'STIG Systems'?

Can we Link Policy Practice with Research on ‘STIG Systems’? Toward connecting the analysis of science, technology and innovation policy with realistic programs for economic development and growth By

Philippe Aghion*, Paul A. David** and Dominique Foray*** * Department of Economics, Harvard University ** Department of Economics, Stanford University & All Souls College, University of Oxford *** College of Management of Technology, Ecole Polytechnique Fédérale de Lausanne Version 1: 2 June 2007 Version 2: 13 Jun3 20007 Version 3: 20 April 2008

________________________-___________________________ Published as Ch. 5 in The New Economics of Technology Policy, D. Foray, ed., Cheltenham, UK: Edward Elgar, 2009: pp. 46-71 ____________________________________________________

1. Introduction: an overview of the argument The conceptualization of “science, technology and innovation (STI) systems” has gained acceptance among social scientists and other policy analysts. The appeal of this perspective has grown with the widening recognition of the existence of a multiplicity of interdependencies among the processes of scientific discovery and invention, technological change, and innovative economic activities, and the intricate connections that the former have with specific features of any given society’s political, legal and social institutions. Behind much of the interest that presently focuses upon that that intricate and still far from thoroughly understood nexus of dynamic interrelationships is the supposition that its structural properties play a powerful role among the determinants of the nature, pace and direction of macroeconomic growth. The processes of long-run growth and development, however, are themselves complex and no less intricately entangled with institutions affecting the growth of knowledge and the distribution of information that touch many aspects of human creative activity besides the advancement of scientific and technological knowledge. It cannot reasonably be imagined, even for theoretical exercises that resource investments in the “STI sub-system” will automatically yield steady flows of “innovation” that somehow immediately “plug into” economic production systems to yield “growth” -even if that is what is depicted simplistically by many of the still fashionable macroeconomic growth models. What is called for, instead, is a more concerted effort to explicitly articulate

-2the multiplicity of dynamic linkages characterizing institutionally grounded “science, technology, innovation and (economic) growth systems”—STIGS, and thus to focus policyoriented studies and proscriptive analyses upon the complex realities of seeking to stimulate development and improve long-run macroeconomic performance through those channels of influence. To thereby begin breaking down the conceptual walls that still compartmentalize STI policy discussions, on the one hand, and “economic growth and development policy” analyses, on the other, and which unfortunately inhibit those areas of sub-disciplinary specialization from more fully informed and fruitful discourse with each other, forms the larger purpose towards which this essay is directed. In order to have any realistic hope of achieving that objective it is necessary to seek a suitably detailed yet manageable integrating framework of analysis, and that exploratory search forms the second and more immediate objective that occupies much of the following discussion. Alternative candidates are available as analytical platforms upon which to begin constructing the sort of expanded conceptual framework that is required. The two starting points that spring most readily to mind here are, firstly, the familiar class macroeconomic growth models and the associated “growth accounting” calculus that draw upon neoclassical production and capital theory; and, secondly, the variety of nonneoclassical models that are more accommodating to Schumpeterian and explicitly evolutionary insights into the dynamics of complex, non-linear systems. Certainly it can be argued that each can commend itself -- albeit on somewhat different grounds, for further adaptation as vehicles of analysis that are logically consistent with the pursuit of enlightened public policies aimed at managing elements of a STIG system that is beset by poorly performing markets. Whether that constitutes a compelling recommendation is an issue that will resurface from time to time in the following text, but should be initially broached here. The traditional preoccupation of contributors to the literature on the economics of “technology policy” has been with market-mediated behaviors and competitive interactions among business firms affecting aggregate or sectoral levels of investment in R&D. Those entities are depicted typically as responding to market-generated signals of profitability, and also to exogenously imposed regulatory incentives and constraints that alter market conditions; firm are thus represented to being predictably and passively responsive to the application of a variety of fiscal and institutional instruments that can be wielded by government policymakers. Conventional usage refers to the latter as “interventions” in the market for R&D investment, thereby reinforcing a general proposition that each public policy action should be regarded as discrete a departure from some norm, and therefore warranted only where the outcomes of resource allocation directed by untrammeled market processes have been found to be in some sense “socially inefficient”. Like other bold abstractions from reality, this one has its uses and its drawbacks, and the intention throughout this essay is to keep both properties more-or-less continuously in view. Since holding two seemingly conflicting viewpoints in one’s head concurrently generally poses something of a challenge, it perhaps will prove helpful to emphasis at the outset the less frequently stated drawbacks of the foregoing all too familiar abstractions. One may start by noticing that while public policy actions are quite often

-3precipitated by specific events, only rarely do they emerge as unheralded isolated responses, and more typically they find justifications as extensions of previously established policyprecedents. The specific instrumentalities employed, and the domains of their application are likely to be interrelated, both politically and administratively. As a practical matter, therefore, the latter are difficult to freely design to be “fit for purpose” despite the urgings of economic advisers; instead, they are constrained by the capabilities of existing government agencies and public institutions, while often reflecting the aspirations of the leaders of those organizations to alter their future capabilities, their sphere of influence, and if nothing else the size of their budgets.. Such recognition and acceptance of the continuing roles and extensive involvements of those public sector agencies in the economy lends some force to Nelson’s (2007) recent critique of the habit of referring to government actions in support of innovation as “interventions” justified by particular “market failures”; and likewise to his contention that the appropriate “orienting question of innovation systems theory” should be concern the identification of the modes of interaction between the array of public and private entities that would best promote innovation, rather than “focusing attention on the question ‘what do markets not do well’”. Turning then to the other element of bold abstraction, the “black-boxing” of the firm for purposes of analytical convenience, clearly, it is pertinent to recognize that the strategies of the large firms’ that collectively are responsible for the bulk of private expenditures on R&D (and nowadays for the larger portion of national R&D outlays in the major OECD economies) are far from passive. Their strategies evolve endogenously, being shaped in part by the ways in which they deal internally with the problems of coordination and informational asymmetries – some of which impart a lack of plasticity and path dependent momentum that render these organizations less than perfectly responsive market signals and government- initiated incentives which (optimistically) are expected to alter their behaviors. Indeed, by acting alone or in concert these firms frequently attempt and sometimes succeed in re-shaping both their markets and their regulatory environments, thereby invalidating the supposition that would-be designers of innovation policies have a free hand to plot and steer an optimal course for the industry and the economy at large. Greater acceptance of these fundamental realities also would have the desirable effect of containing one of the unfortunate consequences of economic theorists readiness to gloss over the “internal life of the firm” when specifying models of the R&Dinnovation- growth connection: the progressive relegation of studies of the nexus of decision involving business innovation strategies, R&D investment commitments, and research management practices to specialists working in the sister disciplines of economics -management studies, sociology of knowledge and organizational science. As Steinmueller (2007) points out, as much as has been learned thereby, a side-effect of this now advanced trend has been to confine these socially significant decision processes within the firm to being explicitly examined and evaluated primarily, if not exclusively in reference to the private objectives of business enterprise. What has been lost in this “division of intellectual labor” is the important broader social welfare-analytic perspective that industrial organization and growth economics would more naturally introduce as prominent topics for discussion.

