J Evol Econ (1995) 5:243 268 - - J o u r n a l of

Evolutionary
Economics
9 Springer-Verlag 1995

Technological paradigms, patterns of learning

and development: an introductory roadmap
Mario Cimoli 1 and Giovanni Dosi 2

1 Department of Economics, University of Venice, Ca'Foscari, Dirsodurro 3246,1-30123 Venezia,

Italy
2 Department of Economics, University of Rome, "La Sapienza", 1-00100 Roma, Italy

Abstract. This paper presents an evolutionary microeconomic theory of innovation

and production and discusses its implications for development theory. Using the
notions of technological paradigm and trajectory, it develops an alternative view of
firm behavior and learning. It is shown then how these are embedded in broader
national systems of innovation which account for persistent differences in techno-
logical capacities between countries. Finally, this "bottom-up" evolutionary analy-
sis is linked with an institutional "top-down" approach, and the potential fruitful-
ness of this dialogue is demonstrated.

Key words: Technological paradigms - Technological change - Theory of

innovations - National systems of innovation

JEL-classification: O 14-O30-O53-O54

1. Introduction

Deep relationships of some sorts between technical change and economic develop-
ment are now generally acknowledged in both economic history and economic
theory. Still, their nature is matter of debate concerning the precise causal links. For
example, it is quite intuitive that improvements in the efficiency of techniques of
production or in product performances may be a determinant or at least a binding
precondition of growth in per capita incomes and consumption. But, intricate

Correspondence to: M. Cimoli

We thank the discussants and participants at the workshop on Technology and Competitiveness in
Developing Countries, Venice, 26/11/93, for their useful comments.
The research leading to this work has benefited at various stages from the support of the Italian
National Research Council (CNR, Progetto Strategico "Combiamento Technologico e Crescita
Economica") and of the International Institute of Applied System AnaIysis (IIASA, Austria).
244 M. Cirnoliand G. Dosi

debates concern "what ultimately determines what...": e.g. is it resource accumula-

tion that primarily fosters the exploration of novel innovative opportunities, or,
conversely, does innovation drive capital accumulation?; do new technological
opportunities emerge mainly from an extra-economic domain ("pure science") or are
they primarily driven by economic incentives?; should one assume that the institu-
tions supporting technical change are sufficiently adaptive to adjust to whatever
underlying economic dynamics emerges from market interactions; or, conversely, are
they inertial enough to shape the rates and directions of innovation and diffusion?
Clearly, these and a few other, related, questions are at the core of many
controversies regarding growth patterns: for example, is convergence the dominant
tendency? How does one then interpret observed phenomena of forging ahead or
falling behind? Is it legitimate to exclude from the analysis at least in a first
approximation the specificities of institutions and corporate organizations? Even
more so, all these questions and controversies underlie the political economy of
development.
Obviously, one would not do justice to these intricate questions in a single paper
even if one had achieved thorough answers (that indeed one is far from having).
However, there has been over at least the last two decades a flourishing of studies on
the sources, mechanisms and patterns of technological innovation. And, the opening
of the technological blackbox has often gone together with important insights into
innovation-driven market competition. Business historians have finally achieved
some cross-fertilization with (some breeds of) economic theorizing. And the institu-
tional understanding of the socio-economic fabrics of contemporary societies starts
showing fruitful complementaries with other analyses stemming from the econom-
ists quarters.
Quite a few of these contributions have been proposed by scholars who would
call themselves evolutionists or institutionalists. Many, others have come within
different theoretical perspectives. Still, there is a sense that these diverse streams of
research show a few common threads, highlighting - to paraphrase Richard Nelson
the co-evolution of technologies, corporate organizations and institutions. These
threads - linking evolutionary analyses of the microeconomics of innovation all the
way to (daring) generalizations on some invariant features of the process of
development - are the subject of this paper. Far from being a comprehensive survey,
it is rather a sort of "roadmap" with an inevitable degree of idiosyncrasy.
We start by discussing the theoretical implications of what we know about the
often patterned dynamics of innovative activities at a micro level. The notions of
technological paradigms, trajectories (and largely overlapping ones such as domi-
nant designs) entail a representation of technologies centered on the cognitive and
problem-solving procedures which they involve.
Another major implication of this view is in terms of theory of production. It is
rather straightforward to derive some sort of non-substitution properties, in the
short-term, and, also in the long-term, technological asymmetries or gaps as
permanent features across firms and, even more as, across countries.
Do these micro technological properties bear consequences at broader levels of
observation, i.e. whole industrial sectors and whole countries? Or, putting it another
way, can one identify invariances and patterns at sectoral or national level which can
be interpreted in terms of some underlying specificities in the processes of collective
learning, market selection and institutional governance of both?
This is the subject of the second part of the paper, and it is also where the
roadmap inevitably bifurcates into different discourses. Some will be persued in
Technological paradigms, patterns of learning and development 245

reasonable detail and while others will only be sketched out, just flagging the
elements of consistency with the rest of the argument. For example, there are sound
theoretical reasons and a growing empirical evidence that the observed patterns of
evolution of industrial structures are the outcome of specific modes of access to
innovative opportunities and market selection mechanisms. However, we shall not
dwell here on this aspect of the co-evolution between technologies and production
structures. Rather more attention shall be devoted to the links between micro
learning and economy-wide accumulation of technological capabilities and, in
particular, to the existence of specific national system of production and innovation.
The argument needs to be built through several steps. First, it follows from the
microeconomics of innovation that firms are central, albeit by no means unique,
repositories of technological knowledge. Hence, also their specific organizational
and behavioral features affect the rates and direction of learning. Second, firms
characteristics are not randomly distributed across sectors and across countries. On
the contrary, particular traits tend to be reinforced through their interactions with
the environment in which they are imbedded. Third, broad institutional mechan-
isms of governance of interactions further enhance the possibility of collective
lock-in into particular modes of learning. Somewhat in analogy with the earlier
microeconomic analysis we shall call these patterns as national trajectories.
Far from reviewing an immense historical evidence on these issues, we shall only
draw from selected examples from developed countries and, in particular, from
a somewhat archetypical comparison between the experiences of Latin America and
the Asian Far East.
Along this tour de Jbrce from micro technological studies to the political
economy of development, we shall on purpose raise many more questions than we
shall able to answer. The major task here is to show that they can be consistently
linked together in a broadly defined evolutionary interpretation.

2. The fundamental properties of technology

Technological paradigms and trajectories

A variety of concepts have recently been put forward to define the nature of
innovative activities: technological regimes, paradigms, trajectories, salients, guide-
posts, dominants designs and so on. The names are not so important (although some
standardization could make the diffusion of ideas easier!). More crucially, these
concepts are highly overlapping in that they try to capture a few common features of
the procedures and direction of technical change (for a discussion and references, see
Dosi 1988). Let us consider some of them. 1
The notion of technological paradigms is based on a view of technology
grounded on the following three fundamental ideas.
First, it suggests that any satisfactory description of "what is technology" and
how it changes must also embody the representation of the specific forms of
knowledge on which a particular activity is based. Putting it more emphatically,
technology cannot be reduced to the standard view of a set of well-defined
blueprints. Rather, it primarily concerns problem-solving activities involving - to

i •nthef••••wing•wesha••stickt•thecateg•ries•f•aradigmsandtraject•ries•butthereaderwh•
is found of other names should still recognizefamiliar ideas.
246 M. Cimoli and G. Dosi

varying degrees-also tacit forms of knowledge embodied in individuals and

organizational procedures.
Second, paradigms entail specific heuristic and visions on "how to do things"
and how to improve them, often shared by the community of practitioners in each
particular activity (engineers, firms, technical societies, etc.), i.e. they entail a collec-
tively shared cognitive frames (Constant 1985).
Third, paradigms generally also define basic models of artifacts and systems,
which over time are progressively modified and improved. These basic artifacts can
also be described in terms of some fundamental technological and economic
characteristics. For example, in the case of an airplane, these basic attributes are
described not only and obviously in terms of inputs and the production costs, but
also on the basis of some salient technological features such as wing-load, take-off
weight, speed, distance it can cover, etc. What is interesting is that technical progress
seems to display patterns and invariances in terms of these product characteristics.
Similar examples of technological invariances can be found e.g. in semiconductors,
agricultural equipment, automobiles and a few other micro technological studies.
The concept of technological trajectories is associated to the progressive realiza-
tion of the innovative opportunities associated with each paradigm, which can in
principle be measured in terms of the changes in the fundamental techno-economic
characteristics of artifacts and the production process. The core ideas involved in
this notion of trajectories are the following. 2
First, each particular body of knowledge (i.e. each paradigm) shapes and
constraints the rates and direction of technological change irrespectively of market
inducements. Second, as a consequence, one should be able to observe regularities
and invariances in the pattern of technical change which hold under different market
conditions (e.g. under different relative prices) and whose disruption is correlated
with radical changes in knowledge-bases (in paradigms). Third, technical change is
partly driven by repeated attempts to cope with technological imbalances which it
itself creates.3
A general property, by now widely acknowledged in the innovation literature, is
that learning is local and cumulative. Local means that the exploration and
development of new techniques is likely to occur in the neighborhood of the
techniques already in use. Cumulative means that current technological develop-
ment - at least at the level of individual business units - often builds upon past
experiences of production and innovation, and it proceeds via sequences of specific
problem-solving junctures (Vincenti 1992). Clearly, this goes very well together with
the ideas of paradigmatic knowledge and the ensuing trajectories. A crucial implica-
tion, however, is that at any point in time the agents involved in a particular
production activity will face little scope for substitution among techniques, if by that
we mean the easy availability of blueprints different from those actually in use, which
could be put efficiently into operation according to relative input prices.

