4. An Evolutionary View of Technology Driven Long-run Growth
Author: Kenneth Carlaw (Associate Professor, Department of Economics, University of British Columbia and the Department of Economics, University of Waikato and UBC)
4.1 Introduction: Technology as Revolution14
We live in a world of rapid economic and social change. Any one change typically causes others, which in turn cause others, and so on in a concatenation of linked causes and effects. For example, the invention of the dynamo in 1887 allowed for the practical generation of electricity. The use of electricity allowed a separate power source to be attached to each factory machine (rather than being driven by a central power source through a system of shafts and belts as in the steam-powered factory). The "unit drive" electric motor allowed the machines in the factory to be rearranged to coincide with the flow of production through the factory. This arrangement allowed Henry Ford to mechanise production with a moving assembly line. In Ford's hands, the assembly line, together with standardized parts (themselves the result of another key invention in the machine tool industry), enabled mass-produced and affordable automobiles. The Model T and its successors transformed America (and later Europe) in myriad ways. It allowed people to move about more quickly and more cheaply. It provided high paying work to many immigrants who could not easily converse in English. It enabled the suburb, the shopping centre, the domestic tourist industry, and the motel. It altered sexual norms (as dating couples were freed from the supervision of parents and chaperones) - to mention only a few of its far-reaching effects.
We argue that such change is best understood as an evolutionary, historical process driven by endogenous innovative activity. Indeed, the evolution of technology drives much of the economic, social, and political change that we experience. Consequently, in our research we pay much more attention to technology than is usual in the writings of most growth theorists who most often focus on economic growth (usually measured by increases in Gross Domestic Product, GDP) rather than economic change. This is understandable since growth in GDP is relatively easy to measure and its cumulative effects are dramatic. However, a full understanding of the causes and consequences of long-run economic growth requires an appreciation of the qualitative changes induced by technological innovations—a point stressed by Joseph Schumpeter many years ago (Schumpeter (1934 and 1943)). People living at the beginning of the 21st century have measured real consumption that is over ten times as much as the consumption of those living at the beginning of the 20th century. But this measurement does not capture that fact that they consume completely new commodities made with new techniques. Technological advance not only increases our incomes; it transforms our lives through the invention of new hitherto undreamed of products that are made in new hitherto undreamed of ways.
Humans are technological animals. Through many millions of years of biological evolution, technology has been fundamental in making us the physical beings that we are today. Through many thousands of years of economic and social evolution, our adaptations to the technologies that we have created have helped to mould and re-mould our economic, social and political institutions and our behavioural patterns.
Homo sapiens share the use of tools with a dozen or so other animals that routinely make use of one or more simple tools. What distinguishes us from all others, however, is our routine use of a wide range of tools and our ability to invent new tools consciously and persistently in the face of environmental challenges and also driven by our own latent curiosity.
New technologies largely result from activities of profit motivated agents making technological change significantly endogenous to the economic system. Furthermore, scientific and technological knowledge is cumulative. Today's knowledge could not have been discovered or invented in the absence of many earlier discoveries and inventions. Thus, growth and technological change is an historical process in which there is a clear arrow of time. Outcomes are not reversible: introducing a shock and then removing it will not return the economy to its original, pre-shock position because the reaction to the shock will typically lead to the accumulation of new knowledge that will affect future outcomes. Since agents' behaviour and choice sets are path dependent, technological change is replete with multiple outcomes, lock-ins, and possible "butterfly effects." To understand where the system is today, we need to know where it has been in the past. In the study of innovation and economic growth, we need explanations that contain an arrow of time, explanations in which past history does exert an influence on the present¾explanations and theories in which history matters.
4.2 Technology Driven Evolutionary Growth
Evolutionary approaches to understanding technology driven economic growth date back at least to Nelson and Winter's (1982) Evolutionary Theory of Technological Change. Their work represented a fundamental departure in approaches to understanding growth. Explanations of economic growth split roughly along the lines of those in the Neoclassical tradition and those in what we call the Structuralist-Evolutionary view. There are several key assumptions on which the two views differ that lead to critically different predictions.