-4The foregoing prefatory remarks should suffice to make it evident that an appropriate framework within which to link STIG policy research and practices would not only transgress the unmarked present boundary between macroeconomic growth policy and science, technology and innovation policy analysis; it also would call for explicit integration of insights from the genre of political economy research that is now being undertaken in the field of development economics, and the perspectives formed by organizational science and sociological studies of company management affecting research and innovation. To reach beyond those lines of disciplinary demarcation will be quite a challenge, especially while trying to keep a foot on one or the other of the two established conceptual platforms that presently are available as initial bases for systems analysis in this area. We are aware, and so should emphasize that to meet even that challenge will not be enough. Quite obviously it would leave open the question of whether or not the resultant expanded framework for empirical and theoretical analysis would be one within which it is feasible to design and evaluate appropriate policy measures that harness the economic and political resources of particular societies to support a creative STI-subsystem, and then mobilize the individual agents and organizations of the economy to effectively exploit that creativity for welfare-enhancing economic growth. Raising this practical question in concrete ways constitutes the modest third aim of this paper, for, we cannot suppose that we are capable of providing any suitable answer from a priori considerations. To approach that objective, however, we can at least attempt to indicate both the difficulties and the importance of the issues involved. This requires revisiting several familiar science and technology policy “themes” and introducing some of the specific complications that typically are put to one side, but which will be seen to belong within any “STIG-systems perspective” that makes even minimal contact with some of the awkward realities that face the policy practitioner in this field. 1.1 Organization of the discussion The discussion directed towards those three main goals is organized in the following four main parts of the paper. Section 2 begins with an overview of the thrust of contemporary science, technology and innovation policy discussions -- informed by applying the familiar the “market failure” rationale for public policy actions to the sphere knowledge production and distribution. This takes account (in section 2.2) of a larger, and rather more complicated system perspective than was contemplated by the seminal formations of Nelson (1959) and Arrow (1962), because “market failure” justifications now are offered in connection with a variety of research and diffusion problems that involve innovation complementarities, coordination system failures, and phenomena such as system-lock-in to sub-optimal configurations -- associated with the economics of path-dependent process of technological and institutional evolution. Each of those specific instantiations of the modern “market failure” to public action in support of innovation-driven growth calls for a corresponding search in Section 3 for appropriate policy-designs and instruments, including fiscal tools and institutional mechanisms. When the complex nature of the problems arising from informational externalities and asymmetries in the context of research investments are adequately acknowledged, it is seen (in Section 3.2) that a number of basic guidelines for public policy

-5offered by the existing literature turn out to be rather less useful than would appear at first sight. Further complications are confronted in Section 4, which begins by explicitly recognizing a number of critical interdependencies between the subjects of narrowly focused STIG policies and other important spheres of public sector action. These are conceived of primarily from the standpoints of their impacts upon education and training, the distribution of economic opportunity, the efficiency and flexibility of labor markets, the stability and responsiveness of financial markets, effective product market competition. That they may impinge indirectly upon long-run macroeconomic performance is certainly acknowledged in a general way, but Section 4.1 makes the point that their relevance for STIG policy-making warrants greater attention that it typically receives. The potential limitations of narrow growth strategies of “technologically driven innovation” are especially likely to be exposed where insufficient consideration is given to policies that would concurrently address those complementary components of the larger dynamic system. To avoid such errors it seems essential to escape the confines of conventional ”market failure” analysis and try (in Section 4.2) to take into account the reality that market processes in modern economies are powerfully shaped by, and embedded in particular institutions and organizations. Those structures exhibit distinctive evolutionary dynamics that may lead to their having dysfunctional interactions with other parts of the economy system’s organizational ecology, as well as in inefficiencies in their purely internal operations. Exploring that perspective brings the morphology of “institutional failures,” and the logic of institutional reform as a development and growth policy tool, within the ambit of the broader STIG-system perspective. Section 5 then confronts the practicalities and costs of actual policy design and its implementation. Understanding the basic principles of market failures does not carry one very far in the direction of deriving practical recommendations about the construction of effective policy “interventions” (or decisions to defer intervention), particular as these have to be executed in real time, and sometimes in particular sequences if they are to be effective. The difficulties of designing “interventions” for a system of such complexity pose formidable challenges, because at least some among the conditions that call for government policy interventions also imply that important aspects of the system’s behavior may be “emergent properties” that cannot reliably be deduced from a knowledge of the properties of its constituent parts. An attractive path of escape from this conundrum is indicated (in Section 5.2), where it is suggested that greater recourse should be made to the approach and tools that are being developed and deployed in the field of system dynamics particularly the methods of “virtual experimentation” using agent-based stochastic simulation models. The essay concludes (in Section 6) with a number of cautionary word for those, ourselves included, who may hope to become visibly effective in “directing” the processes of scientific advance, technological change and innovative activity along trajectories that improve economic welfare and material well-being of whole societies and regions of the world. The “would-be- managers” here do not stand outside the game, they are inevitably a part, and at best a small and transiently influential part of these proceedings. We suggest why hoping to do more than avert particularly wasteful, or pernicious policy errors that arise from

-6the neglect of empirical evidence in favor of ideological pre-commitments, and from disregard for long-run systemic thinking in the enthusiasm for politically expedient shot-run policy impacts, seems extraordinarily ambitious; and that claiming to have been able to do more than that is quite likely to be a risky exercise in professional hubris.

