1 23 Biological Theory ISSN 1555-5542 Biol Theory DOI 10.1007/s13752-012-0050-6 Evolution Beyond Biology: Examining the Evolutionary Economics of Nelson and Winter Eugene Earnshaw
1 23
Biological Theory ISSN 1555-5542 Biol TheoryDOI 10.1007/s13752-012-0050-6
Evolution Beyond Biology: Examining theEvolutionary Economics of Nelson andWinter
Eugene Earnshaw
1 23
Your article is protected by copyright and
all rights are held exclusively by Konrad
Lorenz Institute for Evolution and Cognition
Research. This e-offprint is for personal
use only and shall not be self-archived in
electronic repositories. If you wish to self-
archive your work, please use the accepted
author’s version for posting to your own
website or your institution’s repository. You
may further deposit the accepted author’s
version on a funder’s repository at a funder’s
request, provided it is not made publicly
available until 12 months after publication.
THEMATIC ISSUE ARTICLE: HOW EVOLUTIONARY IS EVOLUTIONARY ECONOMICS?
Evolution Beyond Biology: Examining the EvolutionaryEconomics of Nelson and Winter
Eugene Earnshaw
Received: 1 May 2010 / Accepted: 13 June 2012
� Konrad Lorenz Institute for Evolution and Cognition Research 2012
Abstract Nelson and Winter’s An Evolutionary Theory
of Economic Change (1982) was the foundational work of
what has become the thriving sub-discipline of evolution-
ary economics. In attempting to develop an alternative to
neoclassical economics, the authors looked to borrow basic
ideas from biology, in particular a concept of economic
‘‘natural selection.’’ However, the evolutionary models
they construct in their seminal work are in many respects
quite different from the models of evolutionary biology.
There is no reproduction in any usual sense, ‘‘mutation’’ is
directed as opposed to blind, and there is no meaningful
distinction between phenotype and genotype. Despite these
substantial departures from the conceptions of evolutionary
biology, I argue that the ‘‘evolutionary’’ economics of
Nelson and Winter is indeed a legitimate extension of
Darwinian evolutionary principles to a novel domain, and
that the traditional conception of evolution by natural
selection must be revised. The novel features of evolu-
tionary economics models reflect the distinctive theoretical
requirements faced by economists. I further contend that
reproduction, heredity, blind variation, and the genotype/
phenotype distinction are all inessential to evolutionary
theory, and that their role in evolutionary biology is a
domain-specific feature of biological theory.
Keywords Darwinism � Evolution � Evolutionary
economics � Evolutionary theory � Lamarckism � Natural
selection � Philosophy of biology
Nelson and Winter’s (1982) book An Evolutionary Theory
of Economic Change presents itself as a challenging
alternative to the neoclassical orthodoxy of the discipline.
Drawing on Darwinian evolutionary principles for inspi-
ration, the authors develop what they term ‘‘evolutionary
models’’ of economic change that bear explicit and con-
scious analogies to models of population genetics. For this
reason, their work bears considerable interest as an object
of philosophical analysis. An examination of attempts at
evolutionary model-building in other domains than biology
is valuable in shedding light on the nature of evolution
more generally, precisely because of the historical con-
nection between biology and evolutionary theory.1 In
broadening our analysis, we can minimize the danger that
our understanding of evolution mistakenly includes aspects
which are particular to biology rather than evolution. We
can also understand more specifically how evolutionary
economics may instantiate the broader patterns of evolu-
tion by natural selection in an appropriately domain-spe-
cific fashion. Rather than attempting to take evolutionary
theory as practiced in biology as given and attempting to
extend it to a non-biological domain, we will examine an
extension of evolutionary modeling to a non-biological
domain to see what it can tell us about evolutionary theory.
There are striking similarities between the evolutionary
models of Nelson and Winter (henceforth N&W) and the
typical structure of biological models of evolution by nat-
ural selection. However, there are other aspects of the
N&W’s models which diverge so sharply from traditional
biological evolution that their models arguably are missing
features necessary for evolution by natural selection (ENS).E. Earnshaw (&)
Institute for the History and Philosophy of Science
and Technology (IHPST), University of Toronto, Toronto,
ON, Canada
e-mail: [email protected]
1 Naturally, this assumes that the target of analysis is a genuine case
of evolution, or at least close enough to be interesting. That N&W
indeed do present such a case will be argued subsequently.
123
Biol Theory
DOI 10.1007/s13752-012-0050-6
Author's personal copy
This raises a question: either N&W are not really modeling
ENS after all, or the example of N&W demonstrates that
conventional evolutionary theory fails to accurately capture
the evolutionary character of certain non-biological mod-
els. Although in either case, analysis may prove fruitful, I
intend to argue that an examination of the ‘‘evolutionary’’
models of N&W suggests that we should revise aspects of
the traditional understanding of ENS.2
I will focus on three distinct aspects of N&W’s evolu-
tionary models which can be seen as departures from
genuine evolution. N&W’s models involve directed vari-
ation, they lack a distinction between phenotype and
genotype, and they lack reproduction and heredity. In
examining the evolutionary models of N&W, I will suggest
that the traditional understanding of the nature of evolution
by natural selection is too narrowly based on biology, and
as such may reflect specifically biological concerns that are
not properly part of natural selection considered abstractly.
N&W’s models represent a population undergoing evolu-
tionary change via natural selection (among other things),
and by examining the explanatory structure of their
account, we can refine our broader understanding of evo-
lution by natural selection.