4.3 Neoclassical versus Evolutionary Approaches
In this section, we compare and contrast the specific elements of what we call the canonical versions of the two theories. These are generalisations of the main elements (tastes, technology and technological change, information and motivation of agents, equilibrium, competition, structure and the role of the market) of the two bodies of theory. "Neoclassical" is our collective term for the well-known body of theory based on rational maximizing agents operating under a well-defined exogenous scarcity constraint with fixed technology and tastes. It has been an extraordinarily successful theory. When dealing with the microeconomic issues surrounding innovation and long run technological change, however, the canonical general equilibrium version of neoclassical economics is largely silent. "Structuralist-Evolutionary" (S-E) is our collective term for the body of theories developed explicitly to analyse long term growth using dynamic evolutionary concepts. Instead of focusing on models of stationary equilibrium states, these theories have sought to model the dynamic processes by which actual technologies evolve under the impact of successive innovations.
Tastes
The treatment of tastes is one of the few places where the neoclassical and S-E views are similar. Few economists in either camp have tried to model explicitly the formation of tastes.15 It seems, however, that if one is to understand long term growth, one must accept a substantial endogeneity of tastes - an endogeneity that probably also exists over shorter periods of time but is ignored in the interests of obtaining tractable models. Consumers buy many goods that did not exist in the past and it seems to us unreasonable to assume that they have tastes defined over the unknown (although some economists insist that they do). For example, could a medieval peasant in 800 have had tastes defined over the range of electronic devices and communications technologies (e.g., ipods, iphones, email, the Internet, etc.) available to consumers in 2007?
Technology and technological change
Neoclassical growth theories employ the concept of a "black box" aggregate production function, which implies that the process and the structure of technological change are observable only by their results. For example, given quantities of all inputs may be associated with larger quantities of output. Conceptually, this phenomenon is observed by measuring the amount of the change in output that cannot be statistically associated with a change in the inputs. The remaining change is referred to as the Solow residual, or total factor productivity (TFP) growth.16
In S–E theories, technology is observed through its embodiment in such things as physical and human capital, infrastructure, the legal system, social norms and practices, etc. Technology has a hierarchical structure of engineering complementarities and technological change is modelled explicitly as evolving endogenously. Also, because S-E theories attempt to incorporate many of the awkward facts surrounding the microeconomics of innovation, they often treat the economic, social and political structure of an economy explicitly. Institutions are seen as co-evolving with technology. The firm is seen as inhabiting a specific point in input space with the possibility of moving to other points but only in real time, at significant cost, and under conditions of uncertainty.
Information and motivation
In neoclassical models, agents are assumed to have complete information sets, sufficient to allow them to make maximising decisions. This implies that all decisions are made either with perfect foresight or with foresighted rational expectations. For the latter, agents need to know all possible outcomes of their choices and to have well defined probability distributions about the likelihood of each possible outcome. This implies that in situations of less than perfect information agents operate in situations of "risk" rather than "uncertainty." Agents need not learn from experience since all information that is relevant to their decisions is known by them initially. In this view, two individuals with the same endowments and tastes, faced with the same choice between two alternative courses of action and possessing the same set of relevant information, are predicted to make the same maximizing choice.
In S-E theory, innovation is typically seen as endogenously determined by decisions taken by individuals in search of profits. The theory does not endow agents with perfect information or perfect foresight. Instead, agents face uncertainty when making decisions, particularly those decisions associated with innovation. Since innovation means doing something never done before, it is often impossible to enumerate in advance the full set of possible outcomes of a particular line of research. In such situations, agents will be unable to assign probabilities to alternative future states in order to conduct risk analysis. Therefore, groping in a purposeful, profit-seeking manner is the usually assumed behaviour of agents. The key implication of uncertainty is that two individuals with the same endowments and tastes, faced with the same choice between two courses of action, and possessed of the same bounded set of relevant information, may make different choices. Given the uncertainty, neither individual's choice can be said to be ex ante irrational, even though it may turn out ex post to be inferior to an alternative.