2. Toward a larger dynamic system perspective for policy analysis The modern economic case for policy intervention in this area (as in others) rests first on establishing persuasive grounds for concluding that in its absence the outcomes would be suboptimal. That step, which is necessary but not quite sufficient for practical policy purposes, is rooted in the now classical formal statements about the problematic functioning of competitive market processes when they deal with information, itself both an input and an output of “research”, as an economic commodity. 2.1 The market failure rationale for policy: public goods and “appropriability problems” Modern economists have followed Nelson (1959) and Arrow (1962) in arguing that the potential value of an idea to any individual buyer generally would not match its value to the social multitude, since the latter would be the sum of the incremental benefits that members of society derived from their individual use of the idea. Those private benefits, however, will not readily be revealed in a willingness to pay on the part of everyone who would gain thereby; once a new bit of knowledge is revealed by its discoverer(s), some benefits will instantly spill over to others who are therefore able to share in its possession at little incremental cost. Why should they then offer to bear any of the initial sunk costs incurred in bringing the original thought to fruition? Commodities that allow themselves to be used simultaneously for the benefit of a number of agents, are sometimes described as being non-rival in use (see Romer (1990)), or has having the property of unbounded “expansibility” (see David (1993: 217)), or to generate “intertemporal knowledge spillovers” (see, e.g., Dasgupta and David (1994), Aghion and Howitt (1998)). This characteristic is a form of non-convexity, or an extreme form of decreasing marginal costs as the scale of use is increased: although the cost of the first instance of use of new information may be large, in that it includes the cost of its generation, further instances of its use impose at most a negligibly small incremental cost. Sometimes this formulation it thought to be defective in ignoring the costs of training potential users to be able to find, or to grasp the import of information, or to know what to do with it. But, although it is correct to recognize that developing the human capability (knowledge) to make use of data and information are processes that entail fixed costs, the existence of the latter does not vitiate the proposition that reuse of the information will neither deplete it nor impose significant further (marginal) costs. A second peculiar property of ideas that deserves to be underscored here is the difficultly and cost entailed in trying to retain exclusive possession of them when, at the same time, undertaking to put them to use. Although it is possible to keep secret a new bit of information or a novel idea, the production of visible results that were not otherwise achievable will reveal (at very least) that a method exists for obtaining that effect.

-7The dual properties of non-rival usage and costly exclusion of others from possession of ideas (or other commodities) define what economists mean when they speak of “pure public goods.” While the term has become familiar, confusion lingers around its meaning and implications. It does not imply that such commodities cannot be privately supplied, nor does it mean that a government agency should or must produce it, nor does it identify “public goods” with res publica, the set of things that remain in “the public domain.” What does follow from the nature of pure public goods is the proposition that competitive market processes will not do an efficient job of allocating resources for their production and distribution. Where such markets yield efficient resource allocations, they do so because the incremental costs and benefits of using the commodity are assigned to the users. In the case of public goods, however, such assignments are not automatic and they are especially difficult to arrange under conditions of competition. The disclosure even of novel commodity’s general nature and significance (let alone its exact specifications) in the course of negotiations for a market transaction can yield valuable transactional spillovers to the potential purchaser, who would remain free to then walk away. Complex conditional provisions in the contracts and a considerable measure of trust are required for successfully “marketing an idea”, and both of these are far from costless to arrange especially in “arms length negotiations” among parties that do not have symmetrical access to all the pertinent information. Contracting for the creation of information goods whose specifications may be stipulated but which do not yet exist is fraught with still greater risks; and, a fortiori, fundamental uncertainties surround transactional arrangements involving efforts to produce truly novel discoveries and inventions. This leads to the conclusion that the findings of scientific research, being new information, could be seriously undervalued were they sold directly through perfectly competitive markets, and the latter would therefore fail to provide sufficient incentives to elicit a socially desirable level of investment in their production.

3. STIG policies and the appropriability problem in complex dynamical system contexts The foregoing describes what has come to be referred to as the “appropriability problem,” the existence of which is invoked ubiquitously in the mainstream economics literature as the primary rationale for government “interventions” to correct the sub-optimal provision of “public goods” of widely disparate sorts, ranging from airline safety, to control of infectious disease, to protection from nuclear attack to scientific discoveries. The recommended policy response to the specific diagnosis of chronic under-investment in scientific and technological research by the private sector is that the public sector should first undertake to increase R&D expenditures, using general tax revenues for the purpose, and then have recourse to subsidies that would have the effect of altering the relative prices and private rates of return so as to create incentives for increased private investments. 3.1 The limits of generic guidelines for pubic policy action A number of principles are advanced as guidance for such interventions, some which turn out to be less compelling than would appear at first sight. The prescription to act so as to bring marginal social rates of return into equality in all lines of investment (public and

-8private) can be helpful in knowing when to stop, but less so if one can’t decide where to start. Should one begin by trying to boost research investments where the positive gap between social and private discounted expected rates of return are largest, accepting private time discount rates or uniformly imposing a social discount rate; and what rate, for that can matter for the policy choice when the investment payoff streams are not monotonic. That these are well-known issues in public finance does not make then any the easier to resolve (as evidenced in the heated recent debates provoked by the Stern Report’s recommendations regarding the appropriate way to value to the benefits that future generations will derive from present expenditures to halt global warming). More troubling still is the absence of any theoretical warrant for the presupposition that the public goods properties of information, and therefore of research outputs, imply that socially inadequate levels of R&D investment will be found everywhere, in all lines of business, firms, branches of industry throughout the private sector. Quite the contrary, inefficient over-investment in R&D is likely to be the condition that emerges where there are “common pool problems” (arising from failures of firms to adequately take into account the likely consequences of others’ investments on their own rates of return); or there are tournament- like pay-offs structures (“winners taking all”, or nearly all in patent races and slaloms down the industrial learning curve); or when imperfect inferences from the observable behaviors of potential rivals induces “herding” in the selection of R&D projects and consequent excess correlation of firms research portfolios (on which, see, e.g. Dasgupta and Maskin (1987), Dasgupta and David 1994). Taking complications of the foregoing sort into account leads this discussion to a significant but sobering pair of conclusions. First, although there may be good reason to suppose that the aggregate level of private R&D expenditures will be socially sub-optimal, intermingled regions of excess and deficient levels of expenditure may characterize large zones of research landscape, making it hard to justify reliance upon any uniform, generic guidelines when allocating public research subsidies to further stimulate private investment. Good public policy in this area cannot be constructed without detailed analysis of specific industrial conditions. Second, and possibly still more discouraging, for government programs to take their cues from the intensity of private sector research interest in favoring particular areas of scientific discovery and technological activity is especially likely to result in further augmenting the tendencies toward social over-investment that, as was just pointed out, are prone to arise endogenously from the interactions of business decisions among rival firms seeking competitive advantage through innovation. The latter proposition is just a particular instance of the more general need to listen with a skeptical ear to the advocates of “neutral” implementations of pro-innovation policies, who maintain that the proper policy course is for governments to fix the aggregate level market failures by providing generic research subsidies but then back off, and leave it to private agents to be guided by local technical knowledge and market signals in making the best use of the resources placed at their disposal. The internal contradiction in that position is apparent: since it is granted that competitive markets can not be relied upon to get the economy to an appropriate aggregate level of investment, by what magic will those same processes manage to allocate a “policy-corrected total” in ways that will turn out to be socially optimal?