ENS in Biology
There are many distinct accounts of evolution by natural
selection available in the biological and philosophical lit-
erature. The dominant analysis in the philosophy of biol-
ogy, versions of which can be found in Lewontin (1970),
Sober (1984), and Godfrey-Smith (2009) among many
others, conceives ENS as having three key elements: var-
iation, heredity, and fitness differences. The main ‘‘rival’’
account, the replicator-focused view associated with Hull
(1988, 2001) and Dawkins (1976, 1982) and applied to
economics by Hodgson (2002), differs from some aspects
of this view, but accepts analogues of each of these ele-
ments. Hodgson speaks of ‘‘variation, inheritance, and
selection’’ (2002, p. 270) as the key ingredients. While
there is some dispute among the various authors over the
best definition of each term, and over the extent to which
these elements are individually necessary and jointly suf-
ficient for ENS, some version of these three elements
seems widely accepted—perhaps as widely accepted as
anything about ENS.
These elements presuppose a population of individuals;
evolution will consist of some change over time in the
composition of this population. Each of the three elements
of ENS are then understood as additional characteristics of
this population. The individuals must vary: in other words,
not every individual can have exactly the same traits as
every other. There must also be fitness differences, which
are normally understood in terms of propensities for sur-
vival and reproduction: individuals that fail to survive are
gone from the population (and can no longer reproduce),
and new individuals enter the population via the repro-
duction of existing individuals. If individuals with a par-
ticular trait tend to be better at this3 than individuals with
an alternative trait, then there is a fitness difference
between the traits. If individuals tend to be similar to their
parents (heredity), then the higher fitness trait will tend to
become more prevalent in the population compared to the
lower fitness trait.
Novelty is another key concept involved in evolutionary
modeling. It must be distinguished from variation: varia-
tion refers to the existence of actual differences between
individuals at a particular time, whereas novelty refers to
the possibility (or the actual occurrence) of a trait emerging
in the population that did not previously exist. Novelty may
be seen as a certain sort of failure of heredity, and in
biological contexts may be thought of as the possibility of
mutation. Its importance has been emphasized by some
authors—for example Hull et al. (2001) make it a consti-
tutive part of evolution by natural selection. However,
Sober (1984) and Godfrey-Smith (2009) conceive of
selection as operating on the variation as it exists in the
population, and therefore while selection may in a certain
sense be sterile without an accompanying source of novelty
(in that one expects the variation in the population to
gradually decrease under the influence of natural selection
in most circumstances), it can nevertheless operate. How-
ever, it is clear that the actual explanation of evolution by
natural selection that one gets from Darwin (1859) and
others (e.g., Fisher 1958) essentially employs a mechanism
of novelty. The gradual emergence of complex adaptive
mechanisms is incomprehensible without it, as is the pro-
cess of speciation. So while in a certain sense it may be
accurate to separate novelty from ENS, in the explanatory
contexts where natural selection is most indispensable,
novelty plays a key role. As we shall see, novelty also plays
a key role in N&W’s models.
I take the question of the key explanatory contexts of
evolution by natural selection to be an important one.
There are bound to be disanalogies between models that
operate in different domains. The relevant question is how
significant these differences are for our explanatory pur-
poses. In his recent book Darwinian Populations and
Natural Selection (2009), Godfrey Smith categorizes pop-
ulations according to how prototypically Darwinian they
2 Insofar as any account of ENS can be ‘‘traditional.’’ I sketch what I
take to be key features of the dominant analysis in the next section.
3 That is, at survival and reproduction. Specifying ‘‘better’’ very
precisely is tricky; see, e.g., Sober (1984).
E. Earnshaw
123
Author's personal copy
are along several axes of similarity. So, for example, it is
typically Darwinian (to Godfrey-Smith) to at some point in
the life cycle have a ‘‘bottleneck’’ where there is just one
reproducing entity (such as a fertilized egg, or what have
you), and it is also typical to have a distinction between
germ and soma. It turns out that human reproduction is
very typical on these criteria. But for such measures to
have interest from the point of view of philosophy, they
have to impact our explanatory projects in a significant
way—in a way that ‘‘makes a Darwinian difference’’ (p.
104), as Godfrey-Smith puts it. And I take it that so long as
a population’s change over time is explained in a charac-
teristically Darwinian way, such disanalogies as may exist
between the population as modeled and a more prototypical
population are not of broad significance. This, at least, is
how I propose to consider N&W’s models: the test of their
status as evolutionary models is in the nature of the
explanations they offer.
Nelson and Winter’s Models
In the course of An Evolutionary Theory of Economic
Change, N&W develop several distinct models.4 While the
models differ in various particulars, they share broad
similarities, which are motivated by N&W’s own views
concerning the nature of firms, innovation, and economic
change. The models posit a population of firms, each of
which is characterized by, at minimum, a capital stock and
a production technique. The production technique deter-
mines the production characteristics of the firm’s capital
stock such as the amount of labor used and the output of
‘‘goods.’’ The production technique of each firm in the
population together with other specified parameters such as
a demand curve for the goods produced by the industry
determines the costs and sales of the firms per unit of
capital stock. From this, profits can be calculated. Given
profits, some function determines the change in each firm’s
capital stock—for example, firms that lose money might
probabilistically shrink, and profitable firms might proba-
bilistically grow. When all of these values are specified, the
model is a Markov process; that is, a mathematical system
where the state of the system at a given time suffices to
determine the probabilities of all possible future states of
the system.
