 Evolutionary developmental economics: a rough outline of a new field
The main potential contribution of the evo–devo generalization of Darwinism
appears to be methodological. By interrelating several areas studied by
different specialized fields, the above map offers a solid support through
which the fields may orderly be interconnected into a single field of evo–devo
economics. It is only in such a broad field that economic change can be studied
comprehensively, without missing any of its important aspects.
This field may be here only very roughly and very tentatively outlined.
The map suggests to structure it into three main areas: (1) the evolution of
institutional rules; (2) the development of economies; (3) the behaviors of
individuals.
Each area can be seen to contain several topics. In area (1) they include:
(1.1) the evolution proper, which includes (1.1.1) the evolution of formal law
and (1.1.2) the evolution of informal socio-cultural norms; (1.2) the properties
of the institutional rules evolved, which include (1.2.1) their coordination
effects and (1.2.2) their agent-quality effects. Moreover, each of these topics
may concern entire economies, or organizations within economies.
The topics of area (2) include: (2.1) macroeconomic performance and
growth; (2.2) the development of markets, firms and industries; (2.3) the
development of technologies; (2.4) the effects of government policies and the
development of government bureaucracies.
The topics of area (3) include: (3.1) bounded rationality; (3.2) human
learning with its cultural and genomic limitations; (3.3) the differences in these
boundaries and limitations across individuals and across cultures.
The verb “include” means that the list of the topics is only rough and
incomplete. It is also only roughly and incompletely that the different fields
can be identified. Many of them do not clearly define the scope of their
specialization, and several of them overlap, studying basically the same topics
under different names, without communicating with each other. The list may
perhaps best be seen as a public announcement of topics related to economic
change, addressed to all the specialist who already understand, or seek to
understand, some of these topics, to come forward and join, be it only parttime,
the new broad field.
364 P. Pelikan
There are many fields from which competent volunteers might come, but
three appear particularly promising: (1) the new institutional economics that
studies institutional change in the spirit of Douglass North; (2) the “evolutionary”
economics (that, as noted, should be renamed to “developmental”)
that studies the development of markets, firms and technologies in the spirit of
Joseph Schumpeter, Richard Nelson and Sidney Winter, and (3) the relatively
new field of behavioral economics with its old roots in the works on bounded
rationality by Herbert Simon.
To succeed, the specialists in one field must search for what may be
interesting about their topics to the specialists in the other fields. In other
words, they must be willing to collaborate by trying to answer each other’s
questions—rather than enclose themselves, as many of them now appear to
do, in separate chapels.
Examples of what behavioral economists will be asked about by others are
(1) the variety of alternative institutional rules that humans may be able to
invent and adopt, (2) the bounds of human rationality and the inequalities of
their distribution in society; (3) the limits of adaptability of human preferences
and values.
Institutional economists will mainly be asked about (1) the evolution of
institutional rules and the possibilities of influencing it by reform policies; (2)
the effects of different institutional rules on development processes, including
the so far little explored effects on agent qualities.
What the Schumpeterian developmental economists will mainly be asked
about is how the development of firms, industries and technologies depends
(1) on the prevailing institutional rules, and (2) on the unequal abilities of
the individuals involved, as implied and limited by their cultural-genomic
endowments.9
All this is only a very rough outline of how the field of evo–devo economics
may be built. But not much more can be said about it in a short paper. To build
it properly must be expected to take a long time and require close cooperation
of many different specialists.
4 Concluding comment
The evo–devo generalization of Darwinism was here only applied to economic
change, which is only a part of change of entire societies. This raises the
9A particularly weak link on which much work will have to be done is the one between
Schumpeterian economists and new institutional economists. Still only few of the former take
into account institutional rules, and most of the latter only study the static effects of institutional
rules on incentives and transaction costs. My earlier attempts to interconnect the two fields are in
Pelikan (1992, 2003b)—although then I was still following Schumpeterian economists by calling
the processes they study “evolution,” as opposed to the present finding that it is more proper to
call these processes, in agreement with Schumpeter himself, “development.”
Evolutionary developmental economics 365
question of how this part interrelates with the other parts–such as political
change, cultural change, and the above-mentioned gene-culture co-evolution.
My conjecture is that the generalization can help answer also this question.
But showing it must be left to another occasion or to other in generalizing
Darwinism interested social scientists.