-9Of course, the problem of achieving the right distribution of research expenditures among different kinds of projects has not passed un-noticed, even in theoretical discussions of optimal R&D policy. Part of the conventional “market failure” justification offered for government intervention in the sphere of scientific and technological research and development recognized a difference between exploratory, fundamental investigations, sometime labeled “basic research”, on one side, and “applied” or “commercially-oriented” R&D, on the other. Following Arrow (1962) a special need to subsidize research of the first kind has been found in its greater uncertainties and the longer time horizons that at typical in exploratory (“blue skies”) projects. That rationale, however, abstracts from the existence of two quite different organizational and incentive mechanisms that have become thoroughly institutionalized, and through which modern governments have tended to furnish economic support for different classes of research activity. The distinct institutionalized regimes of “open science” and “proprietary R&D” each address the same appropriability problem, but they do so in contrasting ways that serve quite different purposes that can be complementary in their effects and hence interact at superinstitutional macroeconomic level to sustain long-term innovative capabilities a country’s potential for economic growth. Open science, a cooperative mode of research that treats new findings as tantamount to being in the public domain is able to fully exploit the “public goods” properties of data and information, permitting these to be concurrently shared in use and re-used indefinitely. This is an efficient and effective recipe for promoting faster growth of the stock of knowledge. Maintenance of the key “open science” norm of information disclosure within publicly funded universities and research organizations works in conjunction with the tying researchers’ rewards to the achievement of “priority” in new discoveries; by inducing more rapid and complete disclosures, the collegiate reputational reward system abets faster validation of findings, reduces excess duplication of research effort, enlarges the domain of complementarities. It thus yields positive “spill-overs” among research programs in the public sector, as well as externalities that enhance the rate of return on private sector R&D investments. The evident limitations of this mode of advancing knowledge are two-fold. Firstly, because rapidly disclosing what you have discovered makes it very difficult to directly appropriate any of the economic benefits that are derived from the newly generated knowledge, the enterprise of open science is completely dependent upon material support from the public purse (and the patronage of private foundations, especially those exempted from taxation). But, secondly, the research programs pursued in open science communities, being funded ultimately by political and administrative mechanisms, to that extent remain less closely tied and responsive to the market signals arising from the utility that producers and consumers derive when new knowledge is exploited for commercial ends. Tax-payers may be prepared to tolerate a certain level of “science for science’s sake,” but the conduct of research at the frontiers of science has grown to be an increasing expensive proposition and the indulgence of even the most enlightened of tax-payers can be exhausted. Propriety R&D is the mode of pursuing research the can be seen as the answer to the problems posed by open science. Inasmuch as the unlimited entry of imitative rivals would tend to erode the private profitability of investing in commercially-oriented applications research, discoveries and inventions made by researchers in proprietary R&D labs need to be either kept secret, or be exploited under the “protected” provided intellectual property rights

- 10 monopolies. But this is not a perfect solution: although the prospective award of exclusive “exploitation rights” is conducive to the maximization of private wealth stocks that reflect current and expected future flows of economic rents (extra-normal profits) gained by responding innovatively to perceived market demands, the restrictions that IP monopolies impose on the actual utilization of innovations had a perverse consequence. They curtail the both the immediate social benefits and the externalities that wider diffusion could create for future innovative activity. Whereas the proprietary model of research operating in isolation is likely soon to exhaust profitability exploitable discoveries and inventions and run into declining rates of return and shrinking R&D budgets, the contributions of the open science sector, by contrast, are particularly conducive to maximization of the rate of growth of society’s stocks of reliable knowledge and thereby to supporting both the social and private marginal rates of return from current and prospective innovation-oriented research investments. This functional juxtaposition suggests a logical explanation for the co-existence, and the perpetuation of institutional and cultural separations between the two organizational regimes: the publicly supported research pursued in “the Republic of (Open) Science” and the commercially-oriented R&D conducted under proprietary rules in the private business sector. . Maintaining these sub-systems in a productive balance with each other, therefore, is one of the central tasks, if not the central task towards which informed science and technology policies should be directed. Implicit in the foregoing is the important point that balancing the allocation of resources at the macro-institutional level is a very different undertaking than trying to combine key features of the two regimes within a single public institution or private organization. These alternative resource allocation mechanisms are not fully compatible with one another when conjoined within a common institutional setting, because they require different organizational policies regarding the control of information, and they involve distinct and conflicting incentive structure affecting the behaviors of those engaged in research at the micro-level. A fortiori, within same research groups and institutes an unstable competitive tension tends to emerge between the conflicting organizational norms, and the likely outcome is that if the groups adhering to distinct norms do not break apart, the more fragile micro-level structures of cooperation and the informal peer-esteem based incentives that support those behaviors will be undermined (see, e.g. Owen-Smith and Powell, 2001, David and Hall, 2006) . 3.2 Positioning policy between responses to coordination failure and excess momentum The inability of private agents to coordinate their investment plans in order to create mutual positive externalities, and thereby to increase both private and social returns from their respective innovations, has been recognized historically as a feature of periods of profound technological transition in capitalist economies --such as the dawning of the canaland the railway-building eras in the economies of the West. A rather newer perception is that such inabilities reflect a generic source of “market failure” that calls for corrective policy responses. This reflect a conceptualization of the economy as an evolving complex system, exhibiting properties of increasing returns and self-reinforcing mechanisms in which the management of innovational complementarities plays a major role in determining the

- 11 motivation for and the performance of decentralized private investments, including those in R&D. It is attractive therefore to consider using the structure of micro-level incentives created by complementarities in technical systems and organizational mechanisms as a means of amplifying the effects of key policy interventions. In that way it might be feasible, with a smaller expenditure of public resources to propel the economy, or some large sectors thereof, toward development along a new techno-economic trajectory that would shift resources away from lower productivity uses and expand the future opportunity set of still higher productivity investments. This vision encourages the view that STIG policy should seek to identify and encourage certain classes of technology that provide “natural levers” to lift the economy’s rate of economic growth. The recent popularity of the concept of a “general purpose technology” (GPT), and its relationship to innovation, productivity improvement and acceleration of economic growth (David, 1991; Bresnahan and Trajtenberg, 1995; Helpman, 1998; David and Wright, 2003) could then be seen as an attractive ground on which to build support for governmentally initiated programs of that kind. It will be seen, however, that there are some pitfalls awaiting incautious travelers along this particular policy route.. The aspect of GPTs that should render them attractive for public policy planners is that they often give rise to noticeably “hot” areas of private technological research, where those engaged are enthusiastic about investing in commercialization opportunities that they believe soon to be within reach (biotech, nanotech, synthetic biology, and so on). If the “GPT rationale” for focused programs of public investment is to be invoked persuasively, one should be able to make the case that the dynamics of development and diffusion of the new class of technologies is likely to be characterized by strong innovation complementarities between inventions and the “co-invention of applications.” Thus, in examining the mechanisms through which a GPT in the shape of information technology has contributed to late twentieth-century economic growth, Bresnahan (2003) stresses that the phenomenon of socially increasing returns of scale that is manifested at the economy-wide level rests upon the complementarity of quite different forms of innovative activity. Positive feedbacks between the invention of new information technologies and co-invention of applications in new domains appear concurrently in many particular markets. Where there are innovative opportunities in two domains of invention, the process is one resembling “cross-catalysis,” with positive feedback flowing back and forth and sustaining a temporally extended flow of advances. The development of very general scientific and technological knowledge, emerging from explorations of certain fundamental physical phenomena in a number of distinct domains where their potential applicability is recognized, in turn, forms a common foundation for specialized engineering advances in distinct industrial clusters. Opportunities are thereby created for further innovations that realize new functionalities and technological affordances from the design of products and systems than entail the convergence of previously distinct technological clusters, sometimes exploiting the complementarities between older and newer clusters. But these are just the conditions in which dynamic coordination failures are likely to arise from the very structure of complementarities in which the social increasing returns associated with the GPT-based development are rooted. “Chicken and egg” situations do not