Thus far, we have a population of entities with varying
traits, we can track the change in prevalence of the traits in
the environment, and changes in the environment can
advantage some traits over others. Depending on the initial
conditions and parameters, we may see the gradual
extinction of all variants but the most profitable, or a move
to an equilibrium situation, or an ongoing oscillation. We
can explain these changes by appealing to the interaction of
the production technique of each firm with the environ-
ment. Given the cost of labor, some techniques may be
profitable, others will lose money, and the profitability of
the firms causes them to grow or shrink, and may even
drive some firms out of business entirely.
However for several of their models, N&W add a further
element that is essential to their purposes: a source of
novelty in the system. As they argue, ‘‘To fill in the
ranks… decimated by competitive struggle at earlier times,
or to make possible the appearance of entirely new pat-
terns, some mechanism analogous to genetic mutation must
be posited. Otherwise, selection can only bring about the
dominance of the best of the patterns that started the con-
test’’ (1982, p. 142). We have already seen that a source of
novelty also plays a key explanatory role in biological
evolution by natural selection.
For N&W’s economic models, ‘‘innovation’’ plays this
role of introducing novelty into the system. They model
innovation as a ‘‘search’’ by firms for a new production
technique to replace their old technique. Firms that engage
in search may or may not find a new production technique;
if they do, the relevant characteristics of the technique are
generated via model assumptions probabilistically from a
space of possible techniques. Once a firm has found a new
technique, it will adopt the technique only if it judges it to
be more profitable than the old technique—and firms are
generally modeled as liable to error in judging the char-
acteristics of new techniques. The addition of innovation to
the evolutionary economic models is a key factor that
allows them to model economic change and technical
innovation over time. It also is another respect in which
their evolutionary models both parallel biological evolution
in some respects while diverging in others.
Analogies
Although it may already be fairly clear, I wish nevertheless
to specify more precisely why we ought to take seriously the
claim that N&W’s models involve evolution by natural
selection in just the same sense that biological models can. I
will highlight what I take to be the key evolutionary aspects
of their most complex model, which simulates the historical
development of the production techniques of an industry.
The fact that the model is a Markov process means that,
given a specification of the state of the population at a
particular time, the probability of each possible subsequent
state can be calculated. This means that their model gives
4 I will characterize their models generally to begin with, but much of
my more specific analysis will be focused on one of the more complex
models they discuss, presented in chap. 9, which attempts to recreate
the change over time of an actual historical industry.
Evolution Beyond Biology
123
Author's personal copy
us a complete picture of the possible ways in which the
system can change, and how likely each such path is to
occur. They are sufficient to simulate the trajectory of the
entire system.
Given the environment, some types are less successful
and others more successful. Over time, trait types (pro-
duction techniques) that are less well suited to the envi-
ronment will tend to decline and eventually vanish. This
means that over time, the population will change in an
incremental, directional, broadly predictable fashion,
insofar as we expect the combination of novelty and
selection to gradually drive the population to increasingly
‘‘efficient’’ production techniques—efficient in the sense of
producing more output for lower costs. And we can
therefore appeal to the force of evolution by natural
selection to explain the gradual increase in efficiency of the
industry as a whole.5 This is strikingly similar to how one
would explain the gradual increase in the average value of
any trait that gives a competitive advantage via biological
evolution—whether it be height, or fecundity, or what have
you.
This is not to suggest that evolutionary models must
always behave in directional, incremental fashion, or that
efficiency in evolutionary models must always increase. It
is merely to suggest that we can explain such phenomena,
when observed in a population, by appeal to ENS. N&W’s
model seems to do just that: it generates the historical
pattern of change observed in the industry by modeling the
introduction of novel variants together with the competitive
success of more efficient variants.
Over time, techniques develop in the industry that would
have been extremely improbable given the initial config-
uration of the industry. This corresponds to what Godfrey-
Smith identifies as one of the key roles of natural selection
in biological evolution: changing ‘‘the population-level
background’’ in such a way that ‘‘traits that are otherwise
very unlikely to arise via the immediate sources of varia-
tion, become much more likely to arise.’’ (2009, p. 50). In
N&W’s model of industry change in chap. 9, this is exactly
what occurs: the expansion of successful firms and the
pressure on less successful firms to innovate leads to the
development of increasingly efficient productive tech-
niques. Because search for techniques is local, the most
efficient techniques in the possibility space are initially
inaccessible, but over time, the process of search and
selection brings the industry within striking distance.6 We
see a cumulative process of stepwise ‘‘adaptation’’ where
each step represents a small improvement, but which
eventually results in a situation where a ‘‘mutation’’ that
had previously been impossible becomes likely.
N&W can explain the spread of an advantageous trait, the
increase of an advantageous measurable character, the
achievement of initially improbable novelty. These are all
fairly distinctive successes of evolution by natural selection,
and N&W seem to explain them in broadly the same way as
biologists would. There are definitely disanalogies, which
must be examined more closely, but I take it that the analogies
are sufficiently striking to justify our subsequent inquiry.
Disanalogies
Despite the similarities between the explanatory resources
of N&W’s models and those of evolutionary biology, there
are distinct differences that might undercut the extent to
which N&W can accurately be conceived as offering
evolutionary explanations. I will focus on three disanalo-
gies in particular. In what I consider to be ascending order
of importance, these are that variation is not blind, there is
no distinction between phenotype and genotype, and that
there is no reproduction, and therefore no inheritance.