Equilibrium
Much neoclassical theory is Newtonian in conception. Forces balance each other to produce equilibriums that are typically stationary, unique, optimal, and rendered stable by negative feedback. Small perturbations are dampened so that the system returns to its initial equilibrium position. When technology is changing in this view (often characterised as a stock of accumulating knowledge that enters a given production function), the equilibrium concept is either a steady state or a dynamically stationary optimal growth path characterised by a constant growth rate. Many neoclassical economists have been interested in institutions and have modelled many aspects of the economy's structure such as the location of industry and the internal management of firms. Nevertheless, the general-equilibrium, Arrow-Debreu-type theory on which many of the most influential neoclassical policy prescriptions are based, usually focuses on an equilibrium end state with little or no attention being given to the characteristics of and structure of institutions that experience suggests influence behaviour. Even where such institutions are modelled the equilibrium concept is stationary.
In contrast, the purposeful groping behaviour, endogenous and evolving choice sets, and endogenous and evolving technology of S-E theory imply the absence of a unique, welfare-maximizing equilibrium. The innovation process is replete with non-convexities—such as once-for-all costs of developing and acquiring technological knowledge, positive feedbacks from current market success to further R&D efforts, and complementary relations among various technologies. S-E models with their uncertainty and non-convexities incorporate path-dependent processes. Some formulations of the resulting behaviour yield punctuated equilibriums: long, stable periods alternating with bursts of change, the timing and substance of which are not predictable in advance. Others yield multiple equilibriums, in which historical accidents determine which equilibrium will be reached or approached at any one time. Still others yield only perpetual change or non-stationary equilibria. Considerations such as these put an arrow of time into S-E theories.
Competition
Neoclassical theory treats competition as the end state of the competitive process. There is no ongoing process of rivalrous behaviour. Instead, what is modelled is the static state in which firms are all perfectly adjusted to their stationary environment. In the market structure of perfect competition, firms have no power over the market and so no means to engage in rivalrous behaviour vis a vis each other.
S-E theory treats competition as a process that takes place in real time. Behaviour takes the form of active struggling of firm against firm each seeking a temporary advantage over the others. In this type of competition, technological innovations are a major tool by which firms strive to gain competitive advantages. However, no such advantages are permanent and so none will show up in a stationary, long run equilibrium.
Structure
The neoclassical view tends to display the world as smooth, subject to incremental alterations, with a featureless technology and homogeneous agents whose behaviour is adequately displayed by that of a single representative for each class of agent. The S-E view tends to display the world as lumpy, subject to discrete alterations, with a structured technology and heterogeneous agents. Institutions are themselves hierarchical. It is the co-evolution of technology and institutions that determine the growth dynamic. Evolution is driven by differences among agents and it is often the outlier, not the median agent, who drives change.
The role of the market (evolutionary selection)
As we have seen, the neoclassical market is one in which suitably informed agents acting to maximize their own objective functions subject to well defined feasibility constraints arrive instantaneously at the optimal market equilibrium. In contrast, the S-E view is one of imperfectly informed agents groping under uncertainty towards outcomes they perceive as better, and thus driving an historical, path dependent process that never settles into a stationary equilibrium but is, instead, continually jostled by new endogenously created innovations. One of the great issues in the economics of long run growth is to explain why the whole economy behaves in a more or less ordered way although the key decisions are made by many unrelated agents.
The neoclassical explanation is that the price system does the coordinating by producing publicly available signals that reflect relative scarcities to which individuals respond in a self-interested manner and, in the process, produce order in the system. In this view, agents have the all of the relevant information and do the maximising calculations themselves. The markets' function is to generate information in the form of price signals, which is all that the agents require from them. Without them, decentralised decisions of individual agents would not produce the emerging property of an economy that looks as if it had been consciously coordinated.
The S-E approach commonly assumes that agents lack the relevant information that would be required to make optimizing decisions. Furthermore, when operating under uncertainty it is unclear what maximizing behaviour even means. So the market has a much more important coordinating role to play. Agents do the best they can, often forming mistaken expectations about the underlying processes and often being subject to bandwagon thinking, and various other misdirecting influences. Sometimes they get "it" right but often they get "it" wrong. So the job of the market is to direct behaviour towards more value-creating activities by rewarding successes and punishing failures. In this way markets act as the evolutionary selection device, just as survival functions for biological evolution. Those who, by luck or good judgment (or both), get it right are awarded big profits, much larger than the normal return on capital that is all that is needed to direct resources in static perfect competition. Those who get it wrong lose and, if their losses are sufficient, they disappear from the system.