- 12 automatically resolve themselves into action; excess inertia and the inability of the system to fully exploit the potentialities of the GPT are the “down-side” of this bright coin. Appropriate policy responses in such complex settings are correspondingly more difficult to prescribe than those discussed in connection with cases involving essentially isolated “market failures” (in Section 2.1). They are closer in nature to the strategies for designing coordinated policies interventions in product and factor input markets that are closely coupled with scientific research and market-oriented R&D. It will be seen (in Section 4) that of devising an integrated set of mutually compatible and preferably mutual reinforcing policy actions to reinforce the impact of such a program. But, in addition, it is likely to be necessary for government interventions to be coordinated not only on the supply side, but also to align the development of demands for complementary innovations with the development of supply capacities that will allow them to come to the market concurrently, so that their diffusion into use can be mutually reinforcing. (This was Ragnar Nurske’s seminar contribution to the “big push” strategy of development, which, in the 1950s and early 1960s was a popular rationale for development policies featuring complementary import-substitution investment.) The policy design problem we are considering is thus especially tricky, both because the issues of timing are more delicate and the dynamic processes themselves are fraught with uncertainties, and because one cannot ignore the intricacies of constructing a technically interrelated system through the self-coordinated actions of decentralized innovators and producers of system components. This challenge for policy-making is a particularly critical one where network externality effects are a dominant source of positive feedbacks. Special attention has to be given to the timely creation of conditions of interoperability or technical compatibility, as these permit the realization of economic complementarities and fruitful market and nonmarket interactions among organizationally and temporally distributed researchers, inventors, innovators, and end-users.

4. Policy complementarities and institutional dynamics – and broadened systems perspective The economic payoffs from public programs that aim to promote innovation by supporting private R&D investments are more likely to be disappointing, if indeed they materialize at all, when program design and implementation decision fail to take account of the interdependence of the STIG subsystem with the economy as a whole. There is evidently a need to focus on the more “tightly coupled” elements and give priority to identifying the ones that are strong complements of the activities or institutional structures that the policy intervention seeks to affect. Complements call for complementary policy interventions in order to promote positive feedback responses in the tightly-coupled parts of the economy, or at least to mitigate the force of negative feedbacks that can damp, or effectively counteract, the intended effects of the policy intervention targets to improve the performance in the STIG subsystem.

4.1 Crossing some boundaries of intra-disciplinary specialisation

- 13 We must therefore take note of the need for some coordination across well-defended boundaries of specialization within the economic policy community, inasmuch as R&D subsidies strategies have been found to be rather ineffective when attention fails to be paid to the context that is set by policies for education and training, labour market policies, competition policy, and macro-economic stabilization policies (see Aghion and Howitt, 2005). In respect to each of those distinct domains of policy formation, it would be a signal error to concentrate on detailing a single policy measure while ignoring others that could be in conflict with its contributions toward the objectives that particular science and technology policies are intended to secure. Thus, in conceiving of an integrated program to leverage the positive feedbacks of an emerging general purpose technology (along the lines considered in the previous section), it would be appropriate to examine how it would fit within, or require alterations in government-sponsored research and public funding of basic research in university and government labs, with the criteria used in awarding R&D subsidies and tax credit incentives, and how it could be reinforced by institutionally grounded policies that trained researchers in new specialities, or by adjustments in visa and immigration regulations to recruit those with required skills from abroad. If it were anticipated that the emerging technological systems would significantly alter production and distribution organizations, attention to the measures that would render labor markets more responsive and industrial relations more accommodating of the adjustments in occupational ladders and working conditions that the introduction of new innovations would be likely to set in motion. Setting out to effect “policy complementarities” of these kinds, however, raises non-trivial problems of coordination among different policy objectives, and the concerns of different policy audiences, a subject that calls for the more institutionally based discussion that is undertaken in the following sub-section. instruments

4.2 Institutional mechanisms: autonomously evolving structures or policy -- or both?

Institutions and organizations engaged in the creation and transmission of technological knowledge, like institutions for other purposes, are neither fixed nor exogenously determined. They emerge and evolve endogenously, shaped by the nature and the economic and social significance of the type of knowledge with which they are concerned, the interests they serve and the resources they are able to command through both market and political processes. But because institutional and organizational structure are less plastic and incrementally adaptable than technologies, they mobilize and deploy resources to stabilize those parts of their environment in which changes would otherwise be likely to undermine the economic rents being enjoyed by agents within them although not necessarily by all the agents (see, David 1994). Auto-protective responses of this kind may reinforce the stasis of other complementary elements of the institutional structure and so can work to impede beneficial innovation elsewhere in the system. Conglomeration is another strategy that may serve similarly defensive purposes: institutions sometimes find it attractive to take on new functions that actually do not have strong complementarities with the core

- 14 functionalities and deeply embedded routines of the organization, yet provide additional access to resources, including coalitions of convenience with other entities. Yet, being resistant to disruption of their learned internal routines, and on that account less plastic, it is also the case that formal institutions that seek to stabilize their external environments may become blind to the strength of the forces against which they are working. They are consequently vulnerable to drifting perilously close to the boundaries of their continued viability; becoming dysfunctional in devoting their resources to resisting forces that are driving transformations in the system around them, they are subject to abrupt and catastrophic alteration: subjected to politically imposed “reforms”, captured and absorbed by other organizations, or dissolved and supplanted by newly created institutions. The economic case for “reforms” of institutions that directly affect the performance of the STIG-system therefore may be developed along two separate branches: “interventions” to change institutions that are seen to be contributing to inefficient outcomes of market-directed processes, and reforms in the internal organizational structures and incentives of public institutions that perform badly in delivering services through non-market channels. “Market failures” may be traced to obsolete institutions or perversely functioning procedures. Nonmarket institutions and organizations, i.e., those whose resource support is not drawn from their ability to sell goods and products to private parties on competitive markets in order to fund their own operations, nonetheless are not free from pressures that may transform and even extinguish them. Obviously, the same may be said for specific government organs and agencies. Inasmuch as the research and training “products” of public sector research organizations, including government institutes, universities, polytechnics and the like, are not priced and distributed through market channels, the criteria for determining where and when to make targeted interventions are vague, and tend to be arrived at ad hoc. Being readily tied to the appropriation of public funding, the policy analysis tends to be framed in terms of tactical choices between decentralized guidance with well-defined incentives and performance targets, or centralized “command and control.” General theoretical insights from the economics literature on organizational design (see, e.g., Sah and Stiglitz (1988)) suggest that where the program involves high inputs of specialized expertise, where information on which resource allocation should be based is not symmetrically distributed, and where activity planning is highly contingent on the uncertain outcome of sequential production stages, decentralization of agenda control and flat organizations are preferable. This principle seems a reasonable rationale for large focused national programs that seek to mobilize the efforts of multiple public (and subsidized private) research and training organizations, including research universities, to create a knowledge infrastructure supporting innovation in a new research domain – nanotechnologies, for example. But, by the same token, it invites substantial coordination problems and inertial drag in the responsiveness of the system to sudden shifts that may occur in the external scientific and intellectual environments, or in the conditions affecting governmental or private sector investment support.