In N&W’s models, firms do not always have a chance of
innovating. Generally, they employ what is termed a ‘‘con-
servative’’ assumption that only unprofitable firms attempt to
change their production technique. Furthermore, there is a
definite bias towards the emergence of beneficial variations,
in the sense that firms are able to, in biological terms, eval-
uate (albeit imperfectly) the effectiveness of a prospective
‘‘mutation’’ and decide whether to mutate or not.
This is certainly a disanalogy from how mutation is gen-
erally conceived in biology—although research has sug-
gested that bacteria mutate differentially in response to
stressful situations (Foster 2007), which is analogous to firms
attempting to innovate when they are unprofitable. Donald
Campbell (1960) made the ‘‘blindness’’ of variation a key
element in his conception of evolutionary epistemology,
which has been widely influential. Campbell was concerned
with explaining knowledge acquisition as an evolutionary
process: he conceived variation as emerging out of a back-
ground context of already existing adaptation, but as insen-
sitive to environmental context, to the correct solution, and to
the relative ‘‘correctness’’ of previous trials. Variation in
N&W is sensitive to the environment, because a prospective
5 It should be observed that in most of the scenarios they consider,
much of the increase in efficiency over time is attributable to
innovation rather than selection. As I observe elsewhere, that natural
selection is not the whole story does not mean it is not part of the
story. But additionally, in N&W’s model, unprofitable firms are
motivated to innovate. Since selective disadvantage (i.e., unprofit-
ability) plays a key role in the process of novelty generation, it should
be credited with contributing to increased efficiency in this regard too.
It is rather as if the unsuccessful genotypes started to mutate rapidly—
fitness still plays a role in this context as well, albeit an unusual one.
6 The search for innovation is local in only some versions of N&W’s
models. In others it depends on an exogenous factor—the independent
‘‘advance of scientific knowledge,’’ more or less.
E. Earnshaw
123
Author's personal copy
variation is tested with regard to its profitability in the current
environment. And the variation is also sensitive to the correct
solution, in that variations that emerge in the population are
biased towards more efficient techniques. It might therefore
appear that the source of novelty is insufficiently ‘‘blind’’ to
fit with Campbell’s account.
However, N&W’s model does not really fall afoul of the
blindness criterion. The mechanism of novelty they employ
involves first a blind draw from a possibility space, fol-
lowed by a ‘‘test’’ for profitability. This test can itself be
understood as a selective process. So the apparent bias in
the novelty that emerges in the system can be understood as
due to an initial selection that takes place on the novel
variants, not as inherent to the variation itself.
However, I also think that Campbell’s emphasis on the
importance of ‘‘blind’’ variation can be somewhat mis-
leading. Contra Campbell, I do not take blindness to be a
necessary feature of evolutionary explanations generally.
This is because natural selection will still continue to
operate regardless of whether mutations tend to be advan-
tageous, disadvantageous, or neither. In his foundational
book, The Genetical Theory of Natural Selection (1958),
Fisher argued that most mutations would tend to be disad-
vantageous. But he did so in the context of showing that
despite this fact, evolution by natural selection could still
lead to the fixation of increasingly advantageous genotypes.
There are good reasons why biologists believe that muta-
tions do not tend to be systematically beneficial for organ-
isms. However, these are specific to their understanding of
the causal processes responsible for mutation, and are
independent of the formal structure of evolutionary theory.
And if systematically disadvantageous mutations are com-
patible with ENS, symmetrically, systematically advanta-
geous mutations should also be compatible. Although
systematically advantageous mutations will contribute to
the evolutionary trajectory of the system and therefore form
part of the explanation for any adaptations that occur, this is
nevertheless compatible with part of the explanation for the
trajectory (and any resulting adaptations) being due to
selection. Just as in the more standard case, we see adap-
tation emerging from the effect of selection overwhelming
the maladaptive bias of mutation, we can understand
selection and an adaptive mutational bias as complementing
one another in the alternative case. Recall that for authors
such as Sober (1984) natural selection operates only on
existing variety: if this is accurate, it is clear that any bias in
the process that produces the variety is irrelevant to the
existence and operation of selective process.
It might nevertheless be argued that the problem is not
whether systematically beneficial mutations are compatible
with ENS, but whether variation is ‘‘directed’’ in the sense
of responding sensitively to the needs of the organism. A
random process might generate systematically beneficial
outcomes (as with free lottery tickets). If systematically
beneficial outcomes are generated by a kind of intelligent
process, however, the situation appears different, and per-
haps more problematic, particularly if the novelties are
generated in response to the needs and desires of agents
(Cordes 2006). Variation that responds to the needs of the
individual is Lamarckism, not Darwinism, it might be
claimed. On this view, evolutionary economics begins to
look distinctly Lamarckian. After all, not only did Lamarck
conceive of variations as arising in response to the needs of
the organism, but he saw these directed variations as being
inherited. While some authors accept economic and cul-
tural evolution as unproblematically Lamarckian (Mani
1991; Dopfer 20017), for others who conceive Lamarckism
as incompatible with Darwinian evolution, Lamarckism
disqualifies anything from being legitimately evolutionary.
I regard concern over Lamarckism as a substantial red
herring. The Lamarckian mechanism of evolution involves
two aspects, corresponding to the two ‘‘laws’’ of his Phi-
losophie Zoologique. One law posits the enlargement and
complexification of organs when used and the diminish-
ment of unused organs. These acquired changes are then
thought to be inherited—his second law (Lamarck 1914).
