1. Introduction

Standard economic theory regards innovations as exogenous, reducing the problem to the decision to adopt new techniques and to the ensuing effects. Other approaches, in particular the evolutionary and the sequential approaches, are more interested in the processes generating the innovation, that they consider as being, at least in part, economically endogenous and deserving to be studied from such a viewpoint.

Although it happens that innovations are set up by single firms, many scholars share the feeling that innovative processes emerge or take place within, or involve, or are allowed by, the complementarity or the co-operation of a cluster of different but someway connected firms. There is the feeling that the emergence of innovative processes has to do with that of innovative environments.

The economists, however, often tend to have a short memory. Most of the conceptual categories which are used for the investigation of the case of innovative systems have been created by the scholars dealing with underdevelopment. The economists tend also, sometimes, to overlap their concerns in a disorderly way, forcing old-fashioned issues, such as the need to promote development, into more trendy labels, such as that of innovative systems. Most of the efforts and the studies concerning, for example, scientific parks, in particular those created in otherwise backing areas, hide the bare need to foster the development of such areas and appear as having little to do with the inducement of the creation of innovations. Furthermore a continuous confusion between the case of the adoption of new technologies (and its promotion) and that of the creation of innovation is made.

This explains why we felt the need to convey order into the scattered efforts pivoting around the issue of innovative systems, providing a more accurate definition of the conceptual categories and the involved analytical tools.

2. Differences of focus

In this section we will point out, in a very essential way, how different theories deal with the analysis of innovation. Next, we will shift our attention to the more neglected development theory.

Standard economic theory focuses its attention upon basically synchronic, competitive interactions between agents. The bulk of economic phenomena is regarded as the act of purchasing or selling. Production takes place as an act itself, occurring within the sphere of a single agent, the firm, whose organisational features are basically regarded as unimportant. All forms of relations between firms are reduced to market transactions (selling and purchasing of goods and services). The only considered form of co-operation among firms is the one embodied in the exchange. Co-operation overlaps thus continuously with the mechanisms of competitive forces, despite most recent attempts to embody within the neoclassical framework some of the implications deriving from asymmetries of information, and some of the features brought in by the institutional school, above all by Williamson 1985.

Innovation is consistently regarded as an act occurring in given moments of time; in a world made of basically synchronic interactions among the agents, what is investigated is the decision to adopt them and how this affects the equilibrium.

This way of regarding innovation reflects the way of looking at capital accumulation and its effects. Firms choose among a set of available techniques, solving a problem of long-run profit maximisation. No attention is given to the fact that production takes time, and, in particular, that the building up of productive capacity takes a relevant time-span; equally overlooked are thus the problems of the uncertainty and of sunk costs connected to such processes of construction.

Those who are instead mostly interested to the processes of innovation, above all those who follow the evolutionary approach and those who regard innovations in terms of the feasibility of ex ante appropriate sequences, sometimes referred to also as neo-Austrian approach ([1]), regard as important other features, such as co-operative, diachronic interactions between agents, and not only competitive and synchronic ones.

The firm is no more a black box; both its external and internal relations are considered as important, together with the internal and external organisational features that allow such interactions and contribute in shaping them.

They regard all these phenomena as long-lasting activities which feed cumulative processes; activities and processes are regarded as being highly characterised by changing information sets and learning ([2]).

Non-trivial relations are regarded as specific and being constructed through time; since each agent builds up them with well-specified external and internal agents, the process of construction of relationships is specific itself; the investments associated to it, imposing sunk costs, are specific too, and usually cannot be substituted by spot market transactions.

Among the most interesting dynamic aspects which are considered by the evolutionary approach there is the alternation between changes occurring within a technological paradigm (adoption and diffusion of technologies belonging to the paradigm) and ruptures, or discontinuities associated with a change of paradigm (Nelson, Winter 1982). Although the innovations emerging along a paradigmatic path might be associated in broad terms to a learning process, there is something more than bare "learning by doing": innovations are intentionally pursued in order to obtain infra-paradigmatic innovations (Dosi 1982; Dosi 1988a; Dosi 1988b; Pavitt 1984).

The resources allocated to such efforts are either specific or made specific through intentional efforts; innovations are thus clearly regarded as a process, involving time and non-anonymous relations among agents.

The analysis of the discontinuity generating a new paradigm is less clear-cut. The basic idea appears to be that sooner or later the exploitation of the innovative potential of a given paradigm faces decreasing returns to further investments and efforts. This should push the agents to increase their efforts to search for new paradigms. Although the way of posing the issue of a paradigmatic change lets glimpse the existence of a process generating it, neither the features of such a process are fully developed, nor a full-fledged consideration of the economic factors influencing its emergence are deepened.

As a result we have a framework where innovation is almost completely exogenous when it concerns the change from one paradigm to another, while it is ancipital when it occurs along a paradigmatic path. The latter one is endogenous, in fact, in so far as it might be seen as a result of specific economic efforts and mechanisms. It is exogenous, instead, so long as the occurrence of actual innovations responds to laws of "probabilistic mechanics", which strictly depend on the scientific-engineering features of the given paradigm. These features suggest dynamic behaviours characterised by a substantial degree of determinism, associated however to stochastic disturbances: there is a long cycle, the one between two paradigmatic ruptures, and "patterned" sequences of shorter cycles corresponding to the different phases of exploitation of the potential of a given paradigm.

In the sequential approach, instead, the focus of the attention is brought on the economic aspects and mechanisms which shape a process of creation of technology ([3]). The process is allowed and conditioned by an appropriate sequence of interactions between expectations, decisions, availability and activation of human resources and of money assets.

The system is put into motion by changes of expectations. The perception of new possible needs and options might lead the producer to make new processes start off, hiring and putting at work the labour which is, or is made available in the construction of new specific inputs. The transformation through time of the inputs will carry with it corresponding specific learning processes, which will contribute to give shape, on the way, to the new technology-product.

Sunk costs are necessarily met during this time/process, whose feeding in terms of workers and liquid assets conditions its viability and generates feed-backs in the economy that might contribute to the financing and to the qualitative shaping of demand.

The attention is shifted from the scientific-engineering aspects of innovation to its economic aspects and features. Innovation is made as far economically endogenous as it is conceivably possible ([4]). Thus, the dynamic path shown by a neo-Austrian system is not bound to the mechanical alternation of long cycles of the evolutionary approaches (although it can admit them); short-run cycles are instead possible, but they do not depend necessarily on scientific/technological dynamics, but on the interaction, strictly economical, between expectations, decisions and events (which might or not confirm the expectations which led to the decisions generating the events).

Not only the confirmation of expectations, but the perception "on the way" of new options might relaunch the creation of technologies, even out of the case of crisis of previous paradigms.

On the other hand, the sequential approach finds itself presently in a relatively backward stage for what concerns micro, and micro-macro relationships; the need to deepen the dynamics of the system's variables has necessarily restricted, at least for the moment, the possibility to enquire about inter-subjective relationships. The few efforts made in this direction (Gaffard 1987; Queré 1987; Amendola, Bruno 1990; Must 1990; Gaffard 1990) are not only insufficient to fill the gap, but not yet strictly interlocked with the time-sequential structure such as it has been defined by Amendola, Gaffard 1988, although strongly influenced by it.

3. Development versus innovation theories: overlapping hints

The point we want to raise is whether development -intended as a process through which a virtuous circle of expansion of the economic magnitudes is put into motion and kept going in a previously stagnant environment- has something in common with a process of innovation. In order to investigate about it, we shall cross-survey a series of hints and outlooks developed by the two groups of theoretical contributions.

Innovation theories make a sharp distinction between the creation of technology and of its adoption/diffusion. Creation implies the appearance of something completely new. The technology has to be original, and since it is hard to conceive presently the existence of a completely insulated system, such an originality should apply at the world-wide scale ([5]). In all of other cases we should strictly speak of adoption/diffusion.

However, no specialised scholar would be content of such a sharp statement. Plenty of empirical investigations have shown that the processes of adoption and even more of imitation involve creative efforts, which are the more evident the more the environment where the adoption or the imitation take place differs from that in which the primitive innovation had been created.

Furthermore, attempts of adoption or imitation happen to fail or to succeed; suggesting that the presence and the patterns of a certain amount of creative attitudes have a role and can make the difference between success and failure.

Amendola, Bruno 1990 offer some explanation for this, when they define a major innovation "as one that produces a displacement of competing firms"; this implies "(i) the classification of the innovation on the mere basis of its economic meaning, independently of its source; and (ii) its connection with the quality and the intensity of the rupture produced on the specific environment of the innovating firm" ([6]).

In this sense, thus, innovation implies that some environmental agent breaks a routine and that such a breaking involves a relevant part of his environment; the environment is put into motion along some constructive paths. Could not such propositions apply to the emergence of any development process?