2 The interpretation of technical change and a number of historical examples can be found in
pioneering works on economics of technical change such as those by Chris Freeman, Nathan
Rosenberg, Richard Nelson, Sidney Winter, Thomas Hughes, Paul David, Joel Mokyr, Paolo
Saviotti and others; see for a partial survey Dosi (1988).
3 This is akin to the notion of reverse salients (Hughes 1992) and technological bottlenecks
(Rosenberg 1976): to illustrate, think of increasing the speed of a machine tool, which in turn
demands changes in cutting materials, which lead to changes in other parts of the machine...
Technological paradigms, patterns of learning and development 247

Technological dominance, micro heterogeneity and non-substitution

The notion of paradigms contains elements of both a theory of production and
theory of innovation. In short, we shall call it henceforth an evolutionary theory.
Loosely speaking, we should consider such a theory at the same level of abstraction
as, say, a Cobb-Douglas production function or a production possibility set. That is,
all of them are theories of what are deemed to be some stylized but fundamental
features of technology and, relatedly, of production process. 4
In fact, one finds a few remarkable assumptions underlying conventional pro-
duction theories. As already mentioned, technologies - at least in a first approxi-
mation - are seen as a set of blueprints describing alternative input combinations.
Moreover, at any one time there must be many of them, in order to be able to
interpret empirical observations as the outcome of a microeconomic process of
optimal adjustment to relative prices. Information about these blueprints is gen-
erally assumed to be freely available (except those circumstances whereby they are
privately appropriated via the patent system). Finally, one assumes to be able to
separate the activities leading to the efficient exploitation of existing blueprints from
those leading to the development of new ones (exogeneity of technical progress is its
extreme version). Of course, this is only a trivialized account of a family of models
that can be made much more sophisticated, by e.g. adding details on how blueprints
are ordered with respect to each other (more technically, issues like continuity and
convexity come under this heading). However, it still seems fair to say that the basic
vision of production- also carried over in aggregate growth and development
models - focuses on questions of choice among well defined techniques, generally
available to all producers, who also know perfectly well what to do with all the
recipes when they see them.
Well, to put it very strongly, the theory of production based on paradigms
develops on nearly opposite theoretical building blocks. And indeed many of the
latter yield empirically testable hypotheses.
Here, we shall argue that a paradigm-based theory of technology may perform
the same interpretive tasks, at the same level of generality, and do it better, in the
sense that it is more in tune with microeconomic evidence and also directly links
with theories of innovation. Our theory would predict the following.
a) In general, there is at any point in time one or very few best practice techniques
which dominate the others irrespectively of relative prices.
b) Different agents are characterized by persistently diverse (better and worse)
techniques.
c) Over time the observed aggregate dynamics of technical coefficients in each
particular activity is the joint outcome of the process of imitation/diffusion of
existing best-practice techniques, of the search for new ones, and of market
selection amongst heterogeneous agents.
d) Changes over time of the best practice techniques themselves highlight rather
regular paths (i.e. trajectories) both in the space of input coefficients and also in
the space of the core technical characteristics of outputs (see the earlier example
on aircrafts).

Few believe that a production possibility set literally exists. Many would however probably
maintain that such a notion enhances the understanding of the observed technical coefficientsin
the economy and also how they change over time. We claim the same for the evolution theory.
248 M. Cimoli a n d G. Dosi

A representation of production and technological activities

Let us further illustrate the previous points with a graphical example.
Start from the notion that each technical coefficient observed at the microlevel is
the outcome of codified information (something resembling blueprints), but also of
more tacit and firm specific forms of knowledge. Suppose that, for the sake of
simplicity, we are considering here the production of an homogeneous good under
constant returns to scale with two variable inputs only, x 1 and x2 .5
A paradigm-based theory of production predicts that, in general, in the space of
unit inputs, micro coefficients are distributed somewhat as depicted in Figure 1.
Suppose that at time t the coefficients are cl ..... c,; where 1. . . . . n are the various
techniques/firms labelled in order of decreasing efficiency at time t. It is straightfor-
ward that technique/firm c~ is unequivocally superior to the other ones no matter
what relative prices are: it can produce the same unit output with less inputs of both
x 1 and x 2. The same applies to the comparison between c 3 and c,, etc ....
Let us call this property technological dominance, and call some measure of the
distribution of the coefficients across heterogeneous firms as the degree of asymmetry
of that industry (for example, the standard deviation around the mean value C).
The first question is why doesn't firm n adopt technique cl? To simplify a more
articulated argument (see Freeman 1982, Nelson and Winter 1982, Dosi 1988 and
Dosi, Pavitt and Soete 1990), the answer is "because it does not know how to do
it...". That is, even if it is informed about the existence of cl, it might not have the
capabilities of developing or using it. Remarkably, this might have little to do with
the possibility for c I to be legally covered by a patent. The argument is much more
general: precisely because technological knowledge is partly tacit, also embodied in
complex organizational practices, etc., technological lags and lead may well be
persistent even without legal appropriation. The opposite also holds: if the two firms
have similar technological capabilities, imitation might occur very quickly, patent

X2
: I~

/ 0r . .....
Ix oct3 , ,,"" ....... .... 1 ~

Xi

Fig. 1. Microheterogeneity and technological trajectories

5 N o t e that fixed inputs, vintage effects and economies of scale would iust strenghten the argument.
Technological paradigms, patterns of learning and development 249

protection notwithstanding, by means of "inventing around" a patent, reverse

engineering, etc..
We are prepared to push the argument further and suggest that even if firm
n were given all the blueprints of technique c x (or, in a more general case, also all the
pieces of capital equipment associated with it), performances and thus revealed
input coefficients might still widely differ. Following R. Nelson, it is easy to illustrate
this by means of a gastronomical metaphor: despite readily available cooking
blueprints and, indeed, also codified rules on technical procedures, unavailable in
most economic representations of production ("... first heat the oven, then after
around ten minutes introduce some specified mixture of flour and butter,.., etc."),
one obtains systematically asymmetric outcomes in terms of widely shared stan-
dards of food quality. This applies to comparisons among individual agents and also
to institutionally differentiated groups of them: for example, we are ready to bet that
most eaters randomly extracted from the world population would systematically
rank samples of English cooks to be "worse" than French, Chinese, Italian,
Indian .... ones, even when performing on identical recipes!!. If one accepts the
metaphor, this should apply, much more so, to circumstances whereby performan-
ces result from highly complex and opaque organizational routines (Incidentally,
Leibenstein's X-efficiency rest also upon this widespread phenomenon).
Suppose now that at some subsequent time t' we observe the distribution of
t
microcoefficients c3,... , c,,. How do we interpret such a change?
The paradigm-based story would roughly be the following. At time t, all
below-best-practice firms try with varying success to imitate technological leader(s).
Moreover, firms change their market shares, some may die and other may enter: all
this obviously changes the weights (i.e. the relative frequencies) by which tech-
niques/firms appear. Finally, at least some of the firms try to discover new
techniques, prompted by the perception of innovative opportunities, irrespectively
of whether relative prices change or not (for the sake of illustration, in Figure 1,
firm-3 succeeds in leapfrogging and becomes the technological leader while firm-m
now embodies the marginal technique).
How do relative prices fit into this picture?
In a first approximation, no price-related substitution among firm-known
blueprints occurs at all. Rather, changes in relative prices primarily affect both the
direction of imitation and the innovative search by bounded-rational agents.
However, the paradigm-based story would maintain that, even if relative prices
change significantly, the direction of innovative search and the resulting trajectories
would remain bounded within some relatively narrow paths determined by the
nature of the underlying knowledge base, the physical and chemical principles it
exploits, the technological systems in which a particular activity is embodied. Still
more importantly, persistent shocks on relative prices, or, for that matter, on
demand conditions, are likely to exert irreversible effects on the choice and relative
diffusion of alternative technological paradigms, whenever such an alternative
exists, and, in the long term, focus the search for new ones.
In a extreme synthesis, a paradigm-based production theory suggests as the
general case, in the short term, fixed-coefficient (Leontieff-type) techniques, with
respect to both individual firms and industries, the latter showing rather inertial
averages over heterogeneous firms. In the longer term, we should observe quite
patterned changes, often only loosely correlated with the dynamics of relative prices.
In fact, the available evidence - admittedly scattered, due also to the economists
propensity to avoid disturbing questions - is consistent with these conjectures: there
250 M. Cirnoli and G. Dosi

appear to be wide and persistent asymmetries in efficiency among firms within the
same industry (cf. for a survey and discussion Nelson 1981). This applies to
developed countries and, more so, to developing ones. Moreover, persistent asym-
metries appear also in profitability (Geroski and Jaquemin 1988, Muller 1990).
Finally, several industrial case studies highlight technology-specific regularities in
the patterns of technical change hardly interpretable as direct responses to changes
in relative prices and demand conditions: in this respect, the case of the semiconduc-
tors (Dosi 1984) is only an extreme example of a more general phenomenon.
Let us now expand the space over which technologies are described and include,
in addition to input requirements, also the core characteristics of process and
artifacts, hinted earlier: e.g. wing-load, take off weight etc. in airplanes; circuit
density, processing speed in semiconductors; acceleration, fuel consumption in
automobiles; etc. The conjecture is that also in this higher-dimension space,
trajectories appear and that discontinuities are associated with changes in knowl-
edge bases and search heuristics. Indeed, the evidence put forward by e.g. Devandra
Sahal and, more recently, by Paolo Saviotti at Manchester University show
remarkable regularities in the patterns of change within the space of core product
characteristics: for example, in commercial aircrafts, one can observe a well defined
trajectory leading from the DC-3 to contemporary models. (Interestingly, models
which turn out to be technological or commercial failures often happen to be far
from the trajectory itself).
These findings bear implications also for the economic analysis of the relation-
ship between supply and demand dynamics. Start from a Lancasterian view of final
demand (i.e., consumers demand characteristics which satisfy their "needs"). With
rising incomes and heterogeneous preferences, one might have expected product
variety to grow and be distributed over the whole space of characteristics. In
fact, one obviously observe an enormous product variety. However, at a closer
look, it appears that product innovation explore only a minor sub-set of such
a space. Putting it differently, the nature of each paradigm appears to be a powerful
factor binding the variety in the technical features and performances of observed
products.

