Funded Grants

Evolving ecologies of market agents and their effect on social welfare: Scientific substance and societal significance

Prices are one of the primary goal setting mechanisms for society. The financial system is a distributed self-organized control system whose function is to process information and set prices. Prices in turn determine allocations of resources. When this control system fails to set prices correctly the result can be large inefficiencies in economic production and major social dislocations and inequality. This was dramatically illustrated by the recent technology bubble in American stocks. During 1999 the price of NASDAQ technology stock index rose by roughly 100%, then in the following year dropped by 60%. A popular view is that prior to mid-2000 any competent person with a decent idea for an internet product could easily get venture capital and raise money in the stock market, but since then, even many very solid enterprises have foundered for lack of funding. Real estate prices in Silicon Valley skyrocketed and then plummeted. All this happened without any apparent new information about the technology industry that should reasonably have affected rational valuations. Rather, it seemed that it was driven by investor sentiment and the internal dynamics of the financial system. Even more harm occurred in the 1998 Asian crisis, when a default of Russian bonds triggered major reallocations by American hedge funds which nearly caused a major global financial crash and resulted in several year-long recession for many Asian economies.

Despite the recurring nature of financial bubbles, there is still a widespread belief that markets are almost perfectly efficient. Kenneth Lay, until recently the head of Enron, put it well when he said "I believe in God and Free Markets". At a recent workshop at the Santa Fe Institute, the consensus view among the mainstream financial economists present was that "markets are about 98% efficient", even though there was no justification or consensus on the reasoning that word support such an estimate. Financial economists commonly think in terms of informational efficiency, i.e. the property that prices filly reflect available information. But this also has implications for allocative efficiency - if prices don't properly reflect information, then allocations must be wrong. Wild swings in prices suggest underlying inefficiencies. Either small informational inefficiencies are strongly amplified in their effect on prices (and hence allocations), or informational inefficiencies are much larger than is commonly believed.

Winston Churchill once said that "democracies are the worst form of government, except for all the others". I believe a similar statement might apply to markets. In engineering, a heat engine that is 40% efficient is considered extremely good. The principles of heat engines are well understood, are under the complete control of their designers, and have been perfected over two centuries. Why should we suppose that market agents who satisfy none of these criteria and have the complexities and vagaries of human behavior, are so much more perfect than carefully designed heat engines? While markets are very useful, indeed essential, their design is still imperfect: Once we understand how they really work, we can do a lot to improve them. To do this we have to make models of markets in which optimality is not assumed a priori, and so efficiency is not effectively assumed at the outset.

In this proposal we do not intend to make better predictions about price bubbles. Rather, we propose to develop a new conceptual framework that explains some of the statistical regularities in markets and allows us to understand the broader implications of these statistical properties to problems such as risk planning and market instabilities. Some of the specific questions of interest are:

o What determines how much prices fluctuate? In particular, how does this depend on other properties, such as the flow of orders?

o What is the explanation for the highly non-normal distribution of short term price changes?

o What does market liquidity depend on? That is, why is it that sometimes a given flucutation in demand produces a small change in price, and other times it produces a very large change in price?

o Can we explain the distribution of wealth quantitatively based on simple statistical theories?

o How do we measure the efficiency of a market? Can we say precisely how market A might be more or less efficient than market B?

o Are some financial agent behaviors detrimental to society, and if so, is it possible to suppress them without inducing greater harm in other respects?

Almost all financial models assume a strong decoupling from the economy. Production is taken to be given, and the focus is on the price adjustment process. This throws out the baby with the bath water when it comes to addressing questions of social welfare. Prices are important precisely because they affect production. The fact that there are feedbacks between production and price setting dramatically complicates the task of the financial system, a problem that is usually neglected. Even though we do not imagine we can make accurate quantitative models for the economy, we do intend to explore this question in a qualitative way, by coupling simple "toy economies" to our simulations of price setting agents.

A line of recent work that is related to ours is behavioral finance. The practitioners of this discipline have incorporated information from psychology to argue that not all market participants are rational, have documented price inefficiencies, and have produced toy models that qualitatively illustrate how irrationality can affect markets. Our approach is complementary to this. We differ in several respects:

o We use simulation extensively.

o The style of theory is strongly influenced by physic, computer science, and biology.

o We do not necessarily assume that any of the agents in our models are perfectly rational.

o We place a strong emphasis on the effort to match statistical properties of financial data.

The work proposed here represents a new style for economics. We are proposing a mixture of applications - while most of the focus is on finance, we also look at some broader economic problems, such as the distribution of wealth (for the reason that the style of modeling is similar). If this approach demonstrates success in these problem areas, then one might hope that it could be exported to others.

Speaking as an outsider who has recently entered economics, I have been struck at how much the epistemology of economics differs from that of the physical sciences. This is partly a matter of language, and partly a matter of real differences in substance. Some of it is justified because the object of inquiry is so different, but some of it is not. One of the tangible benefits of this proposed work world be a reworking of economic theory, making the relationship to the theoretical Foundations in other fields more transparent. The connections to thermodynamics, statistical mechanics, and ecology are much bigger than the naive observer might think - it is only necessary to elucidate them. This is partly a matter of translation, but it is also a matter of doing new original work. Thus, we think that the proposed work has direct value for society because of its potential to improve social welfare, and as well as indirect value for its contribution to the broader goal of the unification of knowledge.