Getting out from the routine and putting things into motion, however, is not enough. Building up a self-feeding process of construction of new activities, be them a brand-new innovation or the setting up of activities which are new only for the specific environment, presupposes some form of co-ordination between agents and through time. And this, again, is a problem which is faced by any development process taking off in an otherwise stagnant environment.

Further suggestions may be derived by comparing a certain number of overlapping between the great literature on underdevelopment of the '40s and the '50s and some pieces of the studies on innovation, mainly those concerning local systems. We can read the literature on development now (with the advantage of the perspective gained since then) as a continuous swing of focus between two poles: that of the motivations and incentives to take off and that of systemic co-ordination. Concerned with the first are mainly the theorists of the "unbalanced growth", such as Hirschman 1958 and Streeten 1959, with the second one those of the "balanced growth", such as Rosenstein Rodan 1943 and Nurkse 1953.

Rosenstein Rodan is the one who stresses more the problem of horizontal co-ordination, i.e. the co-ordination among different agents and activities. The famous example of the shoe factory ([7]) is used for arguing the need to create reciprocal external economies between a set of simultaneous complementary investments. A plan is needed as a co-ordination factor.

The need to pay attention to complementarities as something requiring intentional co-ordination, and thus to be planned is considered also by the literature on innovative systems. These appear to be connected someway to the specifities of the complementarities, while their structuring and their persistence appear to be favoured by explicit descriptive factors that have become thus a focus of investigation: the co-operation agreements among firms and the territorial proximity.

Local systems of production have been indicated recently as possible innovative systems, refreshing Marshallian hints on local districts and external economies (Marshall 1920). Marshall regarded external economies as cost reductions for the firms, due to the intensive or concentrated growth of the industry to which they belong.

External economies for single firms amount thus to economies of scale for the whole industry, occurring in local districts because of the territorial concentration of self-reinforcing factors and relationships ([8]), such as the formation and the availability of skills, the intensive use of specialised technologies, the networking of complementary activities, etc. (Becattini 1987).

More recently other complementary activities have been indicated as behaving in synergy with territorially concentrated industries, such as finance and specialised services (Antonelli 1986), while the reference to external economies has been someway "specialised" using the term "agglomeration economies", in order to stress the importance of the factors of proximity for meaningful evolutionary interactions between firms.

Proximity, however, could hardly be considered an explanatory variable for the emergence of innovative systems (Bellandi 1987; Tinacci Mossello 1987), since the observation of a set of specificities connected to a particular territorial environment should lead rather to reflect upon the intrinsic features of the specificities than upon their being agglomerated around a certain territory; proximity is a pervasive feature of human society, while the emerging of innovative systems arises only occasionally.

Rosenstein Rodan's external economies have no specificities, while in the literature on local systems specific systems of relationships are considered, deriving from factors such as technologies, cumulative personal linkages, institutional actions, etc. What appears to emerge is thus the interlocking between objective and subjective factors, and thus the issue of motivations and incentives both to move and to establish adequate mechanisms of co-ordination.

Among the scholars dealing with underdevelopment, it is Hirschman who appears to pay more attention to such topics. A great attention is devoted to constraints and bottlenecks as a possible source of incentives and motion. The joint effect of investment indivisibilities, shortage of capital and lack of incentives rule out, as fundamentally unrealistic, the hypothesis of a big balanced push.

Investments cannot but occur according to unbalanced patterns; many of these unbalances, however, will be the source of further investments aimed at correcting prior distortions, in particular those introduced by previous leaps in the construction of productive capacity.

The likelihood of such a sequence would be favoured by the existence of strong forward and backward linkages embodied in the particular activities that have been set up with the "distorted" investment, i.e. by the need or by the option which are felt to "complement" such activities with other ones to be set up.

If we look at these analytical considerations with the lenses provided by Amendola, Gaffard 1988, it is easy to discover the blinking of two of their central analytical categories, that of sequential processes and that of vertically integrated process; however the two ingredients appear to have been too loosely associated, so that the feasibility of the dynamic process remains totally undefined.

The concept of external economies is used by the unbalanced growth theory in a more interesting way than in the theory of balanced growth, being regarded rather as dynamic complementarities stimulating new undertakings than as mere synchronic complementarities (Streeten 1959). In the dynamic complementarities (both in consumption and in production), and in the forward and backward linkages the advocates of unbalanced growth end up embodying factors for stimulating new undertakings, but also for easing the co-ordination among complementary agents.

In unbalanced growth theory, finally, there is a consciousness, although confuse, of the possible unbalance between savings and investments (Streeten 1959, final section) ([9]).

This can be regarded as the intuition of possible dynamic difficulties concerning the need to keep consumption sufficiently high to maintain the incentive to invest, but also savings sufficiently high to allow faster accumulation. Besides, we are conscious by now (Amendola, Gaffard 1988; Amendola 1991; Amendola, Gaffard 1991a; Amendola, Gaffard 1991b) that it is a change, either implying the creation of new goods or an acceleration in the rate of capital accumulation, which implies, in itself, a "distortion" in the previous state of flow-magnitudes, and this appears again as someway (but obscurely) perceived by such scholars.

If we now go back to the innovation literature, we find a similar attention to the specificities of complementarities, together with their more subjective aspects concerning the motivations and the organisation of complementarities, in the contributions dealing with the co-operative agreements between firms for innovation and in some more recent perspective in the relationships between local systems and innovation.

The constant reference of this literature is to Richardson 1972, according to whom we should regard industry as a set of "activities .. related to the discovery and estimation of future wants .. which require appropriate capabilities, or, in other words, .. knowledge, experience and skills" ([10]). Such activities "are complementary when they represent different phases of a process of production and require in some way or another to be co-ordinated" ([11]), both quantitatively and qualitatively. "Now this co-ordination can be effected in three ways: by direction, by co-operation or through market transactions" ([12]).

For "closely complementary activities" the co-ordination requires "the matching, both qualitative and quantitative, of individual enterprise plans", since the market transactions would not allow for the required specificities, while the "direction" within the same organisation would fail because of the lack of differentiated capabilities ([13]).

Anyway, the extension of the use of this kind of complementarities to innovative systems appear to stem from Richardson's further observation that "manufacturing activity is technology producing as well as technology dependent .. but technology is a very special commodity and the market for it a very special market" (sect.VI). Later on we will provide a critical evaluation of these suggestions ([14]).

Concerning the local systems of innovation, the "thickening" of positive social and economic relations observed in industrial districts (Becattini 1987) might rather be the result (Gordon 1990; Rallet 1991) of a process of construction of relations led by a firm or by a cluster of firms in a given territory, than a feature of the territory itself, as it might appear from the numerous approaches which use the territorial analysis of externalities in a normative way.

We need thus to go back to the features of the innovative process itself, considering the local systems only as the specific environment where a set of conditions -to be investigated according to the styles of co-ordination and to the incentives which motivate it- have allowed or favoured the conditions for the feasibility of the innovation process (Must 1990). The spreading of learning processes and the width of the collective appropriation of the innovative outcome appear to be correlated more to forms of horizontal co-ordination, evocative of those analysed by Aoki 1986 for the Japanese industry, than to "vertical" (hierarchical) relations, which tend to occur when the system of relationship is established by a company playing the role of the leader (Must 1990).

It is thus the organisational patterns of the structure of relationships and their "styles" which appear of major interest, rather than the specific content of the relationships themselves.

4. Richardson's contribution

A remarkable forward step in the clarification of many of these issues may be derived from the insufficiently acknowledged contribution which had been made by Richardson 1960.

He criticises the theories of general equilibrium in perfectly competitive markets arguing that in such markets the information which is needed to reach the equilibrium is missing. The relevant degree of stability and order of the real world is the factor which instigate the economists to believe that equilibrium can prevail. The point, however, is that the real world is not made of perfectly competitive markets (Ch.1, section 4).

The key argument used by Richardson is someway the reverse of that of Rosenstein Rodan: the potential option for profits generated by a shortage of a specific commodity and signalled by the increase of its price might remain unexploited because of lack of co-ordination. Since investing takes time, the commonly available information about the investment option may remain unanswered so long as none of the potential investors would dare to invest without a previous reliable knowledge of the intentions of other potential investors; should all of them decide, in fact, to invest, a systemic overinvestment would result and losses would be incurred.

Only co-ordination, or equivalent pieces of information, might solve such a stand-still. This explains why a great part of his book is devoted to the analysis of how so called "imperfect markets" should be interpreted as attempts to establish the co-ordination mechanisms that are necessary to allow market economic activities to work and evolve.

The most relevant part of Richardson's analysis is, however, the one concerning complementary investments. The individuals are independent, but their activities are related. In particular, concerning investments:
(i) the investments for the production of interchangeable commodities undertaken by different agents will reduce the returns from the investments for each of the undertaking agents (competitive investments);
(ii) investments are instead complementary (or reciprocally supporting) when the profitability of one of them is increased by the realisation of the other ones;
(iii) a further type of complementarity stems from the fact that any type of production produces income, and this feeds the demand of other commodities (Ch.2,2).