Technical change, international asymmetries and development

Naturally, there is an alternative interpretation of all the evidence discussed so far
drawing on standard production theory. Let us consider once more Figure 1. Take
for example the average technical coefficient C at time t by reading it from published
industrial statistics. Assume by definition that C is the equilibrium technique
(whereby average and best practice techniques nearly coincide). Relatedly, draw
some generic and unobservable downward-sloped curve through C (say, in Fig. 1 the
II curve) and also the observed relative price ratio. Do the same with point C'
corresponding to the average values at t', and again with the subsequent average
observations. Next assume a particular functional form to the unobserved curve
postuled to pass through C, C', C',..., etc. and call it the isoquant of a corresponding
production function. (The same method can be applied, of course, over time or
cross-sectionally). Then, run some econometric estimates based on such postulated
function, using data derived from the time-series of relative prices and C, C',...
Finally, interpret the relationship between the values of the estimated coefficients in
terms of elasticities of substitution (i.e. some notional movement along the II curve,
as equilibrium responses to relative price changes), and attribute the residual
Technologicalparadigms, patterns of learning and development 251

variance to a drift in the technological opportunity set, as represented by the

movement from II to IT, etc..
For the purpose of this argument, one can neglect whether such a drift is meant
to be an exogenous time-dependent dynamics - as in Solow-type growth models -,
or is in turn the outcome of some higher level production function of blueprints - as
in many new growth models. In any case, i f - for whatever reasons - relative prices
present some intertemporal regularity and so do pattems of technological search (for
example because they follow paradigm-driven trajectories), then one is likely to find
a good statistical fit to the postulated model, even when no causal link actually exists
between distributive shares and factor intensities. This is a well established point,
convincingly argued in different perspectives by F. Fischer, R. Nelson, L. Pasinetti,
A. Shaikh, H. Simon. Even if the evolutionary microdynamics described above were
the true one, one could still successfully undertake the standard statistical exercise of
fitting some production function. But the exercise would in fact obscure rather than
illuminate the underlying links between technical change and output growth.
Take the illustration of Fig. 1 and suppose that the evidence does not refer to two
distributions of micro-technical coefficients over time within the same country, but
instead to two countries at the same time: after all, paraphrasing Robert Lucas, we
only need informed tourists to recognize that most countries can be ranked in terms
of unequivocal average technological gaps. With some additional assumptions on
the nature of production function, one can still claim that C, C', etc. remain
equilibrium realizations of country-specific allocation processes. Conversely, in the
context of an evolutionary approach, one would suggest as we do - that optimiz-
ing choice among technical alternatives commonly shared by all agents in the two
countries have little to do with all this, and that one should rather look for an
explanation of such inter-national differences within the process of accumulation of
technological competence and also within the institution governing market interac-
tion and collective learning. The contrast between (imperfect) learning vs optimal
allocation of resources as the fundamental engine of development has indeed been
repeatedly emphasized among others by Kaldor, Pasinetti and earlier by Schum-
peter, but to our knowledge, no-one has yet fully explored its consequences for the
theory and policy of development. Needless to say, we are dramatizing the differen-
ces. After all, learning is intertwined with the process of resource allocation. Still, it is
useful to distinguish between what is assumed as having first order or second order
effects.
All this has also an empirical counterpart: indeed, the economic discipline has
undertaken far too few exercises at the highest available disaggregation on interna-
tional comparisons among sectoral technical coefficients. Our conjecture is that, at
this level, one could observe a good deal of evidence conflicting with the standard
theory of production: less developed countries may well show higher utilization of
all or most inputs per unit of output and perhaps even higher relative intensity of
those inputs that the theory would consider more scarce (that is, some loose
equivalent of what euphemistically the economic profession calls in international
trade the Leontieff paradox). Conversely, an evolutionary interpretation is straight-
forward: unequivocal technological gaps account for generalized differences in input
efficiencies. Moreover, if technical progress happens to involve also high rates of
saving in physical capital and skilled-labour inputs, one may also observe less
developed countries which do not only use more capital per unit of output but
also more capital per unit of labour input as compared to technological leaders
(Figure 1 illustrates a similar case: compare for example, techniques c 3 and cn).
252 M. Cimoli and G. Dosi

Some important implications emerge from this approach.

First, the theory would predict persistent asymmetries among countries in the
production processes which they are able to master (this of course also shows up in
terms of different inputs effficiencies: see Dosi, Pavitt and Soete 1990). Thus, at any
point in time, one can draw two major testable conjectures: (i) different countries
might well be unequivocally ranked according to the efficiencies of their average
techniques of production and, in the product space, of the (price-weighted) perform-
ance characteristics of their outputs, irrespectively of relative prices, and (ii) the
absence of any significant relationship between these gaps and international dif-
ferences in the capital/output ratios. Wide differences apply also to the capabilities
of developing new products and to different time lags in producing them after they
have been introduced into the world economy. Indeed, the international distribu-
tion of innovative capabilities regarding new products is at least as uneven as that
regarding production processes. For example if one takes international patents or
the number discrete innovation as a proxy for innovativeness, the evidence suggest
that the club of the innovators has been restricted over the whole past century to
a dozen developed countries with only one major new entry, Japan (more on the
evidence in Dosi, Pavitt and Soete 1990).
Second, the process of development and industrialization are strictly linked to
the inter- and intra-national diffusion of"superior" techniques. Relatedly, as already
mentioned, at any point in time, there is likely to be only one or, at most, very few
"best practice" techniques of production which correspond to the technological
frontier. In the case of developing economies, the process of industrialization is thus
closely linked with the borrowing, imitation, adaptation of established technologies
from more advanced economies. These process of adoption and adaptation of
technologies, in turn, are influenced by the specific capabilities of each economy. 6
In this context, we suggest that evolutionary micro-theories are well apt to
account for the processes by which technological gaps and national institutional
diversities can jointly reproduce themselves over rather long spans of time. Con-
versely, in other circumstances, it might be precisely this institutional and techno-
logical diversity among countries which may foster catching-up (and, rarely
leapfrogging) in innovative capabilities and the per capita incomes. Rigorous
demonstrations of these propositions would indeed require many intermediate
steps, linking the externalities and positive feedback mechanisms based on techno-
logical learning with the institutional context in which microeconomic agents are
embedded, and also the economic signals they face. We shall briefly come back to
this issue later on. Here let us just emphasize that systematically different rates of
learning may have very little to do with "how well markets work". Rather, the
incentives and opportunities which agents perceive in a particular context are
themselves the result of particular histories of technologies and institutions.
The importance of the institutional dimension for evolutionary theories of
production and innovation should come as no surprise, supported by a growing
evidence from both micro and macro patterns of technological change. After all, at
the micro level, technologies are to a fair extent incorporated in particular institu-
tions, the firms, whose characteristics, decision rules, capabilities, and behaviors are
fundamental in shaping the rates and directions of technological advance. In turn,

6 Abramovitz notion of differentiated "social capabilities" is quite consistent with this view,
Abramovitz (1989).
Technological paradigms, patterns of learning and development 253

firms are embedded in rich networks of relations with each other and with other
institutional actors - ranging from government agencies to universities etc...7
But how did particular technologies come into exist existence in the first place?
Let us turn to this question.

Economic and social factors in the emeryence of new paradigms

It is useful to separate the genesis of new paradigms from the processes leading to the
dominance of some of them. Let us first consider the emergence of new potential
paradigms; that is, the generation of notional opportunities of radical innovation
involving new knowledge bases, new search heuristics, new dominant designs.
In the literature one find quite different interpretative archetypes. A first class of
models entails a lot of "techno-scientific determinism": advancements in pure
science determine advancements in technological opportunities which in turn
determine realized technological achievements. In fact, in order to find the most
naive literature along these lines one should mostly search in the archives of defunct
socialist countries. There, one is likely to find plenty of examples of Engels-type
vulgata on the simplest linear models from science to technology to production.
The interpretation that students of economics find in textbook production is
more sophisticated although basically of the same type. It maintains the basic linear
sequence from science to technological opportunities to production but it claims
that science only generates those notional blueprints discussed earlier, while some
optimizing microeconomic algorithm selects among them. Proper economic
analysis begins indeed by stating some daring assumptions on the nature of such
blueprints which maintain in principle an empirical nature albeit little empirical
micro support (e.g. on continuity, convexity, etc.). From then onward, production
theory is generally presented as an application of methods of constrained maximiza-
tion which intends to capture the purposed behavior of the homo economicus facing
alternative allocative choices, and most often, also the aggregate properties of
industries or whole economies.
Yet more sophisticated recent modelling on new-growth, new-trade theories,
while attempting to endogenize the generation of blueprints themselves, push
further upstream that same notional process of optimizing allocation involving
some sort of production function for the blueprints themselves. This is not the place
to discuss the (rather important) achievements and the (equally important) limita-
tions of such theories. What we simply want to emphasize is the persistence across
ample streams of micro and macro literature of two basic ideas: first, the linear
representation of the innovative process, running from science to technology to
production; and second, the focus upon an explicit deliberation, equivalent in every
respect to an allocative choice, by supposedly rational agents.