Compared with the static world of neoclassical welfare economics, the problem of coordination is much more complex in an S-E world of continuous change. How does a system that is continually changing and destroying much of what it has, and that is subject to cumulative causation, path dependence and increasing returns to scale and a host of non-linear dynamic structures, produce relative order? Our answer is that first, technologies develop along relatively structured paths shaped by their technical characteristics, and evolutionary history of accumulated knowledge derived from inventing and applying technology; second, when technologies are evolving endogenously, the evolutionary hand of the market is the major selection mechanism for choosing those strategies that will be reinforced by profits and those that will be discouraged by losses; and, third, the uncertainty associated with technological change requires institutions in the private and public sectors to shape behaviour and organize the interactions of agents, which serves to stabilize the system.
4.4 Technology, Structure and Change
One of the most important features of the S-E approach is that it explicitly models the microeconomic features of technology, the structure (including institutions and culture) into which it is introduced and the evolving interaction between these two. In order to discuss the interaction, which is of ultimate interest to those seeking to understand technology driven growth, we need to define technology and its characteristics, as well as define the economic and social structure into which technology integrates. This will allow us to articulate the co-evolutionary processes of technological and structural change.
Technology
Technological knowledge, technology for short, is the idea set specifying all activities that create economic value. It comprises: (1) knowledge about product technologies, the specifications of everything that is produced; (2) knowledge about process technologies, the specifications of all processes by which goods and services are produced; (3) knowledge about organisational technologies, the specification of how productive activity is organised in productive and administrative units for producing present and future goods and services (which thus includes knowledge about how to conduct R&D).17
Technological change runs the whole gamut from continuous, small, incremental changes, through discontinuous radical inventions, to occasional new general purpose technologies (GPTs) that evolve to pervade much of the economy. All types of technology display the three related characteristics of building on accumulated knowledge, emerging in crude form, having few complementarities with other technologies, but subsequently developing a wider range and variety of use.
New knowledge builds upon existing knowledge. One does not invent the dynamo without an understanding of magnetism, conductivity and so on. No society has ever discovered the cam shaft without first discovering the wheel, and since the cam shaft is the key to harnessing rotary motion, no society without the idea of the wheel has managed the generation and harnessing of rotary motion. The nature of knowledge and discovery is inherently historical, and in accumulating knowledge "signposts" (crucial discoveries that enable whole research trajectories) matter.
It is always the case that newly invented technologies emerge in crude form, usually applied to a single activity and usually designed for a single purpose. For example, Newcomen's simple steam engine had the single purpose of pumping water out of ever deepening coal mines in Britain. Then via innovation and diffusion these technologies are refined, applied to more activities and adapted to more uses. Refinements to the steam engine resulted in it being able to withstand pressure up to several atmospheres and deliver vastly more horse power than the original Newcomen engine which ultimately resulted in a vast variety and range of application.
Elements of technological knowledge integrate with other new or existing elements of knowledge. Integrated capital systems are made up of many components. The components themselves consist of many sub-components and these sub-components are made up from sub-sub-components, and so on. An implication of the interrelated structure of capital is that components are complementary to one another, as well as to the integrated capital good itself. These complementary relationships range from the extreme of a component being necessary for the function of a technology to a range of weaker versions where the component merely enhances other components to varying degrees.
GPTs
In our research, a class of technology, general purpose technology (GPT), warrants special comment in the discussion of evolutionary long-run growth. A GPT is a technology that initially has much scope for improvement and eventually comes to be widely used, to have many uses, and to have many spillover effects. Almost every technology one would care to identify possesses at least some of these four characteristics we have just identified. However, no one of the above characteristics is sufficient to identify a GPT. A GPT must possess all four of the characteristics in abundance.