From this, Richardson derives the important hint that any single investment will produce profit only at the condition that (a) the amount of competitive investments will remain below the limit of actual demand; (b) the amount of complementary investments will go beyond a given minimal threshold.

This further explains why a great part of his book is devoted to the analysis of how so called "imperfect markets" should be interpreted as attempts to establish the co-ordination mechanisms that are necessary to allow market economic activities to work and evolve.

From a single company viewpoint, the firm should rely upon the fact that the competitive investments are not in excess, and that the complementary ones are sufficient; in the first case what is desired is the lack of action of other agents, in the second one their positive action (Ch.4,sect.2).

Richardson appears to be conscious of balanced growth theories concerning complementary investments ([15]); but these - according to him - need not to be simultaneous; a satisfactory result might be achieved through a series of capital expansions undertaken sequentially, unconsciously following Streeten 1959 (in particular sect.2).

However, ordinary expectations of expansion, and even implicit collusion, may be insufficient in order to induce complementary investments. Firms must be reassured on the likely profitability of such investments before venturing in costly and irreversible undertakings.

The unlikelihood of spontaneous co-ordination varies inversely with the size of the system and/or the number of potentially involved firms (for a given state of indivisibility of the minimum investments).

The larger the group, the more accustomed are the firms to anonymous relations: firms tend in fact, in such a case, to relate their expectations to the behaviour of the system rather than to that of specific firms ([16]), as it happens instead when the relevant system is small.

When the behaviours within a small group matter, various forms of non implicit collusion are needed. The form of collusion must be more stringent (up to contractual agreements) the more the conditions of complementarity, in terms of temporal and qualitative specificity, are stringent and/or less flexible.

However, the more the terms of agreement are stringent (and thus the more an agent is reassured on other agents' behaviours), the more rigid will the firm find itself in face of residual uncertainty (unforecasted variances and disturbances), because of the reciprocal nature of the agreement, that will impose on him equally stringent conditions (Ch.4,sect.3).

If the problem of co-ordination is an important one in the case where a purely quantitative change of a given bundle of final or intermediate commodities is postulated, it becomes with no solutions, out of explicit agreements, when the possibility of qualitative competition is considered (Ch.6,sect.7).

Richardson's enlightening contribution deserves some comment. Although he does not say it explicitly, his analysis implies that the more an economy is made to rely upon the information and the co-ordination which is spontaneously generated by pure market transactions, the more such an economy runs the risk of stagnation and of slow growth and evolution.

On the contrary, the more the actors tend to pose intentionally, i.e. in an explicit and positive way, the issue of co-ordination, above all for complementary investments, the more likely their economy will be of experiencing growth and evolution.

This stems from a contradictory attitude of Richardson about equilibrium. While consciously criticising it as something that may be generated by perfect markets, what he really challenges in practice is the possibility of getting out from what we might call an "equilibrium trap". More appropriate would be, according to us, to talk of "phases of rest", to mean those periods during which the agents do not feel pushed to undertake relevant changes of their decisions, with the likely result that not only the majority of the agents, but the system as a whole, is at rest ([17]). Extrapolative expectations prevail since they are systematically confirmed. The real problem is thus why and/or how a system gets out of such a state of rest.

This hint appears to us not only perfectly consistent with Richardson's underlying vision of economic interactions, but also surprisingly up-to-date.

For Richardson, for a systemic equilibrium to hold, (i) individual agents must have no reasons to change their previous behaviours and (ii) the behaviour of each of the agents must be consistent and compatible with that of all of the other agents.

Since each agent's behaviour depends on his vision and on his expectations, what is needed is simply that he firmly believes that his vision and expectations are optimal ([18]), not that they actually are optimal according to an optimally informed external observer. However, in order to allow the possibility of such a belief, the agent's plans and behaviours have to be consistent with the plans and behaviours of all of other agents (Ch.1) ([19]).

This way of posing the problem admits multiple equilibria. We prefer to say, in a more general way, that many routines may result individually and systemically stable. If things go this way, stability, which uses to be the major concern of economists, amounts to be the trap against which to react when we want a system to change, as it happens when we want a stagnant economy to begin to develop.

Any kind of extrapolative expectations tend to confirm and maintain through time a routine to which the system happened sometimes to establish itself. Even random shocks appear at a first glance to have a high probability of being absorbed, so long as they are recognised as truly random, since the desire of each agent to re-establish the unforgotten satisfactory previous routine will push each of them to reproduce the previous patterns of behaviour.

This has nothing to do with Walrasian equilibrium, but simply with the inertia that might affect a reproductive system kept alive by fairly rational, but non omniscient, agents who decide on the basis of their systemically confirmed individual feelings and expectations.

5. No-man's time

The distinction between competitive and complementary relations is of overwhelming importance. Its basic good sense cannot hide the bare fact that standard economic analysis is focused on competitive relations, despite the fact that most of non trivial market transactions are performed in order to implement complementary activities.

These, and the relations which link them, are instead more interesting for the study of the innovative systems ([20]). Furthermore, a better understanding of complementary relationships allows to overcome most of the confuse references usually made to the concept - or rather to the several and overlapping concepts ([21])- of external economies.

On the whole, however, those who have paid attention to the problem of co-ordination among agents have overlooked the relations between the performance of economic activities and the flowing of time, and in particular the fact that producing, and even more the building up of productive capacity, takes a relevant time span ([22]). This applies to development theories, but also, though in a lesser degree, to the evolutionary approaches, which limit their attention to the cumulative nature of most of relevant economic processes, and thus on their path-dependence.

Because of this, the relationships of which co-ordination and co-operation consist necessarily end up in being conceived as substantially synchronous relations, missing thus the most interesting question of how such relations are generated and constructed and whether and how are they likely to change through time, at least in part because of endogenous reasons.

On the other hand, the scholars who have made some attempt to pay attention to the problems raised by the flowing nature of time are those who appear to treat in the loosest and more vague way the issue of complementarities. This applies - at least apparently, as we will see - to the unbalanced growth supporters ([23]), but also somehow (and paradoxically) to Richardson and to the evolutionary approaches.

This is likely to depend on the tendency to keep together too many but poorly defined and anyway heterogeneous objects and factors which are at the basis of complementary relationships, while paying a still insufficient attention to the time specification of the involved phenomena. In order to shed some light on this issue, a radical change in the way of considering the inter-subjective relationships from the temporal viewpoint must be developed first.

The shape and the performance of the socio-economic environment are mostly determined by the behaviours of the agents of whom the environment itself consists. Most of these behaviours have the nature of decisions, or, being their implementation, depend on them. The implementation of most decisions, however, takes some relevant span of time. Things take time to occur and to be perceived, so to stimulate further decisions.

However small, such a real time-interval should not be analytically neglected, since it constitutes the very essence and source of uncertainty. This irreversibility interval -that we shall call "no-man's time" ([24])- is the very root of the fundamental uncertainty which affects any decision maker.

Since many agents have to interact and it is from this interaction that the shape of environment depends, this source of uncertainty for the individual decision makers becomes socially relevant and raises for society serious problems of co-ordination. So long as decisions are taken on the basis of local - and anyway limited - knowledge about what other agents are simultaneously deciding, and since each decision contributes in shaping the systemic outcome, the multitude of decisions may result incompatible. This makes it clear why any form of "horizontal" co-ordination, that is, co-ordination among different human agents, cannot but have intrinsically and unavoidably a diachronic nature. Synchronic relationships between agents can certainly be observed ex post, but can hardly, in themselves, teach us something about the ways the mechanisms of co-ordination actually work: synchronic co-ordination is basically an optical illusion, which might appear to have a meaning only in a world which reproduces the very same features, period after period, as in the steady-state paths of standard dynamic models. Such an artificial world, unfortunately, is of little interest for those who are interested in understanding the real one.

Traditional approaches, be them static, comparative static or dynamic, are based upon equilibrium analysis. Interactive equilibrium implies full horizontal co-ordination. The specific nature of uncertainty, such as we outlined it earlier, suggests that the opposite outcome is the more likely one.

Co-ordination does not occur spontaneously; it is a point of arrival and not a point of departure. Co-ordination has to be constructed through time, experience and learning, and is unlikely to ever become complete and optimal.

Rosenstein Rodan's saying that we should treat the economy as it were a single firm, advocating thus a principle of central planning, is nothing else then the stressing of the need to construct a system of co-ordination, implicitly considered out of the power of the spontaneous forces of the market.

However he appears to have completely missed the diachronic nature of complementary relationships ([25]). The critical considerations developed against balanced growth by Hirschman and Streeten, and in particular their advocacy in favour of unbalanced growth may be regarded as a reaction to this way of regarding the problems concerning the issue of time and of change.