The inheritance of acquired characters is perfectly com-
patible with Darwinism (as can be illustrated by the fact
that Darwin believed that acquired characters were indeed
inherited). From an evolutionary perspective, an ‘‘acquired
character’’ is just a novel character. Indeed even in stan-
dard neo-Darwinism certain acquired characters are
inherited—namely characters that are acquired in the
genetic material, that is, mutations. And while in standard
neo-Darwinist biological scenarios there is little reason to
suppose that heritable acquired characters will be adaptive,
other authors have emphasized that the evolved robust
goal-directed nature of the developmental processes of
organisms tends to make the novel responses of the
organism to environmental insults an adaptive one (Jab-
lonka and Lamb 1995). Evolution that takes place via
inherited adaptive novelty is not evolution via natural
selection. But that does not mean that a model that incor-
porates inherited adaptive novelty cannot also incorporate
evolution by natural selection. It is true that, if individuals
could guide their variation perfectly in response to their
environment, there would be little role for natural selection
to play. Natural selection is a process whereby change in
the population is explained as due to the success of some
variants at the expense of others. Directed variation
explains change in a population as the result of the
7 For both Mani and Dopfer, most emphasis is on the inheritance of
acquired characters, rather than the ‘‘sensitive response’’ to circum-
stance, although I take it both would accept that ‘‘sensitive response’’
in the production of novelty is a common feature of economic
behavior.
Evolution Beyond Biology
123
Author's personal copy
individual response to its environment. Therefore, the lar-
ger the role directed variation plays in explaining popula-
tion change, the smaller the role that is left for natural
selection to fill. However, this should not be seen as a
serious problem. In biology, change can come about from
many sources other than fitness differences.8 The more
change is due to those sources, the less it is due to fitness
differences. Not every model has to assign a large role to
natural selection. However, there is no reason that a fairly
large amount of directed variation cannot coexist in a
model with a large amount of change due to natural
selection. Indeed, that is precisely what I take to be going
on in N&W’s models.
Lamarckism should not be seen as being an alternative to
Darwinian evolution, at least, not in the broad sense of
evolution by natural selection that we are concerned with
here. Darwin himself was a Lamarckian in the sense of
believing in the inheritance of acquired characters. The fact
that in biology many acquired characters cannot be inherited
is because of specific facts about the inheritance mechanisms
of living things in the world (particularly the germ/soma
distinction), not because we could not construct an evolu-
tionary theory that incorporated inheritance of acquired
characters. And while guided variation and natural selection
are in some sense alternative mechanisms of population
change, there is no reason an evolutionary model cannot
include both mechanisms, any more than a biological model
is less ‘‘evolutionary’’ for including migration and mutation
alongside fitness as drivers of population change.
Phenotype and Genotype
Another striking disanalogy between N&W’s evolutionary
economics models and popular views concerning the nat-
ure of biological evolution is that N&W do not have any
real analogue of a distinction between phenotype and
genotype. This distinction plays a central role in many
accounts of biological evolution (Dawkins 1982; Sober
1984; Stephens 2004), as well as some attempts to recon-
struct more general evolutionary principles applicable to
economics (Faber and Proops 1991, Nelson 2001). The
difficulty of shoehorning the distinction between phenotype
and genotype into the economic sphere has struck some
commentators as generally undermining the relevance of
‘‘Darwinian’’ evolution to economic contexts (Witt 2001).
In N&W, firms have production rules and capital stock
and that is it. There is no separate set of traits that are
selected compared to those that are inherited. This is
connected to the already remarked upon ‘‘Lamarckian’’
appearance of N&W’s models: any change that a firm
acquires over time continues to be possessed by that firm at
subsequent times by default. While Nelson in a subsequent
paper reconstructs the routines of the firm as the genotype
and the phenotype as the firm itself, he observes that
‘‘unlike phenotypes (living organisms) that are stuck with
their genes, firms are not stuck with their routines. Indeed
they have built in mechanisms for changing them’’ (Nelson
2001, p. 172). In addition to this disanalogy, in biology the
concept of genotype is linked to the notion of traits that are
inherited via reproduction, and of traits that exist prior to
the developmental process. When biologists speak of
phenotypic traits, they are referring to the traits of the
organism that come about via the process of development.
These are, on the standard understanding, not directly
inherited by the organism’s offspring, but may be indirectly
transmitted via the inheritance of genotypic information
that brings about a similar developmental process in the
progeny. The fact that firms do not develop nor reproduce
in N&W’s models would therefore seem to undercut the
substance of the identification of ‘‘firm’’ with ‘‘phenotype.’’
Since the lack of reproduction by firms is related to the
absence of the phenotype/genotype distinction, part of the
issue must be deferred until we consider the lack of
reproduction more specifically. However, it is still worth
mentioning that it is not uncommon to employ biological
models of evolution that abstract away from the genotype/
phenotype distinction.9 Such models are generally of
asexually reproducing populations where ploidy can be
ignored. Indeed, what may make the distinction between
genotype and phenotype pressing for biologists are factors
such as diploid inheritance, dominance, sexual reproduc-
tion, and (most importantly) the mechanisms of develop-
ment.10 These biological facts demand to be incorporated
into all but the most abstract biological models, and
guarantee that the traits relevant to selection are not just
identical to the heritable traits of an organism’s parents, but
a complicated function of them.
But surely it is clear that the central role of develop-
ment, diploid inheritance, dominance, and sexual repro-
duction are very domain-specific features of terrestrial
biology—and most pressing with regard to animal biology,
as opposed to fungi, bacteria, etc., which make up a large
proportion of the living world. There is no reason to expect
that evolutionary models in other domains must employ
8 So for example, Sober (1984) might say that there are lots of other
evolutionary forces than natural selection, such as drift, migration,
etc., any of which may be responsible for part of the evolutionary
change.