- 15 There are many instances where a case can be made for internal institutional “reforms” because the performance of private R&D labs and public sector research organizations is being adversely affected by the “rent-protecting” behaviors of agents with vested interests. Another paper would be needed to fully develop and present the genesis and possible solution approaches to such situations, especially where the organization in question are buffered against the pressures of market competition or external “takeovers”; or where such extreme remedies are likely to disrupt functionally effective subunits that are “trapped” within a larger dysfunctional system. “Reforming” macro-institutional arrangements, such the legal regime of intellectual property rights, the legislative and administrative law frameworks that structure government university industry R&D programs and projects, and the financing of research training in science and engineering, is generally an undertaking beset by formidable difficulties. These are structures (perhaps “systems” implies too much in the way of order and intentionality) that have evolved in increment, path-dependent fashion, responding at the margins to current pressures and opportunities to garner external support by taking on new missions for which they may not be particularly well suited. The modern patent and copyright systems offer a striking case of legal institutions whose role in the economy has evolved far from their initial historical purposes, and to which other organizations have become adapted even to the point of utilizing them for strategic ends quite inimical to the ostensible purposes on which their claim to legitimacy rests. “Institutional policy” is surely as important as other classes of government interventions that figured more prominently in the preceding discussion (Section 2, especially), but institutions are neither technologies nor commodities, and although economists have much to contribute by analyzing the internal incentives and rule structures of specific existing organizations and institutions, and have developed techniques for evaluating alternative mechanism designs in similarly concrete situations, the present state of economic research on institutional dynamics offers few if any general, a priori points of guidance for policy reformers. Those who seek to stimulate innovation, say, by reforming intellectual property law, or the workings of patent offices, or the organization of research universities, are well advised to study closely the organizations’ histories and professional cultures, as these shape individual behaviors and institutional performance, as well as the specifics of the material incentive structures that have evolved (endogenously) within them. In other words, development policy experiences, which involve some immersion in the local culture and a grasp of the inherited constraints on the melioration of dysfunctional performance (without disrupting the routines that permit continuing fulfillment of vital functions upon which external agents and agencies rely) seem a no less promising practical route to success in addressing needs for institutional reform in developed economies.

5. From theory to practice: toward a more limited role for governments? The general concept of market failure is no longer a controversial issue and the various generic causes of market failures provide a theoretical framework to identify circumstances – indeed, in some respects, too many circumstances -- warranting the provision of public assistance to R&D and other innovation-related activities. While in theory some cases of market failures are obvious, there is a second issue to be considered: the practicality

- 16 and cost of the policy intervention, because it may well be the case that there are some forms of “market failure” that, however serious they may be, are just too costly (or too difficult) to try to correct.

5.1 The challenges of practical implementation A prime example of this is the case of a bad coordination equilibrium involving the selection of technical artifacts or organizational practices characterized by strong network externality effects., Such a situation may come into being as a result of some particular sequencing of events in the incremental evolution of a complex technological system, leaving a majority of those affected wishing they could free themselves to adopt an alternative that now appeared to have offered a better choice. But, having gotten into the existing position along with all the other users, and thus benefitting from the externalities that coordination affords, everyone wants to remain where they rather than bear the burden of attempting an uncoordinated escape to an alternative technical system. The end result, a system that remains “locked in” to technically dominated practices that individuals find costly to simply abandon and replace. Of course, there are situations where the problem is intractable, because even if it was politically possible to organize and execute a collective escape to a new and better position, the prospective collective gains would not be of sufficient magnitude to justify the undertaking the entailed social costs the migration to a better technology. Some bad technological outcomes might have been avoidable, but regrettable as they there existence may be, they are not necessarily worth undoing. This is more likely to be the case once a decentralized system has been allowed to become entangled with organizational and institutional practices that have adapted to it, and correspondingly reshaped business practices, and other technologies that form to accommodate or exploit it special properties. The end-to-end architecture of the Internet, a key technical feature of its design that has posed innumerable problems for previously elaborated fee for service business models and forms of contracting, may be a good case in point (see, e.g., David 2007a) ; in the end even businesses that are not taking full advantage of the accommodation to innovative applications that the Internet’s “connection-less” system of offers, will have entrenched themselves in viable niches that they will resist seeing disturbed by a radically reconfigured high-speed network design. The lesson for thinking about STIG policies in an historical framework is that one is led away from a static analysis of whether or not to intervene, on the evidence that there is market failure and a better arrangement is conceivable if one could start again with a clean slate. Policy decisions will look differently when the options are evaluated at different points in time, that is to say, at different moments in the development of a new scientific field, or in the diffusion of a novel technology. In general, thinking ahead and exercising some leverage on the process in its early stages entails smaller resource costs than will be required for corrective actions subsequently. The only problem with acting on advice is the comparative dearth of information about what one should do at the moments when policy actions would have greatest potency. But is interesting to observe that it is in just such situations, where