However, there is something more to it. Not only Streeten and Hirschman appear as being someway conscious that the transition from stagnation routines to development necessarily implies the need to pass through various forms of unbalances, but they strongly perceived that the issue of co-ordination should have been connected with that of the incentives and the motivations which might push people to innovation and to co-ordination.

6. Hard and soft complementarities

If we regard again Rosenstein Rodan 1943 with the advantage of today's lenses, his contribution appears related to three further, more or less contemporary contributions: the Keynes of the General Theory, von Neumann of the general equilibrium model, and Leontief's input-output analysis ([26]).

Production generates income for the involved factors of production, and the increased available income feeds the demand for consumption goods (income/consumption interdependencies, reminding of Keynes); such an additional demand, however, will be likely to reproduce (consumption interdependencies), if not for slight changes, the same patterns of the pre-existing one (von Neumann).

Finally, in order to allow the aforementioned increase in the vector of consumption goods, the production of the direct and indirect inputs which are technologically required for that production must increase correspondingly (Leontief: production interdependencies).

Should the whole story depend on such a strict definition of complementary relationships, none of the arguments carried on in favour of unbalanced growth would resist the criticism that unbalanced growth is impossible to achieve, unless the lack of production of complementary commodities were compensated by their availability through increased imports for what concerns production interdependencies, and through imports ([27]) or consumption rationing for what concerns consumption interdependencies.

The advocates of unbalanced growth, however, have a broader and more dynamic concept of complementary relationships. The key for shedding some light may be found in Hirschman 1958 ([28]). The complementarities, such as we have narrowly defined them above, concern commodities that must be produced together in a more or less given proportion. In such a case, according to Hirschman, it makes little sense to speak of reciprocally reinforcing demands.

The demand for the commodity which requires such strictly complementary inputs is the source from which their demands derive. A non trivial case is instead that of an activity, or commodity, whose appearance does not compel the simultaneous appearance of another activity or commodity, but instead may slowly induce their emerging.

This suggests to us the need to distinguish between "hard" (or rigid) and "soft" complementarities. While the former ones are trivial, the latter ones deserve attention because of their ability to induce the setting up of new supplies (usually "new" for the involved system, but possibly completely new) ([29]).

The unbalances due either to bottlenecks or to excess capacity, in other words, have a high probability of inducing reactions ([30] ) which, on their turn, are likely to result in an overshooting, and thus in a new unbalance, and so forth. The kind of swinging development resulting from all this would have the advantage of maximising the inducements to invest; that is, it allows to economise the most important of resources, the ability to take decisions.

This analytical statement deserves some comment. The first is that the process Hirschman describes has some interesting overlapping with Amendola, Gaffard 1988. The activity of production has qualitative feedbacks on the cultural patterns from which changes of demand emerge.

Production, in particular the process of setting up innovations, not only produces learning for those who are involved in their actual production but produces learning for other agents acting in the environment.

In Amendola, Gaffard 1988 this feed-back operates upon the demand for the good whose creation is in process; this depends, however, on the contingent specification of their model (which consists of one single producer and a bundle of homogeneous consumers).

The essential point, which is more stressed in Amendola, Bruno 1990, is that along an innovation process the boundaries between the firm and its environment tend to fade away; impulses to learning are thus continuously generated, affecting not only the involved firm but also the other agents who are, or become related to the innovation, either because they are actually called to actively co-operate in the innovative activity or because their potential needs and cultural patterns are the very roots of the innovative activity.

This makes it clear that "soft" complementarities imply that unbalances have the ability of spreading seeds for further learning. Such ability appears to work more or less according to the following patterns. The excess capacity promotes almost naturally an active search for setting up new activities able to saturate earlier the installed capacity ([31]); the bottlenecks, on their turn, appear to perform - according to Hirschman and Streeten - as a kind of natural signalling device.

The feeling such a picture conveys is that an unbalanced system is able to produce, someway "physiologically", incentives to act and to invest that a balanced system is not, or less able to produce, while the same incentives appear to perform also the role of co-ordination mechanisms. A balanced system, on the contrary, appears to produce less incentives to change, less able to orient itself about the directions of change, more in need of explicit co-ordination.

Despite its richness of suggestions, the above picture does not appear as wholly convincing about two issues: the signalling and inductive properties of the unbalances, and the inverse correlation between incentive factors and co-ordination needs. Let us consider both of them.

While the case of excess capacity is quite convincing, that of bottlenecks needs further clarification. If the case occurs with "hard" complementarities, the bottleneck should correspond either to the activation of imports or to the use of substitute inputs of inferior quality (succedanea).

In such a case the signalling role of the bottleneck is evident, but we are in the cases that Hirschman considers as trivial. In the non trivial cases, instead, a generic feeling that "something is lacking" does not appear as so immediately and almost automatically able to signal a void to be filled.

The feeling must be first perceived by some suitable agent able to creatively interpret it, and then such an agent has to translate it into the gradual definition and setting up of an actual supply option. In other words a learning process has to be constructed, and corresponding investments and risk-taking activities should be set up, as suggested by Amendola, Gaffard 1988.

It is quite understandable that a successful innovation might appear ex post as a reaction to a bottleneck, but this cannot mean that the connection observed ex post corresponds to something equally clear and well defined when perceived ex ante.

A parallel argument might be developed for what concerns co-ordination. In the case the unbalance is due to an excess supply, it is almost natural, for the agent who has made the excess investment, to look for the creation of new activities suitable to saturate the earliest possible the excess capacity, either by setting up himself new options or by establishing new external relationships for inducing them.

In the case of bottlenecks, instead, not only the incentives to act are less evident and ready to operate - as we held it earlier - but the arguments developed by Richardson 1960, on the paralysing effects of unfilled needs of co-ordination for competitive activities, are likely to be at work, exactly as they would in an equilibrium situation.

7. Hard complementarities, vertical integration and time

There is, however, a second source of confusion; it is to ascribe, this time, to the balanced growth advocates (and to Richardson 1960). The production complementarities amount to Leontievian input-output relationships. What standard input-output matrices embody is a system of linkages which connect flows of currently produced inputs among themselves and with the flows of final outputs. This might be expressed as

(I-A)x=y

([32]). What the balanced theory scheme says is that, if the economy has to move from a final demand flow y at time t - that we shall express with y_t - to a final demand flow y_t+1 in t+1, then in t+1 there ought to be a flow production x_t+1= (I-A)^-1y_t+1.

However, this amounts to saying how two situations should be - omitting to explain how, i.e. through which time-consuming adjustments - one moves from the first to the second situation.

If in t+1 there must be more production flows of many of the goods, from t to t+1 more productive capacity must be constructed. Now, not only the time length which is required for the production of different capital goods may, and usually does differ, but such a constructive activity for the capital goods required by each sector normally needs a set of inputs which normally differs from the set which the same sector requires in order to feed its current flow production.

This double problem, which introduces unavoidably a set of "distortions" in the structure of the economy when a change in the pace of development is occurring, appears to be completely neglected in balanced growth theories.

In static input-output models the addition to the capital stocks are taken care of in the form of a component part of the final demand vector. This is of some help, but if, as it is likely, the gestation period of capital takes more than one period and different capital goods have different gestation periods, not only additional artificial sectors should be added in order to allow to take into account the goods in process which are passed through from one period to the further ones, but such a complex system of relations should be "backwards worked out", in order to plan the necessary start-ups in t-z (being z=1,...,n) for obtaining the necessary structure of capacity at the beginning of period t.

In dynamic input-output matrices, instead, the problem of capital formation is taken care of in a different way. The additions to the capital stocks are treated as inputs ([33]) and "described explicitly as serving technologically determined capacity expansion required for a rise in the level of output" ([34]). However, the way this is done does not appear to be suitable for our case, since the capital additions components of the technical matrix cannot but be defined in function of a given rate of steady state growth, while our concern is a change.

Rosenstein Rodan emphasised the problems of horizontal co-ordination connected with the planning of a given mix of final goods, and Richardson reinforced his arguments. What both overlooked, however, was the problem of the implicit time dimension of the technological matrix. This results more evident if we reshape the representation of our economy in terms of vertically integrated sectors, each of them producing a final consumption good.

We can sketchy represent each of the vertically integrated sectors, on its turn, in the form of a PERT along a calendar time scale, whose unit is, for the sake of simplicity, the shortest of the gestation periods ([35]) concerning the involved goods.

In Figure 1, the delivery in the 14th date of the final good produced by sector 8 implies to let start the processes of sectors 4,5 and 7 in moment 0; the process of sector 3 in moment 2; the process of sector 1 in two distinct moments, 2 and 9 (feeding two distinct sectors, 3 and 8), etc.