9 See, for example, equations A1 and A2 in Godfrey-Smith (2009,
pp. 165–166).10 This is not to suggest that these are the only possible motivations
for a phenotype/genotype distinction, merely that they may be salient
for biologists.
E. Earnshaw
123
Author's personal copy
these concepts. And the fact that simple biological models
often dispense with the phenotype/genotype distinction
further supports the idea that it is inessential to biological
theorizing generally, and to evolutionary theorizing as
well.11 The key requirement here is that of heredity, which
is related to but independent of the existence of a genotype/
phenotype distinction in the population in question.
Reproduction
The most significant difference between N&W’s evolu-
tionary economics models and the conception of ENS in
biology is the lack of reproduction in N&W’s models.
Biologists might forgivably be tempted to dismiss the
analogy between evolutionary economics and evolutionary
biology entirely at this point, given the centrality of
reproduction to biological evolution. How can natural
selection occur if there is no reproduction? Heritable var-
iation in fitness is supposed to be the recipe for Darwinian
evolution, but the concept might seem nonsensical if there
is no fitness and no heredity. However, I will argue that
there is an analogue for reproduction in N&W’s models,
and that this analogue, while differing from biological
reproduction in some important ways, suffices for natural
selection. That is, the analogue plays a role in N&W’s
models that is explanatorily equivalent to reproduction in
biological models, and the difference, while making a
difference in certain contexts, doesn’t make a difference to
the main evolutionary explanatory projects.
The analogue for reproduction in N&W’s models is the
change over time of the capital stock of each firm. Each
firm at any time has a whole-numbered quantity of capital
stock, and it is the capital stock which, given the produc-
tion technique of the firm, determines the labor inputs and
production output of the company as a whole. As we have
previously observed, N&W provide in their models for the
growth or decrease of firms via change in their capital
stock: unprofitable firms have a tendency to lose capital
stock over time; profitable firms add to it. Firms with zero
capital stock no longer produce any goods and for most
practical purposes can be seen as being ‘‘extinct.’’12 The
capital stock can be seen as being in many respects similar
to individual organisms that share their genotype: each unit
of capital stock has both a cost and revenue; these deter-
mine the profitability, which, analogously to fitness,
determines the probability that there will be more or fewer
individuals with that genotype in the next generation.
If one considers the firm in N&W’s work as analogous
to a population of individuals with a shared genotype, the
apparent divergence between the economic and biological
models largely disappears. Many biological models are
already indifferent to the question of whether each indi-
vidual is replaced from generation to generation,13 or
whether some continue to exist and others enter or leave. It
is just the same with capital stock: N&W assume that the
vast majority of the capital stock continues to exist from
time period to time period, but it is sufficient from the point
of view of the model that the capital stock of the firm at one
time is a function of the capital stock at a previous time.
Both organisms in evolutionary biology and units of capital
stock in evolutionary economics can play the role of
‘‘individuals’’ in a model of evolution by natural selection.
And we can conceive of the change over time in the pop-
ulation as taking place via reproduction, or not, as we see
fit; what is significant from the point of view of our models
is that the ‘‘genotype’’ of the individual fix an average
‘‘fitness’’ value whereby one can determine the change in
size of each distinct genotype group.
In both evolutionary economics and evolutionary biol-
ogy, therefore, we can model change over time as the
growth or decrease of groups of individuals that share a
trait in common: this trait being the genotype for evolu-
tionary biology; the production technique for evolutionary
economics. So long as the model is sufficiently specified,
we can determine the probabilities that some genotypes
shrink and others grow relative to one another, leading to
the evolution of the population via natural selection. We
can thereby explain all the important things that ENS is so
handy for: the spread of a superior trait, the progressive
change towards more and more of whatever is systemati-
cally advantageous, the accumulation of adaptations that
open up the possibility space to the previously inaccessible.
However, what this illustrates is not that reproduction
features in N&W’s models. It shows, rather, that repro-
duction as such is inessential to evolution by natural
selection: it is a mode whereby ENS can occur, but it is not
required for ENS, unless one means reproduction in the
very broadest sense of ‘‘producing more of the same.’’ The
key idea is of growth, and of variation in growth rates
11 My view is that the emphasis on it, and on the interactor/replicator
distinction, and the unit of selection/unit of heredity distinction, is
parasitic on the actually important distinction with regard to ENS, the
distinction between individual and trait. Even the most abstract
characterization of selection appears to require three related elements:
the population, made up of individuals, who possess distinguishing
traits. Change in the distribution of traits in the population constitutes
evolution, but can only occur via changes in the individuals.12 Although in N&W’s chap. 9 model, provision is made for firms
with zero capital to perform research and possibly enter the
marketplace.
13 See, again, equation A1 from Godfrey-Smith (2009, p. 165). It tells
you what the new frequency of the type is, and it is derived from the
assumption of asexual, discrete generations, but it does not specify
whether a ‘‘discrete’’ generation involves the parent dying, or budding
off children and surviving itself, or engaging in cellular division such
that there is no answer to the question of whether the parent survives.
Evolution Beyond Biology
123
Author's personal copy
(Matthen and Ariew 2002). After all, an evolutionary
biology model can remain silent about how exactly indi-
viduals are entering the population, and whether the parent
individuals are replaced by children or continue to survive
themselves. Indeed, in cases such as cell division, it is not
clear whether such questions even have answers.