- 17 public policy-makers are most inclined to hesitate, that business entrepreneurs infused with robust “animal spirits” will be inclined to plunge into risky innovative ventures –especially when they have the use of other people’s money to do it. Another important practical challenge concerns the correction of coordination failures, which was identified above as an important potential obstacle to the full deployment of a GPT (Klette and Moen, 2000). Understanding the basic principles of coordination problems does not lead directly to useful conclusions about how to construct a suitable technology policy response. The practical implementation of a policy involves answering more than a simple set of questions: what activities in what firms need to be coordinated, and in what way? Appropriate choice of policy tools also requires a detailed technical grasp of the externalities and the innovative complementarities involved. Some economists have emphasized that the informational requirements at a practical level raise serious questions about the feasibility of government policy to correct coordination failures in the real world. For instance, Matsuyama (1997) argues that coordination problems are pervasive phenomena, and that economists’ articulation of these problems by means of simplistic game theoretic models tends to trivialize the coordination difficulties that face policy makers. In real coordination problems, the nature of the ‘game’, the payoff structure, the identity of the players and even their number are often unknown to the policy maker. But while policy makers are seen to face immense difficulties in the course of the practical program implementation, it is not at all obvious that managers of large firms are always better situated; they may be unable to implement cooperative solutions through negotiations and contractual relationships. The latter is the Coasean route to solving such coordination problems through market mechanisms. As a result, the appreciation of the costs of practical implementation and the appreciation of a possibility of a solution provided through market mechanisms point to a similar conclusion about the limited role for governments to act effectively to overcome coordination failures that diminish the returns on public and private investments in science, technology and innovation. The US government role as a successful coordinator in the case of IT often is taken as an example of what government should do in other fields. That case, however, involved a very particular context characterized by a strong identification of R&D investments in computer and computer networking technologies with a specific, high priority government mission (national security). It seems that the US government has had difficulties replicating that performance in other areas. Perhaps the repeated failures in energy technology R&D and diffusion policy (see, e.g. Jaffe et al. (2003)) are attributable to the absence of a strong link between R&D public spending and a government mission that can mobilize broad political support (Mowery, 2006).

5.2 Some tools to enhance the art of managing the complex system dynamics of innovation The theory of technology policy is pretty good. Unfortunately, understanding the basic principles of market failures, coordination failures and policy complementarities does not take one very far in the direction of useful, practical conclusions about how to construct

- 18 technology policy. There is a broad open research agenda which has to address such implementation issues. “System dynamics” theory offers a method for understanding the dynamic behavior of complex systems. The basis of the method is the recognition that the structure of any system, the many circular, interlocking, sometimes time-delayed relationships among its components, is often just as important in determining its behavior as the individual components themselves. It has been pointed out that there are some features that are especially prominent in STIG and other tightly coupled subsystems of modern economies, particularly nonconvexities due to indivisibilities and externalities that create a multiplicity of ‘attractors’ or local equilibrium states (or paths in a dynamical system). In addition, the amplifying effects of positive feedback can produce strong nonlinearities in the responses of agents, or whole subsystems, making it possible that the instabilities created by these feedbacks result in unexpectedly abrupt and discontinuous transitions, formal mathematical “catastrophes”, between markedly different states of the system. Thus, it would be reckless to ignore the potentiality for surprising and perverse outcomes to emerge from what may appear to the unschooled policy-planner to be smooth, “incremental” adjustments in incentives, or local targets, or a program of gradual modification of regulatory constraints intended to improve the performance of a particular regional market or institutions. Recognizing the possibility that things may go badly awry, without being able to explore how sensitive the system is to modifications in one or several of its structures, may not be such a good thing as it sounds at first. The problem is that a “little bit of knowledge” is likely to encourage policy inaction. Yet, as business decision-makers understand, or come to be taught, inaction is itself a strategy that can be punished severely by unfolding events that are driven by forces outside the decision-maker’s control. Suspending action in a battle requires suspending time as Joshua’s command (“Sun stand Thou Still”) sought to do; but without being able to halt time and other’s actions can be far more dangerous than experimenting with policies, and especially if one acts in ways that are reversible, or subject to subsequent corrective modifications. So, we might conclude that an options-theoretic approach is called for: the expected costs of deferring irreversible investments that would seize the gains from existing knowledge (in order to collect more information) should continually be weighed against the expected costs of “prematurely” making commitments that will turn out to be mistaken. This sounds reassuring, but how to assess those costs, and how to identify those situations in which a policy commitment that can be effectively reversed at reasonable costs becomes essentially infeasible to undo? The area of environmental policy is fraught with such traps: lakes that become so polluted that they cannot clean themselves, and so on. The policy can be reversed, perhaps, but by then the action will be ineffectual, or will entail far greater resource costs than were sunk when it was first introduced. It was relatively costless to remove the system of institutional patent agreements whereby US universities could obtain patents on the results of federally funded research, as was done in 1980 by the passage of the Bayh-Dole Act. A proposal today to modify the terms of the Act, let alone undo it, is likely to encounter fierce lobbying resistance, if not from the administrators of some of the universities that were lucky and smart enough to learn how to benefit from the new regime, then from an

- 19 entire new profession of university technology managers who have their own professional association (AUTM), complete with a newsletter, offices in Washington, DC, under plans to open branches in Europe (see, David 2007b, for further discussion). Clearly, some among these effects can be modeled in anticipation, and simulation exercises would provide a framework in which to assemble and integrate empirical information about the behavior of various parts of the institutional, environmental, demographic, and governmental systems that will interact. Moreover, the construction of the apparatus for such modeling exercises will force researchers to pay attention not only to how subsystems are linked with one another, but also to the vital question of the time lags and adjustment speeds that govern the propagation of responses throughout the system. This will expose many of the worst conceits and delusions of policy advocacy that involve abstracting from the question of how long it would take before the promised effects are realized. That will not make getting government ministers and legislators to adopt sound STIG policies any easier, because most of the policies results will emerge much too far in the future to be of immediate political interest. But, at least, it would contribute to clearing the air of the promises that this or this particular legal, institutional reform, administrative rule or tax measure affecting the funding of academic science or corporate R&D, or both, will combat current unemployment, stimulate new firm growth, or reduce infant mortality in time for the next election campaign.

6. Concluding cautions about the ambitions of STIG policy research and practice Technology and innovation policy for growth is widely accepted, but it immediately becomes politically controversial when its implementation goes beyond the support of “exploratory” and “far-from-commercialization” research, and enters into specific details that are perceived to have differential effects on particular markets, institutions and industries. There are good reasons for caution in entering those realms, but the growth potential of R&D and innovation is too clear to abandon policy efforts simply because they are difficult to implement, or politically too charged. It is thus critical to try different ways of structuring policy in this area so as to minimize the array of conceptual and practical policy challenges that are entailed. This essay has sought to confront these challenges by addressing the issue of practical implementation of correcting market failures, and policy coordination failures, by finding an appropriate systems paradigm and (simulation) tools to work within it to assess the dynamics of interactions among policy initiatives, and finally, by addressing the problems of practical policy evaluation. The last words are saved for those who aspire to become visibly effective agents in “directing” the processes of scientific advance, technological change, and innovation along trajectories so as to contribute to improving economic welfare and material well-being of whole societies and nations. Palpable effects of public agency interventions in STIG processes are not likely to translate into political credits within the time frame within which practical politicians and public servants in representative democracies have to function, except if their objectives are confined to redistributing claims of resources gathered by taxation among their respective constituencies. In the realms where creating new scientific