The importance of GPTs is found in their capacity to rejuvenate and sustain economic growth over the long run. New GPTs present agents with a whole new research program to develop new process, product and organisational technologies that make use of the new technologies. As long as transforming GPTs continue to be invented, there is no reason why growth cannot proceed into the indefinite future. Scientific and technological history gives no reason to suspect that the flow of new GPTs will dry up. Indeed, several new GPTs can be seen emerging at present, in particular biotechnology and nano-technology, both of which give promise of transforming products, processes and organisations across a wide spectrum of the whole economy.
Structure
Structure is the realisation of technological knowledge; it embodies technological knowledge; all technological knowledge must be embodied in the structure to create economic value. To be useful, the great majority of technologies must be embodied in one way or another. Structure is comprised of the following18:
- all physical capital,
- consumers' durables and residential housing,
- people, and all human capital that resides in them and is related to productive activities, including tacit knowledge of how to operate existing value-creating facilities,
- the organisation of production facilities, including labour practices,
- the managerial and financial organisation of firms,
- the geographical location of productive activities,
- industrial concentration,
- all infrastructure,
- all private-sector financial institutions, and financial instruments,
- all public sector institutions, parliament, courts, civil services, regulatory bodies, and other government bodies,
- humans who staff these organisations and whose human capital embodies the knowledge related to the design and operation of public sector institutions, i.e., institutional competence.
The agents who take most of the decisions concerning these elements are firms, governments and households.
Technological and Structural Evolution
A critical dynamic in understanding the process of economic growth, driven by technological change is to understand how the endogenous actions of agents groping under uncertainty and constrained by the characteristics of technology and pre-existing structure generate change in the economic system. As we have noted, everything that is known about the evolution of technology suggests that its course is uncertain. This uncertainty is involved in more than just making some initial technological breakthrough. Most development expenditures are on product, not process, development, largely because new technologies come into the world in crude form, after which they are slowly developed as their range of applications is expanded in ways that are impossible to predict in advance. Another cause of uncertainty is that two or more technologies sometimes prove, to everyone's surprise, to be complementary and to produce when operating together much more than the sum of the parts when they operate independently. There are also uncertainties about how long a technology will continue to be useful before it is replaced by a superior technology. In addition there is uncertainty about how new technology will interact with pre-existing institutional, legal and cultural structures in a society. Because of such pervasive uncertainties, technologies evolve along trajectories that are path dependent in the sense that what seems a possible, next step depends on the successes and surprises in the previous attempted steps.
Similar comments apply to diffusion, which is a slow, costly and often uncertain business. Just to discover what is current best practice around the world is a daunting task. Even if an agent can identify best practice techniques, this (at most) provides it with a blueprint; learning how to produce what is described in a blueprint successfully implies acquiring all the tacit knowledge that goes with adopting something new. It follows that the existing set of technologies does not provide a freely available pool of knowledge. Learning about technologies in use elsewhere and adapting them to one's own uses is a costly process - typically requiring innovation in its own right - innovation and diffusion shade into each other rather than being clearly distinct activities.
This adaptive learning process applies to all agents operating in all elements of the economic structure. Firms create, search for and adapt best practice techniques to compete with other firms, consumers make choices over ever widening bundles of consumptions seeking to increase their well being and governments adapt existing property right and criminal laws to meet the ever changing technological environment to increase social welfare. In the process structural and institutional mechanism are adapted to better fit emerging technology. All of these processes are costly in the sense that it takes resources to create learn and adapt. Critically the coordinating mechanism of the market selects the best strategies of all agents and rejects the worst.
4.5 Policy implications
The two theoretical views of the growth process discussed above lead to two different views of policy.
In the Neoclassical view, maximizing agents equate the expected returns from a marginal unit of expenditure everywhere in the economy, including all lines of R&D. Given all the other standard assumptions, a welfare-maximizing equilibrium exists. Departures from this equilibrium are caused by market failures, which take three general forms; externalities, imperfect information, and non-convexities. The removal of these market failures is the main object of neoclassical microeconomic policy advice. There is nothing in the general models that distinguish one economy from another such as different specific technologies, different institutions and different histories or stages of development, such as an economy that is catching up technologically or one that is at the technological frontier. As a result, its policy advice is general, applying to all market economies operating at all times. The advice is to remove market imperfections wherever possible.