The key requirement is that the model must have some
mechanism whereby the prevalence in the population of a
type of individual can change, and that the tendency of a
type to change is influenced by some kind of competitive
advantage vis-a-vis its alternatives. In standard evolution-
ary models, this is satisfied by assigning each type a ‘‘fit-
ness’’ in the environment. In N&W, this is satisfied by
calculating the profitability of each production process in
the environment. So long as fitness and profitability feed
back in to affect the prevalence of the type in the popu-
lation, we have evolution by natural selection.
This approach offers a solution to the difficult problem
of distinguishing reproduction from growth. Because var-
ious plants can create new ‘‘individuals’’ via processes that
look a lot more like growth than reproduction, it is an
extremely difficult question of how to understand the nat-
ure of populations such as dandelions or aspen trees in the
context of evolution. If what counts is prevalence in the
environment, however, we do not have to keep rigid track
of an arbitrarily defined individual, nor do we have to
discount the spread of a type through processes of growth
rather than reproduction. Whether we count the aspen as
one large growing individual or many reproducing indi-
viduals is analogous to whether we treat the corporation as
one large individual that grows via adding capital stock, or
as many units of capital stock that add to their number in
some fashion or other. In either case, a very large tree (or
corporation) represents success, however we choose to
analyze it in terms of growth or reproduction. Indeed, the
mode of representation can be chosen depending on the
purposes of the modeling enterprise.
This reconstituted, very abstract conception of repro-
duction is matched by a similarly revised and attenuated
notion of heredity. N&W state that, ‘‘The tendency for…routines to be maintained over time in our theory plays the
role that genetic inheritance plays in the theory of bio-
logical evolution’’ (p. 142). In other words, heredity simply
consists in the tendency for firms to exist over an extended
time period, and to maintain the same routine by default.
This continuity over time is required if natural selection is
to play a role in explaining evolutionary change. Without a
backdrop of stability, the success of one variant over
another is a meaningless concept: the very concept of the
success of a variant presupposed that a variant continues in
existence and that the prevalence of a variant at one time
depends on its prevalence at previous times. N&W quite
rightly adopt this criterion of stability as the key
contribution of biological heredity to evolutionary models.
Biological heredity is the means by which biological sys-
tems maintain the stability necessary for evolution; but it is
stability, not heredity, which is essential to natural selec-
tion considered in the abstract.
A contrary view has been advanced by Godfrey-Smith
(2009) who argues that growth and persistence are both very
marginal cases of ‘‘reproduction,’’ in the sense required for
ENS. His objection is centered around the notion of mul-
tiplication. Without multiplication of entities in the popu-
lation, he argues that the Darwinian possibilities of the
population are extremely limited—in particular, ‘‘origin
explanations’’ of adaptive novelty (which proceed, as dis-
cussed above, by making the emergence of an adaptive
novelty more likely) are ruled out. For cases where the
population can only shrink, Godfrey-Smith’s point is a fair
one. However, as N&W’s model demonstrates, there is no
reason why a model lacking reproduction cannot involve
growth in the population. This perhaps illustrates an
advantage of looking at cases outside of biology, although
species selection suggests a similar moral to me. The focus
on reproduction as the only mode of population change
obscures in Godfrey-Smith’s analysis that growth can be
understood as increasing the size of the population. One can
measure the size of a population in terms of head count, but
that isn’t the only mode available. When we say the pop-
ulation of America is larger than it was thirty years ago,
head count is not the only available interpretation. A species
can grow without speciating: it can increase its overall
prevalence, and that can legitimately be understood as an
evolutionary change. This strikes me as a rather more robust
understanding of the evolutionary success of a species than
one that looks purely at ‘‘child species.’’ And we have
already seen that N&W’s model does support origin
explanations, which undermines the idea that growth and
persistence can’t do the job. So it seems to me that, on the
explanatory criterion that Godfrey-Smith himself employs,
the centrality of reproduction to ENS cannot be defended.
This is not to say that the lack of reproduction and
heredity in N&W is insignificant. I will discuss some of the
significant differences momentarily. In particular, it has
consequences in terms of how we understand novelty to be
realized. But such differences are, one might say, mere
variants; the models are all models of ENS, by virtue of
specifying the population dynamics of a group of individ-
uals divided into classes of variants via the feedback of
competitive advantage into prevalence.
Beyond Biology
Probably the most significant difference from the biologi-
cal to be found in N&W’s models has to do with the
E. Earnshaw
123
Author's personal copy
association of units of capital stock in the larger entity of
the firm. Earlier I interpreted the units of capital stock as
the ‘‘individuals’’ in the model in order to draw the clearest
parallel with individuals in evolutionary biology. However,
that is not how N&W conceive their models: they see their
model as treating a population of firms, not a population of
capital stock, and they conceive natural selection as
applying to these firms via the growth or contraction of the
firm. I hope it is clear that these two perspectives are not
contradictory: insofar as a firm’s growth or contraction is
constituted by its gaining or losing units of capital stock,
both perspectives identify natural selection with such gains
or losses. The difference is in the terminology used to
describe the same process.