In a steady state situation such a time structure is completely opaque, since it has no occasions to play a relevant role; each of the involved sectors (firms) is fed with continuity at each relevant interval ([36]). It is only when a change is considered that the time structure becomes relevant (Amendola, Gaffard 1988): in order to have an increase in the production of the final product in t, for example, a change in the quantities produced in sectors 4, 5 and 7 must be scheduled since (t-14).

It is evident that this cannot but raise much more serious problems of planning and co-ordination than those which were frightening Rosenstein Rodan and suggesting him a sort of central planning scheme, basically performing the role of a full integration of different sectors (firms) into a unique company.

He thought basically to a horizontal integration, but he should have thought as well to the vertical integration, to be conceived in terms of the planning of intertemporal, and not only intersectoral (or inter-firm) co-ordination.

There are problems, however. Let us discuss them with reference to a two-sectors model of the economy, one producing a consumption good and the other sector a capital goods. In order to accelerate the rate of growth starting from a steady state situation, the sector producing capital goods must increase the deliveries made to itself, and this might imply the reduction of those made to the sector producing consumption goods, which may then find itself obliged to slow down its envisaged rate of growth ([37]).

The flexibility of the system, which can be regarded in terms of the number of periods which are required to pass from the initial situation to the final envisaged one, varies proportionally to the excess of capacity at which the sectors customary work (Bruno, De Gregorio 1994).

On the other hand, the longer is the adjustment length and the larger are the "distortions" brought in by the adjustment process ([38]), the higher becomes the probability that false interpretations, false signalling and ill-based formation of expectations can emerge, producing abortive feed-backs upon the initially envisaged expansion. This makes more manifest how complex the problem of co-ordination is.

8. Planning, co-ordination and specificities are environment-dependent

The PERT is a planning technique used when the problem to be solved requires the time scheduling of activities not belonging to a technically necessitated sequence. Such a scheduling is an organisational device which helps in making a project feasible and possibly shortening its implementation time. The way a PERT is laid down strictly depends on the features of the environment where the project is to be accomplished.

If the project requires generic inputs and the environment where the project is to be implemented is such that the required inputs can easily be bought (because they are in store) in the nearest shop, the PERT is at the most a simple ordered shopping list. If some of the inputs is not in store, it must be ordered in advance; the operations concerning the generic and ready available inputs have thus to be delayed of a time length which depends on the delay interval between the order and the delivery.

The likelihood of having to order in advance the inputs depends on many factors and circumstances: (i) in the case of a currently available good, whether the extra quantities required are within the range of what is considered by the suppliers as a normal inventory swing; (ii) still in the same case of a currently available good, whether the need occurs in a period of sharply growing demand; (iii) whether the good, although well known, is usually not customary used in the specific concerned environment; (iv) whether the good needs to be tailored upon specific requirements of the buyer.

Although better classifications may be conceived for particular purposes, the one exemplified above has the advantage of showing on the one side an increasing degree of specificity of the required inputs, and on the other one that such a specificity is not an absolute attribute of the input, but instead it depends on the relationships between the features of the input and the quantitative and qualitative features of the environment.

One might be tempted to add a case (v) to the list above: "whether the required input does not exist, and it is thus a still undefined "something" which is simply outlined in terms of the functions it should be able to perform". However, as we will argue in the next sections, this case cannot be added, since it would be contradictory.

The cases (i) to (iv) are referred to activities which could be clearly specified ex ante and therefore more or less accurately planned, at least in principle. Instead innovation - in the sense of creation of innovation - cannot be planned in the same sense, since it is not defined at the beginning, and only along the sequential steps of its construction it becomes increasingly defined. This makes the difference, from an analytical viewpoint, between the case of mere growth or adoption of new technologies on the one side and that of the innovation creation on the other.

We are now thus in a position to attempt some tentative generalisation about the cases of growth and of new technology diffusion. The more the setting up of an activity requires specific inputs, the more the involved agent is obliged to

(a) take care and plan the time scheduling of his undertaking;

(b) establish reliable and specific relationships with the agents from whom he wants to obtain the desired inputs;

(c) meet the costs and face the risks which are connected to (a) and (b) ([39]).

Furthermore, the more specific some, or many of the involved inputs are (up to the point of being new and unique, as in the case of absolute innovations), the more (d) the relationships with the other agents have to be personal, non anonymous (as basically are the pure market transactions) ([40]); (e) the object of such relationships must be the establishment of co-operative activities (as opposed to simple "acts of purchasing and selling") the aim of which is the setting up of an environment for common learning (see in this direction Ciborra 1992).

These last observations have carried us to the threshold separating, from a pure logical point of view, the case of development and technology diffusion on the one side and that of technology creation on the other one. It is our deep belief that such cases should be kept well distinct from an analytical and logical viewpoint, though we are equally strongly convinced that such a distinction would be hard to be maintained in practice, because of the continuous shift from one case to the other and the continuous overlapping of such phenomena. The way the co-ordination of activities may be planned, at least in principle, and organised in the case of development and in that of the adoption of new technologies highly differs from the way the co-ordination of activities may be set up in the case of the creation of innovation.

In order to stress the relevance of such analytical distinction, it will be useful to make reference to a precise issue: that of the inter-firm co-operative agreements for innovation.

According to Richardson 1972, co-operation agreements are needed when the problem is that of "the matching, both qualitative and quantitative, of individual enterprise plans" ([41]). However, the possibility of co-ordinating plans is actually open only at the condition that some fairly detailed knowledge about what should be planned does exist already within the cluster of companies which are deciding to co-operate.

This case might thus occur either in the case of development or in that of the adoption of already existing innovations, but it cannot occur by definition in the case where firms decide to co-operate in order to find out the way of setting up a new innovation, that is, in order to create an innovation in the sense of Amendola, Gaffard 1988 ([42]).

If firms decide to co-operate for the creation of innovation, therefore, the content of the agreement must be else than the planning of the activities - regarded as actual pieces of production along an established time schedule. It cannot be but

(a) the intention to establish a long-run co-operation in certain "directions" for not precisely specified "common aims",

(b) the generic and specific resources they decide to put in common, at the purpose of engaging them in the innovative effort (and thus rendering them even more specific),

(c) the rules and the procedures according to which to take future decisions "on the way", and finally

(d) the rules and the procedures concerning the exploitation of the commonly produced outcome.

That is why we find it subtly misleading, for example, to indicate the network systems of firms - intended as networks of firms linked by strict and well detailed agreements concerning the performance of well defined activities and the exchange of information through well structured languages, formats and standards- as a good model for the analysis of innovative co-operation ([43]). The innovative co-operation is characterised instead by the very impossibility of defining ex ante the content and the structure of the connections, and the involved firms can only agree on intentions and procedures.

This does not imply, obviously, that a network of firms, defined in the strict sense spelled out above, cannot establish a co-operative agreement for the creation of innovation; quite on the contrary, the accumulated mutual knowledge and trust is likely to favour it.

The point, however, is that such a co-operative agreement will be different in its nature and content from, although overlapping with the agreements establishing and keeping alive the previously existing network.

As we said, in fact, development, technology diffusion and technology creation continuously overlap in the real world, and often feed and cross-fertilise each other, so that it is extremely difficult, if not impossible, to single out the more intimate nature of different phenomena of economic change.

This does not eliminate, but instead strengthens the need to make the conceptual distinctions stressed above. Although the practical overlapping stems from the difficulty to make sharp distinctions among the different types and the different width of the learning processes which are always involved in any form of social change, the features, the incentives and the content of co-ordination efforts are different in the different cases.

9. The analysis of ex ante co-ordination

Dealing with innovations and the processes generating them implies to shift the attention from how economic structures and systems are and function to how do they change and why. This shifts the focus from the competitive relations ([44]) to the co-operative ones, which are of interest from the viewpoint of the processes aimed at constructing them through time, efforts and investments.

Furthermore, the attention should be shifted from organisational patterns to the processes of "organising", that is, the activities having as their object the organisational structuring of the relationships among the firms and within them. This way of rearranging the analytical targets allows to shed some light upon the emergence of innovative systems and about what such systems amount to, at least from an objective and material point of view. This should be integrated then by a deepening of the more subjective factors - that is, concerning the cultural, social, political, organisational and learning aspects - which have been left in the shadow in our effort here. According to Amendola, Bruni 1990

"...during the process of transformation the borders defining the environment itself cannot exist since the innovation consists exactly in the fact that some of the elements previously external to the activity performed by the firm are now made internal to its strategy, in accordance with the changes actually taking place in the process of production".

This proposition, if cross-referenced with what we said earlier, provides a useful starting point. It refers, in fact, to the view-point of a single firm. When not a single one, but several firms are involved in the creation of an innovative process, and such firms are connected because they are building or re-moulding interrelationships among them, the proposition applies to all of them. ([45]).