A further implication of the equilibrium concept in Neoclassical models is that all policy interventions are reversible because historical path dependence is ruled out by the assumptions of this view. A policy intervention could be put in place for a time causing a change in the equilibrium while the policy is in place. And upon its removal the economy would revert to its original equilibrium.
In contrast the S-E view highlights a number of possible, context specific roles for policy intervention at a number of different stages in the path dependent process of complex interrelation between technology and structure.19 The S-E view sees a role for policy in exploiting the differential and context specific technological complementarities among the various elements of technology and the economic structure. There is a role for policy to strike a balance between innovation and diffusion that is also potentially specific to different classes of technology. There are roles for adapting existing structure to accommodate new technologies and to aid other agents in the economy in overcoming sunk information costs about new technologies and best practice used elsewhere. The over arching view is that inducing economic growth through technological change is good but given the uncertainties that are inherent in the process how much of each kind of change to try to induce must come down to an irreducible element of judgement on the part of the policy maker.
Because of uncertainty, complementarity, the accumulation process that knowledge growth follows and the resulting path dependency, policy decision have the capability of altering the development trajectories of research agendas from minor incremental process in specialized lines of activity to the entire development trajectory of an economy, sometimes with devastating effect.20 There is a fundamental message here. In the timeless Neoclassical framework a policy mistake (to the extent that such things are possible) can be reversed. In the S-E view a mistake can have lasting effects and may even eventually lead to success because it diverts the system onto a new and fruitful trajectory. Critically, policy makers must understand that they operate under the same veil of ignorance about the future as the agents whose behaviour they are seeking to influence. In an ever evolving system "doing nothing" is actually "doing something" because governments and policy makers exist and even the action of doing nothing can have lasting effects on how the system evolves. An understanding of the complex interrelationship between technology and economic structure is, therefore, essential to delivering good policy, but so too are independent assessment mechanisms for policy (i.e., an appropriate selection mechanism), institutional competence, and institutional flexibility. All of these are necessary because, given the uncertainty of the process mistakes will be made.
Having given the requisite warning it is also important to note that institutional structure, legal conventions, societal norms and culture all must evolve along side technological change in order to produce economic growth. Governments play a critical role as the agents of change in much of this evolutionary process.
4.6 References
Carlaw, K. I. and R. G. Lipsey (2003) "Productivity, Technology and Economic Growth: What is the relationship?" Journal of Economic Surveys, 17(3), 457-495.
Landes, D.S. (1998) The wealth and poverty of nations: why some are so rich and some so poor, (New York; London: W.W. Norton).
Lipsey, Richard G., Kenneth I. Carlaw and Cliff Bekar (2005) Economic Transformations: General Purpose Technologies and Long-run Economic Growth (Oxford University Press: Oxford, UK).
Lipsey, R. G. and K. I. Carlaw (2000)) "Technology Policy: Basic Concepts" in Edquist, C. and M. McKelvey (eds.) Systems of Innovation: Growth, Competitiveness and Employment (Edward Elgar: United Kingdom), 421-455.
Lipsey, R. G., C. Bekar and K. I. Carlaw (1998) "What Requires Explanation?" in E. Helpman (ed) General Purpose Technologies and Economic Growth, MIT Press, Cambridge, Massachusetts (pp. 15-54).
Lipsey, R. G. and K. I. Carlaw (2004) "Total Factor Productivity and the Measurement of Technological Change", Canadian Journal of Economics 37(4), 1118-50.
Nelson, R. and S. Winter (1982) An Evolutionary Theory of Economic Change, (Cambridge, Mass.: Belknap Press of Harvard University Press)
Schumpeter, J. A. (1934) The theory of economic development: an inquiry into profits, capital, credit, interest, and the business cycle, (Cambridge, Mass.: Harvard University Press).
Schumpeter, J. A. (1942) Capitalism, socialism and democracy, (New York: Harper).
Back to Top