In this co-evolutionary perspective on technologies, corporate organizations and institutions

(Nelson 1994),it is straightforward to acknowledge also a bi-directional relation between market
structures (as proxied by measures on the distribution of different characteristics such as firm sizes,
innovative competences, ownership, persistent behavioral traits, etc.) and patterns of technological
learning. Different rates of learning influence the ability of firms to survive and expand and thus
affectindustrial structures. Conversely any particular structure with its associated distribution of
corporate features - influencesand constrains what and how fast firms are able and willingto learn.
Formal applications of this general idea are in Nelson and Winter (1982), Winter (1984), Dosi,
Marsili, Orsenigo and Salvatore (1993).
254 M. Cimoli and G. Dosi

However, as Chris Freeman, Nathan Rosenberg, and others have convincingly

shown, historical evidence rules out the general applicability of linear models of
innovation. One can find plenty of counter examples. First, the lag between scientific
advancements and their technological application can vary between a few months
(as in case of the transistor) to centuries. Second, technological innovation may
actually precede the scientific discovery of the general principle on which those very
technologies work (as in the case of electric lamps). Third, scientific advancements
may actually be based on the invention of new machinery and not the other way
round (think of the importance of the electronic microscopes for the subsequent
scientific discoveries in biology), s
As regards the behavioral foundations of innovative decisions, we are quite
skeptical about their reduction to deliberate allocative choices. As emphasized not
only by evolutionary economists by also by rational choice theorists like K. Arrow,
almost by definition innovation concerns the generation of something new and at
least partly unexpected. Relatedly, the genesis of exploratory ventures into novel
paradigms is more the domain of institutional and organizational inquiries on the
conditions fastering entrepreneurial activities rather than rational choice models. 9
Indeed, there are good reasons to believe that one will not be able to find
anything like a general theory of the emergence of new technological paradigms.
However, what might be possible is a) an analysis of the necessary condition for such
emergence; b) historical taxonomies and also appreciative models of the processes
by which it occurs; and c) taxonomies and models of the processes of competition
amongst different paradigms and their diffusion.
Regarding the first heading, one is like to find that the existence of some
unexploited technological opportunities, together with the relevant knowledge base
and some minimum appropriability conditions, define only the boundaries of the set
of potential new paradigms: those which are actually explored within this set might
crucially depend on particular organizational and social dynamics. So for example,
there is good evidence that the micro electronics paradigm as we know it (silicon-
based, etc ) was shaped in its early stages by military requirements (Dosi 1984, Misa
1985). David Noble argues that the NC machine-tool paradigm - although he does
not use that expression- has been influenced by power considerations regarding
labour management (Noble 1984). In the history of technology one finds several
examples of this kind. The general point is that various institutions (ranging from
incumbent firms to government agencies), social groups and also individual agents
(including, of course, individual innovators and entrepreneurs) perform as e x a n t e
selectors of the avenues of research that are pursued, the techno-economic dimen-
sions upon which research ought to focus, the knowledge base one calls upon. Thus,
they ultimately select the new paradigms that are actually explored.

8 See Rosenberg (1991)

9 Of course this is not to say that the economic variables governing the incentives and penalties to
entrepreneurial endeavours are irrelevant. The point is that the former tend to set only some lose
incentive-compatibilityconstraints. Given these constraint, explanations of willingness ofincubent
firms to explore new paradigms, of the rate of the birth of new start-up firms, etc. requires a much
more detailed understanding of specificcorporate and institutional histories. Working backward
from observed outcomes to some rational expectations model does not do the trick: on the
contrary, there is evidence that in many new industries, had entrants rational expectations of their
future profit streams, entry would not have occurred at all! (this seems to emerge also from
a research, in progress, by Don Lovallo and Giovanni Dosi).
Technological paradigms, patterns of learning and development 255

There is a much more general theoretical story regarding the development,

diffusion and competition among those (possible alternative) paradigms that are
actually explored. It can be told via explicit evolutionary models (as in Nelson and
Winter 1982 or in Silverberg, Dosi and Orsenigo 1988), via path-dependent stochas-
tic models (as in Arthur 1988, Arthur, Ermoliev and Kaniovski 1987, Dosi and
Kaniovski 1994 and David 1989), and also via sociological models of network
development (as in Callon 1991). The basic ingredients of the story are i) some forms
of dynamics increasing returns (for example in learning); ii) positive externalities in
the production or the use of the technology; iii) endogenous expectation formation;
iv) some market dynamics which selects ex post amongst products, and indirectly
amongst technologies and firms; v) the progressive development of standards and
relatively inertial institutions which embody and reproduce particular forms of
knowledge and also the behavioral norms and the incentives to do so.

3. Learning and trajectories in the process of development

Techno-ecnomic paradigms or regimes:from micro technologies
to national system of innovation
So far, we have discussed paradigms, trajectories or equivalent concepts at a micro-
technological level. A paradigm-based theory of innovation and production - we
have argued - seems to be highly consistent with the evidence on the patterned and
cumulative nature of technical change and also with the evidence on microeconomic
heterogeneity and technological gaps. Moreover, it directly links with those theories
of production which allow for dynamic increasing returns from A. Young and
Kaldor to the recent and more rigorous formalizations of path-dependent models of
innovation diffusion, whereby the interaction between micro decisions and some
form of learning or some externalities produces irreversible technological paths and
lock-in effects with respect to technologies which may well be inferior, on any welfare
measure, to other notional ones, but still happen to be dominant loosely speak-
ing- because of the weight of their history (cf. the models by B. Arthur and P.
David). However, paradigms are generally embodied in larger technological systems
and in even bigger economic-wide systems of production and innovation.
The steps leading from a microeconomic theory of innovation and production to
more aggregate analyses are clearly numerous and complex. A first obvious question
concern the possibility of identifying relative coherence and structures also at these
broader levels of observation. Indeed, historians of technology-T. Hughes, B.
Gilles and P. David, among others-highlight the importance of technological
systems, that is in the terminology of this paper, structured combinations of micro
technological paradigms (see for example, the fascinating reconstructions of the
emerging system of electrification and electrical standards in David 1992).
At an even higher level of generality, Freeman and Perez (1988) have suggested
the notion of techno-economic paradigms as a synthetic definition of macro-level
systems of production, innovation, governance of social relations. So, for example,
they identify broad phases of modern industrial development partly isomorphic to
the notion of "regimes of socio-economic Regulation" suggested by the mainly
French macro institutionalists literature (see Aglietta 1976, Boyer 1988a and b).
In an extreme synthesis, both prospectives hold, first, that one can identify rather
long periods of capitalist development distinguished according to their specific
engines of technological dynamism and their modes of governance of the relation-
256 M. Cimoli and G. Dosi

ships amongst the major social actors (e.g. firms, workers, banks, collective political
authorities etc), and, second, that the patterns of technological advancement and
those of institutional change are bound to be coupled in such ways as to yield
recognizable invariances for quite long times in most economic and political
structures. Just to provide an example, one might roughly identify, over the three
decades after WW II, across most developed economies, some "Fordist/Keynesian"
regime of socio-economic "Regulation", driven by major innovative opportunities
of technological innovation in electromechanic technologies, synthetic chemistry
and relatively cheap exploitation of energy sources, and reproduced by some specific
forms of institutional governance of industrial conflict, income distribution and
aggregate demand management. Analogously, earlier in industrial history, one
should be able to detect some sort of archetype of a "classical/Victorian Regime"
driven in its growth by the full exploitation of textile manufacturing and light
engineering mechanization, relatively competitive labour markets, politically driven
efforts to expand privileged market outlets, etc..
These general conjectures on historical phases or regimes are grounded on the
importance in growth and development of specific combinations among technologi-
cal systems and forms of socio-economic governance. The approach can be applied
also to the analysis of the differences and similarities of development patterns in the
late-industrializing countries. One has focused for example on the interplay between
the modes of governance of the labour market and the pattern of technical
accumulation, showing how the specificities in labour market institutions originate
virtuous or vicious circles of development in different historical periods. 1~
As an intermediate step toward the identification of national socio-economic
regimes let us consider the anatomy and development of particular systems of
innovation and production at national level, embodying distinctive mechanisms
and directions of learning, and grounded in the micro theory of production and
innovation sketched above.
Even if micro paradigms present considerable invariances across countries, the
ways various paradigms are combined in broader technological systems and, more
so, in national systems of production and innovation highlight-we suggest- a
considerable variety, shaped by country-specific institutions, policies and social
factors. The hypothesis here is that evolutionary microfoundations are a fruitful
starting point for a theory showing how technological gaps and national institu-
tional diversities can jointly reproduce themselves over rather long spans of time in
ways that are easily compatible with the patterns of incentives and opportunities
facing individual agents, even when they turn out to be profoundly suboptimal from
a collective point of view.
In order to detail this hypothesis, however, one requires to analysis of the
composing elements and properties of these national systems which in the recent
literature have been referred to with a variety of largely overlapping concepts, such
as global technological capability of each country, national innovation systems,
national technological capabilities and national systems of production.1 a
In our view, the major building blocks in an evolutionary account of the
specificities of national systems of production and innovation are following.