Regardless of how we choose to describe the process, it
has an interesting property as compared to standard bio-
logical models: because ‘‘mutation’’ (that is, the research
and adoption of a new production technique) is a process
carried out by the firm as a whole, N&W model it as being
instantaneously and costlessly applied to all of the firm’s
units of capital stock. If we view the model with units of
capital stock as our individuals, this means that all indi-
viduals that share a genotype mutate in tandem. Such a
linkage is clearly not something we would expect to see in
biology. This has the consequence that more prevalent
production techniques are not subject to more production
of novelty (unless one adds the idea of a ‘‘research budget’’
to the model), in the way that we would expect a successful
biological type to be subject to more variation in virtue of
its ubiquity (Darwin 1859). However, it is perfectly clear
why N&W choose to model change the way that they do.
They explicitly avow that instantaneous and costless tran-
sition to a new technique is an idealization, and suggest
that exploring more realistic models might be worthwhile,
but state that the assumption is harmless and warranted by
gains in simplicity and tractability. And it is clear why
N&W believe that a firm adopting a new practice will try to
implement it across the entire company, just as it is clear
why the cost and slowness of this process might be treated
as a sort of ‘‘friction’’ that can in some cases be ignored for
simplicity’s sake.
To be more specific, the choices made by N&W in
building their model, while in this case diverging quite
dramatically from what would seem realistic in the context
of biology, are made based on the specific context of the
domain they are investigating. Reproduction and heredity
are handled very differently—to the extent that both
essentially vanish from the models. But it is precisely
because N&W do away with reproduction and heredity in
any standard biological sense that examining their models
is so instructive. Both these concepts are revealed to be
domain-specific artifacts of biological theory specially:
neither is required for a functional evolutionary model.
The ‘‘recipe’’ for natural selection is often said to be
‘‘heritable variation in fitness.’’ I suggest that this is really
just the recipe for selection in biology. If fitness is under-
stood as referring to reproduction, it is dispensable.
‘‘Heritable variation in fitness’’ could perhaps be better
rendered, therefore, as ‘‘variation in stable tendencies to
growth.’’
Conclusion
N&W were impelled to their evolutionary models by dis-
satisfaction with economic theorizing about change and
innovation. While inspired by Darwinian evolutionary
biology, they developed sophisticated models distinctively
their own. These models are evolutionary in the full sense
of the term, despite diverging strikingly from models of
evolutionary biology. Analysis of their models suggests
that we revise some traditional conceptions of the nature of
evolution by natural selection. Variation need not be blind,
there is no need for a firm phenotype/genotype distinction,
and reproduction and heredity are required only in a highly
attenuated sense. These results shed light both on the
prospects and methods by which evolutionary thinking may
be extended to yet further domains, as well as revising our
view of the nature of selection within biology itself.
References
Campbell DT (1960) Blind variation and selective retentions in
creative thought as in other knowledge processes. Psychol Rev
67(6):380–400
Cordes C (2006) Darwinism in economics: from analogy to conti-
nuity. J Evolution Econ 16:529–541
Darwin C (1859) On the origin of species. John Murray, London
Dawkins R (1976) The selfish gene. Oxford University Press, New
York
Dawkins R (1982) The extended phenotype. Oxford University Press,
Oxford
Dopfer K (2001) Evolutionary economics: framework for analysis. In:
Dopfer K (ed) Evolutionary economics: program and scope.
Kluwer, Dordrecht, pp 1–44
Faber M, Proops JLR (1991) Evolution in biology, physics and
economics: a conceptual analysis. In: Saviotti PP, Metcalfe JS
(eds) Evolutionary theories of economic and technological
change: present status and future prospects. Harwood,
Philadelphia
Fisher RA (1958) The genetical theory of natural selection. Dover,
New York
Foster PL (2007) Stress-induced mutation in bacteria. Crit Rev
Biochem Mol Biol 42(5):373–397
Godfrey-Smith P (2009) Darwinian populations and natural selection.
Oxford University Press, New York
Hodgson GM (2002) Darwinism in economics: from analogy to
ontology. J Evolution Econ 12:259–281
Hull DL (1988) Science as a process: an evolutionary account of the
social and conceptual development of science. University of
Chicago Press, Chicago
Evolution Beyond Biology
123
Author's personal copy
Hull DL, Langman R, Glenn S (2001) A general account of selection:
biology, immunology, and behaviour. Behav Brain Sci 24:511–
528
Jablonka E, Lamb MJ (1995) Epigenetic inheritance and evolution:
the Lamarckian dimension. Oxford University Press, Oxford
Lamarck JB (1914) Zoological philosophy: an exposition with regard
to the natural history of animals (trans: Hugh E). MacMillan,
London
Lewontin RC (1970) The units of selection. Annu Rev Ecol Syst
1:1–18
Mani GS (1991) Is there a general theory of biological evolution? In:
Saviotti PP, Metcalfe JS (eds) Evolutionary theories of economic
and technological change: present status and future prospects.
Harwood Academic Publishers, Philadelphia
Matthen M, Ariew A (2002) Two ways of thinking about fitness and
natural selection. J Philos 99(2):55–84
Nelson R (2001) Evolutionary perspectives on economic growth. In:
Dopfer K (ed) Evolutionary economics: program and scope.
Kluwer, Boston
Nelson R, Winter S (1982) An evolutionary theory of economic
change. Belknap Press of Harvard University Press, Cambridge
Sober E (1984) The nature of selection. MIT Press, Cambridge
Stephens C (2004) Selection, drift, and the ‘‘forces’’ of evolution.
Philos Sci 71:550–570
Witt U (2001) Evolutionary economics: an interpretative survey. In:
Dopfer K (ed) Evolutionary economics: program and scope.
Kluwer, Boston
E. Earnshaw
123
Author's personal copy