Professor W. Brian Arthur focused his remarks on the application of complexity theory to economics. Arthur noted that two characteristics of complexity theory are of interest to him and his colleagues: non-linearities, and adaptive elements. Complex systems have many elements interacting. The patterns that form are complex and non-linear. Arthur noted that complexity is not a theory in itself, or even a body of literature, it is a change in the way science is conducted. The old paradigm of scientific inquiry tended to be top down; with the insights gained from research into complexity, the sciences are now seeking common patterns and points of interaction to find what unexpected things might emerge.

The ability to discover and study complex systems has been made possible by desktop workstations. In the early 1980s, Arthur asserted, desktop computing power reached the point at which researchers had at their fingertips the power to create and fiddle with patterns that form with the interaction of certain agents. This created a seachange in the sciences from stasis to process, from equilibrium to dynamism.

The science of economics has benefited from this revolution by beginning to understand the non-linearity in the economy and its impact on economic growth and development. (Understanding adaptation is also in the early stages of research, Arthur said.) In his research at Stanford University and the Santa Fe Institute, Arthur has found that economic agents/entrepreneurs who can dream up new things are the analog of reactive elements in chemistry. Many systems are composed of multiple elements and reactive agents; in any complex system there are positive and negative feedbacks that cause different outcomes even under the same or similar conditions.

The outcome of any one event may not be optimal or the most efficient pattern, and the patterns may not be fully predictable. (Arthur used the example of water beading on a tray. An experiment performed over and over will result in different patterns of water beads each time, even though conditions may be the same. While an observer could predict statistical variations, the pattern of any one event could not be predicted.)

In the economy, positive feedback might be defined as the phenomenon of increasing returns. An increased knowledge base, for example, leads to greater economic growth. As knowledge becomes embedded in technology, economies experience increasing returns from the acquisition and investment of new knowledge. High technology industries are different from industries that experience diminishing returns in that the knowledge input, and usually the costs, are up front. As a result, in high technology industries, network effects accrue as technologies become standardized in the marketplace.

As an economy becomes more high tech, it experiences deep sources of increasing returns promoting highly unstable competition. A company may learn by using a product and thus find that development and production must constantly adapt. In this unstable market, chance events can cause a product to lock-in around a standard, but lock-in may not be optimal (i.e., the DOS operating system).

Increasing returns from lock-in can mean that countries may take over in certain industries, such as Japan in consumer electronics, the U.S. in aircraft. Cells within the economy can increase productivity, but under the paradigm of increasing returns, the ability of an agent to gain in productivity depends upon their willingness to invest in education. (Arthur cited a study by Steve Durloff (sic) which examined the cycles that form around education, productivity, and income cycles.)

While many observers have noted the role of increasing returns in high technology industries over the years, until the advent of dynamic modeling and complexity theory, there has not been a way to model this phenomenon. There currently exists a rudimentary ability to analyze these phenomena with rigorous computer experiments.

Within these computerized experiments, analysts are able to pull the levers and dials and examine possible outcomes.

Clearly, there are a multiplicity of possible outcomes in any complex adaptive system, Arthur asserted, but it is not clear how to relate these outcomes to public policy. Small events can drive a system into a cycle of activity. Within these cycles of activity, there may be timing windows when new paths of action are possible. In other cases where the system is locked-in, policymakers may attempt to solve a problem, but no change may be possible. Arthur likened the problem to water streaming into a rugged landscape. At different times one could imagine tilting the system to achieve desired outcomes. When the water has settled into valleys, little change may be possible.

The science of complexity can offer insight into the old notions of industrial policy versus a laissez-faire attitude towards the economy. In some cases where a timing window may open, a light touch at the right time nudging technology development, dissemination, or implementation towards a preferred outcome may have a significant impact. Moreover, it is important for policymakers to note that, because of the phenomenon of lock-in and positive feedbacks, the economy may not be in the best of all possible worlds. This opens a role for policy to help achieve better outcomes.

In addition, the science of complexity emphasizes the importance of history as determining paths a system might take in the future. The economy is much more path dependent and history dependent than earlier economists thought, Arthur said.

The changes wrought by the science of complexity is sweeping through macroeconomics. Its impact is most notable in the studies conducted on the role of expectations in interest rate fluctuation by Tom Sergeant and others. Paul Romer has done important work using complexity theory to examine the sources of economic growth, showing that growth leads to growth as an economy builds knowledge and invests in fixed costs. Paul Krugman has made excellent contributions in trade theory. In addition, antitrust theory is coming under new scrutiny with the findings of complexity theory. Antitrust policy was developed for diminishing returns industries, Arthur said. Firms that operate under the rules of increasing returns would have very different trust relationships. Antitrust theory must take this into account.

Author: Caroline Wagner

Rand Corporation

Go Back to Homepage