lo See Aboites(1988)Boyer(1993),Cetrangolo (1988a),(1988b),Cimoli (1988),(1990),Coriat and

Saboia (1987).
i t See Cimoli and Dosi (1988),(1990),Chesnais (1993), Ernest and O'Connor (1989), Lall (1984),
(1987), (1992),Lundvall (1992),Nelson (1993),Zysman (1994).
Technological paradigms, patterns of learning and development 257

First, there is the idea that firms are a crucial (although not exclusive) reposito-
ries of knowledge, to a large extent embodied in their operational routines, and
modified through time by their higher level rules of behaviors and strategies (such as
their search behaviors and their decisions concerning vertical integration and
horizontal diversification, etc.).
Second, firms themselves are nested in networks of linkages with other firms and
also with other non-profit organizations (such as public agencies etc.). These
networks, or lack of them, enhance or limit the opportunities facing each firm to
improve their problem-solving capabilities.
Third, national systems entail also a broader notion of embeddedness of
microeconomic behaviors into a set of social relationships, rules and political
constraints (Granovetter 1985). Even at a properly micro level, the momentum
associated with single technological trajectories is itself a largely social concept: "it
points.., to the organizations and people committed by various interests to the
system, to manufacturing corporations, research and development laboratories,
investment banking houses, educational institutions and regulatory bodies" (Misa
1991: p. 15). And, in turn, these interests and institutions are sustained by the
increasing-return and local nature of most learning activities. Even more so, at
a system-level, the evolutionary interpretation presented here is consistent and
indeed complementary with institutional approaches building on the observation
that markets do not exist or operate apart from the rules and institutions that
establish them and that "the institutional structure of the economy creates a distinct
pattern of constraints and incentives", which defines the interests of the actors as
well as shaping and channeling their behaviours (Zysman 1994: pp. 1-2). lz

Paradigms, routines, organizations

A locus classicus in the analysis of the profound intertwining between technological
learning and organizational change is certainly Alfred Chandler's reconstruction of
the origins of the modern multi-divisional (the M-form) corporation and its ensuing
effects on the American competitive leadership over several decades (Chandler
(1990), (1992a) and (1993)). And, as Chandler himself has recently argued, there are
strict links between story and evolutionary theories (Chandler (1992b). While it is
not possible to enter into the richness of the Chandlerian analysis here, let us just
recall one of the main messages:
[...] it was the institutionalizing of the learning involved in product and process development
that gave established managerial firms advantages over start-ups in the commercialization of
technological innovations. Development remained a simple process involving a wide variety of
usually highly product-specific skills, experience and information. It required a close interac-
tion between functional specialists, such as designers, engineers, production managers, mar-
keters and managers [...]. Such individuals had to coordinate their activities, particularly
during the scale-up processes and the initial introduction of the new products on the market
[...]. Existing firms with established core lines had retained earnings as a source of inexpensive
capital and often had specializedorganizational and technical competence not available to new
entrepreneurial firms (Chandler 1993: p. 37).

12 Note incidentally that the second building block - i.e. networks, etc. - in so far as it is equivalent
to an externality or to some economy-wide mechanism for the generation of knowledge, is also
captured in a highly simplifiedform by the new growth theories. (More on the general spirit of the
latter in Romer 1994a and b). Conversely the first and third are distinctive of evolution-
ary/institutionalist analyses. These represent also a major point linkage between evolutionary
theories, organizational economics and business history.
258 M. Cimoliand G. Dosi

As thoroughly argued by Chandler himself, this organizational dynamics can be

interpreted as an evolutionary story of competence accumulation and development
of specific organizational routines (Chandler (1992b)).
Did seemingly superior organizational forms spread evenly throughout the world?
Indeed, the Chandlerian enterprise diffused, albeit rather slowing, in other
OECD countries (Chandler 1990, Kogut 1992). However, the development of
organizational forms, strategies and control methods have differed from nation to
nation, because of the difference between national environments (Chandler 1992a: p.
283). Moreover, the diffusion of the archetypical M-form corporation has been
limited to around half a dozen already developed countries (and even in countries
like Italy, it involved very few companies, if any). Similar differences can be found in
the processes of international diffusion of American principles of work organiz-
a t i o n - e.g. Taylorism and Fordism-(for an analysis of the Japanese case, see
Coriat 1990). For the purposes of this work, it is precisely these differences and the
diverse learning patterns which they entail that constitute our primary interest.
So, for example, a growing literature identifies some of the roots of the specifici-
ties of the German, the Japanese or the Italian systems of production into their early
corporate histories which carried over their influence up to the contemporary form
of organization and learning (see Chandler 1990, Coriat 1990, Kogut 1993,
Dursleifer and Kocka 1993, Dosi, Giannetti and Toninelli 1992).
Even more so, one observes quite different organizational initial conditions,
different organizational histories, and together, different patterns of learning across
developing countries. Let us consider them at some detail.
During the last three decades, developing countries have shown increased
technological dynamics associated with a subsequent development of their indus-
trial structures, thus some significant technological progress did indeed occur in the
NIEs and some of them also became exporters of technology.13
The evolutionary path of technological learning are related to both the capacity
to acquire technologies (capital goods, know how etc) and the capability to absorb
these technologies and adopt them to the local conditions. In these respects, one has
now a good deal of microeconomic/micro technological evidence highlighting the
mechanisms which stimulate and limit endogenous learning in the NIEs. 14
Without doing any justice to the richness of these contributions, they seem to
suggest the existence of some characteristics in the paths of technological learning at
the firm level (see also Cimoli 1990 and Cimoli and Dosi 1988). In particular, one
might be able to identify some relatively invariant sequences in the learning
processes, conditional on the initial organizational characteristics of the firms and
the sectors of principal activity.
A first set of regularities regards the varying combinations between acquisition
of outside technologies and endogenous learning. 1S As well know, the transfer of
technology to developing economies is a common source for the subsequent

13 See Lall (1982), Teitel (1984) and Teubal (1984).

t4 See, among others, Bell (1982), Dahlman-Westphal(1982), Hobday (1984), Herbert-Copley
(1990),Justman-Teubal(1991), Katz (1983),(1984a),(1984b),(1986)and (t987),Teubal(1987),Pack
and WestphaI(1986).
15 The technologyflowsto developingeconomiesshowa rapid expansionin the 1960sand 1970s;
duringthe 1980sthisprocessdecreasedits intensity(UNCTAD 1991).During the wholeperiod the
Asian countriesshow an increasingrole as the major recipient of foreigndirect investmentand
capital goods. The flow of capital goods to Latin Americancountries remain stable during this
period.
Technological paradigms, patterns of learning and development 259

development of learning capabilities at the firm and sectoral levels. Possibly with too
extreme an emphasis, Amsden and Hikino identify the ability to acquire foreign
technology as a central characteristic,
[...] of late industrialization at the core of which is borrowing technology that has already been
developed by firms in more advanced countries. Whereas a driving force behind the First and
Second Industrial Revolutions was the innovation of radically new products and processes, no
major technological breakthrough has been associated with late-industrializing economies.
The imperative to learn from others, and then realize lower costs, higher productivity, and
better quality in mid-tech industries by means of incremental improvements, has given
otherwise diverse 20th century industrializers a common set of properties (Amsden and Hikino
1993: p. 37).t6
At a general level, learning patterns can be taxonomized according to the relative
importance of the corporate activities involved, 17 namely a) the acquisition of an
existing technology associated with the paradigm prevailing in the developed world,
b) its adaptation and modification in the local environment and c) the creation of
new innovation capabilities with respect to products and processes.
The importance of the three often follows a temporal sequence. Already the
modification of the adopted technology implies learning of new production skills
which grows through the adaptation of this capabilities to local specificities. Note,
however, that there is no inevitability in the learning-by-doing process which, on the
contrary, requires adequate organization conditions, both within each firm and each
environment. Interestingly, the initial characteristics of corporate organizations
appear to exert a strong influence on subsequent dynamics. For example, evidence
on the last four decades (1950-1990) concerning Latin American countries
(Argentina, Brazil, Colombia, Mexico and Venezuela) indicate that the evolutionary
sequence of organizational and technological learning can be distinguished among
four types of firms, taxonomized mainly in terms of the nature of ownership:
subsidiaries of MNCs, family firms, large domestic firms and public firms.~8
The family firm appears to be characterized by a high "propensity to self-
sufficiency and self-financing" and the "mechanical ability of an individual", which
frequently stems from immigrant entrepreneurs. 19 The technology acquired is
related to the technical background of the entrepreneur and the initial phase is
characterized by the adoption of a discontinuous mode of production, z~ At the