All this might thus be re-stated saying that when a process of creation of innovations appears to be rather a feature of a specific environment - that is, of a specific cluster of actors - than that of a single one, then there must be either (1) a reorienting of the connections among the firms and other innovative agents, or (2) a quality change in such connections; furthermore, (3) at least a part of such changes must be actively pursued by some of the agents belonging to the cluster.

When dealing with processes of development and/or innovation, it is usual to say that the conditions are such as to allow the existing firms to exploit the external economies which exist in the surrounding environment. However, what such external economies really are and how their capture is actually obtained, tend to be left in the shadow.

If one accepts the critical analysis developed here, it is clear that the external economies are something which is possible to observe only in a ex-post perspective ([46]). If we shift to an ex ante perspective, only rather vague options for building-up something new appear to the firms which are involved in a process of innovation (Amendola, Gaffard 1988). Such options might acquire increasingly definite features only as the process of exploration and experimentation takes place and sunk costs are met.

However, in order to proceed in the construction of innovative options, the involved firms have to construct or re-shape their organisational relations responding to the type of stimuli, according to the learning patterns, constructing the systems of links, and facing the sunk costs that have been considered earlier. And it will be such an intertemporal organisational activity which will appear to be, in an ex-post perspective, the capturing and the exploitation of the external economies.

External economies are, by definition, something which exist out of the narrow borders of each of the existing firms. Their capture cannot but occur when firms, because involved in the attempt to build up an innovative process, set up organisational strategies aimed at re-shaping through time the relationships of which the environment is made; that is, when and because "some of the elements previously external to the activity performed by the firm are now made internal to its strategy" (Amendola, Bruno 1990). Therefore such a capture is the fruit of both, a strategic and an organisational change, met in order to develop an innovative choice.

10. Concluding remarks

In this essay we have carefully outlined the objective and the time connected factors of which the process of construction of a system of complementarities is made.

In particular we have shown its intrinsically sequential and time-consuming nature, establishing thus a meaningful link between inter-subjective and inter-temporal relations, hopefully filling some of the gaps which had been left open by both, the literature dealing with innovation and that dealing with development. We have left in the shadow, instead, a certain number of important elements, though we think we have provided some useful frame for their further investigation. We find it interesting to indicate them.

1) The subjective aspects which play a role in the construction of innovative systems ([47]): the activity of structuring organisational relationships requires at least the cultural leadership of one or more agents. It would be interesting to investigate about the reasons and the effects which make for leadership and leadership permanence: size, finance, technology, experience, etc. ([48]). Furthermore, it is within this frame of reference that the specific roles played by institutions should be singled out, being institutions nothing else than one of the possible innovative and/or co-ordinating agents.

2) The descriptive features of the most frequent types of induction of sequential complementarities; in other words, which types of activities or productions are more likely to have the ability to induce sequentially which types of further activities or productions, in which environments, in which phases, according to which mechanisms and patterns ([49]). This is another fashion to put the issue which Hirschman referred to when talking of backward and forward linkages ([50]).

3) Transversal to the two above elements there is a third one concerning the intra-firm organisational patterns and how do they interact with the inter-firm organisational relationships. These appear to be connected to how the firm relates to what is descriptively external to itself (Amendola, Bruno 1990) and to the way certain organisational features favour or instead discourage intra-firm communication and processes of learning and skills formation (Aoki 1986; Nonaka 1992).

4) The ancipital role of imports ([51]). When an innovative system is being built up in a given environment, any demand for complementary inputs addressed outside the concerned environment amounts to a leakage out of the system, so that it provides incentives to production out of the concerned system. The use of imports, however, may have different effects and establish different dynamics according to different types of complementarity ([52]).

5) The relevance or the unfitness of the game theory aspects of co-operation: game theory, at least in its present stage, tends to predict the unlikeliness of the emergence of co-operative strategies, despite contrary empirical evidence. Apart from the counter-arguments carried on by Axelrod 1984, there are further elements in the analysis developed in this essay which suggest the inapplicability of the game theoretic arguments on the unlikeliness of co-operative strategies. They concern on the one side the cumulative, time consuming and sequential aspects of the processes, and, on the other one, the object of co-operation, and thus the features of the pay-offs.

The traditional arguments against co-operative strategies may be put in terms of the expected likelihood of opportunistic behaviours. However these arguments appear to be based on static perspectives ([53]). We made clear instead, along the development of our analysis, the difficulties and the time consuming nature of the processes of building up co-ordination and co-operation. We recalled that the main object of co-operation is the construction of a learning environment. We indicated the cumulative nature of such processes. On the basis of this we can better conjecture about the likely opportunity costs that have to be matched by the gains deriving from the exploitation of an opportunistic behaviour.

Breaking an existing co-operation is likely to imply to meet two distinct elements of costs which are logically interdependent, but which add up one to the other from the economic value viewpoint. The first one is given by the giving up of the gains coming from possible future innovations which might derive from the continuation of co-operation. The second one is given instead from the costs, the risks and the efforts which would be associated to the process of construction of new corporations, after the opportunistic breaking of an old one; in a type of environment and for a kind of productive activity for which co-ordination and co-operation are a real advantage, to have the reputation of an opportunistic dealer might cost really too much.

Such costs may be such as to perform as a strong deterrent against opportunistic behaviours, and the common sharing of this consciousness should work in favour of the reaching of fair co-operation deals.