- 20 and technological knowledge and finding ways to use it are essential, the advances are incremental and cumulative, and the assignment of responsibilities for significant successes are retrospective rather than contemporaneous. Moreover, in complex, contingent, and at best partially understood dynamical processes, individuals who hope to claim responsibility for changing the system’s “performance” for the better are all too likely to find that they are the recipients of blame (albeit in many instances equally unjustified) for outcomes that were unanticipated and unwanted. Acknowledgments This essay was prepared for presentation at the Monte Veritas Conference, held in June 2007 in Ascona, Italy. It draws upon the authors’ longer article, “Science, Technology and Innovation Policy for Economic Growth: Linking Policy Research and Practice in ‘STIG Systems’,” forthcoming in Research Policy, 2008 (Special Issue from the SPRU 40th Anniversary Conference on The Future of Science, Technology and Innovation Policy). [Preprint available as SIEPR Policy Paper No. 06-039., July 2007, at: http://siepr.stanford.edu/papers/pdf/06039.html]. Comments and suggestions received from W. Edward Steinmueller, Nick von Tunzelmann, Giovani Dosi and others participants on the occasion of the SPRU conference, and further useful commentaries, on subsequent drafts, from Goddard, Lawrence Goulder, Ben Martin, Richard Nelson, and Luc Soete are acknowledged gratefully. Not all of this help could be absorbed, and those who kindly offered it should not be implicated in the views expressed herein.

- 21 References Aghion, P. and Howitt, P. Endogenous Growth Theory, Cambridge, MA: MIT Press, 1998 Aghion, P. and Howitt, P., Appropriate Growth Policy: a Unifying Framework, the 2005 J. A. Schumpeter lecture, 20th Annual Congress of the European Economic Association, Amsterdam, 2005

Arrow, K.J., "Economic Welfare and the Allocation of Resources for Inventions," in R.R. Nelson ed., the Rate and Direction of Inventive Activity: Economic and Social Factors. Princeton: Princeton University Press, 1962 Bresnahan, T. F. , Mechanisms of Information Technology’s Contribution to Economic Growth, Ch.5 in Institutions, Innovation and Growth, J.-P. Touffut, Cheltenham, Eng.: Edward Elgar, 2003. Bresnahan, T. F., and M. Trajtenberg General Purpose Technologies: Engines of Growth,” Journal of Econometrics, 65, 1995: pp. 83-108. Dasgupta, P. and P. A. David, “Toward a New Economics of Science”, Research Policy, 29, 1994. Dasgupta, P. and E. Maskin, “The Simple Economics of Research Portfolios,” Economic Journal,97, 1987. David, P. A., General Purpose Engines, Investment, and Productivity Growth: From the Dynamo Revolution to the Computer Revolution, in Technology and Investment— Crucial Issues for the 90s, (E. Deiaco, E. Hörnel and G. Vickery, eds.), London: Pinter Publishers, 1991. David, P. A. Knowledge, Property and the System Dynamics of Technological Change," in Proceedings of the World Bank Annual Conference on Development Economics: 1992, [ Supplement to the World Bank Economic Review], L. Summers and S. Shah, eds.. Washington, DC.: International Bank for Reconstruction and Development, 1993. David, P.A., “Why are institutions the ‘carriers of history’? Path-dependence and the evolution of conventions, organisations and institutions,” Structural Change and Economic Dynamics, 5(2), 1994. David, P. A., “The Economic Logic of ‘Open Science’ and the Balance between Private Property Rights and the Public Domain in Scientific Data and Information: A Primer,” in The Role of the Public Domain in Scientific and Technical Data and Information -- A National Research Council Symposium, J. Esanu and P. F. Uhlir, eds., Washington, D.C.: Academy Press, 2003. David, P. A., “Economic policy analysis and the Internet: Coming to terms with a telecommunications anomaly,” Ch. 6 in Oxford Handbook on Information and Communication Technologies, edited by C. Ciborra, R. E. Mansell, D. Quah and R. Silverstone, Oxford University Press, 2007a David, P. A., "Innovation and Europe’s Universities: Second Thoughts about the Embracing the Bayh-Dole Regime," in Perspectives on Innovation, F. Malerba and S. Brusoni, eds.,Cambridge U.P., 2007b. [Preprint available as SIEPR Policy Paper 04-027 at: http://siepr.stanford.edu/papers/pdf/04-27.html.] David, P. A. and B. H. Hall, “Property and the Pursuit of Knowledge: An Introduction,” Research Policy, 35 (June-July) [Special Issue, Guest-edited by P. A. David and B. H. Hall], 2006. David, P. A. and G. Wright, “General Purpose Technologies and Surges in Productivity: Historical Reflections on the Future of the ICT Revolution,” in The Economic Future in Historical Perspective, P.A. David and M. Thomas, eds., Oxford University Press for The British Academy, 2003. Helpman, E., ed., General Purpose Technologies and Economic Growth. Cambridge, MA:MIT Press, 1998. Jaffe, A. B., R. G. Newell and R.N. Stavins, “Technology Policy for Energy and the Environment,” Papers from the NBER Meeting on Innovation Policy and the Economy, April, 2003. Klette, J. and Moen, J., “From growth theory to technology policy: coordination problems in theory and practice”, University of Oslo, draft, 2000

- 22 -

Matsuyama, K., “Economic development as coordination problems”, in M. Aoki (ed.) The role of Government in East Asian Economic Development, Clarendon Press, Oxford, 1997 Mowery, D.C., Lessons from the history of federal R&D policy for an “Energy Arpa”, Committee on Science, U.S. House of Representatives, 2006 Nelson R.R. , “The simple economics of basic scientific research”, The Journal of Political Economy, 67, pp.297-306, 1959 Nelson, R.R., “Market Failure Versus Innovation Systems as Theoretical Orientations for Analysis of Technology Policies,” Paper presented at the Monte Verita Conference on the Economics of Technology Policy, Ascona, Switzerland, 17-22 June 2007. Owen-Smith, J., and W. W. Powell, “Careers and Contradictions: Faculty Responses to the Transformation of Knowledge and its Uses in the Life Sciences,” Research in the Sociology of Work, 10 [Special Issue on “The Transformation of Work”, Steven Vallas, ed.], 2001:pp.109-140. Romer, P. M., Endogenous Technological Change, Journal of Political Economy, 98(5), October 1990: pp. S71-S102 Sah, R.K. and J.E. Stiglitz, “Committees, Hierarchies and Polyarchies,” Economic Journal, 1988. Stemmueller, W. E., “Technology Policy: The Roles of Industrial Analysis and Innovation Studies,” Paper presented at the Monte Verita Conference on the Economics of Technology Policy, Ascona, Switzerland, 17-22 June 2007.