6 Although we share their view on the curent importance of technological assimilation of outside
technologies, one should not underestimate the degree to which this occured also in the past
experiences of late-coming industrialization and catching-up, for example in the case of the USA or
Continental Europe vis-a-vis Britain.
17 On a similar point Teitel (1987).
~8 Information on the different phases ofthe technological accumulation of firms has been taken
from the case studies of the IDB, ECLA and UNDP programmes and from the overviews of the
research findings in Katz (1983), (1984a), (1984b), (1986) and (1987), Berlinski, Nosier, Sandoval and
Turkieh (1982), Teitel (1984) and (1987), Teubal (1987).
19 See Katz (1983).
2o Tw• a•ternative m•des •f pr•ducti•n name•y• c•ntinu•us and disc•ntinu•us• appear t• be re•evant
for the analysis of learning patterns. Continuous methods imply I) specialization of production along
precise product lines;2) production planning for each line of business; 3) relativelyhigh scale economies;
4) relatively low flexibility in product design and rates of throughput. Conversely, discontinuous
methods involve 11low standardization of production; 2) low economies of scale; 3) the organization
of production into multi-product "shops"; 4) general purpose, low cost machinery. It is remarkable
that in many Latin American examples, (but not in Far Eastern ones) at least until the 1980s,
incremental learning appeared to be more successfulin discontinuous batch-production as compared
to continuous and mass-production activities (such as chemicals, many consumer durables, etc.).
260 M. Cimoli and G. Dosi

beginning, production is characterized by low economies of scale (also as a conse-

quence of the limitations of the domestic market and the diff• in exploiting
export possibilities).
A sort of ideal learning trajectory for a South American family-stablised firm
that is technologically progressive (which is not by any means a general characteris-
tic of the whole population) would run more or less as follows. First, the effort is
concentrated on product design activities (most likely due to the incentive provided
in the past by import substitution policies), and increasingly, on quality improve-
ments and product differentiation.
Next, attention is focused also on process engineering, the organization of
production and the exploitation of some economies of scale, until (in some empiri-
cally not too frequent cases) highly mechanized production is achieved. And, along
the process, it might happen (again, not too often) that the organization is developed
beyond the original family hierarchy and "managerialized".
The story concerning subsidiaries of foreign firms emerges from the set of cases
studies cited earlier is quite different. The bulk of competences and technologies
derives from the parent company and learning mainly concern the adaptation to the
local environment, adjustments in product mixes and re-scaling of production lines.
In some cases, this holds throughout the history of the subsidiary, while in others an
autonomous capability in product and process design is developed. (Note also that
in Latin American foreign subsidiaries tend to be concentrated in mass production
activities like vehicles, consumer durables, food processing, etc.).
State-owned firms display yet another archetypical learning story. First, they
have been concentrated in sectors that have tended to be considered "strategic" and
often happened to be continuous process industries such as bulk materials, steel,
basic petrochemicals, in addition - in some countries - to aerospace and military
production. Second, the strategies have generally be dictated also by political
considerations. Third, learning has often started via agreements with international
suppliers of equipment. In the "healthy" scenario - which is not the rule - inter-
national technology transfer agreement became more sophisticated, involving
adaptation of plants and technologies to local circumstances, while the emphasis
was kept on personnel training and learning by using. Finally, autonomous
capabilities of plant upgrading and process engineering were sometimes developed.
As regards large domestic firms, it is hard to trace any modal patterns. Scanning
through the case studies, they sometime appear to follow patterns not too different
from the family firms, in other cases they seem to perform like East Asian business
groups (see below), and yet in others learning appears to be much more directed
toward he exploitation of political rents and financial opportunities rather than
technological accumulation.
It is interesting to compare these sketchy Latin American "corporate trajecto-
ries" with other experiences, such as the Korean one. zl To make a long and
variegated story very short, in Korea it seems that the major actors in technological
learning have been large business groups - t h e c h a e b o l s - which have been able at
a very early stage of development to internalize the skills for the selection among
technologies acquired from abroad, their efficient use and adaptation, and, not
much later, have been able to grow impressive engineering capabilities.

21 As discussedat greater depth in Amsden(1989),Amsedenand Hikino (1993and 1994),Enos and

Park (1988),Bell and Pavitt (1993), Lall (1992), Kim, Westphal and Dahlman (1985).
Technological paradigms, patterns of learning and development 261

Conversely, the Taiwanese organizational learning has rested much more in

large networks of small and medium firms very open to the international markets
and often developing production capabilities which complement those of first world
companies (Dahlman and Sananikone 1990, Ernest and O'Connor 1989).
This impressionistic list of stylized organizational patterns of learning could be
of course very lengthy. For our purposes, it should be understood only as an
illustration of the multiplicity of evolutionary paths that organizational learning can
take. The fundamental point here is that the rates and directions of learning are not
at all independent from the ways corporate organizations emerge, change, develop
particular problem-solving, capabilities, diversify, etc. It is the core co-evolutionary
view emphasized by Nelson (1994).
The analysis in term of paradigms, trajectories, technological asymmetries, etc.
outlined in the first part of this paper is the most abstract level of description of
production pattern and technical change, whereby the "primitives" of the analysis
itself are entities like "bits of knowledge" and the outcomes of their implementation
in the spaces of production process and output characteristics. However, knowledge
is to a large extent embodied, reproduced and augmented within specific organiz-
ations. Thus, a lot of the action is reflected in the behaviours, evolution and learning
of these organizations - in primis, business firms. At this level, evolutionary analyses
match and cross-fertilize with investigations on organizational dynamics, industrial
demography and business history.

Inter-sectoral networks and production capacity

Of course the multiple business histories of learning and organizational change (or
lack of them) in each country, as mentioned earlier, is nested into flows of
commodities and knowledge across different sectors and different institutional
actors. At this level, can one identify some broad regularities in the processes of
system-construction.
It is useful to maintain the distinction emphasized by Bell and Pavitt (1993)
(which indeed bears some Listian flavour!) between the development of a "produc-
tion capacity" and of "technological capabilities". Production capacity concerns the
stocks of resources, the nature of capital-embodied technologies, labour skills,
product and input specification and the organizational routines in use. Technologi-
cal capabilities rest on the knowledge and resources requested for the generation
and management of technical change.
These seem to be some patterns, albeit rather loose, in the development of
a national production capacity. For example, practically every country starts with
manufacturing of clothing and textile, possibly natural resource processing, and
moves on - if it does - to more complex and knowledge intensive activities. How-
ever, the tricky question is whether there are some activities which hold a special
status in the construction of a national system of production and innovation, also
due to the property that having a production capacity in them makes it easier, other
things being equal, to develop technological capabilities. The conjecture is quite old
(and goes back at least to List, Ferrier and Hamilton) and is present in contemporary
notions such those of fili re or Dahmen's "development blocks", but it might gain
strength on the grounds of the evolutionary microeconomics outlined above. 12

22 Classic contributions of the importance of intersectoral linkages are Dahmen (1971) and
Hirscbman (1992).
262 M. Cimoliand G. Dosi

Whenever one abandons a view of development exclusively shaped by endow-

ments, the degrees of perfection of market signals, and the like, but focuses on the
conditions for technological/organizationallearning, then it also becomes easier to
appreciate the diversity of the sources in learning opportunities and their different
economic potentials.
In fact, there is a good circumstantial evidence from contemporary as well as
previous late-industrializing countries (such as, in their days, the USA, Germany,
the Scandinavian countries, Japan, etc.). Suggesting that there are technologies
whose domains of application are so wide and their role has been so crucial that the
pattern of technical change of each country depends to a large degree on the national
capabilities in mastering production/imitation/innovationin a set of crucial knowl-
edge areas (e.g. in the past, mechanical engineering, electricity and electrical devices,
and nowadays, also information technologies). Moreover, the linkages among
production activities embody structured hierarchies whereby the most dynamic
technological paradigms play a fundamental role as sources of technological skills,
problem-solving opportunities and productivity improvements. 23 Thus, these core
technologies shape the overall absolute advantages/disadvantages of each country.
In other words, the pattern of technical change of each country in these technologies
does not average out with the technological capabilities in other activities but are
complementary to them. These core technologies often also imply basic infrastruc-
tures and networks common to a wide range of activities (such as, for example, the
electricity grid, the road system, telecommunications and more recently the informa-
tion network). Many pieces of empirical evidence strongly convey the idea that
a proper technological dynamism in developing countries is impossible without
major structural changes and a sequential construction of a widening manufactur-
ing sector involving also indigenous skills in a set of "core" technologies.
We do not at all suggest that there is any invariant sequence of industrial sectors
which account for the upgrading of national technological capabilities. 24 However,
one might still be able to identify some rough sequences in the predominant modes
of technological learning. In this respect, the taxonomy of the sectoral patterns of
acquisition of innovative knowledge suggested by Pavitt (1984) is a good - albeit
somewhat theoretically fuzzy - point of departure. As known, Pavitt distinguishes
four groups of industrial sectors, namely (i) supplier dominated - where innovations
mainly enter as exogenously generated changes in capital and intermediate good,
and where learning is primarely associated with adoption and production skills; (ii)
specialized s u p p l i e r s - p r o v i d i n g equipment and instruments to the industrial
system, and relying in their innovative activities on both formal (more or less
scientific) knowledge and more tacit one based also on the user-producer relation-
ships; (iii) scale-intensive s e c t o r s - whose innovative abilities draw, jointly, on the
development adoption of innovative equipment, on the design of complex products,
on the exploitation of some scale economies, and on the ability of mastering complex
organizations; (iv) science-based sectors - whose innovative opportunities link more
directly with advances in basic research.