REFERENCES

AMENDOLA M., 1990; "Liquidity, Flexibility and Processes of Economic Change" in McKenzie L., Zamagni S., (eds.) 1990; Value and Capital: Fifty Years Later, MacMillan, London.
AMENDOLA M., 1991; "On the Time Dimension of Production" paper.
AMENDOLA M., BRUNO S., 1990; "The behaviour of the innovative firm: Relations to the environment" Research Policy 19, n.5.
AMENDOLA M., BRUNO S., INGRAO B., PIACENTINI P., 1984; "Valutazione di fattibilita' di una analisi economica delle interdipendenze funzionali in relazione ai processi di innovazione tecnologica", Dipartimento di Scienze Economiche, Roma, mimeo.
AMENDOLA M., GAFFARD J.L., 1986; "Technology as an Environment: a Suggested Interpretation" Economie appliqué n.3.
AMENDOLA M., GAFFARD J.L., 1988; The Innovative Choice. An Economic Analysis of the Dynamics of Technology, Basil Blackwell, Oxford.
AMENDOLA M., GAFFARD J.L., 1991a; "Efficiency and the Firm Out of the Equilibrium" paper.
AMENDOLA M., GAFFARD J.L., 1991b; "Intertemporal complementarity and money in an economy out of equilibrium", mimeo.
ANTONELLI C. (ed.), 1992; The Economics of Information Networks, North Holland.
ANTONELLI C., PENNACCHI L. (a cura di), Cespe, 1989; Politiche dell'innovazione e sfida europea, Milano, F.Angeli.
AOKI M., 1986; "Horizontal versus vertical information structure of the firm" American Economic Review 76.
AXELROD R., 1984; The Evolution of Co-operation, Basic Books, New York.
BAIN J.S., 1956; Barriers to New Competition, Cambridge, Harvard University Press.
BECATTINI G.(a cura di), 1987; Mercato e forze locali: il distretto industriale, Bologna, Il Mulino.
BELLANDI M., 1987; "La formulazione originaria" in Becattini (1987).
BOHM P., 1987; "External Economies", The New Palgrave, Macmillan, London.
BRUNO S., 1987; "Micro flexibility and macro rigidity", Labour.
BRUNO S., DE GREGORIO C. 1994; "Time and the Dynamics of Macroeconomic Change", Dipartimento di Scienze Economiche, Università di Roma La Sapienza, Mimeo.
BRUNO S., DE LELLIS A., LUCCARINI A., 1993; "The Economic Space between Market and Organisation", in Commission of the European Communities, Fast-Monitor, Cohérence et diversité des systèmes d' innovation en Europe, Annexe 4, Juin , FOP 351, FAST dossier, vol.23-24
CHANDLER A.D., 1990; Scale and Scope: the Dynamics of Industrial Capitalism, The Belknap Press of Harvard University Press.
CIBORRA C.U., 1992; "Innovation Networks and Organizational Learning", in ANTONELLI 1992.
DOSI G., 1982; "Technological Paradigms and Technological Trajectores: A Suggested Interpretation of the Determinants and Directions of Technical Change" Research Policy 11.
DOSI G., 1988a; "The nature on innovative process" in Dosi et alii (1988).
DOSI G., 1988b; "Sources, Procedures and Microeconomic Effects of Innovation" Journal of Economic Literature 26.
DOSI G., FREEMAN C., NELSON R., SILVERBERG G., SOETE L., (eds.) 1988; Technical Change and Industrial Innovation, F.Pinter, London.
GAFFARD J.L., 1987; "La création de technologie. Stratégies d'entreprises et politiques pubbliques" Programme de recherche "Technopoles et dévéloppement", tome 3.
GAFFARD J.L., 1990; Economie industrielle et de l'innovation Paris, Dalloz.
GORDON R., 1990; "Systèmes de production, réseaux
industriels et régions: les transformations dans l'organisation sociale et spatiale de l'innovation" Revue d'économie industrielle n.51, 1er tr.
HAYEK von F.A., 1937, Economics and Knowledge", in Individualism and Economic Order, Henneley, Routledge and Kegan
HICKS J.R., 1974; The Crisis in Keynesian Economics, B. Blackwell, Oxford.
HIRSCHMAN A. O., 1958; The Strategy of Economic Development, Yale University Press, New Haven.
KALDOR N., 1957a; "Inflaçao e Desenvolvimento Economico", Revista Brasileira de Economia, March.
KALDOR N., 1957b; "O Problema do Crescimento Accelerado", Revista Brasileira de Economia, March.
KEYNES J.M., 1936; The General Theory of Employment, Interest and Money, London, Macmillan.
KLEIN B.H., 1977; Dynamic Economics, Harvard University Press, Cambridge Mass.
KLINE S.J., ROSENBERG N., 1986; "An Overview of Innovation" in Rosenberg N., Landau R. (eds.) 1986; The Positive Sum Strategy, National Academy Press, Washington.
LANGLOIS R.N., 1986; Economics as a Process, Cambridge University Press
LEONTIEF W., 1941; The Structure of American Economy, 1919-1929, Harvard University Press
LEONTIEF W., 1987; "Input-output Analysis", The New Palgrave, MacMillan
MALERBA F., 1988; "Apprendimento, innovazione capacità tecnologiche: verso una nuova concettualizzazione dell'impresa" Economia e politica industriale n.58.
MARCHIONATTI R., 1989; "Forme di cooperazione tra imprese nell'economia capitalistica. Una rassegna critica della letteratura" in Antonelli, Pennacchi (1989).
MEADE J. E., 1952; "External Economies and Diseconomies in a Competitive Situation", Economic Journal
MUST, 1990; J.L. GAFFARD (ed.), "Cohérence et diversité
des systèmes d'innovation. Les cas des systèmes d'innovation localisés en Europe", nov.
NONAKA I., 1992; "Managing Innovation as a Organizational Knowledge Creation Process" paper for the Conference "Technologies Strategies in the Nineties" Eni-Isvet, Rome 21 may.
NURKSE R., 1953; Problems of Capital Formation in Underdeveloped Countries, Oxford.
PAVITT K.; "Sectoral Patterns of Technical Change" Research Policy 13.
PENROSE E.T., 1959; The Theory of the Growth of the Firm Oxford University Press.
QUERE' M., 1987; "Evaluation du parc de Sophia-Antipolis: stratégies d'entreprises et milieu technopole" tome 2, Programme de recherche Technoples et Développement, Cnrs-Lest & Latapses, Sophia-Antipolis.
RALLET A., 1991; "Théorie de la polarisation et technopoles", paper for the Conference "Technopoles et autres actions visant à favoriser les transfert de technologie", Nancy 16-18 oct.
RICHARDSON G.B., 1960; Information and Investment, Oxford University Press.
RICHARDSON G.B., 1972; "The Organization of Industry" The Economic Journal, sept.
ROSENSTEIN-RODAN P.N., 1943; "Problems of Industrialization of Eastern and Southern-Eastern Europe" Economic Journal june-sept.
SAPELLI G., 1990; L' impresa come soggetto storico, Mondadori.
SCITOVSKY T., 1954; "Two Concepts of External Economies" The Journal of Political Economy, avril.
SCITOVSKY T., 1959; "Growth: Balanced or Unbalanced", The Allocation of Economic Resources, Stanford University Press.
STREETEN P., 1959; "Unbalanced Growth" Oxford Economic Papers, june.
TEECE D., 1986; "Innovazione tecnologica e successo imprenditoriale" L'industria n.4, oct-dec.
TEECE D., 1989; "Concorrenza e cooperazione nella strategie di sviluppo tecnologico" in Economia e politica industriale n.64, dec.
TINACCI MOSSELLO M., 1987; "Economie di agglomerazione e sviluppo economico" in Becattini (1987).
TORRISI S., 1988; "Apprendimento da cooperazione tra imprese e innovazione tecnologica" Economia e politica industriale n.59.
Von HIPPEL E., 1988; The Sources of innovation, Oxford University Press, Oxford.
Von NEUMANN J., 1937; "Uber ein oekonomisches Gleichungssystem und eine Verallgemeinerung des Brouwerschen Fixpunktsatzes", Ergebnisse eines matematische Kolloquiums 8, Ed. Karl Menger; translated as "A Model of General Economic Equilibrium", Review of Economic Studies, 1945-46
WILLIAMSON, O.E. 1985, The Economic Institutions of Capitalism, The Free Press, New York.