Acknowledgements I thank Christian Cordes, Jan Gecsei, Geoff Hodgson, Philippe Huneman,
Deborah Rogers, Hiroki Sayama, Jack Vromen, Ulrich Witt, an anonymous reviewer, and the
participants of seminars at Centre pour la Recherche en Epistemologie Appliquée and Institut
des Systèmes Complexes, Paris Île-de-France, for inspiring objections and suggestions; Andrea
Pelikan Conchaudron for lessons on molecular biology; René Doursat for calling my attention to
the evolutionary developmental biology and helpfully commenting my attempts to generalize it
for applications in economics; and the Grant Agency of the Czech Republic for support under
grant 402/09/1991. The usual caveat applies.
References
Camazine S, Deneubourg J-L, Franks NR, Sneyd J, Theraulaz G, Bonabeau E (2001) Selforganization
in biological systems. Princeton University Press, Princeton
Campbell DT (1965) Variation, selection and retention in sociocultural evolution. In: Barringer
HR, Blanksten GI, Mack RW (eds) Social change in developing areas: a reinterpretation of
evolutionary theory. Schenkman, Cambridge. Reprinted 1969 in General Systems 14: 69–85
Carroll SB (2005) Endless forms most beautiful: the new science of evo devo. Norton, New York
Carroll SB, Prud’homme B, Gompel N (2008) Regulating evolution. Sci Am 298(5):38–43
Cordes C (2006) Darwinism in economics: from analogy to continuity. J Evol Econ 16:529–541
Cordes C (2007) Can a generalized darwinism be criticized? A rejoinder to Geoffrey Hodgson.
J Econ Issues 41:277–281
Dawkins R (1976) The selfish gene. Paladin, London
Doursat R (2008) Organically grown architectures: creating decentralized, autonomous systems
by embryomorphic engineering. In Würtz RP (ed) Organic computing. Springer, Heidelberg
Hodgson GM (2002) Darwinism in economics: from analogy to ontology. J Evol Econ 12:259–281
Hodgson GM (2004) Darwinism, causality and the social sciences. J Econ Methodol 11:175–194
Hodgson GM, Knudsen T (2006a) Why we need a generalized Darwinism, and why generalized
Darwinism is not enough. J Econ Behav Organ 61:1–19
Hodgson GM, Knudsen T (2006b) Dismantling Lamarckism: why descriptions of socio-economic
evolution as Lamarckian are misleading. J Evol Econ 16:343–366
Knudsen T (2004) General selection theory and economic evolution: the Price Equation and the
replicator/agent distinction. J Econ Methodol 11:147–173
Laland KN, Odling-Smee J, Myles S (2010) How culture shaped the human genome: bringing
genetics and the human sciences together. Nat Rev Genet 11:137–148
Mattick JS (2004) The hidden genetic program of complex organisms. Sci Am 291(4):30–37
Nelson RR, Winter SG (1982) An evolutionary theory of economic change. Harvard University
Press, Cambridge
North DC (1990) Institutions, institutional change, and economic performance. Cambridge
University Press, Cambridge
North DC, Thomas RP (1973) The rise of the western world. Cambridge University Press,
Cambridge
Pelikan P (1988) Can the imperfect innovation system of capitalism be outperformed. In: Dosi G
et al (eds) Technical change and economic theory. Pinter, London
Pelikan P (1992) The dynamics of economic systems, or how to transform a failed socialist
economy. J Evol Econ 2:39–63; reprinted in H.J. Wagener, ed., On the theory and policy of
systemic change, Physica-Verlag: Heidelberg, and Springer, New York
Pelikan P (2003a) Why economic policies require comprehensive evolutionary analysis. In: Pelikan
P, Wegner G (eds) The evolutionary analysis of economic policy. Edward Elgar, Cheltenham
366 P. Pelikan
Pelikan P (2003b) Bringing institutions into evolutionary economics: another view with links to
changes in physical and social technologies. J Evol Econ 13:237–258
Pinker S (2001) The blank slate: the modern denial of human nature. Penguin, London
Popper K (1963) Conjectures and refutation. Routledge, London
Ridley M (2003) Nature via nurture: genes, experience, & what makes us human. HarperCollins,
New York
Rosenberg N, Birdzell LE (1986) How the west grew rich. Basic Books, New York
Schumpeter JA (1912/1934) The theory of economic development. Harvard University Press,
Cambridge
Wilson DS, Wilson EO (2007) Rethinking the theoretical foundation of sociobioogy. Q Rev Biol
82:327–348
Witt U (2003) The evolving economy: essays on the evolutionary approach to economics. Edward
Elgar, Cheltenham
Witt U (2004) On the proper interpretation of ‘Evolution’ in economics and its implications for
production theory. J Econ Methodol 11:125–146