23 See Rosenberg(1976), and, for contemporarylate comers,Chudnovsky,Nagao and Jacobsson

(1984) and Fransman(1986).
24 Indeed,any detailedcomparisonof the sectoral compositionof output of e.g. USA,Germany,
Japan, France,Italybetween1850and 1950,or Korea,Brazil,Taiwan,and Singaporebetween1950
and 1990- we conjecture- would show enough dispersion.
Technologicalparadigms,patterns of learning and development 263

The important issue here is whether one may use that taxonomy in order to
detect some patterns over the development process. The emergence of manufactur-
ing sector is generally characterized by an initial stage where supplier-dominated
sectors dominate accompanied by the emergence of specialized suppliers. The
process of technical change in these sectors is characterized by a sequential
development of various forms of tacit and incremental learning related to the
transfer and acquisition of foreign technology. These learning activities are mainly
related to the use of equipment, development of engineering skills in machine-
transformation, adaptation of existing machines and final products to specific
environmental conditions.
The emergence of "scale-intensive" industries entails further forms of learning
related to the development and use of capital equipment. Moreover, unlike the
supplier-dominated sectors, technological efforts are also focused on (i) the develop-
ment of technological synergies between production and use of innovations, often
internalized via horizontal and vertical integration; (ii) the exploitation of static and
dynamic economies of scale; (iii) the establishment of formal institutions undertak-
ing search (typically, corporate R&D laboratories), and complementary to informal
learning and diffusion of technological knowledge.
Sectoral learning patterns are clearly nested into broader ("macro") conditions
such as those definining the educational system. For example, in "supplier-
dominated" and "specialized supplier" sectors, a significant role is played by the
levels of literacy and skills of the workforce, and the skills and technical competence
of engineers and designers in the mechanical and (increasingly) electronics fields. In
scale-intensive sectors, the existence of managers capable of efficiently running
complex organizations is also likely be important. In science based sectors, the
quality of higher education and research capabilities is obviously relevant.
Moreover, sectoral learning patterns and overall national capabilities are dy-
namically coupled via input-output flows, knowledge-spillovers, complementarities
and context-specific externalities. 25 Together, they contribute to shape the or-
ganizational and technological context within which each economic activity takes
place. In a sense, they set the opportunities and constraints facing each individual
process of production and innovation including the availability of complementary
skills, information on intermediate inputs and capital goods, and demand stimuli to
improve particular products.
This links straightforwardly with the analyses focusing on structural change and
development (here within a vast literature, contributions that come immediately to
mind range from Hirschman to Rosenstein Rodan, Gerschenkron, Prebisch-
notwithstanding his sometimes extremist interpretations -, Lowe, Kuznets, Che-
nery, Sirquin among others). Certanly, the dynamics of development also rest upon
major structural transformations which entail a changing importance of different
branches of economic activity as generators of both technological/organizational
innovations and demand impulses. So, for example, in this interpretative frame-
work, it does not sound so outragions to conjecture that the "quality" (in terms of

25 An obviousquestionconcernsthe unit of analysisto whichtheseexternalitiesand systemeffect

apply. Why shouldn't international trade compensate for spatially circumscribed specificities?
What is unique of a nation distinguishingit from a geographicalregion, or a firm,or a group of
individuals? Far from any intent to reduce the importance of other levels of description (e.g.
regionaldynamics),wemaintainthat nationsare also specifiedby particularmodes ofinstitutional
governancewhich extent make them diverseauto-reproducingentities.
264 M. Cimoli and G. Dosi

technological opportunities and demand elasticities) of any one structure of produc-

tion and export is bound to influence the relative ability of a country to absorve its
labour supply, meet its foreign balance constraints, grow in its per capita income...
At this level of analysis, one empirically finds, for example, that Latin American
countries have increasingly biased their structure of production, in the 1980s, in
favour of resource-based sectors, while East Asian NICs have move toward scale-
intensive and science-base sectors. Obviously, in some ideal General Equilibrium
world, all this is just an irrelevant epiphenomenon. Conversely, under the micro-
foundation sketched above, there might be reasons to worry (even if naive techno-
logically deterministic conclusions should be obviously discussed). Indeed, some
fundamental trade-off between "static allocative efficiency" and "dynamic effi-
ciency" of any one pattern of production might plausibly emerge (more on all this in
Dosi, Pavitt and Soete 1990 and Dosi, Tyson and Zysman 1989).
Moreover, the specificities of each system of production interact with those of
each national system of innovation - as throughly analyzed in Nelson (1993) - and
tend to yield recognizeable national patterns or trajectories shaped by the institu-
tions supporting technical advances and reproduced through time also by processes
of lock-in into particular knowledge bases, corporate organizations and sectoral
specialization.

4. From micro technological paradigms to national systems of production

and innovation: some tentative conclusions and many research avenues

The third major element, mentioned earlier linking microeconomic learning with
national patterns of development was the embeddedness of the thread of incentives,
constraints, forms of corporate organization into the broader institutional frame-
work of the political economy of each country. It is beyond the scope of this paper to
discuss that issue at any satisfactory detail. For our purposes, let us just mention that
the micro- and meso-economic theoretical building blocks sketched above and
drawn from an evolutionary perspective are in principle consistent with broader
institutionalist analyses of national systems of production, innovation and govern-
ance of socio-economic relations. Indeed, one can see multiple links which one is
only beginning to explore. For example, an evolutionary perspective is quite at ease
with the idea that markets are themselves "social constructs" which - depending on
their rules and organizing principles- shape microbehaviours and adjustment
mechanisms. The emphasis on patterned and local learning, and bounded rationaly
assumptions, go well together with the view of political economists and sociologists
of development according to which a major ingredient of development is the process
of change in social norms, expectations and forms of collective organization. The
patterns of socio-economic Regulation (Boyer 1988a and 1988b) can be in principle
microfounded into underlying evolutionary processes of self-organization, learning
and selection. In fact, there seem to be a large domain where more "bottom-up"
evolutionary theories and more "top-down" institutional analyses can develop
a fruitful dialogue.
Notions like those of technological trajectories, path-dependencies, organiza-
tional competences, self-organization, learning and selection dynamics - and many
others stemming from evolutionary investigations- are becoming part of the
tool-kit of many social disciplines. As regards more specifically development issue,
they start becoming building blocks which might provide firmer grounds to the
Technological paradigms, patterns of learning and development 265

broad intuitions of an earlier generation of development theorists - from Myrdal to

Hirschman, from Rosenstein-Rodan to Gerschenkron... In this respect our tour de
force from technological paradigms to national systems should just be considered as
a tentative r o a d m a p over still largely unexplored terrains.

References

Aboites J (1988) Regimen de acumulacion, relacion salarial y crisis en mexico. Paper presented at
the Conference on the Theory of Regulation, Barcelona, June 1988
Abramovitz M (1989) Thinking about growth, Cambridge University Press, Cambridge
Aglietta M (1976) Regulation et crises du capitalisme. Calmann-Levy, Paris
Amsden A (1989) Asia's next giant: South Korea and the last industrialization. Oxford University
Press, New York and Oxford
Amsden A, Hikino T (1993) Staying behind, stumbling back, sneaking up, soaring ahead:
late industrialization in historical perspective. In: Baumol W, Nelson R, Wolff E (eds) Inter-
national convergence of productivity with some evidence from history. Oxford University
Press, Oxford
Amsden A, Hikino T (1994) Project execution capability, organizational know-how, and conglom-
erate corporate growth in late industrialization. Industr Corp Change 3, 11 i-147
Arthur B (1988) Competing technologies. In: Dosi G, Freeman C, Nelson R, Soete L (eds) Technical
change and economic theory. Frances Pinter, London
Arthur B, Ermoliev Y, Kaniovski Y (1987) Path dependent proccesses and the emergence of macro
structure. Eur J Operat Res 30:294-303
Bell M (1982) Technical change in infant industry: a review of empirical evidences. SPRU, Sussex,
mimeo
Bell M, Pavitt (1993) Technological accumulation and industrial growth: contrasts between
developed and developing countries. Industrial and Corporate change, 2, 157-210
Berlinski J, Nogueira da Cruz H, Sandoval D, Turkieh M (1982) Basic issues emerging from recent
research on technological behavior of selected Latin American metalworking plants. CEPAL
Working Paper n. 56
Boyer R (1988a) Technical change and the theory of regulation. In: Dosi et al. (1988)
Boyer R (1988b) Formalizing growth regimes within a regulation approach. In: Dosi et al (1988)
Boyer R (1993) Do labour institutions matter for economic development? Paper presented at the
Workshop on Labour Institutions and Economic Development in Asia, Bali, February 4-6,
1992
Callon M (1991) Technological development and adoption networks. Conference on Constructive
Technology Assessment, Twente, The Netherlands, 1991
Cetrangolo O (1988a) Import substitution and industrialization policy in Argentina under Peron.
IDS, Sussex, mimeo
Cetrangolo O (1988b) State and industrial development in Argentina. MPHIL thesis, IDS, Sussex
Chandler AD (1990) Scale and scope: the dynamics of industrial capitalism. Hardvard University
Press, Cambridge, Mass
Chandler AD (1992a) Corporate strategy, structure and control methods in the United States
during the 20th century. Industr Corp Change 1:263--284
Chandler AD (1992b) Organizational capabilities and the economic hislory of the industrial
enterprise. J Econ Perspect 6:79-100
Chandler AD (1992) Learning and technological change: the perspective from business history. In:
Thomson (1993)
Chudnovsky D, Nagao D, Jacobsson S (1984) Capital goods production in the Third world: an
economic study of technical acquisition. Frances Pinter, London
Cimoli M (1988) Industrial structures, technical change and the modes of regulation in the labour
market. The case of Latin American countries. DRC discussion paper n. 60, SPRU, University
of Sussex, Sussex
Cimoli M (1990) Technology, international trade and development: a north south perspective.
PhD. Thesis, Unversity of Sussex, Sussex
266 M. Cimoli and G. Dosi