NOTES

[1] Because of the use, by Amendola, Gaffard 1988, of a neo-Austrian model for the representation of the productive process, which vertically and intertemporally integrates the phases of construction of productive capacity and those of production of current output.
[2] Ciborra 1992 points out effectively the difference of focus: "The problem with the transactions cost approach (Williamson 1985) is that it considers trade offs between institutional arrangements in a static way.. However, it may well be that in a turbulent environment an arrangement that has bad transaction costs properties is far more efficient .. than an organisation with good transaction costs in equilibrium .. in high-tech industries the trade off that must be considered is the one between efficiency and flexibility" (p.92). Not transaction costs savings, but "new knowledge is the outcome, and the goal.." of the establishment of specific organisational strategies.
See also Langlois 1986 and Klein 1977, referred to by Ciborra.
[3] It is interesting to notice that Amendola, Gaffard 1988 (and elsewhere) do not consider deliberately the factors that subjectively motivate the starting off of the innovative process. Though certainly interesting, they do not consider such an enquiry within the range of issues that should be investigated by the economic analysis. The "Innovative Choice" (which is the title of their main contribution in the line considered here) is considered thus from the viewpoint of its viability and not of its origin.
[4] With the proviso established in the previous footnote. The deepening of the more technological and organisational aspects are left to more specific and contingent enquiries (Must, 1990; Queré, 1987).
[5] Should anyway a technology be dependant on a knowledge base which could be kept completely secret, also some cases of autonomous imitation might be treated someway as an absolute innovation.
[6] Amendola, Bruno 1990, p.423.
[7] "Let us assume that 20,000 unemployed workers ...... are taken from the land and put into large shoe factory.
...... If these workers spent all their wages on shoes, a market for the products of their enterprise would arise an expansion which does not disturb the pre-existing market, and 90% of the problem (assuming 10% profit) would be solved. The trouble is that the workers will not spend all their wages on shoes. If, instead, one million unemployed workers were taken from the land and put, not into one industry, but into a whole series of industries which produce the bulk of the goods on which the workers would spend their wages, what was not true in the case of one shoe factory would become true in the case of whole system of industries: it would create its own additional market,....." (Rosenstein Rodan 1943, p.205).
[8] Some of the ambiguities of Marshall, who tends sometimes to overlap synchronic factors, such as the economies of scale, and diachronic ones, such as those connected with improvements and innovation emerging through time, have been indicated by Bohm 1987.
[9] Not only an excess of investments with respect to either profitable investments, or private savings (and private investments), but even a pure excess of investments on savings, obtained through inflation and changes of distributive shares (Kaldor 1957a; Kaldor 1957b), are regarded as a possible source for future development.
[10] Factors which are overlooked by the traditional approaches to the analysis of firms and industries (p.888; italics of Richardson).
[11] P.889; italics are ours and are aimed at underlining a certain analogy of focus with Amendola, Gaffard 1988, in particular the implicit reference to a concept of vertical integrated sector.
[12] P.890; italics of Richardson.
[13] Ibidem, sect.V. It is worthwhile noticing the implicit inadequacy of the traditional alternative "make or buy".
[14] Teece 1986 exploits further the complementarity between resources and activities when exploring the advantages possibly accruing to a first-comer under different conditions concerning the appropriability of an innovative technology and the specificity of resources with respect to innovation. In such an occasion more precision is made about the cases where co-operation is needed, but at the cost of making of it a residual option, since co-operative agreements would emerge only when the other forms of co-ordination result as inapplicable (the fear for opportunistic behaviour appears to play a certain role; see on this our last section). Of particular interest along the same line are also the contributions dealing with the co-operation which is embodied in the producers-users interactions as a source (Von Hippel 1988; Teece 1989) or as a factor for the development of innovation (Kline, Rosenberg 1986). For a more systematic treatment see Marchionatti 1989.
[15] Richardson recalls the concept of balanced growth (Nurkse 1953; Rosenstein Rodan 1943) as a contribution.
[16] Richardson argues that this is because of the larger reliability of extrapolative expectations.
[17] This state of rest looks like equilibrium and its features recall those of steady state; however what must be made clear is that this state has none of the properties usually attached to equilibrium in general equilibrium theory, the least one that of optimality.
[18] But to say satisfactory would be sufficient, according to us, since the important thing is that the agent decides not to change.
[19] This vision of equilibrium is very close to that of Hayek 1937.
[20] Furthermore, despite the scarce interest manifested by Richardson, of remarkable interest is also the third type of relationships he considers, namely, those concerning the feedbacks of production upon demand. The reasons for this have been shown by Amendola, Gaffard 1988, and the importance of its role will better appear later on, when the problem of intertemporal complementarities will be further deepened.
[21] Scitovsky 1954 is particularly effective in denouncing the many ambiguities associated to the conceptual category of external economies. In particular Scitovsky shows the strong dependence of external economies, such as classified and analysed by Meade 1952, on an equilibrium situation in a pure competition environment, that is regarded as a static situation where no indivisibilities exist. External economies such as they are discussed with reference to industrialisation processes depend instead on indivisibilities and are fitted into a dynamic and disequilibrium problematical context.
[22] Hicks 1974; Bruno 1987; Amendola, Gaffard 1988
[23] Except for a few, really enlightening pages by Hirschman, as we will see in the next section.
[24] In clear analogy with the no-man's land, the dead land between two countries' borders or two war fronts.
[25] From a dynamic point of view, what Rosenstein Rodan had in mind was probably a rough specification of a von Neuman's type of growth model. Therefore, although he was formally dealing with a problem of change -a change from a stagnant situation to a positive rate of development- what he had in mind was a steady-state path, the transition to which was completely beyond his perception and perspective.
[26] Keynes 1936, Von Neumann 1937, Leontief 1941.
[27] This appears to have been well understood by Scitovsky 1959 , acknowledged by Hirschman 1958 , but fundamentally overlooked by Streeten 1959.
[28] While Streeten 1959 is in many senses the main responsible for the confusion.
[29] The supply of new commercial stores requires the simultaneous supply of cement, steel and bricks; it may require new modern furniture; it can induce the setting up of new activities, such as the opening of restaurants
[30] It would be wrong to induce, on this basis, that Hirschman is thinking only at the market reactions induced by prices. He is very positive in indicating that the society on the whole is able to manifest unsatisfied needs and, when deserved, politically calling for reactions to be undertaken by the public sector, as for example in the case of education.
[31] A good example is provided by the French experience in the Videotex "technology". While other countries (for example UK who began its development earlier) attempted a gradual diffusion of the service, France experienced a clear path of unbalanced development. France had made -probably for strategic reasons- an excess investment in a Transpac network, whose capacity of transporting digital messages was highly under-utilised because of the scarcity of messages to be exchanged. A massive policy for the promotion of a market for such messages was thus initiated, through the gratuitous (or underpriced) distribution of the terminal equipments (Minitel) to the potential customers, and on the inducement of the independent supply of new value added services, whose additional corresponding demand might be transported on Transpac network.
[32] Where A is the square part of the structural matrix describing the unitary input requirements of all producing sectors, x is the column vector of total outputs, and y the column vector of final deliveries of all goods.
[33] And thus transferred in the left hand side of the equation provided above.
[34] Leontief 1987 .
[35] Or the shortest of the "no-man's times" considered earlier.
[36] An interval which does not depend on the time of gestation but on the frequency according to which the sector producing the final output must make its deliveries.
[37] This had been very well perceived and considered by Kaldor 1957. For a careful analysis of these and related problems at the macroeconomic level, see Bruno, De Gregorio 1994.
[38] We consider as distortions due to the adjustment process not only the initial reduction of the external deliveries of capital goods mentioned earlier, but also those affecting real savings and relative prices (Bruno, De Gregorio 1994).
[39] The more the setting up of an activity is innovative, the more is it likely to involve long periods of gestation, and thus higher sunk costs. Amendola, Gaffard 1988 lay down the time patterns of the various magnitudes involved in the process of creation of an absolute innovation.
[40] This was well perceived by Richardson 1960. When the environment is large relatively to a single agent who decides to invest and the inputs are not specific, pure anonymous market transaction would be sufficient. If the magnitude of the devised investment is high relatively to the environment, as it happens in underdeveloped countries or areas and for investments presenting important indivisibilities, the relationships which need to be established tend to be unavoidably more personal. But even in a well developed environment, when the inputs must be specific and/or the magnitude of the orders is relatively large, explicit and direct co-ordination and co-operation will be likely to be needed or preferred.
[41] P.892.
[42] It is obvious that this apply also to what we said about PERT, whose implications should not be equivocated.
[43] There are wide examples of such a tendency in the articles collected in Antonelli 1992. In particular, while we do find it of the utmost interest the study of the network systems (such as telecommunication, energy, transport, water, etc), since they are a sort of geometrical locus of the most intriguing conceivable microeconomic problems, we find it a misuse to exploit the mere descriptive analogy between information networks and co-operating clusters of firms for improving the study of the latters. Information networks are rigid systems of connections, and it is basically because of this that the needed qualitative and quantitative flexibility concerning the flows processed by the system must be assured on the one side through highly structured softwares, very strict standards or specific adapters, and anyway specific pieces of added intelligence, and, on the other side, through a remarkable degree of excess capacity. For a deeper analysis see Bruno, De Lellis, Luccarini 1993.
[44] Which are instead the focus of standard economic theory and that become interesting from a dynamic and transformation point of view only in so far as a change of the competitive rules is concerned.
[45] Obviously not all of the involved firms will play the same role, in terms of leadership, intensity and quality.
[46] Gordon 1990 makes a similar observation about the agglomeration economies.
[47] How does a cluster of innovative actors form itself? Which are the patterns of clustering, their reasons and the factors affecting the clustering process, how are the actors selected (why does it happen that certain firms choose other specific firms to co-operate and skip others), are all relevant questions opened to further investigation.
[48] It would be interesting to see when or whether such leadership is permanent, or whether the firms involved in the innovative cluster alternate each other in the leadership role during different phases of the building up of the innovative process. Equally interesting would be to see whether single or multiple leadership tend to occur in different environments and/or for different types of innovation. As a limiting case we can have spread leadership, as it appears to occur in the case of local industrial districts; in other cases only some of the involved firms will play an hegemonic role.
[49] This has to do also with changes affecting the clusters. There are no reasons why a cluster should be stable and formed once and for ever. Which is the kind of inter-cluster mobility? As we noticed, the roles within the cluster may change through time and through different phases of the process; which are the patterns of such an infra-cluster mobility?
[50] The accumulation of knowledge and methods of investigation which has been achieved by evolutionary approaches might result of enormous advantage for this purpose; of equal importance might be investigations along the lines of industrial history (Chandler 1990; Sapelli 1990).
[51] We refer here to imports in a given environment, not necessarily to imports in a given country.
[52] For the case of the "hard" complementarities, imports may in some case be the condition for a positive take off of the system. The use of succedanea instead of imports, when possible, might have exerted, however, an induction pressure upon the domestic environment that the activation of imports may risk to forestall. Furthermore, the fact that imports assure the availability of certain goods or activities, may even withdraw incentives to the future emergence of an equivalent domestic production. On the other hand, the availability of imported goods and activities may be the basis for learning processes that would be missing otherwise, and constitute thus the prerequisite either for future imitations and import-substituting activities or for the taking off of brand-new activities, allowed by the previous learning processes.
For the case of "soft" complementarities, by definition the absence of imports should not forestall the taking off of the activities which are ready to start, so that it might seem that their appearance should play always a negative role. This, however, might be a way of overlooking the strategic importance of the learning processes that might be activated by such imports, though it would be difficult to deny that when certain options are already exploited and filled by "foreign" companies, this might reduce the incentives to get in the play for the domestic ones, when not forestalling their entry altogether. This, however, should be unlikely when dealing with the kind of complementarities the development of which requires, or is advantaged by, the setting up of a network of strong, specific and actively pursued co-operation among several firms. In such a case, thus, it is more likely that the "foreign" company might play the role of an innovative agent, or otherwise that its readiness or its reluctance to play such a role might make the difference between the emerging or not of an innovative system.
[53] The co-operation for innovation should be likely to produce an actual innovation at a certain point of time; should this happen, each of the involved agents would be tempted by opportunistic behaviours, consisting in the autonomous exploitation of the commonly produced outcome by the agent who will find himself in the most favourable position at the moment of the emergence of the innovation, unless the innovation were legally protectable. Knowing this in advance, the agents would not get involved in co-operation if not in very limited cases. Further risks of opportunistic behaviours might be at work, such as those put into evidence by the principal/agent literature: each of the agents would be brought by selfish attitudes to conceal crucial pieces of his knowledge and information and to cheat about his efforts in favour of the common concern, while being ready to maximise the exploitation of the other agent's knowledge, information and efforts. And the shared expectation of this would, again, prevent co-operation.