Cimoli M, Dosi G (1988) Technology and development some implications of recent advances in the
economics of innovation for the process of development. In: Wad A (ed) Science, technology and
development: a critical perspective, Westview Press, Boulder, Col
Cimoli M, Dosi G (1990) The characteristics of technology and the development process: some
introductory notes. In: Chatterji M (ed) Technology transfer in the less developing countries.
Macmillan, London
Coriat B (1990) Penser l'envers. Paris, Bourgeois
Coriat B, Saboia J (1987) Regime d'accumulation et rapport salarial au Bresil (des ann6es 1950 aux
ann6es 1980). University of Paris VII, mimeo
Constant E (1985) The turbojet revolution. Johns Hopkins University Press, Baltimore
David PA (1989) Path-dependence: putting the past into the future of economics. Discussion paper,
Stanford University, Stanford
David P (1992) Heros, herds and hysteresis in technological history: Thomas Edison and the battle
of the system reconsidered. Indust Corp Change 1:129-180
Dahlman C, Westphal (1982) Technological effort in industrial development: a survey. In: Stewart
F, James J (eds) The economics of new technology in developing countries. Frances Pinter,
London
Dahlman C, Sananikone O (1990) Technological strategy in Taiwan: exploiting foreign linkages
and developing local capabilities. Discussion paper
Dahmen E (1971) Entrepreneurial activity and the development of Swedish industry, 1919-1939.
Irwin, Homewood, III
Dosi G (1984) Technical change and industrial transformation. Macmillan, London
Dosi G (1988) Sources, procedures and microeconomic effects of innovation. J Econ Lit 26:
1120 1171
Dosi G, Freeman C, Nelson R, Silverberg G, Soete L (1988) Technical change and economic theory
(eds) Frances Pinter, London
Dosi G, Giannetti R, Toninelli PA (1993) Technology and entreprise in a historical perspective. The
Clarendon Press of Oxford University, Oxford
Dosi G, Kaniovski Y (1984) On "badly behaved" dynamics. Some applications of generalized urn
schemes to technological and economic change. J Evol Econ 4:93-123
Dosi G, Pavitt K, Soete L (1990) The economics of technical change and international trade.
Harvester Wheatsheaf Press, London
Dosi G, Tyson L, Zysman J (1989) Trade, technologies, and development: framework for discussing
Japan. In: Johnson C, Tyson L, Zysman J (eds) Politics and productivity. Ballinger Publishing,
New York
Durleifer B, Kocka J (1993) The impact of the pre-industrial heritage. Reconsiderations on the
German patterns of corporate development in the late 19th and early 20th century. Industr
Corp Change 2
Enos J, Park W (1988) The adoption and diffusion of imported technology: the case of Korea.
Croom Helm, London
Ernest D, O'Connor D (1989) Technology and global competition. OCDE, Paris
Fransman M (1986) Technology and Economic Development, Wheatsheaf Brighton
Fransman M, King K (1984) Technological capability in the Third World. Macmillan, London
Freeman C (1982) The economics of industrial innovation 2nd edn. Frances Pinter, London
Freeman C, Perez C (1988) Structural crises of adjustment: business cycles and investment
behavior. In: Dosi et al (1988)
Geroski P, Jacquemin A (1988) The persistence of profits: a European comparison. Econ J 98:
375-389
Granovetter M (1985) Economic action and social structure: the problem of embeddedness. Am
J Socio191:481-510
Guerrieri P (1993) International competitiveness, trade integration and technological interdepen-
dence in major Latin American countries. Presented at the 9x Convegno Scientifico, 30 Sept. 1
Oct. Milano
Herbert-Copley B (1990) Technical change in Latin American manufacturing firms: review and
synthesis. World Devel 18:1457-1469
Hirschman AO (1992) Linkages in economic development. In: Rival views of market society and
other recent essays. Harvard University Press, Cambridge, Mass
Technological paradigms, patterns of learning and development 267

Hobday H (1984) The Brazilian telecommunications industry: accumulation of microelectronic

technology in the manufacturing and service sectors. Report prepared for Unido, Brighton,
SPRU, Sussex
Justmann M, Teubal M (1991) A structuralist perspective on the role of technology in economic
development. World Devel 19:1167-1183
Katz J (1983) Technological change in the Latin American metalworking industry. CEPAL Rev.
19:85-143
Katz J (1984a) Domestic technological innovations and dynamic comparative advantage. J Devel
Econ 16:13-38
Katz J (1984b) Technological innovation, industrial organisation and comparative advantages of
Latin American metalworking industries. In: Fransman M, King K (1984) Technological
capability in the Third World, Macmillan, London.
Katz J (1986): Desarrollo y crisis de la capacidad technologica Latinoamericana. CEPAL, Buenos
Aires
Katz J (1987) Technology generation in Latin American manufacturing industries. Macmillan
Press, London
Kim L, Dahlman C (1992) Technology policy for industrialization. An integrative framework and
Korea's experience. Res Pol 21:437-452
Kogut B (1992) National organizing principles of work and the dominance of the American
multinational corporation. Industr Corp Change 1:285 326
Kogut B (1993) Country competitiveness, Oxford University Press, Oxford
Lall S (1981) Developing countries in the international economy. Macmillan, London
Lall S (1982) Developing countries as exporters of technology. A first look at the Indian experience.
Macmillan, London
Lall S (1984) India's technological capacity: effects of trade, industrial science and technology
policies. In: Fransman M, King K (eds) Technological capability in the Third World. Macmil-
lan, London
Lall S (1992)Technological capabilities and industrialization. World Devcl 20:165-186
Lundvall B (1992) National systems of innovation: towards a theory of innovation and interactive
learning. Pinter Publishers, London
Misa TJ (1985) Military needs, commercial realities, and the development of the transistor, 1948-
1958. In: Smith MR, Military enterprise and technological change. MIT Press, Cambridge,
Mass
Misa TJ (1991) Constructive technology assessment: cases, concepts, conceptualization. Presented
at the Conference on Constructive Technology Assessment, Twente, The Netherlands, Sept.
1991
Muller DC (1990) The dynamics of company profits. Cambridge University Press, Cambridge
Nelson R (1981) Research on productivity growth and productivity differences: dead ends and
departures. J Econ Lit 19:1029-1064
Nelson R (ed) (1993) National systems of innovation. Oxford University Press, Oxford
Nelson R, Winter S (1982) An evolutionary theory of economic change. Harvard University Press,
Cambridge, Mass
Noble D (1984) Forces of production, Knopf, New York
Pack H, Westphal L (1986) Industrial strategy and technological change: theory versus reality.
J Devel Econ 22:87-128
Pavitt K (1984) Sectoral patterns of technological change: towards a taxonomy and a theory. Res
Pol 13:343-375
Romer P (1994a) New goods, old theory, and the welfare cost of trade restrictions. J of Devel Econ
43:5-38
Romer P (1994b) The origins of endogenous growth. J Econ Perspect 8:3-22
Rosenberg N (1976) Perspectives on technology. Cambridge University Press, Cambridge
Rosenberg N (1990) Why do firms do basic research (with their own money)? Res Pol. 19:165 174
Rosenberg N (1991) Critical issues in science policy research. Sci Publ Pol 18:12-18
Silverberg G, Dosi G, Orsenigo L (1988) Innovation, diversity and diffusion: self-organization
model. Econ J 1032-1054
Teitel S (1981) Towards an understanding of technical change in semi-industrial countries. Res Pol
10:127-147
268 M. Cimoli and G. Dosi

Teitel S (1984) Technology creation in semi-industrial economies. J Devel Econ 16:39 61

Teitel S (1987) Towards eonceptualisation of technological development as an evolutionary
process. In: Dunning J, Mikoto U (eds) Structural change, economic interdipendence and world
development. Macmillan, London
Teubal M (1984) The role of technological learning in the exports of manufactured goods: the
caseof selected capital goods in Brazil. World Devel 12:349-865
Teubal M (1987) Innovation and development: a review of some work at the IDB/ECLA/UNDP
programme. In: Katz J (ed) Technology generation in Latin American manufacturing indus-
tries. Macmillan Press, London
Thomson R (1993) Learning and technical change. Macmillan, London
UNCTAD (1989) Impact of technological change on patterns of international trade. Report by
UNCTAD secretariat, Geneva
UNCTAD (1991) Transfer and development of technology in a changing world environment: the
challenges of the 1990s. Report by UNCTAD secretariat, Geneva
Winter S (1982) An essay on the theory of production. In: Hymans SH (ed) Economic and the world
around it. University of Michigan Press, Ann Arbor
Winter S (1984) Schumpeterian competition under alternative technological regimes. J Econ Behav
Organiz 5:287-320
Zysman J (1994) How institutions create historically rooted trajectories of growth. Industr Corp
Change 3:243-283